...

This is a sumamry of Francisco Rodriguez and Dani Rodrik’s 1999 paper, “Trade Policy and Economic Growth: A Skeptic’s guide to Cross-National Evidence”, published by the National Bureau of Economic Research. It is available in full at http://www.nber.org/papers/w7081

Several recent studies have investigated the systematic relation between growth and trade or growth and liberalization using empirical data comparing countries with different trade regimes.

Rodriguez and Rodrik, in their 1999 paper, review four influential papers which are typical of this recent literature. The papers are Dollar 1992, Ben-David 1993, Sachs and Warner 1995 and Edwards 1998. Dollar 1992 is the most heavily cited paper on the link between growth and openness. Sachs and Warner 1995 is a close second, but it also develops an index of openness which has since been widely used in the cross national research on growth. The two other papers are also widely cited and are representative of different types fo methodologies. Ben-David 1993 considers income convergence in countries which have integrated (such as the European Community) and is representative of a strand in the literature which investigates the effect of trade on income convergence. Edwards 1998 undertakes a robustness analysis, using a wide range of trade policy indicators to see whether the relation between trade openness and growth is robust across indicators.

Rodriguez and Rodrik 1999 argue that these studies, which typify the approach of recent studies devoted to this task, have several methodological deficiencies which mean that the existing studies cannot be taken as evidence of a systematic link between trade openness and economic growth. This summary avoids most of the technical economics and econometrics details. For a fuller treatment, consult the paper itself.

Dollar 1992:

Dollar constructs two separate indices which are demonstrated to be negatively correlated with growth from 1976-85 for a sample of 95 developing countries. The indices are ‘index of real exchange rate distortion’ and ‘index of real exchange rate variability’. The indices aim to capture the notion of outward orientation. Outward orientation is discussed as a combination of two things. One, the level of protection, especially for production process inputs, is relatively low, resulting in a sustainable level of exchange rate which is favorable to exporters (the exchange rate is thus not too ‘distorted’ by the level of protection). Two, there is relatively little variability in exchange rate so that incentives are consistent over time (Dollar 1992: 524).

If the indices are a good measure of outward orientation, their correlation with growth would be potentially informative about the relation between trade openness and growth. However, Rodriguez and Rodrik criticize the indices as poor measures of outward orientation. Whatever correlation there may be between the indices and growth, if the indices are poor measures of outward orientation, then the correlation reveals little of significance about the relation between growth and free trade.

The distortion indexis sensitive to the form in which trade restrictions are applied. A country can seem more outward oriented in the index just by replacing import barriers with export taxes, even though they have the same relevant effect (of changing the domestic price ratio between import competing products and exportables) (Rodriguez and Rodrik 1999: 10). The first criticism relates to the fact that for the index to do a good job in ranking trade regimes according to restrictiveness, it is necessary that export policies play a comparatively small role. This necessary condition is not met in the world.

The adequacy of the index also depends considerably on the Law of One Price. The Law of One Price(LOP) holds that, over the long run, the price of a tradable good produced in different countries can diverge from one another when expressed in a single currency only when there exist trade restrictions or transport costs. The distortion index is based on a comparison of the price levels in various countries expressed in terms of a single currency[1]. The assumption that LOP holds allows us to regard differences in price levels across countries as indicative of trade restrictions.

However, LOP does not hold in the world and thus the suitability of the index as a measure of trade orientation is further weakened. A recent review of empirical research on the issue suggests that tradables for which the LOP holds are the exception rather than the rule (Rogoff 1996: 650). It seems that differences in price levels between countries can be significantly affected by factors such as exchange rate policy which affects nominal exchange rate as well as by geographical factors such as proximity to sea routes and to global markets (Rodriguez and Rodrik 1999: 12).

For these reasons, the distortion index cannot be considered a good measure of outward orientation. The index is affected by factors other than trade restrictions, such as nominal exchange rate policy, geographical factors and the presence of export taxes and subsidies. Let us turn to the variability index.

The distortion index ranks countries by what it claims is the degree of outward orientation (though, we have seen reasons to question whether the index measures outward orientation). For many developing countries, this ranking more or less matches the impression created by other works which have sorted countries by trade orientation. However, there are some significant anomalies, where the distortion index judges some country X to be more outward oriented than country Y when previous works which sort countries by trade orientation have judged the relation between X and Y to be the opposite (Dollar 1992: 530). The number of anomalies falls sharply if we include the variability indicator as well as the distortion one (Dollar 1992: 531).

Thus Dollar uses a weighted average of the distortion and variability indices to create an orientation index (Dollar 1992: 532). However, as Rodriguez and Rodrik point out, the variability index has little to do with trade orientation (Rodriguez and Rodrik 1999: 12). Its inclusion in Dollar’s study appears to be somewhat ad hoc in order to match his rankings of trade orientation more closely to the rankings suggested by existing work and is not well motivated as capturing any aspect of outward orientation.

