Foreign Exchange Flows and Economic Growth in Developing Countries
Master Thesis for Ms International Economics and Business
University of Groningen, Faculty of Economics and Business Corvinus University of Budapest, Faculty of Economics
Siavash Radpour (S2060159) S.Radpour@student.rug.nl
Supervisors : Dr. Bart Los, Dr. István Magas
Academic Year 2010 - 2011
2 Summary
In this study, we build a model to show how foreign exchange flows as different elements of balance of payment are related to growth rates of developing countries with possible limitations in obtaining foreign exchange required for essential imports. We estimate the coefficients of variables representing changes in exports and changes in current account deficit with GDP growth in a sample of 40 developing countries over the years between 1980 and 2005. We found positive and significant coefficients for the variables representing both current account deficits and exports in short and long run tests for developing countries and showed that growth in any form of foreign exchange flows is positively related to output growth in developing countries. We show that this relation is stronger in developing countries compared to developed countries, concluding that developing countries are more dependent on foreign exchange flows for their growth. We also argue that variables representing current account deficit and exports should be used in the analyses together; otherwise the results of analyses will be biased and unreliable.
Key Words: Economic growth, developing countries, Imports, Exports, Current account deficit, Foreign Exchange Gap
1. Introduction
Imports, exports, trade openness, foreign capital flows and financial openness have been some of the main topics studied by several researchers interested in economic growth in developing countries. All of these issues have their own specific characteristics and channels of affecting the economy, but they have one thing in common: for a developing economy with a weak currency, all of them are an issue of foreign exchange. While most of international trade and investment is happening in form of major currencies like US dollar or Euro and advanced economies, developing countries are facing various problems handling the lack of liquidity and exchange rate fluctuations. In order to maintain and increase their economic growth rates, developing countries try to gain access to foreign exchange by increasing international flows and at the same time control and manage these flows to prevent their negative effects on the economy. Although these flows have the similar function of providing foreign exchange, their other functions and effects are totally different. For the two major foreign exchange inflows, exporting and capital inflow, the difference is obvious:
while export is a form of production and may induce some externalities on the industry,
3 capital inflow in forms of short term and long term investment has a totally different target.
However, at the end of the day, the foreign exchange ends up in the foreign exchange markets. Exporters have to sell to foreign exchange to pay for the costs of production factors and Investors have to sell the foreign exchange to buy capital goods and assets in domestic currency.
This model of flows suggest that not only there are possible relations between economic growth and each type of foreign exchange flows, but also suggests that there are connections between the flows that allows them to replace each other in some aspects. For countries with limited foreign exchange resources and strong demands for imports, the main function of the flows, providing foreign exchange to the market, is stronger while in countries with lack of investment or weak domestic demand foreign capital or exporting can have different effects on the economy.
In this study we answer two main questions about this issue: (1) How are foreign exchange flows related to the economic growth of developing counties? Is there any positive relation between volume of foreign exchange flows and economic growth in developing countries?
This question is too general and cannot really show the nature of the relation between the foreign exchange flows and economic growth. We should also investigate the possible differences/similarities between how different forms of foreign exchange flows are related to the economic growth of developing countries. (2) Are the possible relations with economic growth similar between different forms of flows?
Any answer to the second question can be important from two different points of view: it
is useful for policy makers in developing countries, who have to choose policies like export
promoting or incentives for foreign investment to see f they can replace each other or which
one is more effective and efficient for economic growth. But it is also an important question
from theoretical pint of view. If these two types of foreign exchange inflow have the similar
relations with economic growth, they should be studies and analyzed together and in an
integrated model, rather than separately. While there are few researches conducted in such
integrated framework, we can find several studies about specific relations of each of these
flows with economic growth.
4 Since the Second World War, with a lot of former colonies gaining independence and joining other “non-industrial” economies, dynamics of economic growth and growth policies became one on the most important research areas for economist. One of the main issues in this context was the different paces of growth and development between developing countries and heavily damaged European countries after the Second World War. Researchers explained these differences in growth rates of these two groups of countries by defining the “gaps”
between developing and industrial countries that hold developing countries from catching up with the industrial world. Three of the most important defined gaps are saving constraints, foreign exchange constraints and fiscal constraints. Bacha (1990) used these three gaps to model the economic growth problem in a developing economy. Saving gap is the limitation on capital formation in developing countries. Fiscal gap states the limitations faced by government on external borrowing and running a budget deficit necessary for growth, and Foreign exchange gap shows the limitation on importing necessary capital and intermediate goods caused by lack of foreign exchange resources.
For most of the developing countries, except major natural resources and fuel exporters, foreign exchange had been a rare resource during the 20
thcentury. They needed foreign exchange to essential commodities as well as import capital and intermediate goods and buy technology from developed countries. They also needed the foreign currency to facilitate their transactions with rest of the world, since they we too risky, unstable and unknown to do business backed by credit from international financial markets.
During these years, many researchers focused on the trade as the most important way of interaction between developing countries and developed economies, since before 1990’s the volume of capital flow to these countries was small and assumed not to have major effect on most of the developing economies
1.
During 1990’s, as it is shown in figure 1, financial flows in form of FDI and portfolio investment, switched the main argument from trade openness to financial openness. Variables representing capital flows, which were omitted in previous studies, became dominant in the analysis and this time trade variables were omitted in these new studies. This also changed the focus of researches from foreign exchange gap to saving gap in developing countries, and the foreign exchange gap was somehow neglected in most of the studies.
1
We will review some of these studies later in the section 2 of the paper.
5 Figure 1 - FDI flow to Developing countries (source: UNCTADstat)
The main contribution of this study is to develop a model including both trade related and capital related foreign exchange flows, using the literature on effects of both trade and foreign capital on economic growth.
