You are on page 1of 14

The Endogenous Growth Model

Case of Sweden and Norway

Paper within Advanced Macroeconomics and


Growth
Author: Lisa Wassén (901122)
Orsa Kekezi (910909)

Table of Contents Tutor: Pia Nilsson


Jönköping March 2014

i
1 Introduction ............................................................................... 1
1.1 Historical background .............................................................................1
2 Theoretical Framework ............................................................. 2
2.1 Previous research...................................................................................3
3 Method ....................................................................................... 4
4 Analysis ..................................................................................... 5
4.1 Descriptive Statistics ..............................................................................5
4.2 Discussion of results...............................................................................5
5 Conclusion ................................................................................ 8
Appendix ...................................................................................... 10

ii
1 Introduction
Norway and Sweden, two developed countries in the north, is the focal point of this paper.
Even though our analysis starts in the 1970’s up to recent years, both countries have ob-
served large changes in growth over the last 100 years. This paper will therefore try to
check where the growth comes from by the use of the endogenous growth model.
The endogenous growth model introduces research and development (R&D) as a new
variable. Innovation and new technology is a large part of this variable that should lead
to growth in a country. The endogenous growth model assumes that the higher a country’s
expenditure on R&D is, the higher will the growth be. With other words, a country with
a high growth rate is assumed to have a substantial stock of R&D.
The purpose of this paper is then to test if the endogenous growth model is good for
explaining the growth of Norway and Sweden when using real world data. We also
wanted to see whether the model fits one country’s data better over the other.
We run three different regressions. One for Norway, one for Sweden and one where we
merge the data of the countries into one data set, where a country dummy is used in order
to investigate the results. The outputs from the regression differ between the countries
and it can be observed that the model fits Sweden better than Norway. We observe some
variables to be insignificant in our model which one would in fact assume to effect
growth. We discuss possible explanations of this in the analysis section where we also
talk about the overall results of the regressions. Both data sets have statistical issues that
one need to deal with. How we solve the problems and the effects of it can be read in the
analysis section.
Although the growth of Sweden is explained better by the model than Norway, it does
not mean that Sweden’s growth is well-explained by this growth model.
The overall result of this paper is that the endogenous growth model does not explain
growth as well as one would assume. Even though the model has a strong theoretical
background, other factors than the one included in the model are useful for explaining
growth.

1.1 Historical background


Norway has gone from being one of the poorest countries in the world a century ago, to
be one of the economically leading countries in the world. Norway was neutral in the
WW1 but it had still some difficulties in the period between the wars. In the Second World
War (WW2), Norway was invaded and had a tough period. In the beginning of the 1970’s,
Norway became an oil nation, and with the wealth, the economic crisis also appeared.
Norway came later out of the crisis rich and with power. Norway has voted several times
regarding the EU-membership but the nation said no, mostly due to the natural resources
of the country (Sweden Abroad, 2014).
Sweden had a similar starting point as Norway 100 years ago. Sweden has, over the dec-
ades observed both strong recessions and expansions. Due to Sweden not being a part of
the WW2, increased international trade, as well as structural changes were reasons for
Sweden being one of the richest countries in the world in the middle of 1970’s. After that,
Sweden experienced a deep recession with lowered economic growth during both the last

1
part of 1970’s as well as in the beginning of the 1990’s. It was not until 1994-95 that
Sweden somehow recovered from the latter recession. Sweden joined the EU in 1994 and
during 1995-2004, Sweden’s GDP increased by 2.6%, which was higher than other Eu-
ropean countries for the same period (Regeringen, 2014).

2 Theoretical Framework
The endogenous growth model introduces a new variable which explains growth and that
is research and development. This is the sector of the economy where new ideas are es-
tablished. The production function in this model now becomes a combination of labour,
capital and technology which all contribute in improving the technology. Ceteris paribus,
larger resources in R&D lead to more discoveries and therefore to a larger growth. It is
important to notice that the endogenous growth model assumes a Cobb-Douglas function
meaning constant returns to scale to capital and labour. This implies that if the inputs are
doubled so will the output. Moreover, savings and the fraction of capital stock and labour
forced used in R&D are exogenous and do not change over time.
The endogenous model comes in different forms, but the basic one includes four varia-
bles, labour, capital, technology and output:

𝐴̇ (𝑡) = 𝐵[𝑎𝐾 ∗ 𝐾(𝑡)]𝛽 ∗ [𝑎𝐿 ∗ 𝐿(𝑡)]𝛶 ∗ 𝐴(𝑡)𝜃 β≥0, B>0, Υ≥0 (Equation 1)
Where, B is a shift parameter, K is capital, L is labour, A is the stock of knowledge, al
and ak are respectively the fraction of labour force and share of the capital stock used in
R&D. θ reflects the effect of the existing shock of knowledge on the success of R&D.
Its value can be positive if the past discoveries make available ideas for new discover-
ies. In the meantime, it can also be negative if the effect is the opposite.
As it can be seen, the only variable which is outside the Cobb-Douglas function, is
knowledge. That is logical since doubling the inputs for R&D does not always lead to
doubling the number of discoveries.
Depending on the values of the exponentials in the model explained above, the growth
of the economy can take different directions. In this paper, we use the proxies of: al=1/3,
ak =2/3 (Due to C.R.S al+ak=1) and θ=1.
A θ=1 means that the knowledge already existing, is just enough to help in creating new
knowledge proportional to the stock. If this is the case, the economy always shows
steady growth. If capital is not present, this kind of model is also known as Y=AK
which is a linear growth.
However, in the case of taking capital into account, θ+ β becomes automatically over
than 1 due to the restriction of β in the basic model. In this case, the economy experi-
ences increasing growth rates and does not quite converge. This type is known as a fully
endogenous model.
Knowledge, on the other hand, comes in different forms and therefore, it is logical to as-
sume that the accumulation process is not going to be the same in all the forms. All
types of knowledge have one thing in common, they are non-rivalry meaning that the
same knowledge can be used simultaneously in a couple of activities. Despite non-ri-
valry, knowledge is at the same time excludable, implicating that it is possible to pre-
vent others for using it with the help of property rights like patents or trademarks. These

2
means give a specific company monopoly powers and therefore a much higher profit
compared to a perfect competitive market. After some time, when the power of the
property rights ends, the knowledge becomes available to everyone.
In reality we are aware that the growth of knowledge has been a crucial factor in the
higher standard of living that we are experiencing nowadays. However, the fact that
there still exists both rich and poor countries is not explained by this growth model. De-
spite the non-rivalry of the knowledge, developing countries still do not have access to
the same technology. And it seems like this is the main reason to why there are huge
economical differences between different countries (Romer, 2011).

2.1 Previous research


Several studies and reports have been made on the topic of the endogenous growth model
and its impact on growth. In this section, studies will be presented, where the model look
either similar or, due to variable extensions, slightly different.
In Tallman and Wang’s article from 1994, they study how well the endogenous model
could explain the growth of Taiwan with the help of human capital. They came up with a
result saying that having an index measuring the labor quality included in the labor input
would improve the growth of Taiwan. According to the authors, their robust results come
from two reasons. First, they come from improvement in measuring the raw labor input.
Second, they got it from using more relevant variables which are correlated with the eco-
nomic growth.
Another article that deals with the endogenous growth model is the one written by Grif-
fith, Redding and Van Reenen (2004). The aim of the paper was to create econometric
evidence of the importance of R&D by investigating different factors of productivity
growth. By inspecting industries of twelve OECD countries, they find that both R&D and
human capital is central for both the catch-up and the innovation rate of countries. They
concluded that a larger R&D in one country would lead to an overall increase in the total
factor productivity (TFP) due to technological transfer.
The paper written by Benhabib and Spiegel, (1994) studies growth by analyzing human
capital with cross-country estimates. They tested two different models in order to see
which one that explained growth the most. In the first model, their results show that when
using human capital as a factor, human capital accumulation does not manage to explain
growth and enter insignificantly in the model. In the second model, they tried a model
where human capital effects TFP. Here, the human capital affects growth both through
domestically produced technological innovation as well as the speed of technological
adoption from foreign countries. The second model gave the authors a better method of
how human capital affects growth.
Another paper that deals with endogenous growth theory is the one by Jones (1995). In
this paper, it is argued that R&D endogenously affects growth. However, long-run growth
comes from other factors rather than R&D, such as growth in the rate of population. How-
ever, the large increase that one would expect based on the theory, from increase in R&D
is not observed.

