Professional Documents
Culture Documents
A T hesis
Presented to
T he Faculty of Graduate Studies
of
T he University o f Guelph
by
JOSE LUIS JARAMILLO VILLANUEVA
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Library and
Archives Canada
Bibliotheque et
Archives Canada
Published Heritage
Branch
Direction du
Patrimoine de I'edition
NOTICE:
The author has granted a non
exclusive license allowing Library
and Archives Canada to reproduce,
publish, archive, preserve, conserve,
communicate to the public by
telecommunication or on the Internet,
loan, distribute and sell theses
worldwide, for commercial or non
commercial purposes, in microform,
paper, electronic and/or any other
formats.
AVIS:
L'auteur a accorde une licence non exclusive
permettant a la Bibliotheque et Archives
Canada de reproduire, publier, archiver,
sauvegarder, conserver, transmettre au public
par telecommunication ou par I'lnternet, preter,
distribuer et vendre des theses partout dans
le monde, a des fins commerciales ou autres,
sur support microforme, papier, electronique
et/ou autres formats.
i*i
Canada
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
A BSTRACT
EX C H A N G E RATE M O V EM EN TS AND M EX IC O -U N ITED STATES
A G RI-FO O D TRA D E: A C O IN TE G R A TIO N ANALYSIS
Advisor
Dr. Rakhal Sarker
The favourable Mexico-United States exchange rate has been cited as an important factor
for the growth in exports of agri-food products from Mexico to the United States. Some
analysts have also emphasized the importance of NAFTA and the exchange rate volatility
in the performance of the Mexico-US agri-food trade. An important policy question is the
extent to which changes in the Mexico-US exchange rate and its volatility have
contributed to the growth in agri-food trade between these two countries. The purpose of
this study is to quantify the effects of the Mexico-US exchange rate changes and its
volatility on Mexico-United States agri-food trade flows using the Maximum Likelihood
Cointegration analysis.
To guide the empirical analysis, an expected utility maximization model has been
developed in this study. The comparative statics results derived from this model show that
while changes in exchange rate have a positive effect on imports, exchange rate volatility
has a negative influence on trade flows. A set of reduced form equations was developed
and estimated to determine empirically the effects of changes in exchange rate, its
volatility, and other relevant factors on trade flows of tomato, maize, sorghum, and milk
powder from 1989 to 2004.
The results from cointegration analysis show that while changes in exchange rate
have a positive effect on trade flows, volatility of the exchange rate has a negative effect
on trade flows. The specification of the exchange rate volatility does matter because
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
different estimates are obtained when alternative measures of volatility are used. The
results indicate that the volatility measure generated from a GARCH (1,1) model provides
more consistent results in terms of signs and sizes of the estimated coefficients than those
from other volatility measures.
While VEC models provide results supporting the findings of the cointegration
results on the effects of exchange rate and its volatility on trade flows, the short-run
elasticities are smaller than the corresponding long-run elasticities. So, the results satisfy
the Le Chatelier Principle. The free trade agreement seems to have played a significant
role in shaping the trade flows of tomato and maize, but not in the case of sorghum and
milk powder.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
A C K N O W LE D G EM E N TS
First of all I thank God for helping me to complete this dissertation. I would like to
extend sincere thanks to my wife Adriana Palacios and my daughter Adriana Jaramillo
who supported me emotionally and took care of most family issues during my doctoral
studies at the University of Guelph. Your endless assistance, love and care are very much
appreciated. Without your help, I could not have accomplished my degree. I would also like
to thank my parents, Roberto and Maria, for their love and teachings of being hard working
and perseverant to be successful in life.
I express my gratitude to my advisor, Dr. Rakhal Sarker, who was not only an
advisor, but also a friend to me. I thank him for always being attentive, for his guidance,
encouragement and support in carrying out different phases of my PhD studies at Guelph. I
would also like to extend special thanks to the members of my advisory committee, Dr.
John Cranfield, Dr. Maury Bredhal, and Dr. Thanasis Stengos for their helpful comments at
various stages of this research. Thanks Dr. to Alfons Weersink for chairing my final oral
examination. Also thanks to Dr. Karl Meilke and Dr. Stephen Clark for serving as internal
and external examiners.
I wish to thank my friend Juan Cabas, a fellow PhD student for helpful discussions
and support. Finally, I am very grateful to my sponsors, El Consejo Nacional de Ciencia y
Tecnologia (Conacyt) and my Institution, El Colegio de Postgraduados, for their generous
support without which it would not have been possible to complete my PhD at the
University of Guelph.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
TABLE OF CONTENTS
Page
CHAPTER 1
INTRODUCTION
1.3 Purpose
1.5 Hypothesis
CHAPTER 2
LITERATURE REVIEW
11
2.1 Introduction
11
11
13
15
19
20
20
24
25
ii
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Page
2.3.1.4 Time Horizon of the Model
26
26
28
32
32
35
35
38
40
43
44
2.4 Summary
47
CHAPTER 3
CONCEPTUAL FRAMEWORK
49
3.1 Introduction
49
50
52
52
53
54
56
iii
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
57
Page
3.4.2 The Armington Trade Model
58
60
62
63
65
69
70
74
79
83
83
84
87
88
89
93
94
96
Exchange Rate
3.7.4.3 The Respond of Prices and Output to Changes in Exchange Rate
3.7.5 Empirical Implementation
97
98
100
3.9 Summary
108
iv
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
CHAPTER 4
Page
110
112
121
130
132
134
141
145
145
Exchange Rate
4.5.3.2 Absolute Percentage Change of the Exchange Rate
145
146
4.7 Summary
149
CHAPTER 5
RESULTS OF UNIT TOOT TEST AND VOLATILITY MEASURE
5.1 Introduction
150
150
152
152
154
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Page
5.2.1.3 Results of Unit Root Test for Com and Sorghum Models
155
155
157
157
159
168
5.4 Summary
172
CHAPTER 6
MAXIMUM LIKELIHOOD COINTEGRATION RESULTS
6.1 Introduction
174
174
176
181
182
182
185
187
188
189
191
vi
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
191
Page
6.4.2.2 Results of Cointegration Analysis for Maize Using V2
192
193
193
195
195
195
197
198
199
200
2001
201
203
204
206
207
208
211
212
214
6.5.3 Test on the Equality of the Coefficients of Prices and Exchange Rate
215
6.6 Summary
217
vii
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
CHAPTER 7
Page
219
219
223
228
230
235
7.5 Results of the VECM: The M exicos Imports Demand for Sorghum
7.5.1 Main Findings for Sorghum Models
7.6 Results of the VECM: The M exicos Import Demand for Milk Powder
7.6.1 Main Findings for Milk Powder Models
7.7 Hypothesis Testing
236
240
240
244
245
246
247
7.7.3 Test of the Equality of the Coefficient of Foreign Prices and Exchange
248
Rate
7.8. Comparison de Short-Run Results
249
7.9 Summary
252
viii
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
CHAPTER 8
Page
CONCLUDING REMARKS
8.1 Introduction
254
257
259
264
265
266
References
268
Appendices
285
ix
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
LIST OF TABLES
Page
Table 2.1
12
US dollars)
Table 2.2
17
Dollars)
Table 2.3
18
Dollars)
Table 3.1
74
Table 3.2
102
Table 3.3
104
Table 3.4
106
Table 3.5
Table 4.1
122
Testing Cointegration
Table 4.2
Table 5.1
Unit Root Test Results for US and Mexico Macroeconomic Variables 153
Table 5.2
155
Table 5.3
156
Table 5.4
156
Table 5.5
161
Table 5.6
164
Table 5.7
167
Table 5.8
167
Table 5.9
169
Table 5.10
171
Table 6.1
177
Table 6.2
178
Table 6.3
179
Table 6.4
180
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
148
Page
Table 6.5
211
Table 6.6
213
215
Table 7.1
225
Table 7.2
226
Table 7.3
231
Maize
Table 7.4
Error Correction Model Results for the Mexicos Import Demand for 233
Maize
Table 7.5
237
Sorghum
Table 7.6
Error Correction Model Results for the Mexicos Import Demand for 239
Sorghum
Table 7.7
242
Table 7.8
243
Table 7.9
246
247
248
xi
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
251
LIST OF FIGURES
Page
Figure 2.1
12
Figure 3.1
64
Figure 3.2
73
Figure 5.1
158
163
165
Figure 5.4
166
Figure 5.5
172
xii
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
CHAPTER 1
INTRODUCTION
Mexico is an important partner in the North American Free Trade Agreement (NAFTA).
However, unlike Canada and the United States, agriculture in Mexico is less developed.
Agricultural transformation is still underway and domestic production of major staples
such as white com and beans are still dominated by small and mid-sized farmers. The
agri-food sector employs about 25 percent of the Mexican population. About 50 percent
of farmers in Mexico produce for self-consumption and 80 percent of grain and oilseed
farmers possess less than 5 ha. (Yunes-Naude, 2002).
Since the early 1980s, Mexico has initiated significant re-orientation of various
economic sectors and has re-instrumented its development policies. Chronic imbalances
in public finances, growing threats of inflation and skyrocketing external debt during the
early 1980s forced the government of Mexico to initiate economic liberalization (Bank of
Mexico, 1996). This economic liberalization was reinforced by other developments. In
fact, the Mexican economy and its agricultural sector were influenced by four key
factors: (i) the unexpected devaluation of the Mexican peso in 1982; (ii) progressive
liberalization of trade with the accession of Mexico to the General Agreement on Trade
and Tariffs (GATT) in 1986; (iii) the conclusion and implementation of North American
Free Trade Agreement in 1994; and (iv) the Mexican peso crisis of 1995 (Mora-Flores,
et. al., 2002).
The decline in oil prices during the early 1980s coupled with rising foreign debt
eventually led to the Mexican peso crisis in 1982. This crisis forced the Mexican
Government to adopt a new development strategy focused on opening and liberalizing
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
most economic activities and trade. The new trade policy was implemented in 1985 and it
was reinforced institutionally in 1986 when Mexico formally joined the GATT as a
member (Bank of Mexico, 1996).
The importance of trade in the Mexican economy has been growing since it joined
the GATT in 1986. The trade sector received a significant boost in 1994 with the
successful implementation of NAFTA. The share of agri-food trade relative to total
agricultural supply in Mexico almost doubled from about 19 percent during 1990-93 to
35 percent during 1994-2002. Agricultural trade between Mexico and the United States
increased significantly since the implementation of NAFTA. W hile the volume of
Mexican exports of major fruits and vegetables grew by 75 percent to 100 percent under
the NAFTA, imports of grains such as com, soybean, wheat and sorghum increased by
over 80 percent (Yunez and Barceinas, 2002).
Due to the growing dependence on trade, the Mexican agri-food sector is
becoming more vulnerable to changes in exchange rate of Mexican peso relative to the
United States dollar. Also important in this context are the changes in exchange rate
volatility which have been documented by the Bank of Mexico since 1995 (Bazdrech,
2002).
The importance of the effects of exchange rate changes on agriculture and
agricultural trade was first emphasized by Schuh (1974) who argued that changes in
exchange rates have a large impact on the price received for farm products and the price
paid for imported tradable inputs. His analysis drew attention to the importance of an
overvalued exchange rate on agricultural exports from the United States, during the
1970s. A number of empirical studies measuring the effects of changes on exchange rate
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
in agricultural trade followed. Kost (1976) and Chambers and Just (1981) found that
exchange rate fluctuations have significant impact on exports and domestic use of
commodities in the United States. Coleman and Meilke (1988) found that the devaluation
of the Canadian dollar led to a significant increase in net exports of beef and a small
increase in net exports of pork from Canada to the United States.
Since the beginning of the floating exchange rate regime the volatility of the
exchange rate has increased and the potential impact of exchange rate volatility on trade
was recognized. Economic theory suggests that unexpected changes in the exchange rate
volatility adversely influence the decisions made by risk-averse commodity traders.
Ethier (1973) showed theoretically how uncertainty about an exchange rate would reduce
the volume of trade, and Hooper and Kohlhagen (1978) tested empirically the effects of
volatility on trade. The results, however, are mixed.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
while the misalignment problem has not been dramatically mitigated under the new
system, significant increases in volatility created additional problems (Bazdrech, 2002).
Since the economic and trade liberalization took place in the mid 1980s in Mexican
agricultural trade, both exports and imports have grown dramatically in terms of volume
and value.
It has been argued that exchange rate played an important role in the performance
of Mexican agricultural trade since the 1980s. Some studies of United States-Mexico
agricultural trade have generally emphasized the importance of NAFTA in the dramatic
increase in exports and imports (Rosenzweig, 1996; Schwentesius, et. al., 1996; and
USDA-ERS, 1999). Others have emphasized the role played by exchange rate changes in
explaining Mexican trade performance (Diaz-Garces, 2002; and Mora-Flores et. ah,
2002). Since the shift from the fixed to a floating exchange rate system in 1995, the
Mexico-United States exchange rate has also been characterized by unexpected periods
of calm followed by episodes of high volatility.
While the border between Mexico and its important trade partners, the United
States and Canada, became increasingly open due to NAFTA, Mexican agri-food trade
has also been influenced by changes in exchange rate and exchange rate volatility.
However, it is yet to be determined the extent to which NAFTA, the continuous
devaluation of the Mexican peso, and the exchange rate volatility contributed to the
growth in Mexican agri-food trade during the last decade.
Unexpected changes in exchange rates influence the decisions made by riskaverse traders and may act as an impediment to international trade (McKenzie 1998). In
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
the short-run, exchange rate risk may be hedged. In the long-run, however, international
traders are exposed to higher and possibly unhedgable exchange rate risk.
The available literature on Mexican exchange rate and agricultural trade
(Espinoza-Arellano 1998; Malaga et al 2001; Diaz-Garces, 2002; and Mora-Flores 2002)
provide evidence that changes in exchange rate do have significant effect on trade. This
is surprising given the gradual devaluation of Mexican peso since the 1980s, moreover,
no research has been conducted to investigate the effects of exchange rate volatility on
Mexican agri-food trade. This issue is also very important for Mexican traders, agri-food
policy makers and agri-food producers in Mexico since evidences exist that a floating
exchange rate regime has been characterized by unexpected changes in exchange rate,
which may have increased the levels of risk traders face (Bazdrech, 2002). If individuals
are risk-averse, they would be willing to incur an added cost to avoid this risk. So, if risk
is not hedged properly, it acts as an implicit cost.
A study on the extent to which changes in exchange rate and its volatility affect
Mexican
agricultural
trade
would
provide
valuable
information
to
producers,
international traders and policy makers in Mexico. They might use such information to
devise more appropriate marketing strategies to reduce risk and improve competitiveness
of Mexican agri-food products in the international market.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
of Mexico in this context has been remarkable not only in manufacturing but also in agri
food products.
Krueger (1999) and Konno et. al. (2003) investigated whether NAFTA
significantly influenced bilateral trade between the United States and Mexico. They
found that NAFTA had no additional impact on the long-run trade relationship between
these two countries because a gradual switching was in progress since 1985. They argued
that as market integration under the NAFTA continues, Mexican exports would become
even more sensitive to changes in economic conditions of the United States economy in
the future.
Since the end of the 1970s when the world markets became increasingly open,
several theoretical and empirical studies have focused on the effects of both changes in
exchange rate and changes in exchange rate volatility on agri-food trade flows.
Unfortunately the results have been mixed and controversial as both positive and negative
effects of exchange rate on trade have been found in the trade literature.
Empirical studies on the effects of exchange rate movements on Mexico-Unites
States agri-food trade have either theoretical or empirical limitations. Existing studies
considered just estimation of either an import demand or an export supply function to
determine the effects of exchange rate movements on the volume and the value of trade
flows and produced negative results. They focused on the short-run effects of exchange
rate on aggregate trade and ignored the effects of exchange rate volatility on trade. The
previous studies have ignored the time series properties of the macroeconomic data used
in the estimation. Therefore, the results from these studies could be spurious. In view of
these limitations, an attempt is made in this research to perform a more comprehensive
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
study of the effects of exchange rate changes and its variability on Mexico-United States
agri-food trade. In particular, this study deals with an investigation of the nature and
extent of the influence of changes in exchange rate and exchange rate volatility on
Mexican agri-food trade flows with the Unites States the last decade.
1.3 Purpose
The purpose of this study is to quantify the effects of changes in the Mexico-United
States exchange rate and its volatility on the Mexico-United States agri-food trade flows
using the Maximum Likelihood Cointegration analysis.
1.4 Specific Objectives
The specific objectives of this study are as follows:
1.
2.
To determine the long-run and short-run effects of changes in exchange rate and its
volatility on Mexico-United States agri-food trade. The Johansens Maximum
Likelihood Cointegration Analysis is used to examine the long-term effects of
exchange rate variability on trade. However, the short-term variability is assessed
using the vector error correction model (ECM).
3.
To determine if real exchange rate volatility has an adverse effect on the volume of
the Mexico-Unites States agri-food trade flows using a set of four volatility
measures.
4.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
1.5 Hypothesis
To accomplish the above objectives, the following set of hypothesis stated in null form
was tested in this study:
1.
2.
3.
Exchange rate volatility has an insignificant effect on the volume of the MexicoUnites States agri-food trade flows in the long-run.
4.
Exchange rate volatility has an insignificant effect on the volume of the MexicoUnites States agri-food trade flows in the short-run.
5.
The coefficients of exchange rate and foreign price is equal (i.e., it is not
appropriate to use exchange rate as a separate variable in the trade model).
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
rate determination and the relationship between exchange rate and international trade.
The theoretical trade model developed in this research is also presented in this chapter.
Chapter 4 is an exposition of the empirical model and data issues. Specifically, the
equations that make up the econometric model, specifications of variables, sources of
data, procedure to test for the presence of unit root in each data series, the Johansens
Maximum Likelihood Cointegration approach and the specification of the Error
Correction Model (ECM) are all outlined in this chapter. The last section of this chapter
deals with the procedures followed to derive four alternative measures of exchange rate
volatility used in this study. Chapter 5 presents the empirical results of unit root tests
pertaining to individual series as well as the empirical results from the four different
approaches to estimate the volatility of the exchange rate. Chapter 6 presents and
discusses the results of the Maximum likelihood cointegration analysis. Chapter 7 deals
with the results of the short run dynamic relationships estimated from the Error
Correction Models. Chapter 8 provides a summary of the major findings of this study,
highlights some limitations and offer suggestions for further research in this area.
10
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
CHAPTER 2
LITERATURE REVIEW
2.1 Introduction
This chapter provides a brief overview of the Mexican economy and its agri-food sector
and a critical assessment of the relevant published literature on the effects of exchange
rate movements and trade flows. First a background of the most recent developments in
the Mexican economy, Mexican agri-food sector and trade policy is provided. This is
followed by a revision of the theoretical literature on the relationship between exchange
rate movements and trade flows. The third section deals with the empirical published
research on the effects of exchange rate movements on international trade flows. This
section is divided in two sub-sections: the effects of changes in exchange rate on
agricultural trade and the effect of changes in exchange rate volatility on trade flows. The
fourth section is devoted to studies of Mexico-United States agri-food trade. The chapter
is concluded with a summary of the main points and synthesis.
11
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
maintain price stability and economic growth. This reform together with the above
changes positively affected the growth of the Mexican economy and the international
trade performance of Mexico (See figure 2.1).
35000 -i
30000 O 25000 ^
20000
15000 -
**
10000
5000 LO
CO
CD
CO
hCO
CO
CO
O)
CO
O
o>
*
05
CM
co
a>
0) 0
o
0
o
CM
cm
CM
CM
o o
o
o
Years
Total Exports
Total Imports
During the 1987-1994 period the exchange rate regime evolved from a fixed rate
until 1988, to a transitory crawling peg and eventually, from January 1991 to 1994, to the
adoption of a band that was widened gradually.
inflows and financial liberalization took place in Mexico, but the current account deficit
grew quickly from 6 billion dollars in 1989 to 20 billions dollars in 1993 and the
overvaluation of the domestic currency, gave away to the balance of payments problem
and the financial crisis of 1994-1995.
12
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
13
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
14
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Mexico can impose TRQs on the imports of barley, dry edible beans, maize and
powdered milk, this right was never exercised for maize. The United States imposed
seasonal tariffs as well as TRQs for several fresh vegetables and fruits imported from
Mexico. Quota levels were established based on 1989-91 trade flows between Mexico
and its two North American partners. In 1994, the TRQs were set at 2,500,000 metric
tons (Mts.) for United States maize and 1,000 Mts. for Canadian maize. Although the
above-quota base or consolidated tariff on maize from both countries was fixed at 215
percent (or 206.4 US$/Mt.), Mexico did not impose any TRQ on maize imported from
the United States.
15
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
movement of the exchange rate will have a strong influence on trade outcomes. After the
trade liberalization, the participation of Mexico in the international market has increased
remarkably, not only in manufactured products but also on agricultural commodities
(Table 2.1).
Table 2.1 Mexican Exports and Imports by Economic Activity (Million US dollars)
1992
1994
1995
1998
2002
1985
1990
2 000
Exports
Agriculture and Forestry
1184.5
1720.7
1679.3
2220.9
3323.4
3435.7
3615.4
3285.6
224.4
441.7
433
457.3
692.8
360.9
601.6
606.3
Mining
13818.9
9537.5
7775.7
6994.4
7974.8
6865.4
15427.4
13509.9
Manufactures
6427.9
14861.2
36168.8
51075.3
67382.9
106550.4 146497.4
143185.2
8.1
277.4
138.6
134.3
167.6
247.4
183.7
Total FOB
21663.8
26838.5
46195.6
60882.2
79541.6
313
Imports
Agriculture and Forestry
1296
1829.9
2402.1
2993.3
2478.8
4280.6
4304.9
4872
311
241.1
456.3
378.1
164.9
492.2
493.9
503.4
Mining.
212.8
388.8
520
438
600.5
916
1325.6
1966.9
Manufactures.
12581.9
28523.1
58235.2
74424.8
67500
116431.2
165135.6
160613.9
131.4
289
515.9
1111.6
79345.9
1708.9
3252.8
3196.7
713.7
72453
125373
174457.8 168678.7
Total FOB
14533.1
31271.9
Source: Ministry o f Economy-Bank o f Mexico
62129.3
16
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
exception was 1995 which Mexico reached a trade surplus of 844.6 millions. According
to Calva (1996), this surplus was due to depreciation of the Mexican peso, as it changed
from $Mex 3.5/$US in 1994 to $Mex 7.5/$US in 1995.
Historically most of the Mexican trade occurred with United States and the
volume of trade has increased since the economic liberalization of the 1990s. The share
of total exports to North America rose from 85.6 percent in 1993 to 90.7 percent in 2002,
while Mexican exports to other destinations decreased slightly during the 1990s. The
most important agricultural commodities exported by value are fresh vegetables, coffee,
dry chickpeas, strawberries, watermelon and other fresh fruits. Malaga et al, (2001)
reported that vegetables consist of about 15 percent of total agricultural production and
employ nearly 18 percent of the agricultural labour force. Fresh vegetables are by far the
leading agricultural exports accounting for 48 to 65 percent of the value of Mexican
agricultural exports. The case of tomato is remarkable since tomato exports rose by 252
percent between 1990 and 2002 (Table 2.2).
1991
1992
1994
1995
1996
1998
2000
2001
2002
Coffee
368.1
258.1
359.7
706.2
677.2
617.3
613.7
211.2
176.7
Strawberries
20.7
12.2
31.5
43.9
53.4
68.5
50.3
42.6
79.7
Dry checkpeas
32
35.5
44.5
72.9
103.3
67.2
105.8
127.9
84.7
Tomatoes
261.8
166.8
394.4
585.7
539.9
589.3
462.3
532.5
661.1
Fresh Vegetables
489.3
550.9
689.4
929
742
1095.7
1352.2
1421.6
1348
W ater melon
142.3
89.4
89.1
114.4
128.3
135.6
156.7
157.9
129.1
319.4
350.2
401.7
418.3
481
486.3
486.9
530
Others
1432.3
1958.8
2853.8
2662.4
3054.6
3227.3
2980.6
3009.3
Total
1876.8 1679.3 2220.9 3323.4
Source: Secretaria de Economia-Bank of Mexico
3197.3
3435.7
3615.4
3325.9
3408.1
1597.8
17
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
In case of imports, Mexico decreased their imports in term of value from United
States from 71 percent in 1993 to 66 percent in 2002. Imports from the rest of the world
increased slightly. Fresh vegetables, live-cattle, dry chick peas, avocados, tobacco, coffee
and sugar account for about 85 percent of total agricultural exports in the 1990s, while
com, wheat, feed grains, potatoes, cotton seed and milk powder accounted for about 80
percent of total agricultural imports into Mexico. It is worth noting the remarkable
increase in imports of com and soybean since the 1990s. Imports of com increased from
178.5 million dollars in 1991 to 840 million in 2002, while the imports of soybean
increased from 348 million in 1991 to 891.8 millions in 2002. These two commodities
accounted for 35% of total agricultural imports into Mexico during the 1991-2002 period
(Table 2.3).
Table 2.3 Mexican Agricultural Imports by Main Commodities (Million US Dollars)
Commodity
1990
1992
1994
1995
1996
1998
2000
2001
2002
Com
426.1
155.7
337.3
361.0
1051.0
610.2
539.0
628.2
622.5
Soybean
217.4
512.1
640.4
542.3
897.5
861.4
782.9
851.2
891.8
Wheat
46.2
163.5
189.2
217.1
427.1
339.4
333.6
422.6
467.4
Sorghum
331.4
542.1
394.8
254.3
331.2
348.9
469.4
514.5
506.8
Milk Powder
295.4
271.3
194.7
211.4
299.5
166.4
233.0
308.6
200.0
Fresh Meat
69.9
236.0
269.8
81.28
144.8
446.1
712.9
831.6
953.3
Potatoes
2.06
6.50
16.42
13.12
15.10
26.27
43.68
48.89
57.79
Cotton Seed
8.51
34.68
31.61
19.39
23.03
29.71
46.67
50.31
44.16
2828.6
3161.6
3656.2
3743.9
Sub Total
1397.3
1922.0 2074.4
1700.1
3189.6
Source: Mexican Ministry of Economy and Bank of Mexico
Since 1994 Mexico adopted a floating exchange rate system and peso depreciated
significantly against the United States dollar. It has been argued in different studies
(Malaga et al, 2001; Mora-Flores, 2002) that depreciation of the Mexican peso has
18
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
contributed to the Mexican trade performance. Schwentesius and Gomez (1996) pointed
out that in Mexico the performance of agricultural trade can be better explained
considering the effects of devaluation on production cost and domestic and international
commodity price level. Furthermore, the agricultural imports of grains, meats and dairy
products increased in volume and value after 1995.
19
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Thursby and Thursby (1987); Cushman (1988); Anderson and Garcia (1989); Lastrapes
and Koray (1990); Chowdhury (1993); Qian and Varangis (1994); and Cho et al, (2002)
have found a negative relationship between exchange rate volatility and international
trade flows. On the other hand, studies by Franke (1991), Asseery and Peel (1991) and
Kroner and Lastrapes (1993) suggest that a positive relationship exists between exchange
rate volatility and trade flows. Controversial empirical findings of the effects of exchange
rate volatility on trade may be associated with the estimation method employed, the use
of nominal versus real exchange rates, and the specification of the volatility measures
used in these studies.
20
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
than a risk-neutral firm. As a consequence, the export supply curve will shift up and to
the left and this movement implies a reduction in the volume of trade.
Ethier (1973) found support for the negative effect of exchange rate volatility on
trade flows. He examined the effect of the forward exchange rate on international trade
and the relationship between the market for goods and the forward exchange market.
Ethier represented the exchange rate risk as the standard deviation of the spot exchange
rate and modeled the decision of a risk-averse importers regarding both the amount of
forward exchange rate cover and the volume of goods to be imported. He argued that (i)
to the extent that firms know how their revenues depend on the future exchange rate;
exchange rate uncertainty influences only the degree of forward cover and not the level of
trade; (ii) the pattern of trade will be inefficient to the extent that the forward rate does
not accurately reflect traders expectations concerning the future behaviour of the spot
rate; and, (iii) uncertainty about how the firms revenue depends upon the future
exchange rate will cause the volume of trade to become sensitive to exchange
uncertainty, which reduces the level of trade and will increase the terms of the trade of
expected profit for a reduction in risk.
While the Ethier (1973) and Clark (1973) models are static in nature and assume
fixed production, Kawai (1983) studied the behaviour of a competitive firm in a dynamic
context. The firm is engaged in domestic production and foreign trade and financial
transactions in domestic and foreign currencies. Since the firm faces uncertainty about
the future exchange rates and future prices of output and inputs, it attempts to hedge
against exchange rate risk if it is risk-averse. He argued that a risk-averse firm hedges in
the forward exchange market against not only the exchange risk but price risk as well. If
21
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
the forward exchange market is unavailable, the firm can still use the international
financial market for hedging. If the firms expectations about the future spot exchange
rate are different with respect to the forward exchange rate or interest rate parities, its
decisions on production and investment becomes inefficient. If the firm has access
neither to the forward exchange market nor the international financial market, an increase
in exchange rate uncertainty will curtail the firm s optimal investment and foreign trade
in the long run.
Demers (1991) showed basically the same results as Ethier (1973) but instead of
risk-averse traders, he assumed a risk neutral competitive firm which is uncertain about
its demand but has prior beliefs which it updates each period by Bayes law after
receiving information signals. He builds upon Arrow (1970) by demonstrating that the
irreversibility of investment in physical capital in conjunction with the anticipation of
receiving information and of learning the true state of demand lead to (i) lower
investment levels than otherwise since the firm can not disinvest if market conditions turn
out to be less favourable than anticipated; (ii) a time varying risk premium, or marginal
adjustment cost; (iii) a gradual adjustment of the capital stock to the desired level. He
argued that in an uncertain world the irreversibility of investment in physical capital lead
to reduced production levels and so quantity traded over time.
Frankel (1991) also analyzed the exporting strategy of a risk neutral firm. The
firm was assumed to operate in a monopolistically competitive environment and
maximize the net present value of expected cash flows from exports, which was specified
as an increasing function of the real exchange rate. The export strategy was determined
by the transaction costs incurred whereby the firm weighs the entry/exit costs associated
22
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
23
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
asset in their model is a nominal unhedged trade contract, which contains a risk element
in the form of an exposure to changes in the exchange rate. Their analysis examines a
single individual who consumes as well as imports and exports both available goods (no
production is incorporated in the model). The results of this model indicate that an
increase in the riskiness of the return on these assets depends on the nature of the risk
aversion parameter assumed. Assuming a convex function, an increase in risk increases
the level of exports. This result is found to be robust to the presence of a forward market
with non-zero transaction costs and the introduction of production.
24
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
On the other hand W olf (1995) considered a utility maximizing and risk-averse
competitive import agent. This agent is assumed to face uncertainty about not only
changes in the exchange rate during the period in which decisions are made (measured as
the variance of the exchange rate) but also uncertainty about the prices of imported
commodities. Faced with a multiplicative risk structure, W olf proves explicitly that the
variance of the exchange rate has a negative effect on the level imports. It is also shown
that demand by the importing firm of inputs and hedging instruments is less compared to
derived demand when the input price is known.
In a similar fashion, Gagnon (1993) modeled the response function of traders who
buy goods in one country and sell in another to maximize discounted expected future
utility. These traders are assumed to be risk averse as well as face a convex cost structure
in response to adjusting the level of trade because of contracting and marketing costs. In
this model, both exchange rate variability (measured as the unconditional variance) and
exchange rate uncertainty (measured as the conditional variance) are shown to reduce
trade flows. If adjustment is assumed costless, then the former relationship is lost but the
latter remains with no changes.
25
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
of forward markets, an increase in exchange rate volatility reduces both imports and
exports. A forward market was then incorporated into the equations of this model. The
results show that the effect of a change in exchange rate volatility on imports and exports
are opposite to each other. This result is attributed to the fact that importers and exporters
are by definition, on opposite sides of the forward market as is their exposure toward
movements in the exchange rate. Hence, increased exchange rate volatility may enhance
or reduce trade flows depending on the net currency position of that country.
26
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
the prevailing exchange rate is crucial to international business and decision-making. The
difference in relative prices determines the flow of agricultural products and the patterns
of trade. Based on trade theory, the real exchange rate is viewed as the relative price of
traded to non-traded goods. Real exchange rate movements accommodate changes in
technology, income levels, or borrowing from abroad that require either higher or lower
relative price of non-traded goods to clear those markets.
Rosson (2001) points out that proponents of the current flexible exchange rate
system argue that since floating exchange rates are allowed to fluctuate according to the
supply and demand with a minimum of government intervention then they respond
quickly to changes in market conditions. Market disequilibria are less likely to occur
under a floating rate system and business and resource allocation decisions are made in a
timely, efficient manner. Flexible exchange rate regimes have been criticized because
they may be subject to destabilizing speculation. Some analysts argue that floating
exchange rates result in additional risk and uncertainty for market participants and that
while most of this risk can be hedged transactions costs are higher than under a fixed
exchange rate system.
The trade literature related to the effects of exchange rate movements on
agricultural trade focused on whether exchange rate matter to agricultural trade flows,
and if it matters, what is the magnitude of this effect on agricultural trade. Among the
proponents that exchange rates do not matter for trade performance are Batten and
Belongia (1986), Fuller et al. (1992), Bessler (1986), and Greenes (1975). Contrary to
this view, others have argued that exchange rate does matter and that it is a key variable
in explaining trade performance. These studies include Cho et al. (2002), Anderson and
27
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Garcia (1989), Chambers and Just (1981), Thraen, et al. (1992), Carter, et al. (1990),
Mora-Flores (2002), Collins, et al. (1980), and Guzel and Kulshreshtha (1995). M ost of
the research supporting that exchange rate matter has found a positive effect of exchange
rate on trade flows. As the theoretical and empirical literature on this topic evolved a
number of issues have come to the forefront (see appendix 1 to 4).
Schuh (1974) called attention to the relationship between exchange rate and
agricultural products. He argued that the overvalued dollar caused the decline in
agricultural exports due to their relative expense in other countries. The overvalued dollar
led to depressed prices and lower farm profits causing an under valuation of farm
resources and excess supply of output. He argued that changes in monetary policy
induced international capital flows, which in turn caused changes in the value of the
dollar. These changes in the value of the dollar had an impact on the level of imports and
exports. The net result of these changes was that agriculture, as an export-oriented
industry, must bear the majority of the burden caused by changes in monetary and fiscal
policies.
28
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
linkage of real exchange rate movements to underlying causes accounted for market
equilibrium of traded and non-traded goods and provided somewhat more support for real
exchange rate effects on agriculture.
Kost (1976) developed a simple but consistent theoretical framework to assess the
effects of exchange rate changes (devaluation or appreciation) of a countrys currency on
commodity trade. He traces the effects of changes in exchange rates on commodity
production, consumption, trade levels, and price for any two trading partners. Examining
the effects of a devaluation of the exporters currency which has the same effects as that
of an appreciation of the importers currency, Kost stated that either of these actions
cause an increase in quantity exported and prices in the exporter country, which causes an
increase in production and decrease in consumption in the exporting country.
Kost points out that there is an upper limit on how much price and quantity can
change in response to an exchange rate change. The maximum that price and quantity can
change is by the same percentage of the exchange rate change, and the price maximum
would only occur if the export supply curve is perfectly inelastic, while the quantity
maximum would ocuur only if the export supply curve in perfectly elastic. The impacts
of an exchange rate change on imports and exports depend on the magnitude of the
exchange rate change.
Vellianitis-Fidas (1976) tested the hypothesis that exchange rate changes have a
significant effect on the demand for Unites States agricultural exports. To test her
hypothesis, a cross-sectional study of the demand for Unites States agricultural exports
was performed for the period 1971-1973. Additionally, the exchange rate changes in
importing countries were examined to determine if changes in these rates explained
29
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
variations in imports over time, both from the Unites States and the world in the period
1954-1969. Her results supported the hypothesis that special characteristics of the
agricultural sector negate the effect of exchange rate changes in demand for Unites States
agricultural exports. Exchange rate was not significant in the wheat equation and not
important in the com and soybean equations. Almost none of the variation in changes in
quantities exported for 1971 to 1972 and 1972 to 1973 was explained by the variation in
the exchange rates.
An alternative approach to empirical modeling adopted the time-series analysis to
seek causal relationships and dynamic impacts from monetary indicators to agriculture.
Sims (1980) pioneered the use of small dynamic models without too many a priori
restrictions as an alternative to over-identified structures imposed either by traditional
Keynesians or by the neoclassical rational expectations models. W hile it was appealing to
think that monetary effects on agricultural prices and trade could be measured easily in
small dynamic models where they were important, it turned out to be a difficult task.
Orden and Fackler (1986), using the VAR approach, detected little effect of
money supply on real Unites States agricultural prices or export values. Shocks to
financial market variables such as a short-term interest rate or the exchange rate had
larger impacts. These shocks explained about 20 percent of forecast error variance for
exports and 10 percent for real agricultural prices one year ahead, and over 50 percent
and 25 percent respectively, for a three-year forecast horizons. An increase in the interest
rate or appreciation of the dollar had a depressing effect on agriculture.
Dorfman and Lastrapes (1996) implemented time-series methods to measure the
effects of monetary variables on agriculture. They imposed the long-run restriction of
30
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
31
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
32
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Collins, Meyers, and Bredahl (1980) used a simple analytical method to analyze
the impact of multiple exchange rate changes, rates of inflation, and trade restrictions on
real Unites States commodity prices. Their study included two steps: (i) an expression of
the short m n Unites States commodity price changes caused by both nominal exchange
rate changes and adjusted exchange rate and (ii) calculated annual changes in Unites
States prices of wheat, com, soybeans, and cotton attributed to exchange rate changes and
inflation rates in 1971-1977. They compared these estimated changes with observed
changes in prices to determine how large the exchange rate impact was on Unites States
agriculture. The authors noted that there were some other issues involved. The size of the
exchange rate impact depends on crop, year, country, governmental influence in markets,
elasticities, measured price variables and alternative prices considered. They argued that
if the exchange rate changes reflect only differential rates of inflation then under free
trade, nominal commodity prices change but the underlying supply and demand do not. If
the exchange rate is fixed, differential inflation rates cause supply and demand changes
and the use of nominal price insulation policies increase the impact of inflation and
exchange rate changes on Unites States export demand and real commodity prices
increase significantly.
Using monthly data for 1973-mid- 1980s, Chambers (1981) found some evidence
of a causal relationship between money supply and agricultural imports and exports,
while there was little evidence of a causal relationship between the interest rate and
agricultural exports and imports. He also tested the null hypothesis that money supply did
not cause the level of wheat exports for the period of 1892-1952 and found that there is
limited evidence of causality in this instance as well.
33
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Batten and Belongia (1986) provided empirical support for the exchange rates do
not matter position. They argued that the real stimulus for export demand comes from
income enhancements in importing countries. Looking at a changing monetary policy,
they found that neither monetary policy nor a government deficit has an impact on the
value of the Unites States dollar. Focusing on the factors that affect changes in real
exchange rate, they failed to discover any evidence that monetary policy or budget
deficits have had effects on the real value of the dollar.
Schwartz (1986) compared the effects of changes in exchange rate in a simple
competitive versus a noncompetitive market for wheat. In the simple competitive case,
under a floating exchange rate, a change in exchange rate in one country will cause a
short run adjustment in price, output, trade flows, market share of exports and export
volume for two countries competing with one another. The country that experiences
depreciation in currency value will see a decrease in exports and the other country will
see an increase in exports. The more the exchange rate fluctuates, the more variable
short-run changes in domestic prices and trade shares are.
Bradshaw and Orden (1990) tested the Granger Causality of exchange rates on
agricultural prices and exports. They examined the impact of the real agricultural tradeweighted exchange rate for monthly forecasts of real cash prices and export sales
volumes comparing out-of-sample forecasting performance of univariate models to
bivariate model. Bradshaw and Orden found that their bivariate models outperformed the
univariate models in statistically significant ways. They argued that model specification
and the choice between in sample and out of sample tests are important in determining
Granger Causality among the exchange rate, prices and exports sales.
34
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Dorfman and Lastrapes (1996) disaggregated Unites States agricultural data into
crop and livestock for the period 1952 to 1993 and used a Bayesian approach to test the
responses of agricultural prices to money supply shocks. The results suggest short to
medium-mn benefits to both sectors from an expansionary monetary policy.
35
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
reports nine different measures of volatility that have been empirically tested. Most
published research has tested a variance or standard deviation-based volatility measure,
specifically the moving average of the standard deviation (Cushman, 1983; Akhtar and
Spence-Hilton, 1984; Gotur, 1985; Bailey et al, 1987; Kenen and Rodrik, 1986; BiniSmaghi, 1991). Belanger et al (1992) proposed a non-parametric measure of volatility.
They argued that nonparametric measures are robust to assumptions regarding the
distribution of the data and offer additional insights than measures that assume Gaussian
errors. Recently ARCH and GARCH models have also been used to generate a measure
of exchange rate volatility (Mckenzie, 1998; Kroner and Lastrapes, 1989; Qian and
Varangis, 1994).
Akhtar and Spence-Hilton (1984) analyzed multilateral United States-Germany
trade over the floating rate period using a nominal volatility measure specified as the
standard deviation of daily exchange rate over a three month period. Their results reveal
that exchange rate risk reduces the volume of trade.
In contrast, some recent studies have rejected the hypothesis that exchange-rate
volatility has had a systematically adverse impact on trade. Gotur (1985) tests the
robustness of Akhtar and Hilton's empirical results and finds their methodology to be
flawed in that it included a nominal volatility measure. Gotur stated that over the medium
term, the real exchange rate is the more relevant measure because the effects of
uncertainty on a firms revenue and costs arise from fluctuations in the exchange rate are
likely to be offset in large part by movements in costs and prices.
Kroner and Lastrapes (1989) estimate the reduced form impact of exchange rate
volatility on international trade quantities and prices using a joint estimation technique in
36
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
37
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
agents. The results showed no stable economically and statistically significant negative
impact of exchange rate risk on Unites States imports from Canada. It was true for all the
sectors examined except for capital goods that accounts for only 12 percent of imports
from Canada. They results seem to challenge results found in some recent bilateral trade
studies such as Maskus (1987), Thursby and Thursby (1987), and Cushman (1988).
Peree and Steinherr (1989) estimated export equations on aggregate and bilateral
basis; the general functional form was quite traditional. They used data for the period
1960-1985 for five developed countries. Their paper focuses on the problem of
approximating
meaningfully
exchange
rate
uncertainty
without
tackling
other
shortcomings. In view of the interest in long run uncertainty they proposed a measure for
projections several years into the future. Variances over past periods are of very limited
relevance for appreciating uncertainty over periods of several years in the future.
Consequently, two measures of medium term exchange rate uncertainty were constructed.
One is a weighted function of the magnitude of past movements in nominal exchange
rates and the current deviation of the exchange rate from equilibrium. The second
depends on both the duration and the amplitude of misalignment from equilibrium
exchange rates. The empirical evidence suggests that when exchange rate uncertainty was
defined over a medium term period it does adversely affect trade flows of the country
under review, with the exception of the Unites States.
38
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
process of the traders. Some of the theoretical and empirical research in this area focused
on the use of nominal exchange rate volatility included Ethier (1973), Hooper and
Kohlhagen (1978), Koray and Lastrapes (1989), Anderson and Garcia (1989), and
McKenzie and Brooks (1997), while others focused on the use of real measures of
exchange rate volatility (such as Cushman (1986), Caballero and Corbo (1989), Assery
and Peel (1991), Chowdhury (1993), Lastrapes and Koray (1990), McKenzie (1998), and
Fabiosa (2002)). Under both specifications negative and positive effects of exchange rate
volatility on international trade have been reported.