The structure of this thesis is as follow: In section 2 we will discuss the theory and literature on different forms of foreign exchange flows and economic growth. In chapter 3 we will state the hypotheses to answer the research questions based on the general idea formed through the literature review. We will develop the integrated model to fill the gap in the existing literature and test the hypotheses in chapter 4. Section 5 will deal with the statistical methods and tools used for the analyses and data collection on foreign exchange flows in balance of payments and economic growth from a group of 40 developing countries during 1980-2005 period. In section 6, we will discuss the results of the analysis and in section 7, we will draw my conclusion and talk about imitation of the study and give indications for future studies.
2. Literature Review
There are three sets of studies that should be reviewed in order to form a proper idea about the subject of economic development and foreign exchange flows.
The first group is consisted of general and basic articles that discuss the growth models.
To understand more complicated models used by other researchers, which used these simple general models and adjusted them to study specific aspects of growth or the relation of economic growth with different factors, we need to have a good understanding of assumptions and conditions of the simple basic models first.
0 100000 200000 300000 400000 500000 600000 700000
1970 1972 1974 1976 1978 1980 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010
Current US Dollar in Millions
Developing economies: Asia Developing economies: America Developing economies: Africa
6 The second set is a group of articles discussing the possible connections between trade and economic growth. They are relating exports and imports to economic growth and argue about the different possible channels for such relations (like foreign exchange gap or externalities) and look for causalities in the relationships.
Studies of the third group are about financial openness and effects of capital flows on economic growth.
The second and third groups of studies are important for our research, as we are working with the same variables as they used before, but in a new framework. Although they analyzed the trade and capital flows separately, looking at their works can help us on both developing our model and empirical methods and possible problems that we will face while gathering data and running the tests.
2.1.Growth Models and Development Gaps
One of the first developed growth models, the Harrod-Domar model (Harrod, R. F.
(1939), Domar, D. (1946)). Harrod-Domar model assumes that aggregate output of a country, in presence of labor unemployment and unlimited supply of labor, is a function of capital stock of that country with constant marginal capital productivity. Denoting aggregate output with “Y”, capital stock as “K” and constant marginal productivity of capital as “a” we can write:
(1) = ,
=
In Harrod-Domar model, investment or change in capital is equal to saving, S, and saving is a constant proportion of income. This constant proportion is saving rate or s.
(2) = = × (3)
= ×
= ×
As equation (3) shows, in Harrod-Domar model growth rate is equal to product of saving rate and marginal capital productivity which is assumed to be constant over the time.
So the only way to increase the growth rate seems to be increasing the saving rate which
could affect both demand (through increasing the saving) and supply (through increase in
investment).
7 This model can show that in an economy with low income, people don’t earn enough to save much money, so the saving rate will be low. This low saving rate limits the growth rate and “saving gap” emerges in the less developed country.
Chenery and Bruno (1962) and Chenery and Strout (1966) added another gap, the foreign exchange gap, to the Harrod-Domar model and formed the two-gap model. They recognized to forms of imports: M
k, as the capital goods imports and Mo, imports of final consumer goods. Denoting exports with X, they defined:
(4) = −
Here, E is the amount of foreign exchange gained with exporting that can be used for importing capital goods. Now, denoting foreign capital inflow and outflow by F and J, and defining m as import content of investment (the amount of capital goods that should be imported from other countries and there is no domestic substitution for them), we can write:
(5)
= ×
(6) =
+ −
Using these equations, Chenery and others showed that in full factor employment, investment and therefore growth is constrained by amount of foreign exchange available for importing capital goods. They also acknowledged the dual nature of foreign investment which can work for the economy as both investment and foreign exchange reserve.
Solow (1956) extended the Harrod-Domar model by questioning some of its assumptions. He adopted a more complex production function that could allow more flexibility on the basic assumptions. In his new production function, he replaced the constant marginal productivity of capital with diminishing return to capital and added labor as another production factor with its independent diminishing return, and gave both together a constant return to scale. He also added factor productivity to the model that could change with time.
By removing the “fixed-proportions” of the Harrod-Domar model, Solow allowed new equations of price-wage-interest to enter the model and make it more accurate and useful.
While Solow growth model was introduced in mid 50’s, researchers like Chenery
were still using Harrod-Domar model in their studies, neglecting its limitations. Models like
two-gap model could be easily transformed to match with the Solow model, because their
basic definition were not totally dependent on Harrod-Domar model and the concept were
8 adoptable to the new model. To understand the reason behind this persistency on the old Harrod-Domar model, we should take a look at the general understanding of those economists about developing countries.
Krueger (1997) gives us a good overview of early stages of development theories that generally led to policy recommendation toward protectionism and import-substitution. As she explains, most of the theories were based on a set of standard facts and premises later found out to be invalid or questionable: (1) Development economies are oriented toward primary commodity production and there is no opportunity for manufacturing, (2) Development economies don’t have any competitive advantage in manufacturing sectors and trade openness will eliminate the possibility of domestic manufacturing, (3) Global income and price elasticity of demand for primary commodities is low and there is no chance for developing economies to increase their export income, (4) Marginal productivity of labor in developing economies is low and even negligible, (5) Capital accumulation is crucial for growth, and the only way to do that is to import capital goods instead of consumption goods by producing consumption goods domestically, (6) Consumers in developing countries are behaving “traditionally” and they are not sensitive to price incentives. Based on these premises, economists and policy makers developed the idea of closing the economy by implementing import, export and capital control.