3
3 Method
The endogenous growth model that we analyse is the same as the one of Tallman and
Wang (1994) in their paper about the case study of Taiwan:

𝑌𝑡 = 𝐴𝑡 𝐾𝛼𝑡 (𝐻𝑡 𝐿𝑡 )𝛽 (Equation 2)

After transforming it into a log form by taking the natural logarithms of each variable, we
receive the following:
𝐿𝑛𝑌𝑡 = 𝐿𝑛𝐴𝑡 + 𝛼 ∗ 𝐾𝑡 + 𝛽 ∗ 𝐻𝑡 + 𝛽 ∗ 𝐿𝑡 (Equation 3)

Where: Yt is the GDP/capita; At is the amount of money spent on R&D, Kt is calculated


by the net inflows of FDI, Ht is the enrollment in tertiary school and Lt is the Labour
force; α is 1/3 and β is 2/3. The last two approximations come from previous research and
as the endogenous model assumes, this is a Cobb-Douglas function. Here, as mentioned
in the theoretical part, the value of θ is one and therefore, taken away from the equation.
The chosen countries for testing the endogenous growth model on, are Sweden and Nor-
way. The time period is from 1970 – 2011 and the data over these dates is in yearly fre-
quency.
We run three OLS regressions, one with variables from Sweden, one from Norway, and
the third one with both countries simultaneously. In the third case we include a country
dummy in the regression which takes the value of 1 for Norway and 0 for Sweden. Since
we only analyse two countries, we include just one dummy or since a use of two dummies
would lead to a problem called a dummy trap.
The number of observations for the regression for Norway is 37 even if there are 40 years
observed. This comes because for some years, the net FDI inflows for Norway were neg-
ative and one cannot take the natural logarithm of a negative number. Therefore they had
to be excluded from the regression. The number of observations for Sweden is 40 and the
one for the last regression is 75. This is not a very large number of observations, but yet
still above the rule of thumb for the central limit theorem which is around 30 observations.
However, one can say for sure that an increase in the number of observations would lead
to a more clear result due to the higher number of degrees of freedom.
However, there were quite many problems encountered in the data.
The first one we noticed was the extremely high correlation that there was between FDI
and R&D. Therefore we decided to exclude R&D from the regression since that would
lower the effect of multicollinearity and we could draw the same conclusions for R&D
as the ones we observe for FDI. Multicollinearity means that the independent variables
are correlated over time (Guajarati, 2009).
The second common problem was the one of autocorrelation which was present in all
three regressions. Autocorrelation means that the error terms are correlated over time.
One can fix for it by running the regression with the first differences of the variables in-
stead of the one with the variables in their level form. However, after doing that, the in-
dependent variables became insignificant and therefore we decided to not take autocor-
relation into consideration in all three cases. (Guajarati, 2009)
We also tested for misspecification but the F-statistics showed that that was not the
case.

4
The data for FDI, enrollment to tertiary school and GDP per capita came from the
World Bank. However, the data for R&D was collected from NIFU even if it was not
used in the end and the labor force was retrieved from OECD.

4 Analysis
4.1 Descriptive Statistics
Before we analyze the outputs from the regressions, we start by having a look at the de-
scriptive statistics, for both Norway and Sweden, which can be seen in Table 1 and Table
2 in the Appendix.
The values in the descriptive statistics are not logged while the values for the regressions
are in their logged form. Among the observations in the data sets, one can observe varia-
bles taking on various values, for example GDP/capita for Norway has a minimum of
3708 US$ and a maximum of 99143US$. The same variable for Sweden takes on a min-
imum value of 4869 US$ and a maximum value of 56755 US$.
It can be seen that minimum value of GDP/capita is lower in Norway than in Sweden but
the growth in Norway is larger than Sweden. This can most probably be explained by oil
resources that Norway has. Differences can also be seen in other variables as well.
We also conducted a correlation analysis in order to see the variable relationship with
each other. The correlation should range between -1 and 1 with 1 being a perfectly posi-
tively correlated and -1 being perfectly negatively correlated. The results in general
showed relatively high correlation between the regressors.
Graph 1 and 2 in the Appendix show the pattern that the variables have during the time
period tested. It is noticeable that all of them have an upward trend and this is something
which one would expect due to the growth of the economies in general.

