Is international trade more beneficial for small open developing economies
compared with small open developed economies?
Alexander Zonjee* January 2016
Abstract
The aim of this study is to identify the effects of international trade on income and growth for developing and developed small open economies. This paper analyse whether international trade benefits the developing world more compared with the developed world. To answer that question an empirical model is created to determine the effect of international trade on income and growth. The outcome of the paper is that international trade benefits the small open developing countries more.
JEL Classifications: F10, F43, F63
*Student BSc Economics and Business, specialization Economics and Finance, Faculty Economics and Business, University of Amsterdam, Amsterdam, The Netherlands.
Statement of originality
This document is written by Alexander Zonjee who declares to take full responsibility for the contents of this document.
I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it. The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.
Tables of Contents
1. Introduction……….4 2. Literature Review………5 2.1 Existing Models….……….5 2.2 Economic Growth………...6 2.3 Difference in Trade………..………...7 3. Hypotheses………..………...……….84. Data and Methodology………...………...9
4.1 Data………..……….10
4.2 Methodology………..………...12
4.2.1 The Effects of Trade on Income……….………..12
4.2.2 The Effects of Trade on Growth………..………....……….13
4.2.3 Endogeneity………..…………...……..…………14 5. Statistical Results………..………15 5.1 Descriptive Analysis………...………15 5.2 Empirical Analysis……….………...16 5.2.1 First Hypothesis………..…..…..………..…..………..……....16 5.2.2 Second Hypothesis………....18 6. Conclusion……….……….……...20 7. Limitations……….………....……...21 8. References………..………...23 9. Appendix……….………..25
1. Introduction
International inequality has been a subject of debate for a long time. Economics and trade organizations try to find answers of what is the best strategy to lower international inequality and how to create more overall welfare. According to Irwin & Terviö (2002) trade is a variable that creates a higher income. The income of a country measured in the Gross Domestic Product (GDP) will rise when a nation participates more actively in international trade. When according to Irwin & Terviö a country has more openness to trade, the total income in the country rises.
During the last decades it has been shown that the Brazil, Russia, India and China or also known as the BRIC countries grew very fast towards the developed world and that the average income measured in GDP per capita of the people living in those countries grew faster than the GDP per worker of developed world countries. Increased international trade is one of the reasons why the BRIC countries were able to grow faster. In the last decades the BRIC countries created higher growth rates and they developed themselves from lower-income countries toward middle-lower-income countries.
Rassekh (2007) investigated whether international trade raises income of developing countries more than the income of developed countries. His findings suggest that international trade benefits the lower income economies more than it benefits the higher income economies. In his paper there was no distinction between large open economies and small open economies. His findings suggest that international trade benefits all developing economies more than they do for developed economies. In this thesis this outcome is questioned, because it has been shown that large economies like China, Brazil, Russia and India grew very fast over the last decades. But are the findings of Rassekh not driven by the fact that these large countries create an upward biasfor developing economies? I wonder that large economies create an upward bias for small economies. The outcome of Rassekh that more international trade benefits developing economies more might be wrong for part of the sample. I think that large developing economies started to grow faster compared with small developing economies and that large developing economies therefore created an upward bias for small open developing economies. If the large countries are removed from the sample, it is wondered whether international trade still benefits the small open economies more compared with small open developed economies. Therefore this thesis will re-run the regression of Rassekh, but this time with only small open economies to see whether international trade still benefits developing small open economies more.
This thesis uses a total of 91 small open developing and developed countries in a cross-country regression. It tries to answer the question whether or not international trade benefits developing small open economies more in terms of income and growth. To answer the research question this thesis will conduct a regression analysis.
2. Literature review
The research for the effect of international trade on income and growth is not new. There has been done a lot of research on this subject and many models have been created which all try to answer the question whether international trade improves income and growth. This literature review will start to examine what the existing models tell about the effect of international trade on income and growth. The last part shows how the effects of international trade differ among countries.
2.1 Existing models
One of the first who thought about international trade were Heckscher and Ohlin (1933). Samuelson (1948) wrote a paper where he mathematically described the findings of Heckscher and Ohlin. The model is about factor endowments and tells that countries should produce the product with the factor they are abundant in. The factors that are used are labour and capital. When a country is abundant in labour it is assumed that the trading partner is abundant in capital. To make the model simple, there are only two goods produced. The first goods that is produced, is machinery. The production of machinery is capital intensive and is produced in the country that is abundant in capital. The second product that is produced is cloth. The production of cloth is labour intensive and is produced in the country that is abundant in labour. If nations start opening up to international trade they can gain by producing the good they are abundant in. The nation that is abundant in capital starts producing more machinery and the country that is abundant in labour starts producing more cloth. The inhabitants of each country can consume more when the countries start opening up to international trade, because the total world production of the two goods are higher under free trade. This creates a gain in welfare for both nations.
After opening up to international trade the country that is abundant in labour is producing more cloth and uses more labour. Therefore the demand for labour rises what will lead to a rise in the wages of labour. Higher wages increases the labour to capital ratio in the first country. Simultaneously the price of capital rises in the country that is abundant in
capital. The price of capital rises, because the second country is demanding more capital for the production of more machinery. The increase in the price of capital leads to a decrease in the labour to capital ratio. Before there was international trade, the first country had a low labour to capital ratio and the second country had a high labour to capital ratio. When the countries start opening up to international trade, the labour to capital ratio’s of the two nations grow towards each other and in time these countries are similar in income from labour and income from capital.
This model can be seen as trade between developing and developed countries. Developing nations are abundant in labour and developed nations are abundant in capital. If both nations begin to produce the good they are abundant in and start trading, these nations grow towards each other and international inequality is lowered in time.