Such a view on the structure and fundamentals of developing economies made the developments of Solow model rather useless. In economies with rigid production structure (premises 1,2 and 3), inefficient and unimportant role of labor (premise 4) were nullifying the flexibility Solow added to the growth model, and premises 5 and 6 were emphasizing on the role of capital accumulation where traditional consumers had no intention of saving and increasing the investment.
Indeed, later on, there were researchers that questioned these premises and used the concepts of saving gap and foreign exchange gap in the framework of Solow model to investigate the effects of trade and foreign investment as foreign exchange flows on growth.
We will take a look at some of them in the next two parts of literature review.
2.2.Trade and Economic Growth
Michaely (1977) is the oldest study on the relation between trade and economic
growth we will present here in the literature review. He examined the relation between
9 expansion of exports and economic growth of 41 developing countries for years 1950 to 1973, and found a positive correlation between expansion of exports and economic growth of these countries. Instead of working with a complex model that include all possible sources of growth and capture the possible effects of exports, he performed a simple test that (he argued) could test his hypothesis in an isolated condition. He simply examined the relation between increases in exports/total output and growth ratio. He argued that using exports/total output ratio eliminates the autocorrelation caused by exports being part of national production. If we use the change in exports directly, it will have an inevitable positive correlation with changes in national products and the results will be useless. We will use his method later to deal with part of the endogeneity of our variables.
Balassa (1978, 1985) based his methods on the work of Michaely and repeated the test for 11 developing countries for years 1960 to 1973 and found the same positive correlation. He used limited number of countries to have a more homogenous sample. He also ran a multiple regression including other variables like foreign and domestic investment and labor, but in this regression he used the change in value of exports instead of changes in exports/total output ratio and he did not present any proof to validate his regression, despite the previous warning given by Michaely. He neither gives any reasons for putting exports in the regression other than a general assumption that “exports tends to raise total factor productivity”. His work is similar to what we intend to do, but we have to make sure not to repeat his mistakes and be more careful about the model and methodology.
Tyler (1981) followed Michaely and Balassa. He criticized the work of Balassa because of his small and selected sample that could generate biased results. Tyler used the data from 55 middle income developing countries
2for years 1960 to 1977. Unlike Michaely and Balassa, he used a production function to define his regression equation. The production function is:
(7)
=
!
"Where Y
i, K
i, L
iand X
iare denoting country i’s output, capital stock, labor force input and exports and A is a technological constant. Tyler adds Exports to the production function on the grounds that “there are scale effects and externalities associated with export production and sales. His final regression equation is the differentiated form of equation 7, which is:
(8)
##
=
$$+ %
##
+ &
''##
+ (
))##
2
He used the World Bank classification of the countries in 1977.
10 He also found a positive coefficient between exports growth ratio and output growth ratio and an increase in explanatory power of the regression in presence of exports variable.
Feder (1982) tried to capture the externalities that were mentioned in these studies, but never actually evaluated. He used the data of 31 developing countries for years 1964 to 1973.
The most important aspect of his study is that he followed Chenery and adopted his model and even used the same set of countries Chenery used in some studies. However, Feder used Solow model and didn’t repeat Chenery’s mistakes. His work opened the way for other researchers, like Esfahani (1991) to use the Chenery’s two-gap model in new forms.
Esfahani examined exports’ externalities as well as the effects of exports through filling foreign exchange gap for the same 31 developing countries as Feder, during a longer period of 1960-1986. He found that the coefficient between increase in export and GDP growth rate is an outcome of foreign exchange gap getting filled with export incomes, not externalities of export industries. He used the same definition of foreign exchange gap as in two and three gap models, but allowed for variation in exports and added an imported intermediate good to capture the effect of limitations in imports.
His work is very important for our study, as he put imports in his production function and let the foreign exchange flows enter the equations by replacing imports. Following the concept of foreign exchange gap, he argued that imports of intermediate goods and services are necessary for production. Denoting imports of intermediates with M, he wrote his basic production function as below:
(9)
=
!
"His conclusion was that higher rates of exports provide necessary foreign exchange to import capital and intermediate goods and facilitate the economic growth. He also mentioned that there is no difference between filling the foreign exchange gap with export promotion or borrowing from international capital markets, but he didn’t really test this statement. His model was too complicated, as he was trying to isolate different effects of exports from each other, and it is not easy to add another variable representing foreign capital to it.
In this study we will use the concepts developed and studied by Chenery, Feder and
Esfahani, but with the simpler methods used by Michaely, Balassa and Tyler. Still, to add the
variable representing foreign capital flows, we need to add some more studies to our
literature review.
11 2.3.Foreign Capital and Economic Growth in Developing Countries
After more than two decades of research and discussion, there is still no clear cut conclusion on effects of foreign capital flows on growth in developing countries. The empirical analyses by different researchers have different and opposite results. As a result of these puzzling empirics, different theories are proposed in favor and against financial openness of these countries.
Kose et al. (2006) in a literature review paper, showed that how different approaches on dealing with the issue of foreign capital and financial openness can lead to different and opposite results. They listed some of the most important empirical studies in their review and discussed the main aspects of these researches. Some of these studies used measures of trade openness in their analyses as a control variable, but none of them included trade or exports as a main variable in addition to foreign capital flows.
One of the most important studies in Kose Et al. literature review is the work of Bosworth and Collins (2003). They construct the growth accounts of 84 countries over 1960- 2000 period. They also categorized these countries and ran the separate test for 62 developing countries on their sample. Their study has several important characteristics: other than using a large database including many countries in a long period, they tried to “standardize” the process of growth accounting by comparing different factors and variables used in different studies and choosing a coherent set of indicators for them to reduce the variations in methodologies and results of growth accounting researches. We will refer to some of their findings in our research.
