You are on page 1of 10

Impact of Exchange Rates on Bilateral Trade Flows: A

Gravity Model Analysis


by
Anurag Siroliya and Dr. S.K. Mathur

Abstract
This study is divided in two sections. The first section examines the impact of exchange rate on the level of Indias
bilateral trade flows with its trade partners. The study uses panel data estimation on Anderson and Wincoops
gravity model for India and its 176 trading partners for the years 1996 to 2006. The exchange rate used in the panel
analysis is the currency rate between the US dollar and INR. The results show that exchange rate volatility have a
significant and negative impact on bilateral trade flow which is consistent with the past literature. This suggests that
observed forethought management of the real exchange rate is highly critical for trade promotion and
macroeconomic stability. In the next section we look at Novys work and calculate trade cost measure and
multilateral resistance term for India and its trading partners. Along with trade growth decomposition, we observe
that declining trade cost is favourable for bilateral growth but declining multilateral resistance leads to decline in
bilateral trade growth due to trade diversion effect, i.e. trade with the rest of the world increases if multilateral
resistance term is relaxed for two bilateral trading partners. For some countries, Indias bilateral trade
decomposition is not consistent with Novys results. An intuitive explanation for this can be that the determinants
suggested by Novy in trade growth decomposition are not sufficient enough to explain the bilateral trade growth.

Keywords: Exchange Rate, Gravity Model, Novys Trade cost, India

Page | 0

1. Introduction
The research paper tries to answer the following questions:
a) Does exchange rate have a significant impact on bilateral trade for India between India and its major
trading partners? Is this impact positive or negative on bilateral trade?
b) What are the Novys tariff equivalence and multilateral resistance term for India and its trading
partners during the period 1996 to 2006? What is the individual contribution of factors in the Novys
decomposition of growth of trade?
In the recent years, international trade has surged so much so that now even countries which had heavy
restrictions on trade earlier, have relaxed their barriers and trade much more than what they used to a few
decades ago. This can primarily be accrued to reduction in international trade costs. Broadly defining,
Trade costs include all costs incurred in getting a good to a final user other than the marginal cost of
producing the good itself: transportation costs (both freight costs and time costs), policy barriers (tariffs
and non-tariff barriers), information costs, contract enforcement costs, costs associated with the use of
different currencies, legal and regulatory costs, and local distribution costs (wholesale and retail). Both
domestic and international trade costs are included because one should also compute trade costs once
goods cross a border.
Novy (2011), in his paper uses Anderson and Wincoops gravity equation and manipulates it to derive a
micro-founded measure of aggregate bilateral trade costs. As a result, he obtains the trade cost parameters
as a function of observable trade data that can be observed over a time period. This measure is highly
useful since it is easy to implement and captures a variety of trade cost components ranging from
transport costs to tariffs to non- tariff components such as language barriers, institutional factors,
contiguity, etc.
In the past few decades, various changes are impacting the international monetary system. China is
considering floating Chinese Yuan, many countries in Eastern Europe are joining the euro area, El
Salvador and Gautemala have established peg to the dollar and Ecuador has accepted dollar as its
currency. These changes have caused significant fluctuations in the exchange rate system which has
raised debates over significant determinants of bilateral trade flows, particularly effect of exchange rate
variability on international trade. A number of studies indicate that real effective exchange rates
significantly impact bilateral trade flows and for most cases the relation between them is found to be
negative.
The paper is divided into five sections with the present one being introductory. The analysis part in the
paper is further divided in two sections. In the first section we refer to Novys paper to compute measure
of multilateral resistance term and bilateral trade cost which can be used in the next section to estimate
Anderson and Wincoops gravity model. We also use Novys decomposition of trade growth to compute
individual contributions of factors in trade growth accounting. Furthermore, in the second section we
study the impact of exchange rates on bilateral trade flows using Novys version of Anderson and
Wincoops gravity model.