Qian and Varangis (1994) provided evidence which suggests that the nominal and
real exchange rates have moved closely together during the floating period. Hence, the
authors argued that the distinction between real or nominal volatility should make no
difference to the results derived. Some authors have tried to test this proposition by
retesting their equations using alternate volatility measures. Thursby and Thursby (1987)
find that their results for nominal exchange rate volatility are indistinguishable from those
based on real volatility.
Bini-Smaghi (1991) supports this view of nominal measure of exchange rate
volatility by arguing that risk should be regarded as nominal rather than real exchange
rate risk as the latter depends not only on the variance of the nominal exchange rate but
on that of relative prices. The proxy for exchange rate risk was based on the standard
deviation of the effective nominal exchange rate index for each quarter. In the estimation
procedure, the author included up to eight lag periods on both the real exchange rate and
the risk proxy. The results indicate that exchange rate variability did serve to reduce
39
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
German bilateral trade in general and also Unites States exports. Other volatility
measures were also tested and the results were found not to change significantly.
40
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
aggregation necessarily dampens exchange rate variability, which may make identifying
any true trade-risk relationship more difficult (Wang and Barret 2002). Moreover, where
different sectors have different conventions for contracting and delivery or payment lags,
intersectoral and intertemporal aggregation together could necessarily mute real traderisk effects.
The literature indicated that many of the studies in this research area have
undertaken an examination of aggregate trade flows. These studies implicitly assume that
the impact of exchange rate volatility is uniform between countries and across
commodities both in terms of direction and magnitude. If this assumption is incorrect,
then, the examination of aggregate trade data is likely to dilute the true nature of the
relationship and lessen the probability of deriving a significant result. Recognising this
possibility researchers in recent years have adopted trade models which focus on
disaggregated trade data in the form of bilateral and sectoral trade flows.
In this spirit, Klein (1990) analyzed the effects of real exchange rate volatility on
specific categories of bilateral exports by the United States over the period 1978 to 1986.
The volatility measures used the standard deviation of the monthly percentage change in
the bilateral real exchange rate. The importance of the nature of cost and demand
functions in determining the effect of exchange rate volatility on trade suggests
undertaking an analysis disaggregated by industrial sector while differences across
destination markets in the behavior of exchange rate volatility calls for an analysis of
bilateral, as opposed to multilateral, trade flows. Kleins results supported the hypothesis
that volatility affects differentially by category of exports. It is found that real exchange
rate volatility significantly affects the value of United States exports for six of the nine
41
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
sectors studied and in five of these sectors the effects are positive. These results suggest
that real exchange rate volatility may stimulate export supply by risk-neutral firms
through its effects on their expected profits.
Hooper and Kohlhagen (1978) adopted the Ethier (1973) model of market
equilibrium for traded goods and extended it by relaxing the assumption of infinitive
elasticity of the export supply function. Hooper and Kohlhagen formally tested the
impact of volatility on trade flows and prices using quarterly multilateral and bilateral
trade flow data for Germany, Japan, Canada, France, the United States and the U.K. for
the period 1965-1975. When aggregated trade data were the focus of the model, no
significant impact of volatility on volume or price was found. However, for bilateral trade
flows, the authors found evidence that exchange rate uncertainty has a significant impact
on prices but not significant impact on volume of trade.
McKenzie (1998) analyses the impact of exchange rate volatility on both
aggregate and sectoral trade flow data for the Australian economy over the period 19881995. ARCH models were used to generate a measure of exchange rate volatility and
then volatility was tested in a standard model. In this Model he specified imports or
exports as a function of prices, income, the exchange rate and its volatility. He concluded
that in terms of aggregate, volatility of the exchange rate negatively affected Australian
imports. These results indicates that the aggregate nature of trade data may dilute the
effects of volatility and also that the effects of volatility may potentially be sensitive to
the nature of the industry producing the traded commodity.
42
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
43
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
44
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
imports increased by 11, 7, and 2 percent respectively. They argued that exchange rates
do have an important effect on trade, in particular, the weakening of the exporters
currency increases export opportunities in the short run.
Kapombe and Colyer (1998) used a multiple equation structural time series model
as the theoretical framework for supply and demand equations to analyze Unites States
broiler exports. They used quarterly data from 1970-1995 for Unites States broiler
production, demand, inputs, exports, meat prices and annual values for broiler import
equations for Mexico, Japan, Hong Kong and Canada. The results indicated that Mexican
broiler demand is negatively influenced by the exchange rate, as a one percent increase in
the peso-dollar exchange rates will result in a 0.58 percent decrease in import demand.
Japanese demand for Unites States broiler exports were negatively influenced by the
exchange rate, a one percent increase in the yen-Unites States dollar exchange rate will
cause a 0.96 percent decrease in import demand. Also Hong Kong import demand is
negatively influenced by the exchange rate.
Diaz-Garces (2002) examined the performance of the Mexican trade flow for the
period 1980 to 2000, which includes important institutional changes such as accession of
Mexico to GATT and the implementation of the NAFTA agreement. Dias pointed out
that the remarkable performance of the Mexican trading sector has been frequently
discussed in political sectors but serious studies on the topic have been rather scarce. He
estimated long-run import and export demand functions for the period 1980-2000 using
co-integration analysis. He found that the export equation, as a function of the Unites
States industry output and the real exchange rate, is stable for the period 1990-2000. The
import demand function, as a function of the Mexican industrial output, real exchange
45
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
rate, and exports, showed instability for the same period. It was not possible to find a
unique cointegration vector for the whole sample period because of the structural change
in 1989-1990.
Konno and Fukushige (2003) assesses the long-run bilateral trade relations in the
import function between the US and Mexico by applying the Dynamic Ordinary Least
Squares (DOLS) method proposed by Stock and Watson (1988) together with a gradual
switching model for each countrys long-run import function. They used domestic
income, domestic price, foreign price and exchange rate as determinants of bilateral
import functions. The general foreign export price is used as the foreign price because
exact export prices for each country were not available. The data used in this paper was
quarterly data from 1981(1) to 1999(4). Their results suggest that no additional impact
was caused by NAFTA in either import functions when NAFTA came into effect. There
was a gradual shift in the long-run relation for each import function much earlier than the
operation of NAFTA. The estimated result also suggest that Mexican exports would
become more sensitive to the condition of the Unites States economy through gradual
increases in income elasticity and the domestic price elasticity of the Unites States
imports, while Mexican exports to the Unites States still account for about eighty percent
of the total Mexican exports.
Mora-Flores et al (2003) assessed the effect of changes in export prices on
Mexican crop and livestock gross value. Changes in quantities due to variations of prices
were modeled by means of a Cobb-Douglas price function while the total change in
exports gross value was estimated using the logarithmic differential between both
variables. They used monthly data from 1998 to 1996. A direct relationship between
46
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
export price and export quantities was found. The response for cattle exports was greater
than the response for other agricultural commodities, in terms of price elasticity. A
devaluation of the exchange rate implies higher prices for importers and higher gross
value for exports.
2.4. Summary
The economic and trade liberalization undertaken by the Mexican government during the
1980s and 1990s positively the growth of the economy and the Mexico-United States
trade flows of agri-food products has grown twice since 1990. To explain the
performance of the Mexican agricultural trade, different hypothesis have been advanced
such as the contributions of NAFTA and the peso devaluation of 1994-1995 but no
consensus exist on this respect. The review of the published literature on the effects of
changes in the exchange rate and its volatility on agricultural trade reveals some
interesting features; first, most of the previous studies have ignored the time series
properties of data. Thus, these studies may be subject to spurious relationship bias
(Granger (1981), and Engel and Granger (1987)). Second, different data sets have been
used by different studies, so the results of the effects of exchange rate on trade are not
comparable among studies. Third, some previous studies have concentrated on aggregate
trade flows that essentially mask the effects of exchange rate on a particular sector and no
attention on disaggregated trade flows or particular groups of commodities. Finally, the
focus of most empirical studies has been on the specification of a standard deviation or
variance-based measure of exchange rate volatility.
47
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
The review of the literature also revealed that a number of empirical studies
dealing with effects of exchange rate changes on trade flows have explicitly focused on
the Mexico-Unites States trade but no conclusive evidence is provided to explain the
Mexico-Unites States trade performance. Finally, none of the studies investigated the
effects of exchange rate volatility on Mexican trade flows and the specification of the
exchange rate volatility remains an unsolved issue.
48
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
CHAPTER 3
CONCEPTUAL FRAMEWORK
3.1 Introduction
The purpose of this chapter is to develop a theoretical framework to be used to explain
the effects of changes in exchange rate and its volatility on international agri-food trade
flows. This framework is then used to develop an empirical specification of the trade
model used to determine empirically the effects of changes in exchange rate and its
volatility on the Mexico-United States agri-food trade flows.
The first part of this chapter presents a brief overview of the modem theory of
exchange rate determination based on the monetary approach. In this framework,
exchange rates are determined as part of the general theory of asset price determination.
This is followed by a short discussion of the main trade models in the literature and the
role of the exchange rate on international trade. The second part of this chapter consists
of two sections: the first section describes the microeconomic foundations of the
theoretical model and the derivation of the import demand functions. The second section
deals with the derivation of comparative statics results from the theoretical model used in
this research and interpretation of these results. The third part describes observed
peculiarities
in
specific
commodities
(i.e.
specific
institutional
and marketing
49
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
50
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
51
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
of changes in exchange rate and its volatility on Mexican agri-food trade. However, for
the sake of better understanding, this section provides a brief exposition of two main
theories of exchange rate determination; (i) the monetary approach (the flexible-price
model and the sticky-price differential formulations) and (ii) the portfolio balance
approach.
(3.1)
m f = p* +</>* y* - A* r * t
(3.2)
Equilibrium in the traded goods market ensures that there are no further profitable
incentives for trade flows to occur. This occurs when prices in a common currency are
equalized and PPP holds. That is, st = p t - p *. Where st is the logarithm of the nominal
52
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
exchange rate. Thus, if PPP holds continuously, the logarithm of the real exchange rate
q, say (qt = st - p t + p * )is a constant. The world price, p *, is exogenous to the
domestic economy, being determined by the world money supply. The domestic money
supply determines the domestic price level and hence, the exchange rate is determined by
relative money supplies. Algebraically, substituting equations (3.1) and (3.2) into the PPP
equation yields:
st = (m s - m s*)t - <pyt + <p* y* + Art - X * r*
(3.3)
Equation (3.3) is the basic flexible-price monetary model. The above equation says that
an increase in the domestic money supply, relative to the foreign money stock will lead to
a rise in st - that is, a fall in the value of the domestic currency in term of the foreign
currency. An increase in domestic output, as opposed to the domestic money supply,
appreciates the domestic currency. Similarly, a rise in domestic interest rates depreciates
the domestic currency.
53
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
54
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
determinant of the current account of the balance of payments. A surplus (deficit) on the
current account represents a rise (fall) in net domestic holdings of foreign assets, which in
turn affects the level of wealth, the level of asset demand, both of which affect the
exchange rate. Thus, the portfolio balance model is an inherently dynamic model of
exchange rate adjustment, which includes asset markets, the current account, the price
level and the rate of asset accumulation.
The portfolio balance model, like the sticky-price model, allows one to
distinguish between asset market short-run equilibrium and the dynamic adjustment to
long-run equilibrium. Unlike the sticky-price model, it also allows for the full interaction
between the exchange rate, the balance of payments, the level of wealth and stock
equilibrium. In this model the exchange rate is determined solely by the interaction of
supply and demand in the asset market in the short run.
The key feature of this model is that it assumed imperfect substitutability between
domestic and foreign assets. In this model, the net financial wealth of the private sector
(W) consists of money (M), domestically issued bonds (B) and foreign bonds
denominated in foreign currency (B*). Then, the definition of wealth is W= M + B +
SB*. With foreign and domestic interest rates given by i and i* and Se the expected rate
of depreciation of the domestic currency, the demand functions are defined as follow:
(3.4)
(3.5)
(3.6)
B* = T (S/P) + i*B*,
(3.7)
55
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
The rate of change of B* given by equation (3.7), the capital account, is equal to
the current account which is in turn equal to the sum of the trade balance, T(/) and net
debt service receipts, i*B*, which assumes static expectations and an open market
purchase of bonds by government. In order to induce agents to hold more money and
fewer bonds the domestic interest rate falls and, as agents attempt to compensate for the
reduction of their portfolios of domestic assets by buying foreign bonds, the exchange
rate will depreciate, driving up the domestic currency value of foreign bonds.
56
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
57
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
58
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
share and that expenditure elasticities are the same and unitary. These assumption are too
restrictive for meaningful empirical analysis.
A concern in using the Armington model is whether the restrictive assumption
should be imposed on data that do not support the assumptions. If the data do not satisfy
these assumptions, the estimated results would be biased. A number of studies re
specified the Armnigton model to derive more sensible results.
In general, a two-stage budgeting procedure assumes that consumers allocate their
total expenditures in two stages. In the first stage, total expenditure is allocated over
broad groups of goods, while in the second stage group expenditures are allocated over
individual commodities. It is well known that weak separability of the direct utility
function over broad groups of goods is a necessary and sufficient condition for the second
stage of a two-stage budgeting procedure (Deaton and Muellbauer, 1980). However,
weak separability imposes restrictions on consumer behaviour. First, the marginal rate of
substitution between two goods from the same group is independent of the consumption
of goods in other groups. Second, the substitution effects between goods in different
groups are limited. A price change of a commodity in one group affects the demand for a
commodity in another group only through the group income effect. Third, separability
implies a restrictive relationship between price and income effects.
The first two-stage of Armingtons framework is, in general, equivalent to that of
two stage budgeting process. That is, in the first stage the importer decides how much of
a particular commodity to import. In the second stage, given the total amount to be
imported, the importer decides how much to import from each supplier. Thus, the
59
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
where wt is the marker share of import from source i,bi is a constant, Pt is the price of
the commodity from the ith source, P is the import price index depending only on the
within-group prices and o is the constant elasticity of substitution parameter. The CES
specification implies weak separability between different import sources.
The Armington approach also assumes homotheticity of the sub-utility or withingroups utility functions. This implies that importers market share is independent of
group expenditures. Consequently, all expenditure elasticities within a group are equal
and unitary and import market shares change only in response to relative price changes.
Thus, the Armington framework implies that in the second stage (within-groups
allocations) market shares do not vary with expenditures and that different import sources
are separable (Alston et al., 1990).
success, it was also criticized because it lacked a coherent theoretical foundation. These
theoretical foundations were subsequently developed by Anderson (1979) and Bergstrand
60
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
(1985), who derived gravity models from models of monopolistic competition and
Deardorff (1998), who demonstrated that the gravity model can be derived from the
Heckscher-Ohlin framework. Typical gravity models contain the following components:
(i) economic factors affecting trade flows in the origin countries, (ii) economic factors
affecting trade flows in the destination countries, and (iii) natural or artificial factors
enhancing or restricting trade flows.
A gravity model is a reduced form equation generated from a partial equilibrium
model. The demand equation for the specific commodity is derived from utility
maximization using a constant elasticity of substitution (CES) utility function subject to
an income constraint. The supply equation is derived from the firms profit maximization
approach in the exporting countries with resource allocation assuming a constant
elasticity of transformation (CET) during the production process (Bergstran, 1989). The
commodity-specific gravity model, under market equilibrium conditions of demand and
supply systems, can be derived as follows:
X, = c c ^ Y f C f T ^ P ^ P f E f e ^
i = 1,2,
, N and j = 1,2,...., Af
61
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
62
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
the theoretical model based on the Batra and Ullah (1974) model is developed and fully
described.
63
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Assuming initially that the exchange rate between these two countries is at par
(i.e. one Mexican peso equal to one US dollar), as shown by the 45 line OA in panel (c)
of Figure 3.1, the volume of trade is equal to the distance OQo. From panel (a) the initial
price, Po is higher than the autarky price in Mexico. This creates an excess supply of a
of this good in Mexico. Since exchange rate is at par and that Po is lower than equilibrium
price in the U.S., the United States is willing to import from Mexico the same amount of
the commodity (a). If a change in exchange rate causes the value of the Mexican peso
to depreciate, the OA rotates to OA, the excess demand curve rotates upward from ED to
ED in panel (b). The new equilibrium is attained where ES intersects the new excess
demand (ED) function. The intersection of ES and ED" functions results in an increase
in the exporters price to OPi and an expansion of trade to OQi (panel b).
$MX
'Mex
ES
ED
ED
Mex
64
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
The depreciation of the Mexican peso relative to the U.S. dollar causes Mexican
commodities to become cheaper relative to U.S. products. At the new price Pi, U.S. is
now willing to import b units of the commodity from Mexico. As can be seen from
Figure 3.1, the trade equilibrium price lies between equilibrium price of the exporting
country and that of the importing country. The exact location of this price in the two
countries depends on the relative elasticities of the excess supply and demand functions,
which in turn depend on the elasticity of demand and supply in Mexico and in the U.S.
Domestic Demand;
(3.8)
SMex = c + dPMex
(3.9)
<0 + (b + d)P\iex
(3.10)
(b + d ) > 0
(3.11)
(3.12)
65
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
(3.13)
E D = D us - S us = e - f P u s - ( 8 + hPu s)
ED = ( e - g ) - ( f + h)Pus
(3.14)
dED,
'U S
(3.15)
=
-(f+h)<
dPuS
Pus =
(Price equilibrium)
(3.16)
ES = ED
(3.17)
where: Pmcx is price in Mexico (Mexican Pesos), Pus is price in U.S. (US dollars)and r is
the exchange rate between Mexico and U.S. currency. Substituting (3.10), (3.14) and
(3.16), into (3.17) gives
ES(PMex) = ED(PMex r ) .
(3.18)
V^Mex
dED
dED
dPUa = r dPMex + PUa dr
vPuS
vPuS
im
(3.19)
Equation (3.19) is then solved for the elasticity of the exporters price with respect to a
change in the exchange rate, E p r to obtain
p
Mex-r
dPMex
dr
ur
r
P
M ex
----e E S - e ED '
Similarly, the elasticity of the importers price with respect to a change in the exchange
rate, e p
is
= dP,JS J _
P u s ,r
Pus
,3 2 l)
es
ES
ED
66
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
where, e ED is the elasticity of excess demand and s ES is the elasticity of excess supply.
The elasticity of Mexican market price with respect to exchange rate (e Pu r ), as
well as that of U.S. market price (s p^ r ), depends on the elasticity of excess demand for
U.S. and that of the excess supply for Mexico. Since the Mexican supply elasticity is
positive and that of the U.S. demand elasticity is negative, it follows that Pu r < 0. The
effects of changes in the exchange rate on the quantity traded is found by taking the total
derivative of the U.S. (excess) demand and substituting result (3.19) result into an
elasticity expression:
dED r
dr ED
dP r
3 ED Pm
(3.22)
dr Pus dPus ED
(3.23)
(3.24)
Thus, the elasticity of the equilibrium quantity traded with respect to the exchange
rate ( s P r) also depends on the elasticity of excess demand for U.S and that of excess
supply for Mexico. Again, since e ES > 0 and s ED < 0 , then the demand and supply
functions are well behaved, s EDr < 0, (i.e. the elasticity of import demand with respect to
exchange rate will be negative).
From equation (3.20) one can deduce that a devaluation of the Mexican peso,
represented by a decline in r, will increase PmCx ranging between zero and the percentage
change in the exchange rate. But at the same time it will increase the quantity traded.
Given e ES = e ED ^ 0, it is clear from the above trade model that for a given change in
67
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
the exchange rate there is an impact on both price and quantity traded. This model is
simplistic in that both supply and demand are functions of only the commodity price. In
reality supply and demand are also influenced by other factors such as prices of related
goods, consumers income, tastes and technology. Due to theses limitations the above
model is not suitable for this research. Consequently, a more general model is developed
next.
To determine the effects of changes in exchange rate and its volatility on Mexican
agri-food trade, this study relies on standard international trade theory. To represent the
effects of exchange rate uncertainty, income and other prices on trade flows, it is assumed
that a risk-averse competitive firm is faced with uncertainty in the international market
and that it maximizes the expected utility of its profits (Ethier, 1973; Kawai, 1973; and
Batra andUllah, 1974).
The firm faces exchange rate uncertainty both through the domestic cost of
imports and through the value of exports. Since there is no forward market for foreign
exchange, the firm faces an exchange rate risk for future conversion of its sales revenues
from the importing country into the currency of the exporting country. If the firm is riskaverse, it would be willing to incur an added cost to avoid this risk, so that the un-hedged
risk serves as an implicit cost. Thus, in the presence of risk, the firms supply price at
each quantity would be higher than if there was no risk.
A few previous studies dealing with exchange rate effects on agricultural trade
have included a measure of volatility in their analysis. Inclusion of a volatility measure in
a trade model is likely to minimize bias in the estimated coefficients due to omitted
variables and allows the trade response to changes in exchange rate volatility. There is,
68
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
69
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Mexico is also the world leading importer of non-fat dry milk. The following section is
devoted to examining some distinguishing features of these commodities in Mexico, so
that some institutional features can be incorporated into the model.
70
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
71
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
production. Fresh tomato production has two perfectly differentiated destination markets;
80 percent of the production goes to the domestic market, 70 percent of this production is
marketed through three regional wholesale markets.
United States is the leading external supplier of tomatoes, onions, avocados,
lettuce, oranges and potatoes to Mexico (USDA-ERS, 2003). Mexico is normally
considered a source of these products, not a destination. However, Mexico is an
expanding market for fresh fruit and vegetables. While Mexico will supply much of the
growing demand itself, it will not supply all of it and certainly not all products. Given its
location, the United States is the natural supplier. California and Texas are closer to
Mexico City than to Chicago or New York. And northern Mexico, where demand is
growing the most, is even closer. Shipping produce from the United States to Mexico
should be no more remarkable than shipping produce within States. Indeed, as the two
economies become more closely integrated, channels of distribution have expanded and
barriers to free exchange are being dismantled (USDA-ERS, 2003).
Tomato exports from Sinaloa, Mexico, directly compete with tomato produced in
South Florida. Mexican shippers in Sinaloa produce mainly extended-shelf-life tomatoes
that are harvested as vine-riped that helped them to gain additional market share relative
to Florida growers over the last decade. On the other hand, the export production of Baja
California competes with the production of California that goes to the market from May
to December. Mature-green tomatoes are produced domestically in both Florida and
California. Florida tomatoes are shipped predominantly to eastern and mid-western
markets, while the western half of the country is served primarily by tomatoes from
California and Mexico (Sexton and Zhang, 2003).
72
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
nationally
and internationally
(Cook,
2001).
In the grower-shipper
combination the shipper often joint ventures with growers to obtain the necessary
production and then markets for a fee, often advancing cartons, controlling the harvest
operations and imposing grades and standards to generate a consistent quality. In fact, the
process from production regions in Mexico to consumption centres in the US the
commodity receives marketing services that add value to the final consumers of the
product. While sometimes multinationals act as the grower-shipper, several large
Mexican growers have expanded their operations in this way in recent years (Lacroix, et
al., 2001). To illustrate the above, Figure 3.2 describes the tomato supply chain.
Mexican
Growers
Broker/shipper
Exports
Wholesaler
Supermarkets
Terminal market.
73
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Fewer than a thousand farms dominate fresh tomato production in North America
and fewer than 50 shippers control the first marketing stage as tomatoes move into the
wholesale, retail and food service sectors. Often these shippers are also the growers. The
participation in the U.S. market by the main supplying regions is summarized in table 3.1.
Number of
Size (acreage)
Grower-Shippers
Shipments
California
15
650-6500
80
Florida
800-7500
75
Mexico
14
500-6000
60 (Exports)
74
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
75
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
crop industry. Many changes in production and consumption trends have their origins in
changes in international prices. The economic policy implemented in the 1980s and
continued in the 1990s has tended to link domestic prices to international prices so that
the latter may serve as a reference in the allocation of resources. According to the above
reasoning, the maize economies cannot be understood without examining what has
happened to world market prices of maize, wheat (a substitute in consumption), sorghum
(a substitute in production) and fertilizer.
The price of maize in real terms fell in the international market during the 1970s
and most of the 1980s. By 1988 prices stabilized and then began to show a slight
increase during the 1990s. W ith respect to the price of maize substitutes, during the 1980s
and 1990s, wheat has become more expensive relative to maize in the international
market while sorghum has remained more or less stable (Sainz and Lopez-Pereyra, 1999).
With respect to maize consumption, total consumption is the aggregate of two
principal components: direct human consumption and indirect consumption as a
component of livestock feed (most in poultry, egg and pork production). White maize is
used for direct human consumption, while yellow maize is used mainly for feeding
livestock in Mexico. Growth in total maize consumption results from growth in both
components.
Given the preferences of people in Mexico for white maize, which is produced in
limited quantities in other regions and traded only in very small quantities on
international markets, growth in industrial processing of white maize may represent a
powerful incentive for increasing maize productivity. The maize price and the prices of
substitute crops and other relevant factors are involved in determining the demand for the
76
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
final product (that is, for poultry meat, eggs and pork since the demand for grains is
regarded as derived demand). Since maize and sorghum are the main components (in
volume) of concentrated feeds for the poultry industry, demand for both grains increased
proportionally. Maize and sorghum volumes used in manufacturing balanced feeds
depend fundamentally on the domestic availability of sorghum and its relative prices.
Maize and sorghum both provide substantial energy to animal diets.
In Mexico, growth in productivity rose considerably over the last decade.
Together with moderate growth in cultivated area, yield growth contributed to M exicos
high growth rate in maize production. Compared with Latin American countries and the
world in general, maize production and yields in Mexico during the last ten years grew at
a higher rate. Maize production in central and southern Mexico (the states of Jalisco,
Mexico, Oaxaca, Veracruz, Tabasco, Chiapas, Campeche, Yucatan and Quintana Roo)
represents approximately 60 percent of the countrys total production (SAGARPA,
2004). Growth rates vary among the states of centre and southern Mexico, but maize
production in the area as a whole rose by 4.0 percent per year.
While yellow com accounts for the bulk of U.S. com production, white com
dominates production in Mexico. In the feed market, yellow com from the United States
is supplementing Mexican production, which is clearly insufficient to meet growing
domestic demand.
The majority of Mexican imports of com consist of yellow com (around 80
percent), which is primarily used in Mexico as an ingredient in animal feed. From 1998
to 2002, Mexico also imported significant quantities of white com, which is used to make
traditional food. Of total imports, the U.S. supplies about 95 percent of total com
77
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
imported into Mexico. Mexico has pursued a transitional policy toward the United States
com which is more liberal than that required under the NAFTA. With the loosening of
Mexican trade restrictions, U.S. com exports to Mexico have increased dramatically to
about 7.7 million metric tons in 2003 (USDA 2004). Over the next decade, Mexican com
demand is expected to follow recent trends. The demand for feed com is expected to
increase rapidly with per capita income as Mexicans incorporate more meat into their
diets, while the demand for food com is expected to expand with population growth,
(Zahniser and Coyle, 2004)
Since the NAFTA took effect in 1994, Mexican grain imports have increased,
particularly com imports have increased by 240 percent relative to the average annual
imports from 1984 to 1993. Similar trends are also observed for sorghum, wheat and
soybean imports. To ensure that domestic demand for com is fully met, however, the
Mexican Government has issued additional import permits beyond the amount required
by NAFTA ((Zahniser and Coyle, 2004). In the late 1990s, white com emerged as a
significant component of Mexican imports from U.S. From 1998 to 2002, white com
made up about 15 percent of U.S. com exports to Mexico, compared with roughly 2
percent during 1991- 93 (USDA-ERS, 2004). Mexican com imports follow a different
channel depending on the end-use. The hog and poultry industry import most of the
yellow com as well as sorghum and soybean, while large flour companies account for
most of the white com and wheat imports.
78
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
79
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
subsidies. Agricultural performance was quite favorable in the early eighties, strongly
supported by subsidies as part of the Sistema Alimentaria Mexicana (SAM). In 1982,
following a drought, agricultural GDP fell 2.9 percent and fixed investment, wages and
cultivated land diminished significantly. Also agricultural imports fell by 54.6 percent in
the same year (Villa-Issa, 1990).
In 1982, several government policies were instituted which influenced the dairy
sector. Price controls were put in at every level of milk production and marketing
(production, processing and consumption). Retail prices were fixed despite rising
production costs. This led to a slow expansion in milk production, chronic production
shortfalls and increasing imports.
In 1990, the milk system had a nutritional dependency coefficient of 38 percent
(calculated as imports/disposable milk (imports + production)), which has been steadily
increasing over the past four years. Milk and dairy products as a group were the principal
food imports in 1993 with a volume of 445 000 t and a value of $626.3 million. This
shows a tendency to increase dependency on imports in both volume and value terms
(Munoz et al., 1994). Mexican dairy policy has tended to favour the consumers at the
expense of the dairy producers. Consequently, Mexican milk has the lowest producer
subsidy equivalent (PSE) of -5 6 percent (Munoz and Odermatt, 1993). This negative PSE
reflects a tax on producers rather than a subsidy. The principal factor that explains the
negative PSE is low producer prices.
Until 1988, producers were forced to sell their milk at below market prices.
Decapitalization forced many farmers out of production, especially those who produced
low quality milk. Price controls on milk led to the diversification in the use of fluid milk
80
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
by using milk for products whose prices were not controlled (such as cheese, yogurt, and
butter), adulteration of products by substituting up to 80 percent vegetable fat instead of
butterfat, distribution of milk through less controlled channels, and increased vertical
integration of dairy farmers. All these decreased the production of milk which led to an
increase in social programs of CONASUPO (Munoz and Odermatt, 1993).
Dairy-product consumption shows distinct patterns in Mexico, where per-capita
consumption of fluid milk is much less when compared to that of the EU or the US.
However, per-capita consumption of NFDM is two to three times more than in the US or
EU. In 1991, per-capita fluid milk consumption was 46 percent of US per-capita
consumption and NFDM consumption was about 288 percent of US consumption. Milk
is not a traditional consumption item and protein source in Mexico. The high
consumption of NFDM occurs because it is easy to store and transport without
refrigeration and also because of its price.
Cheese is a very important item in the daily diet of the average Mexican
household. It is consumed either as an appetizer or topping, or as a main dish. Different
types of cheese are used for different purposes. However, the majority of the Mexican
consumers prefer fresh cheeses to aged ones. Tastes and preferences are also influenced
by income level where fresh cheeses are consumed at lower-income level and aged
cheeses consumed more at middle and high-income level (Munoz and Odermatt 1993). In
1992, 10 percent of total cheese consumption was made up of imports which were mostly
hard or semi-hard cheeses.
CONASUPO has existed in Mexico since 1938 under various names, adopting its
current name in 1965. CONASUPO is involved in the distribution of basic foods to the
81
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
82
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
market duty free. Imports over the quota level will be assessed at a tariff of 133.4 percent.
This tariff will gradually be phased out by the year 2008.
83
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Individuals with such utility functions will be risk-averse. Absolute risk aversion captures
the above features. A unique measure of the direction of bending and the rate of change
in slope of the utility function is the absolute risk-aversion defined by Arrow and Pratt as:
U\x)
Hence, the degree of absolute risk aversion is higher as the curvature of the utility
function is larger. The degree of absolute risk aversion may change as the wealth level
changes, i.e. RA may be a function of the level of wealth. In this way, it is possible to
classify utility functions with diminishing absolute risk aversion (DARA), where RA(x) is
decreasing in wealth. In other words, the degree of absolute risk aversion diminishes as
the wealth level increases. When there is constant absolute risk aversion (CARA), RA(x)
is constant in wealth, which means that the degree of absolute risk aversion remains the
same as the wealth level increases. However, with Increasing Absolute Risk Aversion
(IARA), the degree of absolute risk aversion increases as the wealth level increases.
Arrow (1970) showed that with decreasing absolute risk aversion, the absolute
demand for risky investment increases as the wealth level increases. Furthermore, with
increasing absolute risk aversion, the demand for risky investment decreases as the
wealth increases. Because the demand for risky investment should increase as wealth
increases an agents utility function should have a feature of DARA.
The
assumptions of the basic model are as follows: the firm in this model is not large enough
84
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
to influence its subjective probability distribution about its sales price, which is assumed
to be a random variable. Therefore, the firm is a price taker. Input prices are given for the
firm. The firms decision regarding input utilization and volume of output must be made
prior to the knowledge of the market price. The firm seeks to maximize expected utility
from profits and its attitudes toward risk can be described by a Von NewmannMorgenstem utility function, i.e., the utility function is such that: U = U (tc), where U (7i)
> 0 and U (7t) is less, equal or greater than zero depending on whether the firm is riskaverse, risk neutral or risk referrer. The firms profit is given by
n = pq - w L - r K ,
(3.25)
where P is the product price which is assumed to be random, with density function f(p)
and the mean E(p) = p. The expected utility of profits is given by
E {U [P f(K,L) - wL - rK]} ,
(3.26)
where E is the expectation operator, f ( ) have isoquants that are convex to the origin, that
is, 2 f Kf Lf K i - f t f x K ~ f x f u . > 0 The firm chooses input quantities so as to maximize
expected utility from profits. The first order conditions for the maximum are
oL
= E[U \ n ) { p f L - w)\ = 0 ,
(3.27)
and
(3.28)
= E [{ /W O * -r)] = 0 .
= ^ = E[u
and
85
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
(3.29)
(3.30)
and W
- B f = D > 0,
(3.31)
In the certainty case these conditions ensure utility maximization assuming that
the production function is strictly concave. Under a random price, assuming risk
aversion, strict concavity of the production function is sufficient but not necessary to
ensure expected utility maximization.
It can be shown that under price uncertainty, the risk-averse competitive firm
demands smaller quantities of inputs and produce lower output than in the case under
uncertainty. The first order conditions (3.27) and (3.28) are
E [ U \ k )P - f L] = E[U' ( x) - w]
and
(3.32)
E [U 'Or) p - f K] = E[ U Or) r] ,
(3.33)
( 3 -34 )
then,
K = E(ju) + ( p - fi) q .
U'(7t) <U'[E{7tj\
for
p > //
Assuming
or
strict
multiplying
concavity
of
through
by
U (-),
then
(p - j u ) f L,
ju ) ] <
0.
Using this result in (3.34) implies that E[U '(7r)(w- jUfL] < 0 and since marginal utility is
86
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
always positive, w < j l f L. Similarly can be shown that r < jufK . The above exposition
shows that under uncertainty, the expected marginal value product of each factor exceeds
its price or exceeds marginal cost. Thus, the optimal output produced under uncertainty
will be lower than that produced under certainty.
(3.35)
B<B
11 =-E\ V - r > 4
off + ^ off
(336)
+h V W ] =
dK _ AtF2- B t F x
dff~
Since
(3.38)
and A 2 are less than zero, and D > 0 the problem now is to sign the term Bi Fi
87
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Using (3.36), the expression for Bi and the expression for Fi and F 2 in (3.37) and (3.38)
are
3L E ( U ' p ) { E ^ W ) H q l h ) E [ U ' ( p f K- r ) } } ( f Kf KL- f J KK)
= ----------- 1------------------------------------------de
d
dK
and
(3.39)
D
Since E [ U " ( p - f K - r ) ] and E[U"( p- f L- w ) \
E W p ^ E i J J d H q l f K) E [ W ( p f K - r j \ } ( 2 f J J KL- f Lf KK)
'
88
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
dw
D -fL
dK
(3 .4 2 )
(3.44)
The signs of the expressions given by (3.42), (3.43) and (3.44) depend on the
assumptions about the firms attitude to risk as well as the sign of f KLand f... In the
certainty case we have d L / d w < 0 due to concavity, but the signs of dK / dw and d q / d w
depend on the sign of f KL in addition to concavity. The main result of this discussion is
that with a well-behaved production function and the hypothesis concerning the firm s
risk attitude, a rise in the price of any input induces the firm to reduce the demand for that
input and lower its output. However, the firm s response toward the demand for the other
input is indeterminate. Under risk-aversion, firms utilize smaller quantities of inputs and
thus produce a lower output than a firm operating under certainty, even though the former
minimizes the unit cost of production of whatever output it chooses to produce.
89
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
function of real income and the price of the traded goods relative to the price of domestic
substitutes and other relevant factors. Depending upon whether traded goods are
considered as intermediated goods or finished products, such demand functions for
imports can be derived from conventional production or utility theory. It is clear,
however, that imports can be expressed as a function of output or income and one relative
price only if imports and domestic goods can be aggregated.
In this model production theory is used to theoretically derive the import demand
functions by treating imports as inputs to the domestic technology. It is assumed that
import and export decisions are made by profit maximizing firms which operate under
perfect competition in all commodities and factor markets. Firms choose their optimal
output mix and their input requirements subject to a vector of output and import prices.
These domestic factors are assumed to be mobile between firms and their rental rates are
determined by their marginal products.
The assumption that the commodities in this study (tomatoes, maize, sorghum and
milk powder) are intermediate goods is based on the fact that tomatoes in the United
States and maize, sorghum and milk powder in Mexico go through a value added process
as explained in section 3.6.
Most of the empirical literature dealing with trade under uncertainty has evolved
from finance literature. Thus, most authors regresses some measure of exchange rate
volatility and sometimes the level of the exchange rate on the volume of imports or
exports. In most cases, the functional form is either linear or log-linear. This amounts to
using a rather simplistic specification of import and export functions. Indeed, until two
decades ago, nearly all the empirical work on import and export determination relied on
90
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
such ad hoc specifications, where the quantities of import and exports were written as
linear or log-linear functions of relative prices and other factors. These approaches have
some disadvantages. In particular, they are based on little or no economic theory. More
seriously, such demand and supply functions are not consistent with well-behaved utility
or production functions, except in very restrictive cases.
For the model used in this research imports are used together with domestic inputs
and services to produce output that can be absorbed at home or exported. In this line of
research, using production theory, Kohli (1982) disaggregated output and modeled the
supply of exports. Appelbaum and Kohli (1979) allowed for non-competitive behaviour
and tested for departures from price taking behaviour. Furthermore, Appelbaum and
Kohli (1997) modeled import demand function under uncertainty. They argued that
uncertainty must be taken into account explicitly if import decisions and its role as a
deterrent of international trade need to be assessed.
framework, an attempt is made to model import demand under exchange rate uncertainty.
Exchange rate uncertainty enters the model affecting both firms revenue and firms
expenditure on imported inputs.
Import demand function will now be derived. Let q = f (Xy Xm, K ) be a
neoclassical production function. Assume that f(-) is continuous, non-decreasing, linearly
homogeneous and strictly quasi-concave or concave, (The assumption of a well-behaved
production function hold for both the theoretical and empirical derivations). Where X l is
the input labour, Xm represents the quantity of the imported input needed to produce
output q and K is capital. Output can be absorbed in the domestic market or can be
exported. It is assumed that the only source of risk is the uncertain exchange rate.
91
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Therefore the exchange rate is a random variable and by implication, so is the foreign
price and profits. The profit maximization problem of the firm can be represented as:
| = = - [ /' (*)] y .
Therefore the firms input demand and output supply functions are given by
v /, -a
v
x , = -d
dw (.
ow K
and
(3.46)
yK
92
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
(3.48)
(3.49)
w l
and
(3.50)
(R + 91) P - f M( X h, X M, K) = ( R + 9 M)-wM ,
(3.51)
(3.52)
X ,M =X*M(wL,wM, p , R , 9 M).
(3.53)
These relations indicate the amount of each factor that will be hired as a function
of the factor and product price; they are the choice functions of this model. Assuming that
93
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
it is possible to solve for equations (3.52) and (3.53), it becomes meaningful to perform
comparative statics of the profit maximization model to know the changes in factor
employment due to given changes in prices and exchange rate.
Substituting equations (3.52) and (3.53) back into equations (3.50) and (3.51) results in
(R + 0l ) P - f L(X*L(wL,wM, p ) , X tM(wL,wM, p), K ) - w l = 0 and
(3.54)
(3.55)
(3.56)
( R + 0 , ) P f KL^ + ( R + 0 t ) P f u t f ^
O = O.
owL
dwL
(3.57)
dX
To solve for - multiply equation (3.56) by / mm and equation (3.57) by/uvi to get
dwL
94
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
(3.58)
(3.59)
dw,
'J y y L
= ^ ----------fj m ----------
<Q
(3.60)
This expression is evidently negative under the hypothesis of risk-aversion and a well
behaved production function since denominator is positive. In other words, demand of
labor responds negatively to changes in its price. Differentiation of equations (3.54) and
(3.55) with respect to wm yields
( R + d1) p . f lL^
^
+ ( R + 0 l) p . f lM^ L ^ - O
dwM dR
owM dR
= O and
(3.61)
( + g , ) f - / , a f Ct
+(R + el) P - f m
~(R + e,)= o.
dwM dR
dwM dR
(3.62)
3X *
To solve for - ----- multiply equation (3.61) b y / lm and equation (3.62) by/L L to get
dw2 dR
( R + e , ) P - f u , f u. ^ - ^ - + (R + ^ ) P - f m ^ JL^ S - - 0 = 0
dwM dR
dwM dR
and
( R + e l) P . f l j KL^ . ^ M . + ( R + e l) p - f u. f m ^ - ^ - < . R + e l) f u. = o .
@R
OWj^ oR
(3.63)
(3.64)
(R+0,)U
<0
(3.65)
( * + 0 ,W u / , - / i )
This expression indicates that demand of imported inputs responds negatively to changes
in its price and the exchange rate.
95
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
3.7.4.2 The Response of Import Demand to Output Price and Exchange Rate:
Differentiation of equations (3.54) and (3.55) with respect to p yields
(3.66)
d x l dP
( R + <?,)P f ML
+ (R + 0,)P
1 dP dR
0-67)
+ )/
d x[ dP
To solve for - multiply equation (3.66) by /mm and equation (3.67) b y / m to get
dP dR
(R + %
+W
- f
(it + 0 , ) P - f L 1 / 1 | + ( R + 4 ) P f u , S mm
f w
f /
~ (* + 3 ) / /
(3 6 8 )
(3.69)
dX'L dP
dP dR
- ( R + f f ,)[/,/ + / , / , J
(R +
- f m)
> 0
a x ; dp
dp dR
yields
- q + 3 )[/ / + / , / , . ] ^ 0
(+a,)F(/u / - / i )
Again, under the hypothesis of risk-aversion and assuming negative sign for the cross
partial derivative, the demand for imports respond positively to changes in output price
and the exchange rate.
96
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
3.7.4.3 The Response of Import Prices and Output to Changes in Exchange Rate
To know the effect of changes in exchange rate on the price of imports differentiate
(3.54) with respect to wm , yielding
dwM dR
+ ( R + 0 l) P - f ul
dwM dR
-0 = 0.
(3.72)
Solving (3.66) for , gives an expression that indicates the effect on the input price
dR
of changes in exchange rate
dWM ~
dR
(R + 0 J P
>0
1
dX^
' LL
dwM
(3.73)
dxl
J LM
dwM
Equation (3.73) indicates that the price of imports responds positively to changes in
exchange rate. To know the effect of changes in exchange rate on price of output, solve
dP
equation (3.67) for to obtain expression (3.68), which indicates that the output price
dR
respond positively given a devaluation of the Mexico-US exchange rate
d X M dP _
(R + 9\)\ f m f mm + f t f i M ] > q
dP dR
2 dX L
74 )
dp
Summary on Comparative Statics
The comparative statics results the profit maximization model show how the import
demand function behave as input prices, output prices, the random exchange rate and its
volatility change values. The main results from the comparative statics are as follows:
97
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
The comparative statics results show theoretically how the import demand function
change due to changes in a number of explanatory variables. These results are used to
guide the empirical analysis presented in chapter six and chapter seven.