First of all, they conclude that cross-country regressions “can yield consistent and
useful results”. In a cross-country regression, which is one of the main tools for empirical
analysis of cross-national differences in economic growth, we assume that all of the countries
have the same production function and changes in different factors have the same effects in
every country. Therefore, differences in countries’ growth rates are only a result of
differences in their factor endowment. This assumption allows us to run a regression on data
from different countries during the same time period, and observe the coefficients of different
elements with economic growth. However, this assumption has been criticized by many
researchers, questioning the validity of such methods.
12 All of the studies mentioned in this paper used some kind of cross-country analysis in their work, and we will also follow their methods. So any discussion about consistency and usefulness of this method is very important for us. We will come back to this discussion later in section 6, when we talk about the methodology of our analyses.
Bosworth and Collins also argue about using per capita values of output and capital instead of total output and total capital stock of countries. They also discuss that change in capital stock is a better indicator than investment rate to capture the effects of capital in economic growth. We will also discuss about these two issues late in methodology section of the thesis.
In another empirical study using the work of Bosworth and Collins as their basis, Prasad et al. (2007) tried to find the benefits of financial globalization, studying a group of 59 developing countries during years 1970-2004. They expected a positive correlation between current account deficit, as a measure of capital inflows, and economic growth of countries, but their result was in total contrast with their expectations. They found a positive correlation between growth rate and current account surplus instead of a negative correlation which could explain the benefits of foreign capital inflows. They proposed that appreciation effect of capital inflow can harm the exporting industries of a developing country and lower the growth rate instead of boosting it. Rodrik (2007), in line with their results, argued that the main driver of growth in successful developing countries is the currency undervaluation, which can only be kept in place with export promotion and capital control policies. Although they had current account as their main variable, they didn’t mention possible connection of their findings with trade. In another paper, Rodrik and Subramanian (2008) again tried to explain this result. They defined two different types of countries, “Investment-constrained”
and “Saving-constrained”, to explain why inflow of foreign capital doesn’t have any positive effect on some economies. They argue that in Investment-constrained countries there is not enough investment opportunity for local or foreign investors, so any capital inflow will result in more consumption (instead of investment) and can only have negative effects on the economy. In other words, if we look at the discussion in the framework of three-gap model, they argued that “saving gap” is not the main problem in developing countries.
While Rodrik and Subramanian’s work is generally theoretical, they used what they
called “a simple test” to show that many developing countries are not saving constrained, as
most of theories would say and conclude the importance of foreign investment to fill the gap
13 of domestic saving and investment, but rather there is no investment opportunity in those countries, and foreign capital inflow won’t lead to increase in domestic investment. They used the correlation between domestic investment rates of developing countries with change in U.S. interest rates to show if the real limitation for development is the saving or not. The argument is that if these countries are constrained with lack of domestic saving, lower interest rate in U.S. will encourage investors to move their capital from U.S. to those countries, and therefore their domestic investment should increase. If the U.S. interest rate increases, investment will move from developing countries back to U.S. and domestic investment rate in developing countries should fall. So, if developing countries are saving constrained, we should expect a negative correlation between change in U.S. interest rates and domestic investment in developing countries. Surprisingly, the correlation was negative only for China and India, and domestic investment rates of other developing countries in the sample were positively correlated with changes in U.S. interest rate. That means with lower interest rates in US, the investment opportunity and saving in these countries also get lower. This is happening because most of the savings were going toward investment in foreign countries rather than in domestic projects, and these countries are constrained by lack of domestic investment opportunity rather than domestic saving.
These studies have some interesting issues that are useful in our research. First issue is about nature of current account balance used by Prasad et Al. Current account balance is the difference between values of exports and imports. While in definition it is only the difference between these two (and not necessarily related to each of them individually) current account balance is correlated with both imports and exports, so any positive coefficient of current account balance in a growth regressions can be simply a result of omitting imports or exports and transferring their effect on current account balance.
Therefore putting exports or imports may increase explanatory power of our regressions and gives us more understandable results.
The second point is about the main causes of overvaluation of currencies, that in both
Prasad et al’s and Rodrik and Subramanian’s studies mentioned as the main reason behind
negative effects of foreign capital inflow to developing countries. Excessive supply of foreign
exchange will cause overvaluation regardless of sources of supply. Countries with high
dependency on exporting and low demand for imports (and foreign currencies) can face the
same problem as a country with high amounts of capital inflow. This also motivates us to
14 look for the reasons other than currency overvaluation behind the positive coefficient of current account balance and growth.
The proposition of Rodrik and Subramanian is also interesting for us, since they argue that saving gap is not the main problem and allow us to focus on foreign exchange gap instead.
3. Hypotheses
According to the previous researches mentioned in literature review, there are three possible ways that foreign exchange flows can affect the economic growth through them: (a) relaxing the foreign exchange gap, (b) increasing (foreign) investment and (c) externalities.
We are not able to catch each of these different effects separately in a simple model and developing a complex model and testing it will be out of the scope of this thesis. We will follow Michaely (1977) and examine the aggregate relations of the foreign exchange flows with economic growth.
If the assumptions about the role of foreign exchange gap is true (as in Chenery et al (1962, 1966), Bacha (1990) and Esfahani (1991)) we should be able to see a positive relation between foreign exchange flows and growth in both short run and long run, since in short run imports of intermediates and capital goods immediately (or at least in a one year span) increases the domestic production and leads to economic growth and in long run we can also see the possible externalities or effects of direct investment on the growth rate.
We state our first hypothesis based on the expected coefficients of foreign exchange flows in short run:
Hypothesis1: An increase in any form of foreign exchange flow to a developing country is associated with an increase in the economic growth rate in short term.