Another old model about the income distribution in the world is the catch up theory. This theory was created by early neoclassical economics and it argued that poor countries would grow faster than wealthy countries. According to Solow (1964) the developing world will grow faster because of technological advances and diminishing returns to capital in the latter. Knack and Keefer (1998) explained why poor countries do not catch up with the wealthier countries. According to them the reverse has occurred. They find multiple reasons why poor countries do not catch up with the wealthier countries. Their paper shows that the ability of poor countries to catch up is determined in large part by the institutional environment in which economic activity in these countries takes place. The risk and problems in the developing countries create a situation where there is less economic activity compared with developed countries.
2.2 Economic growth
Frankel and Romer (1999) searched for the relation between international trade and economic growth. According to them a country’s geographic characteristics have important effects on trade. Frankel and Romer created an income regression with geographic characteristics to measure the effect of trade on income. The results show that international trade raises income. The relation between the geographic component of trade and income suggests that a rise of one percentage point in the ratio of trade to GDP increases income per person by at least one-half percentage. According to Frankel and Romer the size of a country does matter for the GDP per worker in a country.
Dollar and Kraay (2001) came with remarkable conclusions. They found a strong positive effect of trade on growth. It is also argued that the post-1980 globalizers are catching
up to the rich countries while the rest of the developing world is falling further behind. Increases in growth rates that accompanies expanded trade leads to proportionate increases in income of the poor. The paper concludes that globalization leads to faster growth and poverty reduction in poor countries.
Edwards (1992) used a cross-country data set to analyse the relationship between trade and growth. Using nine alternative indicators of trade orientation he found that the data supports the view that more open economies tend to grow faster than economies with trade distortion. The results also support the existence of a catch-up effect. Countries that had an initial lower level of income per capita tend to grow faster than other countries. His findings also conclude that there is a negative relationship between instability and growth. The higher the instability in the country, the lower the growth of the country.
Noguer and Siscart (2005) tried to find the effect of trade on income. They re-run the findings of Frankel and Romer (1999) and added more geographical control variables to the equation. Their findings are that geographical controls must enter the income equation to avoid an upward bias on the trade coefficient. They also found that countries that have more trade reach a higher level of income. The results show that a one percent point increase in the trade share leads to a one percent increase in income per capita. This estimate is remarkably robust to the inclusion of a wide array of geographical and institutional controls.
The research of Zhang (2004) is built on the trade model that Frankel and Romer (1999) created earlier. Zhang used a larger dataset compared with Frankel and Romer. Zhang tried to find the differences between import- and export openness to trade. His first finding was that the total exports plus imports over GDP have a positive effect on income. This is in line with all earlier studies about the effect of trade on income. Separating import- and export openness show that only export openness is positively associated with income levels. The effect of export openness to income was statistically significant. The effect of import openness to income was negative, but not statistically significant. Therefore Zhang concluded that only export openness raises income.
2.3 Different effects on international trade
Paudel (2003) found that countries that are landlocked have up to 80% less international trade compared with countries that have direct access to the sea. The main reason why these countries trade less with other countries is because of the high transportation cost of land transport compared with sea transport. The effect of being landlocked is even
larger for developing countries. Being further away from the nearest sea is also an indicator of a lower trade ratio.
Another reason why trade is different between countries comes from Streeten (1993). He argues that there are large differences between small and large economies. Small economies have most of the time less diversity in raw materials and natural resources. Many of them are also landlocked. Most of these countries have to specialize in only a few goods. And at last, international trade is more important for small economies because this trade is the only possibility to gain benefits from economies of scale. Therefore they are more vulnerable for a world economic downturn, because they are producing only a few goods what makes them more vulnerable compared to large diversified economies.
3. Hypotheses
After revising the existing literature, some questions arise. Rassekh (2007) argue that trade is more beneficial for developing countries compared with developed countries. The effects of trade for developing countries outweigh the trade effects for developed countries. Rassekh’s sample consists of 150 countries for the year 1985. The sample consists of large open economies and small open economies. According to Streeten (1993) there are differences between small open economies and large open economies. Small open economies are more dependent on international trade because the total population in the country is smaller and with international trade they can still benefit from economies of scale. Therefore, the question arises whether small open developing economies compared with small open developed economies still show higher effectiveness of trade. This is done in a cross-country dataset to see whether the positive trade effect that Rassekh found for developing economies still occurs for small open developing economies. And if it occurs, does it occur in the same size? It may also happen that the trade effect is a lot larger or smaller for small open developing economies compared with small open developed economies.
The second reason why there might be differences in the effect of trade between small open economies and large open economies is by the theory of Heckscher-Ohlin. In this theory it is stated that small open economies cannot determine the world market price. The world market price is the price for which a nation can trade their goods on the world market. A large open economy can often influence the world market price. Fluctuations in prices of the exporting goods have huge effects on the income of a small open economy. That is because they are more specialized in a few products compared with large open economies. These large
open economies are more diversified and therefore are less dependent on the export of a few goods.
H0: The effect of international trade on income does not depend on the economy’s level of development.
H1: The effect of international trade on income does depend on the economy’s level of development.
The first hypothesis tests whether an increase in international trade leads to a higher level of income for small open developing economies compared with small open developed economies. If the zero hypothesis is rejected, than an increase in the trade share is more effective for small open developing economies.
The second hypothesis that is tested, is whether an increase in the trade share generates more economic growth for small open developing economies compared with small open developed economies.
H0: The effect of international trade on growth does not depend on the economy’s level of development.
H1: The effect of international trade on growth does depend on the economy’s level of development.
If from the data follows that trade creates more economic growth for small open economies, than it can be concluded that small open economies should trade more to increase their growth perspectives. The conclusion of Rassekh (2007) was that developing countries have a higher growth rate than developed countries. But the sample of Rassekh included the large economies as well. Therefore it is questionable that small open developing economies grow at a faster pace than small open developed economies do.