98
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
the exchange rate variable are introduced directly into the import demand functions to be
estimated. The general formulation of this imports demand function is:
Qd = F(Y, ER, V, Pc, Ps )
(3.75)
The import demand function is a function of income, prices, exchange rate and
exchange rate uncertainty. In this research uncertainty of the exchange rate is represented
by the volatility of exchange rate. The random variable is distributed according to the
density function g(6), with E(d) = 0, so that E(R) = R and Var(R)= Var (d)= cr2. Since the
moments of the distribution g(6) are unknown, they need to be estimated in order to
proceed with the estimation of the import demand. The moments of the g(0) distribution
are estimated using a GARCH and a nonparametric method, and two commonly used
procedures.
As for the appropriate specification of import demand, the theory does not provide
any specific direction on the best functional form and the most appropriate measures of
the variables to be used in the analysis. Thursby and Thursby (1984) tested nine most
common specifications of aggregate import demand functions for the United States and
found the appropriate functional form to be log-linear rather. Import demand equation is
specified in logarithmic form also because of its convenience and ease of interpretation of
the results.
99
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
(3.76)
where:
dQ *
_
>0
dUSY
dQ * _
-^ -< 0
dPt
dQ *
- ^ > 0
dPus
dQ *
- ^ > 0
dW
dQ * _
>0
dER
dQ * _ dQ *
< 0 >0
dVt
dDt
Definition of variables:
Q* is quantity of imports; it represents the volume of fresh tomato (excluding green
house tomato) imported from Mexico to United States. It is measured in metric tons.
USY is United States per-capita income (importers income), the nominal US per-capita
disposable income was deflated using the United states Consumer Price Index (CPI) to
obtain real per-capita disposable income in the United States. W is the ratio of United
States-Mexico farm wage rate (expressed in US dollars per hour and has been deflated
using the United States CPI). It represents a ratio of unit cost of production, Pt is the
border fresh tomato price (in US dollars per kilogram), it is the quotient of the total value
of Mexican exports to US and the volume (in metric tons) of Mexican exports. All border
prices and Unites States values were converted to real terms using the US CPI. Pus is the
100
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
price of a substitute, in the case of tomatoes is the tomato price in the Unites States (in
US dollars per kilogram). This variable was selected basically because using the domestic
tomato price in the Unites States provides theoretically expected signs. D; is NAFTA and
seasonal dummy variables to indicate the effects of liberalization (the reduction or
elimination of tariffs) and seasonal effects respectively. ER is the Mexico-Unites States
exchange rate (pesos per US dollar). V) are measures of exchange rate volatility. Four
different volatility measures were tested for each commodity. Table 3.2 provides a
description of the variables and a summary statistics.
The expected sign for income variable (USY) is positive indicating that as real
income in the importing country rise so will the quantity imported. The expected sign for
Pt is negative and for Pus is positive indicating an inverse relationship between quantity
demanded and own price and a direct relationship between import demand and related
commodity price. The ratio of wage rates is expected to have a positive coefficient. A
positive coefficient is expected for the foreign exchange variable because it is assumed
that appreciation of the Unites States dollar lead to an increase in imports as Mexican
products become relatively cheaper to those in the United States. Assuming risk-averse
traders, the coefficients of the measures of volatility are expected to be negative.
The trade liberalization dummy variable is expected to have a positive coefficient
because the hypothesis of free trade is that trade liberalization increases trade flows and
wealth of nations.
101
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Description
Units
M ean
Variance
St Dev.
Qt*
Metric Ton
46,337
98334425
31,358
USY
US$/Person
6401.617
844031.79
918.712
Pt
US$/kg
0.75029
0.07672
0.27699
Pus
US$/Kg
0.77361
0.30482
0.55211
US$/hour
6.7392
0.9424
0.9708
ER
MX$/US$
6.84389
10.2334
3.19897
(3.77)
3 g . <()
dQ'
dPc
dPcd
>0
dQ'
>0
dINV
dQ '
dER
>0
dQ *
dV;
<0
102
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
dQ*
dD,
>0
Definition of variables:
Here Qc* is quantity of imported maize for Mexico from the Unites States. It represents
the volume of maize that Mexican importers import from the Unites States. It is measured
in metric tons. MXY is importers income, the original series was in constant pesos
(1993=100) so, series was divided by the population to obtain the real Mexican GDP per
person. Pc is the border price (in US$/kg) of the commodity, it is the quotient of the total
value (in US dollars) of Mexican imports from Unites States and the volume (in metric
tons) of Mexican imports. All border pieces and Unites States values were converted to
real terms by using the US CPI. Pcd is the price of a substitute. The substitute of
imported com is domestic com. The variable INV is the inventory of hogs on feed.
Dj
is
103
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Units
M ean
Variance
St Dev.
Metric Ton
328953
61987948
248973.7
MX$/person
2977.21
1869674.5
1367.36
Pc
US$/kg
0.13667
0.003009
0.05485
Pcd
Units
11971
0.1016E+9
10083
Million
Units
6.74318
1.20296
1.0968
ER
MX$/US$
6.84389
10.2334
3.19897
Di
Zero / one
...
Variable
Qc*
MXY
INV
104
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
(3.78)
where:
dQ *
^
dMXY
>0
dQ *
-^-<0
dPs
dQ *
- ^ > 0
dPsd
dQ *
dINV
>
dQ * _
^ > 0
dER
dQ *
<
dVt
dQ * ,,
-^ > 0
dDi
Definition of variables:
Here: Qs* is quantity of imported sorghum. It represents the volume of sorghum that
Mexico imports from United States. It is measured in metric tons. MXY is importers
income; the original series was in constant pesos (1993=100) so, the series was divided
by the population to obtain the real Mexican GDP per person. Ps is the border price (in
US$/kg) of the commodity; it is the quotient of the total value (in US dollars) of Mexican
imports from United States and the volume (in metric tons) of Mexican imports. All
border pieces and United States values were converted to real terms by using the US CPI.
Psd is the price of a substitute. Because the bulk of sorghum imports go to feed industry,
the substitute of sorghum is the price of domestic sorghum. It is the price of domestic
sorghum in Mexico. D; is NAFTA or seasonal dummy variable to indicate the effects of
liberalization (the reduction of tariffs). ER is the Mexican peso per US dollar exchange
rate. V) is measures of exchange rate volatility. Four different volatility measures were
tested, one at a time, in each equation (Table 3.4).
The expected sign for income variable (MXY) is positive indicating that as real
income in the importing country rises so the quantity imported. The expected sign for Ps
is negative and for Psd is positive indicating an inverse relationship between quantity
demanded and own price and a direct relationship between demand and related price
105
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
respectively. A positive coefficient is expected for the foreign exchange variable because
appreciation of the US dollar leads to a decrease in imports from United States as
sorghum in the United States become relatively expensive to those in Mexico. Assuming
risk-averse traders, measures of volatility are expected to be negative. The dummy
variable is expected to have a positive coefficient.
Mean
Variance
St Dev.
Metric Ton
286800
22199414
148994
MX$/person
2977
1869674
1367.3
US$/kg
0.114
0.00030
0.01736
US$/kg
0.130
0.0027
0.0521
Million
Units
6.743
1.2029
1.0968
MX$/US$
6.843
10.2334
3.1989
Zero / one
...
...
Description
Quantity o f imports represents the volume
imported by Mexico from US.
MXY
Ps
Qs*
Psd
INV
ER
Di
Units
106
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
the case of milk powder from the United States into Mexico the following equation was
estimated:
dQ *
= <
dPm
dQ *
- ^ >0
dPca
dQ *
- >0
dMSF
dQ *
dER
>
dQ *
^< 0
dVt
(3.79)
dQ *
= >
dDt
Definition of variables:
Here Qm* is quantity of imported milk powder. It represents the volume of imports into
Mexico from United States. It is measured in metric tons. MXY is importers income, the
original series was in constant pesos (1993=100) so, the series was divided by the
population to obtain the real Mexican GDP per person. Pm is the border price (in
US$/Kg) of the commodity, it is the quotient of the total value (in US dollars) of Mexican
imports from United States and the volume (in metric tons) of Mexican imports. All
border prices and United States values were converted to real terms by the US CPI. Pea is
the price of a substitute, it is represented by the price of milk powder imported from
Canada into Mexico. MSF is the difference between total production and consumption of
milk products in Mexico. D; is NAFTA or seasonal dummy variable to indicate the
effects of liberalization (the reduction of tariffs). ER is the Mexican peso per US dollar
exchange rate. V) are measures of exchange rate volatility. Four different volatility
measures were tested, one at a time, in each equation (Table 3.5).
The expected sign for income variable (MXY) is positive indicating that as real
income in the importing country rises so the quantity imported. The expected sign for Pm
is negative and for Pea is positive. A positive coefficient is expected for the foreign
exchange variable because appreciation of the US dollar leads to a fall in Mexican
107
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Table 3.5 Description and Specification of Variables for Milk Powder Equations
Variable
Qm*
Description
Quantity o f imports represents the
volume o f milk powder imported
by Mexico from US.
Units
M ean
Variance
St Dev.
Metric Ton
3842.6
1826667
4274
MXY
MX$/person
2977.211
1869674
1367.3
Pm
US$/kg
1.4227
0.2655
0.5153
US$/kg
1.5622
0.2771
0.5490
Hundreds
Liters
306.24
20158
141.98
MX$/US$
6.843888
10.2334
3.198
Pea
MSF
ER
Di
3.9 Summary
Exchange rate variability has increased since the 1970s after the termination of the
Bretton-woods agreement on fixed exchange rates. The exchange rate variability has been
identified as a source of important negative effects on international trade flows. As a
consequence, it has been a concern for trading nations. This issue, together with changes
in exchange rate, is the main concern of this research. In this chapter, the effects of
108
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
exchange rate changes on commodity price were illustrated. First, graphically and then,
an algebraic treatment was presented. For the purpose of explanation the graphical and
algebraic treatment is appropriate, but since this approach does not allow for the
treatment of income, related prices and exchange rate volatility, then, a more general
framework was described.
Given the nature of the traded commodities dealt with in this research, the
theoretical framework developed treats import as intermediate good that enter into the
production process, together with domestic inputs, to produce final goods. Uncertainty of
exchange rate enters the model at both exporters revenue and exporters expenditure on
inputs. The comparative statics results show that the direction of the effects of exchange
rate movements on both import demand and export supply. The next step in this research
is to select the econometric techniques to be used for the estimation of import demand
functions and the appropriate treatment of data non-stationary and exchange rate
volatility, all of which are considered in chapter four.
109
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
CHAPTER 4
FORMULATION OF THE EMPIRICAL MODEL
4.1 Introduction
The purpose of this chapter is two fold. First, to discuss the selection of suitable
econometric methods for estimating the effects of changes in exchange rate and its
volatility on Mexico-US agri-food trade, second, to describe data used and to develop
four alternative exchange rate volatility measures to be used in estimation. The chapter
concludes with the definition and specification of the empirical model based on the
theoretical model described in chapter three.
The determination of the effects of the exchange rate on agri-food trade can be
performed using alternative approaches. The literature reports the use of two different
techniques: the structural approach and the reduced form technique. In the structural
approache a set of equations in each country representing the market for a homogeneous
commodity is formulated. This approach was very popular in the 1970s and 1980s
especially for agricultural products. Ordinary least squares and two stage least squares are
the most common econometric methods employed to estimate structural models.
The structural approach has the advantage that institutional, production and
marketing information about the specific commodity can be incorporated relatively easily
in to the structural model. Since the idea is to represent the underlying economic
conditions relevant for a commodity, this approach is highly demanding on specific data.
One of the principal limitations of using OLS to estimate a trade model is that stationarity
of the macroeconomic data is often ignored.
110
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Regression analysis based on time-series data assumes that the underlying timeseries are stationary. They are stationary in the sense that mean, variance and covariance
of the individual series are time-invariant. The classical t-test and F-test are based on this
assumption. This assumption can be verified by determining if the time-series in question
contains a unit-root. A time-series that has a unit-root is known as a random walk, and a
random walk is an example of a non-stationary time-series. It is well documented in the
literature that many macroeconomic time series, like exchange rate, commodity prices,
incomes, etc., follow a random walk process (Banerjee et al., 1986; Meese and Singleton,
1983; and Nelson and Plosser, 1982). W hen non-stationary variables are used in
regression analysis the classical t and F-test are not appropriate and may lead to
misleading conclusions (Hendry, 1986).
Most of the empirical work on the effects of exchange rate and its volatility on
agricultural trade have been done without considering the time series properties of data.
This is particularly true for empirical studies on Mexican trade. No previous study on
Mexican trade performance has made any attempt to test for the existence of nonstationary data and use appropriate econometric analysis in view of data non-stationarity
Consequently, all previous studies may suffer from: a) a spurious regression problem
often characterized by a high R and low value of the Durbin-Watson statistic b)
inconsistent and less efficient ordinary least squares (OLS) parameters estimates unless
the variables are co-integrated (Engle and Granger, 1987); and c) not having a valid error
correction representation (Phillips, 1986).
The reduced form approach uses a time series model to study the economic
relationship between the exchange rate movements and the variables entering the model
111
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
such as prices,
quantities
expressing institutional
information. In the reduced form approach, the endogenous variables, are expressed
explicitly in terms of lagged values of itself and of other variables in the model. The main
advantage of the reduced form model is that meaningful economic hypotheses can be
tested without estimation of structural parameters and it can accommodate non-stationary
data. Also, reduced form model is less demanding in terms of data.
The structure of this chapter is as follows: Section 4.1 discusses the different
approaches proposed in the literature to tests for the presence of a unit root in the data.
Section 4.2 deals with the methods for determining cointegration and discusses
advantages and disadvantages of each technique to finally propose the appropriate one for
this particular study. The Error correction model (ECM) is discussed in section 4.3. Also
it presents an overview of the estimation technique. Four different volatility measures are
developed in section 4.4. The chapter ends with a discussion of the data and data sources
and a conclusion.
112
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
with independent and identically distributed errors, and they modified the statistic nonparametrically along the lines of Phillips (1987) to deal with more general time series
(Oya and Toda, 1995).
The unit root tests proposed in the literature can be classified into parametric and
non-parametric groups. Simulation studies performed by several authors (e.g. Schwert,
1989a; DeJong et al., 1992) reported that the non-parametric unit root test of Phillips
(1987) and Phillips and Perron (1988) do not work well for typical sample sizes available
for economic analysis. The Dickey-Fuller (DF) and Augmented Dickey-Fuller (ADF)
unit root tests are often applied to test whether a time series has a unit root.
Since cointegration analysis is interesting only for non-stationary time series, the
first step in cointegration analysis is to verify that all variables in question are integrated
of the first order (i.e., there is a unit root in the data set).A variable y r is said to have a
unit root in its autoregressive process if it has the following autoregressive representation:
(4.1)
(1 - L ) y t = a 0 + a, yr_, +
1=1
(i - L)- + ,
113
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
(4 -2)
A negative and significant estimate of axis inconsistent with the null hypothesis
of a unit root in y t . The t-ratio on the estimated ax, however, it does not have a standards
t-distribution. The critical value provided by Dickey and Fuller (1979) needs to be used.
A number of econometric studies have found that standard test for a unit root,
such as the ADF (Dickey and Fuller, 1979, 1981) and the Phillips and Perron (1988)
tests, have low power against stationary alternatives in the relatively small samples. (See
Dejong et al., 1992). This is especially true when a series under investigation is a near
integrated process. Since the low power of the univariate unit root test is the primary
problem, it is important to investigate whether or not the null hypothesis of a unit root is
rejected by a more powerful test.
Oya and Toda (1995) developed a new procedure for testing the presence of a unit
root test by combining features of the ADF and the LM tests. Through Monte Carlo
simulations they examined the dependence of the finite-sample distributions of the DF
and LM test statistics on the estimated parameters and initial values of AR processes.
They argued that by combining two different test statistics one may obtain a unit root test
that is more robust to the initial value (and the value of coefficient parameters). Note,
however, there is no unique way of combining the DF and LM tests to produce a more
robust unit root test. It would be interesting to seek an optimal way of combining these
two unit root tests. The asymptotic local power for different methods of combination
could presumably be compared, for example, in a local-to-unity asymptotic framework,
formulating the initial value as an increasing function of the sample size.
To overcome the problem of low power and severe size distortions of
conventional methods, Ng and Perron (2001) propose new model selection procedures
114
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
and estimation strategy that produce good power and reliable size. They suggested new
modified information criteria (MIC) with a sample dependent penalty factor, along with
the GLS detrending of Elliott et al. (1996) to increase the power of their tests. This test
was applied by Yoon (2003) to Brazilian data. He concluded that the Ng and Perron
procedure for testing unit root yield reliable results. Brazilian inflation rate is not
stationary, while application of the ADF test yields stationarity for the same data set.
It is a well-established empirical fact that standard unit root tests fail to reject the
null hypothesis of a unit root for many economic time series. This was first argued
systematically in the influential article of Nelson and Plosser (1982) who applied DickeyFuller type tests (Dickey and Fuller (1976) and Dickey and Fuller (1979)) to 14 annual
U.S. time series and failed to reject the hypothesis of a unit root in all but one of the
series. These results do not changed by allowing for error autocorrelation using the
augmented tests of Said and Dickey (1984) or the test statistics of Phillips (1987) and
Phillips and Perron (1988). Similar results are obtained for many other macroeconomic
time series. The standard conclusion that is drawn from this empirical evidence is that
many or most aggregate economic time series contain a unit root. However, it is
important to note that in this empirical work the unit root is the null hypothesis to be
tested, and the way in which classical hypothesis testing is carried out ensures that the
null hypothesis is accepted unless there is strong evidence against it. Therefore, an
alternative explanation for the common failure to reject a unit root is simply that most
economic time series are not very informative about whether or not there is a unit root, or
equivalently, that standard unit root tests are not very powerful against relevant
alternatives. Several more recent studies have argued that this is indeed the case.
115
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
116
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
sample Bayesian inference and that contrasts with the non-standard asymptotic
distribution theory of classical estimators and tests.
As can be seen from the above discussion almost no method meets all the
desirable characteristics in terms of power and robustness. In most cases, it is ultimately
the reserchers preferences that determine what test should be applied to actual data.
Thus, the ADF test statistic is used in this research instead of Dickey and Fuller and
Phillips and Perron tests.
yt = 3V-i + t
(4-3)
where yt and yt./ are the present and the immediate past values of a variable, respectively;
and et is a stationary error term. Equation A .l can be expressed in the following form:
(4.4)
y , = a + 0 )y
where 6 is an arbitrary parameter. W hen 6 =0, then Equation A.2 equals Equation A .I.
After rearranging Equation A.2, the following equation is obtained:
(4.5)
If 6 equals 0 and et is stationary, then yt ~ 1(1), and if - 2 < 6 <0, then yt is a stationary
process. Based on the above modification, Dickey and Fuller (1976) proposed a test of
117
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Ho: 9 = 0 against Ho: 6 < 0. If the null hypothesis is accepted then the process is 1(1),
i.e. yt ~ 1(1). Dickey and Fuller considered the following three different equations to test
for the presence of unit roots:
A y ,= 0 y ,-i+ ,
(4.6)
Ay, = a 0 + 6yt_x + e ,
(4.7)
Ay, = a0 +9y,_l + a 2t + e t
(4.8)
The differences among the above regression equations depend on the presence of
ao, constant (drift) and ci2 t, deterministic term (time trend), all of which are called
nuisance parameters. Test results can be based on OLS estimations. The above equations
represent the first order autoregressive process (a process depending only on one lag
value). The test can be extended for higher order autoregressive processes. In conducting
the DF test based on equations 4.6 to 4.8 it was assumed that the error term was
uncorrelated, but in case the errors are correlated, Dickey and Fuller developed a test
known as the Augmented Dickey and Fuller test. It is conducted by augmenting to
equations 4.6 to 4.8 with lagged values of the dependent variable to get the following
equations:
P
(4.9)
j=i
P
(4.10)
Ay, = a 0 + a,y,_! + / M - j + ,
(4.11)
118
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
where ao and t are the constant and the time trend, respectively. There are t-critical
values related to Equations 4.9, 4.10 and 4.11, respectively. Enders (1995) suggests that a
procedure on Equations 4.9, 4.10 and 4.11 should always be followed to test for unit
roots when it is not known whether that process includes a trend term or drift. Unit root
tests utilize the following three main null hypothesis tests based on Equations 4.9, 4.10
and 4.11:
Ho: a\ = 0 (testing unit roots in Equation 4.9
Ho: a\ = <32 = 0 (testing unit roots with the time trend in Equation 4.10
Ho: a\ = ao = 0 (testing unit roots with the constant term in Equation 4.11.
Analyses are conducted based on the results of DF, ADF tests. A summary of the
procedure of the steps involved in each test is as follows: Check for unit roots in each
variable with the time trend and the constant terms in Equation 4.10. If a null hypothesis
of Ho'. a\ = 0 is not rejected (at the Dickey-Fuller critical value), there are unit roots. If the
null is rejected, then check for the presence of the time trend, ci2 , in Equation 4.10. If the
time trend is significant and if the presence of unit roots is not rejected according to the
conventional t-value, it can be concluded that the process of the variable has unit roots
with the time trend. If both are rejected, then it can be concluded that the process is
stationary.
If there is no time trend, i.e. null is rejected in Equation 4.10, then check for the
unit roots and the constant term in Equation 4.9. First, check the process for unit roots at
the Dickey-Fuller critical value. If there are no unit roots, it can be concluded that there
are no unit roots in the process, which means that the variable is stationary. If a constant
term is significant, then check the results for unit roots. If Ho is not rejected according to
119
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
the t-value, it can be concluded that the process of the variable has unit roots with the
constant. If Ho is rejected, it can be concluded that the process has no unit roots.
If there is no constant, check the process with neither constant nor time trend in
Equation 4.8. If there are no unit roots, the process is stationary, otherwise it would have
unit roots. If there are unit roots in any of these hypothesis tests, check the variable in
first difference form to check for two unit roots. If there are no unit roots, then it can be
concluded that the variable can be represented by an 1(1) process. Otherwise, the above
steps are repeated for the process in first difference values to test for the presence o f 1(2)
process.
120
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
It is possible to determine the lag length truncation for each series using an
information criterion such as the Akaike Information Criterion (AIC) or the Schwarz
Bayesian Criterion (SBC). These methods are proposed for this research. O f course, in a
very large sample with normally distributed errors, the methods should all select the same
lag length. In practice, the Schwarz (1978) criterion (SBC) will select a more
parsimonious model than will either the Akaike (1973) Information Criterion (AIC).
121
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Johansen approach has better properties than the other alternatives (see Table 4.1).
Johansen
Maximal eigenvalue and trace
statistic
Engle-Granger
Assumptions
-Common
restrictions
Advantages
Super consistent
estimation o f P
cointegrating vectors),
Disadvantages
|3 and a
Sources
for Johansen (1988, 1995)
critical
values Johansen and Juseluis (1990)
and p-values
Osterwald-Lemus (1992)
_________
McKinnon et al (1999)
Source: Ericcson (1993)
Weak
exogeneity
is
assumed.
Usually r < 1 is imposed
Banerjee et al (1993)
Banerjee et al (1998)
Harbo etal (1998)
Pesaran et al (2000)
factor
122
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
study indicated that finite sample properties are consistent with the asymptotic results.
This is so even when the errors are non-Gaussian or when the dynamics are unknown
Before performing cointegration analysis, Gonzalo (1994) suggests that it is
important to discuss three elements present in any cointegrated system. First, the
existence of unit roots, second the multivariate aspect and third the dynamics. Not taking
these elements into account may create problems in estimation. In general the coefficient
estimates will be biased in mean and median as well as inefficient. The distribution will
be nonsymmetric and nonstandard and there will be nuisance parameter dependencies.
The appropriate cointegrating method should have the following characteristics:
Incorporate all prior knowledge about the presence of unit roots. This eliminates
median bias, non-symmetry, part of the nuisance parameter dependencies, and
increases efficiency.
Full system estimation. This eliminates the simultaneous equation bias and
increases efficiency.
compared only maximum likelihood in an error correction model that satisfies these
requirements. This approach ensures that coefficient estimates are symmetrically
distributed and median unbiased and that hypothesis tests may be conducted using
standard asymptotic chi-squared tests. None of the other methods considered has these
properties. Although the above properties are based on asymptotic theory, Gonzalo
(1994) showed, via a Monte Carlo study, that this conclusion is still valid for finite
samples. Based on results of Gonzalo (1994), Phillips (1991) and Hubrich (2001) among
123
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
124
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
likelihood ratio tests, Lagrange multiplier and Wald type test, lag augmentation tests,
tests based on canonical correlations, the Stock-Watson tests and Bierens nonparametric
tests. Asymptotic results regarding the power of these tests and previous small sample
simulation studies were discussed.
Hubrich (2001) compared a wide range of systems cointegration tests with respect
to their asymptotic properties as well as their small sample performance. It is well known
that in applied work, it is important to know the assumptions under which a test
procedure is supposed to work in order to choose the appropriate test. An important
aspect in this context is the assumption about the trend terms of the process and whether
the test statistics depend on these terms. Therefore, he suggested to focusing on analyzing
the sensitivity of the test performance to the trending properties of the data generating
process. He found that a trade-off between the generality of the assumptions for the
validity of a test on the one hand and its local power on the other hand.
The aim of the simulation study performed by Hubrich (2001) was to explore the
strengths and weaknesses of the different systems of cointegration tests to compare on the
basis of some simple DGPs. The idea is that tests which dont work well for a simple
DGP cannot be trusted in more complicated situations in general even if they have been
shown to perform well in specific complicated situations. Using this criterion, it was
found that the lag augmentation tests, the canonical correlation tests, the Stock and
Watson tests as well as the Bierens test perform very poorly even under ideal conditions.
Therefore their use cannot be recommended for applied work. In General likelihood
based methods are found to perform better than other methods.
125
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
126
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Xt = c + 7iiXt_i + 712X ^2 +
Where:
(t = 1,..., T)
(4.12)
(4.13)
where:
Tj = - 1 +7ii + .......... + 7ii, and
It is interesting to note that the reparameterized model is a traditional firstdifference VAR model except for the term II Xt.k. The coefficient matrix of Xt.k, II,
contains information about the long-run relationships among variables in the vector of
data. If II has a full rank, then X is a stationary process. In this case, a non-differenced
VAR model is appropriate. If II has a zero rank, then II is a null matrix and Xt is an
integrated process; only in this case, a traditional first-difference VAR model is
appropriate (Orden and Fisher 1991). If, however, 0 < (rank (II) = r) < p, cointegration
127
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
holds and n can be represented as the product of two p x r matrices a and (3, such that II
= a(3'. The /?'s are the cointegrating vectors and a s are the weights. In this case, f i fX t
is stationary. The long-run equilibrium is unique only when r =1.
The maximum likelihood estimation of II consists of two sets of regressions; one
set generates the residuals Rot from the regression of VXt on VXt_i, ...., VXt.k+i, and the
other set generates Rkt from the regression of Xt.r on VXt-i,
The concentrated likelihood function in terms of the product moment matrices of the
residuals can be expressed as:
(4 . 14 )
where Soo, Sok, Sko and Skk are the product moment matrices of the residuals defined as:
Sij = T ' Z * . * ;
Vij = 0,k
t=1
It is clear from equation 4.14 that maximizing the concentrated likelihood function is
equivalent to minimizing |QA(/?)|. This minimization amounts to solving the following
eigenvalue problem:
l-V -C -X O -C '-'h O
(4.15)
where C is a (p x p) matrix such that Skk = C'C. The vector of eigenvalues is given by X
while the corresponding eigenvectors can be derived as Vi = C'~l ei, where e is are the
eigenvectors from equation 4.15. The estimates of a and Q. can be obtained by using the
estimated value of /?. The null hypothesis that there are r cointegrating vectors is tested
using two likelihood ratio tests called the trace test and the maximum eigenvalue test. If
Hi: is a special case of H 2: for r = p, then the trace statistic is defined as:
128
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
(4 .1 6 )
2 In(Q; H 2/ H i) = - T ^ In{ 1 - M)
Similarly, the maximum eigenvalue statistic for testing H 2 (r) in H 2 (r+1) can be defined
as:
-2In (Q; r| r +1) = -T in (1 -Ar+1)
(4.17)
The asymptotic distributions of these likelihood ratio tests do not follow the
standard chisquared distribution. They actually represent multivariate versions of the
Dickey-Fuller distribution. The critical values for these tests are generated through
simulation and are reported in Johansen and Juselius (1990), Owterwald and Lenum
(1992) and MacKinnon (1999).
Johansens approach provides a convenient framework for testing linear
hypotheses expressed in terms of the coefficients p., a and (3. In particular, it is possible to
test for the presence or absence of linear trends in the stochastic part of the model. It is
important to choose the appropriate model formulation (i.e., with or without a trend
variable), because the asymptotic distributions of the test statistic and estimators depend
on which assumption is maintained (West 1988). The null hypothesis of the absence of a
linear trend in the model is equivalent to H q: |i. = e t/? ' Notice that when Hq: is true:
+M = uft-k + a f t = a f i* 'X *,_t i ,
where /3* = (/?,/?) (3* = and X*t-k = (Xt.k< 1).
The test statistic can be computed as:
In [ (1 -i* ) / (1 - Xt) ]
129
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
(4.18)
Likelihood ratio tests can also be formulated to test a variety of linear restrictions
on a and p. Theoretical and empirical economic knowledge can be used to formulate
these restrictions. These restrictions essentially limit the space spanned by the r
cointegrating vectors to lie in the s-dimensional space. If s = r, then the cointegration
space is said to be fully specified (Johansen 1991). Let / / 3: y5 = Hg represent a formulation
of a linear restriction on the cointegrating vectors, where H is a p x X matrix of
restrictions designed to restrict the space spanned by /3 to lie in s-dimensional space and 8
is a set of cointegrating vectors (see Johansen 1991 for details). The likelihood ratio test
can be computed as:
-2In (Q; H3 1H2) = -T Er In [(1-X,A3.i)/(l - 1* 0]
(4.19)
130
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
Tests for either or both short run or long run causality can be conducted using a
vector error correction model (VECM). Cointegration among the variables implies that
there is a long run error correction process, so, any deviation from long run equilibrium
will be restored by the correction of the equilibrium error back towards its long run
equilibrium.
Having obtained the long-run cointegration relations using the Johansen approach,
it is now possible to reformulate the above model and estimate the VECM with the errorcorrection terms explicitly included in it. Although Engle and Grangers (1987) two-step
error-correction model may also be used in multivariate context, the JohansensVECM
yields more efficient estimators of cointegrating vectors. This is because the VECM is
estimated with the full information maximum likelihood method, which allows for testing
for cointegration in a whole system of equation in one step and without requiring a
specific variable to be normalized. This allows one to avoid carrying over the errors from
the first step into the second, as would be the case if Eangle-Grangers methodology
used. It also has the advantage of not requiring a priori assumptions of endogenity or
exogenity of the variables. Consider a VECM that is of the form:
*-i
(4.20)
k-1
where , = ^ F y AY t_j and a/3'Yl k are the vector autoregressive (VAR) component in
j =i
131
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
the jth lag. /?' is a p x r matrix of cointegrating vectors, and A denotes first differences.
a is a p x r matrix of speed of adjustment parameters representing the speed of error
correction mechanism.
In estimating the VECM, stationary and unit roots are first checked by performing
the selected test on the variables in levels and first differences. Only variables integrated
of the same order may be cointegrated and the unit root test will help to determine which
variables are 1(1). For this research the Error Correction Model will be used to analyze
the short-run dynamics among the variables given the desirable properties of short-run
forecasting based on this procedure.
The VECM is estimated using OLS since diagnostic test is performed such as
goodness of fit, the F-test, the DW test for first order autocorrelation, the BreuschGodfrey Lagrange multiplier test for serial autocorrelation, Ranseys RESET test for miss
specification and the Jarque-Bera test for normality.
132
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
variance-based volatility measures (the standard deviation, variance of the spot exchange
rate around its trend, the difference between previous forward and current spot rates, the
coefficient of variation, the standard deviation of the annual percent change of a bilateral
exchange rate around the mean), (ii) Autoregressive Conditional Heteroskedasticity
(ARCH) model or Generalized Autoregressive Conditional Heteroskedasticity (GARCH)
models, and (iii) non-parametric techniques. Each of these measures has its advantages
and shortcomings. Appendix 5 provides an overview of the most widely used models to
measure volatility in empirical studies.
Economic theory does not provide a clear guidance on features of the most
appropriate volatility measure (Clark, et al 2004). In an attempt to shed light on the
efficacy of different volatility measures on trade flows, this research attempts to test the
effects of four alternative exchange rate volatility measures. These volatility measures on
trade flows are as follows: (i) a volatility measure based on an ARCH or GARCH model;
(ii) a non-parametric-based exchange rate volatility measure, and (iii) two variance or
standard deviation-based volatility measures (moving average of the standard deviation
of the exchange rate and the average absolute difference between the previous forward
and the current spot rate; Hopefully, this comparison help to identify what constitutes a
good measure of exchange rate risk in the context of Mexico-US agri-food trade. In the
next section each of the volatility measures is described in terms of their advantages and
shortcomings, empirical justification and specification and what each measure intends to
capture.
Previous research on exchange rate volatility and trade has failed to recognize that
exchange rates are generated through a stochastic process and that exchange rate data
133
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
have exhibit certain set of characteristics that violates the assumptions of a standard
regression analysis. As indicated by by Bollerslev et al. (1992), financial time series are
typically heteroskedastic, leptokurtic and exhibit volatility clustering. He suggests that
these features could be handled more succesfully by modeling the volatility of the time
series as conditional on past behaviour. Since their introduction by Engle (1982) ARCH
models and their subsequent generalizations (see Bollerslev, 1986) have proved to be a
very useful technique to measure exchange rate volatility. The GARCH models have
been particularly used to describe the properties of high frequency exchange rate time
series. But also ARCH and GARCH models have been recently used in different studies
of exchange rate volatility using monthly data (McKenzie, 1999; Lastrapes and Koray,
1990; Mckenzie, 1998; Qian and Varangis, 1994).
134
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
iVr+1 = ,+iX
,
(4-21)
is
Var(yl+1 / x t ) = xfcr2, here the conditional variance of yt+i is dependent on the realized
values of x,. If the magnitude of (jct)2 is large (small), the variance of yt+i will be large
(small) as well. Then, in this way, the introduction of the {xt } can explain periods of
volatility in the {yt } sequence. In fact, equation (4.18) can be generalized as:
Ln (y ,) = a0 + ax ln(xM) + et
(4.22)
where et is the error term, et = ln(r). But a major difficulty with this strategy is that it
assumes a specific cause for the changing variance. Also in this approach {v,}is forced to
affect the mean of the Ln(yt ). Instead of using ad hoc variables choices for x, and/or data
transformation Engle (1982) showed that it is possible to simultaneously model the mean
and the variance of a series.
To elaborate the Engle methodology further, consider the case that the variance of
{ t } is not constant; then, one can estimate any tendency for sustained movements in the
135
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
variance using an ARMA model. Let \et} be the estimated residual from the following
model; y, = a 0 + axy t__x + e t , so, the conditional variance of yt+i is:
Var(yt+l / y t) = Et [(y(+1 - a0 - axy t) 2] = E,(t+l) 2
(4.23)
where Et(et+1) 2 = o 2. Assuming that this conditional variance is not constant, one
strategy is to forecast the conditional variance as an AR(q) process using squares of the
estimated residuals.
(4.24)
where vt is a white-noise process. If the values ax, a2, ...., an all equal zero, the
estimated variance is simply the a0. Otherwise the conditional variance of yt evolves
according to the autoregressive process given by (4.24). Then, we can use model (424) to
forecast the conditional variance at t+1 as:
E
, f+ \
2 ^ f - l
q f + \- q
(4.25)
136
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
Maximum Likelihood estimation, gives a systematic way to estimate the optimal weights.
Once the weights are determined, the dynamic model of time-varying volatility can be
used to measure the volatility at any time and to forecast it into the near and distant
future.
137
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
o f =co+Ya a if-i + X Pi
1=1
1=1
=co+ a (L )f +
(4.26)
assumed that the roots of the polynomial /3(A) = 1 lie outside the unit circle (see Nelson
and Cao, 1991). For a GARCH (1,1) process this amounts to ensuring that both a
and ft
138
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
tail
behavior)
and
volatility
clustering,
two
important
characteristics
of
139
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
rebounds), and other highly unanticipated events that can lead to significant structural
change (Gourieroux, 1997). GARCH models often fail to fully capture the fat tails
observed in asset return series. Heteroscedasticity explains some of the fat tail behavior,
but typically not all of it. To compensate for this limitation, fat-tailed distributions such
as Student's t have been applied to GARCH modeling.
In this research the GARCH technique is applied to the real monthly exchange
rate time series for the period 1989:1 to 2004:12. Utilizing a GARCH model to capture
the conditional variance of real rates is particularly appropriate for the flexible exchange
rate period because this period has generated more volatile real exchange rates than the
fixed exchange rate regime. The next section describes the specification of the ARCH
and GARCH models and how they can be employed to generate exchange rate volatility
measures.
The estimation of GARCH models basically consist of three steps: the first
concerns to the selection of (p, q); the second deals with testing for ARCH effects; c) the
third one is the estimation of the model. To apply the GARCH technique, it is necessary
to find the best fitting time series model for the real exchange rate; an ARIMA model
could perform this task. Once this is done a GARCH (p, q) model to capture the time
varying conditional variance of the real exchange rate is be used, the proposed model will
have the following specification:
a) dLnER =ao+ iet-i+ et,
b) e, 1 1 - N (0, A,).
= a + z m 2- , + 1 ; m - ,
1=1
<4 -27>
1=1
140
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
where: dLnER is the difference in the log of the real exchange rate with respect to the
previous period.
I t-i is all relevant and available information at time t-I;
et-i is the moving average term;
et is the disturbance term;
ht is the conditional variance.
Equations in (4.27) need to be estimated jointly for the sample period. If
coefficients of e t-i and h t-i are statistically significant, significant ARCH and GARCH
effects are said to exist in the data. The prediction value of equation (c) in (4.27) provides
a measure of real exchange rate uncertainty.
141
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
Pagan and Ullah (1988) suggested an instrumental variable method to estimate the
model with a non-parametrically specified risk term. Pagan and Hong (1991) argued that
one can also use least squares after the plug-in risk term is estimated non-parametrically.
However, Pagan and Hong (1991) did not develop the asymptotic distribution of their
proposed estimator. Baltagi and Li (2001) established the 4 n -normality result of the
Pagan and Hong (1991) estimator with generated regressors. Following Pagan and Ullah
(1988) and Baltagi and Li (2001) the departure of the proposed approach is that in
modeling macro-economic models agents are assumed to optimally choose a variable yt
on the basis of some information set Ft.
E( yt \xt Ft_l) = xta + a f S
(4.28)
where xt e F is a p x l vector variable and cr2 is the variance of some variable say wt,
conditional on
Ft., i.e. a f = E [(w, - E(wf | Ff_, ))2 1Ft_x] = E(wf \ Ft_x) - [E(wt \ Ft_x) f .
<T2 represents a risk term arising from the failure of agents to be able to correctly predict
some variable wt, which may be yt, (Pagan and Ullah, 1988). The size of crfS indicates
the impact of the risky environment upon yt. A test for S = 0 is a test for the significance
of the risk premium. Writing (4.28) as a regression model gives:
yt = x toc+o-*S+t =( xt,<jf)
+ ,
(4.29)
j
s x j + t , with E(t I xt Ft_{) = 0 ,
where
X t = (xcr,2)'
and
142
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
In
practice
of ,
is
not
observable because
the conditional
mean
(4.30)
Xu = (nhqy l^ w sK (^ L^ L)/ f t f,
Itt
h
(4.31)
hx = ( n W y ' Y w i K i ^ ^ ) / f ,
itt
h
(4.32)
/, = ( n h i y ^ K i ^ ^ )
Y!
h
Therefore, a] - E(wf \ z , ) - [ E ( w t \ zt) f = <t2(z;) is estimated by erf =
143
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
- i^)2
The above results are applied to monthly observations of bilateral exchange rate
between the U.S. dollar and the Mexican peso for the period of 1989:1 to 2004:12. The
data consist of monthly spot rates, and one-month forward rates. The sample consists of
192 monthly observations. Hai, Mark and Wu (1997) provide convincing evidence that if
the spot and forward rates, St+k and fk,t (k = 1,3) are cointegrated, it is possible to use this
information to obtain a more powerful test. The proposed approach was implemented by
Baltagi and Li (2001) as follows:
st+k- f kj=<Zo + f o t + ,
(4 -33)
Equation (4.33) is exactly the same as considered in Pagan and Ullah (1988)
where they used it to study the risk premium of $U.S.= Canadian exchange market.
Following Baltagi and Li (2001) this research chooses:
wr = ,+t - /* , and z, = (zu z2ty = (sI+k_hst+k_2 - f k t_2y .
Since z, e R 2 (q = 2), condition (A2) in page. 449 of Baltagi and Li (2001) require a
higher order kernel to be used. However, as they argued (in Remark 2.1 page 449) one
can show
that a second order kernel can be used for q=2. Therefore, it is possible to use
a product second order normal kernel function. In the normal kernel, the smoothing
parameters are chosen via ht = czt sdn~U5 where ZjiSd js is the sample standard deviation of
(st+k - f k , )"_* (i = 1,2) and different values of c can be used. The window width h is an
important parameter, and its choice determines the size or the interval around z over
which the observations are averaged. Usually the larger the h is, the less is the variance
and the smoother the curve, but the larger the bias. The parameter h can be calculated as:
h = n~ll(n+q). And use the (zn ,
Z jq )
144
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
4.5.3.1 Moving sample standard deviation of the growth rate of the exchange rate
For the estimation of the sample standard deviation of the growth rate of the exchange
rate the following formula is used:
1/ 2
(4.34)
where m is the order of the moving average, and Z the (log) relative price of foreign
consumer goods in terms of US consumer goods. This measure have been used in the
literature by Cushman (1983; 1986); Gotur (1985); Kenen and Rodrik (1986); Bailey
Tavlas and Ulan (1987); Koray and Lastrapes (1989); Bini-Smaghi (1991); Chowdhury
(1993). This proxy of exchange rate volatility intends to capture temporal variations in
the absolute magnitude of changes in exchange rate.
145
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
146
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
arrangements. So, it is expected that exchange rate movements will affect them in
different manner. Furthermore, more than 90% of the traded value (exports/imports)
takes place with the U.S., so it allows for work with just US-Mexico exchange rate data.
Commodities were identified based on their codes under the Harmonized Commodity
Description and Coding System (Harmonized System, or HS).
Monthly data from January 1989 to December 2003 have been used to estimate
the models. This period was selected based on the availability of data. Also most of the
Mexican trade liberalization took place during this period. Since the focus of this study is
to determine the effects of exchange rate variations on agri-food trade flows, the main
component of the data consists of commodity data and the macroeconomic variables
related to them. Table 4.2 describes the main characteristics of the data and respective
sources of the data used.
All US domestic prices, including the wage rate and border prices were converted
to real values using the general US consumer price index which includes all main cities
and all items. Similarly, all domestic prices in Mexico were transformed to real values
using the general consumer price index in Mexico.
147
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
Description
Sources
Exchange Rate
USDA-ERS.
http://www.ers.usda.gov/Data/Exc
hangeRate.
US CPI
US department of labor
http://www.bls.gov/cpi/cpifaq.htm
#Question%2015.
US Income
US W age
Rate
US National Agricultural
Statistics Service (NASS).
http://www.usda.gov/nass/pubs/re
portname.htm#Farm Labor.