The indices used by Dollar are poor measures of outward orientation. The relation between countries ranked by the indices and their rates of growth thus shed no light on the correlation between free trade and growth.

Sachs and Warner 1995

The authors devise an index of openness which combines several aspects of trade policy. The rationale for combining the various policies into a single index is that these are different ways an economy can close itself to trade. The Sachs-Warner (SW) openness indicator (OPEN) is a zero-one dummy which takes the value of zero if the economy was closed according to any of these (Sachs and Warner 1995: 22):

1. it had average tariffs rates above 40% (TAR)

2. its non-tariff barriers covered on average more than 40% of imports (NTB)

3. it had a socialist economic system (SOC)

4. it had a state monopoly of exports (MON)

5. its black market premium exceeded 20% during either the 70s or the 80s. (BMP)

The black market premium is a measure of exchange control, and a large BMP indicates rationing of foreign exchange, which tends to be a form of import control (Sachs and Warner 1995: 25). The rationing of foreign exchange creates a black market for the currency.

The authors find that there is a difference in growth rates between economies which their variables class as open and as closed. In particular, economies classed as open by the OPEN indicator tend to have higher growth rates than ones classified as closed (Sachs and Warner 1995: 36-7).

Rodriguez and Rodrik discover that the correlation between OPEN and growth rates is largely due to the combination of BMP and MON. Little of the indicator’s statistical strength is lost if the other three policy aspects are excluded. Recall that these other three include the two most direct measures of trade policy – tariff (TAR) and non-tariff barriers (NTB), as well as SOC. An indicator constructed using only BMP and MON (let’s call this new indicator BM) very closely matches the results yielded by OPEN in classifying countries as closed or open to trade. By contrast, an indicator constructed out of TAR, NTB and SOC performs quite poorly in matching the classification yielded by OPEN (Rodriguez and Rodrik 1999: 16-9). BMP and MON do most of the work in the OPEN indicator’s classification of economies as either closed or open.

The degree of correlation between economies classed as closed by OPEN and lower rates of growth is matched to a large extent by the same correlation between economies classed as closed by BM and lower rates of growth. Rodriguez and Rodrik narrow their focus to BMP and MON and their relevance for rates of growth, since the other three components of OPEN do little work. Recall that the context is to analyze the relation between trade openness and growth via the OPEN indicator. The authors investigate whether the two variables MON and BMP correlate with other variables which have a detrimental effect on growth. If they do, and if these other variables have little to do with trade policy, then the correlation demonstrated by Sachs and Warner between OPEN and rates of growth cannot establish that differences in trade policy are the relevant factor in the differences in growth rates. For, these other variables which have little to do with trade policy (and which are yet to be discussed) could be responsible for the correlation between rates or growth and economies classed as closed by the OPEN indicator.

Their investigation reveals that, except for one country (Mauritius), the inclusion of MON in the SW dummy is indistinguishable from the use of a Sub Saharan Africa dummy for the countries studied by Sachs and Warner. The only info we can extract from this is that Sub Saharan economies have grown slower than the rest of the world in the period studied by Sachs and Warner (Rodriguez and Rodrik 1999: 20).

As for BMP, it is not a good measure of trade policy because it is also a proxy for many other variables unrelated to trade policy. Black market premia correlate highly with level of inflation, with debt/exports ratio, with wars and with institutional quality. It is reasonable to expect these latter things to affect growth (Rodriguez and Rodrik 1999: 22).

BMP and MON are overwhelmingly responsible for the effect of the SW index on growth. Yet, the two indicators are not very robust at controlling for variables which are not related to trade policy, such as an Sub Saharan Africa geographical dummy or indicators of macro-economic and political distress or poor institutions. Thus, the Sachs Warner study cannot justify inferences about the effect of trade openness on growth.

Edwards 1998:

Edwards 1998 does not construct a new indicator of openness as was the case in the previous two articles considered. Rather, Edwards considers the robustness of the openness-growth relationship to the use of different indicators. The idea behind this is that although particular indicators can be faulted, this is less of an issue if the estimated positive coefficient on openness is robust to ways in which openness is measured (Edwards 1998: 386).

Edwards carries out his analysis using nine different indicators of openness present in the literature, as well as a measure of schooling and a measure of initial income. Edwards concludes that the results suggest with tremendous consistency that there is a significantly positive relationship between openness and productivity growth (Edwards 1998: 391).

Rodriguez and Rodrik will argue that the results do not warrant such a strong conclusion. The robustness of the regression results are largely due to weighting and identification assumptions which Rodriguez and Rodrik argue are inappropriate. Edwards tests 19 specifications of the theory that openness promotes growth, spanning the nine indicators and a constructed indicator combining some of these indicators, and calculating regressions using weighted and instrumental weighted least squares.