4. Data and Methodology
In order to accomplish the stated hypotheses, this thesis uses a dataset of 91 countries on international trade, income and growth of developing- and developed small open economies. In the following, it is described which dataset will be used in this thesis. This is followed by the explanation of the models used. Finally, the problem of endogeneity will be discussed.
4.1 Data
This thesis is focussing to answer the question whether small open economies react differently to international trade while being developed or developing. Therefore only small open economies are used. Frankel and Romer (1999) and Rassekh (2007) used a dataset of 150 open economies for the year 1985. That dataset is containing small and large open economies. The same dataset will be used for this thesis, because this thesis wants to develop the findings of the former paper of Frankel and Romer and the paper of Rassekh. The large open economies will be removed from the dataset to make a dataset that contains only small open economies.
There are large differences between small- and large open economies. This thesis determines whether an economy is small or large on the basis of three criteria. The countries that are considered as small in this dataset are small in the working population. The country will be considered small if the total working population is less than ten million people. The larger the population, the more a firm can trade within the country. That makes them less dependent on trade with other countries. When having a smaller population makes the country more dependent on international trade to be able to gain from economies of scale.
The second determinant for a country being small or large is whether the country is world dominant in the production of a product and being a price setter in that good. Considering the book of Krugman, Obstfeld and Melitz (2015) a small economy has a given world market price. Changing the output of the product in a small economy will not affect the price at the world market, because it is considered that a small economy produces a small amount of the total world demand of a good. A large country produces a large amount of the good and fluctuations in the output of a large country will affect the world supply.
The last determinant from being a small open economy is the word open. Being a small country does not mean that the country is an open economy. To be considered as open, the country needs to have a certain amount of trade to GDP. If the trade to GDP is low, the country in question is considered as a closed economy. Being a closed economy means that trade takes place mainly within the country and not outside of the country. All the countries with a low openness to trade are removed from the dataset. The minimum level of openness to trade to be considered as an open economy is 50% of GDP.
From the 150 countries in the sample of Rassekh (2007) that had an economically active population of more than ten million people were removed from the sample. Secondly, the countries that were price setters in highly demanded products were removed from the sample. For this reason Australia and Saudi-Arabia are removed from the sample. Australia is
the largest producer in iron. This is one of the most important natural resource in the modern world. Changing supply of iron in Australia makes the world demand for iron change. Saudi-Arabia is removed from the dataset because Saudi-Saudi-Arabia is the largest producer in crude oil in the world. Again, changing supply of crude oil in Saudi-Arabia will affect the total world demand and supply. Therefore according to the book of Krugman, Obstfeld and Melitz (2015) these two countries are no longer considered as small and therefore they are no longer valid for this dataset.
The last determinant that is used to distinguish a closed economy from an open economy was the required level of trade of at least 50% of GDP. The countries with less than 50% openness to trade are removed from the sample, because these countries are considered as small closed economies. The countries that are removed are Algeria, Ecuador, Iraq and Somalia. After this separation the dataset contains 91 small open economies. The countries that are used are in table two of the appendix.
At last, the countries needed to be listed as developed or as developing. This is done at the same way as Rassekh (2007) have done this. He took the median of the 150 countries and classified the 75 countries under the median as developing and the 75 countries above the median as developed. To make the outcome of Rassekh and this thesis comparable, the countries that were classified as developing in the sample of Rassekh are also classified as developing in this sample. After removing the invalid countries, the dataset contains 46 developing countries and 45 developed countries. From the 59 countries that were removed, 29 were developing and 30 were developed.
The second regression needs to have data from the GDP per worker from the year 1960. This thesis uses the year 1960 because that was 25 years earlier. To measure economic growth you want to have a large timeframe, because in a small timeframe there would be less economic progress. In a longer timeframe a country can develop and show a lot of progress. This is less likely in a smaller timeframe. Not all the countries have got the Gross Domestic Product per worker for the year 1960. Because of data limitations, the total sample size for the second regression is reduced from 91 to 71 countries. In this sample there are 35 developing countries and 36 developed countries. Again it was needed to remove one more developing country from the sample. There was a reason why not an equal amount of developed- and developing economies were removed. That was because I did not want to change the classification of an economy and I did not want to remove more countries than necessary. The countries that were classified as developed in the paper of Rassekh (2007) are also classified
as developed in this dataset to compute the GDP per worker and the growth in GDP. This is done to make the best possible comparison between this thesis and the paper of Rassekh.
4.2 Methodology
4.2.1 Effects of trade on income
The first regression equation is:ln Yi,1985 = β0+ β1∗ D + β2∗ Ti+ β3∗ D ∗ Ti+ β4∗ lnNi+ β5∗ lnAi+ β6
∗ Landlocked + ∈
In this equation Yi will be the Gross Domestic Product per worker in the economy i.
The total GDP is the total income of a country plus net interest payments from abroad. Dividing this by the number of workers in the country generates the average income of a worker in that country.
D is a dummy variable for if the country is considered to be developing. This determination is done in the same way as in Rassekh (2007). This dummy is zero if a country is developed and the dummy is one if a country is developing
Ti is the measure for the trade openness of a country. It is measured as the total
imports plus exports divided by the GDP. By dividing with the GDP it is shown what happens with the openness to trade of a country when the import, exports or GDP grows. If the GDP grows faster than trade, than the openness of the country to trade is lowered. If trade grows faster than the GDP, than the openness rises. International trade is calculated as exports plus imports. Either import or export can make the openness to trade variable rise.
Ni is the measure of total population living in the country. If Ni is larger it is assumed
that trading within the country is larger. It is harder to generate economies of scale within a country when the population of the country is smaller. Therefore small countries have to trade more with other nations to create more economies of scale.
Ai is the size of the country. If a country is larger, it is harder to trade with another
country and also within a country. The amount of trade per square mile is smaller when a country is larger with the same population.