M exico Wage
Rate
M exico CPI
2002 = 100 ).
Mexico GDP
Corn Volume
Corn
Price
Border
Claridades Agropecuarias.
http://www.infoaserca.gob.mx/cla
ridades/
Sorghum
Volume
Sorghum
Border
Price
M aize
Domestic Price
Domestic
Sorghum
Price
Claridades Agropecuarias.
http://www.infoaserca.gob.mx/cla
ridades/.
148
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
D escrip tio n
S o u rces
Tomato
Volume
USDA-ERS. FAS,
U.S. Trade Internet Svstem
Tomato
Price
USDA-ERS. FAS,
U.S. Trade Internet Svstem
M ilk Powder
Canada
USDA-ERS.
U.S. Trade Internet Svstem
M ilk Powder
Volume
USDA-ERS. FAS,
U.S. Trade Internet Svstem
M ilk Powder
Border Price
USDA-ERS. FAS,
U.S. Trade Internet Svstem
4.7. Summary
This chapter described the selection of the econometric techniques for the estimation of
the model outlined in chapter three. The selection was guided by an interest of using the
appropriate econometric methods to incorporate data non-stationary and the appropriate
treatment of exchange rate volatility. To this end, Johansen Maximum Likelihood
cointegration analysis has been selected. This econometric technique is a broadly
accepted in the literature as it accommodates non-stationary data in econometric analysis.
This approach also provides ample opportunity to test meaningful economic hypothesis
without estimating structural parameters.
The appropriate specification and estimation of exchange uncertainty has been a
concern in the literature. This concern is addressed in this research using four alternative
measures including a GARCH and a non-parametric technique to estimate exchange rate
volatility.
149
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
CHAPTER 5
RESULTS OF UNIT ROOT TEST AND VOLATILITY MEASURES
5.1 Introduction
The purpose of this chapter is to present and discuss the results of unit root tests as well
as the estimates of different exchange rate volatility measures. The chapter is divided in
to three sections: the first section deals with the results of unit root test on individual
series and macroeconomic variables for selected commodities (fresh tomato, yellow com,
sorghum and milk powder). The analysis was carried out using SHAZAM econometric
package (version 10 professional edition). The second section presents and discusses the
estimates volatility measures. The final section concludes the chapter.
150
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
c5-1)
Ay, = a0 +a x
Y Y j ^ t - j + t
(5-2)
7=1
P
(5-3)
i =i
where
ao
and
yj
are the
parameters. With an ADF test, the null hypothesis of Ho: a, = 0 is tested against the
alternative hypothesis of Ha:
a,
inconsistent with the null hypothesis of a unit root in the data. If the null hypothesis is not
rejected then this is taken as evidence in favour of the presence of a unit root. Standard t
statistics are not valid for testing the null hypothesis. The appropriate critical values were
derived by Dickey and Fuller (1979). There are different critical values associated with
Equations 5.1, 5.2 and 5.3 respectively. The appropriate statistic to use depends on the
deterministic components included in the Dickey-Fuller regression equation. Without an
intercept or trend, the use of T statistic is suggested; with only the intercept, use the
statistic and with both intercept and trend, use the t T statistic.
The unit root test uses the following three null hypotheses depending upon which
of 5.1, 5.2 or 5.3 is used. These are Ho: a\ = 0 equation 5.1, Ho: a\ = <32 = 0 equation 5.2,
Ho: a\ = ao = 0 equation 5.3. The critical value of the ADF test for model 5.3 is -3.44 at
the 5 percent level of significance. Values of ADF test smaller than the critical value
indicates that the series has a unit root while ADF values higher than the critical -3.44
value confirm that the series is integrated of order zero.
151
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
1 This test is implemented by adding a dummy variable that takes a value o f (0,1) to allow for shifts in the
intercepts assuming a single crash . Critical values are found in Perron, 1990.
152
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
Table 5.1 Unit Root Test Results for US and Mexican Macroeconomic Variables
Variable in Level Form
Estimated
Coefficient
Lag Length
(Months)
ADF
Statistic*
Real US-Mexico ER
-0.014912
- 1.092
-0.018528
-0.9442
- 0.026235
- 1.076
-0.192870
12
- 2. 772
-0.326130
12
-2 .7 1 0
-0.184490
- 1.315
-0.175660
- 3.068
- 0.966270
- 7.283
- 0.502867
- 2.028
- 0.672479
- 3.087
- 0.97241
- 7. 162
-1.0096
-8.253
- 1.59250
- 9. 156
- 4.48070
12
- 8. 573
-4.53180
12
- 5.860
- 1.01620
-2 9 .1 4
- 1.70210
- 8.462
-5.87510
- 8.935
- 1.23845
- 8.089
- 1.37812
-7 .2 1 0
The unit root test in first difference form is performed to confirm whether or not
the series is integrated of order one. Since the absolute value of all the Augmented
Dickey and Fuller (ADF) statistics are larger than the critical value at five percent level
153
R ep ro d u ced with p erm ission o f the copyright ow ner. Further reproduction prohibited w ithout perm ission.
all the macroeconomic variables included in the analysis are integrated of order one, i.e.,
they are 1(1).
The above is true also for the volatility of the exchange rate variables, moving
average standard deviation of the exchange rate (VI), percentage change of the exchange
rate (V2), GARCH measure of the volatility of the exchange rate (VG) and the non
parametric measure of the volatility of the exchange rate.
5.2.1.3 Results of Unit Root Test for Corn and Sorghum Models
The optimal lag-length and ADF test results for the variables included in the trade
equations for com and sorghum are presented in Table 5.3. The results indicate that the
null hypothesis of a unit root cannot be rejected when all the variables are in level form.
However, they become stationary after first-differencing. Hence, they are integrated of
order one, 1(1).
154
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
Table 5.2 Unit Root Test Results for the Tomato Models
Estimated
Coefficient
Lag Length
(Months)
ADF*
Statistic
- 0.27698
12
-2.188
- 0.96221
- 3.068
- 0.28591
-2.125
- 0.55501
-2.881
-5.1510
12
-7.313
- 3.0371
- 9.027
- 3.0371
- 9.027
- 3.0234
-9.321
Variable
Level Form
155
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
Table 5.3 Unit Root Test Results for the Corn and Sorghum-Models
Variable in Level Form
Estimated
Coefficient
Lag Length
(Months)
ADF Statistic*
-0.1325
-3.164
- 0.3334
- 3.160
-0.3185
12
-2 .6 8 1
-0.3058
-3.420
-0.1762
12
- 1.958
-0.1343
-3.139
- 1.2311
-3.1122
8
5
- 4.941
- 11.53
- 4.5644
12
-5.912
- 2.5840
-11.59
- 3.8964
12
- 6.009
- 1.2470
- 4.945
Table 5.4 Unit Root Test Results for the Milk Powder-Mod els
Variable
Estimated
Coefficient
Lag Length
(Months)
ADF Statistic*
-0.31842
- 3.399
-0.36167
-4.122
- 0.00964
- 1.597
-1.7752
- 8.992
- 2.4022
- 10.61
-0.38018
-5.964
L ev el
Form
156
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
1/ 2
(5.4)
v ,= . ( rn
- ) Ei=1< Io8 z - . - | o8 z -2)2
where m is the order of the moving average, and Z is the exchange rate variable. The
application of this formula to the data, using three period moving average (three months)
yields a series that is integrated of order one. The second volatility measure is the
absolute percentage change of the exchange rate defined as;
(5.5)
157
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
where e is the spot exchange rate and t refers to time. The resulting series from the
application of these formulas to the exchange data are shown in appendix 5. The plot of
the standard deviation-based volatility measures reveals that there was a period of
tranquility, from 1989 to 1994. This period corresponds to the government-managed
period of exchange rate determination in Mexico while the period of extremely high
volatility in exchange rate is observed during the Mexican peso crisis in 1994-1996. After
this episode, the peso is floated and the Mexico-US exchange rate is characterized by
high but manageable volatility periods (Figure 5.1). Figure 5.1 also shows that there are
no appreciable differences in the plot between both measures of volatility during the first
sixty months, from 1989 to 1994, but from 1995 to 2004, the variance of the volatility
measure V I is larger that the variance of the volatility measure V2.
15.0 4
io .o
5.0
0 .0 -f
-5.0 -r
-
10.0
Jan-1989
Dec-2004
Months
rerceniage Change of ER; V2
158
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
The general form of the GARCH (p, q), model proposed by Bollerslev (1986) is as
follow:
a ) = (0 + Y .^ e li +
;= i
1 = co+ a (L )? + P (L )a f ,
(5.6)
1=1
The literature indicates that many forms of volatility can be represented using the
particular GARCH (1, 1) specification (Bollerslev 1986; Enders 2003). The (p, q) in
parentheses is a standard notation in which the first number refers to how many
autoregressive lags or ARCH terms appear in the equation, while the second number
159
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
refers to how many moving average lags are specified, which is often called as the
number of GARCH terms.
Although this model is directly set up to forecast for just one period, it turns out
that based on the one-period forecast, a two-period forecast can be made. By repeating
this step, long-horizon forecasts can also be constructed. For the GARCH(1,1), the twostep forecast is a little closer to the long-run average variance that is the one-step forecast,
and the distant-horizon forecast is the same for all time periods as long as a + P < 1, i.e.
the coefficients of the ARCH and GARCH terms sum up to less than one. The GARCH
model is mean reverting and conditionally heteroskedastic but has a constant
unconditional variance (Engle, 2001). In this research, the GARCH (p, q) model has been
applied to real monthly exchange rate time series for the period 1989:1 to 2004:12. The
next section describes results from this model.
Following the empirical specification in chapter four, the procedure to calculate
volatility of the exchange rate was as follows: the residuals from a preliminary OLS
estimation were tested for the presence of ARCH behaviour in data. In particular the
following tests were performed.
Tests for Non-Normality. If the normality assumption is used to describe the conditional
error distribution then a property of ARCH is that the unconditional error distribution will
be non-normal with high values for kurtosis;
Test for Autocorrelation. The autocorrelation structure of the residuals and the squared
residuals were inspected. An indication of ARCH is that the residuals will be
uncorrelated but the squared residuals will show autocorrelation. Test statistics are given
160
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
p-value
59.7942
27986.7
0.000
Q(20)
31.14
0.053
Q2(20)
7.33
0.995
Test
Excess kurtosis
Jarque-Bera test
The high value for excess kurtosis indicates that the distribution is characterized
by leptokurtosis and an indication that the unconditional error distribution will be non
normal. The standard normal value for kurtosis is three (Engle, 2001). The Jarque-Bera
test statistic provides clear evidence to reject the null hypothesis of normality for the
unconditional distribution of the monthly percentage changes in the Mexico-US exchange
rate at the five percent significance level. Both kurtosis and the Jarque-Bera statistic
suggest the presence of ARCH errors.
Q(20) and Q (20), for up to twentieth order serial correlation in the residuals and the
squared residuals have been used. The sample autocorrelation function of the residuals
shows no autocorrelation while the squared residuals showed autocorrelation, both of
which indicate the presence of ARCH errors.
161
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
(5.7)
The estimation of the GARCH model requires the use of a numerical optimization
algorithm . To apply the GARCH technique, it is required to find the best fitted time
series model for the real exchange rate. An ARIMA model could perform this task. Once
it was done, a GARCH (p, q) model to capture the time varying conditional variance of
the real exchange rate can be formulated. The estimated model in this study has the
following specification:
a) ALnER = ao + ait.,-+ eu
b) e, 1 1 - N (0, h, ),
0 h, = A + Z A A , + S A V ,
1=1
(5.8)
1=1
where ALnER is the difference in the log of the real exchange rate with respect of the
previous period, I t-i is all relevant and available information at time t-i. st.j is the moving
average term; et is the disturbance term and hxis the conditional variance. Equations (5.8)
need to be estimated jointly for the sample period. If coefficients on e ,_i and h t-i are
2 The algorithm works as follows: starting values for the parameters o f the mean equation are obtained from
an OLS regression. For an ARCH(q) or GARCH(p,g) process the starting values for the parameters o f the
conditional variance equation are obtained from a regression of the OLS squared residuals on a constant
and q lags. For a GARCH process, the starting values for the parameters on the lagged conditional
variances are set to zero. Pre-sample estimates are required for the squared errors and the conditional
variances (McCullough and Renfro, 1999). An initial Hessian estimate is constructed from the outer
product of the gradient, analytic expressions for the derivatives are used (Bollerslev, 1986). Once initial
values are calculated, the estimation can proceed. SHAZAM uses a quasi-Newton algorithm. A description
of the quasi-Newton method is given in Judge, Griffiths, Hill, Lutkepohl and Lee (1985; pp. 958-960).
162
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
statistically significant, significant ARCH and GARCH effects are said to exist in the
data. The predicted values from in (c) of equation 5.8 provide a measure of real exchange
rate uncertainty.
Figure 5.2 Exchange Rate Data Expressed as First Difference of the Logarithm of
Real Mexico-US Exchange Rate (direr)
0.20
Months
Different ARCH and GARCH specifications were tested from ARCH (1) to
ARCH (8) and from GARCH (1,1) to different values of (p, q). Only the results from the
model with values p=l and q= l are reported in the text. Based on significance of the
parameters of the lagged squared error term and the lagged forecast variance (conditional
variance), a GARCH (1,1) specification was chosen. For this model, a test for non
normality and a test for autocorrelation were conducted. The data for this model is the
difference in the log of the real Mexico-US exchange rate and it is plotted in figure 5.2
163
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
shows high variation of the exchange rate particularly during the period of 1994-1996
which represents the Mexican Peso crisis period.
As can be seen in table 5.6, the GARCH term is highly significant. For this
specification, the coefficients of the ARCH and GARCH terms sum up to less than one
and both are positive. Thus, the model is mean reverting and conditionally
heteroskedastic, but has a constant unconditional variance (Engle, 2001).
Table 5.6 Results for the GARCH (1,1) Measure of the Exchange Rate Volatility
Dependent Variable: ALRER
Method: M L - ARCH
Sample: 1989:01 2004:12
Convergence achieved after 48 iterations
Coefficien
Std. Error
z-Statistic
Prob.
0.019748
0.005118
3.858659
0.0001
Variance Equation
0.000589
0.000233
2.531241
0.0114
A R C H (l)
0.003133
0.001578
1.985747
0.0471
G A R C H (l)
0.967598
0.017016
56.86385
0.0000
R-squared
-0.00262
0.0176
S.E. o f regression
0.04325
-3.5852
0.35166
Schwarz criterion
-3.5173
Log likelihood
Durbin-Watson stat
348.1810
1.7819
The conditional variance function from table 5.6 is specified in the following equation:
(5.9)
164
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
The predicted values from this equation, namely conditional variance, provide a measure
of Mexico-US exchange rate volatility. The series of the conditional variance used as a
measure of exchange rate volatility are reported in appendix 8.
The series of the conditional variance is used in the estimation of the import
demand equations (as a measure of exchange rate uncertainty). It is calculated using the
prediction values from equation 5.9. The resulting volatility values are plotted figure 5.4
0.007 -|
0.006 -
0.005 0.004 -
0.003 1
0.002
0 .0 0 1
ON
OO
On
ON
On
ON
OO
ON
o
o
Months
Looking at the normality test, the model shows non-normal errors since the
Jarque-Bera test is significant at 5 percent level. Thus the residuals are highly leptokurtic.
The Jarque-Bera statistic rejects the hypothesis of normal distribution. Furthermore, the
165
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
high values of kurtosis (greater than 3) and negative skewness (see table 5.3) suggest the
presence of ARCH residual in the data.
Figure 5.4 Results of the Normality Test for GARCH (1,1) Model
100
Series: Standardized Residuals
Sample 1989:01 2004:12
Observations 192
80.
60.
40-
Mean
Median
Maximum
Minimum
Std. Dev.
Skewness
Kurtosis
0.036896
0.178402
2.351028
-10.26257
1.000310
-6.178925
61.47810
Jarque-Bera
Probability
28579.24
0.000000
20 .
-10
-8
-2
166
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
Table 5.7 Results for the ARCH LM Test Using One Lag
F-statistic
Obs*R-squared
0.064932
0.131857
Probability
Probability
Test Equation:
0.937152
0.936198
Coefficient
Std. Error
t-Statistic
Prob.
0.978232
0.573764
1.704939
0.0899
0.026198
0.073127
0.358252
0.7206
STD R ESIDA2(-2)
0.002161
0.073125
0.029554
0.9765
R-squared
0.000694
S.E. o f regression
7.781392
6.957011
11322.86
Schwarz criterion
7.008280
Log likelihood
-657.9161
F-statistic
0.064932
2.000633
Prob (F-statistic)
0.937152
1.006770
Table 5.8 Results for the ARCH LM Test Using Sixteen Lags
ARCH Test using 16 lags
F-statistic
0.273465
Obs*R-squared
4.713546
Probability
Probability
0.997720
0.997004
Test Equation:
Dependent Variable: STD_RESIDA2
Sample; 1990:05 2004:12
Coefficient
Variable
0.908177
C
STD_RESIDA2(-1)
0.018690
STD_RESIDA2(-2)
-0.003214
STD_RESIDA2(-3)
0.149832
0.037721
STD_RESID A2(-4)
STD_RESIDA2(-5)
0.017332
STD_RESIDA2(-6)
-0.032438
STD_RESID A2(-7)
-0.018622
STD_RESIDA2(-8)
-0.004250
STD_RESIDA2(-9)
-0.002147
STD_RESID A2(-10)
0.044182
STD_RESID a2(-11)
0.010525
STD_RESIDA2(-12)
-0.009333
STD_RESIDA2(-13)
-0.021199
STD_RESIDA2(-14)
-0.016366
STD_RESID A2(-15)
-0.012280
STD_RESIDA2(-16)
-0.005713
Log likelihood
-613.7478
Durbin-W atson stat
2.000125
Std. Error
t-Statistic
0.690415
1.315407
0.079302
0.235682
0.079307
-0.040523
0.079297
1.889504
0.080165
0.470548
0.080218
0.216067
0.080225
-0.404341
-0.232236
0.080188
0.080199
-0.052990
0.080200
-0.026771
0.080187
0.550985
0.080223
0.131194
0.080214
-0.116350
0.080161
-0.264451
0.079293
-0.206406
0.079304
-0.154845
0.079296
-0.072052
F-statistic
Prob(F-statistic)
Prob.
0.1903
0.8140
0.9677
0.0606
0.6386
0.8292
0.6865
0.8167
0.9578
0.9787
0.5824
0.8958
0.9075
0.7918
0.8367
0.8771
0.9427
0.273465
0.997720
167
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
(5.10)
3 This method performs a local averaging o f the observations when estimating the regression function.
168
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
where ER is the Mexico-US real exchange rate, M l is the money demand, GDP is real
gross domestic product, i is the interest rate and CPI is the general consumer price index.
The star (*) indicates the difference between the fundamentals in Mexico and United
States. The R-square in table 5.10 implies that the selected variables explain a high
proportion of Mexico-US exchange rate variability, and also that all coefficients are
statistically significant at 5 percent.
Variables
Ml
GDP
TB
CPI
CONSTANT
0.9645
R-Square
adjusted
0.9638
Estimated
Coefficient
Standard
Error
t-ratio
P-value
-0.9590
-14.0840
-0.1950
-0.0597
226.2400
0.1199
0.3374
0.0314
0.0220
5.6780
-7.9960
-41.7400
-6.2040
-2.7090
39.8500
0.0000
0.0000
0.0000
0.0070
0.0000
4 In general, non-parametric regression analysis traces the dependence o f a response variable (Y) on one or
several predictors (Xs) without specifying in advance the function that relates the response to the predictors
169
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
(5.11)
where:
R = ER/SQRUHAT
M = M l/SQRUHAT
G = GDP/S QRUHAT
T = TB/SQRUHAT
C = CPI/SQRUHAT
170
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
The results of the OLS regression from equation 5.11 are presented in table 5.10. The
above regression generates an R-square of 0.71 which indicates that the selected variables
explain the exchange rate changes fairly well. The estimated parameters of all variables
but one are statistically significant at 5 percent.
Std Error
t Ratio
P-Value
1.8800
0.3042
6.1800
-0.9003
0.1151
-7.8190
0.0000
0.0106
0.0929
0.1143
0.9090
0.3991
0.0483
8.2670
0.0000
-0.6115
1.1660
-0.5247
0.6000
CONSTANT
0.0000
The values from the results presented in table 5.10 are taken as a measure of volatility of
the Mexico-US exchange rate. Figure 5.4 shows the behaviour of the exchange rate
volatility generated by this method for the studied period. It is clear that volatility rose as
expected after month 78, which correspond to the beginning of the floating period of the
exchange rate in Mexico.
171
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
25.0 0
20.00
15.00 -
10.00
5.00 -
0.00
OO
Ov
CN
M onths
5.4 Summary
This chapter discusses the main results of unit root test and estimated four alternative
measures of Mexico-US real exchange rate volatility. The ADF test was conducted to
determine the integration properties of individual series used in each commodity model.
The ADF statistic from the unit root test for the variables in level form is used to test for
the presence of unit root in the data, while the ADF statistic for the series in first
difference form is used to test and confirm if the series is integrated of order one.
To determine optimum lag length for each series, the Akaike (1973) Information
Criterion (AIC) and Schwarz (1978) Criterion (SBC) were applied. Both criteria selected
the same optimal lag length. The results for the ADF test show that all macroeconomic
172
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
variables, as well as the price and quantity variables included in each of the four
commodity models are non-stationary, have a unit root in the univariate representation.
The results from the estimation of four different volatility measures show that the
volatility measures based on standard deviation, moving average standard deviation (V I)
and percentage change of the exchange rate (V2) are very similar in tracking the
exchange rate fluctuations. With respect to the GARCH model, it was found that
GARCH(1,1) describes the exchange rate data well based on a set of standard tools .
Finally, the semi-parametric regression was performed using the kernel regression
approach with a normal kernel function to obtain nonparametric measures of the MexicoUS exchange rate volatility. These four volatility measures are incorporated in the
estimation of the import demand functions for the selected commodities in chapter six.
The next chapter reports and discusses the results of the cointegration analysis.
173
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
CHAPTER 6
MAXIMUM LIKELIHOOD COINTEGRATION RESULTS
6.1 Introduction
The purpose of this chapter is to present and discuss the results of Maximum Likelihood
Cointegration analysis. This chapter is divided into four sections. The first section deals
with an overview of the econometric modeling strategy used to obtain the results. Section
two is devoted to determine the optimal lag structure of each model. The third section
focuses on the determination of the cointegrating vectors and the discussion of the longrun cointegrating relationships. The final section summarizes the main results and
concludes the chapter.
174
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
Dhrymes (1998) and Clements and Hendry (2002), Harris and Solis (2003)) and are
becoming increasingly popular in time series econometric analysis. The results obtained
using the VAR formulation are sensitive to the lag-length choice and alternative
specifications of the lag structure of a given VAR model generate quite different results
(Haffer and Sheenan, 1991). Therefore, it is important to determine the optimal lag
structure of each of the VAR models.
176
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
Table 6.1 Optimal Lag Length Determination for the Tomato Models
Number of
periods lags
Number of Lags
Log Likelihood
M L-Test
LRm X2 *(d>
Test Result
14.02
Reject Lag 8
14.02
Accept Lag 9
14.02
**
8
9
9
10
10
8
9
9
10
10
-39.503
27.967
-40.202
-40.202
11.508
-40.645
-40.645
6.758
TOMATO MODEL V2
-36.938
-37.705
-37.705
-38.015
-38.015
30.675
14.02
Reject Lag 8
8.045
14.02
Accept Lag 9
7.534
14.02
**
8
9
9
10
10
-40.907
-41.834
-41.834
-42.434
-42.434
20.403
14.02
Reject Lag 8
3.599
14.02
Accept Lag 9
-332.696
14.02
**
17.413
14.02
Reject lag 9
7.507
14.02
Accept lag 10
-2.866
14.02
**
9
10
10
11
11
-7.318
-7.988
-7.988
-8.614
-8.614
Eleven to Twelve
* df = 7, 5 percent.
** Once an optimal lag is achieved, no further lags are tested
For the tomato models, the optimal lag length using volatility measure V I, V2
and VG is nine lags (%2= 14.02, 83 d f which is significant at the five percent level),
while it is ten using the volatility measure VN. For the maize models, the optimal laglength using V I is ten lags while using V2, VG, and VN it is nine lags (Table 6.2).
177
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
Table 6.2 Optimal Lag Length Determination for Maize Models (New data)
Number of
periods lags
Log Likelihood
M L-Test
LRmx2 *(d )
Test Result
14.294
14.02
Reject Lag 9
7.586
14.02
Accept Lag 10
-67.062
14.02
**
21.808
14.02
Reject lag 8
13.074
14.02
Accept Lag 9
8.636
14.02
**
17.495
14.02
Reject lag 8
13.403
14.02
Accept Lag 9
9.249
14.02
**
18.308
14.02
Reject lag 8
10.757
14.02
Accept Lag 9
8.080
14.02
**
MAIZE MODEL V I
Nine to Ten
Ten to Eleven
Eleven to Twelve
-37.353
10
-37.902
10
-37.902
11
-38.535
11
-38.535
MAIZE MODEL V2
Eigth to Nine
Nine to Ten
Ten to Eleven
-34.713
-35.258
-35.258
10
-35.761
10
-35.761
MAIZ M ODEL VG
Eigth to Nine
Nine to Ten
-9.830
-10.267
-10.267
10
-10.782
Ten to Eleven
10
-10.782
Eigth to Nine
-4.483
-4.941
-4.941
10
-5.354
10
-5.354
MAIZE MODEL VN
Nine to Ten
Ten to Eleven
* df = 7, 5 percent.
** Once an optimal lag is achieved, no further lags are tested
The results of the optimal lag-length determination for the sorghum models are
presented in table 6.3. For all sorghum models, the optimal lag-length is ten. For milk
powder models using V I and V2 the optimal lag-length appears to be nine, but for
models using VG and VN it is ten (Table 6.4). While it is tempting to attribute the
differences in optimal lag-lengths to the specifications of volatility measure, a definitive
answer in this regard requires additional investigation which is beyond the scope of this
research.
178
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
Table 6.3 Optimal Lag Length Determination for Sorghum Models (New data)
Number of
periods lags
Log Likelihood
M L-Test
LRm x2 *(d)
Test Result
SORGHUM M ODEL V I
Nine to Ten
Ten to Eleven
Eleven to Twelve
-37.335
10
-38.060
10
-38.060
11
-38.649
11
-38.649
18.854
14.02
Reject Lag 9
7.077
14.02
Accept Lag 10
-1.220
14.02
**
14.686
14.02
Reject lag 9
10.516
14.02
Accept Lag 10
-1.473
14.02
**
16.164
14.020
Reject lag 9
6.955
14.020
Accept Lag 10
-1.978
14.02
**
881.0
14.02
Reject lag 9
7.580
14.02
Accept Lag 10
-1.489
14.02
**
SORGHUM MODEL V2
Nine to Ten
Ten to Eleven
Eleven to Twelve
-35.127
10
-35.692
10
-35.692
11
-36.568
11
-36.568
SORGHUM MODEL VG
Nine to Ten
Ten to Eleven
Eleven to Twelve
-10.367
10
-10.989
10
-10.989
11
-11.568
11
-11.568
SORGHUM MODEL VN
Nine to Ten
Ten to Eleven
Eleven to Twelve
-39.403
10
-5.519
10
-5.519
11
-6.151
11
-6.151
* df = 7, 5 percent.
** Once an optimal lag is achieved, no further lags are tested
179
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
Table 6.4.0ptimal Lag Length Determination for Milk Powder Models (New data)
Number of periods
lags
Number of
Lags
Log Likelihood
M L-Test
LRmx2 *(d)
Test Result
14.218
14.02
Reject Lag 9
9.269
14.02
Accept Lag 10
-1.290
14.02
**
15.302
14.020
Reject lag 9
10.247
14.020
Accept Lag 10
-1.471
14.020
**
1355.2
14.02
Reject lag 8
12.529
14.02
Accept Lag 9
7.620
14.02
**
30.737
14.02
Reject lag 8
9.235
14.02
Accept Lag 9
10.671
14.02
**
Ten to Eleven
Eleven to Twelve
-31.684
10
-32.231
10
-32.231
11
-33.003
11
-33.003
Ten to Eleven
Eleven to Twelve
-29.314
10
-29.902
10
-29.902
11
-30.756
11
-30.756
Nine to Ten
Ten to Eleven
-38.502
-4.621
-4.621
10
-5.103
10
-5.103
Nine to Ten
Ten to Eleven
-32.949
-33.718
-33.718
10
-34.073
10
-34.073
* df = 7, 5 percent.
** Once an optimal lag is achieved, no further lags are tested
180
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
a,
would
indicate slow adjustment while a high coefficient indicates rapid adjustment to the long
run equilibrium (Johansen and Juselius, 1990).
181
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
The critical values of the trace and maximum eigenvalue statistics at the five
percent level of significance reported by MacKinnon (1999) are used in this study
because they provide more refined critical values for the these tests than those available
elsewhere. The next section presents and discusses the results of the Johansens
maximum likelihood cointegration analysis applied to the four commodities. Then, an
attempt is made to bring out informally which volatility measure provides better results
than others and why. The above discussion is conducted taking into account the
properties of different volatility measures as well as consistency of the results with theory
in terms of both signs and magnitude of the long-run coefficients.
1 While the theory suggests that both the trace and maximum eigenvalue test select the same number of
cointegrating vectors, in a small sample, trace test tends to have more than the maximum eigenvalue test. It
is suggested in the literature to employ both tests in empirical work to determine the number of
cointegrating vector in a given system (Lutkepohl et al, 2001).
182
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
Following the above criteria, it was determined that for tomato model using V I,
there are two cointegraing relationships. The long-run relationships between United
States tomato imports from Mexico and the set of explanatory variables can be
represented as linear functions.
(6.1)
(6.2)
The first cointegrating vector, shown in equation 6.1, is normalized using the
coefficient of import quantity while the second cointegrating vector represented by
equation 6.2, is normalized using the coefficient of own price of the commodity. In
equation 6.1 all estimated coefficients but one appear to be consistent with demand
theory. The own price has the expected negative sign implying that if the price of fresh
tomatoes imported into the United States from Mexico increases by 10 percent, the
quantity imported would fall by 6.8 percent. The substitute price does not exhibit the
expected positive sign. A possible explanation could be that the tomato produced in the
United States serves as a complement to the imports of tomato from Mexico by supplying
the eastern seaboard region of the United States.
The elasticity of income exhibits the expected positive sign. This implies that
when per capita personal disposable income in the United States increase, say by 10
percent, imports of fresh tomatoes from Mexico increase also by 10 percent. The high
value of income elasticity could be due to the high quality of Mexican tomatoes imported
183
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
into the United States and may also reflect the fact that during the winter season high
quality vine-ripped tomatoes from Mexico are highly appreciate by consumers in the
United States.
The positive sign of the exchange rate variable indicates that the bilateral MexicoUnited States exchange rate exerts a positive long run effect on the United States imports
of tomatoes from Mexico. If there were a 10 percent devaluation of the Mexican peso, the
United States import demand for fresh Mexican tomatoes would increase by about 14.2
percent. The exchange rate volatility variable exhibits an expected negative sign,
suggesting that exchange rate uncertainty reduces the volume of tomato imported from
Mexico into the United States. Note that all results are consistent with the expected utility
maximization model outlined in chapter three, so that ^ 0 > 0 and
< 0. These
dER
av,.
results are also consistent with the findings in the literature (e.g. Chowdhury (1993),
Koray and Lastrapes (1989), Qian and Varangis (1994), and Gagnon (1993)) about the
relationship between volatility and trade flows.
The Mexico-United States wage ratio has a positive sign as expected. This
indicates that if agricultural wage rate in the United States increases, imports of fresh
tomatoes from Mexico would also increase. Thus, an increase in the price of labour in the
importing country is followed by an increase in tomato imports from Mexico. Since the
wage rate in the United States was eight to nine times higher than that in Mexico during
the sample period, this result is quite reasonable. Malaga et al (2001) found that wage
rate differential between Mexico and the United States explained about 2 percent of the
growth of Mexico-United States tomato trade flows during 1990 to 1998.
184
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
(6.3)
(6.4)
Equation 6.3 shows that all variables have coefficients with theoretically expected signs.
The negative sign of the own price variable implies a negative relationship between the
185
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
United States imports of fresh tomatoes from Mexico and tomato import price. Since the
coefficient of own price variable is less than one, it implies that the demand is inelastic.
Thus, for a given increase in import price, the imported quantity will decrease
proportionally less than the change in own price of tomato. Based on the positive sign of
the related price, tomato produced in the United States is a substitute for tomato imported
from Mexico.
The income elasticity of import demand is inelastic. It implies that when percapita personal disposable income in the United States increases, say by 10 percent, the
imports of fresh tomato from Mexico increases by almost 6 percent. The above value of
income elasticity of demand suggests that tomato is a normal good.
The positive value of the exchange rate variable indicates that a devaluation of the
Mexican peso exerts a positive long run effect on the United States imports. The
exchange rate volatility exhibits an expected negative sign. Some studies in the literature
dealing with agricultural trade flows and volatility have reported a negative relationship
between imports and exchange rate volatility (e.g. Anderson and Garcia, 1989; Sarker,
1993; Fabiosa, 2002; and Cho, et.al. 2002).
The elasticity of the Mexico-United States wage ratio shows the expected positive
sign, i.e., as the agricultural wage rate in United States increases, then imports of
tomatoes from Mexico also increases. It is expected that when rural wage rates in the
United States become expensive relative to the price of labour in Mexico, tomato imports
from Mexico would also increase. The positive impact of a lower wage rate in Mexico
with respect to that in the United States were also reported in the literature by Malaga et
al, (2001); Love and Lucier (1996); and Schwentesius and Gomez Cruz (1992) among
186
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
others. They argued that the United States imports of fresh vegetables were sensitive to
labour cost differentials between Mexico and the United States.
The estimated weights ( a ' s ) for fresh tomato model with V2, suggest that in the
event of any shock to the system exchange rate volatility, the tomato import price as well
as the price of tomato imported from Canada will respond faster than any other variables
to bring this system back to the long run equilibrium.
(6.5)
Equation 6.5 shows that all variables but volatility have coefficients with theoretically
expected signs. Since the coefficient of own price variable is less than one, it implies that
the demand is inelastic. Thus, for a given increase in import price, the imported quantity
will decrease proportionally less than the change in own price of tomato. Based on the
positive sign of the related price, tomato produced in the United States is a substitute for
187
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
tomato imported from Mexico. Also the income elasticity is less than one, which implies
a low response of import demand to changes in income.
The positive value of the exchange rate variable indicates that a devaluation of the
Mexican peso exerts a positive long-run effect on the United States imports. Since the
exchange rate volatility exhibits a positive sign, it implies that a positive relationship
exists in the long-run between Mexico-United States trade tomato flows and volatility of
the exchange rate. The above result is consistent with the hypothesis of risk-averse
importers examined in the theoretical model in chapter three of this study. The sign of the
wage rate ratio is consistent with the theoretical model and with tomato models using V 1
and V2 in this study. The estimated weights ( a 's ) for imports of fresh tomato suggest
that if an external shock affects the long run equilibrium, the price of tomato produced in
the United States respond faster than any other variables to bring this system back to the
long run equilibrium.
188
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
( 6 .6 )
(6.7)
______
The above equations show that all parameters exhibit the expected theoretical
sign, the exception being the price of tomato produced in the United States which has a
negative coefficient, the own price elasticity is almost unity in equation (6.7). On the
other hand, the related price has a negative sign, suggesting a complementary relationship
between tomatoes produced in the United States and tomatoes imported from Mexico into
the United States. While exchange rate indicates a positive relationship between trade
flows and Mexico-United States exchange rate, the volatility of exchange rate exert a
negative impact on trade flows. The sign of the wage rate ratio is positive and consistent
with the previous tomato models in this study. The estimated weights ( a ' s ) for imports
of fresh tomato suggest that if an external shock affects the long run equilibrium, the own
price of tomato and the import volume will respond faster than any other variables to
bring this system back to the long run equilibrium.
189
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
and Gutierrez (1983) found elasticities of -0.31 and -0.44 respectively while Malaga et al
(2001 ) found a value of -0.62.
The magnitude of income elasticity of import demand across models ranges from
0.22 to 1.00. The above values are consistent with income elasticity values found in the
literature, even though the estimation methods and samples were quite different. The
values of income elasticities found in the literature for the United States tomato import
demand range from 0.23 (Mitterhammer, 1978), to 1.47 (Salcedo-Baca, 1990). Malaga et
al. (2001) reported an intermediate value of 0.92 for the elasticity of income.
While the coefficients of exchange rate variable in all four models have the
expected sign, it is not the case for the volatility variable which has an unexpected
positive sign in one model. The positive sign is inconsistent with the theoretical model
outlined in chapter three. The negative sign of the coefficient of the exchange rate
volatility variable is obtained in three out of four cases which are consistent with the
comparative statics results presented in chapter three of this thesis.
It is worthwhile to note that Arellano et al (1998) and Malaga et al (2001) also
found a positive effect of the devaluation of the Mexican peso on the United States
imports of fresh vegetables. They reported that the United States fresh vegetable trade
resulted in higher United States imports in periods of devaluation of the Mexican peso,
particularly when it was allowed to move freely against the United States dollar. Malaga
et al (2001 ) also argued that in the case of fresh tomato, the peso devaluation explained
87 percent of the short-run changes in the United States tomato imports from Mexico
during the period of 1993-1996.
190
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
(6 .8 )
Equation 6.8 can be interpreted as the long-run import demand function for maize in
Mexico. In this equation, all variables exhibit a priori expected signs except for the
variable INV (hogs inventory) which shows a negative sign. The magnitude of the own
price coefficient suggests that the import demand function for the com from the United
States into Mexico is inelastic.
The exchange rate elasticity indicates that the bilateral exchange rate devaluation
exerts a positive long run effect on Mexicos maize imports from the United States.
Given the positive sign of the price of a substitute commodity, the domestic price of
maize produced in Mexico (Pmx) is considered a substitute for maize imported from the
United States. The coefficient of the exchange rate volatility variable exhibits a negative
sign as expected. Thus, in the long-run, uncertainty of the exchange rate exerts a
significant negative effect on maize trade flows between Mexico and the United States.
The coefficient of INV variable shows a negative sign which is contrary to the theoretical
191
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
expectation because M exicos maize imports consist of yellow com (80-90 percent)
which is mostly used as an input by hog producers, poultry producers, starch
manufacturers and flour millers (Avalos-Sartorio, 1998).
The estimated weights ( a ' s ) for imports of maize suggest that in the event of any
disturbance or shock to the system that affect the long run relationship, the quantity
imported and volatility of the exchange rate will respond faster than any other variables
to bring this system back to the long run equilibrium (Appendix 25).
(6.9)
Equation 6.9 can be interpreted as the long-run import demand function for maize in
Mexico. In this equation, all variables have coefficients with the expected signs. The
volatility variable shows that a negative effect exists between imports of maize from the
United States into Mexico and exchange rate uncertainty.
192
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
------------------------------------
(6.10)
Equation 6.10 represents the long-run import demand function for maize in Mexico. In
this equation all the variables have coefficients with the theoretically expected signs. All
elasticity values of import demand for maize are inelastic with the exception of the
elasticity of the related price.
The sign of the exchange rate volatility indicates a negative relationship between
Mexico-US trade flows of maize and exchange rate uncertainty. The estimated weights
(ex's) for imports of maize using VG suggest that in the event of any disturbance
affecting the long run equilibrium, the quantity of maize imports and the import price
variable will responds faster than any other variable to bring this system back to the long
run equilibrium
193
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
cointegration at the five percent level of significance. Both statistics indicate that there is
just one cointegrating relationship in this model. This cointegrating vector represents the
long-run relationship among the variables in this system and can be specified as
(6.11)
Equation 6.11 can be interpreted as the import demand function for maize in
Mexico. The coefficients of all variables but volatility of the exchange rate exhibit
expected signs. Similar to previous maize models, the import demand for maize is
inelastic. The positive sign of the related price suggest that maize produced in Mexico
can be treated as a substitute of maize imported from the United States. W ith regard to
the income variable, the positive sign suggests that as per capita personal disposable
income in Mexico increases, the import demand for maize from the United States will
rise.
Volatility of exchange rate does not have the expected negative sign in this model.
However, the positive sign of the exchange rate coefficient is an indication that, in the
long-run, devaluation of the United States currency relative to the Mexicos currency will
lead to an increase in the volume of maize imported into Mexico from the United States.
The estimated weights ( a ' s ) for maize import demand with VN suggest that in the event
of any disturbance affecting the long run equilibrium, the volume of imports, the INV
variable and the own price of maize will respond faster than any other variable to bring
this system back to the long run equilibrium.
194
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
(6.12).
Equation 6.12 can be interpreted as the long-run import demand function for
sorghum in Mexico. The results show that all the variables in this equation have
coefficients with the expected signs. As expected, exchange rate and its volatility had a
positive and a negative effect respectively on import demand for sorghum in Mexico. The
demand is price inelastic, indicating that changes in prices are followed by proportionally
smaller changes in imports of sorghum. This result is true also for the cross price
elasticity. Given the positive sign of the related price (Psd) for Mexicos imports of
sorghum, domestic sorghum is a substitute for sorghum imported from the United States
into Mexico. The coefficient of hog inventory (INV) variable has a positive sign. A
priori, it was expected that given an increase in inventory, import demand of sorghum
would also increase. This is because all the sorghum imports are allocated to animal feed
industries in Mexico (Zahniser and Coyle, 2004).
The estimated weights ( a ' s ) for imports of sorghum suggest that in the event of
any disturbance or shock to the system that affect the long run relationship, inventory of
hogs in Mexico and import volume of sorghum will respond faster than any other
196
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
variable to bring this system back to the long run equilibrium. These are followed by
volatility of the exchange rate.
(6.13).
From the above relationship, equation 6.13 can be interpreted as the long-run
import demand function for sorghum in Mexico. In this equation all variables exhibit a
priori expected signs. The negative sign of the own price variable implies a negative
relationship between the Mexican imports of sorghum from the United States and
sorghum import price. The magnitudes of the own price and cross price coefficients
suggest that the long-run import demand function for United States sorghum in Mexico is
inelastic. The exchange rate elasticity indicates that a devaluation of the bilateral
exchange rate exerts a positive long run effect on Mexicos sorghum imports from the
United States. The coefficient of the exchange rate volatility exhibits a negative sign as
expected. Thus, in the long-run, uncertainty of the exchange rate exerts a negative effect
197
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
on sorghum trade flows between Mexico and the United States. The coefficient of INV
variable shows a positive sign, as expected. A probable explanation is that M exicos
sorghum imports are mostly used as input by hog producers in Mexico (e.g. AvalosSartorio, 1998).
The estimated weights ( a ' s ) for imports of sorghum suggest that in the event of
any disturbance or shock to the system that affect the long run relationship, the inventory
and volatility will respond faster than any other variables to bring this system back to the
long run equilibrium.
(6.14).
Equation 6.14 represents the long-run import demand function for sorghum in Mexico. In
this equation all variables have coefficients with the expected signs and the magnitudes
indicate inelastic responses.
198
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
The sign of the exchange rate variable indicates a positive relationship between
Mexico-US trade flows of sorghum and the bilateral Mexico-United States exchange rate.
On the other hand, the sign of the volatility variables indicates a negative relationship
with respect to sorghum import volume, as expected. The estimated weights ( a ' s ) for
imports of sorghum using VG suggest that in the event of any disturbance affecting the
long run equilibrium, the INV and the import volume as well as the own price of the
commodity will respond faster than any other variable to bring this system back to the
long run equilibrium.