2. Literature Review
One of the most crucial factors affecting international trade has been trade costs that can be incurred
locally as well as across the borders. There have been several attempts to measure this trade costs by
Page | 1

various researchers. Tinbergen (1962) was the first to account for this trade cost and he used distance as a
proxy for it. This further led to other economist expanding the set of proxies used in measuring trade cost
by incorporating common border, tariffs, common language, common currency, remoteness, etc in it.
Anderson (1979) used a system of expenditure equations to derive the gravity model of international
trade. With Wincoop in 2003, Anderson challenged McCallums (1995) use of distance and border
proxies for trade cost in the gravity model of US-Canada bilateral trade and proved that the selection of
proxies used by McCallum was inappropriate. Anderson and Wincoop thus used a particular trade cost
function, exogenous to the model, to represent trade cost in the gravity model. Moreover, they claimed
that apart from bilateral trade barriers, multilateral trade barriers also impact international trade and used a
specific term called multilateral resistance term to represent them. After a couple of years, Novy (2008)
derived a micro-founded measure for international trade costs by applying certain transformations to
Anderson and Wincoops gravity model. Unlike Anderson and Wincoop, Novy assumed asymmetric
trade cost which allowed him to directly calculate international trade cost from observable data.
Moreover, past work shows mixed results about the impact of exchange rate volatility on trade flows but
most of them show a negative impact. Recently, Dell Ariccia (1999) estimated the impact of exchange
rate volatility on the bilateral trade of 15 EU member states and Switzerland over the 20 years, from 1975
to 1994 using fixed effects in panel data. Ariccia found that exchange rate volatility has a small but
significant negative impact on trade. Furthermore, he predicted that eliminating exchange rate volatility to
zero in 1994 would have increased trade by 3 to 4 percent. Other empirical studies that discovered a
negative relationship between exchange rate volatility and bilateral trade include Thursby and Thursby
(1985), Kenen and Rodrik (1986), Koray and Lastrapes (1989) and Onafowora and Owoye (2007). In
contrast to the above, Asseery and Peel (1991) and Todani and Munyama (2005) reported a positive
relationship between exchange rate volatility and exports. Surprisingly, Gotur(1985), Solakoglu (1998),
Rey (2006), and Boug and Andreas (2007) could not find any significant relationship between the two.
The takeaway from the empirical literature is that earlier studies done in this area resulted either in
insignificant relationships between exchange rate volatility and bilateral trade, or significant relationships,
most of which were negative in nature.

3. Model, Data and Variables


3.1 Novys Measure of Trade Costs
Trade costs are defined as the difference between the marginal cost of a traded commodity and
its price paid by the ultimate consumer.[2] They capture variables such as tariff and non-tariff
barriers, transport costs, exchange rate volatilities, etc. Novy uses Anderson and Wincoops
gravity equation as a framework to derive measure of trade costs by relaxing the assumption of
symmetric bilateral trade costs to non-symmetric.
Anderson and Wincoop (2003) Gravity Model

Here,
xij is the total volume of trade between two trading partners i and j; tij is the transport cost of the iceberg
type; yi and yj are the GDP of country i and j respectively; yw world GDP ; is the elasticity of
substitution; Pi and Pj are the CES price indices of country i and j.
Page | 2

Pi in the above equation can be interpreted as:

Novy uses the above model and applies certain transformations to get the following bilateral trade cost
measure:

Here, ij are the tariff equivalents of trade costs, xii and xjj represent the intra-national trade in countries i
and j respectively. xij denotes the bilateral trade flow from country i to j and xji denotes the bilateral trade
flow from country j to i. is the elasticity of substitution across goods. We observe that trade costs (ij)
are directly proportional to intra-national trade (xiixjj) and inversely proportional to international trade
(xijxji). Thus, if bilateral trade flow (xijxji) increases in comparison to intra-national trade flows (xiixjj), the
bilateral trade between the two countries rises. Moreover, this is reflected in the decline of bilateral trade
costs and vice-versa.