To make a valid income model Noguer and Siscart (2005) show that geographical controls must enter the income equation to avoid a bias. Larger countries show more within country trade and less trade with other nations. Therefore the openness to trade is smaller compared with larger countries. Since this thesis wants to measure the benefits from international trade, it is needed to control for the size of the country, which is easiest
measured by area and population. Not controlling for these variables creates a bias and makes a valid outcome less likely.
Landlocked is the second dummy variable. This counts for one if a country is landlocked and zero if the country has free access to the sea. This is important because being landlocked makes it harder for a country to sell their goods abroad. Trading over land is more expensive than trading over sea. Therefore landlocked countries might trade less, because it is too expensive for them to sell their products abroad. For landlocked countries it is also more expensive to ship through other countries. According to Paudel (2014) Landlocked countries trade less compared with countries that are not Landlocked.
The term Landlocked is the only term that is new in the equation compared with the former paper of Rassekh (2007). The regression will be tested without the term landlocked first. The second time, the regression is re-done with the term Landlocked to see whether this term should be part of the regression equation.
Increases in the trade share are more effective for small open developing economies when Beta 3 is statistically different from zero. If Beta 3 is statistically significant, than an increase in the trade share benefits small open developing economies more. This is because Beta 3 only captures the effect of trade for developing countries.
4.2.2 Effects of trade on growth rates
The regression to generate the outcome of the second hypothesis will look as follows: ln Yi,1985− lnYi,1960 = β0+ β1∗ D + β2∗ Ti+ β3∗ Ti∗ D + β4∗ lnNi+
β5∗ lnAi+ β6∗ Landlocked − α ∗ lnYi,1960
The equation is very similar to the first equation. Adding the term GDP per worker from 1960 is the only difference. The formulation of the estimated model is built on the neoclassical Solow model from Mankiw et all. (1992). The coefficient α tells that when the GDP per worker is larger in 1960, the growth will be lower. Developed countries in 1960 will grow at a slower pace, because they already have a high stock of capital. Developing economies will grow faster, because the capital stock is low and the marginal gain of an increase in capital is a lot larger for these developing countries.
The logarithm of the GDP per worker in 1985 minus the GDP per worker in 1960 is the growth of the economy. This thesis is investigating whether Beta 3 is statistically significant different from zero. If Beta 3 is statistically significant different from zero, than an
increase in the trade share for developing economies has a larger positive effect for small open developing economies compared with small open developed economies.
4.2.3 Endogeneity
The problem with the two equations above is the endogeneity. Former papers all tried to solve this problem, but eventually all failed by creating a valid instrument. Endogeneity occurs when the dependent variable correlates with the error term. Stock and Watson (2015) explains that this problem can be solved with Instrumental Variables. A valid instrument does correlate with the independent variable, but does not correlate with the dependent variable. A good instrument creates smaller deviations from the estimated regression line.
In this regression the gross domestic product per worker is possibly affecting the trade share as well. A country with a higher GDP per worker will likely trade more. According to Fink et all. (2005) the reason why richer countries trade more, is because richer countries trade more with other rich countries. This is due to a closer match of exporter supply structure and import preferences.
Frankel and Romer (1999) tried to solve the problem of endogeneity with the use of instrumental variables. They used the gravity model to construct a variable that does correlate with the trade share, but does not correlate with the GDP per worker. They thought that distance between two countries, size of the countries, landlocked, and sharing the same border are determinants of trade between two countries. After re-running the regression with Instrumental Variables they found that the estimate of the trade share is considerably larger than the estimate of the trade share with the ordinarily least squares regression. An increase of one percent point in the trade share had a larger effect on GDP per worker in the regression with Instrumental Variables.
Rodriguez and Rodrik (2001) immediately refute these findings. They are concerned that a geographical constructed trade share is not a valid instrument. The reason is that geographic factors are likely to be determinants of income through a multitude of channels, of which trade is only one. Geography affects public health through exposure to various diseases. It influences the quality of institutions through the historical experience of colonialism, migrations, and wars. It determines the quantity of quality of natural endowments, including soil fertility, plant variety, and the abundance of minerals. The geographically determined component of trade may be correlated with all these other factors, imparting an upward bias to the Instrumental Variable estimate.
Rodriguez and Rodrik (2001) re-run the same dataset from Frankel and Romer in the OLS and the IV regression and added binary variables to the equation to show that the IV regression made by Frankel and Romer (1999) is not correct. Adding these binary variables should not affect the significant of the trade share by a lot.
Rodrigues and Rodrik (2001) re-run the regression three times with three different binary variables from former papers. These former papers all created a regression equation of trade on GDP. The first time they added distance from the equator from Hall and Jones (1998) to the OLS and the IV regression. The second time they added the percentage of a country’s land area that is in the tropics from Radelet, Sachs and Lee (1997). In the third regression they added regional dummies. The findings were that the IV estimate went from significant to insignificant when these binary variables were added. The estimates of the trade shares under IV were even lower compared with the OLS estimates of the trade share. Therefore they argue that the IV findings from Frankel and Romer are wrong and create an upward bias.
Other instrumental variables are all correlated with the error term. There is no instrument that correlates with international trade and does not correlate with the GPD per worker. Therefore there will always be a possibility of endogeneity in this regression. This thesis uses an OLS regression because using a wrong instrument creates an upward bias in the estimate of the trade share.
5. Statistical Results
In this chapter first there will be a brief discussion about the descriptive analysis. The second part of this chapter will describe the empirical results of this research.
5.1 Descriptive Analysis
Table 1 in the appendix, summarizes the descriptive statistics of the important variables concerning the GDP per worker and growth in GDP. The average of the GDP per worker in the year 1985 is 11139 Dollar. The standard deviation is almost as large as the mean. That means that the sample with 91 countries has a wide spread and the differences in GDP per worker are large between countries. The GDP growth shows that on average the countries in the sample grew by 76 percent during 1960 and 1985. But there are exceptions. The economy of Chad reduced by 40 percent during the timeframe of 1960 till 1985. This was the largest reduction. There are only four countries in this sample where the GDP per worker reduced during this timeframe.
