Trade Structure and Economic Growth

Raja Kalia, Fabio Mndez a, and Javier Reyesa,b

Department of Economics, Business Building Room 402, Sam M. Walton College of Business, University of Arkansas, Fayetteville, AR, 72701, USA. Corresponding author. Tel: +(479) 575-6079. E-mail address: [email protected]

October 30th, 2006.

Acknowledgements: The authors greatly benefited from comments and suggestions provided by participants of the Midwest International Economics Group Fall 2005 Meetings and the Fifth Workshop of the Regional Integration Network (sponsored by LACEA), as well as Rossitza B. Wooster, Christopher Laincz, Christopher P. Ball, and Ari Van Assche. Additionally the authors would like to acknowledge that partial research funding was provided by the Summer Research Grant program of the Sam M. Walton College of Business at the University of Arkansas.

Trade Structure and Economic Growth

Abstract How do the number of trade partners and the concentration of trade among partners affect the economic growth of a country? We refer to these characteristics as the structure of trade, and explore this question empirically in this study. We find that the structure of trade, independently from of the level of trade itself, has an important effect on the rate of economic growth. The results of the study suggest that the number of trading partners is positively correlated with growth across all countries, and this effect is more pronounced for rich countries. Trade concentration is positively correlated with growth for all

countries, and the effect is concentrated in poor countries. Previous work has overlooked these characteristics of trade, though we find them to be quite relevant and that they could lead to new ways of understanding the trade-growth relationship.

JEL classifications: O1; F43; O47; F0; O4 Keywords: Trade; Growth; Trade Structure; Trade Concentration.

1 Introduction
The effects of international trade on economic growth have been the subject of a rich debate. Still, the main question of whether (and how) trade enhances growth remains

unanswered, as the conclusions of both theoretical and empirical studies are highly sensitive to changes in the assumptions made, the variables used to measure trade openness, the sample data used, and the econometric technique employed (see, for example, Grossman and Helpman (1991), Matsuyama (1992), Walde and Wood (2005), Rodriguez and Rodrik (2000), Yannikkaya (2003)). Noticeably, most of the analysis within this debate involves trade measures regarding export and import volumes or shares, trade policies regarding tariffs or custom barriers, and related measures of trade openness1. Little or no attention has been given to the trade strategies followed and the types of trade relations established; even though they too have been subject to noticeable changes. Just in the last ten years, over 130 regional trade agreements have been reported to the World Trade Organization (compared to 124 reported during 1948-1994)2 and key players, like the United States and the European Union, have adopted a more aggressive pace in their bilateral trade negotiations. In fact, the changes experienced over the last decades of globalization are more pronounced in terms of the structure of trade than in the volume of trade itself (or in the average trade policies). As shown in Table 1, while the share of total trade in GDP for the average economy went from 58.3% in 1970 to 88.5% in 2003, the average number of trading partners more than doubled as it went from 46.4 to 93.9 in that same period. Similarly, as shown in Table 2, when measuring the trade concentration with the standard Herfindahl index, trade concentration decreases from 2201 to 1580 for the average

economy between 1970 and 2003. Today, most countries are less reliant on one trading partner (or group of partners) than they were 30 years ago. There are reasons to believe that such characteristics regarding the structure of trade have significant implications for economic growth. In this paper, we study these

implications. Specifically, in this paper, we assemble a dataset that tracks the number of trading partners for a given country and the volume of exports and imports. We then construct a measure of the concentration of trade and use these variables to examine the relationship between the international trade structure and economic growth in a panel of over 155 countries for the period 1980-2000. Our results provide new insights into the mechanisms that make international trade a positive influence for growth. We find that the structure of trade, independently from trade openness, has an important effect on the rate of economic growth. The number of trading partners is found to be positively correlated with growth across all countries; this effect is found to be greater for rich countries. Trade concentration is also found to be positively correlated with growth for all countries in general; but this correlation is mostly attributed to poor countries only. These results are robust to changes in the econometric specification used, the number of explanatory variables included, and corrections for heteroskedasticity and endogeneity. In terms of policy, understanding the effects of trade structure on growth is important. As our empirical results suggest, the effects of trade structure on growth vary significantly between poor and rich economies. Additionally, the results present

evidence supporting the argument that trade relations established with rich countries are

more beneficial than those established with poor countries. Understanding these differences should yield better policy prescriptions. Even though the structure of trade is a complex concept that might be relevant in more than one sense, we regard our efforts as a first step in this new direction. Empirical measures of trade characteristics or trade patterns and configuration are scarce. Theoretical models that incorporate those elements and guide our intuition are rare as well. As this paper demonstrates, however, the study of trade structure characteristics

could lead to new ways of understanding the trade-growth relationship. Several other studies have gone beyond the simpler measures of trade. Dollar (1992), Sachs and Warner (1995) and Warcziarg (1998), for example, create their own indicators of openness. However, as pointed out by Rodriguez and Rodrik (2000), these measures might not achieve the purpose they were conceived for, since they are likely proxies for a wide range of policy and institutional differences. The measures of the number of trading partners and the concentration of trade used in this paper, in contrast, are clearly related to trade and are simpler to interpret. Additionally, there is a related branch of the literature that has looked for evidence regarding the effects of trade on total factor productivity. Representative papers of this literature include Coe and Helpman (1995), Keller (1998) and Edwards (1998). While these papers look for changes on productivity that arise as the result of increased trade openness (ratio of total trade to GDP), we look for changes that arise as the result of the structure of trade, independently from trade openness. The remainder of the paper is organized as follows: The next section describes the conceptual arguments that characterize the expected empirical relationship between our

measures of trade structure and the rate of output growth.

Section 3 presents the

econometric specifications employed to test these arguments. Section 4 introduces the dataset and discusses the trade structure variables employed. Section 5 presents the empirical results obtained and, finally, the last section summarizes our conclusions.

2 Theoretical Background
The likely effects of international trade on economic growth can be formally derived from endogenous growth models that study the creation, diffusion and adoption of better production technologies. Some of these models are now well known in the literature, as is the case for Romer (1990), Grossman and Helpmann (1991), and Rivera-Batiz and Romer (1991), and the idea that trade promotes growth through the transmission and creation of knowledge is widely accepted among economists3. In these models, the trade of goods and services can generate positive effects on economic growth from at least two sources: First, international trade may generate knowledge spillovers that expand the number of intermediate goods known by domestic producers, where these intermediate goods can be understood as new ideas, designs, or managerial processes that lead to increased productivity. international trade may act as a diffusion mechanism for foreign technologies. Second, international trade may produce incentives for local research and development. Trade expands the number of potential buyers and thus, the potential for economic profits associated with innovation, brand recognition, patent registrations, and any improvements over competing firms products. Trading goods with foreign markets also forces the producer to react to a new environment and innovate in the process. That is,

These innovations might take the form of necessary adjustments to packaging and transportation methods due to local climates and infrastructure, product adjustments made to entice consumers with different tastes, innovations in the production process that comply with foreign sanitary regulations, etc. What is still to be explained are the actual mechanisms that diffuse technology across countries and the process through which countries adopt these newer technologies. In this paper, we argue that both the number of trading partners and the concentration of trade among partners are important elements of those mechanisms.