(6.15)
Equation 6.15 can be interpreted as the long-run import demand function for
sorghum in Mexico. The results show that all the variables in this equation have expected
signs with the exception of the INV variable. Exchange rate and its volatility suggest a
positive and negative relationship with respect to import demand of sorghum
respectively. The own price elasticity suggests an inelastic demand, indicating that
199
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
200
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
The exchange rate has an expected sign in all four sorghum models, suggesting
that a devaluation of the United States will favour imports of sorghum into Mexico from
the United States and vice versa. The elasticity of exchange rate with respect to imported
volume into Mexico is inelastic in all four models and it ranges from 0.10 using VN to
0.16 using VG. The Volatility of the exchange rate has the expected sign in all four
sorghum models. The magnitude of the effects of exchange rate volatility on sorghum
imports into Mexico from the United States ranges from -0.25 using volatility VN to 0.54 using volatility V I and V2. The negative relationship between exchange rate
uncertainty and import volume is consistent with the expected utility maximization
hypothesis assuming risk-aversion.
The sign for the INV variable is the expected in three of four models while model
using VN has a positive sign. The positive relationship between import demand and hogs
inventory is not unexpected given the increasing domestic demand for sorghum in
Mexico to meet the growing demand for feed in the livestock industry. In the case of a
disturbance to the long-run equilibrium, prices, exchange rate volatility and hog
inventory would adjust faster than any other variables to restore the long-run equilibrium.
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
that there is at most one cointegraing vector in this system. The following cointegrating
vector represents the long-run relationship among the variables in this system:
(6.16).
Equation 6.16 can be interpreted as the long-run import demand function for milk
powder in Mexico. In the above equation all the variables have the expected sign. The
magnitude of the own price coefficient suggests that the import demand function for
United States milk powder in Mexico is inelastic. The coefficient of the related price has
the expected positive sign implying a positive relationship between M exicos import
demand and related price. For milk imports, Mexico has a broad range of sources,
including the United States, Canada, Germany, New Zealand, Denmark and Ireland
among others. So, relatively high price responsiveness of import volume to changes in
prices is expected.
The sign of the exchange rate variable indicates that a devaluation of the bilateral
United States-Mexico exchange rate exerts a positive long-run effect on M exicos milk
powder imports from the United States. On the other hand, the negative sign of the
exchange rate volatility suggests a negative relationship between imports of milk powder
from the United States into Mexico and exchange rate uncertainty.
With regard to the response of import demand to changes in the real exchange
rate, it depends on the degree of elasticity of import demand. For this model, the
exchange rate elasticity is smaller than one, which implies an inelastic demand. The
income elasticity is inelastic, implying that given a 10 percent change in Mexican
income, milk powder import demand would change only by 3.28 percent.
202
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
The MSF variable (short-fall of milk and milk products in Mexico) has the
expected positive sign, indicating that as internal demand for milk and milk products rise,
imports from the United States also rise. The positive relationship between milk short-fall
and import volume is consistent with the increasing dependency of Mexico for milk and
milk products from the international market, especially from the United States. The
estimated weights ( a ' s ) for imports of milk powder suggest that in the event of any
disturbance to the system, both the own price and imported volume will respond faster
than any other variable to bring this system back to the long run equilibrium.
(6.17).
(6.18).
203
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
From the above long-run relationship, equation 6.17 can be interpreted as the
long-run import demand function for milk powder in Mexico. In the above equation all
the variables, but the MSF variable, have the expected sign. The magnitude of the own
price coefficient suggest that the import demand function for United States milk powder
in Mexico is inelastic. The coefficient of the related price has the expected positive sign
implying a positive relationship between M exicos import demand and related price. The
sign of the exchange rate coefficient indicates that a devaluation of the bilateral exchange
rate exerts a positive long-run effect on Mexicos milk powder imports from the United
States. The negative sign of the exchange rate volatility suggests a negative relationship
between imports of milk powder from the United States into Mexico and exchange rate
uncertainty.
With regard to the response of import demand to changes in the real exchange
rate, it depends on the elasticity of import demand. For this model the exchange rate
elasticities is less than one, which implies an inelastic demand. The MSF variable (short
fall of milk and milk products in Mexico) has a negative sign which is not as expected.
The estimated weights ( a ' s ) for imports of milk powder suggest that in the event
of any disturbance or shock to the system that affect the long run relationship(s), both
imported volume and volatility will respond faster than any other variable to bring this
system back to the long run equilibrium.
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
are two cointegrating vectors, while the maximum eigenvalue statistic indicated the
presence of just one stable long-run relationship (see appendix 35). Thus, it was
concluded that there is, at least, one cointegraing vector in this system. The following
cointegrating vector represents the long-run relationship among the variables in this
system
(6.1i
Equation 6.18 can be interpreted as the long-run import demand function for milk
powder in Mexico. In the above equation, all the variables have the expected economic
sign. The magnitude of the own price coefficient suggests that the import demand
function for United States milk powder in Mexico is unity. The coefficient on the related
price has the expected positive sign implying a positive relationship between M exicos
import demand of milk powder and related price, but the cross price elasticity is elastic.
The sign of the exchange rate coefficient indicates that the bilateral exchange rate
devaluation exerts a positive long-run effect on M exicos milk powder imports from the
United States. Similar to models with V2, the negative sign of the exchange rate volatility
suggests a negative relationship between imports of milk powder from the United States
into Mexico and exchange rate uncertainty.
With regard to the response of import demand to changes in the real exchange
rate, it depends on the magnitude of the elasticity of import demand. For this model, the
exchange rate elasticity is less than one, which implies an inelastic demand. The MSF
variable (short-fall of milk and milk products in Mexico) has the expected positive sign,
205
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
indicating that as internal demand for milk and milk products rise, imports from the
United States also rise.
The estimated weights ( a ' s ) for imports of milk powder suggest that in the event
of any disturbance or shock to the system that affect the long run relationship(s), both the
own price and the imported volume will respond faster than any other variable to bring
this system back to the long run equilibrium.
(6.19)
From the above long-run relationships, equation 6.19 can be interpreted as the
long-run import demand function for milk powder in Mexico. In the above equation all
the variables have the expected economic sign. The magnitude of both the own price and
related price coefficients suggest that the import demand function for United States milk
powder in Mexico is elastic. The coefficient of the related price has the expected positive
206
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
sign implying that milk powder imports from Canada into Mexico can be considered a
substitute for milk powder imports from the United States. The sign of the exchange rate
coefficient indicates a devaluation of the bilateral exchange rate exerts a positive long-run
effect on M exicos milk powder imports from the United States. The negative positive
sign of the exchange rate volatility suggests a negative relationship between imports of
milk powder from the United States into Mexico and exchange rate uncertainty. The MSF
variable (short-fall of milk and milk products in Mexico) has the expected positive sign,
indicating that as internal demand for milk and milk products rise, imports from the
United States also rise.
The estimated weights ( a ' s ) for imports of milk powder suggest that in the event
of any disturbance or shock to the system that affects the long run relationship, both
imported volume and own price of milk powder will respond faster than any other
variable to bring this system back to the long run equilibrium.
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
The bilateral exchange rate has a positive sign as expected in all four models,
suggesting that a devaluation of the United States dollars will favour imports of milk
powder into Mexico from the United States. The elasticity of the exchange rate is
inelastic and ranges from 0.21 using VG and VN to 0.45 using V2. On the other hand,
volatility of the exchange rate shows the expected negative sign in all models. The
magnitude ranges from -0.27 for models using VN to -0.79 for using V2. The sign for the
MSF variable is as expected. The positive relationship between import demand of milk
powder and MSF reflects the increasing dependency of Mexico for milk powder from the
international market.
For milk models, the fastest speed of adjustment in the case of a disturbance to the
long-run equilibrium, given by the largest value of the coefficients in the weight matrix
(a), suggest that the import volume of the commodity, own price, and milk short-fall
would adjust faster than any other variable to restore the long-run equilibrium.
208
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
and the properties of the volatility measures employed. The results are compared among
models using different volatility specification.
Considering the fact that each of the selected commodities has good number of
substitutes and that small proportion of the consumers expenditure is devoted to these
commodities, they can be viewed as necessary goods. For such goods, it is expected that
the magnitudes of price and income elasticities be less than one. It is also expected that in
the long-run, own price, income and related price have stronger influence on import
demand than exchange rate, volatility and wage rate.
For tomatoes, it was found that while models using V I and VN have all the
coefficients with the a priori expected signs, except that of the related price, models V2
and model VG have all coefficients with the expected signs. In terms of magnitude of the
estimated coefficients, only model using VG has coefficients that are less than one while
the other models for tomato have two or three coefficients with magnitude around two,
which are contrary to the above reasoning.
For the maize model using V 1 two of six coefficients are greater than one and its
signs are as expected. For model using V2, all variables have the expected signs and only
the volatility of the exchange rate has coefficient greater than one. Model with VG has all
its coefficients with magnitude less than one with the exception of the coefficient of the
related price. Model using VN shows consistent magnitudes, but two variables have
coefficients greater than one and volatility has a positive sign. For sorghum, the model
using V I has all its coefficients smaller than one. Also all the coefficients have the
expected signs. Model using V2 has all its coefficients less than one and the signs are as
expected. On the other hand, model using VG not only have all its coefficients with the
expected signs but also the magnitudes are reasonable for a normal good. Models using
209
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
volatility VN has two coefficients greater than one and some of them are smaller in the
long-run with respect to the long-run. It is interesting to notice that model VG has
consistent coefficients in terms of signs and magnitudes.
The literature suggests that the volatility measure generated through the GARCH
method provides more accurate forecasts of conditional variances and co-variances than
those based on variance-standard deviation or non-parametric methods because of its
ability to model time-varying conditional variances. So, this method seems more suitable
for modeling exchange rate risk than others. Based on the above discussion, it can be
concluded that the model using a volatility measure generated by the GARCH method
provides better results than those obtained by using V I, V2 and VN.
Comparing results across the four commodity models, it was found that, for all
import demand functions, models using VG provide more consistent results with theory
than models using other volatility measures. For the four models, all the variables have
the a priori expected signs. Similarly, the magnitudes are reasonable for each of the four
models. The cointegration results suggest that in the long-run, a positive relationship exist
between income and exchange rate with the import volume of the commodity while a
negative relationship exists between own price and the import volume of the commodity.
The above results also suggest that the United States imports of tomato and M exicos
imports of maize, sorghum and milk powder are significantly influenced by the bilateral
Mexico-United States exchange rate.
It is interesting to note that exchange rate volatility (VG) had a negative
relationship with import volume in fifteen of the sixteen models estimated. The above
results imply that volatility of exchange rate have an adverse effect on the Mexico-United
States trade flows of tomatoes, maize, sorghum and milk powder in the long-run.
210
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
Table 6.5 Summary of Long-run Cointegration Results for Four Commodities using
Different Exchange Rate Volatility Measures
Models for Tomatoes
Variable
Own Price
Related Price
US Income
Exchange Rate
Volatility
W age Rate
Variable
Own Price
Related Price
M X Income
Exchange Rate
Volatility
Inventory
Variable
Own Price
Related Price
M X Income
Exchange Rate
Volatility
Inventory
Variable
Own Price
Related Price
M X Income
Exchange Rate
Volatility
M ilk Shortfall
Expected
Sign
(-)
(+)
(+)
(+)
(-)
(+)
Expected
Sign
(-)
(+)
(+)
(+)
(-)
(+)
Expected
Sign
(-)
(+)
(+)
(+)
(-)
(+)
VI
V2
-0.687
-0.532
+2.441
-2.068
+1.008
+0.596
+1.420
+1.810
-0.360
-0.480
+1.420
+0.380
Models for M aize
VI
V2
-0.772
-0.995
0.958
1.099
0.423
0.339
0.405
0.365
-1.082
-1.061
0.461
0.387
Models for Sorghum
VI
V2
-0.526
-0.590
0.342
0.132
0.157
0.129
0.142
0.138
-0.543
-0.541
0.738
0.867
M odels for M ilk Powder
VG
VN
-0.285
+0.844
+0.246
+0.589
-0.454
+0.503
-1.010
-1.878
+0.220
+0.602
-0.529
+0.669
VG
VN
-0.803
1.267
0.582
0.631
-0.421
0.420
-1.212
1.188
0.320
0.481
0.738
0.852
VG
VN
-0.638
0.241
0.769
0.168
-0.416
0.516
-0.090
0.141
0.116
0.103
-0.251
-0.528
Expected
Sign
VI
V2
VG
VN
(-)
(+)
(+)
(+)
(-)
(+)
-1.248
0.924
0.328
0.252
-0.343
0.144
-0.391
0.428
1.245
0.455
-0.799
-0.205
-1.007
1.433
0.785
0.210
-0.347
0.777
-1.095
1.372
0.405
0.208
-0.277
0.443
211
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
the following hypotheses are tested: (a) changes in exchange rate and its variability have
insignificant effect on trade flows (H0: fier = 0,
H 0:
combined effect of exchange rate and its volatility on trade flows is insignificant (H0: Per
= Pv = 0 ) in the long-run.
Finally, the appropriateness of using exchange rate as a separate variable in the
trade model is tested as H0: p = pPrice. The above hypotheses are implemented using the
general model /3 = Hep2 which reflects relevant linear restrictions imposed on the
cointegrating vectors. The matrix H is a pxs matrix with r < s < p and s equal to the
number of variables minus the number of restrictions.
LR = T"ZLn1=1
U -i )
(1-4)
, where A y and
are
the eigenvalues from the restricted and unrestricted models respectively. The test statistic is compared to
2
jjfo 9 5 with (r (p-s)) degrees of freedom, where r is the number o f cointegrating vectors in the original
model, p is the number of variables in the model and V is the total number o f variables minus the number
o f restrictions. For example for model for tomato using V I, the number o f degrees of freedom is (2(7 - 6 ))
=
2.
212
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
1
0
0
0
0
0
0
0
0
1
0
0
0
0
0
0
0
1
0
0
0
0
0
0
0
1
0
0
0 0
0 0
0 0
0 0
0 0
1 0
0 1
The null hypothesis is rejected for all sixteen commodity-specific models at the
five percent level of significance. Similarly, the null hypothesis related to the effects of
the volatility of the exchange rate on the Mexico-United States agri-food trade flows in
the long-run is rejected for all the sixteen commodity-specific models at the five percent
level (Table 6 .6 ). Thus, both exchange rate and the volatility o f the exchange rate have
significant influence on the trade flows of the selected commodities between Mexico and
The United States in the long-run.
Critical
Value*
Models
H 0. Per = 0 ,
LR Statistic
pr = o
Per
Pv
Models
LR Statistic
Critical
Value*
Per
Pv
Tomato V I
5.99
16.86
17.25
Sorghum V I
3.84
6.914
19.966
Tomato V2
5.99
14.34
16.72
Sorghum V2
3.84
8.121
19.917
Tomato VG
3.84
18.60
13.44
Sorghum VG
3.84
7.057
17.277
Tomato VN
5.99
13.93
16.28
Sorghum VN
3.84
4.938
13.782
M aize V I
3.84
9.134
13.175
M ilk Powder V I
3.84
5.821
6.671
M aize V2
3.84
9.558
9.755
M ilk Powder V2
3.84
8.067
13.782
M aize VG
3.84
8.041
27.439
M ilk Powder VG
3.84
9.641
18.093
M aize VN
3.84
7.910
11.023
M ilk Powder VN
3.84
6.945
14.981
213
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
1 0
0 0
1 0
0
0
0
0
1 0 0
0 1 0
The null hypothesis is rejected for all sixteen commodity-specific models at the five
percent level of significance (Table 6.7). The results imply that the combined effect of
exchange rate and its volatility on the Mexico-US trade flows of the selected
commodities in the long-run is also statistically significant. Thus, if either variable is
ignored in an empirical analysis, bias due to mis-specification could result.
214
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
Table 6.7 Long-Run Test Results on the joint Significance of the Exchange Rate and
Volatility of the Exchange Rate on Mexico-US Trade Flows
H 0: Per = Pv 0
Model
Critical Value*
LR Statistic
Critical Value*
LR Statistic
Tomato VI
Tomato V2
9.49
9.49
18.159
21.045
Sorghum VI
Sorghum V2
5.99
5.99
20.847
20.075
Tomato VG
5.99
22.033
Sorghum VG
5.99
18.854
Tomato VN
9.49
19.303
Sorghum VN
5.99
16.848
Maize VI
Maize V2
5.99
5.99
13.208
10.155
Milk Powder VI
Milk Powder V2
5.99
5.99
18.347
12.872
Maize VG
5.99
14.222
Milk Powder VG
5.99
18.219
5.99
14.494
Milk Powder VN
5.99
18.032
Maize VN
Model
6.5.3 Test on the Equality of the Coefficients of Foreign Prices and Exchange Rate
Despite the recent growth in the trade literature focusing on the effects of exchange rate
on trade, it does not provide conclusive evidence on the appropriate way to incorporate
exchange rate in a trade model. Traditionally, exchange rate has been incorporated into
trade models by expressing all prices in common currency units or using a composite
relative price variable (e.g. Carone, 1996). This implies that both foreign prices and the
exchange rate have equal coefficients. Some analysts argued that this is too rectrictive
and used the exchange rate and foreign prices separately into the trade model.
This study incorporated exchange rate directly into the trade equations to estimate
commodity specific import demand functions. To determine if it was appropriate to do
so, an equality restriction on /? coefficients was formulated to verify that the coefficient
of exchange rate is in fact equal to that of the foreign price. The H 3 matrix, as specified
below, is used to test the null hypothesis of H 0 : J32j - fi5j for i , j = 1,2,3
215
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
p; where
1 0 0 0 0 0
0 1 0 0 0 0
0
0
1 0 0 0
0 0 1 0 0
0 0
0 0 0 0 1 0
0 0 0 0 0 1
1 0 0
The null hypothesis of the equality of foreign price and exchange rate coefficients
is rejected for all sixteen commodity-specific models at the five percent level of
significance (Table 6 .8 ). These results imply that when estimating the import demand
function of tomato, maize, sorghum and milk powder, it is appropriate to include
exchange rate directly into the trade model. Therefore, the approach followed in this
study to incorporate exchange rate as a separate variable into the trade model seems to be
appropriate.
Table 6.8 Long-Run Test Results on Equality of Coefficients of Exchange Rate and
Foreign Price on Mexico-US Import Demand
B -r e s tr ic tio n s : H 0: p 2j = Psj f o r i j = 1 ,2 ,3 ,...., p ;
M odel
Tomato V I
Tomato V2
Critical Value*
LR -Statistic
5.99
5.99
8.721
8.620
Sorghum V I
Sorghum V2
M odel
Critical Value*
LR-Statistic
3.84
3.84
9.329
8.821
Tomato VG
3.84
8.742
Sorghum VG
3.84
9.852
Tomato VN
5.99
6.898
Sorghum VN
3 .84
8.202
M aize V I
M aize V2
3.84
3.84
7.944
9.067
M ilk Powder V I
M ilk Powder V2
3.84
3.84
8.733
9.989
M aize VG
3.84
15.657
Milk Powder VG
3 .84
10.520
3.84
9.671
M ilk Powder VN
3.84
8.922
M aize VN
216
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
6.6 Summary
This chapter presented the main results of maximum likelihood cointegration analysis on
commodity models, each using four different volatility measures. Lag-length selection
results reveal that the optimal lag-length for tomato is nine. Also for maize models the
optimal lag-length is nine, with the exception of model V I. For sorghum models the
optimal lag-length is ten. For milk powder models using volatility V I and V2 the optimal
lag-length is ten while for models using VG and VN is ten. The long-run results for all
commodities are consistent in terms of their expected signs. The only exception is the
coefficient of the hog inventory variable (INV) in one of the sorghum models and the
variable MSF in one of the milk powder models.
Although many factors affect agricultural trade, changes in exchange rate play a
significant role. While the magnitude of the impact of exchange rates on Mexico-United
States trade flows varies by commodity, they are, nevertheless, significant in the longrun. With regard to consistency of the results with theory, it was found that models using
VG seem to provide better long-run results than those provided by models using V I or
V2, both in terms of signs and magnitudes of the estimated coefficients.
Comparing results of the four models for each commodity as well as across the
four commodity models, it was found that, for all import demand functions, models using
VG provide more consistent results with theory than models using other volatility
measures. For the four commodity models using VG, all the variables have the a priory
expected signs with the exception of the volatility of the exchange rate variable which is
positive for maize and milk powder models. Similarly, the magnitudes are fairly
consistent for each of the models. The hypothesis of the significance of the effects of
exchange rate and its volatility on the Mexico-US agri-food trade in the long-run was
217
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
formally tested. The results suggest that both exchange rate and the volatility of the
exchange rate have significant influence on the trade flows of the selected commodities
between Mexico and The United States in the long-run.
With regard to the speed of adjustment for most import demand equations, the
results suggest that own price, the trade volume, and exchange rate would adjust faster
than any other variable to restore the equilibrium condition. The next chapter presents the
results of the Vector Error Correction models for each of the selected commodities. These
results represent short-run effects of exchange rate and exchange rate volatility and other
variables on Mexico-United States trade flows of the selected commodities.
218
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
CHAPTER 7
RESULTS OF THE VECTOR ERROR CORRECTION MODEL
7.1 Introduction
The derivation of the long-run relationship is an important step in understanding the
relationship between trade flows and its main determinants. However, a good
understanding of short-run relationships between variables entering the model is also
important for policy analysis. The estimated parameters from the Johansen Maximum
likelihood cointegration analysis are used to formulate the Vector Error Correction Model
(VECM). The estimated parameters from the VECM are interpreted as short-run
elasticities of import demand with respect to each of the independent variables. This
chapter is devoted to accomplish three objectives: (i) the estimation of the VECM; (ii) the
discussion of the short-run results, with particular emphasis on its main explanatory
variables, such as income, prices, exchange rate, and exchange rate volatility; and (iii) a
brief discussion of the policy implications of both short and long-run results. Before
proceeding with the discussion of the VEC results, an overview of how the VEC
mechanism works is presented.
219
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
showed that if two variables are integrated of order one and are cointegrated, then, they
can be modeled as having been generated by an ECM. Thus, the Engle and Granger (EG)
Representation Theorem established that models with valid ECMs entail cointegration
and, conversely, cointegrated series imply an error-correction representation for the
economic model. In the ECM, the long-run equilibrium relationships are used to impose
constraints on the short-run dynamics in economically meaningful ways.
In cointegration approach, relevant cointegration among the variables implies that
there is an underlying error correction process so that any deviation from the long run
equilibrium will be corrected through changes in the variables and the equilibrium will be
restored by the correction of the equilibrium error. Furthermore, the EG representation
theorem for dynamic modeling ensures that the ECM is not subject to "spurious
regression" problem. Finally, in an ECM derived from the EG Representation
Theorem , all variables are stationary and thus, the OLS is an efficient way to estimate
the parameters of the Vector Error Correction Model (VECM).
Having obtained the long-run cointegration relations using the Johansen approach,
it is now possible to reformulate equation 4.13 presented in chapter four and estimate the
VECM with the error-correction term(s) explicitly included in it. Consider the
reformulation of an unrestricted VAR as a VECM discussed in chapter four, so that,
az ,
= 2 r,A Z _, + nz,_, + ( , + / / + ,
j=1
(7.1)
where n = a(3' , with [3 being a matrix of long-run coefficients and a represent the speed
of adjustment to disequilibrium. For the estimation of the VECM, the number of error
220
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
correction terms which were determined in chapter six, are incorporated in equation 7.1
to obtain an empirical version of the model as follows:
AZ, = r,.AZr_. + a ( fr lZ t_l) + <l>Dt +ju + t
j=i
_ _
* -i
(7.2)
and
;'= i
2002).
221
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
The initial estimation of the error correction model for each commodity included
twelve lags of the first difference of each relevant variable, a free-trade dummy variable,
a constant term, eleven seasonal dummy variables, twelve lags of the changes in the
dependent variable and the error correction term (ECIY0 lagged one period. The ECT is
derived from the cointegrating vector representing the long-run relationship among the
variables in each of the estimated models and it is through this term that the long-run
relationship is embedded in the ECM. Each of the ECT is obtained through the scalar
multiplication of the cointegrating vector and the Z, vector.
The main results of the Vector Error Correction Models applied to sixteen
commodity models are presented and discussed in the next section. Each model differs
from the others in terms of lag specification and the inclusion of a different volatility
variable. For all models, the estimated coefficients and their corresponding t-statistic are
9
reported along with the R , adjusted R and a set of standard diagnostic statistics. The
estimated coefficients represent short-run elasticities. The coefficient of the errorcorrection term must be negative and statistically significant to ensure dynamic stability
of the model. It also re-reconfirms cointegration relationships obtained in chapter six
(Harris and Sollis, 2003).
Since all VEC models include the same set of explanatory variables but different
volatility measures and the ECT, at least, part of the differences in expected signs and
magnitudes of the estimated short-run elasticity values can be attributed to the
specification of different volatility measures.
222
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
7.3 Results of the VECM: the US Import Demand for Fresh Tomatoes
To asses the adequacy of each model, the goodness of fit measured by the adjusted R ,
the F-statistic and a set of standard diagnostic tests are used1. For the tomato model, the
empirical results suggest that the statistical fit is quite satisfactory. The R values ranges
from 0.36 for model using VG to 0.46 for the model using V I and all F-statistics are
significant at five percent level. Since the EC model includes a lagged dependant
variable, the conventional Durbin-Watson test can not be used. Instead, one is required to
use the Durbin "h" statistic to test for autocorrelation. While the Jarque-Bera test results
consistently reject the null hypothesis of normality of the residuals at the five percent
significant level for all four models, the null hypothesis could not be rejected at one
percent. The Lagrange Multiplier (LM) statistic indicates that the null hypothesis of no
autocorrelation cannot be rejected at the five percent significant level. Thus, no serial
autocorrelation is present in the VEC models for tomatoes.
The results of the four VEC models for fresh tomato imports into the Unites
States from Mexico are reported in tables 7.1 and 7.2. All own price coefficients exhibit
the expected negative sign and are statistically significant at the five percent level. These
1 One of the key requirements is that the residuals follow a white noise process i.e., residuals have zero
mean, constant variance, and are uncorrelated with explanatory variables. A number o f tests are used to
assess model adequacy such as: (a) the goodness o f fit of the model (R2 and adjusted R2). i.e., what
percentage o f the total variation in the dependent variable is explained by the independent variables); an
F-test is used to test overall significance of the regression, the null hypothesis in this case is Ho: R2 = 0, the
population R2 is zero (all the regression coefficients are zero, excluding the intercept); (b) the DW-h test is
used for detecting first order autocorrelation (which is strictly necessary since all models includes a lagged
dependent variables; (c) the Jarque-Bera test for determining if the residuals are distributed as a normal
distribution. The test is based on the sample skewness and kurtosis of residuals and is a joint test for the
skewness and kurtosis of the disturbances. It is asymptotically distributed as chi-squared with two degrees
of freedom. The null hypothesis is that residuals are normally distributed. For a normal distribution, the
sample skewness should be zero and the value of the kurtosis is three. The Jarque-Bera test determines
whether the sample skewness and kurtosis are unusually different than their expected values, as measured
by a chi-squared statistic; (d) the Hansen test for coefficient instability. It includes a statistic for the
variance and a second one for the joint instability o f the model; and (e) finally, to test for serial
autocorrelation, the LM statistic is employed.
223
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
results are consistent with the results obtained in the long-run analysis and are also
consistent with the prediction of the theoretical model. The own price elasticity ranges
from -0.092 for the tomato model using VN to -0.15 using volatility V2 and suggest that
an increase in the price of imports leads to a decrease in the quantity of fresh tomato
imported by the Unites States from Mexico.
The price of the related commodity (price of tomatoes produced in the Unites
States) also exhibits the expected positive sign in all four models and it is significant at
the five percent level. The above results suggest than in the short-run, imports of tomato
from Canada into the Unites States are a substitute for tomato imported from Mexico.
The cross price elasticity of the US import demand ranges from 0.195 for tomato model
using volatility VG to 0.536 using volatility V2.
The coefficients of the income variable are all positive and significant suggesting
a direct relationship between income and quantity of fresh tomato imported into the
Unites States from Mexico. The magnitude of the elasticity of income with respect to
import demand ranges from 0.10 for tomato model using VN to 0.206 for tomato model
using V2. All short-run elasticities are smaller than the long-run elasticities, and hence,
are consistent with the Le Chatelier principle.
With regard to the exchange rate variable, all four models yield coefficients with
positive signs as expected. These results support the hypothesis that a depreciation of the
Mexican currency relative to the Unites States dollar makes imports into the Unites States
less expensive resulting in an increase of fresh tomato imports from Mexico. The
elasticity of the exchange rate for the four models ranges from 0.50 for tomato model
using VN to 0.80 for tomato model using V I.
224
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
Table 7.1 Error Correction Model Results for the US Tomato Import Demand
Volatility V I
Variables
Estimated
Coefficient
T-Ratio
5.921
Volatility V2
Variables
Estimated
Coefficient
T-Ratio
QT t-12
USY mi
ERm
0.178
4.051
0.206
4.515
0.706
0.029
-0.152
2.475
3.802
-3.406
0.536
4.487
0.127
3.399
-3.532
QT M2
0.331
USY mi
0.156
3.939
ERm
0.802
-0.064
2.921
-1.983
V 2
-0.117
-3.428
0.457
3.961
PTM2
PCA mi
V I ,.3
P T ,.12
P C A t.
W l t.8
ECT m
0.295
3.275
-0.048
-3.535
W l,.6
ECT m
-0.072
NAFTA
0.495
2.310
NAFTA
0.475
2.125
M 10
M il
-0.007
-1.078
1.078
M6
M il
-0.076
-0.985
0.116
2.804
CONSTANT
0.000
0.306
CONSTANT
0.078
0.348
2.731
R2
R2 Adjusted
0.459
0.426
R2
R2 Adjusted
0.431
0.399
F-Value
13.68
F-Value
12.17
DW -H Statistic
-0.116
DW-H Statistic
1.225
Skewness
0.271
0.000
Skewness
0.161
0.000
Kurtosis
1.259
3.000
Kurtosis
1.464
3.000
J-B Normal
8.946
9.210**
J-B Normal
6.214
9.210**
Instability Test:
Instability Test
Variance
0.467
0.748
Variance
0.528
0.748
Joint
3.849
3.690
Joint
3.265
3.150
20.780
35.172*
29.746
35.172*
L.M Statistic
L.M Statistic
With regard to the volatility of the exchange rate variable, three out of the four
models yield coefficients with negative signs as expected. Thus, in the short-run volatility
has a negative influence on the Unites States imports of tomato from Mexico. Only the
tomato model with volatility V2 indicates a positive relationship between exchange rate
volatility and the Unites States imports of tomato from Mexico. In all cases, the
coefficients of the volatility variable are significant at the five percent level. The above
results lend empirical support to the comparative statics results presented in chapter three
225
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
and suggest that risk-averse importers are discouraged by higher volatility episodes of the
Mexico-Unites States exchange rate.
Table 7.2 Error Correction Model results for the US Tomato Import Demand
Volatility VG
Variables
Estimated
Coefficient
QT t-12
USY,.10
0.198
0.110
ER,.i
V G t2
0.663
-0.043
-0.099
0.195
PTt-io
PCAt.
W l,.6
ECT,.!
NAFTA
M3
M il
CONSTANT
Volatility VN
Variables
T-Ratio
3.274
3.750
3.269
-2.671
-3.360
Estimated
Coefficient
QT M2
USY ,.9
0.169
ER,.!
0.499
-0.108
-0.092
VN,.2
0.098
T-Ratio
3.048
9.370
2.350
-2.123
-2.639
3.438
1.508
P T ,. i
PCA ,_!!
0.130
-0.030
0.427
2.936
-3.453
2.081
W l,.6
E C T ,.,
NAFTA
0.416
0.343
-0.041
0.523
0.011
0.070
0.809
0.809
Ml
M2
0.011
0.021
1.103
1.082
0.309
2.365
CONSTANT
0.375
3.152
4.825
-4.068
2.303
R2
R2 Adjusted
0.364
0.332
R2
R2 Adjusted
0.396
0.359
F-Value
9.198
F-Value
10.577
DW -H Statistic
1.677
DW -H Statistic
1.570
Skewness
0.414
0.000
Skewness
0.120
0.000
Kurtosis
2.221
3.000
Kurtosis
2.359
3.000
J-B Normal
8.209
9.210**
J-B Normal
8.875
9.210**
Instability Test
Instability Test:
Variance
0.704
0.748
Variance
0.732
0.748
Joint
2.777
2.890
Joint
3.280
3.150
26.816
35.172*
29.889
35.172*
L.M Statistic
L.M Statistic
The values of the coefficient of the volatility of the exchange rate range from 0.03
for the model using V2 to 0.11 for the model using YN. To the best of my knowledge, no
information exist in the published literature on the relationship between the exchange rate
226
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
volatility and trade flows in fresh vegetable from Mexico to the Unite States. So, the
results of this study can not be compared to those of a previous study.
The coefficients of the lagged dependent variable are all significant at the five
percent level and the estimated values vary from 0.17 to 0.33. These results illustrate the
importance of previous imports of fresh tomatoes on current import volume and suggest a
stable trade relationship in this commodity (i.e., the Unites States importers of fresh
tomatoes take the volumes they imported in past periods into account when deciding how
much to import in the current period).
Regarding the ratio of the Unites States-Mexico farm wage rate (W l), defined as
farm wage rate in the Unites States over farm wage rate in Mexico, the VECM yields
results in line with those obtained from the long-run analysis as all coefficients are
positive and statistically significant. These results are also consistent with the
comparative static results derived from the theoretical model. These results imply that an
increase in farm wage rate in the United States relative to that in Mexico will lead to an
increase in tomato imports from Mexico. The magnitude of the estimated coefficient
ranges from 0.13 using volatility VG to 0.34 for model using volatility VN. These results
are in line with those of Malaga et al (2001) who showed that increases in M exicos real
wage rates have on Mexicos ability to export fresh vegetables to the United States.
With regard to the Error Correction Term (ECT), the results show that all four
models yield negative and significant coefficients. These results corroborate the findings
of chapter six and reconfirm the presence of cointegrating relationships in the Unites
States import demand function for fresh tomato for Mexico. The negative sign of this
coefficient indicates that the direction of correction is towards the long-run equilibrium
227
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
while the size indicates the speed of adjustment towards the long-run equilibrium (Carone
1996; Harris and Solis 2003). The small values of the ECT coefficient obtained in all four
models suggest a slow adjustment towards long-run equilibrium.
The coefficient of the free trade dummy variable is positive and significant in all
four models, suggesting that NAFTA had a positive effect on the Unites States imports of
tomato from Mexico. Previous researchers argued that NAFTA has a small positive effect
on the Unites States tomato imports from Mexico and in the short-run, the peso
devaluation rather than NAFTA, explains most of the increase in the Unites States
imports of tomato and fresh vegetables from Mexico. Finally, none of the seasonal
dummy variables is statistically significant. In the spirit of the general to specific
approach each model was run using 12-lags for each variable and the 11 monthly
dummies and then, the variables with insignificant parameters were dropped from the
analysis. In the end, none of the monthly dummy variables was found to be significant in
the tomato import demand functions.
228
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
The own price elasticity of import demand is inelastic in all four models while the
cross price elasticity is positive but shows greater variability across models. The own
price elasticity of import demand reported in the existing literature varies from -0.31 to 0.62 (Salcedo-Baca, 1990; Gutierrez, 1983; M alaga et al, 2001), which appears to be
much higher than those obtained in this study.
The income elasticity of import demand across all models is less than one. Once
again the short-run income elasticities obtained in this study are smaller than the values
of income elasticities reported in the literature, these values range from 0.23 to 1.47 for
the US imports from Mexico (Mitterhammer, 1978; Salcedo-Baca, 1990; and Malaga et
al., 2001).
The effect of changes in the ratio of wages is found to be positive in all models.
Results in the same direction were found by Love and Lucier (1996); and Schwentesius
and Gomez Cruz (1996). They reported that the United States imports of fresh vegetables
were sensitive to labour cost differentials between Mexico and the United States.
The error correction term is negative and significant in all four models for
tomatoes. These results reconfirm the presence of cointegrating vectors found in each
model discussed in chapter six. However, the small value of the estimated ECT
coefficient implies that the adjustment towards the long-run equilibrium is rather slow.
Finally, NAFTA had a significant positive impact on tomato trade but surprisingly none
of the monthly dummy variables appeared to be significant. In terms of relative size of
elasticities, the short run elasticities are, in general, smaller than their corresponding long
run elasticities, suggesting a relatively inelastic short run import demand for Mexican
fresh tomatoes in the United States.
229
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
7.4 Results of the VECM: the Mexicos Imports Demand for Maize
All short-run results of Mexican import demand for maize from the United States along
with a set of diagnostic statistics are reported in tables 7.3 and 7.4. The R2 ranges from
0.30 in model using V I to 0.32 in all other models for tomatoes. All F statistics are
significant indicating that the overall goodness of fit is satisfactory. The DW-h statistic
rejects the presence of autocorrelation. W ith regard to the normality test, while the null
hypothesis of normality of the disturbances is rejected at the five percent level, the
hypothesis cannot be rejected at the one percent level of significance. The Lagrange
Multiplier (LM) statistic indicates that the null hypothesis of no autocorrelation cannot be
rejected at the five percent significant level. Thus, the estimated import demand function
for maize seems satisfactory.
The coefficient of the own price of maize exhibits an expected negative sign in all
four models and are all significant at the five percent level. The value of the own price
elasticity ranges from 0.13 to 0.15. The coefficient of the price of the related commodity
(maize produced in Mexico) exhibits an expected positive sign and is significant at five
percent level in all four models. The estimated value of the cross price elasticity ranges
from 0.14 for maize model using volatility VN to 0.18 using volatility V I. The above
results are consistent with the results obtained from the long-run analysis. Maize
produced in Mexico is a substitute for maize imports from the Unites States. Both shortrun own price and cross price elasticity values are smaller than their long-run values. The
small value of the price elasticity reflects the fact that few possibilities of substitution
exist for maize imports from the United States. Other sources of imports of maize into
Mexico are nonviable as long as Mexico enjoys trade preferences because of the NAFTA
230
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
agreement and lower transportation costs because of the proximity to the United States.
The coefficients of the income variable are all positive and significant suggesting a direct
relationship between income and quantity of maize imported into Mexico from the Unites
States. The estimated values of income elasticity range from 0.08 to 0.23 for all four
maize models.
Table 7.3 Error Correction Model Results: M exicos Import Demand for Maize
V o la tility V I
V a ria b les
E stim a te d
C o efficien t
T -R a tio
V o la tility V 2
V a ria b les
E stim a ted
C o efficie n t
T -R a tio
1.774
0.087
0.086
1.986
3.222
QCt-6
MXY ,.9
0.122
0.234
0.221
- 0.222
0.142
2.222
0.090
- 2.809
1.902
E R t-3
V 2m
DINV tl2
-0.147
0.181
- 2.269
3.081
PCOM
PM X m
CONSTANT
0.909
-4.970
2.725
2.054
2.138
4.327
ECT n
NAFTA
M3
M9
-0.018
0.245
0.493
0.331
QCW
M X Y ,.,
E R t3
V I t-t
DINV n2
PCOM
PM X n
ECT n
NAFTA
-0.155
-0.134
-0.156
-0.170
-0.014
M3
M9
0.379
0.379
0.319
CONSTANT
0.859
8.699
2.499
-3.261
2.210
-2.280
2.621
-5.029
2.613
1.944
2.113
4.234
R2
R2 Adjusted
0.307
0.276
R2
R2 Adjusted
0.327
0.290
F-Value
5.261
F-Value
5.785
DW -H Statistic
-1.292
DW -H Statistic
- 1.269
Skewness
-0.645
0.000
Skewness
- 0.741
0.000
Kurtosis
0.576
3.000
Kurtosis
1.543
3.000
J-B Normal
7.870
9.210**
J-B Normal
8.314
9.210**
Instability Test:
Instability Test
Variance
0.232
0.353
Variance
0.214
0.353
Joint
3.785
2.960
Joint
3.854
3.690
27.17
35.172*
31.106
35.172*
L.M Statistic
L.M Statistic
231
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
232
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
as freight future contracts and commodity future contracts among others. The results
suggest that they have not been effective in Mexico-Unites States maize trade during the
study period.
Table 7.4 Error Correction Model Results: M exicos Import Demand for Maize
Volatility VG
Variables
Estimated
Coefficient
T-Ratio
QC,6
0.081
1.702
QC,-6
MXY ,.9
E R t_3
0.125
0.244
8.376
2.646
MXY ,.9
0.133
8.764
2.895
V G t.i
D IN V tl2
-0.057
0.144
-0.391
-1.980
DINV.12
0.137
1.957
PCOm
PMX n
-0.138
0.153
-3.326
1.993
-2.202
ER t-3
VN,.!
0.259
-0.152
-2.248
ECTu
NAFTA
M3
-0.018
PCOm
PM X,.!
ECT,.i
2.628
-5.002
2.556
Volatility VN
Variables
Estimated
Coefficient
0.070
2.464
-5.338
NAFTA
-0.019
1.808
M3
0.438
1.909
M9
CONSTANT
0.308
0.956
1.999
4.491
M9
CONSTANT
0.883
R2
R2 Adjusted
0.319
0.284
R2
R2 Adjusted
0.322
0.294
F-Value
5.010
F-Value
4.795
DW -H Statistic
- 1.932
4.258
2.092
0.144
1.161
0.427
0.308
1.835
1.988
T-Ratio
2.825
DW -H Statistic
- 1.482
Skewness
- 0.732
0.000
Skewness
-0.813
0.000
Kurtosis
1.815
3.000
Kurtosis
1.135
3.000
J-B Normal
7.891
9.210**
J-B Normal
8.861
9.210**
Instability Test:
Instability Test
Variance
0.233
0.353
Variance
0.302
0.353
Joint
4.082
3.690
Joint
3.631
3.510
29.059
35.172*
29.820
35.172*
L.M Statistic
L.M Statistic
The results suggest that NAFTA is having a significant effect on the Unites
States-Mexico maize trade. This support the findings of Zahniser and Coyle (2004). Since
the agreement took effect in 1994, M exicos imports from the Unites States have
233
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
increased by 240 percent compared to their average annual level in the pre-NAFTA
period. In regards to seasonality, two seasonal dummy variables, M3 and M9, are found
to be statistically significant. It suggests that imports of maize increase during the winter
season and before the harvest season since the main region devoted to hog and cattle
production in Mexico (the state of Sonora located in the border with the Unites States) is
affected by winter weather conditions.
The coefficient of lagged dependent variable is positive and significant in two of
the four cases. The values range from 0.08 to 0.12. Even though the responsiveness
appears to be small, these results illustrate the influence of previous trade on current
volumes of maize imports into Mexico. In other words, the Mexican maize importers take
into account the volumes they imported in the past to make their decisions on how much
to import in the current period. The lagged dependent variable becomes significant at six
lags for all four maize models. Thus, it implies that it takes about six months for
importers of maize to adjust to changing market conditions.
With regard to the inventory o f hogs (INV), while the VECM yields expected
positive sign for all models, the coefficients are statistically significant only in models
using V2 and VG. Previous research in Mexico does not suggest a strong link between
imports of maize and the domestic feed industry. Results from a regression that
considered the period of 1990 to 998 did not produced a strong link between imports of
whole maize and animal feed production. However, results indicated a strong link
between imports of broken maize and animal feed produced commercially, and between
imports of maize gluten and feed produced by commercial operations (Avalos-Sartorio,
1998). The free trade dummy variables are positive and significant in all four models
234
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
suggesting that NAFTA has a significant effect on the M exicos imports of maize from
the Unites States.