5.2 Empirical Analysis
5.2.1 The First hypothesis
Table 3 with regression one, two, three and four in the appendix and the table at the end of this sub-chapter, show the outcome of the first hypothesis. The first regression is similar to the regression that is done by Rassekh (2007). This is the OLS regression in its simplest form. Here the natural logarithm of GDP per worker in 1985 is regressed on the trade share, natural logarithm of the population and natural logarithm of area. The outcome shows that increasing the trade share by one percent point, the GDP per worker increases with 0.50% keeping the other variables constant. This outcome is significant at a 5% level. If this outcome is compared with the outcome of Rassekh, a drop is shown in the effectiveness of the trade share. Increasing the trade share by one percent point in the complete sample of Rassekh generates an increase in GDP per worker of 0.85 percent.
Area has a negative effect on the GDP per worker. If Area is increased with 1%, the GDP per worker is reduced with 0.14%. This outcome is significant at a 1% level. This can be explained as trading within the country and with other countries is harder when the people live further away from each other. A larger distance to travel will reduce trade and therefore it will lower the GDP per worker. The coefficient of the population variable is 0.18. This is significant at a 1% level. With a higher population firms can sell more goods within the country and generate more income per worker.
Adding the term landlocked to the equation increases the R2 of this simplest regression by a lot. This is the second regression. Landlocked is significant at a 5% level. This is all conforming the expectations. Rassekh did not add this term to the regression, but for this data it is shown that Landlocked should be part of this regression equation.
In the third regression a distinction is made between developing and developed countries. After making the distinction between developed and developing countries, it is shown that there are differences between these classifications. The estimate of the trade share for developed countries is a lot smaller compared with the estimate for developing countries.
For developing countries an increase in the trade share of one percent point generate an increase in GDP per worker of 0.74 percent. This increase is calculated by adding trade and trade*developing together. The effect of trade plus trade*developing is tested for zero. In this way it can be tested weather this effect is different from zero. If the estimate is different from zero, than there will be an effect on GDP per worker when the trade share rises. The outcome is significant at a five percent level (t=2.56). Therefore an increase in the trade share
of one percent point generates a statistically significant increase in the GDP per worker. Area and population are also significant at a five percent level in. Making this classification in the regression improves the regression a lot. The R2 of the regression increases from 0.198 in the simplest model to 0.747 in the second regression model.
Comparing the third regression with the output of Rassekh (2007) it is shown that the trade share has a lower effect on GDP per worker for small open economies. According to the output of Rassekh a one percent point increase in the trade share, increases the GDP per worker for developing countries by 1.35. For small open economies this effect is reduced to 0.74. Increasing the trade share is less effective for small open economies than for large open economies because the estimate reduced a lot. But the reduced dataset predicts the effect of trade on GDP per worker for small open economies better. This is because the R2 increased a lot compared with the complete sample of Rassekh. The R2 of the dataset with 150 countries was equal to 0.53, while in the sample with only small open economies the R2 is equal to 0.75. So this regression gives lower estimates, but these lower estimates fit the model better for small open economies compared with the full data sample.
Adding the term Landlocked creates the fourth equation. Adding Landlocked makes the regression better through a higher R2, but this increase is not as large as the increase in R2 from the first regression to the second regression. The term landlocked is significant at a ten percentages level.
From the third and fourth regression it is shown that trade is more effective for small open developing economies. The outcome is that trade for developing small open economies is highly significant at a one percent level. Therefore the zero hypothesis is rejected. The effectiveness of international trade depends on whether an economy is a developing economy or a developed economy.
Dependent/ Independent (1) Ln (gdp) per worker (2) Ln (gdp) per worker (3) Ln (gdp) per worker (4) Ln (gdp) per worker Trade 0.5049** (0.2463) 0.5371** (0.2584) 0.0093 (0.1415) 0.0387 (0.1539) Developing -2.2379*** (0.2787) -2.1648*** (0.2637) Trade*Developing 0.7344*** (0.2646) 0.7101*** (0.2480) Ln (Area) -0.1476*** (0.0467) -0.1088** (0.0494) -0.0714** (0.0319) -0.0579* (0.0327) LN (population) 0.1813*** (0.0424) 0.1840*** (0.0375) 0.0645** (0.0266) 0.0694** (0.0269) Landlocked -0.7121** (0.2842) -0.2936* (0.1678) Constant 9.7903*** (0.5971) 9.5212*** (0.6208) 10.3502*** (0.4162) 10.2210*** (0.4263) Countries 91 91 91 91 R2 0.1979 0.2692 0.7471 0.7587
* 10% significant, ** 5% significant, *** 1% significant, (..) standard deviation.
5.2.2 The second hypothesis
In table 3 regression five, six, seven and eight of the appendix and at the end of this sub-chapter, show the outcome of the second hypothesis. The fifth regression shows that the coefficient of trade is 0.31 for the sample of 91 small open economies. This is significant at a 5% level. The outcome of 0.31 is also quite similar to the outcome of Rassekh (2007). He founded that a one percent point increase in the trade share increases growth in the full sample by 0.34 percent.
Adding the term Landlocked does not change the outcome of the full sample regression. The term Landlocked is not statistically significant. Small open landlocked economies do not show lower growth potentials.
In the seventh equation the distinction between developed and developing has been made. A one-percentage point increase in trade share generates a 0.16 percent increase in growth for developed countries. This outcome is significant at a five percent level. The total
effect for developing countries is calculated as Trade plus Trade*Developing, because both estimates captures the effect of higher international trade. The total effect for developing countries is tested for zero. The total effect of trade for developing countries is 0.636 (t=2.80). That outcome is significant at a one percent level.