2.1 Economic growth and the number of trading partners Every country generates knowledge through research or experience; some countries generate more knowledge than others. But when countries trade the goods and services that embody those original ideas, the knowledge is indirectly transferred to the trading partner who learns by induction. In this manner, international trade facilitates the

transmission of technologies across countries; where the amount of technological diffusion depends on how much a country trades and with how many partners. In our econometric specification, besides including a measure of trade openness, we also include the number of trading partners. For any given amount of goods traded, the exposure to new ideas is greater when the number of countries involved in the trade grows larger, as each partner contributes an original amount of knowledge. Thus, ceteris paribus, countries with more trading partners would be expected to face a greater number of foreign technologies and to experience greater economic growth as the adoption of those technologies increases productivity.

In addition, as the number of trading partners increases, the potential market in which domestic producers can sell their products expands both in size and diversity. Both effects are expected to support growth. The expanded size of the potential market attracts local and foreign investments. These investments, in turn, have been shown to play an important role for technological diffusion and innovation (Grossman and Helpman (1991)). Noticeably, the size of the potential market is related to the number of trading partners but is not necessarily related to traditional measures of trade openness that capture actual volumes of imports and exports. In turn, the expanded diversity of the market puts all producers in contact with a wider rage of consumer tastes, government regulations, climates, etc. As local producers seek to compete in more foreign markets, they accommodate to the particular characteristics of these markets and innovate in the process. The greater the number of trading partners, the greater the innovations that become necessary. As mentioned before, these

innovations might take the form of necessary adjustments to packaging and transportation methods, product adjustments made to entice alien consumers, innovations in the production process that comply with foreign sanitary regulations, and others. Finally, as the number and diversity of partner countries increases, the number of potential competitors for the local market increases as well. Competition, in turn, leads to higher productivity and greater economic growth. Evidence on the effects of

competition on productive efficiency can be found in Vickers and Yarrow (1991) and Bourbakri and Cosset (1998). The marginal benefit of an additional trading partner, however, could be different in poor economies than in rich economies. On the one hand, if newer technologies increase

the productivity of older technologies, then the effect of an additional trading partner on growth should be smaller for a poor economy since they have a smaller stock of knowledge to begin with. On the other hand, because the stock of knowledge is smaller in poor countries, the contribution of an additional trading partner in terms of new knowledge (and thus, its impact on growth) is likely to be greater for undeveloped economies. While these effects operate in opposite directions, they both suggest

asymmetry between rich and poor countries. Similarly, the effects of having a greater number of rich (technologically superior) trading partners could be different from those of having a greater number of poor (technologically inferior) trading partners. If the benefits of trade come mainly from technological diffusion, then it could be expected that a country learns more by trading with a rich partner than by trading with a poor partner; as the imports from richer countries are likely to be composed of relatively more technology intensive goods (see Schneider (2005) for an empirical analysis of a similar argument). In our econometric specifications we explore both the possibility that an increase in the number of trading partners generates an effect on economic growth that differs between poor and rich countries, and the possibility that the effect of increasing the number of rich trading partners is different from the effect of increasing the number of poor trading partners.

2.2 Economic growth and the concentration of trade There are also reasons to expect that the concentration of trade among trading partners may have a positive impact on output growth. First, whenever technologies from

different countries act as substitutes and compete with each other, then concentrating trade on a partner(s) with a single technological configuration could facilitate technology adoption. The implementation and dissemination of foreign innovations is made easier when more people become familiar with the language and the conventions of these innovations (see Yanikkaya (2003), for related ideas). Consider the case of an economy faced with two alternative technologies. Both technologies are equally efficient, but they were conceived by different sources and do not produce the desired results when combined. In this case, the economy might be better off by adopting only one of them (instead of both); since this might speed up the learning process and the collaboration across industries that use the same technology. Examples of such technologies can be found in American and European video formats and home heating technologies, alternative medical procedures, alternative drug prescriptions and related drug interactions, or in economic papers that model time in either discrete or continuous form. Second, given the total number of partners and the total volume of trade, it may be cost-efficient for countries to concentrate trade on one or a few partners. Concentrating trade might help minimize trade costs associated with the congestion of insufficient infrastructure like ports, airports, diplomatic posts or costumes personnel. Thus, where the trade related infrastructure is not well developed, concentrating trade might lower transportation costs. These costs have been shown to affect trade significantly. Frankel et al. (1995), for example, point to transportation costs as one main reason why trade blocs may form. Similarly, it is possible that trade-related public policies that promote local investments become easier and cheaper to implement when trade is concentrated

among few partners. Such policies might include, for example, the dissemination of foreign regulation manuals, public information regarding national and foreign trade barriers, the coordination of national and international sanitary regulations, etc. Finally, small economies that depend on certain industries for a significant share of their GDP might benefit from greater trade concentration as far as these specific industries benefit from greater concentration. As suggested by Pickering and Sheldon (1984), concentration at the industry level might lead to economies of scale in production and marketing, as well as stronger competitive positions. As with the number of trading partners, the asymmetry between rich and poor countries can also be expected from the effects of trade concentration on economic growth. Because rich countries hold and create more knowledge to begin with, have a more diversified economy, have better infrastructure and more efficient governments, they would be expected to benefit less from trade concentration than poor countries do.

3 Econometric Specifications
In what follows, we marshal the conceptual arguments presented above to explore the following hypothesis: There is an empirical relationship between trade structure and economic growth. Past empirical studies that fail to control for trade structure may suffer from omitted variable biases. An increase in the number of partners is associated with an increase in the rate of economic growth. This effect differs between rich and poor countries.

 Given the volume of trade and the number of trading partners, an increase on the concentration of trade is associated with an increase in the rate of economic growth. This effect is smaller for rich countries.