4.2 Discussion of results


The endogenous growth model can be interpreted in different ways and other variables
than the ones in our model can both be included and excluded. However, all the included
variables in our regression model are relevant for the endogenous growth theory.
With the help from the theory of the endogenous growth model together with the outputs
from the regressions from Norway and Sweden, we analyze the data and try to see how
well the theory fits the real world data.
As explained in the method part, we run log-log OLS regressions with the GDP/capita as
the regressand and with three regressors trying to fit the theoretical model.
As we ran the three regressions, one for Norway, one for Sweden, and one with data from
both countries with the help of a country dummy. Unfortunately, some statistical prob-
lems were present, such as autocorrelation and multicollinearity.
Equation 1 explains that we assume CRS with a Cobb-Douglas production function. The
indexes used for capital (α) and labor share (β) are respectively 1/3 and 2/3 as assumed
by the Solow Growth Model.

5
In Table 1 below, one can see the regression result for Norway. We tested for heterosce-
dasticity with the Breusch- Pagan-Godfrey test which gave us a low F-statistic. Therefore,
we cannot reject the null hypothesis of homoscedasticity which is good for the model.
We also tested for misspecification in the model with the help of Ramsey Reset test which
showed no sign of misspecification. The problem of this model however is the presence
of autocorrelation and multicollinearity. We tested for autocorrelation with the help of
Breusch-Godfrey Serial Correlation LM test, and some of the problems of autocorrelation
are inefficiency, wrong standard estimates and an inflated R2. When we fix for it, the
variables become insignificant. Therefore we can conclude that the endogenous growth
model does not explain the growth in Norway.
Table 1

Variable Coefficient Std. Error t-Statistic Prob.

C -31.91129 3.481651 -9.165563 0.0000


LOGFDI 0.044175 0.029171 1.514371 0.1395
LOGE 0.201962 0.135364 1.491992 0.1452
LOGL 5.260438 0.520663 10.10334 0.0000

R-squared 0.978143 Mean dependent var 10.11985


Adjusted R-squared 0.976156 S.D. dependent var 0.887807
S.E. of regression 0.137090 Akaike info criterion -1.034554
Sum squared resid 0.620190 Schwarz criterion -0.860400
Log likelihood 23.13924 Hannan-Quinn criter. -0.973156
F-statistic 492.2770 Durbin-Watson stat 0.791685
Prob(F-statistic) 0.000000

There are more factors that one has to take into consideration when analyzing growth. As
mentioned before, the discovering of oil in Norway has played a huge role in their level
of growth. This does not have any significant relationship with education and level of
knowledge. The only variable that is significant before fixing for autocorrelation is the
labor force. Since the endogenous growth model assume that the only growth could come
from innovation and human capital, a large labor force influence it positively. Moreover,
the level by which knowledge increases is proportional to the skilled labor force.
A reason for FDI being insignificant could be that Norway already before the time period
of this paper had a significant growth in GDP. At the same time, Norway is a developed
and rich country and therefore not quite dependent on foreign investors.
However, one would suggest that education would be a significant factor in explaining
development. The insignificance could be solved by using a larger data set. This was
unfortunately impossible for this paper due to the lack of data.
In the methodology part, we say that due to FDI and R&D being strongly correlated, we
analyze the two variables as one. One would however assume that R&D would have a
significant impact on growth. The endogenous growth model uses R&D as one of the
sources of growth since it leads to a growth in knowledge.
Having access to new technology and ideas should increase growth according to theory.
New ideas and technologies should speed up the factor productivity and due to the labor
force being significant, growth should increase.

6
An exogenous model may explain the growth in Norway better than what the endogenous
model does. However, this is something that can be explained in another paper. Norway
is an export driven economy due to its natural resources and that is a highly significant
factor that is not included in the endogenous model.
Table 2

Variable Coefficient Std. Error t-Statistic Prob.

C -46.56119 6.313700 -7.374628 0.0000


LNL 9.878188 1.176256 8.397990 0.0000
LNFDI 0.284285 0.069398 4.096416 0.0002
LNE 0.500419 0.180285 2.775707 0.0087

R-squared 0.922784 Mean dependent var 9.945370


Adjusted R-squared 0.916350 S.D. dependent var 0.640689
S.E. of regression 0.185302 Akaike info criterion -0.439016
Sum squared resid 1.236131 Schwarz criterion -0.270128
Log likelihood 12.78032 Hannan-Quinn criter. -0.377952
F-statistic 143.4091 Durbin-Watson stat 0.565156
Prob(F-statistic) 0.000000