In the long run, we expect current account deficit to lose its importance for two reasons.
First is that in the long run current account deficit will be more expensive as countries have to
pay back the interests and loans. Second, in the long run countries (especially developing
countries) cannot maintain a current account deficit and accumulate more debt, since they
will become more risky and their credibility and ability for paying back their debt will
become questionable. In other words a long term deficit will lead developing countries to
15 crisis, or at least will not help their economic growth. Also in long run, there will be possible externalities of exports, while these externalities for aggregate amount of capital inflows which include short term and long term loans in addition to foreign direct investment (which is suggested as the main source of foreign capital externalities) are not expected to have much externality. But foreign capital inflow will enter the economy and production cycle as investment and it will also have some relations to the economic growth. We should also consider the findings of Prasad et al (2007) about negative coefficient of capital inflow to developing countries in long run.
So, while in the short run we expect the different forms of foreign exchange flows to have approximately similar relations with economic growth, in the long run their specific characteristics will become more important and we expect different coefficient from them.
We cannot predict the possible values of the coefficients, so we will use two hypotheses to test the possible magnitude of coefficients:
Hypothesis3: In long term, foreign exchange flows in forms of exports has more positive correlation with economic growth as a result of the externalities and its coefficient will be larger than coefficient of capital inflows.
Hypothesis4: In long term, foreign exchange flows in forms of capital inflows has more positive correlation with economic growth as a result of the externalities and their role as investment, and its coefficient will be larger than coefficient of exports.
If we find the results in support of hypothesis 3, we can argue that Rejecting both of these two hypotheses will show us that the coefficients of the foreign exchange flows are not related to the assumed mechanisms and we have to explain them only via foreign exchange gap or we should look for other possible explanations for the coefficients.
4. Empirical Model
To test our hypotheses, we need a growth model that includes both exports and foreign
capital inflows. We start with a basic model similar to that of Esfahani (1991). The total
output of the economy (or Gross Domestic Production, GDP), Y, is assumed to be a produced
by means of three factors: K for total capital, L for total labor force, M imports. A represents
total factor productivity (TFP) of Solow growth model. Considering the production function
with a Cobb-Douglas functional form we can write:
16 (10)
=
* **!
*"Esfahani initially added only the imports of intermediate and capital goods into his model.
In the same way that Chenery defined the concept of foreign exchange gap, Esfahani argued that imports of intermediates and capital goods are necessary for production in any country, especially developing country with limited technological ability to produce wide range of products. He doesn’t look at imports as a source of externalities, but a necessary factor for production. Later to run his regression, he replaced the imports of intermediates with total imports as a result of data unavailability. He argued that this is acceptable since according to Khan and Knight (1988) during his period of studies (1970’s) over 75% of total imports to developing countries were consisted of intermediate goods.
We add the total value of imports directly to our model, on the ground of his arguments.
In addition we should mention that in nowadays economies it is difficult to separate intermediate and capital goods from consumer goods, especially when a lot of final products can be used in both service sector (as a capital good) and households (as a consumer good).
Even imports of consumer goods can create a whole distribution and service sector that rely on these imports as intermediates (like sales offices and service shops for imported cars). At the end, as it is repeatedly mentioned in trade openness literature, imported consumer goods generate competition in the market and induce externalities on the domestic production.
All of these arguments support the idea of adding imports into the production function.
To continue, we write the differentiated form of equation (10) as below:
(11)
*=
$#+#+*
+ %
#+#+*
+ &
'#+#+*
+ (
,#+#+*
It is natural for countries with larger population to have more output, capital and imports.
This will cause a level of collinearity between our variables, in addition to increasing the risk of heteroskedasticity in analyses. To eliminate these problems, and also reduce the number of variables in our analyses, we use the per capita (per unit of labor) version of (10):
(12)
'=
',
'=
',
'=
,'(13)
'= -
' ́'!-(14)
'=
$-/- + %́
//'
+ &-
,//'
17 In equations (12), (13) and (14), as in equations (10) and (11), all the variables except %́
and &- are time and country specific and we just ignored i and t indices to make it easy to read.
Equation (14) gives us the required basis for our regression, but to find an answer for our question, we have to put exports, X, and current account deficit, C into our model.
Considering that net flow of foreign exchange to each of the countries is zero, we can replace imports, M, with the value of exports, X, and net capital inflow, F-J, with F as capital inflow (foreign investment) and J as capital out flow and change in the reserves, r. However, reliable data on capital accounts of developing countries is not available in most of the cases.
Therefore, assuming that balance of payment sums up to zero
3, we replace Capital account balance with Current account deficit (C). Using the definition of balance of payments definition, we can write:
(15) (X - M) + (F – J) + r = 0 (16) C = M – X = F – J + r (17) M = X + (M – X) = X + C
Notice that we are using current account deficit as a proxy for capital account balance and net foreign capital inflow, neglecting the change in reserves. We are following Prasad et al in this, as they also used current account deficit as a measure of capital inflow to developing countries. Replacing M in equation (14) with equation (17) gives us:
(18)
'=
)', 0
'=
,1)'(19)
'=
$-/- + %́
//'
+ & -
,//
'
+ & -
2,/ /0
'(20)
//
=
$-$-+ %́
//+ & -
)/,/
+ & -
2 3/ ,/According to equation (14) an the way we derive equation (20) from it, & - 4 &
- should
2be equal to each other and &-, the coefficient of imports. Still, we will let them to have different values, as according to our second and third hypotheses, exports and other ways of obtaining foreign exchange can have different costs.