Empirically, the relationship between trade and growth has been tested in a variety of econometric specifications. The earlier literature, represented by Feder (1982), focused on the relationship between exports and growth. More recent empirical studies, however, have used an empirical formulation closer to that of Barro (1991) or Levine and Renelt (1992). We will use the latter, more recent specification for most of the analysis and use the former only as a robustness test of our results. Following Barro (1991) and Levine and Renelt (1992), the rate of growth of GDP perr & capita ( y y ) is regressed against a vector C of standard explanatory variables and an r additional set of explanatory variables that capture the issues of interest. The vector C is

composed of measures of the initial per-capita level of income (y0), the rate of population & growth ( L L ), the secondary school enrollment ratio (SED), and the ratio of investment to GDP (I/Y)4. With respect to the variables measuring trade, a wide variety of measures have been used in the empirical growth literature. The most commonly used measure is the ratio of total trade (exports plus imports) to GDP (see, for example, Easterly, Loayza and Montiel (1997), or Frankel and Romer (1999)). Other, less common measures include composite indices of openness (Dollar (1992), Sachs and Warner (1995), Wacziarg (1998)) and measures of trade restrictions and policies such as tariffs and export taxes (as in Harrison (1996) and Edwards (1988)). Since our focus of attention is not on the volume of trade or

10

the degree of openness in economic policies, but on the structure of trade, we choose to utilize the simpler and more popular variable of Total Trade as a percentage of GDP to control for trade openness5.

Finally, we include our measures of trade structure in a vector TS of additional explanatory variables. The resulting econometric specification takes the following form:
r & y = + C + Trade + TS + y

(2).

Throughout the empirical results section we either conduct the standard OLS regressions (using the averages for the whole sample consider) to estimate equation (2), or use fiveyear averages for all variables and conduct fixed effect regressions6. When conducting fixed effects regressions, all regressions include a country-specific term as well as a period-specific term. As pointed out by authors like Easterly et al. (1997) and Temple (1999), there are several reasons why it is preferable to take advantage of the time-series aspects of our data in addition to the cross-sectional variation. First, given that the variables of interest vary significantly over time, the use of cross-sectional averages misses out on much of the information emerging from changes that occur over time in each country. Second, the use of panel data allows us to control for omitted time-specific effects and countryspecific effects and, thus, to diminish the endogeneity biases associated with institutional or sociopolitical factors, and world wide shocks. In this regard, the use of five year averages allows one to take advantage of the time series aspects of the data while still capturing long-run aspects of economic growth.

11

As mentioned earlier, we use the alternative specification proposed by Feder (1982) to test the robustness of our results. Feder (1982) tested a two-sector model of growth in which export goods generate positive externalities for domestic production. In his basic & econometric formulation, the rate of growth of output ( Y Y ) is determined by the & investment to output ratio (I/Y), the rate of population growth ( L L ) and the product of & the exports to output ratio (X/Y) and the rate of growth of exports ( X X ). We modify Feders formulation in two ways. First, when studying the effects of exports on growth we also control for the structure of trade as described by our data. Second, we reformulate the specification so that per-capita income growth (y) is used as the dependent variable. The resulting specification is summarized in equation (3):
& L & y I = + ( ) + L y Y
& X X + TS + X Y

(3)

Other authors contemporaneous to Feder (1982), like Balassa (1978) or Tyler (1981) used simpler one-sector models for which the relevant exports-related regressor in the & & econometric specification was (X X ) instead of (X X ) ( X Y ) as in equation (3). In our robustness tests we allow for these possibilities as well.

4 Data Description and Trade Structure Variables

The empirical analysis uses data from a large sample of countries during the period 1980-2000. Values of annual population growth (POP), real income per capita (GDP), annual GDP growth, secondary school enrollment rates (SED), the investment share of GDP (Investment), the share of total trade to GDP (Trade Openness), and the share of

12

government expenditures in GDP (Government) were extracted from World Bank's World Development indicators (2004). With respect to the variables capturing the structure of trade, we incorporate two variables into the analysis: the total number of trade partners a country had at each point in time and the concentration of trade volumes among these partners. The number of trading partners for each country was extracted from the Direction of Trade Statistics database (DOT) from the International Monetary Fund. Also from the DOT database, we use the monetary values of total imports and total exports for each country/trading partner pair and construct a Herfindahl-Hirschman concentration index of trade (HHI) for all countries. This index measures the

concentration of trade among all trading partners; where a low number indicates low concentration. The index was computed as follows:

T i j HHI = N j Ti j j
N

Where N and Ti j denote the total number of trading partners and the total value of trade (exports plus imports) between countries i and j, respectively7. Table 2 presents

summary statistics for the HHI across regions of the world and confirms our initial claim that the changes on the trade structure have been more pronounced than changes on the levels of trade. It should be noted that even though the HHI index described above is a function of the number of trading partners, these two variables are not necessarily related and, a priori, there should not be a multicolinearity problem for the regression analysis. The addition of a new trading partner could result in a higher, lower, or constant degree of

13

trade concentration. The actual result depends on the level of trade emerging from the addition of the new trading partner, with respect to the already existing trade volumes. Having said this, there is a possibility that the relationship between the number of trading partners and trade concentration is such that, by chance, there is multicolinearity between the variables. To explore this possibility, we computed the correlation between the changes in the number of trading partners with the changes in trade concentration (HHI), the results show that the correlation coefficient between these two variables is 0.14, suggesting that multicolinearity between these variables is not an issue for the econometric analysis. Appendix 1 shows the source and the definition of all data variables used in this study. Additionally, Table 3 offers a summary of descriptive statistics for these variables for each of the samples explored; all countries, poor countries and rich countries.

5 Empirical Estimation Results

We start by using five year averages of all relevant variables and conducting fixed effects estimations of equation (2). The specific time periods for the averages are 1980 1984, 1985 1989, 1990 1994, and 1995 1999. All estimations included a country

specific term as well as a period specific term. The presence of both country and period specific effects could not be rejected for any of the estimated regressions presented here based on a Likelihood Ratio test. Additionally, the results presented are corrected for heteroskedasticity using Whites correction method. Tables 4, 5 and 6 present the results of estimating equation (2) for the whole sample of countries, the sample of poor countries, and the sample of rich countries, respectively.

14

In all these tables, column (1) presents the estimation results for the base specification where the trade structure variables are not included, while columns (2) - (6) present those obtained for specifications that include the trade structure variables. Columns (2) (4) present the main results for the trade structure variables and columns (5) - (6) explore the possibility that the effect of increasing the number of rich trading partners is different from the effect of increasing the number of poor trading partners. To this effect we compute the number of rich and poor trading partners and include them separately in the regression analysis. A country was categorized as poor if its income in 1980 was smaller than the income of the median economy in 1980. The results of the basic specification regarding trade openness as measured by total trade to GDP ratio resemble those of other studies that use similar methodologies and data (see, for example, Harrison (1996), Yannikaya (2003)). Trade openness has a positive and significant coefficient for the poor countries sample, a mostly positive but statistically insignificant coefficient for the total sample, and a negative but mostly insignificant coefficient for the rich countries sub-sample. It is only for poor countries that the level of trade openness appears to impact growth positively and significantly. With respect to the number of trading partners, the estimated coefficient is positive and statistically significant for almost all of the econometric specifications across the three samples explored in the analysis8. Additionally, the value of the estimated coefficients for the number of trading partners imply that the effects of this variable on GPD growth rates is of considerable economic magnitude. Focusing on Table 5, where the results of the estimation for the poor countries sample are presented, we see that the estimated coefficient for the number of trading partners, in column (4), is approximately 0.00052.