Adding the term Landlocked to the seventh equation does not make a difference. The term Landlocked is not significant. Landlocked countries do not have a significantly lower growth rate than not landlocked countries.
The coefficient of the trade share on growth for small open developing economies is 0.636, while in the complete sample of Rassekh (2007) this coefficient was 0.86 for developing countries. Again the effect of trade is smaller for small economies. But the R2 of the data set of Rassekh with 150 countries was 0.24. The R2 increased in the reduced sample to 0.45. So the estimate of trade is smaller for small open economies, but the reduced model fit the data better.
To answer the hypothesis whether international trade is more effective for small open developing economies compared with small open developed economies, the zero hypothesis has be rejected. The coefficient for trade*developing is significant at a five percent level. Therefore developing countries do benefit more from international trade. The growth rates of developing countries rise more than the growth rates of developed countries when trading more. The growth in the GDP per worker increases when a developing country has a higher openness to trade.
Dependent/ Independent (5) Growth in gdp per worker (6) Growth in gdp per worker (7) Growth in gdp per worker (8) Growth in gdp per worker Trade 0.3082** (0.1203) 0.3182*** (0.1192) 0.1557** 0.0779 0.1646** (0.0760) Developing -0.9662*** (0.1879) -0.9632*** (0.1882) Trade*Developing 0.4803** (0.2033) 0.4804** (0.2078) Ln (Area) -0.0571 (0.0350) -0.0525 (0.0339) -0.0457 (0.0300) -0.0418 (0.0291) LN (population) 0.0742** (0.0365) 0.0739** (0.0367) 0.0730*** (0.0260) 0.0729*** (0.0262) Landlocked -0.0960 (0.1250) -0.0838 (0.1041) Ln(GDP1960) -0.0348 (0.0402) -0.0441 (0.0381) -0.2654*** (0.0605) -0.2729*** (0.0579) Constant 1.1268** (0.4878) 1.1733** (0.4982) 3.3358*** (0.6041) 3.3657*** (0.6021) Countries 71 71 71 71 R2 0.2385 0.2464 0.4500 0.4560
* 10% significant, ** 5% significant, *** 1% significant, (..) standard deviation.
6. Conclusion
This thesis questioned the outcome of Rassekh (2007) for small open economies. It wondered whether looking at small open economies only would make the outcome different from his findings. To make the best comparison possible, this thesis used a reduced dataset from Rassekh. This dataset contained only the small open economies from the dataset that Rassekh used. Streeten (1993) argued that small open economies are more dependent on trade to benefit from economies of scale. The effect of international trade for small open economies is different than for large open economies, because small open economies are more dependent on it. Large economies can generate economies of scale within the borders and are less dependent on other countries.
The first hypothesis in this thesis examined whether international trade has a higher effectiveness on the GDP per worker for small open developing economies compared with small open developed economies. The outcome shows that increases in the trade share are more effective for small open developing economies. This finding is in line with the finding of Rassekh (2007). The data does show that the outcome of the complete dataset is a lot higher for developing economies compared with the reduced dataset with only small open economies. The effectiveness of trade in the complete dataset of Rassekh was equal to 1.35, while in the reduced dataset with only small open economies this was only 0.74. Therefore small open developing economies have to create more openness to trade compared with large developing economies to benefit the same output gains.
The second hypothesis in this thesis questioned whether trade generate higher growth for small open developing economies compared with small open developed economies. Again, the zero hypothesis has to be rejected and the alternative hypothesis is accepted. At a five percent level trade is more beneficial for the growth of small open developing economies compared with small open developed economies. Creating a higher openness to trade increases the economic growth of a country.
Overall this thesis can conclude that small open developing economies benefit more from international trade than small open developed economies do. This is also in line with the outcome of Rassekh (2007). Rassekh used a dataset of 150 countries and the sample of this thesis was the same dataset, but a reduced one to only the 91 small open economies. The estimates that were found in this thesis are lower compared with the outcome of Rassekh. Therefore separating economies and by only looking at small open economies gives lower estimates, but these estimates better fit the model, because the R2 of the reduced dataset is higher compared with the dataset of the full dataset.
Creating more trade openness for small open developing economies makes the economies better off, because the total GDP per worker grows and economic growth is higher. Trading more can help small developing open economies to catch up faster.
7. Limitations
This thesis is limited in several ways. The largest problem is the problem of endogeneity between trade and GDP. This thesis did not find or create a valid instrument. A new research should try to find a valid instrument.
The other limitation of this study was the designation ‘developing’. This was done according the same way as Rassekh (2007). The top 75 countries in his sample were considered as developed and the bottom 75 countries were considered as developing. This is a rough way of saying a country is developing or not and it does not look at the potential of the country, growing pace or liveability in the country. A more sophisticated designation could be used.
8. References
Dollar, D. & Kraay, A. (2004). Trade, Growth and Poverty. The Economic Journal 114, F22-F49
Edwards, S. (1992). Trade Orientation, Distortions and Growth in Developing Countries.
Journal of Development Economics 39 (1), 31-57
Fink, C., et All (2005). Income-Related Biases in International Trade: What Do Trademark Registration Data Tell Us? Review of World Economics 141 (10), 79-103
Frankel, J.A. & Romer, D. (1999). Does Trade Cause Growth?. The American Economic
Review 89 (3), 379-399
Hall, R. E. & Jones, C.I. (1999). Why do some countries produce so much more output per worker than others? Quartely Journal of Economics 114 (1), 83-116
Irwin, D.A. & Terviö, M. (2002). Does Trade Raise Income? Evidence from the Twentieth Century. Journal of International Economics 58 (1), 1-18
Knack, S. & Kneefer, P. (1998). Why Don’t Poor Countries Catch Up?. Economic Inquiry 35, 590-602
Krugman, P.R., Obstfeld, M. and Melitz, M.J. (2015), International Economics. Theory &
Policy, 10th edition, Pearson Global Edition.