Rodriguez and Rodrik argue against some of the assumptions in Edwards. They add that if we are to take seriously the fact that property rights are not excludable from the productivity growth regressions, (and we must do so, given the extensive literature on the protection of property rights being an important determinant of growth) we face the conclusion that of the 19 different specifications, we find a negative and statistically significant correlation between trade restrictions and productivity growth in only 3 cases (Rodriguez and Rodrik 1999: 25-7). The three cases which survive this initial scrutiny use the following indicators: the Collected Taxes Ratio (ratio of trade tax revenue to total trade), the World Development Report Index (a subjective index), and the Heritage Foundation Index (another subjective index). For these cases, Rodriguez and Rodrik criticize the data as inadequate.

In Edwards 1998, the Collected Taxes Ratio (CTR) is calculated based on raw data from the IMF (Edwards 1998: 389). Rodriguez and Rodrik are skeptical of this data, as many of its numbers for developing countries are implausible. For example, India, known to have one of the highest tariff rates in the world, is found to have an average ratio of only 2.4% according to the raw data. Rodriguez and Rodrik recalculate the CTR using the most recent data (at time of publication of their paper) from the World Bank and obtain very different, and more plausible, results. Using this new data set to perform the same regressions as Edwards and obtain quite different results which undermine Edwards’ result for the CTR indicator. The coefficient on average duties is, contrary to expectation, positive (meaning that higher CTR correlates positively with growth), albeit insignificant (meaning that the coefficient is too small to confidently establish a positive correlation). If import and export duties are considered separately, import duties yield a positive and significant coefficient and export duties are insignificant (Rodriguez and Rodrik 1999: 27).

The two other cases concern subjective indices constructed by the World Bank (the World Development Report Index) and the Heritage Foundation. Subjective indices are commonly criticised as suffering from judgment bias. It is telling, the authors emphasise, that the only indices which stand up their analyses are subjective ones.

In addition, the Heritage index rates trade policies in 1996. Yet, the period over which Edwards considers growth rates is 1980-90. Rodriguez and Rodrik recalculate the Heritage Foundation Index using its methodology but with data about trade restrictions from 1985-88, a period more fitting for Edwards’ study. When they replicate the relevant equation from Edwards using this new data, they find that the coefficient of correlation is insignificant. This suggests that the significance of the Heritage Foundation Index in Edwards’ regression is due to changes in trade policy which occurred after 1990.

The World Bank’s index exhibits several anomalies. For example, high growth Korea is ranked as more open than moderate growth Malaysia despite having higher tariffs and non-tariff barriers, higher exchange rate distortions and lower export/GDP ratio. Indeed, in another paper, Edwards himself had noted serious probems with the index (Edwards 1993: 1386-7). Further, given that the primary data for the World Bank’s index is the same as that used in some of the other work discussed earlier in the Rodriguez and Rodrik paper (where the data was not similarly telling of a connection between openness and growth), the authors conclude that it is likely that the significance of the index stems from subjective biases in the way that countries have been classified (Rodriguez and Rodrik 1999: 29).

All of this combines to leave the Edwards 1998 with little credibility in its conclusion of having demonstrated significant correlation between trade openness and productivity growth.

Ben-David 1993

Ben-David’s paper measures the effect of trade policy on income by investigating whether trade liberalization promotes income convergence. The expectation that liberalization may lead to income convergence is based in the factor price equalization theorem (FPE). The theorem holds that, under various conditions (including zero transport costs, identical technologies and equal numbers of goods and factors), free trade leads to factor price equalization. The idea is present in the common intuition that as a developed country and a developing country open to trade with one another regarding a certain product, the wage (the price of the labour factor of production) will fall in the former country and rise in the latter, and will approach equality.

Ben-David looks primarily at income convergence among countries belonging to the European Economic Community (EEC). He shows that there is a decrease in the dispersion of incomes among these countries in the post war era. Of course, this decrease in income dispersion could be due to many factors. To make the case that it as due to trade liberalization, he argues that (i) the convergence is not merely a continuation of a long term convergence trend unrelated to postwar economic integration; (ii) that the European countries which did not enter the free trade agreement did not experience the same level of convergence; and (iii) other subsets of economies in the world which were not economically integrated did not experience convergence. Rodriguez and Rodrik criticise these three arguments and thereby cast doubt on Ben-David’s conclusion.