15

This coefficient implies that, everything else constant, increasing the number of trading partners by 10 (just a fourth of the standard deviation) is associated with an increase in the rate of growth of 0.52 percentage points9. Given that the average rate of annual growth in most undeveloped countries lies beneath one percent, the potential benefits of increasing the number of trading partners appears to be substantial. When the sample is divided into rich and poor countries, the estimated effect of an additional trading partner is much smaller for the poor countries sub sample than for the rich countries sub sample. The average value for the estimated coefficient is approximately 0.0004 for poor countries and 0.0008 for rich countries (see Tables 5 and 6). As explained before, poor countries are likely to benefit less from additional trading partners when different technologies act as complementary inputs and when the acquisition of knowledge is mostly channeled through one relatively advanced trading partner. We now turn to the discussion of the effects that trade concentration has on economic growth. The estimated coefficients for the trade concentration variable were, for most of the cases considered, positive and statistically significant for both the total sample and the poor countries sub sample. In contrast, for the rich countries sub sample, the estimated coefficient was often insignificant and in some cases negative. Since the level of

concentration increases as the HHI index increases, the results imply that poor countries benefit from more concentrated trade while the evidence for the rich countries is mixed at best. The estimated coefficients for the index of concentration show that the economic magnitude of the effects from trade concentration could also be considerable. For

16

example, the coefficients for the poor and rich countries sub samples, reported in column 4 of Tables 5 and 6, imply that for poor countries a one standard deviation increase in the level of trade concentration (HHI index) leads to an increase in the GDP growth rate of 0.32 of a standard deviation (1.42 percentage points), while for rich countries the corresponding increase in the GDP growth rate is 0.19 of a standard deviation (0.97 percentage points)10. Again, the economic impact is significant. As noted before, columns (5) - (6) in Tables 4 through 6 present the results obtained for the analysis of the hypothesis regarding technological diffusion arising mainly from rich trading partners. The evidence provided from this expanded analysis generally favors this argument. The estimated coefficient for the number of rich trading partners is mostly positive and statistically significant for all samples studied, while the coefficient for the number of poor trading partners is always negative, but only significant for the whole sample and the rich countries sub sample. The conclusion that emerges from the arguments stated in section 2 and the empirical evidence presented above is that the elements that characterize the structure of trade do influence the rate of economic growth in a significant manner beyond the effects of trade openness. The evidence also supports our hypothesis regarding the differences between rich and poor countries. While both groups of countries seem to benefit from a greater number of trading partners, rich countries appear to benefit more and these benefits (for all countries) come mainly from trading with rich countries. Finally, while more trade concentration is beneficial in general, it seems to be more significant for poor countries than for developed economies.

17

5.1 Robustness of the Empirical Findings: Alternative Specification As a first robustness test for our results, we use the alternative econometric specification introduced by Feder (1982). Once again, the main modifications are the inclusion of the number of countries to which a given country exports to and the HHI index as explanatory variables of the economic growth rate. The resulting base econometric specification is that presented in equation (3). We should emphasize here that the specification presented in equation (3) is based on the argument that the effects of trade on economic growth arise from exporting activities, therefore the HHI index used in these regressions is computed only using export data for each country. Tables 7 presents the results for the whole sample of countries, the poor countries sample, and the rich countries sample, respectively. The estimated coefficients for the export growth term in Table 7 (all countries) are not statistically significant. At first glance, this result seems to be at odds with previous studies that reported a positive and significant coefficient for this variable. Noticeably, however, past studies often based their main results on samples of poor or less developed countries (including Feder (1982), Balasa (1978) and Tyler (1981)). In columns (3) through (6), where we consider the two sub samples, poor and rich countries respectively, the results are more similar to prior studies. The coefficient for exports growth is positive and significant poor countries, while it is negative for rich countries. With respect to the effect of the number of trading partners on growth, our previous conclusions remained unaltered. The estimated coefficient for the number of rich trading partners is mostly positive and significant across the three samples considered, while the one for the number of poor trading partners is negative and often insignificant. Thus, the

18

results from this alternative specification also suggest that the benefits of established trading relationships on economic growth arise from trade links with rich countries. Similarly, the results obtained for the trade concentration match those obtained with the original econometric specification to the extent that trade concentration affects growth positively. The estimated coefficients for the trade concentration, in this case measured by export concentration, on the GDP growth rate are always positive and significant.

5.2 Robustness of the Empirical Findings: Addressing endogeneity A separate concern regarding the robustness of our results is that the number of trading partners could be contemporaneously correlated with the error term, i.e., presence of endogeneity. The logic behind this concern is that as a countrys GDP expands, this country will search for other markets in which to sell its products and/or other countries will look to establish trading relationships with this country. If this is so, then the results presented so far might be biased. With respect to the trade structure variables, there are a couple of reasons why we do not think endogeneity biases are too severe in our results. First, although one could be concerned that the rate of growth of output might influence the rate of growth in the number of partners, it does not seem apparent why the rate of growth of output should affect the total number of trading partners. We are considering level effects, while the endogeneity concern seems more plausible for rate of growth effects. Second, trading relationships (and in some cases bilateral trade volumes) are at least partially influenced by governments international trade policies; which can be viewed as exogenous.