3.2 Novys Measure of Multilateral Resistance Term and Trade Growth


Decomposition
Novy raises the question whether the increase in bilateral is only due to secular economic growth or
whether the increase can be accrued to reductions in trade frictions. He uses Anderson and Wincoops
gravity model followed by its first difference:

Substituting tariff equivalence (ij) for tij and tji in the above equation we obtain:

...... (1)
where i is the Novys multilateral resistance term for country i relative to domestic trade costs,

If we divide equation (1) by left hand side we get the following:

............ (2)
Equation (2) represents decomposition of bilateral trade growth into three contributions: (a) the
contribution of income growth, (b) the contribution of the decline in relative bilateral trade costs, and (c)
the contribution of the decline in relative multilateral resistance. If relative multilateral trade barriers fall
(i.e., ln(ij) < 0), then contribution (c) becomes negative. This negative contribution is due to trade

Page | 3

diversion effect i.e., if trade barriers with other countries fall, trade with those countries rises but bilateral
trade between i and j declines.
Moreover, equation (1) can be modified to estimate the coefficients:
Y = a1*X1 + a2*X2 + a3*X3 +
where:

Y=

; X1 =

; X2 =

; X3 =

Novys Multilateral Resistance term:


Novy found an explicit solution for multilateral resistance term (MRT) by using Anderson and Wincoop
gravity model as a starting point and applying certain transformations. The MRT for Novys model is
given below:

3.3 Impact of Exchange Rate Volatility on Bilateral Trade Flows


In this section we study the impact of real effective exchange rate on bilateral trade flows. Exchange rates
are adjusted on the basis of CPI data to convert the nominal exchange rate values to real exchange rate
values. The estimated model for this is given below:
M

lnbtradeij = 1* lnYi + 2* lnYj 3* lnYw + 4* REERit + m=1m*(Zijm)


................ (i=1,2...M and t = 1996-2006)
Here,
lnbtradeij is the log of bilateral trade between country i and j ; i are the parameter estimates; lnYi and lnYj
are the GDP of country i and j respectively; lnYw is the log of the world GDP; REERit is the real effective
exchange rate; Zijm is the set of dummies including contiguity, common currency, common language,
RTA member, GATT member, etc.

3.4 Dataset
This study uses panel data methods on the gravity model developed by Anderson and Wincoop (2003).
The analysis is carried out for India and its 176 trading partners during the post liberalization period 1996
to 2006. The exchange rate used in the panel analysis is the currency rate between the US dollar and INR.
Intra-national trade for country i (xii) is defined as the total income minus total exports, xii = yi - xi.
Total exports is the total of all exports from country i, xii = xij. We only consider contribution of
agricultural and manufacturing goods in GDP to calculate yi. Data for all the variables is obtained from
CEIC database and UNCTAD database. Moreover, similar to Anderson and Wincoops paper, we have
assumed the elasticity of substitution, = 8 which generally lies in the range 5 to 10.

Page | 4

4. Results
4.1 Impact of Exchange Rate Volatility on Bilateral Trade Flows
To analyse the effect of exchange rate on bilateral trade we use panel data study for 176 countries during
the period of 1996 to 2006. To figure out the appropriate panel data estimation technique to be used, we
apply Haussman and Breusch Pagan LM test. These tests do not reject the null hypothesis and thus
favour the use of random effect model. The table below reports the result from the application of random
effects model.
Random-effects GLS regression

corr(u_i, X)

= 0 (assumed)