3
The balance of payments by definition “should” sum up to zero, but because of measurement errors the real
data can sum up to small non-zero values. By assuming that the real data of balances of payments sum up to
zero, we ignore these measurement errors in our study.
18 To use money from exports, foreign loans or reserves, we face different type of transactions with different costs. Foreign exchange obtained from exporting will directly enter the banking system and can be used for imports immediately, while to channelize foreign investment to the country and then into domestic banking system we have several intermediation and insurance costs to cover or the risks of asymmetric information. Using the country’s foreign exchange reserves will also impose cost of government involvement in the market and increase in the overall transaction costs as a result of higher risks. When a country accumulates foreign exchange as reserves, the general risk of defaulting in the country on foreign loans and other types of financial instruments reduces, and the general costs of transactions like borrowing, getting letters of credit or insurance become lower. These sorts of costs may not be visible in micro level, but since we are dealing with macro level data, they can affect the value of & - 4 &
- and make them differ from each other.
25. Statistical Method
5.1.About Cross-Country Regression
As it is mentioned before, the nature of this thesis is to find a global relation between foreign exchange flows into developing countries and their economic growth. Doing that, we have no choice but to use a cross-country analysis.
Many researchers have used cross-country regressions over different time periods for the similar types of hypotheses. There are also several critics on their methodology and conclusions that argue such methods will always have biased and useless results.
Durlauf and Quah (1999) argue that growth models were basically developed for within country analyses rather than cross-country regressions and such regressions will lead to biased results. They try to show the reasons and magnitude of this bias in some models. But at the same time they accept the value of the cross-country regressions as an important tool for economist and just advice them to be more careful with their models and even more careful about interpretation of their possibly biased results.
Lindauer and Pritchett (2002) on the other hand focus on the less statistical sides of the
issue. They argue that the relations between different factors and economic growth can vary
from time to time and from country to country, and cross country regressions can just give us
misleading results. They also point out that many variables used in regressions are
intermediate variables affected by policies and other basic factors, and using such
endogenous factors cannot give any useful and reliable result.
19 Bosworth and Collins (2003) tried to answer this critique by making cross-country regressions as a standardized process. Since most of the regressions have the same core elements or their core elements are correlated with variables used in other researches, they tried to find a standard regression by recognizing the set of main determinants common in most of the researches they try to make a consistent model in line with findings of other researchers. We also follow Prasad et al. (2007) and Kose et al. (2006) that used the same basic set of variables as Bosworth and Collins (2003) in their several recent papers on financial openness and growth to avoid this critique. However, we should acknowledge that our results also are limited to the time period and set of countries in our study.
The other concern about growth regressions comes from endogeneity of macroeconomic variables. We have simultaneous determination problem with all of our main variables, GDP, imports, exports and investment. Since there is no valid instrument for most of these variables, many researchers have ignored the problem and tried to explain their biased results instead of relying on the numerical outputs of their regressions. Although by using variation in ratios, which is more related to the structural changes in economies than changes in level of output, we still acknowledge the problem of endogeneity of our “jointly determined values” like domestic investment, capital flows and exports in relation with GDP.
The other problem in using cross country regressions (or simple panel data analyses) is that they do not check for causality between dependent and independent variables. We are careful enough in developing our hypotheses to mention the “association” of any change in explanatory variables to possible change in growth rate, but still lack of causality tests is the main weakness of our research and other researches based on cross country analyses that can just capture approximate sources of growth.
5.2.Control Group
Because of all the problems mentioned above, any result based on a cross-country
regression can be questionable. We are working with macroeconomic variables, which are
usually jointly determined and have many common determinants like policies or cultural
backgrounds. For this reason, it is possible for the results of our estimation to be derived by
such endogeneity inside the model instead of real relationships between variables and become
biased. Since our specific model has not been tested before in other studies, we don’t have
any idea about existence, size or direction of these biases.
20 To check for robustness of our results in presence of all mentioned issues, we run all of our tests for a control group, selected from high-income OECD countries. As the core issue of the study, the foreign exchange gap, is not present in these countries, we expect the estimation results for the control group to be different from the group of developing countries.
By comparing the results of the cross-country regressions of developing countries with the results from control group we can make sure that the coefficients are not only caused by general endogeneity of variables and a systematic error in our empirical model, and at least there are some country/group specific characteristics that affected the results.
If we find similar results for developing and developed countries, it can show us that our model or the statistical analyses are not valid. However, difference in the results does not validate the model, and just helps us to explain the findings.
5.3.Regression Equations
To test out hypotheses, we need to run two separate tests, one for short run and the other for long run changes in economic growth. For the short run estimation, we use panel data analysis, while for the long run estimation we run a multiple regression on averaged values of the factor during the observed time period.
We base our regressions on equation (20), with growth rate of GDP per unit of labor dY
L/Y
L(that will be shown as y in the test from here on), as dependent variable and change in capital stock per labor dK
L/K
L(which will be called as k), as control variable.
Our independent variables dX
L/M
L(that will be denoted as x) and dC
L/M
L(which will
be denoted as c) are equal to changes in exports and current account deficit divided by value
of imports. Calculating the change in current account and exports in respect to imports shows
that even big variations in these variables are not really important if they are small compared
to total amount of imports. For a variable like current account deficit that changes with high
ratios in some periods because of its small values (when the amount of deficit/surplus is small
compared to volume of trade) this ratio is more useful than ratio of changes in current
account deficit itself (dC/C) that can have really large values when C is small and close to
zero. Using these ratios is also in line with the method of Michaely, as changes of the ratios
do not affect the GDP in a direct way.
21 We will leave the effects of change in factor productivity ( -) to be captured in the constant, 5 (as overall change in TFP for countries in the sample), 6
(country specific changes in TFP) and in the error terms ( 7
*, 7
.