19

Finally, regarding trade concentration levels, there is no reason to believe that higher or lower GDP growth rates could lead to higher or lower trade concentration among trading partners. Therefore we do not believe that endogeneity could be an issue when using trade concentration variable. Still, we make an effort to correct for any potential endogeneity by using Instrumental Variables in our previous estimations. As a first step, we perform an OLS regression for the whole sample using long run averages (1980-2000) for all variables included in the estimation of equation (2). We then compute the IV estimation and compare the results. As for our choice of instrumental variables, we use physical access to international waters and tropical climate, both obtained from the Sachs and Warner dataset11, and a variable for the type of political regime, obtained from Easterly and Lus Global Development Network Growth Database12. Geographic and socio-political variables should serve as good instruments for the number of trading partners since trade relationships are partly determined by geographical characteristics of countries, as well as political relations. The results of these estimations for the total sample are presented in columns (1) and (2) of Table 8. As mentioned above, the fixed effects approach is preferable to OLS, therefore we just use this set of results as a first step in our efforts to address the possibility of our results being affected by endogeneity. The results presented in Table 8 show that the effects of the total number of trading partners are strongly significant for both the OLS and the OLS - IV estimation13. Unfortunately, the instruments used for the OLS regression cannot be used for the fixed effects regressions because they do not change over time. Therefore, following Bloom et. al. (2004) and Ranis et. al. (2000), we use predetermined but not strictly

20

exogenous variables as instruments for the IV estimation. The predetermined variables that are used in the auxiliary regression are the values for the previous period for GDP per capita, trade openness, and the number of trading partners. The argument for using these variables in the auxiliary regression is that endogneity, if present, emerges because an increase in economic growth could result in a higher level of trade and higher number of trading partners. Therefore by using the lag values of the number of trading partners, GDP per capita, and trade openness as instrumental variables, we break the possible contemporaneous effect between economic growth and the number of trading partners. This methodology is justified since the correlation between the residuals at time t are not correlated with the lagged variables. This approach can also be used for the cases where we split the number of trading partners into rich and poor. The only variation is that in that case the auxiliary regression uses the lagged value of the number of rich and poor trading partners. Columns (3)-(8) of Table 8 present the results obtained for the three samples considered in the analysis. We present the results for the regressions for which the number of trading partners was divided into rich and poor partners, but the same conclusions are reached for the case where the total number of trading partners is considered. For comparison purposes, columns (3), (5), and (7) report the results for the original panel regression for the adjusted sample (i.e. removing the first cross section in order to match the time period considered for the IV estimation). Columns (4), (6), and (8) report the results for the IV estimation. It is clear from the results presented that the conclusions regarding the relevancy of the number of rich and poor trading partners for rich countries and the whole sample still

21

hold. A higher number of rich trading partners has a positive effect on the economic growth rate for these two samples, while a higher number of poor trading partners has a negative effect. Regarding trade concentration, the results show that it affects growth positively for the overall sample and the poor countries sub sample, but has a negative effect on economic growth for the rich countries. Once again consistent with the original results presented in Tables 4 through 6. Finally, the results presented in columns (3), (5), and (7) of Table 8 can also be used as a robustness check for our original results when these results are compared to those presented in column (6) of Tables 4, 5 and 6. In this case the robustness check is with respect to the sample period. The results of Table 8 are based on a sample period that would correspond to the cross sections of 1985 1989, 1990 1994, and 1995 1999, while those presented in Tables 4, 5, and 6 also include the cross section of 1980 1984. When the results are compared across the two sample periods, the signs are the same, and the magnitudes and statistical significance are relatively similar. This suggests that our results are robust to sample period selection as well.

6 Conclusions and Future Research

There are reasons to believe that the structure of trade has significant implications for the rate of economic growth. Most empirical studies overlook the conditions under which trade takes place and concentrate mostly on trade volumes and measures of openness. In this paper, instead of concentrating on the level of trade volumes or the degree of openness of an economy, we give attention to the structure of trade as

22

characterized by the number of trading partners a country has, and the concentration of trade among those partners. We find that the number of trading partners is positively correlated with growth across all countries; this effect is greater for rich countries. Trade concentration is positively correlated with growth for all countries in general; this effect is mostly attributed to poor countries. These results are robust to changes in the econometric specification used, the number of explanatory variables included, and corrections for heteroskedasticity. In terms of policy, our findings may inform policy makers as well as the international community that ponders the benefits of international trade agreements. In particular, our findings suggest that high trade concentration, such as is the case for developing countries whose trade is concentrated with a major economic power14, may not be a matter of serious concern.

23

Appendix 1: Variable Definitions

Real income per capita (GDP): Yearly Gross Domestic Product per-capita, constant 1995 USA dollars. World Development Indicators (WDI), 2004 GDP Growth: Annual growth rate of GDP per-capita. WDI, 2004 Annual population growth (POP): Average annual population growth. WDI, 2004 Secondary education (SED): Gross Enrollment Rate in secondary schooling. WDI, 2004 Investment share of GDP (Investment): Yearly Gross Fixed Capital formation (% of GDP). WDI, 2004 Trade Openness: Total Exports plus Total Imports as percentage of GDP. WDI, 2004 Government: General government final consumption expenditures (% of GDP). WDI, 2004 Number of Trading Partners: Total number of countries with whom a given country trades (exporting or importing). Direction of Trade Statistics, IMF 2004

24

References
Balassa, Bela (1978), Exports and Economic Growth: Further Evidence, Journal of

Development Economics, vol. 5, no. 2, pp. 181-189.

Barro, Robert J. (1991), Economic Growth in a Cross-Section of Countries, Quarterly

Journal of Economics, vol. 106, no. 2, pp. 407-443.

Bloom DE, Canning D, Sevilla J. (2004) The effect of health on economic growth: a production function approach, World Development, vol. 32, pp 1-13. Boubakri, Narjess and Jean-Claude Cosset (1998). The Financial and Operating Performance of Newly Privatized Firms: Evidence from Developing Countries, The

Journal of Finance, vol. 53, no. 3, pp 1081-1110.

Coe, David T. and Elhanan Helpman (1995), International R&D Spillovers, European

Economic Review, vol. 39, no. 5, pp 859-887.

Dollar, David (1992), Outward-Oriented Developing Economies Really Do Grow More Rapidly: Evidence from 95 LDCs, 1976-1985, Economic Development and Cultural

Change, vol. 40, no. 3, pp. 523-544.

Easterly, William, Norman Loayza and Peter Montiel (1997), Has Latin Americas postreform Growth Been Disappointing?, Journal of International Economics, vol. 43, no. 2, pp. 287-311. Edwards, Sebastian (1993), Openness, Trade Liberalizations, and Growth in Developing Countries, Journal of Economic Literature, vol. 31, no. 3, pp. 1358-1393. Edwards, Sebastian (1998), Openness, Productivity and Growth: What Do We Really Know?, The Economic Journal, vol. 108, no. 447, pp. 383-398. Feder, Gershon (1982), On Exports and Economic Growth, Journal of Development

Economics, vol. 12, no. 1, pp. 59-73.

25

Frankel, Jeffrey A. and David Romer (1999), Does Trade Cause Growth? The

American Economic Review, vol. 89, no. 3, pp. 379-399.

Frankel, Jeffrey A. Ernesto Stein and Shang-jin Wei (1995), Trading Blocs and the Americas: The Natural, the Unnatural, and the Super-natural.