Mankiw, N.G., Romer, D. & Weil, D. (1992), A Contribution to the Empirics of Economic Growth. Quarterly Journal of Economics 107, 407-437
Noguer, M. & Siscart, M. (2005) Trade Raises Income: a Precise and Robust Result. Journal
of International Economics 65 (2), 447-460
Paudel, R.C. (2014). Economic Growth in Developing Countries: Is Landlockness Destiny?.
Economic Papers 33 (4), 339-361
Penn World Tables (Mark 5). An Expanded Set of International Comparisons, 1950-1988
Received at: http://datacentre.chass.utoronto.ca/pwt56/
Radelet, S., Sachs, J.D. & Lee, J.W. (1997) Economic growth in Asia. Harvard institute for International Development. Development Discussion Paper.
Rand McNally. Quick reference world atlas. Chicago: Rand McNally, 1993
Rassekh, F. (2007). Is International Trade More Beneficial to Lower Income Economies?. An Empirical Inquiry. Review of Development Economics 11, 159-169
Rodriguez, F. & Rodrik, D. (2001). Trade Policy and Economic Growth: A Skeptic’s Guide to the Cross-National Evidence. National Bureau of Economic Research
Samuelson, P. (1948). Trade and the Equalization of Factor Prices. The Economic Journal 58
(230), 161-184
Stock, J.H. & Watson, M.W. (2015). An Introduction to Econometrics, 3th edition, Pearson Education.
Streeten, P. (1993). The Special Problems of Small Countries. World Development 21 (2), 197-202
Zhang, S. (2004). The Link between Trade and Income: Export Effect, import Effect, or Both?. New York
9. Appendix
Table 1
Variable Observations Mean Std. Dev. Min Max
GDP 1985 91 11139.13 10021.62 940 38190 GDP growth 71 0.5677 0.4188 -0.5197 1.7136 Trade 91 97.53 43.06 51.27 318.07 Population 91 1.859 2.060 0.029 8.262 Area 91 84.863 132.448 0.122 604.829 GDP 1960 71 5631.761 5301.466 576 21285 Developing 91 0.5055 0.5027 0 1 Landlocked 91 0.1978 0.4005 0 1
GDP 1985 & GDP 1960 measured in international prices (1985). GDP growth is the logarithm of the growth in percentage.
Trade is the share of trade measured in percentage. Population is the population in millions.
Area is the size of a country in square miles.
Developing if country was in the lower 75 in the sample of Rassekh (2007). Landlocked if a country does not have free access to the sea.
Table 2 Country Trade Share Area Pop. GDP 1985 GDP 1960 Land- locked Deve- loping Angola 69.1 481.354 3.512 1742 1935 0 1 Benin 79.99 43.483 1.874 2391 1969 0 1 Botswana 121.28 231.800 0.37 6792 1224 1 1 Burkina Faso 52.42 105.870 4.15 940 776 1 1 Cameroon 57.67 183.569 3.831 3869 1348 0 1 Cape Verde 118.02 1.557 0.12 2829 1385 0 1 CAR 65.43 241.313 1.309 1266 1166 1 1 Chad 61.43 495.755 1.791 1146 1927 1 1 Comoros 67.06 0.863 0.181 1400 1084 0 1 Congo 112.81 132.046 0.76 6878 2494 0 1 Djibouti 117.06 8.958 0.105 4647 - 0 1 Gabon 100.18 103.346 0.42 9672 3535 0 0
Country Trade Share Area Pop. GDP 1985 GDP 1960 Land- locked Deve- loping Gambia 89.14 4.093 0.358 1609 1142 0 1 Guinea 71.8 94.926 2.243 1583 928 0 1 Guinea-Bissau 62.74 13.948 0.425 1354 927 0 1 Ivory Coast 78.19 124.502 4.03 3740 2028 0 1 Kenya 51.69 224.96 7.98 2014 1451 0 1 Lesotho 152.42 11.72 0.743 2018 576 1 1 Liberia 79.63 43 0.811 2312 1705 0 1 Malawi 54.09 45.747 3.18 1171 765 1 1 Mali 73.6 482.077 2.332 1686 1506 1 1 Mauritania 141.56 397.953 0.533 2674 2128 0 1 Mauritius 109.1 0.787 0.577 7474 5974 0 0 Morocco 58.5 172.413 6.714 6427 2835 0 1 Namibia 119.81 317.818 0.38 8465 4900 0 0 Niger 51.27 489.206 3.343 1098 902 1 1 Reunion 52.14 0.969 0.216 7858 3669 0 0 Senegal 70.63 75.954 2.758 2688 2164 0 1 Seychelles 111.95 0.175 0.029 7058 2524 0 1 Swaziland 118.71 6.704 0.277 5225 2495 1 1 Togo 105.52 21.925 1.277 1516 792 0 1 Tunisia 71.33 63.379 2.28 8783 3931 0 0 Zambia 76.96 290.586 2.274 2399 2662 1 1 Zimbabwe 56.4 150.699 3.135 3261 2241 1 1 Bahamas 124.11 5.382 0.097 29815 - 0 0 Barbados 130.3 0.166 0.127 12212 6724 0 0 Belize 183.27 8.866 0.049 8487 - 0 0 Costa Rica 63.19 19.652 0.92 9148 6830 0 0 Dominica 103.09 0.305 0.03 6163 0 1 Dominican Republic 63.24 18.704 1.912 7082 4130 0 1 El Salvador 52.21 8.26 1.564 5547 4371 0 1 Grenada 120.63 0.133 0.039 4502 - 0 1