By looking at Table 2 above, one can see the output received from the country of Sweden.
Again we tested for heteroscedasticity, autocorrelation, misspecification and multicollin-
earity. In the case of Sweden, the only problem present was autocorrelation. After fixing
for it by taking the first differentials of all variables, we get insignificance for all regres-
sors. However, we decided to not take that into consideration since the estimates derived
using OLS are unbiased. For Sweden, all the variables in the model are significant at 1%
level. Even though a large R2 is a good sign, having it at this level (0.92) is a result from
autocorrelation, hence, an untrustworthy outcome.
However, one can see that all the variables in the model are significant even at the 1%
level. This means that all the variables in the model are highly explanatory for
GDP/capita. All the regressors show a positive relation to the regressand. It seems that
one unit increase in the labor force would lead to an increase of approximately 9.9 units
increase in GDP/capita.
On the other hand, FDI and education have lower coefficient values. Perhaps this can be
explained by the fact that both education and FDI do not give the results immediately. As
mentioned before, we use the enrollment to tertiary education as a variable and it does
take a couple of years until these students go out in the real life and give their inputs.
The model holds better in Sweden for various reasons. First of all, Sweden does not have
natural resources in the same level as Norway. Therefore, the growth here comes mostly
from the skilled labor force and the increase in education. FDI is also important since
more investment leads to more innovation and therefore more growth. Since Sweden is a
developed country FDI might not play a very large share of growth in GDP. One would
think that total domestic investment would lead to a larger growth, however, this was not
proved in the model. Therefore, we analyze FDI instead.
The same problem as for Norway occurred in this case as well regarding the high corre-
lation between R&D and FDI. Again, we had to exclude one of them from the regression.

7
Therefore, we do expect to receive the same results as those regarding foreign investment.
This is due to the very high correlation the variables had. However, by looking at the
coefficient of FDI, one can see that it is not large. One would probably assume the oppo-
site since Sweden is a capital intensive rather than labor intensive country.
To sum up, it seems that the endogenous growth model fits Sweden better than Norway
for explaining the economic development.
The endogenous growth model implies that knowledge leads to a larger rate of growth.
Therefore, we decided to test for correlation between R&D and education. Since R&D
and FDI were so highly correlated, we use FDI and Education instead. The results for
both countries showed values which were close to one. This proves that a higher share of
education would lead to more resources in research for the case of Sweden and Norway.
This means that even if the model itself did not hold perfectly, a part of it still does no
matter the regression output. Many theoretical models make a lot of assumptions and that
does not always hold for real world data. There are other factors which can influence the
growth of GDP that the endogenous model does not take into consideration. Therefore,
just because the output results were not positive, it does not mean that the model is wrong.
It can just mean that it is incomplete or that it holds for some countries or some time
periods better than the others.
Table 3 in the appendix presents the results received from running the OLS regression
with Sweden and Norway simultaneously. In this case, the problems of multicollinearity
and autocorrelation were present. We used the same method as in the two other models
and it can be seen that this model does definitely not follow the theory. Therefore we
decide to separate the countries rather than take them together even if in this case there
are more observations and therefore more degrees of freedom.
If we decide to not take the statistical issues into account, we get insignificant results for
labor and FDI. The only significant variable is the enrollment into tertiary education.

5 Conclusion
In this paper we test whether the endogenous model holds for the real world data of Nor-
way and Sweden. We ran OLS regressions with GDP/capita as the regressand, labor force,
tertiary education, FDI and R&D as the independent variables. By combining the results
and the theory we came up with the following:
One would assume the endogenous growth model to explain the data better than it actually
does, mostly due to the variables being relevant in the growth of one country. From the

8
cases of Norway and Sweden, we conclude that the theory fits Sweden’s data better than
it does for its neighbor in the west. The years considered are 1971-2011 and during that
time period, both countries experienced swings in the economy.
An explanation for the model’s poor result of Norway could be due to the fact it is an
export driven economy that can rely on its natural resources. The growth of the Norwe-
gian economy came mostly from the oil rather than human capital and R&D.
For Sweden on the other hand, the focus was as well on exports, however, not at all on
the same level as the oil driven Norway. This is why the endogenous growth model shows
better results in this case.
The results seemed not very optimistic from the model with the two countries merged.
Even though one would assume the variables of the regressions to have an effect growth,
it seems like there are other factors excluded from the model need to be included.