Journal of

Development Economics, vol. 47, no. 1, pp. 61-95.

Grossman, G. and E. Helpman (1991), Innovation and Growth in the Global Economy, MIT Press, Cambridge, MA. Harrison, Ann (1996), Openness and Growth: A Time-Series, Cross-Country Analysis for Developing Countries, Journal of Development Economics, vol. 48, no.2, pp. 419-447. Keller, Wolfgang (1998), Are International R&D Spillovers Trade-Related? Analyzing Spillovers Among Randomly Matched Trade Partners, European Economic Review, vol. 42, no. 8, pp. 1469-1481. Levine, Ross and David Renelt (1992). A Sensitivity Analysis of Cross-Country Growth Regressions. The American Economic Review, vol. 82, no. 4, pp. 942-963 Matsuyama, Kiminori (1992), Agricultural Productivity, Comparative Advantage, and Economic Growth, Journal of Economic Theory, vol. 58, no. 2, pp. 317-334. Pickering, J.F. and I.M. Sheldon (1984), International Trade performance and concentration in British industry, Applied Economics, Vol. 16, pp. 421-442. Ranis G, Stewart F, Ramirez A. (2000), Economic growth and human development.

World Development, vol. 28, pp 197-219.

Rivera-Batiz, Luis and Paul M. Romer (1991), Economic Integration and Endogenous Growth, The Quarterly Journal of Economics, vol. 106, no. 2, pp. 531-555.

26

Rodriguez, Francisco and Dani Rodrik (2001), Trade Policy and Economic Growth: A Sceptics Guide to the Cross-National Evidence, NBER Macroeconomics Annual

2000. Cambridge, MA: MIT Press, pp. 261-324.

Romer, Paul (1990),Endogenous Technological Change, Journal of Political Economy, vol. 98, no. 5, pp. S71-S102. Sachs, Jeffrey and Andrew Warner (1995), Economic reform and the process of global integration, Brookings Papers on Economic Activity, vol. 1, pp. 1-118. Schneider, Patricia Higinio (2005), International trade, economic growth and intellectual property rights: A panel data study of developed and developing countries, Journal

of Development Economics, vol 78, pp. 529-547.

Temple, Jonathan (1999), The New Growth Evidence, Journal of Economic Literature, vol. 37, no. 1, pp. 112-156. Tyler, William G. (1981), Growth and Export Expansion in Developing Countries: Some Empirical Evidence, Journal of Development Economics, vol. 9, no. 3, pp. 121-130. Vickers, John and George Yarrow (1991). Economic Perspectives on Privatization,

The Journal of Economic Perspectives, vol. 5, issue 2, pp. 111-132

Wacziarg, Romain (1998), Measuring the Dynamic Gains from Trade, The World

Bank Economic Review, vol. 15, no. 3, pp. 393-429

Walde, Klaus and Christina Wood (2005), The Empirics of Trade and Growth: Where are the Policy Recommendations?, International Economics and Economic Policy, vol. 1, no. 2-3, pp. 275-292. Yanikkaya, Halit (2003), Trade Openness and Economic Growth: A Cross-Country Empirical Investigation, Journal of Development Economics, vol. 72, no. 1, pp. 5789.

27

Table 1. Trade Trends 1970-2000

All Countries Trade (%GDP) Number of partners* Europe and USA Trade (%GDP) Number of partners* Latin America Trade (%GDP) Number of partners* Trade (%GDP) Africa Number of partners* Trade (%GDP) Asia Number of partners*

1970-1974 1975-1979 1980-1984 1984-1989 1990-1994 1995-1999 2000-2003

58.30 68.33 75.37 72.06 78.25 83.45 88.52

46.43 51.38 52.91 56.89 67.71 85.94 93.98

65.22 74.05 80.36 78.98 84.95 96.08 108.61

105.54 109.81 95.11 95.58 112.52 146.89 161.49

54.34 67.88 76.24 74.39 78.69 79.10 77.09

38.27 45.53 52.83 57.61 66.91 80.36 87.05

58.96 67.15 69.60 64.78 68.09 74.49 76.47

31.79 35.08 38.38 42.63 51.54 64.56 71.36

54.82 65.65 79.54 75.02 80.91 86.35 94.04

47.45 52.70 62.23 69.78 80.01 96.33 104.33

* Number of partners = number of countries to whom goods where exported to (IZ)

Table 2. HHI Trends 1970-2000

All Countries 1970-1974 1975-1979 1980-1984 1984-1989 1990-1994 1995-1999 2000-2003 2201 1977 1730 1706 1647 1564 1580 Europe and USA 1131 986 979 999 1058 981 912 Latin America 2273 2419 2289 2316 2215 2137 2005 Africa 2216 1927 1626 1446 1373 1383 1485 Asia 2291 1873 1507 1519 1531 1346 1332

28

Table 3. General descriptive statistics (Country-Averages from 1980-2000).

GDP Growth Rate

Initial GDP

Population

Sedondary Education

Investment Trade (%GDP) ALL COUNTRIES

Number of Trading Partners

TOTAL TRADE Number of Number of Rich Trading Poor Trading Partners Partners

Concentration (HHI)

Weighted Exports Growth

Exports Growth

Concentration (HHI)

EXPORTS BASED Number of Number of Trading Rich Trading Partners Partners ALL COUNTRIES 72.2016 188.0000 0.4000 48.6401 742 42.8307 88.4000 0.2000 23.0584 742

Number of Poor Trading Partners

Mean Maximum Minimum Std. Dev. Observations

0.0096 0.3201 -0.4365 0.0493 663

5,766.9260 45,951.9500 49.3231 8,762.2910 641

1.7912 8.7285 -5.0608 1.3685 780

55.7700 146.3200 2.7000 33.2156 635

22.1647 86.7935 2.5255 8.4763 630

77.8120 268.8018 2.3515 43.4732 651

89.7679 198.0000 0.4000 48.0790 742

52.0720 90.2000 0.2000 21.0372 742

37.7468 108.4000 0.2000 28.2960 742

0.1614 0.8190 0.0326 0.1115 742

227.9661 19,401.8000 -1,005.1040 891.8052 562

6.5375 691.3065 -41.1255 30.2144 563

0.2036 0.8193 0.0393 0.1483 742

29.4105 100.0000 0.2000 26.7492 742

POOR COUNTRIES SAMPLE Mean Maximum Minimum Std. Dev. Observations 0.0067 0.2982 -0.3250 0.0448 276 812.2830 3,404.7060 49.3231 667.4884 276 2.2442 7.0285 -5.0608 1.1238 276 33.0677 101.7300 2.7000 22.6379 249 21.3017 70.6529 2.5255 9.3171 262 69.4226 231.9663 12.8760 36.8659 266 80.7333 191.6000 2.0000 37.7185 273 48.7260 89.6000 1.4000 17.8924 273 32.0073 102.2000 0.6000 20.9598 273 0.1475 0.5034 0.0467 0.0921 273 150.0608 2,459.3360 -877.5405 277.5435 250 4.8841 38.7992 -37.0092 8.2488 251