Wald chi2(6)
Prob > chi2

=
=

3268.91
0.0000

-----------------------------------------------------------------------------lnxij |
Coef.
Std. Err.
z
P>|z|
[95% Conf. Interval]
-------------+---------------------------------------------------------------Contig |
279.6871
705.5396
0.40
0.692
-1103.145
1662.519
Lnyi
|
5.841429
2.778467
2.10
0.036
.3957346
11.28712
lnyj
|
2.365787
.1419134
16.67
0.000
2.087642
2.643932
Lnyw
| -11.13885
4.295508
-2.59
0.010
-19.55789
-2.719806
rta
|
.8704819
.3348378
2.60
0.010
-.1218001
1.19074
reer
| -.0673669
.0310455
-2.17
0.030
-.115215
.0064812
gatt_d |
.2362733
.2165503
1.09
0.275
-.1881575
.660704
Comcur | -104.6879
1455.529
-0.07
0.943
-2957.472
2748.096
_cons |
123.5476
48.82639
2.53
0.011
27.84961
219.2455
-------------+---------------------------------------------------------------Table 4.1

From the random effect estimation in table 4.1, we observe that the Dollar-Rupee Real Effective
Exchange Rates (reer) have a negative and significant impact on bilateral trade flow of India with its 176
trading partners. This result is similar to our expectation as well as past literature. This can be explained
by the fact that exchange rate volatility creates uncertainty about future profit from bilateral trade, hence
it causes expectations of gains from bilateral trade to fall which further leads to decline in bilateral trade
between partner countries.
As far as other gravity model variables are concerned, GDP of partner countries show a positive and
significant on bilateral trade while world GDP shows a negative and significant impact on bilateral trade.
This verifies that gravity model holds for India and its trading partners. From all the proxies used to
account for bilateral trade costs, only RTA show a positive and significant impact on bilateral trade since
with introduction of RTA free flow of trade can occur between partner countries. As other proxies for
trade cost are insignificant, we need to look at other methods to measure trade cost such as Novys trade
cost measure.

4.2 Novys Measure of Multilateral Resistance Term and Trade Growth


Decomposition
In the following section we provide tariff equivalence, multilateral resistance term and growth accounting
of trade between India and a specific list of partner countries. The above analysis between India and the
entire list of countries is available in Appendix 2.
Novy stated that bilateral trade growth can be decomposed into three contributions:
Page | 5

(a) The contribution of income growth


(b) The contribution of the decline in relative bilateral trade costs
(c) The contribution of the decline in relative multilateral resistance
Decomposing the Growth of Indian Bilateral Trade with its Trading Partners

Partner
Country

Partner
Country
code

Average
Yearly Tariff
Equivalence

Average
Yearly
Multilateral
Resistance
Term

Average
Yearly
Bilateral
Trade Growth

Contribution
of The
Decline in
Relative
Multilateral
Resistance

Contribution
of The
Growth in
Income
60.10%

Contribution
of The
Decline in
Relative
Bilateral
Trade Costs

Japan*

JPN

0.194021

1.535177

10.16331

Germany*

DEU

0.198776

1.378471

10.73199

-11.60%

49.27%

62.32%

100.00%

Italy*

ITA

0.270599

1.532609

11.24807

-19.37%

67.22%

52.15%

100.00%

Canada*

CAN

0.280792

1.704159

11.80462

-21.97%

137.65%

-15.68%

100.00%

Australia*

AUS

0.288684

1.422783

12.25247

ARG

0.289339

1.882985

13.25102

57.04%
42.93%

62.16%
79.71%

100.00%

Argentina*

-19.20%
-22.64%

Malaysia

MYS

0.295771

1.13938

13.87708

-1.38%

52.93%

48.45%

100.00%

Indonesia*

IDN

0.320463

1.289722

13.17715

-31.63%

88.80%

42.83%

100.00%

France*

FRA

0.347797

1.603185

11.06386

-98.13%

234.89%

-36.76%

100.00%

Maldavies

MDV

0.439574

2.447268

18.80335

-17.32%

71.53%

45.79%

100.00%

Chile

CHL

0.458763

1.894089

14.02676

-13.85%

52.50%

61.35%

100.00%

Madagascar

MDG

0.486733

2.388067

17.02694

-34.25%

57.82%

76.43%

100.00%

Brazil*

BRA

0.491895

1.898079

11.93208

-31.45%

80.98%

50.48%

100.00%

-12.11%

39.61%

72.51%

100.00%

China*
CHN
0.587093
1.344244
11.18799
* Countries belong to G20 list of countries which includes India too.
Average Growth per year is calculated for the time period 1996-2006.
Calculations are based on Novys trade decomposition equation.