5.3.1. Short run estimation:
For the short run estimation, we use panel data analysis with fixed-effects estimator.
Using panel data analysis allows us to check if our model holds within each single country during the observed period, as well as between the countries at each time.
The fixed-effects model assumes similar coefficients for each independent variable across the different countries, but allows for different intercept for them. With these individual intercepts we can capture the differences between factor productivity between the countries that can give us better estimations. The other alternative estimator, the random- effects estimator, is more efficient than the fixed effects estimator, but it uses an extra assumption. Random-effects estimator assumes that differences between individual intercepts are random and normally distributed. Since on the basis of our model we cannot accept such an assumption, we run our estimations with fixed-effects estimator and later examine the possibility of using random-effects estimation instead of it. Our regression equation for fixed- effect panel data analysis will be as:
(21) 8
*= %́9
*+ & - :
*+ & - ;
2 *+ 5 + 6
+ 7
*8 =
' ': =
' '; = 0
' '9 =
' 'We can formulate and test the first hypothesis as a joint test to check if & - 4 &
- both
2are statistically significant. Then if their values were larger than 0 we can accept the first hypothesis. We place our hypothesis as the alternative hypothesis of the test and write:
Hypothesis 1: H0: & - = 0 => &
- = 0
2H1: & - > 0 4 &
- > 0
25.3.2. Long run estimation:
For running the multiple-regression and estimating the coefficients in the long run, we write the regression equation as below:
(22) 85
= %́95
+ & - :̅
+ & - ;̅
2+ 5 + 7
22 85
= A 8
**
B
95
= A 9
**
B
:̅
= A :
**
B
;̅
= A ;
*B
*
where T is the number of observed years.
The second and third hypothesis can be formulated as:
Hypothesis 2: H0: & - ≤ &
-
2H1: & - > &
-
2Hypothesis 3: H0: & - ≥ &
2-
H1: & - > &
2-
5.4.Data and Indicators
The main sources of data for the analysis are World Bank’s World Development Indicators (WDI). We tried to extract most of our data from one source, WDI, to avoid any problem caused by differences in accounting methodology.
5.4.1. Main Variables:
Total Output (Y)(in form of added value-Gross domestic product): Data from World Bank’s WDI database, “GDP (Constant 2000 US$)”. GDP at purchaser’s price in constant 2000 US$ (converted by using 2000 official exchange rates). It is calculated without making deductions for depreciation of fabricated assets or for depletion and degradation of natural resources. For a few countries where the official exchange rate does not reflect the rate effectively applied to actual foreign exchange transactions, an alternative conversion factor is used by World Bank.
Imports (M) and Exports (X): Data from World Bank’s WDI database, “Imports of goods, services and income (BoP, current US$)” and “Exports of goods, services and income (BoP, current US$)”. We converted the data to constant 2000 US$ using the same exchange rates as it was used by World Bank for converting GDP values.
Current account balance (C): Data from World Bank’s WDI database, “Current account balance (BoP, current US$)”. We converted the data to constant 2000 US$ using the same exchange rates as it was used by World Bank for converting GDP values.
Labor (L): Data from World Bank’s WDI database, “Labor force, Total”. “Total labor
force comprises people ages 15 and older who meet the International Labor Organization
23 definition of the economically active population: all people who supply labor for the production of goods and services during a specified period. It includes both the employed and the unemployed. While national practices vary in the treatment of such groups as the armed forces and seasonal or part-time workers, in general the labor force includes the armed forces, the unemployed and first-time job-seekers, but excludes homemakers and other unpaid caregivers and workers in the informal sector”
4.
Many studies use population as a proxy for Labor. It allows them to use GDP per capita growth rates directly in their equation. While population can be a good estimator of labor force in developed countries, in developing countries with big differences in life expectancy and population growth rates (which alter the working age population ratio) it is better to use actual labor data.
Our data about labor force includes unemployed workers, since the the unemployment can be explained by lack of other production factors, and this must be showed in our per labor unit ratios of production factors of our study. For example, consider a country that has a high unemployment rate because there is not enough capital for production. By excluding unemployed workers, the value of capital per unit of labor will be higher and we will not notice the lack of capital in our analyses.
Capital Stock (K): Unlike other variables, Capital stock data is not directly available on WDI database. Estimating total capital stocks of developing countries is a difficult task and we could not find a reliable source that includes the data for all of the countries in our list.
However, data for “Fixed capital” is available in a World Bank study by Vikram and Dhareshwar (1993). We use the “fixed capital” data from their database for our initial year, 1980, and calculate the net change in fixed capital using World Bank data, “Gross fixed capital formation (constant 2000 US$)” and “Adjusted saving: Consumption of fixed capital (current US$)” converted to constant 2000 US$ for estimating the fixed capital for the remaining years. We calculate the fixed capital stock with this equation:
(23)
*=
EFG+ ∑
*9
I IJEFG5.4.2. Time Period:
Choosing period of 1980-2005 is mainly a result of data availability in secondary sources.
However, this is an important period as with including 1980’s and 1990’s, we can see the
4
Definition as given by World Bank.
24 effects of export promoting policies during 80’s and raise in capital inflows during 90’s.
However, this selection of time period forces us to exclude transition economies, countries that gained their economic (and political) independence from Soviet Union during the 90’s, from our study.
Since we are using differentiated values of our variables, the effective time period of our study is the 25 year period of 1981-2005.
5.4.3. Countries:
We have the data for developing countries with population more than 1 million. We filtered these countries on basis of availability of data on our main variables explained before and finally selected 42 countries. It is most likely that there is a strong relation between data availability and development, therefore running a regression on “available data” will probably gives us biased results for the parameters of general production function for all developing countries.