POOR COUNTRIES 0.2022 0.7748 0.0480 0.1417 273 62.3993 183.4000 2.0000 36.9367 273 38.8066 87.2000 1.4000 18.8727 273 23.5927 96.8000 0.6000 19.1981 273

RICH COUNTRIES SAMPLE Mean Maximum Minimum Std. Dev. Observations 0.0118 0.3201 -0.4365 0.0523 387 9,513.4510 45,951.9500 97.6258 10,097.9500 365 1.5431 8.7285 -3.9789 1.4267 504 70.4148 146.3200 3.2700 30.6325 386 22.7790 86.7935 6.1789 7.7777 368 83.6083 268.8018 2.3515 46.6768 385 95.0269 198.0000 0.4000 52.5151 469 54.0196 90.2000 0.2000 22.4573 469 41.0949 108.4000 0.2000 31.3425 469 0.1695 0.8190 0.0326 0.1208 469 290.3903 19,401.8000 -1,005.1040 1,167.9797 312 7.8677 691.3065 -41.1255 39.8876 312

RICH COUNTRIES 0.2044 0.8193 0.0393 0.1522 469 77.9075 188.0000 0.4000 53.5194 469 45.1731 88.4000 0.2000 24.8976 469 32.8043 100.0000 0.2000 29.7960 469

29

Table 4. Dependent variable: per capita GDP growth (All Countries)

(1) -0.000004 **
-2.395678

Initial GDP Population Secondary Education Investment Trade Openness (Total Trade to GDP ratio) Number of Trading Partners Number of Rich Trading Partners Number of Poor Trading Partners Total Trade HHI

ALL COUNTRIES (2) (3) -0.000004 * -0.000005 *

-2.804989 -2.666757

(4) -0.000005 *
-3.119962

(5) -0.000003 **
-2.229205

(6) -0.000003 **
-2.396127

0.003804
1.310745

0.005185 ***
1.736725

0.004564
1.571616

0.005933 **
2.044834

0.005604 ***
1.846349

0.005984 **
2.075951

-0.000434
-1.371912

-0.000489 ***
-1.653094

-0.000341
-1.229374

-0.000396
-1.518711

-0.000052
-0.357029

-0.000035
-0.238816

0.001912 *
4.002104

0.001847 *
4.120541

0.001929 *
3.990859

0.001846 *
4.129383

0.001810 *
3.946391

0.001810 *
3.965089

0.000041
0.471501

-0.000018
-0.173721

-0.000024
-0.234321

-0.000076
-0.717285

0.000109
1.625355

0.000076
1.077607

0.000784 *
3.686365

0.000762 *
4.797275

0.002537 *
3.796425

0.002436 *
3.901482

-0.000976 **
-1.998974

-0.000905 ***
-1.895677

0.122581 **
2.377725

0.139734 *
4.369265

0.090824 *
3.881132

No. Observation Adj. R squared

512 0.4417

506 0.4744

506 0.4572

506 0.4862

506 0.5182

506 0.5227

(*) 1%, (**) 5%, and (***) 10%, t-statistics in italics.

30

Table 5. Dependent variable: per capita GDP growth (Poor Countries)

(1) -0.000051 *
-4.517157

Initial GDP Population Secondary Education Investment Trade Openness (Total Trade to GDP ratio) Number of Trading Partners Number of Rich Trading Partners Number of Poor Trading Partners Total Trade HHI

POOR COUNTRIES (2) (3) -0.000053 * -0.000050 *

-4.384735 -4.618537

(4) -0.000052 *
-4.664286

(5) -0.000049 *
-4.371293

(6) -0.000048 *
-4.498285

0.006432 **
2.240495

0.006781 **
2.356131

0.007270 *
2.777914

0.007924 *
3.082697

0.007176 **
2.555777

0.008262 *
3.108087

0.000029
0.114332

-0.000009
-0.036045

0.000075
0.331289

0.000026
0.115835

-0.000010
-0.041575

0.000027
0.120222

0.002462 *
4.027496

0.002493 *
4.352627

0.002320 *
3.643316

0.002342 *
4.010316

0.002462 *
3.981291

0.002297 *
3.598189

0.000361 **
2.464323

0.000327 **
2.055764

0.000329 **
2.322198

0.000272 ***
1.651022

0.000370 *
2.600000

0.000329 **
2.344537

0.000354
1.504774

0.000524 **
2.349184

0.001163
1.510314

0.001219 ***
1.617521

-0.000197
-0.528160

-0.000039
-0.120612

0.132831 *
5.246587

0.154613 *
5.145123

0.157979 *
5.864371

No. Observation Adj. R squared

236 0.4833

236 0.4848

236 0.4974

236 0.5041

236 0.4921

236 0.5126

(*) 1%, (**) 5%, and (***) 10%, t-statistics in italics.

31

Table 6. Dependent variable: per capita GDP growth (Rich Countries)

(1) -0.000004 **
-2.026270

Initial GDP Population Secondary Education Investment Trade Openness (Total Trade to GDP ratio) Number of Trading Partners Number of Rich Trading Partners Number of PoorTrading Partners Total Trade HHI

RICH COUNTRIES (2) (3) -0.000004 ** -0.000005 **

-2.216887 -2.162535

(4) -0.000004 **
-2.308894

(5) -0.000003
-1.604933

(6) -0.000003
-1.591552

-0.002602
-1.115898

-0.002357
-0.566990

-0.002294
-0.879154

-0.002418
-0.629746

-0.002759
-1.445532

-0.002702
-1.323933

-0.000670 ***
-1.722326

-0.000656 ***
-1.845457

-0.000558 ***
-1.654267

-0.000553 ***
-1.710508

0.000060
0.380469

0.000060
0.385018

0.001255 **
2.027497

0.001028
1.616657

0.001412 **
2.255940

0.001175 ***
1.907210

0.001266 **
2.152944

0.001222 **
2.059598

-0.000225
-1.399505

-0.000261 ***
-1.773236

-0.000308 ***
-1.785954

-0.000322 **
-2.030919

-0.000064
-0.424180

-0.000048
-0.289760

0.000828 *
7.434122

0.000759 *
9.390180

0.003418 *
5.682781

0.003476 *
5.301107

-0.001767 *
-3.179349

-0.001793 *
-3.093731

0.078877
1.198480

0.080765 **
2.341196

-0.033357
-1.000213

No. Observation Adj. R squared

276 0.5069

270 0.5454

270 0.5187

270 0.5473

270 0.6349

270 0.6333

(*) 1%, (**) 5%, and (***) 10%, t-statistics in italics.