71.60%

Total

-31.70%

100.00%

100.00%

Table 4.1

We applied Novys trade growth decomposition for Indian and its major trading partners for the years
1996 to 2007 and presented the results in table 4.1. The table reveals that during the period 1996 to 2007,
growth in income in most countries like Japan, Canada, Brazil and France, etc. explains trade by more
than 50% while in some countries like China and Argentina, income growth explains trade by less than
50%. Similar is the case with decline in bilateral trade cost. For countries like Japan, Argentina, China,
etc. decline in bilateral trade cost explains trade by much more than 50% while for countries like
Indonesia and Maldavies, it explains trade by less than 50%. Thus decline in relative bilateral trade cost
and growth in income are close competitors as far as trade growth is concerned. Moreover, for some
countries like France and Canada decline in bilateral trade cost has a negative impact on trade growth.
This is can be explained by the fact that there must be some external factors not included in the above
table that might have caused a significant effect on trade growth. In general, declining bilateral trade cost
ang growth in income will cause a positive impact on bilateral trade growth which explains the positive
sign of the two variables.
A decline in multilateral resistance term will have a negative impact on bilateral trade since it will divert
away from India to ROW. This explains the negative sign of decline in multilateral resistance contributing
to bilateral growth. Furthermore, decline in multilateral resistance term (MRT) for France poses a
Page | 6

significant negative impact on its bilateral trade growth with India. Thus France favours trade much more
with rest of the world rather than India if its MRT term declines.
We also included three more variables in the above table, namely average yearly tariff equivalence,
average yearly MRT and average yearly bilateral trade growth. For many countries we see that they
satisfy the relation that if the tariff equivalence for a country is higher, then its bilateral trade is low since
the country prefers domestic trade to international trade.
To summarize the above table, we can say that income growth is the one of biggest drivers causing the
bilateral trade to rise between India and its partners. The result obtained follows the findings of Baier and
Bergstrand (2001) who argued that two-thirds of the growth in trade amongst OECD countries between
1958 and 1988 can be explained by the growth of income. The reason for decomposing the growth of
trade with equation (2) is to explicitly account for multilateral trade barriers in the model. They are crucial
as in general equilibrium, both by bilateral and multilateral trade barriers affect the trade flows between
any two countries.

3.3 Trade Growth Decomposition using Regression:


Now, to observe the contribution of trade growth drivers and their significance level individually we
modify Novys trade growth decomposition to an estimated equation with error term given below:
Y = a1*X1 + a2*X2 + a3*X3 +
where:
Y=

; X1 =

; X2 =

; X3 =

We consider a panel data with around 80 countries and calculate their trade growth and individual
contributions of decomposed trade growth elements over a span of 10 years. To check for the correct
panel data estimation technique we apply Haussman and Breusch Pagan LM test. The tests reject the null
hypothesis and thus favour the use of Fixed Effects method for estimation. The results from Fixed Effects
estimation are show in table 4.2 given below:
Fixed-effects (within) regression
Group variable: iso3_id

Number of obs
Number of groups

=
=

570
83

R-sq:

Obs per group: min =


avg =
max =

1
6.9
8

within = 0.9996
between = 0.9975
overall = 0.9994

corr(u_i, Xb)