Rodriguez and Rodrik argue that, despite what Ben-David claims, he has not shown that (i) is the case. They do this partly by including Germany in the calculations of the long term trend. Ben-David had excluded Germany because it has historically been one of the poorest of the six countries which began the EEC and yet is one of the wealthiest in Europe today. Were Germany included in Ben-David’s analysis, it might be claimed by detractors that all of the convergence among the EEC countries is due solely to Germany’s influence. Its exclusion, claims Ben–David biases the conclusion away from convergence, making his job harder (Ben-David 1993: 662).

However, Rodriguez and Rodrik point out that while Germany’s exclusion makes it harder for Ben-David to establish income convergence in post-war liberalization, it also hides Germany’s role in any potential long term trend of convergence. Its exclusion biases results in favour of the hypothesis that there is no long term convergence predating postwar liberalization. When Germany is included, it is much harder to discount the possibility that the income convergence is part of a long term trend predating post-war liberalization. It now appears that income dispersion has been on a downward trend for many decades before the war (Rodriguez and Rodrik1999: 31).

A look at the pre war history of the countries further reveals that the previous major period of trade liberalization from the 1820s until 1880 was actually accompanied by increased income dispersion (Rodriguez and Rodrik1999: 32). Moreover, the retreat from free trade beginning around 1880 was a period of income convergence which continued to the start of WWI.

The analysis of Rodriguez and Rodrik reveals that there is no long run tendency for trade liberalization to be associated with income convergence in the EEC and that there has been a trend towards income convergence beginning around the start of the 20^th century and thus predating post war liberalization (Rodriguez and Rodrik1999: 33).

Ben-David then turns to the next three countries to join the EEC (after the EEC began with the initial members) and to whether their income differentials behaved just as those of the original six (Ben-David 1993: 662-3). He looks at changes in income dispersion amongthese three countries – UK, Ireland and Denmark. Rodriguez and Rodrik point out that the relevant issue is income convergence between these countries and the otherEEC members and not convergence among the three countries themselves, as the relevant policy change is that the three countries relaxed trade restrictions between themselves and the initial EEC members. Looking at this, Rodriguez and Rodrik find that the three countries do not show a marked change in their pattern of convergence with EEC levels of income after the change in their trade policy regarding the EEC in 1965.

Ben-David then compares the income convergence of subsets of world economies which, unlike the EEC group of economies, have not liberalized their trade with one another. From the groups of economies he considers, he obtains the result he desires. Rodriguez and Rodrik criticise Ben-David’s selection of groups opf economies for this comparison. They note that the groups whose incomes are converging in Ben-David’s selection are geographically close, while the diverging groups are geographically distant. The obvious confounder is that what Ben-David claims is the effect of economic integration may in part be the effect of geographical proximity. For a fair comparison, we must ask whether free trade or lack thereof among geographically adjacent economieswould lead to convergence or divergence (Rodriguez and Rodrik 1999: 34).

Looking at geographically adjacent groups of economies, Rodriguez and Rodrik find that the economies of East Asia – canonical examples of open economies – have experienced steadily divergingincomes since the 1960s. Meanwhile, the economies of Latin America – canonical examples of closed economies – have experienced a steady decrease in divergence during their period of import substitution. Moreover, dispersion jumps upwards as countries in Latin America begin to liberalize trade. The authors make the point that while there are areas of the world where convergence has been accompanied by trade liberalization, there are also regions where the opposite is true. Ben-David’s study cannot justify belief in a systematic relationship between convergence and liberalization (Rodriguez and Rodrik 1999: 34-5). The authors also cite Slaughter 1998 as a recent empirical study which shows that liberalization does not lead to convergence in the cases studied (formation or EEC, of European Free Trade Agreement, liberalization between EEC and EFTA and the Kennedy round tariff cuts under GATT).

Ben-David’s paper fails to establish that the income convergence follows trade liberalization as serious criticisms can be brought against each of (i), (ii) and (iii) which were to be established in order to settle the larger question.

This ends my discussion of the papers reviewed in Rodriguez and Rodrik. Let me summarise the sorts of criticisms they have mounted. They questioned the adequacy of openness indicators at measuring the effects of trade policies; pointed out cases where openness indicators may be reflecting factors which influence growth but which are not part of trade policy; noted cases where the data used in studies was unreliable and where the results of a study changed substantially when more plausible data was substituted; they argued that the methodology in some cases was poor and again noted the substantial change in results when better methodology was used.

Rodriguez and Rodrik rightly caution that their paper reviewing recent attempts to demonstrate a link between openness and growth is not intended to bolster support for the position that trade restrictions promote growth. The context for their paper is their belief that there has been a tendency in academic and policy discussions to greatly overstate the systematic evidence in favour of trade openness. The motivation for their paper is to provide a corrective to this, showing that the touted evidence is not convincing and to foster a stance of humility about the state of research into the connection between trade and growth (Rodriguez and Rodrik 1999: 39).

************************************************************************************