9
Appendix
Table 1

Descriptive Statistics Norway

N Minimum Maximum Mean Std. Deviation

41 3708,96983482 99143,1651181 33511,2174117 26252,1308722


GdpCapita
95256 185300 60320000 52117000
41 - 25453237724,0 3899663250,62 6478700383,66
FDI 668114802,999 0000000000000 6578000000000 9277000000000
0610000000000 00
Labor 41 1576,000 2629,000 2154,86588 269,802891
Enrollment 41 15,79487 79,28231 46,5758524 22,22589158
Valid N (listwise) 41

Table 2

Descriptive Statistics Sweden

N Minimum Maximum Mean Std. Deviation

41 4869,61965201 56755,3319349 25090,0574464 13920,4389578


GDPpercapita
4172 43475 5102500 13320000
41 - 23,4248637657 2,76841031984 4,36156645245
FDI ,002008801419 683600 8632 5555
8434
Enrollment 41 21,73277 83,79127 46,1364246 20,85298727
LaborForce 41 3961,0 5016,3 4426,054 250,2318
Valid N (listwise) 41

Graph 1- Sweden
ENROLLMENT FDI
90 25

80 20

70
15
60
10
50
5
40

30 0

20 -5
1975 1980 1985 1990 1995 2000 2005 2010 1975 1980 1985 1990 1995 2000 2005 2010

GDP_PER_CAPITA LABOR_FORCE
60,000 5,200

50,000 5,000

4,800
40,000
4,600
30,000
4,400
20,000
4,200

10,000 4,000

0 3,800
1975 1980 1985 1990 1995 2000 2005 2010 1975 1980 1985 1990 1995 2000 2005 2010

10
Graph 2- Norway
ENROLLMENT FDI
80 3.0E+10

70 2.5E+10

60 2.0E+10

50 1.5E+10

40 1.0E+10

30 5.0E+09

20 0.0E+00

10 -5.0E+09
75 80 85 90 95 00 05 10 75 80 85 90 95 00 05 10

GDP_CAPITA LABOR
120,000 2,800

100,000 2,600

2,400
80,000
2,200
60,000
2,000
40,000
1,800

20,000 1,600

0 1,400
75 80 85 90 95 00 05 10 75 80 85 90 95 00 05 10

Table 3

Variable Coefficient Std. Error t-Statistic Prob.

C 3.654265 1.920109 1.903155 0.0611


LNL 0.256613 0.321035 0.799331 0.4268
LOGE 1.949921 0.127838 15.25304 0.0000
LOGFDI 0.033501 0.088813 0.377213 0.7072
COUNTRY -0.578292 0.301285 -1.919421 0.0590

R-squared 0.835824 Mean dependent var 10.07387


Adjusted R-squared 0.826443 S.D. dependent var 0.727586
S.E. of regression 0.303114 Akaike info criterion 0.514924
Sum squared resid 6.431462 Schwarz criterion 0.669423
Log likelihood -14.30966 Hannan-Quinn criter. 0.576614
F-statistic 89.09299 Durbin-Watson stat 1.184706
Prob(F-statistic) 0.000000

11
REFERENCES
Benhabib. J., & Spiegel, M.M., (1994) “The Role of Human Capital in Economic Devel-
opment- evidence from aggregatecross-country data”, Journal of Monetary Economics,
Vol 34, pp. 143-173, Elsevier Science B.V
Griffith, R., Redding, S., & Van Reenen, J., (2004)” Mapping the Two Faces of R&D:
Productivity Growth in a Panel of OECD Industries”, The Review of Economics and
Statistics, The MIT Press
Gujarati, D.N., & Porter, D.C., (2009) Basic Econometric, Fifth Edition, McGraw-Hill

Jones, Charles I., (1995) “R&D-Based Models of Economic Growth”, Journal of Political
Economy, Vol 103, Nr 4, The University of Chicago Press
NIFU, Retrieved 2012-02-21 from www.foustatistikkbanken.nifo.no/nifu/
OECD, Retrieved 2014-02-20 from www.stats.oecd.org
Regeringen, Tillväxten i Sverige fram till idag, Retrieved 2014-03-02, from:
http://www.government.se/sb/d/3923/a/55727?setEnableCookies=true
Sweden Abroad, “Historisk Bakgrund”, Retrieved 2014-03-02 from:
http://www.swedenabroad.com/sv-SE/Ambassader/Oslo/Landfakta/Om Norge/Histo-
risk-bakgrund/
Tallman, Ellis W., & Wang, Ping., (1994) “Human Capital and Endogenous Growth Ev-
idence from Taiwan”, Journal of Monetary Economics, Vol 34, pp. 101-124, Elsevier
Science B.V
World Bank, Retrieved 2014-02-20, from www.data.worldbank.org

12

You might also like