= 0.0124

F(3,484)
Prob > F

= 405703.21
=
0.0000

-----------------------------------------------------------------------------trade_gr |
Coef.
Std. Err.
t
P>|t|
[95% Conf. Interval]
-------------+---------------------------------------------------------------decl_mrt |
8.222955
.3288708
25.00
0.000
7.576764
8.869145
inc_gr |
1.628453
.0210186
77.48
0.000
1.587154
1.669752
decl_tc |
-14.0172
.0127306 -1101.07
0.000
-14.04221
-13.99218
_cons |
.0178139
.0020886
8.53
0.000
.01371
.0219178
-------------+---------------------------------------------------------------sigma_u | .03156301
sigma_e | .02971344
rho | .53015649
(fraction of variance due to u_i)
-----------------------------------------------------------------------------F test that all u_i=0:
F(82, 484) =
3.16
Prob > F = 0.0000
Table 4.2

Page | 7

For regression we have not taken sign with elasticity values so as to avoid multicollinearity, hence the signs of decline in trade
cost and multilateral resitance term is inverted.

From table 4.2 we observe that all the three drivers of trade growth (trade_gr) namely, contribution of
declining multilateral resistance term (decl_mrt) , contribution of income growth (inc_gr) and decline in
trade cost (decl_tc) are significant at 95% confidence level. Moreover decline in trade costs have a
negative impact on trade growth that can be explained because we have not taken the sign and the
elasticity part of the trade cost decline term during regression so as to avoid multicollinearity. Similarly
for MRT the sign for decline in MRT is negative since a decline in multilateral resistance term suggests
that trade with the rest of the world will increase rather than with India. This is called the trade diversion
effect. From the above table we can also say that declining trade cost and income growth have a
enormously high correlation with trade growth as compared to declining multilateral resistance term.

5. Conclusion
Our analysis on impact on exchange rates on bilateral trade provides the evidence that real exchange rate
fluctuations depress Indias bilateral trade flows with partner countries. The effects are largely due to
exchange rate volatility creating uncertainty about future profit from bilateral trade. Thus this has policy
implications for the government to foster efficient management of the exchange rate. One should not
undermine the consequences of depressed trade flows which also turns out to be volatile, has serious
economic growth implications for a developing country such as India. Therefore it implies that
government has to exercise appropriate tools or monetary policies that ensure macroeconomic stability.
For a developing country like India, with a market determined exchange rate, it is necessary to keep in
mind that the macroeconomic fundamentals that affect the exchange rate are well managed to ease out
the volatility.
Novys trade cost analysis provides an easy to compute and efficient way of calculating bilateral trade
costs. The trade cost measure is good estimate as it captures transport costs, tariff as well as non-tariff
barriers like common language, common border, institutional factors, etc. From trade growth
decomposition we observe that growth in income and decline in bilateral trade cost have been
significant contributors to growth in bilateral trade for India and its trading partners. Moreover, growth
in income and decline in trade cost leads to surge in trade flows while the decline in multilateral
resistance term causes trade growth to also fall due to trade diversion effect as trade with the rest of
the world seems to rise. For some countries, Indias bilateral trade decomposition is not consistent with
Novys results. An intuitive explanation for this can be that the determinants suggested by Novy in trade
growth decomposition are not sufficient enough to explain the bilateral trade growth.

6. References
[1] Novy, Dennis. "Gravity redux: measuring international trade costs with panel data." Economic
Inquiry 51.1 (2013): 101-121.
[2] Singh, Sarbjit, and S. K. Mathur. "Trade Costs of India within Asia: Measurement and Its
Determinants."
[3] Lubinga, Moses Herbert, and Barnabas Kiiza. "Exchange Rate Uncertainty and Bilateral Trade Flows:
Insights from Uganda." Business and Economic Research 3.1 (2013): 227-239.

Page | 8

[4] Tenreyro, Silvana. "On the trade impact of nominal exchange rate volatility." Journal of Development
Economics 82.2 (2007): 485-508.

Appendix 1
Derivation of Novys Trade Cost

Page | 9

You might also like