The coefficient of the Error Correction Term (ECT) is negative and significant in
all four models. These results corroborate the results of the long-run analysis and
reconfirm the presence of cointegrating vectors in the system. The small value of the
ECT coefficient in all four models suggests a slow adjustment towards the long-run
equilibrium.
235
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
7.5 Results of the VECM: the Mexicos Import Demand for Sorghum
'y
For the sorghum models, the R ranges from 0.36 for model using V2 to 0.38 for model
using V I. All the F-values are statistically significant at the five percent level. The DW-h
statistic indicates that there is no autocorrelation in the estimated models. With regard to
the normality test, the null hypothesis of normality of the disturbances is rejected at the
five percent significant level, but it can not be rejected at the one percent level for all
sorghum models. The Lagrange Multiplier (LM) statistic indicates that no autocorrelation
is present in the models for sorghum.
The results of the four ECM models for M exicos sorghum imports from the
Unites States are reported in tables 7.5 and 7.6. The coefficient of the own price of
sorghum imports exhibits an expected negative sign and is significant at the five percent
level in all four models. These results are consistent with the long-run results presented in
chapter six. For sorghum, the long-run price elasticities are, in general, larger in value
than those in the short-run. The estimated own price elasticity from VEC models range
from 0.10 to 0.15.
The price of the related commodity (sorghum produced in Mexico) exhibits an
expected positive sign and is statistically significant in all sorghum models. The cross
price elasticity of import demand is consistent across four models and range from 0.28
for model using volatility VG to 0.30 using volatility V I. These results are consistent
with the results obtained from the long-run analysis.
The coefficient of the income variable is positive and significant suggesting a
direct relationship between income growth and the quantity of sorghum imported into
Mexico from the Unites States. The estimated value of income elasticity implies that if
236
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
Table 7.5 Error Correction Model Results: M exicos Import Demand for Sorghum
Volatility VI
Variables
QS t-12
MXY t_2
E R .6
VI
,.3
D I N V .3
Estimated
Coefficient
0.130
1.798
Volatility V2
Variables
Q S t-1 2
Estimated
Coefficient
0.134
T-Ratio
1.778
0.065
0.144
5.547
1.984
MXY ,.2
0.064
ER,-6
0.153
5.377
2.032
-0.181
0.063
-3.553
2.192
-4.084
V 2 ,.3
-0.091
-2.664
0.069
-0.107
0.292
2.165
-2.682
-0.126
0.306
PS ,.7
PSD ,.io
ECT,.!
NAFTA
M4
M7
M il
T-Ratio
5.698
d i n v
,3
PS ,.7
PSD n o
ECT,.,
NAFTA
M4
M7
MU
CONSTANT
4.255
-4.325
-0.011
0.137
-0.288
-0.168
-4.403
0.4228
CONSTANT
-0.259
0.450
-1.601
3.387
R2
R2 Adjusted
0.385
0.351
R2
R2 Adjusted
0.367
0.332
F-Value
8.496
F-Value
7.912
DW -H Statistic
1.233
DW -H Statistic
- 1.045
Skewness
-0.298
0.000
Skewness
-0.322
0.000
Kurtosis
1.904
3.000
Kurtosis
1.834
3.000
J-B Normal
8.997
9.210**
J-B Normal
9.098
9.210**
-1.987
-1.784
Instability Test:
-0.011
0.101
-0.300
-0.196
-0.266
0.442
0.306
-2.039
-2.003
-1.830
3.267
Instability Test
Variance
1.676
0.748
Variance
1.993
0.748
Joint
4.474
3.690
Joint
4.175
4.070
28.740
35.172*
26.900
35.172*
L.M Statistic
L.M Statistic
Regarding the exchange rate variable, all four models yield the expected positive
sign of the coefficients and they are significant at five percent level. This relationship
between exchange rate and import demand for sorghum is consistent with the long-run
237
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
results obtained in chapter six. The estimated value ranges from 0.12 for model using YG
to 0.15 for sorghum model using V2 and VN.
All four sorghum models yield an expected negative sign for the volatility of the
exchange rate variable, suggesting that in the short-run volatility has a significant
negative influence on M exicos imports of sorghum from the Unites States. Thus, riskaverse Mexican importers are discouraged by high volatility episodes of the MexicoUnites States exchange rate. As a consequence, exchange rate uncertainty decreases the
volumes of sorghum imports into Mexico. The estimated values, however, are small and
range from 0.091 to 0.243.
The coefficients of the lagged dependent variable are all positive and significant
at five percent level. The estimated values range from 0.12 to 0.13. Notwithstanding
small values of the estimates, these results illustrate that the Mexican sorghum importers
take into account the volumes they imported in past periods to make their decisions on
how much to import in the current period. The lagged dependent variable becomes
significant at twelve lags which implies that importers of sorghum takes about twelve
months to adjust to changing market conditions.
Regarding the hog inventory variable (INV), the VECM yields coefficients with
the expected sign and all are statistically significant. The significant coefficient is
reflective of the increasing domestic demand for sorghum in Mexico. Avalos-Sartorio
(1998) found a statistically significant relationship between imports of sorghum and the
domestic production of animal feed. At the present time Mexico is the major sorghum
importer in the world. Mexican imports accounted for about 30 percent of total world
imports of sorghum between 1992 and 2000. However, the countrys domestic
238
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
production of sorghum has been augmented and thus, its ability to supply the domestic
market is improving.
Table 7.6 Error Correction Model Results: M exicos Import Demand of Sorghum
Volatility VG
Variables
0.123
0.061
Q t-1 2
MXY ,.2
0.119
ER t-6
VG.,3
i n v
Estimated
Coefficient
-0.103
0.068
.,3
-0.154
T-Ratio
T-Ratio
0.128
1.772
5.171
M XY ,.2
0.062
2.189
-5.180
2.221
ER,.6
VN,.3
5.289
2.058
ECT,.!
0.288
-0.012
NAFTA
M4
0.039
-0.297
0.119
-1.977
M7
M il
-0.156
-0.290
-1.629
-1.856
CONSTANT
0.435
3.171
R2
R2 Adjusted
P S D t.jo
Estimated
Coefficient
Q t-1 2
1.748
-2.715
4.278
-4.627
P S , 7
Volatility VN
Variables
i n v
0.150
,.3
-0.243
0.102
-2.664
0.111
-2.988
4.342
P S ,.7
P SD no
E C T ,.,
NAFTA
0.291
-0.012
0.053
3.227
-4.379
0.157
-1.978
-1.864
M4
M7
-0.265
M il
CONSTANT
-0.292
0.420
0.365
0.338
R2
R2 Adjusted
0.362
0.336
F-Value
7.809
F-Value
7.740
DW -H Statistic
-1.790
DW -H Statistic
-1.126
Skewness
-0.451
0.000
Skewness
-0.410
0.000
Kurtosis
1.820
3.000
Kurtosis
1.790
3.000
J-B Normal
8.713
9.210**
J-B Normal
29.897
9.210**
-0.176
-1.866
3.046
Instability Test
Instability Test:
Variance
1.826
0.748
Variance
1.897
0.748
Joint
4.441
3.690
Joint
4.052
3.540
23.750
35.172*
29.917
35.172*
L.M Statistic
L.M Statistic
The Error Correction Terms (ECT) are negative and significant in all models.
Similar to tomato and maize models, the small value of the ECT coefficients in sorghum
models suggests a slow adjustment towards the long-run equilibrium.
239
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
The free trade dummy variables are positive but not significant in all four
sorghum models, suggesting that the free trade has not been a key driver of M exicos
imports of sorghum from the Unites States. It may be due to the fact that even before the
NAFTA took place, Mexico was already importing significant quantities of sorghum
required for domestic consumption from the United States and the trade was not hindered
by any policy restrictions. In regards to seasonality, sorghum models show that some
seasonal dummy variables are statistically significant in all four models (M4 for all
models and M7 for model using V2). This seasonal pattern may reflect the fact that
importers build stocks for the winter season in the northern states of Mexico where most
of the livestock production is concentrated.
7.6 Results of the VECM: the M exicos Import Demand for Milk Powder
The VECM for Milk powder in this study yield R values ranging from 0.31 to 0.33. All
F-statistics are significant and the DW-h statistic indicates that there is no autocorrelation
240
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
in any of the milk powder models. The null hypothesis of normality of the residuals is
rejected at the five percent significance level, but, it cannot be rejected at one percent
level for all four milk powder models. So, the estimated import demand functions for
milk powder seem satisfactory.
The results of the four VEC models for milk powder imports into Mexico from
the Unites States are reported in tables 7.7 and 7.8. The coefficient of the own price of
milk powder imports into Mexico exhibits a negative sign as expected in all four models
and they are significant. The own price elasticity ranges from 0.46 for milk powder
model using V I to 0.61 using volatility VN.
The coefficient of the related price (price of milk powder imported from Canada
into Mexico) exhibits an expected positive sign in all four models and all the coefficients
are significant at five percent level. The cross price elasticities range from 0.66 for milk
powder model using volatility V I to 0.88 using volatility VG. Thus, milk powder
imported from Canada into Mexico is considered a substitute for milk powder imported
from the Unites States.
The coefficients of the income variable are positive and significant in all models
suggesting a direct relationship between personal disposable income in Mexico and
quantity of milk powder imported from the Unites States. The income elasticity of import
demand range from 0.06 for model using V2 to 0.28 for model using VN. Similar to the
previous commodity models, all short-run elasticities are smaller than the corresponding
long-run elasticities.
241
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
Table 7.7 Error Correction Model Results: Milk Powder Import Demand
Volatility V I
Variables
Estimated
Coefficient
Estimated
Coefficient
T-Ratio
Q M t i2
0 .2 0 0
2.694
M XY ,.3
0.063
2.326
0.484
3.372
-0.133
-2.343
0.443
-0.530
2.659
-2.234
Volatility V2
Variables
T-Ratio
QM t-12
0.181
MXY t3
E R t.7
0 .1 0 1
0.789
4.682
VI . 5
M SFM
PM Km
-0.253
0.430
-1.905
2.479
E R t-7
V 2
M SF t.4
-0.458
-1.999
p m k ,.4
PM C ,.2
ECT t.i
0.661
2.180
0.752
2.141
-0.109
0.744
-4.687
0.983
PM C t 2
ECT t.i
-0.077
-3.923
NAFTA
1.431
1.721
3.271
NAFTA
M3
CONSTANT
2.300
3.421
0.354
1.250
CONSTANT
1.288
2.773
M3
2.508
0.412
R2
0.318
R2
0.314
R 2 Adjusted
0.273
R 2 Adjusted
0.283
F-Value
5.918
F-Value
4.957
DW -H Statistic
- 1.458
DW -H Statistic
-1.231
1.523
Skewness
0.095
0.000
Skewness
0.036
0.000
Kurtosis
1.502
3.000
Kurtosis
1.141
3.000
J-B Normal
8.335
9.210**
J-B Normal
6.612
9.210**
Instability Test
Instability Test:
Variance
Joint
L.M Statistic
0.7147
0.470
Variance
0.689
0.353
2.431
2.890
Joint
2.853
2.890
20.373
35.172*
17.567
35.172*
L.M Statistic
Regarding the exchange rate variable, all models for milk powder yield the
expected positive sign of the coefficients consistent with the results of the theoretical
model for a risk-averse importer. Thus, a depreciation of the Mexican currency relative to
the Unites States dollar makes imports into Mexico more expensive resulting in a
decrease of milk powder imports into Mexico from the Unites States. The estimated
coefficients range from 0.48 for model using V2 to 0.78 for model using V I. The above
242
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
short-run results support the empirical observations that imports into Mexico from the
Unites States decreased during the Mexican peso devaluation periods.
T-Ratio
Q M .12
MXY ,.3
0.171
0.097
2.046
3.220
QM t-12
M XY t.3
E R ,.7
0.766
2.783
E R ..7
V G ts
M SF t.i
P M K .2
-0.316
-3.368
VN
0.450
2.624
-0.548
0.876
-2.565
2.192
M S F ,.4
p m k ,.2
ECT t-i
NAFTA
CONSTANT
-0.091
1.163
-3.990
1.471
3.144
R2
0.332
R2
0.320
R 2 Adjusted
0.291
R 2 Adjusted
0.290
F-Value
4.910
F-Value
5.737
DW -H Statistic
-0.690
DW -H Statistic
-0.799
Skewness
0.073
0.000
Skewness
0.226
0.000
Kurtosis
1.007
3.000
Kurtosis
1.095
3.000
J-B Normal
5.443
9.210**
J-B Normal
7.766
9.210**
0.456
0.470
1.815
2.890
27.104
35.172*
PM C ,.2
2.008
Instability Test:
PM C ,2
e c t ,.,
NAFTA
CONSTANT
T-Ratio
0.229
0.283
3.737
4.706
0.730
-0.903
2.719
-3.103
0.467
2.859
-2.654
-0.608
0.848
-0.092
0.781
1.288
2.268
-3.948
1.024
2.639
Instability Test
Variance
0.689
0.470
Joint
2.693
2.890
19.950
35.172*
L.M Statistic
Volatility VN
Variables
Estimated
Coefficient
Estimated
Coefficient
Variance
Joint
L.M Statistic
The coefficient of the volatility of the exchange rate variable yields an expected
negative sign for all milk powder models. In all cases, the coefficients of the volatility
variable are significant at five percent level. These results are consistent with the
hypothesis that a risk-averse importer would reduce the volume of imports if faced with
243
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
uncertainty in relative currency values. The estimated coefficients range from 0.13 for
model with V2 to 0.90 for model with VN.
The coefficient of the lagged dependent variable is positive and significant in all
four models. The estimated values range from 0.17 to 0.22. The lagged dependent
variable becomes significant at twelve lags. So, it is reflective of the time it takes for
importers of milk powder to adjust to changing market conditions.
It is worthwhile to note that the milk short-fall (MSF) variable has a positive and
significant coefficient in all milk powder models. The elasticity of the MSF with respect
to import demand ranges from 0.43 to 0.47. In regard to the Error Correction Term
(ECT), results show that all four models yield negative and significant coefficients.
Similar to previous models, the small value of the ECT coefficient in all four models
suggests a slow adjustment towards the long-run equilibrium.
While the free trade dummy variable has a positive coefficient in all models, none
is statistically significant at the five percent level. This result suggests that free trade had
no significant impact on the M exicos milk powder imports from the Unites States. All
seasonal dummy variables yield statistically insignificant coefficients suggesting that
there is no pronounced seasonal pattern in the quantity of milk powder imports from the
Unites States into Mexico.
244
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
volatility. The error correction term is negative and significant in all milk powder models,
reconfirming the presence of cointegrating vectors found in chapter six. In all models, the
small value of the ECT coefficient implies a slow adjustment towards the long-run
equilibrium. A possible explanation for the low adjustment could be the presence of
market interventions in this sector. This hypothesis is plausible for the case of Mexico
since the government maintained price control and other regulation until the 1990s for
milk powder products. While many of the restrictions have been removed, some residual
effects may still persist. The free trade dummy variable appears to be positive and not
significant in all milk powder models confirming that NAFTA did not have a positive
impact on milk powder imports from the Unites States into Mexico. Furthermore, none of
the monthly dummy variables is significant. Finally, the sign of the MSF variable is
positive as expected which reflects the increasing dependency of Mexico for milk powder
from the international market to satisfy growing domestic demand for milk and other
dairy products.
the combined effect of exchange rate and its volatility on trade flows is insignificant (Flo:
Per = Pv
245
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
the trade model is tested using H0: /fer = pPrice- The results of these tests are discussed
below.
7.7.1 Testing the Significance of Individual Coefficients
To perform this, the null hypotheses: H0: fiER = 0 and H0: J3vi = 0 are used where fiER =
estimated coefficient of the exchange rate and f i Vi = estimated coefficient of the volatility
variable and a standard t-test is employed. The null hypothesis is rejected for all sixteen
commodity-specific models at the five percent level of significance. Similarly, the null
hypothesis of the insignificant effect of the volatility of the exchange rate on the MexicoUnited States agri-food trade flows is rejected for fifteen of the sixteen commodityspecific models at the five percent level (Table 7.9). Thus, both exchange rate and the
volatility of the exchange rate have significant influence on the trade flows of the selected
commodities between Mexico and The United States in the short-run.
Models
Per
= 0,
pv= o
t-statistic
t-statistic
Critical
Value*
Per
Pv
Models
Per
Pv
Tomato V I
1.96
2.92
1.983
Sorghum V I
1.96
1.984
3.553
Tomato V2
1.96
2.48
3.80
Sorghum V2
1.96
2.032
2.664
Tomato VG
1.96
3.27
2.67
Sorghum VG
1.96
2.189
5.180
Tomato VN
1.96
2.35
2.12
Sorghum VN
1.96
2.058
2.664
M aize V I
1.96
2.222
2.809
M ilk Powder V I
1.96
4.682
1.905
2.343
M aize V2
1.96
2.499
3.261
M ilk Powder V2
1.96
3.372
M aize VG
1.96
2.646
3.326
M ilk Powder VG
1.96
2.783
3.368
M aize VN
1.96
2.895
1.980
M ilk Powder VN
1.96
2.719
3.103
246
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
sixteen commodity-specific models at the five percent level of significance (Table 7.10).
The result implies that the combined effect of exchange rate and its volatility on the
Mexico-United States trade flows of the selected commodities is also statistically
significant.
Table 7.10 Short-Run Test Results on the joint Significance of the Exchange
Rate and Volatility of the Exchange Rate on Mexico-US Trade Flows
H 0: Per = fiv = 0
Critical Value*
F-Statistic
Sorghum V I
3.00
7.652
6.872
Sorghum V2
3.00
5.161
3 .00
6.965
Sorghum VG
3.00
47.436
Tomato VN
3.00
8.717
Sorghum VN
3.00
6.307
M aize V I
3.00
4.704
M ilk Powder V I
3.00
10.986
Maize V2
3.00
8.076
Milk Powder V2
3.00
7.841
M aize VG
3.00
8.802
M ilk Powder VG
3.00
9.216
M aize VN
3.00
5.969
M ilk Powder VN
3.00
7.940
M odel
Critical Value*
F-Statistic
Tomato V I
3 .00
9.571
Tomato V2
3 .00
Tomato VG
Model
* F-Critical Values at 5 Percent level of significance with two and 180 degrees o f freedom.
247
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
7.7.3 Test on the Equality of the Coefficients of Foreign Prices and Exchange Rate
In this study, the exchange rate has been incorporated directly into the trade model. To
determine if this was appropriate, linear restriction on the (3 parameter is formulated to
verify that the coefficient of foreign price is in fact equal to the coefficient of the
exchange rate.
The null hypothesis of the equality of foreign price and exchange rate coefficients
is rejected for all sixteen commodity-specific models at the five percent level of
significance (Table 7.11). The results imply that when estimating the Mexico-US import
demand functions, it is appropriate to include exchange rate directly into the econometric
model. Therefore, the approach followed in this study to incorporate exchange rate as a
separate variable into the trade model is appropriate.
Table 7.11 Short-Run Test Results on Equality of Coefficients of Exchange Rate and
Foreign Price on Mexico-US Import Demand
B-restrictions: H 0: p 3j = Psj f o r i j = 1,2,3,.
Model
Critical Value*
F-Statistic
Tomato V I
Tomato V2
3.84
3.84
12.950
6.240
Tomato VG
Critical Value*
F-Statistic
Sorghum V I
Sorghum V2
3.84
3.84
8.241
6.670
Model
3 .8 4
6.220
Sorghum VG
3.84
7.774
Tomato VN
3.84
5.055
Sorghum VN
3.84
7.213
M aize V I
M aize V2
3.84
3.84
7.198
9.084
Milk Powder V I
Milk Powder V2
3.84
3.84
17.660
12.615
M aize VG
3.84
9.540
Milk Powder VG
3.84
12.473
M aize VN
3.84
10.941
Milk Powder VN
3.84
12.104
* F-Critical Values at 5 Percent level of significance with one and 180 degrees of freedom.
248
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
249
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
goods and so, traders care more about exchange rate changes than about its volatility. In
the case of the ratio of wage rates, the small value of the coefficient could be an
indication of declining importance wages to total production costs and thus, has a small
effect on tomato trade flows.
Comparing results of M exicos import demand functions for maize, sorghum and
milk powder, the results suggest that the own price variable has higher impact on milk
powder imports than on imports of maize and sorghum from the United States into
Mexico. It could be partly due to the fact that milk powder is a higher value-added good
relative to maize and sorghum and consumer devote a higher income share on milk and
milk products than that on the other two goods. In fact, sorghum models have the smaller
values of the coefficients of own price and income variables. While maize and sorghum
are two of the most subsidized commodities in Mexico, the dairy industry has been
subject to less government subsidy which may also have been reflected in higher values
of the estimated coefficients for milk powder. In fact, Maize has received the highest
levels of support on a per-ton basis than any other commodities in the pre NAFTA era
(World Bank, 2005). The coefficients of exchange rate and volatility are higher for milk
powder model than those for maize and sorghum reflecting the increasing tendency of
dairy imports into Mexico. The value of the elasticity of the exchange rate using YG is in
line with the value of 0.599 reported by Tanyeri-Abur and Rosson (1997). Finally, the
NAFTA has influenced maize and milk powder imports from the United States into
Mexico more significantly than the imports of sorghum. This could be due to the fact that
the levels of sorghum imports were increasing even before the NAFTA agreement as a
results of high growth rates in the hog and poultry industries in Mexico.
250
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
Table 7.12 Results of the VECM for Four Commodities using Different
Specifications of the Volatility of the Exchange Rate
M odels for Tomatoes
Variable
Lagged Volume
US Income
Exchange Rate
Volatility VI
Own Price
Related Price
W age Rate
ECT
NAFTA
Variable
Lagged Volume
MX Income
Exchange Rate
Volatility VI
Inventory
Own Price
Related Price
ECT
NAFTA
Expected Sign
VI
V2
(+)
(+)
(+)
(-)
(-)
(+)
(+)
(-)
(+)
0.331
0.156
0.802
-0.064
-0.117
0.457
0.295
-0.048
0.495
0.178
0.206
0.706
0.029
-0.152
0.536
0.127
-0.072
0.475
Expected Sign
(+)
(+)
(+)
(-)
(+)
(-)
(+)
(-)
(+)
0.122
0.087
0.086
0.234
0.090
-0.155
-0.134
-0.156
0.170
-1.144
0.379
0.221
- 0.222
0.142
-0.147
0.181
-0.018
0.245
VG
VN
0.198
0.663
-0.043
-0.099
0.195
0.130
-0.030
0.427
0.169
0.098
0.499
-0.108
-0.092
0.416
0.343
-0.041
0.523
VG
VN
0.818
0.125
0.244
-0.057
0.144
-0.138
0.153
-0.018
1.161
0.070
0.133
0.259
-0.391
0.137
-0.152
0.144
-0.019
1.808
0.110
Variable
Lagged Volume
M X Income
Exchange Rate
Volatility VI
Inventory
Own Price
Related Price
ECT
NAFTA
Variable
Lagged Volume
M X Income
Exchange Rate
Volatility VI
Milk Short-fall
Own Price
Related Price
ECT
NAFTA
Expected Sign
VI
V2
VG
VN
(+)
0.130
0.065
0.144
-0.181
0.063
-0.126
0.306
-0.011
0.137
0.134
0.064
0.153
-0.091
0.069
-0.107
0.292
-0.011
0.101
0.123
0.061
0.119
-0.103
0.068
-0.154
0.288
-0.012
0.039
0.128
0.062
0.150
-0.243
0.102
-0.111
0.291
-0.012
0.053
VG
VN
0.171
0.097
0.766
-0.316
0.450
-0.548
0.876
-0.091
1.163
0.229
0.283
0.730
-0.903
0.467
-0.608
0.848
-0.092
0.781
(+)
(+)
(-)
(+)
(-)
(+)
(-)
(+)
Expected Sign
(+)
(+)
(+)
(-)
(+)
(-)
(+)
(-)
(+)
0.200
0.063
0.484
-0.133
0.443
-0.530
0.752
-0.077
1.431
251
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
7.9 Summary
The main results obtained from the Vector Error Correction models for the selected
commodities are presented in this chapter. The results indicate that all import demand
functions yield coefficients with theoretically expected signs. In general, short-run
elasticities are smaller than the corresponding long-run elasticities and so the results are
consistent with the Le Chatelier principle. Considering that each regressor in the ECM is
specified in the first-differenced form, the empirical results suggest that the statistical fit
of all import demand functions to data is satisfactory.
Although many factors affect agricultural trade, exchange rates and its volatility
play an important role in the agri-food trade between Mexico and the United States.
While the magnitude of the impact of exchange rates on trade varies by commodity, in
most cases, the responsiveness of agricultural imports to prices is inelastic. The above is
also true for the effect of exchange rate volatility on trade flows. Except for one case, the
exchange rate volatility has a significant negative effect on the trade flows of the selected
commodities. The free trade agreement seems to have played a significant enhancing role
in the trade flows of tomato and maize, but not for sorghum and milk powder.
All error correction terms are negative and significant in all models. As pointed
out by Harris and Sollis (2003), this result supports the validity of an equilibrium
relationship among the variables in each cointegrating equation. This implies that
overlooking the cointegration among the variables would have introduced misspecification in the underlying dynamic structure for each of the four commodities
considered in this study.
252
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
The change in real imports per period (month in this study) that is attributed to
disequilibrium between the actual and equilibrium levels is measured by the absolute
values of the ECT in each equation. There is a considerable inter-commodity variation in
the adjustment speed to the last periods disequilibrium. In other words, the adjustment of
import volume to changes in its explanatory variables may take about 10-12 months for
tomato and milk powder while up to 28-30 months for maize and sorghum.
The ECM results in this study show that exchange rate volatility has a significant
short-run effect on import demand in addition to its long-run effects. To reinforce the
above finding, the null hypothesis that that exchange rate and volatility have an
insignificant effect on the Mexico-US trade flows was tested formally. The null
hypothesis was rejected using a t-test as well as using the F-test. Thus, ignoring such
variables in the analysis of Mexico-United States trade equations could produce biased
results. It can be argued that traditional import demand studies for commodities that do
not include a variable representing the effects of exchange rate uncertainty are potentially
mis-specified. Finally, a comparison between models for each commodity was made to
informally determine which volatility specification provides more consistent results than
others. It is concluded that models using VG have produced more consistent coefficients
both in terms of signs and magnitudes.
253
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
CHAPTER 8
CONCLUDING REMARKS
8.1 Introduction
The objective of this chapter is to summarize the major findings of this study, highlights
the key contributions, and identifies certain limitations of this study. It also suggests
some avenues for future research.
Since the initiation of economic and trade liberalization in the mid 1980s in
Mexican agricultural sector, both exports and imports have grown dramatically in terms
of volume and value. It has been argued in policy and academic circles that exchange rate
played an important role in the performance of Mexican agricultural sector during last
two decades. While some studies of the Unites States-Mexico agricultural trade have
emphasized the importance of NAFTA in the dramatic increase in exports and imports
(Rosenzweig, 1996; Schwentesius, et. al., 2001), others have emphasized the role played
by exchange rate changes in enhancing Mexican trade performance in the agri-food
sector (Espinoza-Arellano 1998; Malaga et al 2001; Mora-Flores 2002). Since the shift
from the fixed to a floating exchange rate system in 1995, Mexico-Unites States
exchange rate has been characterized by periods of unexpected calm followed by
episodes of high volatility, which may also has influenced agri-food trade flows between
the two countries.
While the border between Mexico and its important trade partners, the Unites
States and Canada, became increasingly open due to NAFTA, Mexican agri-food trade
has also been influenced by changes in exchange rate and exchange rate volatility. It is
yet to be determined the extent to which the NAFTA, the continuous devaluation of the
254
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
Mexican peso, and changes in the exchange rate volatility have contributed to the growth
in Mexican agri-food trade during the last decade. Since the Unites States is the most
influential trade partner of Mexico, it is important to investigate the extent to which
exchange rate changes, changes in the volatility of the Unites States-Mexico exchange
rate and the NAFTA have contributed to the expansion of the Mexico-Unites States trade
in agri-food commodities.
While the available literature on the effects of Mexico-Unites States exchange
rate on agricultural trade provide evidence that changes in exchange rate do have an
important effect on trade, no research has been conducted yet to investigate the effects of
exchange rate volatility on Mexican agricultural trade. Empirical studies on the effects of
exchange rate movements on Mexico-Unites States agri-food trade conducted in the past
have ignored the volatility issues and focused only on short-run effects of exchange rate
on trade. Moreover, they concentrated on aggregate trade rather than on trade in specific
commodities. The previous studies have also ignored the time series properties of the data
used in econometric estimation. In view of these limitations in the existing studies and
the fact that they provide inconclusive evidences of the effects of exchange rate
movements on trade flows, efforts are made in this research to undertake a more
comprehensive study of the effects of exchange rate changes and its variability on
Mexico-Unites States agri-food trade flows. Four important traded commodities: tomato,
maize, sorghum, and milk powder, are analyzed for the period from 1989 to 2004.
The economic research problem addressed in this study deals with the nature and
extent of the influence of exchange rate changes and exchange rate volatility on Mexican
agri-food trade with the Unites States during the last decade. The purpose of this study is
255
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
to quantify the effects of the Mexico-Unites States exchange rate changes and its
volatility on Mexico-Unites States agri-food trade flows using the Maximum Likelihood
Cointegration analysis. This research attempted to accomplish the following specific
objectives:
To determine the long-run and short-run effects of changes in the exchange rate
and its volatility on Mexico-United States agri-food trade.
To test if real exchange rate volatility has an adverse effect on the quantity of the
Mexico-Unites States agri-food trade flows using a set of alternative volatility
measures.
To discuss policy implications of the effects of exchange rate and its volatility on
Mexican agri-food trade flows.
To accomplish the above objectives, the following set of hypothesis stated in null form
was tested in this study:
Exchange rate volatility has an insignificant effect on the quantity of the MexicoUS agri-food trade flows in the short-run.
256
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
The coefficient of exchange rate and foreign price is equal (the appropriateness of
using exchange rate as a separate variable in the trade model is tested).
257
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
the own price of imports responds positively to changes in exchange rate (i.e. the output
price respond positively given a devaluation of the Mexico-United States change rate).
The empirical model derived from the theoretical framework has been used to
determine the effects of exchange rate changes, its volatility as well as the effects of
income and prices on the trade volumes of specific commodities.
While both the structural approach and the reduced form approach have been used
in the literature to determine the effects of exchange rate on trade flows of agri-food
commodities, a reduced form model has been used in this study. It allows a simpler
means of studying the economic relationship between the exchange rate movements and
the variables entering the model such as prices, quantities and other variables than the
other approaches. The main advantage of the reduced form model is that meaningful
economic hypotheses can be tested without estimation of structural parameters and it can
accommodate non-stationary data. It is widely known that when non-stationary variables
are used in regression analysis, the classical t and F-tests are not appropriate and may
lead to misleading conclusions (Hendry, 1986). The reduced form approach also offers a
convenient framework for addressing data non stationarity issue in econometric
estimation.
The econometric approach used in this research consisted of testing for unit root
in each data series. To this end, the ADF test was employed. The Johansens Maximum
likelihood technique has been used to test for cointegration. This was followed by the
estimation of Vector Error Correction models and hypothesis testing.
While it is generally recognized in the trade literature that exchange rate volatility
induces additional uncertainty in trade, no consensus exists on how to measure it. A
258
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
number of statistical measures of variability of the exchange rate have been used in the
literature. Economic theory does not provide clear guidance on the features of the most
appropriate volatility measure (Clark, et al 2004). In an attempt to shed light on the
effects of different volatility measures on trade flows, four different volatility measures
were constructed and employed in this study, including a volatility measure based on a
GARCH model, a non-parametric exchange rate volatility measure, and two variance or
standard deviation-based volatility measures (moving average of the standard deviation
of the exchange rate and the average absolute difference between the previous forward
and the current spot rate).
259
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
The results from the estimation of four different volatility measures show that the
standard deviation based measures of volatility (V I and V2) closely track exchange rate
fluctuations. With respect to the GARCH model, it was found that GARCH (1,1)
describes well the exchange rate data based on significance of the parameters and related
statistics. Finally, a non-parametric regression was performed using the kernel regression
approach with a normal kernel function to generate the non-parametric volatility measure.
These four volatility measures were incorporated in the estimation of the import demand
function for each of the four commodities considered in this study.
With regard to the unit root test, the ADF test was conducted to determine the
integration properties of individual series used in each commodity equation. Additionally,
for the exchange rate variable, the Perron test was used to determine if there was a
structural break in these series. The results indicated that no structural break took place in
the sample period despite the outlier.
To determine the lag length truncation for each series, the Akaike (1973)
Information Criterion (AIC) and Schwarz (1978) Criterion (SBC) were applied. The
results of the ADF test show that all economic variables included in each of the four
commodity models are non-stationary and they have a unit root in its univariate
representation.
For the Unites States import demand for tomato, cointegration results suggest that
the Unites States personal disposable income and exchange rate have positive impacts on
trade flows in all models using different volatility specifications. These results imply that,
in the long run, devaluation of the Mexican peso relative to the Unites States dollar works
as an incentive for importers to increase the quantity of tomato imported into the Unites
260
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
States from Mexico. On the other hand, the volatility of the exchange rate produced
mixed results. W hile three tomato models indicated a negative relationship between
volatility and trade flows, one of them suggested a positive relationship.
With regard to the price of tomato, an expected negative relationship is suggested
between own price and trade volume in all models. For the related price, all tomato
models suggest a positive relationship between price of tomato in the Unites States and
Unites States import demand function, but one model yields a coefficient that is not
statistically significant. Thus, the tomato produced in the Unites States is considered a
substitute of the Unites States imports of tomato from Mexico. There is a positive
relationship between farm wage rate ratio and the Unites States import demand. The
increasing wage differential, about 8-9 times, between the Unites States and M exicos
rural wage rates for the analyzed period appear to have enhanced tomato imports from
Mexico.
For M exicos import demand for maize, both the income and exchange rate
variable have significant positive influence on trade flows. The own price of maize and
the price of the related commodity have the expected sign in all models. Thus, M exicos
domestic maize can be considered as a substitute for maize imports. The above result
makes sense since domestic maize is used to complement the requirements for maize
used in feeding animals in Mexico. The inventory variable has a positive sign in all four
maize models. It is interesting to note that exchange rate volatility has a negative impact
on trade volumes in fifteen out of sixteen models. This is consistent with the behavior of
risk-averse importers.
261
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
An important finding of this research is that the specification of the exchange rate
volatility does matter. Not only different elasticity values are obtained for different
exchange rate volatility measures used, but also, for some models, the magnitudes of
some variables differ across models. While the magnitude of the impact of exchange rates
on trade varies by commodity, in most cases, the responsiveness of agricultural exports to
changes in exchange rate is less than one. In regards to the speed of adjustment, the
magnitude of the estimated weights suggest that in the event of a disturbance to the long
run equilibrium, import volume, own price, and volatility of the exchange rate would
adjust faster than any other variable to restore the equilibrium condition.
The overall results from the ECM indicate that almost all the import demand
models yield theoretically expected signs and magnitudes of relevant variables.
Furthermore, most short-run elasticities are smaller than the corresponding long-run
elasticities. Thus, the Le Chatelier principle holds in almost all of the models.
Considering that each regressor in the ECM is specified in first-differenced form, the
empirical results suggest that the statistical fit of most of the models to the data are fairly
good as indicated by the R , F-statistic, and the set of diagnostic tests employed.
While the magnitude of the impact of exchange rate on trade flows varies by
commodity, in most cases, the responsiveness of agricultural trade flows is less than one
in the short-run. The above is also true for the coefficient of the volatility variable. The
sign of the volatility variable is negative and consistent across all commodities. In
general, the short-run income elasticity of import demand is inelastic, as expected for all
commodities.
262
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
The free trade dummy variable has a positive impact on trade flows of tomato and
maize, but this is not the case for sorghum and milk powder where the free trade dummy
yields positive but insignificant coefficients. Finally, seasonality influences sorghum and
maize trade flows but not the trade flows of tomatoes and milk powder.
The error correction term is negative and significant in all models. This result
reconfirms the existence of long-run equilibrium relationships among the variables in
each cointegrating equation. It implies that overlooking the cointegration of the variables
would have introduced misspecification bias in the results.
The change in real imports per period (month in this study) that is attributed to
disequilibrium between the actual and equilibrium levels is measured by the absolute
values of the ECT in each equation. There is considerable inter-commodity variation in
the adjustment speed to the last periods disequilibrium. While the adjustment of import
volume to changes in various explanatory variables takes about 10 months for tomato and
20 months for milk powder, it takes about 30 months for maize and sorghum.
Finally, the ECM results in this study show that exchange rate volatility has a
significant effect on import demand, both in the short-run and in the long-run. Thus,
ignoring such a variable could produce biased results. Thus, the traditional import
demand studies that do not include a variable representing the effects of exchange rate
uncertainty are potentially mis-specified.
The results from hypothesis tests indicate that the exchange rate and volatility of
the exchange rate do have a statistically significant effect on the Mexico-United States
trade flows of tomato, maize, sorghum and milk powder. So excluding these variables
from the trade models could lead to model mis-specification and biased results. The
263
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
results also suggest that for the commodities considered in this study it is appropriate to
include the exchange rate variable directly into the trade model.
Comparing results of the four models across four alternative specifications, it was
found that, for all import demand functions, models using VG provide more consistent
results with theory than those from other models. For the four commodity models using
VG, all the variables have the a priory expected signs and magnitudes. Furthermore, a
comparison of the estimated results for models using VG shows than institutional features
related to each commodity are useful in explaining the magnitudes of trade flows
between Mexico and the United States.
264
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
Secondly, the results from cointegration analysis suggest that trade volumes
between Mexico and the US are sensitive to both short-run and long-run changes to the
bilateral exchange rate. They are also sensitive to exchange rate uncertainty. This means
that assisting the traders to access forward markets could be helpful to mitigate exchange
rate uncertainty, which will have a beneficial effect on trade flows and induce better
allocation of resources in Mexico.
265
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
266
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
short-run results of this study. The dynamic response of international trade flows to
exchange rate and its volatility shocks could be evaluated by employing impulseresponse and variance decomposition analysis. Future research along this line will
enhance our understanding of the short-run dynamic effects of changes in exchange rate
and its volatility on agricultural trade flows between Mexico and the United States.
267
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
References
Akaike, H., (1973) Information Theory and an Extension of the Maximum Likelihood
Principle Proc. Second International Simposium o f Information Theory. 267-281.
Akhtar, M.A and Hilton, S.R., (1984) Effects of Exchange Rate Uncertainty on German
and U.S. Trade FRBNY Quarterly Review!Spring 7-16.
Alston, J.M., Carter, C.A,.Green R, and Pick. D., (1990) Whither Armington Trade
Models? Amer. J. Agr. Econ. 72:455-67.
Anderson, M., and Garcia, P., (1989). Exchange Rate Uncertainty and the Demand for
U.S Soybeans, American Agricultural Economics Associations, August 1989. 721-729.
Anderson, J., E., (1979) A Theoretical Foundation for the Gravity Equation American
Economic Review American Economic Association, vol. 69(1), pages 106-16.
Appelbaum, E., and Kohli, U., (1997) Import Price Uncertainty and Distribution of
Income Review o f Economics and Statistics 79, 620-630.
Appelbaum, E., (1993) Empirical Application of Duality UNDER Uncertainty Paper
presented at The Annual Meetings o f the European Economic Association, Helsinky.
Appelbaum, E., and Kohli, U., (1979) Canada-United States Trade: Test for the SmallOpen Economy Hypothesis Canada Journal Economics 12: 1- 4.
Arellano et al., (1998) Analysis of Forces Affecting Competitive Position of Mexico in
Suplf US Winter Melons International Food and Agribusiness Management Review. 1(4)
495-507.
Arize A. and Steven S. S., (1998) Cointegration, Real Exchange Rate and Modelling the
Demand for Broad Money in Japan Applied Economics, Vol. 25 (June), pp. 717-26.
Armington, P., (1969) A theory of Demand for Products Distinguished by Place of
Production IM F Staff Papers, 16, 159-178.
Arrow, K., and Kurtz, M., (1970) "Optimal Growth with Irreversible Investment in a
Ramsey Model" Econometrica.
Asseery, A., and Peel, D.A., (1991) The Effects of Exchange Rate Volatility on
Exports. Economics Letters 37, 173-177.
Avalos-Sartorio, B., (1998) Structural Developments in the Mexican Grain-Livestock
Subsector: The Grain and Oilseed Subsector. In Loyns, et al.(editors). Economic
Harmonization in the Canadian/U.S./Mexican Grain Subsector. Proceedings of the
268
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
Fourth Agricultural and Food Policy Systems Information Workshop. Texas A&M
University, University of Guelph, pp. 161-186.
Bailey, M.J., Tavlas, G.S., and Ulan M., (1987) The Impact of Exchange - Rate
Volatility on Export Growth. Some Theoretical Considerations and Empiricals Results
in. Journal o f Policy Modeling vol 9, no 1 22: 225-243.
Bailey, M., Tavlas, S.G., and Ulam, M., (1986) Exchange-Rate Variability and Trade
Performance: Evidence for the Big Seven Industrial Countries," Weltwirtschaftllches
Archiv 122 :466-77.
Baltagi, B.H
& Li, Q., (2001) "Estimation O f Econometric Models With
Nonparametrically Specified Risk Terms," Econometric Reviews, Taylor and Francis
Journals, vol. 20(4), 445-460.
Bank of Mexico (1996) La Economia Mexicana Informe Anual. Mexico.
Bamerjee, A., Doblado J.J., Hendry, D.F., and Smith, G.W., (1986) Exploring
Equilibrium Relationships in Econometric Theory Through Static Models: Some Monte
Carlo Eevidence , Oxford Bulletin o f Economics and Statistics, 48: 253-77.
Batra, R.N. and Ullah, A. (1974). Competitive Firm and the Theory of Input Demand
Under Price Uncertainty The Journal o f Political Economy Vol. 82, No 3, 537-548.
Batten, D.S. and Belongia, M.T. (1986) Monetary Policy, Real Exchange Rates and U.S.
Agricultural Exports. American Journal o f Agricultural Economics 68: 422-427.
Bazdresch, S. And Werner, A., (2005) Regime, Switching Models for the Mexican
Peso Journal o f International Economics, Volume 65, Issue 1 pp 185-201.
Bazdresch, S. (2002) El Cmportamiento del Tipo de Cambio en Mexico y el Regimen de
Libre Flotacion: 1996-2001 Documento de Investigacion No. 2 2002-09, Banco de
Mexico.
Belanger, D., Gutierrez, S., Racette, D., and Raynauld, J., (1992) The Impact of
Exchange Rate Variability on Trade Flows: Further Results on Sectoral U.S. Imports
from Canada. North American Journal o f Economics and Finance 3: 888-892.
Bergstran J.H., (1989) The Generalizad Gravity Equation, Monopolistic Competition,
and the Factor Proportion Theory of International Trade The Review o f Economics and
Statistics, Vol. 71, Issue 1, pages 143-53.
Bergstran J.H., (1985) The Gravity Equation in International Trade: Some
Microeconomic Foundations and Empirical Evidnce The Review o f Economics and
Statistics, 1985, 67, (3), 474-81.
269
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
Bessler, D. A., and Kling, J.L., (1986) "Forecasting Vector Autoregressions with
Bayesian Priors." American Journal o f Agricultural Economics 68(1): 144-51.