From these 42 countries with available data, we excluded Panama from our sample, since its currency arrangement and using USD as the second official currency makes the whole argument about foreign exchange invalid. Sierra Leone was also excluded because of its unusual Current account deficit caused by large amounts of Official Development aids that made the country an outlier in our regressions.
For the group of developed countries, we have all of OECD members at 1980 excluding Luxembourg (because of its small population). The final lists of 40 developing countries and 24 developed countries are given in appendix 1.
6. Descriptive Statistics
It is always useful to have a general idea about the values and trends of the variables we use in any study. In this section, we present some descriptive statistics about our main variables. Since we have a panel data for 40 countries over 25 years, the data summery is presented in two forms: averages of years for countries (to make different countries comparable) and averages of countries as time series (to see general trends of different variables over the time). The detailed summaries for all of the variables in our analyses are presented in appendix 2 (for short term analysis) and appendix 3 (for long term analysis).
6.1. Total output, GDP per unit of labor and growth rates
25 Graph 1 shows the summation of GDP’s for developing countries in our sample during 1980-2005 period. The growth is accelerated during the last years of the sample, probably after the 2002 dot-com crisis.
Figure 2: Aggregate output (GDP) in the sample of developing countries
Since this graph is based on aggregate output of the countries in the sample, the visible growth is also partly caused by population growth in these countries. Graph 2 is showing the average GDP per labor unit growth rate in the countries of the sample. The average is weighted by GDP of each economy.
Figure 3: Average GDP per unit of labor growth ratio (Weighted average on countries GDP) in sample of developing countries
In this graph we can see the effect of 1982-83 South American crises, 1998-1999 crisis in
East Asia and Russia and dot-com bubble in 2001 on the growth rate of the countries in the
sample. We can see that majority of the countries have positive but small growth rates.
26 Figures 4 and 5 show the distribution of average output per unit of labor and average per unit of labor growth rates of countries in the sample. Majority of the countries in the sample had low output per unit of labor and growth rates in the period of study.
Figure 4: Distribution of average output per unit of labor in sample of developing countries
Figure 5: Distribution of average per unit of labor growth rates in sample of developing countries 6.2.Imports, exports and current account deficit
Figure 6 shows exports, average imports and current account deficit per unit of labor of
developing countries in the sample of developing countries in the studied time period. In
general, developing countries were running a current account deficit in the whole period.
27 Figure 6: average of exports, imports and current account deficit per unit of labor for sample of developing
countries 6.3.Capital Stock and Capital formation
Figure 7 shows the trend of capital stock accumulation in the developing countries of our sample during the studied period. The steady growth of the capital stock suggests that the average investment rate in these countries had only slow and small changes during this period.
Figure 7: Total capital stock of the developing countries in the sample
The distribution of average of capital per unit of labor during the studied period for
developing countries of the sample is presented in figure 8. Because of big differences
between capital stock of developing countries in the sample, the graph is presented in
logarithmic scales. Turkey with about 9 million US$ per unit of labor has the highest value of
capital per labor, while this value is as low as 800 US$ per unit of labor for Ethiopia.
28 Figure 8: distribution of capital per unit f labor in sample of developing countries (as in 2005)
While capital stock distribution between the countries of the sample has a high variance, the ratios of change in capital stock per unit of labor of the countries are close to each other.
Distribution of this ratio is shown in figure 9. Notice that a negative ratio of change in capital stock per labor does not mean that there is any negative investment, but simply shows that investment could not keep up with population and labor force growth.
Figure 9: distribution of ratio of change of capital stock per unit of labor (average over the period for each country)
7. Regressions, Diagnostic Tests and Results 7.1.Short term analysis
To estimate the parameters in short run, we begin with estimating coefficients of Exports
and Current account deficit in short run to check validity of Hypothesis (1). We use fixed-
effect regression based o equation (11) on our panel. The panel is strongly balanced (with one
set of data missing for Indonesia in 1980).
29 We mentioned before that there might be some overlaps between the effects of changes in k and c, as c also represents foreign investment as a form of change in capital stock. There is also a possibility for correlation between x and c
5. Therefore we need to check for collinearity before running our regression. We check for correlation between out three right hand variables. If the correlation between each pair of them is higher than 0.8, the regression will not be valid as a result of collinearity. The correlation table of variables is presented in table 1. We don’t have any strong correlation and collinearity is therefore not present.
Table 1: the correlation table for variables in short term, fixed-effects estimation
X c K
X 1.0000
C -0.1433 1.0000
K -0.0098 0.0893 1.0000
We also run our tests with and without taking into account the ratio of changes in capital stock, k, and to show that the results are not really different and the overlaps between k and c in respect to imports are neglectable. If the main effect of changes in current account deficit on economic growth was through formation of capital stock by importing capital good, then coefficient of c had to lose its importance in presence of k in regressions, but it does not happen in our model (as it is shown n the regression results) and we conclude that they have different relations with economic growth rate.
Another diagnostic test that should be run before running the panel data analysis is checking for non-stationary time series in our panel. We use Fisher-type unit-root test for panel data to check for stationarity of our panel time series, as it works also for unbalanced panels and allows for different autocorrelations terms for each panel. We run the test for all of our main variables and the test rejects the null hypothesis of unit-roots for all panels for all of the variables in 1% confidence limit.
The first regression only includes x and c as independent variables. Adding k in regression 2 does not change the coefficients of our main variables and only adds the R- squared of the regression by 0.04. It shows that there is no overlap between the role of changes in current account balance and changes in capital stock.
5