32

Table 7. Dependent variable: per capita GDP growth (Feders Specification)

Investment Population Weighted Exports Growth Exports Growth Number of Rich Trading Partners Number of Poor Trading Partners HHI Index (Exports)

All Countries (1) (2) 0.00101 * 0.00056 *

4.92839 3.78343

Poor Countries (3) (4) 0.00201 * 0.00201

8.66811 27.14964

* **

Rich countries (5) (6) 0.00101 * 0.00056

4.92839 3.78343

0.00051
0.14279

0.00029
0.08281

0.00817 *
3.40702

0.00507
2.19853

0.00051
0.14279

0.00029
0.08281

-0.000011 *
-8.06184

0.000046 *
7.35430

-0.000011 *
-8.06184

-0.00036 *
-34.46021

0.00159
5.31426

-0.00036 *
-34.46021

0.00222
2.76417

* *** **

0.00226 *
2.93876

0.00099
1.37622

0.00105
1.41803

0.00222
2.76417 -1.69576
*

0.00226
2.93876 -1.86866

-0.00080
-1.69576

-0.00083 ***
-1.86866

-0.00010
-0.28792

-0.00010
-0.34025

-0.00080 ** -0.00083 *** 0.04146

2.00007
**

0.04146
2.00007

0.03918 **
2.58948

0.04159 *
4.66913

0.05265
4.60104

0.03918
2.58948

**

Adj. R Squared Number of observations (*) 1%, (**) 5%, and (***) 10%, t-statistics in italics.

0.616174 529

0.637905 524

0.547639 247

0.548540 247

0.616174 282

0.637905 277

33

Table 8. Dependent variable: per capita GDP growth (Robustness Regressions)

(1) OLS -0.0000002
-0.793490

Initial GDP Population Secondary Education Investment Trade Openness (Total Trade to GDP ratio) Number of Trading Partners Number of Rich Trading Partners Number of Poor Trading Partners Total Trade HHI

All Countries (2) OLS IV (3) Panel -0.0000011 ** -0.0000050 *

-2.631488 -2.947152

(4) Panel IV -0.0000050 *

-3.335380

Poor Countries (5) Panel (6) Panel IV -0.0000626 * -0.0000609 *

-7.231790 -7.562165

Rich Countries (7) Panel (8) Panel IV -0.0000059 ** -0.0000065 **

-2.758716 -2.349752

-0.005041 **
-2.552059

-0.002078
-0.799102

0.008261 *
3.704539

0.008507 *
3.833978

0.009778 *
4.953186

0.009640 *
4.996621

-0.001020
-0.550121

-0.000069
-0.017446

-0.000180 **
-2.037032

-0.000051
-0.569925

-0.000163
-0.789295

-0.000234
-1.372010

0.000032
0.084817

0.000011
0.028280

0.000013
0.090371

-0.000246
-1.576464

0.001536 *
4.271307

0.001509 *
3.154588

0.001513 **
2.139334

0.001582 ***
1.723113

0.001581 **
1.632060

0.001579 **
1.658527

0.001128 **
2.567255

0.001563 ***
1.793756

0.000107 **
2.125885

0.000192 **
2.167317

0.000134 ***
1.706251

0.000228
1.213165

0.000273
1.373829

0.000261
1.333385

0.000003
0.015130

0.000175
1.217736

0.000206 *
4.441109

0.000369 *
2.930752

0.002960 *
4.381577

0.001096 *
4.243621

0.001269 **
1.913360

0.000949
1.578527

0.003740 *
5.030071

0.001260 *
4.544451

-0.0014020 *
-2.670692

-0.0001860
-0.351714

-0.0000936
-0.495250

0.0002470
0.482186

-0.0021240 *
-3.401437

-0.0005000 *
-2.710275

-0.0132780
-1.079674

-0.0211140
-1.375304

0.0588890 *
3.806073

0.0878930 *
3.516244

0.1763360 *
4.732703

0.1640970 *
4.087528

-0.0796660 *
-3.330870

-0.0541400 *
-6.329680

No. Observation Adj. R squared

120 0.4410

91 0.4016

400 0.5692

361 0.4168

180 0.6177

177 0.5568

220 0.6268

184 0.3681

(*) 1%, (**) 5%, and (***) 10%, t-statistics in italics.

34

For a more details on this see Edwards (1993), Temple (1999) and Walde and Wood

(2005) for literature reviews.

2

Many more are in the works. See the WTO secretariat web link

http://www.wto.org/english/tratop_e/region_e/regfac_e.htm).
3

Other authors like Feder (1982) or Matsuyama (1992) have modeled growth in two

sector models where the exporting sector generates positive externalities for the rest of the economy. While these models are different from the one by Romer (1990), the intuition behind the effects of trade through the exchange of ideas is similar.
4

In order to check for the robustness of the results, additional explanatory variables like

the ratio of government expenditures to GDP or the primary school enrollment were also
r included in C . Including these other variables produced no changes in our results.
5

Authors like Harrison (1996) and Edwards (1998), however, have argued that most

measures of trade openness and trade policy used in the literature capture roughly the same aspects and could be used interchangeably.
6

When conducting fixed effects regressions, the variable for the initial income level per-

capita is the value at the beginning of each five-year period.

7

When equation (3) is estimated, the data used to calculate this index was limited to

Export volumes instead of Trade volumes.

8

The exception is the estimated coefficient for column (2) of Table 5. Throughout the study, gross growth rates were used and the size of the coefficients

reflects this choice. For example, a one percent growth rate was recorded in the data as 0.01.

35

10

The descriptive statistics, Table 3, report that the standard deviation of the GDP growth

rate is equal to 0.044, while that for the HHI index is 0.092. Therefore a one standard deviation of the HHI index implies a change of (0.154*0.092)=0.0142, or 1.42 percentage points, in the GDP growth rate, which is about 0.32 of a standard deviation of the dependent variable (0.0142/0.044).
11

Sachs and Warner data set is published on the Center for International Development

Web site accessible from http://www.cid.harvard.edu/

12

http://www.worldbank.org/research/growth/GDNdata.htm The results for the poor and rich country samples are not presented here for matters of

13

space, but the conclusions follow those discussed for the overall sample, in other words they statistical significance and signs of the relevant variables remain.
14

Mexicos trade with the U.S., for instance.

36