Bini-Smaghi, L., (1991) Exchange Rate Variability and Trade: Why is it so Difficult to
Find any Empirical Relationship? Applied Economics 23, 927-936.
Bollerslev, T., Chou, R. S., and Kroner, K. F., (1992) "ARCH Modelling in Finance a
Review of the Theory and Empirical Evidence" Journal o f Econometrics, 52, 5-59.
Bollerslev, T. (1986) "Generalised Autoregressive Conditional Heteroskedasticity" J.
Econometrics, 51, 307-327.
Bossaerts, P., (1988) Common Non-Stationary Components of Asset Prices Journal
o f Economic Dynamics and Control, 12, 347-364.
Bradshaw, G., and Orden, D., (1990) Granger Causality from the Exchange Rate to
Agricultural Prices and Export Sales. Western Journal o f Agricultural Economics. 15:1
(July), pp. 100-110.
Brodsky, D.A., (1984) Fixed Versus Flexible Exchange Rates and the Measurement of
Exchange Rate Instability Journal o f International Economics 16, pp. 295-306
Brooks C., (2002) Introductory Econometrics to Finance Cambridge University Press
Broil, U., (1994) Foreign Production and Forward Markets Australian Economic
Papers. 62 PP 1-6.
Byerlee, D., and Sain, G., (1991) Relative Food Prices Under Structural Adjustment:
Preliminary Findings from Latin America . Food Policy 16(1): 74-84.
Caballero, J. R., and Corbo, V., (1989) The Effect of Real Exchange Rate Uncertainty
on Exports: Empirical Evidence The World Bank Economic Review, Vol 3 2:263-280
Calva J.L., (1996) La Economia Nacional y la Agricultura de Mexico, A Tres Anos de
O peration
del TLCAN,
NAFTA
and
Agriculture:
Is
the
Experiment
Working Symposium Proceedings.
Carone, G. (1996) "Modeling the U.S. Demand for Imports Through Cointegration and
Error Correction" Journal o f Policy Modeling, Elsevier, vol. 18(1), pages 1-48.
Carter, C.A., Gray, R.S., Furtan, W.H., (1990) Exchange Rate Effects on Inputs and
Outputs in Canadian Agriculture American Journal o f Agricultural Economics. Vol.
72(3).
Chambers, R.G., and Just, R.E., (1981) Effects of Exchange Rate Changes on U.S.
agriculture: a Dynamic Analysis American Journal Agricultural Economics, 61: 249257.
270
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
Chambers, R.G., and Just, R.E., (1979) A critique of Exchange Rate Treatment in
Agricultural Trade Models. American Journal Agricultural Economics, February 63:
32-46.
Cho, G., Sheldon, M. I., and McCorriston S., (2002) Exchange Rate Uncertainty and
Agricultural Trade American Journal of. Agricultural Economics,. 84 (4) November:
931-942.
Chowdhurry, A.R., (1993) Does Exchange Rate Volatility Depress Trade Flows?
Evidence from Error-Correction Models The Review o f Economics and Statistics, 75:
700-705.
Clark, P.B., Tamarisa, N., and Shang-Jin, W., with Sadikow, A., and Zeng L., (2004)
Exchange Rate Volatility and Trade Flows-Some New Evidence International
Monetary fund, May, 2004.
Clark, P. B., (1973) Uncertainty, Exchange Risk, and the Level of International Trade,
Western Economic Journal 11, September, pp. 302-13.
Clements, M. P., and Hendry, D. F (2002) An Overview of Economic Forecasting in.
(eds.)A Companion to Economic Forecasting, Blackwells. 1- 18.
Coleman, J.R. and Meilke, D.K., (1988) The Influence of Exchange Rates on Red Meat
Trade Between Canada and United States Canadian Journal o f Agricultural Economics
401-424.
Cook, R, (2001) "The U.S. Fresh Produce Industry: An Industry in Transition," Chapter 2
in Postharvest Technology of Horticultural Crops, Adel A. Kader (eds.), University o f
California Division o f Agriculture and Natural Resources, Publication 3311, pp.5- 26,
2001 .
Collins, J.K., Meyers, H.W., and Bredahl, E.M., (1980)Multiple Exchange Rate Change
and U.S. Agricultural Commodity Prices Amer. J. Agr. Econ. 62: 656-665.
Cushman, D.O., (1988) U.S. Bilateral Trade Flows and Exchange Risk During the
Floating Period Journal o f International Economics 24,317-330.
Cushman, D.O., (1986) Has Exchange Risk Depressed International Trade? The Impact
of Third-Country Exchange Risk Journal o f International Money and Finance 5, 361379.
Cushman, D.O., (1983) The Effects of Real Exchange Rate Risk on International Trade
Journal o f International Economics 15, 45-63.
271
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
Deardorff, A
(1998) Determinants of Bilateral Trade: Does Gravity Work in a
Neoclassical W orld?, in J.A. Frankel, ed. by The Regionalization o f the World Economy,
(Chicago: University of Chicago Press), pp. 7-22.
De Grauwe, P., and De Bellefroid., (1988) Long Run Exchange Rate Variability and
International Trade, Chapter 8 in S. Arndt and J.D. Richardson, ed. by Real Financial
Linkages Among Open Economies, (London: The MIT Press).
De Grauwe, P., (1988) Exchange Rate Variability and the Slowdown in Growth of
International Trade IMF Staff Paper. 35: 63-84.
DeJong, D.N., and Whiteman C.H., (1989) Trends and Random Walks in MacroEconomic Time Series: A Reconsideration Based on the Likelihood Principle Working
paper No. 1989- 4, University of Iowa.
DeJong, D.N., Nankervis J.C., Savin, N. E., and Whiteman, C.H., (1992) The Power
Problem of Unit Root Tests for Time Series with Autoregressive Errors Journal o f
Econometrics, 53: 323-43.
Dellas, H., and Ben-Zion Z., (1993) Real Exchange Rate Volatility and International
Trade: A Reexamination of the Theory. Southern Economic Journal 59 (April): 641-7.
Demers, M., (1991) Investment Under Uncertainty and the Arrival of Information Over
Time The review o f Economic Studies. Vol 58(2) 333-350
Dhrymes, P. J., (1998) Time Series, Unit Roots, and Cointegration . Academic Press,
San Diego California.
Diaz-Garces. D., (2002) Analysis de las Funciones de Importacion y Exportacion de
Mexico, 1980-2000 Documento de Investigacion, Direccion General de Investigacion
Economica. Banco de Mexico.
Dickey, D. A., and Fuller W.A., (1981) Likelihood Ratio Statistics for Autoregressive
Time Series with a Unit Root. Econometrica 49, 1057-1072.
Dickey, D. A., and Fuller, W.A., (1979) Distribution of the Estimators for
Autoregressive Time Series with a Unit Root Journal o f American Statistical
Association, 74: 427-31.
Dickey, D. A., and Fuller W.A., (1976) Distribution of First Order Autoregressive
Estimator" Proceedings o f the Business and Economic Statistocs Section,American
Statistical Association, 278-281.
Dorfman, J. H., and Lastrapes, D.W., (1996) The Dynamic Responses of Crop and
Livestock Prices to Money-Supply Shocks: A Bayesian Analysis Using Long-Run
Identifying Restrictions. American Journal o f Agricultural Economics. 78:3. pp. 530-41.
272
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
Dombusch, R., (1976) Expectations and Exchange Rate Dynamics Journal o f Political
Economy, Vol. 84, pp. 1161-76.
Elliot, G., Rothenberg T J , and Stock, J.H., (1996) Efficient Tests for an Autoregressive
Unit Root Econometrica, 64: 813-836.
Enders, W., (2003) Applied Econometric Time Series. Publisher: John Wiley and Sons
(WIE), 2003.
Enders, W., ( 1995) Applied Econometric Time Series New York. John Wiley.
Engle, R (2001) "GARCH 101: The Use of ARCH/GARCH Models in Applied
Econometrics," Journal o f Economic Perspectives, American Economic Association, vol.
15(4), pages 157-168.
Engle, R., (2004) "Risk and Volatility: Econometric Models and Financial Practice,"
American Economic Review, American Economic Association, vol. 94(3), pages 405-420.
Engle, R.F. and Granger, C.W.J., (1987) Cointegration and ntegration and Error
Correction: Representation, Estimation and Testing Econometrica, 55: 251-276.
Engle, R.F., and Granger, C.W.J., (1987) Co-integration and Error Correction:
Representation, Estimation and Testing Econometrica 55, 251-276
Engle, R.F. and Yoo B.S. (1987) Forecasting and Testing in Co-Integrated Systems,
Journal o f Econometrics, 35: 143-59.
Engle, R. F., (1982). "Autoregressive Conditional Heteroscedasticity, with Estimates of
the Variance of United Kingdom Inflation" Econometrica, 50, 987-1007.
Ericsson, R. N., (2001) "Forecast Uncertainty in Economic Modeling," International
Finance Discussion Papers 697, Board of Governors of the Federal Reserve System
(U.S.).
Espinoza-Arellano, J.J., Fuller S., and Malaga J., (1998) Analysis of Forces Affecting
Competitiveness of Mexico in Supplying U.S. Winter Melon Market International Food
and Agribusiness Management Review1, 4: 495-507a.
Ethier, W., (1973) International Trade and Forward Exchange Market The American
Economic Review, Vol 63, No.3, 494-503.
Fabiosa, F.J., (2002) Assesing the Impact of the Exchange Rate and its Volatility on
Canadian Pork and Live Swine Exports to the United States and Japan Working Paper
02-WP 355 Center fo r Agricultural and Rural Development, june 2002.
Fendel R., (2002) Taylor-Type Rules and the Term Structure of Interest Rates mimeo.
273
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
Fleming, J. M., (1962) Domestic Financial Policies Under Fixed and Under Floating
Exchange Rates, IM F Staff Papers 9, 3: 369-79.
Fox, J., (2000) Nonparametric Simple Regression: Smoothing Scatterplots , Thousand
Oaks, CA, Sage Publications.
Fox, J., (2000) Multiple and Generalized Nonparametric Regression. Thousand Oaks,
CA, Sage Publications.
Franke. G., (1991) Exchange Rate Volatility and International Trading Strategy Journal
o f International Money and Finance 10, 292-307.
Frankel, J.A., (1991) The Making of Exchangea Rate Policy in the 1980s Economics
Working Papers 91-157 University of California at Berkeley.
Frenkel J.A., and Mussa, M.L., (1985) Asset Markets, Exchange Rate and the Balance of
Payments. Chapter 14. Handbook o f International Economics. Vol 3. Ansterdam Norht
Holland.
Frankel, J.A., (1979) On the Mark: A Theory of Floating Exchange Rates Based on Real
Interest Differentials" American Economic Review Vol. 69, pp. 610-22.
Frenkel J.A, (1976) A Monetary Approach to the Exchange Rate: Doctrinal Aspect and
Empirical Evidence Scandinavian Journal o f Economics. 200-24.
Friedman, D., and Vandersteel, S., (1982) Short-Run Fluctuations in Foreign Exchange
Rates Journal o f International Economics, 13:171-186.
Fuller, S., Gutierrez, N., and Capps, O., Jr (1992) International Dry Onion Trade:
Factors Affecting Import Demands for U.S, Dry Onions Agribusiness. 8:445-455.
Gagnon, J.E. (1993) Exchange Rate Variability and the Level of International Trade
Journal o f International Economics 34, 269-287.
Gallant, A. R., and Tauchen, G., (1989) Semi Non-Parametric Estimation of
Conditionally Constrained Heterogeneous Processes.Asset pricing Applications
Econometrica 57:1091-1120.
Gonzalo, J., (1994) Five Alternative Methods of Estimating Long-Run Equilibrium
Relationships Journal o f Econometrics 60, 203-233.
Gotur, P., (1985) "Effects of Exchange Rate Volatility on Trade; Some Further
Evidence," IMF S taff Papers 32 /475-512.
Gourieroux, C., (1997) ARCH Models and Financial Applications. New York: Springer
Verlag.
274
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
275
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
Hendry, D.F. (1995) On the Interactions of Unit Roots and Exogeneity . Econometric
Reviews, 14, 383-419.
Hendry, D.F. (1987) Econometric Methodology: A personal perspective. In T.F.
Bewley (Ed.), Advances in Econometrics, Chapter 10. Cambridge: Cambridge University
Press.
Hendry, D.F., (1986) Econometric Modelling with Cointegrated Variables: An
Overview, Oxford Bulletin o f Economics and Statistics, 48, 201-12.
Hooper, P., and Kohlhagen, S.W., (1978) The Effect of Exchange Rate Uncertainty on
the Prices and Volume of International Trade Journal o f International Economics 8:
483-511.
Hubrich, K., Lutkepohl, H., and Saikkonen, P., (2001) "A Review Of Systems
Cointegration Tests" Econometric Reviews, Taylor and Francis Journals, vol. 20(3), pages
247-318.
Ian Sheldon (2003) Exchange Rate Uncertainty and Agricultural Trade Agricultural
Competitiveness and World Trade Liberalization: Implications of the WTO and
NAFTA, Fargo, ND, May 29-30, 2003.
INEGI (2004). Banco de Information Economica-Sector Extemo. Grupo de Trabajo:
SHCP, Banco de Mexico, Secret aria de Economfa e INEGI.
Jennings S., Adamowicz, W., and Constantino, L., (1990) The Canadian Lumber
Industry and the Macroeconomy: A Vector Autoregression Analysis. Canadian Journal
o f Forest Research. 21: 288-299.
Johansen, S., (1995) Likelihood Based Inference in Cointegrated Vector Autoregressive
Models Oxford University Press.
Johansen, S., (1991) Estimation and Hypothesis Testing of Cointegration Vectors in
Gaussian Vector Autoregressive Models. Econometrica. 59: 1551-1580.
Johansen, S., and Juselius, K., (1990) Maximum Likelihood Estimation and Inference on
Integration with Applications to the Demand for Money Oxford Bulletin o f Economics
and Statistics. 52: 169-179.
Johansen, S., (1988) Statistical Analysis of Cointegration Vectors. Journal o f Economic
Dynamic and Control. 12: 231-254.
Kapombe, C. M., and Colyer D., (1998) Modeling U.S. Broiler Supply Response: A
Structural Time Series Approach. Submitted to the Review o f Agricultural and Resource
Economics 7:2(Oct 1998):241-251.
276
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
277
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
Kwiatkowski, D., Phillips, P., Schmidt, P., and Y. Shin, Y., (1992) Testing the Null
Hypothesis of Stationary Against the Alternative of a Unit Root Journal o f Econometrics
54:159-178.
Lacroix, L.J.R., Schwentesius, R.R., Gomez, C.M.A., Echanove, H.F., (2001) Marketing
of Fruits and Vegetables The World Bank Group June 2001.
Lastrates, D.W., and Koray F., (1990) Exchange Rate Volatility and U.S. Multilateral
Trade Flows Journal o f Macroeconomics, vol 12, No 3, 341-362.
Levich, R.M., (1985) Empirical Studies of Exchange Rates: Price Behavior, Rate
Determination and Market Efficiency in: R.W. Jones and P.B. Kenen (eds.), Handbook
o f International Economics, Vol. II, Amsterdam: North Holland, 979-1040.
Liang, H., (1998), Real Exchange Rate Volatility: Does the Nominal Exchange Rate
Regime Matter? International Monetary Fund (working Paper).
Love, J., and Lucier, G., (1996) Florida-Mexico Competition in the US Market for Fresh
Vegetables Vegetables and Specialties. April. USDA-ERS.
Lutkepohl, H., (2001) Comment on Essays on Current State and Future Challenges of
Econometrics Journal o f Econometrics, Elsevier, vol. 100(1), pages 81-82.
Macdonald, R., and Taylor, M. P., (1992) Exchange Rate Economics: A Survey IM F
Staff Papers, March
Mackinnon, G.J., (1999) Critical Values for Cointegration Test in Long-Run Tests
Econometric Theory 15: 367-376
Malaga, E. J., Williams, W.G., and Fuller, W.S., (2001) US-Mexico Fresh Vegetables
Trade: the Effects of Trade Liberalization and Economic Growth Agricultural Economic
26, 45-55.
Malaga, E. J., Williams, W.G., and Fuller, W.S., (1996) Effects of NAFTA and Other
Exogenous Factors on the U.S.-Mexico Fresh Vegetables Trade: A Simulation Analysis
Symposium Nafta and Agriculture: Is the experiment working? San Antonio Texas.
Maskus, K.E., (1987) Exchange Rate Risk and US Trade: A Sectoral Analysis. Federal
Reserve Bank of Kansas City Economic Review, 3: 16-28.
McKenzie, D.M., and Melbourne, T.I.M.R., (1999) The Impact of Exchange Rate
Volatility on International Trade Flows Journal o f Economic Surveys vol. 13, 1: 21-38.
McKenzie, D.M., (1998) The Impact of Exchange Rate Volatility on Australian Trade
Flows Journal o f International Financial Markets 8, 21-38.
278
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
McKenzie, D.M., and Brooks D.R., (1997) The Impact of Exchange Rate Volatility on
German-US trade flows Journal o f International Financial Markets 7,73-87.
Meese, R. A., and Singleton, Kenneth J., (1983) Rational Expectations and the Volatility
of Floating Exchange Rates International Economic Review, Department of Economics,
University of Pennsylvania and Osaka University Institute of Social and Economic
Research Association, vol. 24(3), pages 721-33.
Mittelhammer, R.C., (1978) The estimation of domestic demand for salad vegetables
using a priori information. Ph.D. Dissertation. W ashington State University.
Mora-Flores,
S.
J.,
Lara-CovarrubiasD.,
Martinez-Damian,
A.M.,
(2003)
Competitiveness and Comparative Data Advantageas of Milk Production System in
Jalisco State, Mexico Agrociencia, vol. 37 Num 1 enero-febrero.
Mora-Flores, S.J., Matus-Gardea, A.J., and Martinez-Damian, A.M., (2002) Effects on
the Gross Value of Agriculture and Cattle Exports Due to Unit Change in their Prices
Agrociencia 36: 389-399.
Mundell, R., (1963) "Capital Mobility and Stabilization Policy under Fixed and Flexible
Exchange Rates," Canadian J. Econ. And Poli. Sci, Nov.. Or as adapted in his
International Economics, 18, pp. 250-262.
Munoz, M., and Odermatt, P., (1993) E1 Sistema Leche de Mexico en el Tratado
Trilateral de Libre Comercio (TLC) In: Juan Pablos (Editors) La Agricultura Mexicana
Frente al Tratado Trilateral de Libre Comercio 2nd edicion, Universidad Autonoma
Chapingo S.A, Mexico.
Mussa, M., (1979) Empirical Regularities in the Behaviour of Exchange Rates and
Theories of Foreing Exchange Market, in K Brunner and A. H. Melttzer eds. CarnegicRochester series on public policy vol 11 (North-Holland Amsterdams).
Mussa M., (1976) "Nominal Exchange Rate Regimes and the Behavior of Real Exchange
Rates: Evidence and implications," Carnegie Rochester Series on Public Policy, Vol. 25,
117-214.
Nadal, A., Wise, A.T., (2004) The Environmental Costs of Agricultural Trade
Liberalization: Mexico-U.S. Maize Trade Under NAFTA Working Group on
Development and Environment in the Americas Discussion Paper Number 4, June 2004.
Nelson, C.R., and Plosser C.I. (1982) Trends and Random Walks in Macroeconomic
Time Series: Some Evidence and Implications, Journal o f Monetary Economics, 10:
139-62.
Ng, S., Perron, P., (2001), Lag Length Selection and the Construction of Unit Root Tests
with Good Size and Power Econometrica 69, 1519-54.
279
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
Obstfeld, M., and Rogoff K., (1995) Exchange Rate Dynamics Redux Journal o f
Political Economy, 103(3), 624-60.
Orden D., and Fisher, L.A., (1991) Macroeconomic Policy and Agricultural Economics
Research American Journal of Agricultural Economics 73:1348-54.
Orden, D., and Fackler P., (1986) Identifying Monetary Impacts on Agricultural Prices
in VAR Models American Journal o f Agricultural Economics 71: 495-502.
Osterwald-Lenum, M., (1992) "A Note with Quantiles of the Asymptotic Distribution of
the ML Cointegration Rank Test Statistics" Oxford Bulletin o f Economics and Statistics,
54, 461-472.
Oya, K., Toda, H. Y., (1995) Dickey-Fuller Lagrange Multiplier and Combined Tests for
a Unit Root in Autoregressive Time Series Journal o f Time Series Analysis, Vol 19
3:326-347.
Pagan, A. R., and Hong, Y.S., (1991) Nonparametric Estimation and the Risk Premium
in Barnett, William A., James Powell, and George E. Tauchen eds., Nonparametric and
Semiparametric Methods in Econometrics and Statistics: Proceedings of the 5th
International Symposium in Economic Theory and Econometrics, Cambridge University
1991, 51-75.
Pagan, A.R., and Ullah , A., (1988) The Econometric Analysis of Models W hit Risk
Terms Journal o f Applied Econometric.
Peree, E. and Steinherr, A.,(1989) "Exchange Rate Uncertainty and Foreign Trade,"
European Economic Review 33 :1241-64.
Phillips, P.C.B., (1991a) Optimal Inference in Cointegrated System Econometrica, 59:
283-306.
Phillips, P.C.B., and Perron, P., (1988) Testing for Unit Roots in Time Series
Regression, Biometrica, 75 335-346.
Phillips, P.C.B., (1987) Time Series Regression with a Unit Root Econometrica, 55:
277-301.
Phillips, P.C.B., (1986) Understanding Spurious Regressions in Econometrics Journal
o f Economic, 33: 311-40.
Pick, D.H., (1990) Exchange Rate Risk and U.S. Agricultural Trade Flows. American
Journal o f Agricultural Economics 72: 694-700.
Qian, Y., and Varangis, P., (1994) Does Exchange Rate Volatility Hinder Export
Growth? Empirical Economics 19: 371-396.
280
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
281
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
Schwentesius, R. R., and Gomez, C.M., (1996) El Impacto del TLC Sobre las
Exportaciones de Hortalizas Frescas de Inviemo de Mexico a EUA Symposium TriNational NAFTA and Agriculture: Is the Experiment working? San Antonio Texas.
Schwert, G.W., (1989a) Why Does Stock Market Volatility Change Over Time?
Journal o Finance 44:115-1153.
Sercu, P., and Vanhulle C., (1992) Exchange Rate Volatility, International Trade, and
the Value of Exporting Firms. Journal o f Banking and Finance 16: 155-82.
Sexton, R.., Zhang, M., and Chalfant, J., (2003) Grocery Retailer Behaviour in the
Procurement and Sale of Perishable Fresh Produce Commodities Economic Research
Service/USDA.
Sims, C.A., (1980) Macroeconomics and Reality, Econometrica, 48: 1-48
Smith, E.C., (1999) Exchange Rate Variation, Commodity Price Variation and the
Implications for International Trade Journal o f International Money and Finance 18,
471-491.
Stock, J., and Watson, M., (1988) Testing for Common Tyrends Journal o f the
American Statistical Association. 83: 1097-1107
Stock, J., And Watson, M., (1987) Asymptotic Property of Least Square Estimators of
Cointegrating Vectors Econometrica 55: 1035-1056.
Taylor, M.A., and Taylor, P.M., (2004) The Purchasing Power Parity Debate NBER
Working Papers 10607, National Bureau of Economic Research Inc.
Taylor, P.M., (1995) The Economics of Exchange Rates Journal o f Economic
Literature, 33, 13-47, 1995
Thraen, S.C., Hawang, C.T., and Larson, W.D., (1992) Linking of U.S. Monetary Policy
and Exchange Rates to World Soybean Markets Agricultural Economic, 6 (1992) 365384.
Thursby, G.J., and Thursby, C.M., (1987) Bilateral Trade Flows, the Linder Hypothesis,
and Exchange Risk The Review o f Economics and Statistics, 488-495.
Thursby, G.J., (1985) "A Test Strategy for Discriminating Between Autocorrelation and
Misspecification in Regression Analysis: Reply" The Review o f Economics and Statistics,
67(1), February 177-78.
Thursby, G.J., and Thursby, C.M., (1984) "How Reliable Are Simple, Single Equation
Specifications of Import Demand?" The Review o f Economics and Statistics, 66(1),
February 1984, 120-28.
282
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
283
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
284
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
A p p e n d ix 1: Exchange Rate and Agricultural Trade: Exchange Rate Volatility was Included in the Model
Study
Anderson and
Garcia (1989)
Fabiosa (2002)
Major Findings
Remarks
Exports of
Pork and
Live swine
Quarterly data
from 1957 to
1994
OECD
Countries
Imports and
Exports of
Agricultural
And industry
products
1974-1995
10 developed
Countries
Bilateral
Trade of
Machinery
Chemicals
Agriculture
Other
Manufactures
-Negative effect
-Agricultural trade is more
susceptible to ER uncertainty
than the aggregate trade data
would suggest.
-No stationarity of
the data is taken
into account.
-Structural model
is preferred to
VAR.
- Did not take into
account nonstationary
properties o f the
data.
- It uses data from
the fixed and
floating periods.
- Did not take nonstationarity o f the
data used into
account.
- They use gravity
Model and Panel
data.
Ex. Rate
Treatment
Volatility
Measure
Sample, Data
Countries
OLS
Nominal
Exchange
Rate
Absolute %
change in
quarter to
quarter spot
ER.
Quarterly data
from
1974 to 1985
Bilateral
Trade of
Soybean
Maximum
Likelihood
Real
Exchange
Rate
GARCH
Monthly data
from 1994 to
Estimation
Method
2001
Commodities
U.S
Smith (1999)
Cho, et al
(2002)
OLS
Real
Exchange
Rate
NA
OLS
Real
Exchange
Rate
Standard
dev of the
first
differences
in log ex.
Rate
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Appendix 2: Exchange Rate and Agricultural Trade: Volatility of the ER was not Included in the Trade Model
Study
Arellano, et al
(1998)
Carter, Gray
& Furtan
(1990).
Chamber &
Just (1981)
Estimation
method
3-SLS
OLS
3SLS
Ex. Rate
Treatment
Real
Exchange
Rate
Exogenous
Real ER
-Real
ER
Cho, et al.,
(2002)
OLS
Real
Exchange
Rate
Collins
(1980)
Partial
equilibrium
world MK
Model
Structural
time-series
Real
Exchange
Rate
Kapombe and
Kolyer (1999)
Kost (1976)
Algebraic
manipulation
and derivation
Real
Exchange
Rate
Volatility
Measure
None
None
Quarterly data
From
1975 to 1988
Exchange Rate
Pass-through
Inputs and
commodities
None
Quarterly data
From
1969 to 1977
Exports of
Wheat, Com
Soybeans
1974-1995
10 developed
Countries
Bilateral trade of
Machinery
Chemicals
Agriculture
manufactures
Inports and
Exports of
Grains and
oilseeds
Exports of
Broilers
Standard dev
of the first
differences in
log ex. Rate
Short term
ER volatility
None
1970-1977
Annual
37 Countries
70-95
Quarterly
U.S.
None
Exogenous
N.A
Commodities
Sample, Data
Countries
Winter season
data from
1970 to 1994
M exico- U.S.
Caribean
None
Bilateral
Trade of
Melon
Bilateral
Trade of
Agricultural
commodities
Major Findings
Remarks
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Appendix 2: Exchange Rate and Agricultural Trade: Volatility of the ER was not Included in the Trade Model
Ex. Rate
Treatment
Nominal
Exchange
Rate
Time series
Econometric
Methods
Real
Exchange
Rate
N.A
Coleman and
Meilke (1988)
OLS
Nominal
Exchange
Rate
None
Coleman
(1984)
3SLS
Jennings, et al
(1991)
Babula et. al
(1995)
N>
oo
Volatility
Measure
Estimation
Method
Reduced
Form VAR
Study
None
Sample, Data
Countries
Quarterly data
From 1968 to
1987
Canada - U.S.
1978-1989
Montly
US-Row
Quarterly data
From
1972 to 1980
Canada - U.S.
None
-J
Guzel &
Kulshreshtha
(1995)
GAMS
Kaabia and
Gil, (2000)
Cointegration
SVAR Model
Malaga, et.al.
(2001).
3SLS
Nominal
Exchange
Rate
1972-1982
Commodities
Remarks
Lumber
- Non-stationary of the
data used was is taken into
account.
Com
Read Meat
Beef and Pork
Grains livestock,
dairy and
poultry.
Non-Agric
Products
Services
Real
Exchange
Rate
None
Real
Exchange
Rate
None
Quarterly data
From 1978 to
1995
Agricultural
Output
Real
Exchange
Rate
1974-1993
Mexico-U.S.
Winter season
Fresh vegetables
None
1984
No sample
period
Major Findings
- The volatility of
exchange rate is not
considered.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Appendix 2: Exchange Rate and Agricultural Trade: Volatility of the ER was not Included in the Trade Model
to
00
00
Sarker(1993)
Johansens
cointegration
analysis
Nominal
Exchange
Rate
Thraen et al,
(1991)
Structural
Vector
Autoregressio
n (SVR)
Nominal
Exchange
Rate
None.
None
U.S.-Canada
Lumber
Annual data
from 1965 to
1985. Brazil
Argentina
Japan, EC-12
Soymeal
Soy oil
Soybean
- The econometric
procedure used concerns a
small system that omits
other theoretical relevant
variables
- Linkages between
monetary policy ER and
world soybean market are
determined
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Appendix 3 Empirical Literature on Exchange Rate Effects on Trade (Exchange Rate Volatility was Included)
Volatility
Measure
Short Term
Volatility
Moving sample
st. dev. of the
growth rates of
RER
Short Term
Moving st. dev
of the growth
rate of RER
Estimation
Method
Vector
Autoregres
sion (VAR)
Ex. Rate
Treatment
Real
Exchange
Rate
Koray and
Lastrapes
(1989)
Vector
autoregresi
on (VAR)
Nominal
Exchange
Rate
Kroner and
Lastrapes
(1989
Cointegrati
on
Nominal
Exchange
Rate
GARCH
Short Term
Volatility
Lastrapes and
Koray (1990)
Vector
autoregresi
on (VAR)
Real
Exchange
Rate
Short Term
Moving st. dev.
Of RER
OLS
Real
Exchange
Rate
Long Term
ARCH to
generate ER
volatility
Study
Chowdhury
(1993)
McKenzie
(1998)
Qian and
Varangis
(1994)
SUR
ARCH-M
Alkthar and
Spence-Hilton
(1991)
Assery and
Peel (1991)
OLS
Error
Correction
Model
Nominal
Exchange
Rate
ARCH Model
Nominal
Exchange
Rate
Real
Exchange
Rate
Short Term
ARCH-in-mean
Short Term
Variability of an
effective nom.
ER index
Squared residual
from the
ARIMA process
Sample, Data
Countries
Quarterly data
From 1973 to 1990
G7 countries
Commodities
Major Findings
Remarks
Aggregate
- Non-stationarity in data
used is taken into account.
- Economic agents are
assumed to be risk averse.
Monthly data
1959-1972
1973-1985
Developed
Countries
Monthly data
1973-1990
Developed
Countries
Aggregate
Aggregate
1973-1987
DC(US trade)
Montly
Aggregate
- Non-stationarity in data
used is taken into account
-The model restrict the
variance to be the same as
that generated by the data.
-Reduced form model, it
cannot distinguish between
structural hypothesis.
1969-1995
Quarterly
Australia and DC.
Aggregate
and sectoral
trade data
73-90
Aggregate
Six developed
countries
1974-1981
Quartely
Us-Ger
Aggregate
1972-1987
Quarterly
Five DC
Aggregate
- N on-stationarity in data
used is taken into account
- No structural Model.
- Both ER and ER volatility
enter into the equations.
- ARCH-in-mean, it provides
more efficient coefficient
estimates avoids spurious
regression.
-Explicitly considered the
impact of risk on volume
through prices.
- Non-stationarity in data
used is taken into account
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Appendix 3 Empirical Literature on Exchange Rate Effects on Trade (Exchange Rate Volatility was Included)
Study
Bailey et.al,
(1986)
Estimation
Method
OLS
Commodities
Major Findings
Remarks
Ex. Rate
Treatment
Nominal and
Real
Exchange
Rate
Volatility
Measure
Short Term
Absolute%
changes and st.
deviation
Sample, Data
Countries
Quarterly
1962-1974
1975-1985
Nominal
Exchange
Rate
Exogenous
Short Term
St. dev, of
weekly rates of
changes. FER
1976-1984
DCs (intra EMS)
Quarterly
A, B & S
Manufactures
- Non-stationarity in data
used was not taken into
account
Export equations allowing
for uncertainty are estimated
He argued that third country
ER factors must be included
Aggregate
Bini-Smaghi
(1991)
OLS
Caballero and
Corbo (1989)
OLS and
IV
Real
Exchange
Rate
Exogenous
Long Term
Annual St. desv.
of the RER
Quarterly data
6 developing
countries
Aggregate
Cushman
(1986)
OLS
Real
Exchange
Rate
Short Term
Moving four
quarter st. dev.
Of changes in
ER
1965-1977
1973-1983
DC(US exports
Aggregate
Cushman
(1988a)
OLS
Real
Exchange
Rate
1973-1983
DC(US exports)
Aggregate
De Grauwe
(1987)
SURE
Both
Exogenous
1960-1969
1973-1984
EMS Countries
Aggregate
Gagnon
(1993)
OLS
Theoretical
Model
OLS
US- 6 ind
countries
1960-1971
1972-1988.
1965-1975
(US - Ger)
Aggregate
Trade
Hooper and
Kohlhagen
(1978)
Real
Exchange
Rate
Exogenous
Nominal
Exchange
Endogenos
Aggregate
bilateral and
multilateral
Short Times
Bilateral
Exchange Risk
Long Term
Variance of the
yearly% changes
of the bilateral
RER
Uncondditional
variance
Short Term
Variance of the
future spot ex.
Rate.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Literature on Exchange Rate Effects on Trade (Exchange Rate Volatility was Included)
Study
to
so
Estimation
Method
Ex. Rate
Treatment
Klein (1990)
Pooled
Regression
Real
Exchange
Rate
Klein (1990)
Pooled
Regression
Real
Exchange
Rate
Konno and
Fukushige
(2002)
McKenzie
and Brooks
(1997)
Dynamic
OLS
Real
Exchange
Rate
Nominal
Exchange
Rate
OLS
Perre and
Steinher
(1989)
OLS
Nominal
Exchange
Rate
Thursby and
Thursby
(1987)
OLS
Nominal
Exchange
Rate
Chowdhury
(1993)
Multivariat
e Error
Correction
Model
Koray and
Lastrapes
(1989)
VAR
Real
Exchange
Rate
Nominal
Exchange
Rate
Volatility
Measure
Short Term
Volatility
St.dev. of the
montly %
change in RER.
Short Term
Volatility
St.dev. of the
montly %
change in RER.
Long Term,
Volatility
Short Term
ARCH
Long term
function of the
magnitude of
past movements
in nominal ER
Short Term
Variance of the
Spot ER
Moving sample
st. deviation of
the growth rates
of real ER
Short Term
Moving st. dev.
Of the growth
rate of RER
Sample, Data
Countries
Commodities
Major Findings
Remarks
78-86 Montly(US
exports)
9 Sectors
78-86 Montly(US
exports)
9 Sectors
1981-1994
Quarterly
Aggregate
1973-1992
Montly
US-Germany
Aggregate
1960-1985
Five developed
countries
Aggregate
bilateral
1974-1982
D C s (17)
Aggregate
Quarterly data
From 1973 to 1990
G7 countries
Aggregate
- Non-stationary in data
used is taken into account
Monthly data
1959-1972
1973-1985
Developed
Countries
Aggregate
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Appendix 3 Empirical Literature on Exchange Rate Effects on Trade (Exchange Rate Volatility Was Included in the Model)
Estimation
Method
Co
integration
Ex. Rate
Treatment
Nominal
Exchange
Rate
Volatility
Measure
Short Term
GARCH
Lastrapes and
Koray (1990)
VAR
Real
Exchange
Rate
Short term
Moving st.
deviation of real
ER.
McKenzie
(1998)
Co
integration
Real
Exchange
Rate
Long Term
ARCH
Qian and
Varangis
(1994)
SUR
Nominal
Exchange
Rate
Short Term
ARCH
Study
Kroner and
Lastrapes
(1989
Sample, Data
Countries
Monthly data
1973-1990
Developed
Countries
1973-1987
DC(US trade)
Montly
Commodities
Major Findings
Remarks
Aggregate
Aggregate
1969-1995
Quarterly
Australia and DC.
Aggregate
and Sectoral
trade data
1973-1990
Six developed
countries
A&B
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Appendix 4 Empirical Literature on Mexican Agricultural Trade (Exchange Rate Volatility was not Included in the Model)
Study
Estimation
method
Ex. Rate
Treatment
Volatility
Measure
Diaz, (2002)
Co-integration
Error
Correction
Model
Real ex rate
None
Real
Exchange
Rate
Real
Exchange
Rate
Nominal
Exchange
Rate
None
VAR
Real
Exchange
Rate
None
Arellano, et al
(1998)
3-SLS
Real
Exchange
Rate
None
Malaga, et.al.
(2001).
3SLS
Real
Exchange
Rate
None
Mendoza
(1996)
OLS
Mora-Flores, et.,
(2002)
2-SLS
Salas (1980)
OLS
Varella, (20002)
Konno and
Fukushige
(2002)
Dynamic
OLS
Real
Exchange
Rate
None
None
Long Term,
Volatility
Sample, Data
Countries
Yearly data
80-89 and
90-2000
Mexico
Yearly data
1970-1994
Mexico
Yearly data
1988-1986
Mexico
Yearly data
from
1961-1979
Quarterly data
1980-2000
Mexico
Winter season
data from
1970 to 1994
Mexico- U.S.
Caribean
Winter season
data
1974-1993
Mexico-U.S.
Quarterly data
1981-1994
Commodities
Major Findings
Remarks
Aggregate Trade
Flows
Consumption,
capital and in
term goods.