Country Trade Share Area Pop. GDP 1985 GDP 1960 Land- locked Deve- loping Honduras 54.15 43.277 1.307 4652 3268 0 1 Jamaica 131.89 4.411 1.059 4726 4338 0 1 Panama 70.96 29.761 0.76 10039 4739 0 0 Puerto Rico 136.73 3.515 1.101 21842 11388 0 0 St. Lucia 165.77 0.238 0.057 5317 - 0 1 St. Vincent 152.17 0.15 0.042 5796 - 0 1 Trinidad 61.9 1.98 0.441 25529 16901 0 0 Chile 53.85 292.132 4.303 9768 8756 0 0 Guyana 109.95 83 0.28 3573 5608 0 1 Suriname 82.99 63.251 0.124 10883 7110 0 0 Bahrain 188.7 0.24 0.178 22840 - 0 0 Bhutan 62.54 17.954 0.575 1504 - 1 1 Hong Kong 209.52 0.398 3.516 16447 4172 0 0 Israel 85.8 8.02 1.602 21953 9685 0 0 Jordan 113.5 37.297 0.601 15655 4488 0 0 Kuwait 96.45 6.88 0.64 35065 - 0 0 Malaysia 104.69 128.328 6.217 10458 4110 0 0 Mongolia 82.72 604.829 0.894 3966 - 1 1 Oman 87.06 82.03 0.368 31609 - 0 0 Qatar 80.94 4.412 0.166 36646 - 0 0 Singapore 318.07 0.22 1.189 17986 5008 0 0 Sri Lanka 62.93 25.332 5.786 5597 3508 0 1 Taiwan 94.62 13.895 8.262 12701 3374 0 0 UAE 89.66 32 0.694 38190 - 0 0 Austria 81.27 32.375 3.528 23837 10713 1 0 Belgium 151.34 11.781 4.071 27325 14310 0 0 Bulgaria 85.99 42.823 4.417 9662 - 0 0 Cyprus 107.57 3.572 0.31 13918 4967 0 0 Czechoslovakia 69.45 49.383 8.137 7467 3328 1 0 Denmark 72.99 16.631 2.78 23861 14807 0 0 Finland 57.5 130.119 2.493 23700 11577 0 0
Country Trade Share Area Pop. GDP 1985 GDP 1960 Land- locked Deve- loping Greece 53.97 50.961 3.8 16270 5151 0 0 Hungary 82.32 35.92 5.195 10827 - 1 0 Iceland 81.83 39.709 0.127 23256 12585 0 0 Ireland 118.84 26.2 1.342 19197 8391 0 0 Luxembourg 211.94 0.999 0.157 30782 18769 1 0 Malta 160.86 0.122 0.119 15380 4737 0 0 Netherlands 118.76 16.041 5.855 28563 17117 0 0 Norway 86.00 125.049 2.043 28749 14291 0 0 Portugal 77.95 35.55 4.54 11343 4853 0 0 Sweden 69.02 173.8 4.238 26504 17352 0 0 Switzerland 77.69 15.941 3.222 29848 20149 0 0 Fiji 89.13 7.078 0.232 9840 7607 0 0 New Zealand 65.25 103.884 1.438 26039 21285 0 0 Papua New Guinea 94.52 178.704 1.66 3374 2270 0 1 Solomon Islands 123.6 10.954 0.88 5109 - 0 1 Tonga 102.25 0.288 0.03 6022 - 0 1 Vanuatu 123.33 4.707 0.042 5707 - 0 1 Western Samoa 92.17 1.093 0.05 5388 - 0 1
Trade share is import + exports to GDP, 1985 (Penn World Table, Mark 5.6, Series OPEN) Area is from Rand McNally (1993), measured in square miles
Population is economically active population, 1985 (Penn World Table, Mark 5.6)
Income per worker is real GDP per worker, 1985; 1985 international prices (dollars) (Penn World Table, Mark 5.6, Series RGDPCH)
Table 3 * 10% significant, ** 5% significant, *** 1% significant, (..) standard deviation. Dependent / Independent (1) Ln(gdp) per worker (2) Ln(gdp) per worker (3) Ln(gdp) per worker (4) Ln(gdp) per worker (5) growth in gdp per worker (6) growth in gdp per worker (7) growth in gdp per worker (8) growth in gdp per worker Trade 0.5049** (0.2463) 0.5371** (0.2584) 0.0093 (0.1415) 0.0387 (0.1539) 0.3082** (0.1203) 0.3182*** (0.1192) 0.1557** 0.0779 0.1646** (0.0760) Developing -2.2379*** (0.2787) -2.1648*** (0.2637) -0.9662*** (0.1879) -0.9632*** (0.1882) Trade* Developing 0.7344*** (0.2646) 0.7101*** (0.2480) 0.4803** (0.2033) 0.4804** (0.2078) Ln(area) -0.1476*** (0.0467) -0.1088** (0.0494) -0.0714** (0.0319) -0.0579* (0.0327) -0.0571 (0.0350) -0.0525 (0.0339) -0.0457 (0.0300) -0.0418 (0.0291) Ln(population) 0.1813*** (0.0424) 0.1840*** (0.0375) 0.0645** (0.0266) 0.0694** (0.0269) 0.0742** (0.0365) 0.0739** (0.0367) 0.0730*** (0.0260) 0.0729*** (0.0262) Landlocked -0.7121** (0.2842) -0.2936* (0.1678) -0.0960 (0.1250) -0.0838 (0.1041) Ln(gdp1960) Per worker -0.0348 (0.0402) -0.0441 (0.0381) -0.2654*** (0.0605) -0.2729*** (0.0579) Constant 9.7903*** (0.5971) 9.5212*** (0.6208) 10.3502*** (0.4162) 10.2210*** (0.4263) 1.1268** (0.4878) 1.1733** (0.4982) 3.3358*** (0.6041) 3.3657*** (0.6021) Number of Countries 91 91 91 91 71 71 71 71