Manufactures
Tradables Non
tradables
Melon
Fresh vegetables
Aggregate
APPENDICES 5-20
Determination of Optimal Lag Specification for
Selected VAR Models
294
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
M ODEL VI
NLAG S
Log
L ikelihood
-36.841
-37.314
-37.314
-37.841
-37.841
-38.194
-38.194
-38.828
N um ber
V ariables
N*
M LTest
158
58.559
(df=7,
5% )
14.02
Reject Lag 2
151
58.059
14.02
Reject Lag 3
144
33.813
14.02
Reject Lag 4
137
52.050
14.02
Reject Lag 5
130
19.394
14.02
Reject Lag 6
123
21.033
14.02
Reject Lag 7
116
27.967
14.02
Reject Lag 8
109
11.508
14.02
Accept Lag 9
102
6.758
14.02
**
95
-67.053
14.02
L ags
Two-Three
Three-Four
Four-Five
Five-Six
Six-Seven
Seven-Eigth
Eigth-Nine
Nine-en
Ten-Eleven
ElevenTwelve
-38.828
-39.114
-39.114
-39.503
-39.503
-40.202
-40.202
10
-40.645
10
-40.645
11
-41.208
11
-41.208
T est
R esult
LRm X \
34
41
48
55
62
69
76
83
90
295
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
NLAGS
Log
Likelihood
-34.423
N um ber
Variables
N*
M L-Test
158
45.688
14.02
Reject Lag 2
151
42.971
14.02
Reject Lag 3
144
44.432
14.02
Reject Lag 4
137
55.685
14.02
Reject Lag 5
130
19.602
14.02
Reject Lag 6
123
17.630
14.02
Reject Lag 7
116
30.675
14.02
Reject Lag 8
109
8.045
14.02
Accept Lag 9
102
7.534
14.02
**
95
-2.430
14.02
LRm x \
Test Result
(df=7) 5%
Two-Three
Three-Four
Four-Five
Five-Six
Six-Seven
Seven-Eigth
Eigth-Nine
Nine-en
Ten-Eleven
Eleven-Twelve
-34.791
-34.791
-35.182
-35.182
-35.645
-35.645
-36.324
-36.324
-36.612
-36.612
-36.938
-36.938
-37.705
-37.705
10
-38.015
10
-38.015
11
-38.642
11
-38.642
12
-39.857
34
41
48
55
62
69
76
83
90
97
N* N et Number of Observations
** Denotes that once an optimal lag is achieved, no further testing is conducted.
296
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
M ODEL VN
Lags
NLAGS
Log
Likelihood
N um ber
Variables
-37.585
Three-Four
3
3
4
-37.991
-37.991
-38.439
53
-38.439
-39.065
-39.065
-39.972
61
-39.972
-40.423
69
-40.423
-40.907
77
Five-Six
Six-Seven
Seven-Eigth
Eigth-Nine
Nine-en
Ten-Eleven
ML-Test
LRm
Test Result
(df=7) 5%
Two-Three
Four-Five
N*
7
7
8
8
155
47.917
14.02
Reject Lag 2
147
45.669
14.02
Reject Lag 3
139
53.860
14.02
Reject Lag 4
131
63.517
14.02
Reject Lag 5
123
24.330
14.02
Reject Lag 6
115
18.375
14.02
Reject Lag 7
107
20.403
14.02
Reject Lag 8
99
3.599
14.02
Accept Lag 9
91
-332.696
14.02
**
83
-34.096
14.02
37
45
-40.907
9
9
10
-41.834
-41.834
85
-42.434
93
10
11
-42.434
-9.164
101
11
-9.164
-10.476
109
Eleven-Twelve
12
297
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
NLAGS
2
-38.098
3
3
4
-38.745
-38.745
-39.160
4
5
-39.160
5
6
6
7
7
Eigth-Nine
8
8
Ten-Eleven
Number
Variables
N*
ML-Test
LRm x ]
Test Result
(df=7) 5%
Seven-Eigth
Nine-en
Log
Likelihood
-39.651
-39.651
-40.161
-40.161
-6.428
-6.428
-6.740
-7.318
-7.318
-7.988
-7.988
-8.614
11
12
-8.614
80.231
14.02
Reject lag 2
151
45.684
14.02
Reject lag 3
144
47.156
14.02
Reject lag 4
137
41.792
14.02
Reject lag 5
130
2293.819
14.02
Reject lag 6
123
16.807
14.02
Reject lag 7
116
23.150
14.02
Reject lag 8
109
17.413
14.02
Reject lag 9
102
7.507
14.02
Accept lag 10
95
-2.866
14.02
**
41
48
55
62
69
-6.740
9
9
10
10
11
158
34
76
83
90
Eleven-Twelve
-10.047
97
298
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
NLAGS
Log
Likelihood
-33.618
-33.966
3
4
4
5
5
6
6
-33.966
-35.928
-35.928
-36.213
-34.597
Num ber
Variables
-36.213
-36.474
-36.474
10
-37.245
10
-37.245
11
-37.803
11
-37.803
12
-4.197
Test Result
158
43.088
14.02
Reject Lag 2
151
69.469
14.02
Reject lag 3
144
52.785
14.02
Reject lag 4
137
27.610
14.02
Reject lag 5
130
30.220
14.02
Reject lag 6
123
15.401
14.02
Reject lag 7
116
14.411
14.02
Reject lag 8
109
20.064
14.02
Reject lag 9
102
6.693
14.02
Accept Lag 10
95
-67.213
14.02
48
55
-35.484
(df=7) 5%
41
-35.147
-35.484
M L-Test
34
-34.597
-35.147
LRm x ]
N*
62
69
76
83
90
ElevenTwelve
97
299
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
NLAGS
Log
Likelihood
Num ber
Variables
ML-Test
N*
LRm x l
Test Result
(df=7)
2
-34.260
3
3
4
-34.641
-34.641
34
-35.047
41
5
5
-35.047
-35.437
-35.437
48
5
6
6
7
7
8
8
9
9
10
10
11
-35.806
-35.806
55
-36.226
62
-36.226
-36.744
-36.744
69
-37.335
-37.335
-38.060
-38.060
76
-38.649
90
158
47.252
14.02
Reject Lag 2
151
44.693
14.02
Reject lag 3
144
37.392
14.02
Reject lag 4
137
30.252
14.02
Reject lag 5
130
28.573
14.02
Reject lag 6
123
28.008
14.02
Reject lag 7
116
23.602
14.02
Reject lag 8
109
18.854
14.02
Reject lag 9
102
7.077
14.02
Accept Lag 10
-1.220
14.02
83
ElevenTwelve
11
12
-38.649
-39.259
95
97
300
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
Number
Variables
NLAGS
Log
Likelihood
-6.083
3
3
4
4
-6.496
34.000
-6.496
-7.302
41.000
5
5
6
-7.302
-7.854
-7.854
48.000
-8.209
55.000
-8.209
8.310
8.310
62.000
Seven-Eigth
6
7
7
Eigth-Nine
8
8
8.700
8.700
9.130
9.130
Six-Seven
Nine-en
Ten-Eleven
9
10
10
11
LRm x ]
N*
M l Test
(df=7) 5%
Test Result
158
51.105
14.02
Reject Lag 2
151
88.675
14.02
Reject lag 3
144
52.975
14.02
Reject lag 4
137
29.184
14.02
Reject lag 5
130
23.319
14.02
Reject lag 6
123
21.060
14.02
Reject lag 7
116
17.200
14.02
Reject lag 8
109
14.560
14.02
Reject Lag 9
102
6.960
14.02
Accept Lag 10
95
-1.360
14.02
69.000
76.000
9.690
9.690
10.270
83.000
90.000
10.270
10.950
97
ElevenTwelve
11
12
301
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
Num ber
Variables
NLAGS
Log
Likelihood
-35.822
-36.133
3
4
-36.133
-36.688
41
4
5
48
Five-Six
Six-Seven
6
6
7
-37.501
-37.501
-37.981
SevenEigth
-37.981
8
8
-38.302
-38.302
69
Eigth-Nine
-4.159
76
9
10
-4.159
-4.886
83
10
11
-4.886
-5.464
90
11
-5.464
TenEleven
M L -Test
(df=7) 5%
Test Result
158
38.493
14.020
Reject Lag 2
151
61.061
14.020
Reject lag 3
144
52.192
14.020
Reject lag 4
137
22.069
14.020
Reject lag 5
130
32.656
14.020
Reject lag 6
123
17.355
14.020
Reject lag 7
116
1365.735
14.020
Reject lag 8
109
18.908
14.020
Reject lag 9
102
6.938
14.020
Accept Lag 10
95
-1.281
14.02
34
-36.688
-37.231
-37.231
Nine-en
LRm X d
N*
55
62
ElevenTwelve
-6.105
97
302
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
NLAGS
Log
Likelihood
Two-Three
-34.260
-34.641
-34.641
-35.047
-35.047
-35.437
Three-Four
Four-Five
Five-Six
-35.437
-35.806
Six-Seven
-35.806
-36.226
Seven-Eigth
-36.226
-36.744
-36.744
-37.335
Eigth-Nine
Nine-en
Ten-Eleven
-37.335
10
-38.060
10
-38.060
11
-38.649
11
-38.649
12
-39.259
Num ber
Variables
LRm x \
Test Result
N*
ML Test
(df=7) 5%
158
47.252
14.02
Reject Lag 2
151
44.693
14.02
Reject lag 3
144
37.392
14.02
Reject lag 4
137
30.252
14.02
Reject lag 5
130
28.573
14.02
Reject lag 6
123
28.008
14.02
Reject lag 7
116
23.602
14.02
Reject lag 8
109
18.854
14.02
Reject Lag 9
102
7.077
14.02
Accept Lag 10
95
-1.220
14.02
34.000
41.000
48.000
55.000
62.000
69.000
76.000
83.000
90.000
ElevenTwelve
97.000
303
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
NLAGS
Log
Number
Likelihood Variables
-31.946
-32.315
-32.315
-32.671
-32.671
-33.094
-33.094
-33.507
-33.507
-33.885
-33.885
-34.483
-34.483
-35.127
-35.127
10
-35.692
10
-35.692
11
-36.568
11
-36.568
12
-37.305
LRm X l
N*
Test Result
M L Test (df=7) 5%
158
45.711
14.02
Reject Lag 2
151
39.154
14.02
Reject lag 3
144
40.599
14.02
Reject lag 4
137
33.873
14.02
Reject lag 5
130
25.679
14.02
Reject lag 6
123
32.292
14.02
Reject lag 7
116
25.775
14.02
Reject lag 8
109
14.686
14.02
Reject lag 9
102
10.516
14.02
Accept Lag 10
95
-1.473
14.02
34.000
41.000
48.000
55.000
62.000
69.000
76.000
83.000
90.000
Eleven-Twelve
97.000
304
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
NLAGS
Log
Num ber
Likelihood Variables
-7.266
-7.628
-7.628
-8.215
-8.215
-8.560
-8.560
-8.962
-8.962
-9.447
-9.447
-9.816
-9.816
-10.367
-10.367
10
-10.989
10
-10.989
11
-11.568
11
-11.568
12
-12.557
LRm x \
N*
158
44.988
14.020
Reject Lag 2
151
64.485
14.020
Reject lag 3
144
33.201
14.020
Reject lag 4
137
32.912
14.020
Reject lag 5
130
32.968
14.020
Reject lag 6
123
19.961
14.020
Reject lag 7
116
22.035
14.020
Reject lag 8
109
16.164
14.020
Reject lag 9
102
6.955
14.020
Accept Lag 10
95
-1.978
14.02
34
41
48
55
62
69
76
83
90
Eleven-Twelve
97
305
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
NLAG
Two-Three
Three-Four
3
3
Number
Log
Likelihood Variables
-36.491
-36.823
-36.823
-37.142
Four-Five
-37.142
5
5
-37.522
-37.522
48
Five-Six
6
6
7
-38.015
-38.015
55
-38.429
-38.429
-4.438
-4.438
62
Six-Seven
Seven-Eigth
Eigth-Nine
8
8
Nine-en
Ten-Eleven
N*
ML
Test
LRm Xd
(df=7) 5%
Test Result
158
41.145
14.02
Reject Lag 2
151
35.032
14.02
Reject lag 3
144
36.564
14.02
Reject lag 4
137
40.381
14.02
Reject lag 5
130
28.182
14.02
Reject lag 6
123
1835.541
14.02
Reject lag 7
116
1398.620
14.02
Reject lag 8
109
881.0
14.02
Reject lag 9
102
7.580
14.02
Accept Lag 10
95
-1.489
14.02
34
41
69
9
9
10
10
11
-39.403
-39.403
76
-5.519
83
-6.151
90
11
12
-6.151
-6.896
97
-5.519
ElevenTwelve
306
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
M ODEL V I
Lags
Two-Three
NLAGS
2
3
Three-Four
Four-Five
3
4
4
5
Five-Six
Six-Seven
5
6
6
7
Number
Log
Likelihood Variables
-28.165
-28.567
-29.055
-29.055
41.000
-29.409
-29.409
-29.871
48.000
Eigth-Nine
8
8
9
-30.951
-30.951
-31.684
9
10
-31.684
-32.231
10
11
-32.231
-33.003
11
-33.003
12
-33.648
Nine-en
Ten-Eleven
M L Test
(df=7) 5%
Test Result
158
49.813
14.02
Reject Lag 2
151
53.704
14.02
Reject lag 3
144
33.972
14.02
Reject lag 4
137
37.940
14.02
Reject lag 5
130
37.199
14.02
Reject lag 6
123
29
14.02
Reject lag 7
116
29.331
14.02
Reject lag 8
109
14.218
14.02
Reject Lag 9
102
9.269
14.02
Accept Lag 10
95
-1.290
14.02
55.000
-29.871
Seven-Eigth
N*
34.000
-28.567
-30.419
-30.419
LRm
62.000
69.000
76.000
83.000
90.000
ElevenTwelve
97.000
307
R ep ro d u ced with p erm ission o f the copyright ow ner. Further reproduction prohibited w ithout perm ission.
NLAGS
2
3
Three-Four
Four-Five
Five-Six
Six-Seven
SevenEigth
Eigth-Nine
Nine-en
Ten-Eleven
Log
Likelihood
Number
Variables
-25.915
-26.307
34
-26.307
4
4
5
5
6
6
-26.747
-26.747
-27.068
-27.068
-27.447
-27.447
-28.015
7
8
-28.015
-28.578
8
9
9
-28.578
-29.314
-29.314
10
10
11
-29.902
-29.902
-30.756
11
-30.756
-31.492
LRm x ]
N*
M L Test
(df=7) 5%
Test Result
158
48.619
14.02
Reject Lag 2
151
48.342
14.02
Reject lag 3
144
30.795
14.02
Reject lag 4
137
31.115
14.02
Reject lag 5
130
38.616
14.020
Reject lag 6
123
30.371
14.020
Reject lag 7
116
29.457
14.020
Reject lag 8
109
15.302
14.020
Reject lag 9
102
10.247
14.020
Accept Lag 10
95
-1.471
14.020
41
48
55
62
69
76
83
90
ElevenTwelve
12
97
N* N et Number of Observations
** Denotes that once an optimal lag is achieved, no further testing is conducted.
308
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
NLAGS
2
3
3
4
4
5
5
6
6
7
7
8
8
9
9
10
10
11
Number
Log
Likelihood Variables
-35.111
-35.620
-35.620
-36.299
-36.299
-36.789
-36.789
-37.292
-37.292
-37.852
-37.852
-38.502
-38.502
-4.621
LRm x l
N*
M L - Test
(df=7) 5%
Test Result
158
63.174
14.02
Reject Lag 2
151
74.613
14.02
Reject lag 3
144
47.030
14.02
Reject lag 4
137
41.300
14.02
Reject lag 5
130
38.089
14.02
Reject lag 6
123
35.090
14.02
Reject lag 7
116
1355.2
14.02
Reject lag 8
109
12.529
14.02
Accept Lag 9
102
7.620
14.02
95
-2.279
14.02
34
41
48
55
62
69
76
-4.621
-5.103
-5.103
83
-5.738
90
ElevenTwelve
11
-5.738
12
-6.878
97
309
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
NLAGS
Log
Likelihood
Two-Three
-30.421
-30.814
-30.814
34
Three-Four
3
3
4
4
-31.215
-31.215
41
Four-Five
-31.576
-31.576
48
Five-Six
5
5
-31.886
-31.886
-32.359
55
Six-Seven
6
6
7
8
8
-32.949
-32.949
69
10
11
-33.718
-33.718
-34.073
-34.073
-34.962
76
90
11
12
-34.962
-35.578
97
MODEL VN
Lags
Seven-Eigth
Eigth-Nine
Nine-en
Ten-Eleven
9
9
10
Num ber
Variables
LRm x ]
Test Results
N*
M L Test
(df=7)
158
48.741
14.02
Reject Lag 2
151
44.103
14.02
Reject lag 3
144
34.650
14.02
Reject lag 4
137
25.444
14.02
Reject lag 5
130
32.191
14.02
Reject lag 6
123
31.861
14.02
Reject lag 7
116
30.737
14.02
Reject lag 8
109
9.235
14.02
Accept Lag 9
102
10.671
14.02
95
-1.232
14.02
62
-32.359
83
ElevenTwelve
310
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
APPENDICES 21-36
Maximum Likelihood Cointegration Results
311
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
E ig en v a lu es
0.271292
0.22766
0.178369
0.142208
0.122468
0.074314
0.006916
E ig en v ecto rs
USY
-1.0085
-0.5919
-1.1695
4.3220
-0.0067
2.5105
0.1807
ER
-1.4250
-1.2222
-1.0243
-0.0034
0.6317
-2.7650
0.7689
VI
0.3620
-1.9924
-0.6794
-3.3103
0.1215
2.6320
0.8981
Pt
0.6879
1.6578
-2.1761
-1.3886
0.3383
0.2091
-1.3216
Pea
2.0680
-2.1329
-0.1719
2.1181
0.1195
-0.9394
-1.1540
W1
-1.4249
-0.6637
1.0813
-0.7911
0.4424
1.0963
-2.2798
Qt
1.0000
1.0000
1.0000
1.0000
1.0000
1.0000
1.0000
W eig h ts
USY
0.00023
0.00041
-0.00026
-0.00027
0.00074
0.0002
-0.0002
ER
0.01159
-0.00189
-0.00229
-0.00605
-0.00417
-0.0008
-0.0006
VI
0.07999
-0.06329
-0.06878
-0.04569
-0.02019
0.0423
0.0053
Pt
0.00678
0.06511
-0.05019
0.03072
-0.01794
0.0002
-0.0004
Pea
-0.03253
0.00996
-0.03879
-0.01720
-0.01193
-0.0154
-0.0000
W1
0.01161
-0.01060
-0.00455
-0.00210
-0.00408
-0.0026
-0.0016
Qt
-0.02733
-0.00570
0.01222
-0.00658
-0.04890
0.0187
-0.0037
Trace
Statistic
Trace
(0.95)
MAX (X.)
M AX (X)
(0.95)
*1
II
o
Null
Hypothesis
207.38*
150.40
57.60*
50.51
r< 1
149.78*
117.49
47.02*
44.37
r<2
102.77*
88.59
35.76
38.22
r<3
63.01
63.66
27.92
31.99
r<4
39.09
42.70
23.78
25.68
r<5
15.32
25.64
14.05
19.21
r<6
1.26
12.34
1.26
12.34
312
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
0.314072
0.228075
0.171309
0.159458
0.11313
0.08900
0.02026
E ig en v ecto rs
USY
-0.5957
-1.2060
0.8849
-4.5951
-0.7619
3.2666
-0.6717
ER
-1.8126
-0.4431
0.1668
-8.9773
0.2224
-3.5346
0.8252
V2
0.4796
-0.5566
-0.7489
-6.0675
0.4413
4.2511
0.8607
W1
-1.3858
0.6769
-0.0568
-1.2528
0.8627
0.6483
-2.4053
Pea
-2.4410
0.4417
0.0248
-5.7788
-0.0811
-1.7996
-1.0284
Pt
0.5323
-5.3284
0.0466
1.9700
0.3984
-0.7687
-0.2096
Qc
1.0000
1.0000
1.0000
1.0000
1.0000
1.0000
1.0000
W eigh ts
USY
0.00027
-0.00046
-0.00009
0.00029
-0.00073
-0.00023
0.00041
ER
-0.01087
-0.00113
0.00788
-0.00506
-0.00118
0.00150
0.00109
V2
-0.05619
-0.19450
0.10891
-0.28325
0.00974
-0.18688
-0.05002
W1
-0.01223
0.00163
0.00209
-0.01177
-0.00029
0.00276
0.00250
Pea
0.04893
-0.00962
0.02239
-0.02092
-0.00452
0.02318
0.00046
Pt
0.05465
-0.07149
0.00293
0.00373
0.02184
0.01968
0.00376
Qc
0.01444
0.01967
0.03146
-0.00781
0.03945
-0.01271
0.00853
Trace
Statistic
Trace
(0.95)
MAX
(X)
M AX (X)
(0.95)
II
224.05*
150.40
68.61*
50.51
r< 1
155.44*
117.49
47.11*
44.37
r<2
108.33*
88.59
34.19
38.22
r<3
74.13*
63.66
31.61
31.99
r<4
42.51
42.70
21.85
25.68
r <5
20.66
25.64
16.93
19.21
3.73
12.34
* Significance at 5 percent level (MacKinnon, 1999)
3.73
12.34
Null hypothesis
r<6
313
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
0.2400
0.1805
0.1340
0.1060
0.1013
0.0574
0.0210
E ig en v ec to rs
USY
0.4115
-0.2463
-0.6911
1.8000
1.2848
-4.0166
-0.4092
ER
1.0963
-0.5888
4.4093
0.3389
0.5607
1.6184
-0.6003
VG
-0.4492
-0.4604
-3.1409
1.2229
1.0981
3.1982
0.2617
W1
0.4909
-0.5053
0.3450
-0.8676
-0.1159
-0.9232
2.6691
Pt
-2.4009
0.2850
2.2059
3.1834
-0.1577
-0.0887
0.5835
Pea
-1.6991
-0.8440
1.0792
-3.8290
1.1606
-0.3269
-0.0205
Qt
1.0000
1.0000
1.0000
1.0000
1.0000
1.0000
1.0000
W eig h ts
USY
0.00048
-0.00058
-0.00066
0.00031
-0.00001
-0.00070
-0.00021
ER
-0.00502
-0.00501
0.00418
-0.00147
0.00428
-0.00355
0.00126
VG
-0.00044
-0.00920
0.00009
0.00080
0.00243
0.00057
-0.00063
W1
-0.00995
-0.00791
0.00829
0.00263
0.00072
-0.00416
0.00000
Pt
0.04810
-0.01693
0.02731
-0.03221
-0.00068
0.00373
-0.00767
Pea
0.10529
-0.02008
0.03870
0.02992
0.01274
0.00086
0.01151
Qt
0.02564
0.04707
0.04380
-0.00188
0.01092
-0.00894
0.00101
Trace
(0.95)
MAX
II
Null hypothesis
166.80*
150.40
49.95*
50.51
r< 1
116.85*
117.49
36.22
44.37
*1
Trace
Statistic
a)
MAX(X)
(0.95)
r<2
80.63
88.59
26.19
38.22
r<3
54.44
63.66
20.40
31.99
r<4
34.04
42.70
19.43
25.68
r<5
14.61
25.64
10.75
19.21
r<6
3.85
12.34
3.85
12.34
314
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
E ig en v a lu es
0.2916
0.2325
0.1786
0.1234
0.0984
0.0184
0.0160
E ig en v ecto rs
USY
-0.2204
-0.4611
1.1117
-0.5487
2.2797
-0.6308
-1.8070
ER
-0.6020
-1.0109
0.6143
-0.0140
-3.6494
-0.0701
1.1895
VN
0.2290
0.3944
-0.0229
0.3579
-3.4483
2.4819
-2.0620
Pt
1.0104
0.9522
-0.1465
0.8374
1.1661
-1.1215
-0.5534
Pea
1.8788
1.2893
0.3737
-0.3691
-3.6476
-1.9575
0.0285
W1
-0.6695
-0.4636
-0.1434
0.7855
-1.1315
-2.5702
-0.9714
Qt
1.0000
1.0000
1.0000
1.0000
1.0000
1.0000
1.0000
W eights
USY
-0.00056
-0.00010
0.00018
0.00053
0.00054
-0.00023
-0.00029
ER
-0.00005
0.00494
-0.00753
0.00299
-0.00037
-0.00176
0.00066
YG
0.00977
0.01970
-0.01950
-0.00861
0.00010
0.01124
-0.00754
W1
-0.00716
0.05219
0.01924
-0.03046
0.00703
-0.00400
0.00337
Pt
-0.06925
0.00500
-0.00098
0.00504
-0.01279
0.00201
0.00459
Pf
0.00070
0.00841
-0.00470
0.00446
-0.00583
-0.00223
-0.00082
Qt
-0.01826
-0.02681
-0.03591
-0.03739
-0.01402
-0.00649
-0.00259
Trace
(0.96)
M AX
(X)
MAX(X)
(0.95)
*1
11
o
Null hypothesis
194.751*
150.41
62.389*
50.51
r<1
132.361*
117.49
47.890*
44.37
r<2
84.471
88.59
36.603
38.22
r <3
48.868
63.66
23.836
31.99
r<4
25.032
42.71
18.742
25.68
r<5
6.290
25.64
3.367
19.21
r<6
2.922
12.34
2.922
12.34
315
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
E igenvalues
0.2670
0.2305
0.1379
0.1083
0.0527
0.0263
0.0035
E ig en v ecto rs
MXY
ER
VI
INV
Pc
Pcd
Qc
-1.572
-0.751
1.691
0.154
0.584
-0.554
0.072
-0.171
1.000
0.184
0.776
-2.607
1.155
-47.390
-16.054
-0.136
-0.001
0.085
0.341
5.251
2.214
-1.286
2.266
-32.197
57.211
-95.224
-25.792
0.543
-0.370
-0.514
1.000
1.000
1.000
0.687
1.000
16.437
10.419
0.971
11.198
-16.060
-8.455
5.556
-6.554
-6.660
-0.947
-4.149
0.792
1.000
1.000
0.0005
0.0010
0.0028
0.0003
-0.0003
-0.0041
0.0095
-0.0020
-0.0027
0.0036
-0.0010
0.0001
0.0490
0.0001
W eig h ts
MXY
ER
VI
INV
Pc
Pcd
Qc
0.0012
-0.0003
0.1484
0.0000
-0.0002
-0.0432
0.0363
0.0072
-0.0101
-0.0390
0.0150
0.0031
0.0173
0.1913
0.0006
0.0013
-0.0012
0.0001
0.0014
0.1545
0.0084
0.0068
-0.0791
0.0033
0.0100
0.0007
-0.0206
-0.0197
0.0027
0.0008
0.0405
0.0521
-0.0791
0.0541
0.0006
Trace
(0.95)
Max
M ax X
(0.95)
II
o
**
Null
Hypothesis
166.504*
150.4
56.211*
50.51
r< 1
110.293
117.49
47.429*
44.37
r< 2
62.864
88.59
26.852
38.22
r< 3
36.012
63.66
20.740
31.99
r< 4
15.272
42.7
9.803
25.68
r< 5
5.468
25.64
4.828
19.21
0.641
12.34
* Significance at 5 percent level (MacKinnon, 1999)
0,641
12.34
r<6
316
R ep ro d u ced with p erm ission o f the copyright ow ner. Further reproduction prohibited w ithout perm ission.
0.2816
0.2330
0.1343
0.0903
0.0540
0.0280
0.0005
E ig en v ecto rs
MXY
ER
V2
INV
Pc
Pcd
Qc
-1.5078
0.0489
0.8105
-1.7068
-0.1465
3.8035
-5.6579
-0.6609
0.0425
4.4622
0.3479
-0.6384
2.9207
6.1278
1.1106
-0.7290
-0.6720
-1.4157
1.0618
0.8854
3.6993
-0.1382
0.1875
-10.3210
0.0909
-0.3575
1.5111
2.1693
0.2457
0.5568
0.5275
-3.2894
-0.5670
-1.7122
2.3001
-1.3745
0.1448
-0.5379
-1.0283
1.3200
0.0389
0.1777
1.0000
1.0000
1.0000
1.0000
1.0000
1.0000
1.0000
W eigh ts
MXY
ER
V2
INV
Pc
Pcd
Qc
0.0005
-0.0010
0.0012
0.0005
0.0004
-0.0003
0.0001
0.0010
0.0002
0.0008
0.0019
0.0007
-0.0009
-0.0001
-0.2379
-0.3524
-0.0658
0.1247
-0.0886
-0.0338
-0.0056
0.0017
0.0193
0.1524
0.0049
-0.0581
-0.0052
0.0004
-0.0169
0.0047
0.0060
0.0062
0.0036
0.0015
-0.0003
0.0004
0.0085
-0.0034
0.0104
-0.0004
0.0029
0.0000
-0.1900
0.1333
-0.0231
0.0327
-0.0336
-0.0608
0.0013
Trace
Statistic
Trace
r=0
r< 1
r<2
r<3
r<4
r< 5
r<6
166.375*
150.4
106.512
117.49
58.501
88.59
32.406
63.66
15.278
42.7
5.226
25.64
0.082
12.34
* Significance at 5 percent level (MacKinnon, 1999)
59.862*
48.010*
M ax X
(0.95)
50.51
4437
26.095
17.127
10.052
5.143
0.082
38.22
31.99
25.68
19.21
12.34
M ax
317
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
0.234248
0.2032
0.1512
0.1297
0.0872
0.0220
0.0035
-0.5860
-2.3805
3.3061
-11.5885
Eigenvector
MXY
ER
-0.8473
0.9547
-0.1644
-0.5168
2.6810
-0.1812
1.2912
-1.2714
1.6409
20.1984
VG
0.8517
-1.3779
-0.2251
-0.2471
-0.3561
-0.6658
6.9855
INV
-0.3170
2.8230
0.1311
-3.2038
2.6480
0.6237
6.9953
0.3568
-1.4004
-5.7391
-1.0850
3.9987
0.0258
0.6383
-0.6188
-1.3585
-6.4744
1.0000
1.0000
1.0000
1.0000
1.0000
Pc
Pcd
0.3760
-0.4428
-0.6790
7.9606
Qc
1.0000
1.0000
W eights
MXY
ER
VG
INV
Pc
Pcd
0.0007
0.0000
-0.0004
0.0011
-0.0001
-0.0005
0.0003
0.0004
-0.0010
-0.0003
0.0018
0.0014
-0.0007
-0.0003
0.0042
-0.0017
-0.0002
-0.0015
0.0002
0.0010
-0.0001
0.0857
-0.0242
0.0890
0.0926
-0.0807
-0.0008
0.0002
0.0013
0.0143
0.0036
0.0080
0.0054
0.0015
-0.0010
-0.0043
-0.0044
0.0083
0.0033
0.0090
0.0022
0.0002
0.1340
0.0961
0.1315
-0.0662
0.0492
-0.0428
-0.0061
Qc
Trace
Statistic
Trace
(0.95)
Max
Max X
(0.95)
165.420*
150.4
48.308
50.51
117.110
76.003
117.49
88.59
46.321
63.66
41.113
29.681
25.141
21.179
46.581
42.7
25.64
16.521
44.37
38.22
31.99
25.68
4.027
0.631
19.21
12.34
0.631
12.34
* Significance at 5 percent level (MacKinnon, 1999)
318
R ep ro d u ced with p erm ission o f the copyright ow ner. Further reproduction prohibited w ithout perm ission.
0.2610
0.1805
0.1260
0.1096
0.0440
0.0209
0.0028
E igenvector
MXY
ER
VN
-0.1553
0.1438
-1.1758
-2.1803
-0.2343
8.7934
-1.5472
-0.7287
-0.2397
1.2036
-2.2959
-0.9751
2.0795
3.8146
0.7672
0.5498
0.0663
-0.9794
-1.5498
-3.1225
-2.1506
0.7442
INV
Pc
-0.0482
0.0565
-0.5828
8.4852
-0.6842
2.0954
0.2783
0.2783
-3.7386
-1.3613
0.0141
-2.0889
2.0372
Pcd
Qc
-1.9290
-0.6702
-0.7381
1.1919
0.4674
-0.6166
-1.0598
1.0000
1.0000
1.0000
1.0000
1.0000
1.0000
1.0000
W eights
MXY
ER
VN
INV
Pc
-0.0008
-0.0002
0.0004
-0.0008
-0.0006
-0.0006
0.0002
0.0021
-0.0018
0.0011
0.0002
-0.0013
-0.0005
-0.0002
0.0100
-0.0316
-0.0146
-0.0055
-0.0038
0.0090
0.0015
-0.0340
-0.0480
0.0380
-0.1357
0.0301
-0.0225
-0.0045
-0.0109
-0.0112
0.0104
0.0015
-0.0024
0.0023
-0.0005
0.0096
-0.0015
0.0115
0.0023
-0.0001
0.0021
0.0008
-0.0310
-0.1617
0.0399
0.0988
0.0417
-0.0452
0.0003
Pcd
Qc
r<1
r< 2
r<3
r<4
r<5
r<6
Trace
Statistic
Trace
(0.95)
Max X
Max X
(0.95)
148.680*
93.924
57.884
150.4
54.756*
36.039
24.383
21.015
8.152
50.51
44.37
38.22
33.500
117.49
88.59
63.66
12.485
4.332
42.7
25.64
3.823
12.34
0.509
* Significance at 5 percent level (MacKinnon, 1999)
0.509
31.99
25.68
19.21
12.34
319
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
0.3122
0.2203
MXY
ER
VI
INV
Ps
Psd
Os
-0.1576
-0.1425
0.5437
-0.7383
0.5262
-0.3429
1.0000
0.3636
-0.4954
MXY
ER
VI
INVI
Ps
Psd
-0.0003
-0.0009
0.1223
-0.2067
-0.0148
Qs
0.0129
0.0931
0.1259
0.0796
0.0300
0.0021
-0.8377
3.3252
7.3889
8.1246
0.4577
1.1488
-1.9248
-3.8282
6.4436
0.4094
0.4412
-13.193
-6.418
-4.106
5.3014
E igenvectors
1.3462
0.4436
0.1505
-2.4206
0.0191
-0.1365
0.3922
-3.7508
-0.2879
-1.4804
-0.6364
4.1565
-0.3562
4.4573
-2.7023
-3.6274
-3.036
-5.572
1.0000
1.0000
1.0000
1.0000
1.0000
1.0000
0.0023
0.0008
-0.0006
0.0010
-0.0001
-0.0009
0.0003
0.0014
-0.0002
0.0002
0.0866
0.1085
0.0005
0.0565
-0.0872
-0.0339
-0.0005
-0.0002
0.0104
-0.0029
-0.0140
-0.0011
-0.0028
0.0030
0.0011
-0.0054
0.1667
-0.0039
-0.0054
0.0455
-0.0152
0.0006
1.5540
W eights
-0.0107
-0.0101
0.0497
-0.0966
0.0010
0.0117
-0.0068
-0.0635
Trace
(0.95)
M AX (X)
II
o
Null
Hypothesis
175.709*
150.4
r< 1
r<2
r<3
107.979
62.944
117.49
88.59
63.66
67.730*
45.034*
r<4
r<5
r<6
38.591
20.905
5.894
0.385
* Significance at 5 percent level (MacKinnon, 1999)
42.7
25.64
12.34
M AX (X)
(0.95)
50.51
44.37
24.353
38.22
17.685
31.99
15.011
5.508
25.68
19.21
12.34
0.385
320
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
0.0000
0.0006
0.0066
0.2830
0.1938
0.1431
MXY
-0.1299
-0.1389
0.4121
0.2626
-0.2051
-0.2734
0.3113
DINV
Ps
0.5410
0.8675
0.5904
1.2886
0.0921
-1.5099
5.0023
-3.4069
-0.7583
-1.7240
-0.3994
6.3386
0.2252
Psd
-0.1322
0.7131
-7.1816
0.5547
-2.2829
-1.8006
0.0589
-1.6793
-2.4175
Os
1.0000
1.0000
1.0000
1.0000
1.0000
1.0000
1.0000
MXY
ER
-0.0007
-0.0011
0.2772
-0.0023
-0.0003
-0.0004
0.0008
0.0002
-0.0002
0.0000
0.0000
0.1800
-0.0005
-0.0213
-0.0002
-0.0014
-0.2435
-0.0005
-0.1281
0.0016
0.0036
-0.2245
-0.0115
-0.0087
-0.0693
0.0006
0.0058
-0.0763
0.0129
-0.0084
-0.1557
0.0021
0.0010
0.0009
0.0028
0.0000
0.0000
0.0772
-0.0035
-0.0063
0.0163
Ps
Psd
Os
0.0101
-0.0523
-0.0491
0.0518
0.0199
-0.0043
0.0013
0.0876
0.0582
0.0257
0.0001
-1.5373
4.2194
23.0737
-0.5935
-0.1090
-0.1738
4.5946
-0.3373
-26.0202
E igenvectors
ER
V2
-12.3725
-3.3275
-4.1442
-4.2438
W eights
V2
DINV
Trace
Statistic
Trace
(0.95)
M A X (X )
MAX (k )
(0.95)
159.340*
99.128
150.4
117.49
60.211*
38.988
50.51
44.37
60.139
32.192
15.590
4.736
88.59
63.66
42.7
25.64
27.947
16.601
10.853
4.719
38.22
12.34
0.1706
0.1706
* Significance at 5 percent level (MacKinnon, 1999)
31.99
25.68
19.21
12.34
321
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
0.0002
0.1822
0.2340
0.1021
0.0785
0.0274
0.0065
E igenvectors
MXY
-0.769
0.611
0.385
-0.148
-1.120
-28.220
-1.655
ER
-0.168
-0.357
-0.388
0.675
-1.043
-12.170
4.663
VG
0.916
1.053
0.089
-0.804
0.210
9.015
1.929
INV
-0.516
0.361
-2.511
-1.009
1.221
-5.647
0.214
Ps
0.638
0.180
0.190
-1.314
-3.417
7.732
0.242
Psd
-0.241
0.401
-1.489
1.944
-1.828
7.591
-1.188
Qs
1.000
1.000
1.000
1.000
1.000
1.000
1.000
W eights
MXY
0.0004
-0.0019
0.0007
-0.0008
0.0002
-0.0003
0.0003
ER
0.0007
-0.0016
0.0002
0.0013
-0.0009
-0.0011
-0.0004
VG
0.0018
-0.0008
-0.0020
0.0003
-0.0010
0.0011
-0.0001
INV
-0.0826
-0.1750
-0.1440
0.0778
0.1610
0.0077
-0.0042
Ps
-0.0128
-0.0049
0.0038
0.0066
-0.0075
0.0017
0.0003
Psd
-0.0084
-0.0011
-0.0148
-0.0061
-0.0044
0.0025
-0.0011
Qs
0.0192
-0.0177
0.0495
-0.0552
0.0196
0.0095
-0.0139
Statistic
Trace
(0.95)
MAX Q,)
M AX (X.)
(0.95)
150.4
58.267
50.51
r< 1
125.166*
117.49
44.004
44.37
r<2
76.911
88.59
36.415
38.22
r<3
40.496
63.66
19.484
31.99
r<4
21.011
42.7
14.793
25.68
r<5
6.217
25.64
5.032
19.21
12.34
1.185
12.34
II
183.433*
r< 6
1.185
* Significance at 5 percent level (MacKinnon, 1999)
322
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
0.1071
0.1026
0.0300
0.0014
0.2133
0.2558
-0.3582
0.1601
0.0514
-1.0442
2.0914
4.3032
-77.8048
-13.6504
-30.3623
30.1112
3.9483
-6.9823
-20.0611
-29.0076
-2.2258
-0.8946
7.6610
-1.9824
1.8069
-1.4073
4.3531
-20.1605
1.0000
-0.9866
1.0000
0.8817
1.0000
12.2451
1.0000
-13.1376
6.7477
16.8814
1.0000
5.2322
0.1295
-3.1626
1.0000
0.0001
-0.0009
-0.0005
0.1531
-0.0104
0.0011
-0.0010
-0.0092
-0.0576
-0.0003
0.0004
-0.0004
-0.0016
-0.0002
0.0001
-0.2399
-0.0082
-0.0005
0.0002
0.0343
-0.0777
-0.0065
-0.0044
-0.1506
0.0031
-0.0002
0.0017
-0.0059
-0.0412
-0.0036
0.0237
0.0081
-0.0004
0.0074
-0.0004
0.0004
0.3272
0.1723
0.1290
Eigenvectors
MXY
ER
VN
INV
Ps
Psd
Qs
-0.1169
-0.1030
0.2514
0.5286
0.0904
-0.1414
MXY
ER
VN
INV
Ps
Psd
Qs
-0.0015
-0.0015
1.0000
-4.7967
8.5374
39.8987
W eights
-0.0311
-0.0138
-0.0069
-0.0019
0.0792
0.0029
0.0114
0.0098
0.0031
0.0012
Trace
Trace
(0.95)
Max
176.848*
150.4
71.719*
50.51
117.49
105.129
70.890
88.59
45.882
63.66
42.7
25.371
25.64
5.768
0.257
12.34
* Significance at 5 percent level (MacKinnon, 1999)
34.238
25.008
20.510
19.603
5.5108
0.257
44.37
38.22
31.99
25.68
19.21
12.34
Null Hypothesis
r=0
r< 1
r<2
r<3
r<4
r<5
r<6
Max
(0.95)
323
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
0.0055
0.3641
0.2447
0.2337
0.1145
0.0364
0.0168
0.0074
E igenvectors
MXY
ER
-0.328
-0.079
-0.501
0.282
-4.829
3.983
5.234
-0.252
-0.203
1.301
0.091
-1.890
16.421
-2.443
VI
0.343
-0.145
-2.359
0.796
2.637
7.687
-0.002
MSF
Pm
-0.144
-0.196
-0.292
1.006
-4.142
-8.210
-3.613
1.248
-0.463
1.028
0.260
0.883
-3.923
2.790
Pmc
Qm
-0.924
-0.630
1.109
0.836
2.836
-0.544
0.788
1.000
1.000
1.000
1.000
1.000
1.000
1.000
W eights
MXY
ER
VI
MSF
Pm
Pmc
Qm
0.0003
-0.0024
-0.0007
-0.0003
0.0008
0.0004
0.0001
0.0007
-0.0010
0.0011
-0.0003
0.0013
-0.0009
-0.0001
0.0037
-0.0275
-0.0906
-0.0732
-0.0101
-0.0147
-0.0110
-0.0833
0.0213
0.0263
-0.0643
-0.0002
0.0028
0.0170
-0.1495
-0.0433
0.0389
-0.0031
-0.0016
0.0002
-0.0031
0.0226
-0.0305
0.0352
-0.0020
-0.0040
-0.0016
0.0004
-0.1621
0.1060
0.1587
-0.1666
0.0476
0.0549
-0.0441
Max
Max
(0.95)
Trace
"t
II
Null Hypothesis
214.054*
150.41
81.937*
50.51
r< 1
r< 2
132.116*
117.49
40.803
44.37
81.313
88.59
38.189
38.22
r<3
r<4
33.123
63.66
22.002
31.99
11.121
42.71
6.708
25.68
r<5
4.412
25.64
3.059
19.21
r<6
1.353
12.34
1.352
12.34
324
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
0.3657
0.2544
0.2266
0.1549
0.0384
0.0172
0.0121
Eigenvectors
MXY
-1.245
0.958
-0.157
0.135
-3.049
6.880
8.964
ER
-0.455
-1.936
-0.002
0.383
-2.154
-5.784
10.904
V2
0.799
2.006
-0.556
0.292
3.762
-0.084
6.211
MSF
0.205
-0.408
-0.500
1.063
-3.169
0.018
-7.952
Pm
0.391
-3.808
0.112
-0.194
0.825
4.159
0.145
Pmc
-0.428
-2.059
-0.324
0.810
3.294
1.353
0.345
Qmk
1.000
1.000
1.000
1.000
1.000
1.000
1.000
Weights
MXY
ER
0.00001
-0.0004
-0.0026
-0.00001
0.0007
0.0000
-0.0003
0.0009
-0.0008
-0.0009
-0.0006
0.0014
-0.0001
0.0007
V2
-0.1918
0.3238
-0.1742
-0.1593
-0.0188
0.0572
0.0491
MSF
-0.1049
-0.0179
0.0102
-0.0705
0.0003
-0.0227
-0.0095
Pm
-0.1317
-0.0750
-0.0097
0.0041
-0.0002
0.0089
0.0035
Pmc
0.0259
-0.0303
-0.0181
-0.0153
-0.0069
0.0008
0.0029
Qmk
-0.1375
-0.0887
0.1919
-0.1948
0.0575
0.0755
-0.0318
Null Hypothesis
r=0
Trace
Trace
(0.95)
Max
M ax
(0.95)
224.952*
150.41
82.395*
50.51
r< 1
142.556*
117.49
53.140*
44.37
r<2
81.415
88.59
36.519
38.22
r<3
42.896
63.66
30.462
31.99
r<4
12.433
42.71
7.088
25.68
r< 5
5.345
25.64
3.140
19.21
r< 6
2.20
12.34
2.20
12.34
325
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
0.31825
0.23516
0.17705
0.14325
0.08801
0.01900
0.00045
E ig en v ecto rs
MXY
-0.7859
-0.1947
0.3016
-0.6852
0.0788
11.3519
3.2500
ER
-0.2103
-0.5928
0.0905
0.5487
-0.0682
5.6821
-7.7770
VG
0.3470
0.1212
0.1834
-1.2708
1.6822
-2.1786
-2.6896
MSF
-0.7775
0.0378
1.4856
-0.4007
-0.9771
-2.7457
-0.9965
Pm
1.0726
-2.3014
0.3950
-0.2947
0.1152
-1.0433
1.2879
Pmc
-1.4336
-1.1367
0.0966
1.0532
0.8276
-0.4895
1.9529
Qm
1.0000
1.0000
1.0000
1.0000
1.0000
1.0000
1.0000
W eig h ts
MXY
0.00090
-0.00134
-0.00084
-0.00158
-0.00080
-0.00017
-0.00004
ER
-0.00006
0.00027
-0.00061
-0.00001
-0.00049
-0.00129
0.00008
VG
-0.00114
0.00138
0.00011
-0.00096
0.00346
0.00032
0.00002
MSF
0.02425
0.01683
-0.08250
-0.01240
-0.01971
0.01944
0.00312
Pm
-0.10933
-0.05432
-0.04126
0.00977
-0.00347
0.00664
0.00104
Pmc
0.02793
-0.03368
-0.00219
0.01791
0.01083
-0.00183
0.00058
Qm
-0.09139
0.05827
-0.32480
0.11897
0.01531
-0.00135
-0.01063
Trace
Trace (0.95)
Max
Max
(0.95)
II
o
Null Hypothesis
202.457*
150.41
69.722*
50.51
r<1
132.734*
117.49
41.001
44.37
r< 2
83.942
88.59
35.464
38.22
r<3
48.478
63.66
28.138
31.99
r< 4
20.340
42.71
16.767
25.68
r< 5
3.572
25.64
3.491
19.21
12.34
.081
12.34
r< 6
.081
* Denotes significante at 95% (Mackinon, 1999)
326
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
0.2480
0.1923
0.1885
0.1225
0.0420
0.0124
0.0056
-6.9873
-16.4540
3.5215
E ig en v ecto rs
MXY
-0.4054
ER
VN
MSF
-0.0900
0.5935
0.1149
-0.2082
0.0994
-2.7042
0.2995
0.0547
-31.8890
-2.0118
0.2774
-0.3438
-0.9747
1.5159
-9.4245
9.9088
-1.2855
-0.4434
0.7164
-1.9985
-1.2664
-8.0627
5.7002
-0.9992
Pm
1.0959
-0.2546
-6.1878
-0.5194
0.7021
5.4305
1.4688
Pmc
-1.3727
0.2660
-4.3487
1.0407
4.0610
5.2468
0.6190
Qmk
1.0000
1.0000
1.0000
1.0000
1.0000
1.0000
1.0000
W eig h ts
MXY
0.0011
-0.0012
0.0008
0.0013
-0.0009
0.0003
-0.0001
ER
0.0003
-0.0014
0.0009
-0.0018
-0.0017
-0.0003
0.0000
VN
-0.0122
0.0001
-0.0142
-0.0174
-0.0024
0.0095
0.0013
MSF
-0.0652
0.0690
0.0445
0.0068
-0.0056
0.0064
-0.0173
Pm
-0.0934
-0.0157
0.0855
0.0178
-0.0003
0.0016
0.0037
Pmc
0.0301
0.0044
0.0339
-0.0148
0.0034
0.0015
0.0008
Qmk
-0.1286
0.3126
0.0849
0.1116
-0.0732
-0.0072
0.0325
Trace
(0.90)
r=0
r< 1
163.616*
150.41
111.738
r<2
72.879
r<3
r<4
r<5
Null Hypothesis
M ax
Max
(0.95)
51.878*
50.51
117.49
38.859
44.37
88.59
38.004
38.22
34.874
63.66
23.790
31.99
11.084
42.71
7.804
25.68
3.279
25.64
2.262
19.21
12.34
1.010
* Significance at 5 percent level (MacKinnon, 1999)
1.010
12.34
r<6
327
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.