University of Basel
On leave from the IMF

This version: July, 98

Abstract

This paper presents empirical evidence that institutional reform may be a fruitful way of promoting external performance. In a set of 49 countries 15 different indicators of institutional quality are shown to be closely associated with real export growth. These indicators include measures of the accountability of rule making, property rights security , predictability of laws, corruption and political instability. Within a subset of 21 African countries differences in the security of property rights and the rule of law are the most important factors for explaining differences in external performance. This result is holds even if I control for income, policy distortions and other measures of political instability. Furthermore, I show that differences in the perceived security of property rights do not depend on societal factors such as the level of ethnic division.

Introduction

Africa’s general economic performance since 1960 has been described as tragic. Certainly one of the areas where the tragedy is most pronounced is in the area of trade. Of course, this is particularly true when Africa’s performance is compared with the--now battered-- star performers in East Asia. While East Asian countries saw more than two decades of high growth based largely on increasing exports, most African countries have not hardly seen any export growth at all. On average real merchandise export growth of sub-Saharan African countries has been 2.6 percent over the past three decades while the average for many South East Asian countries has been in the region of 10 percent (see graph 1). However, there were also significant differences in export performance within Africa. For instance countries such as Cameroon from 1980 to 1990 had an export performances which is comparable to Indonesia's and Mauritius saw export growth which rivaled South East Asian star performances. Nevertheless, overall the contrast between the regions remains, and it begs for an explanation.

There is by now a substantial body of literature which tries to explain why East Asian countries have managed to expand their external sector so rapidly and successfully. The consensus of this literature is that East Asian countries conduced a conscious strategy of export promotion. This export promotion in many cases consisted not only in leveling the playing field for exporters, but in tilting it in their favor by employing interventionist policies which range from coordination of investment plans to directed credits and infant industry protection. Africa, so it has been suggested, could learn from this experience.

The catch is that many of these selective interventions have already been tried in Africa, however, with very different results than in East Asia. Harrold, Jayawickrama and Bhattasali (1996) conduct a detailed study of the success of East Asian type export promotion policies in African countries. Their conclusion is worth quoting:

This conclusion reflects the general tone of the study. After a careful comparison of schemes in African and East Asian countries the authors suggest that industrial and trade policies have not been successful in Africa because of a more general failure of institutions including bureaucratic failure, mistrust in the relations between governments and the private sector, corruption and political instability.

The conclusion that institutional failure is an important obstacle to better economic performance has also been supported by other recent studies. To name just a two, Fischer, Hernandez-Cata and Khan (1998) note that African countries need far-reaching improvements in governance, the African Competitiveness Report (1998) shows that one of the greatest concerns of local and foreign businesses were corruption, lack of stability and of transparency. Even though there is wide acceptance of the proposition that African institutions are a determinant of external performance there is, to my knowledge, no cross country empirical study which tests this. Presumably, the lack of adequate data on institutional performance is a reason for this gap.

This paper sets out to analyze the relationship between institutions and export performance with special reference to Africa. I proceed in two steps. First I investigate if institutional quality can explain differences in export performance across countries, i.e. not only in the direct comparison between East Asia an Sub-Saharan Africa but also in a larger set of countries. Then I test the same proposition hypothesis within Sub-Saharan Africa. I use data on institutional quality form various sources, the main one some of which have only become available very recently. I find that a number of indicators of institutional are closely associated with export performance in the cross county regressions. In explaining differences within Africa only the most powerful indicators are general property rights security.

The paper is organized as follows. The section 2 discusses the data and the empirical strategy. Section 3 presents the results from regressions on real export growth for a set of 49 developing and developed countries and for a subset of for 21 African countries. Section 4 proceeds to explore whether differences in the security of property rights in Sub-Saharan Africa can be explained with political or cultural factors.

Detailed data on institutional quality for many countries has only become available recently. The problem is that institutional quality is not easy to measure objectively. For instance, there are simply no objective data on the level of corruption or the effective security of property rights enforcement. Data on such issues is typically obtained from surveys of experts or of the private sector. In this paper I rely on three sources of data on institutional quality, one is derived from an expert survey and the other is based on private sector surveys and the third are objective variables on the form of the political system.

The first source of data is the International Country Risk Guide (ICRG), a private firm which produces annual ratings of bureaucratic quality and the rule of law (variable names in italics) based on experts’ surveys. Indicators are rated from 0 (worst option) to 6 (best option).

The second source of data is the WDR+ survey collected by the World Bank and the University of Basel in preparation for the World Development Report 1997. This data is based on private sector surveys in 73 countries. I will use ten different measures from this survey. The first four are all related to the credibility and accountability of rule making. They are: the degree of policy surprises, the credibility of announcements, extent of information on new rules and the degree to which business can participate in making new rules. As a variable which measures the predictability of law enforcement I use the predictability of judiciary enforcement (as well as two variables by ICRG, the quality of the bureaucracy and rule of law). A third set of variables measures the degree to which property rights are perceived to be secure. They are theft and crime, and security of property rights. The last set of variables from this source are corruption variables. They measure the frequency of corruption, the uncertainty of corruption and the extent to which corruption is perceived as an obstacle to business. All indicators are rates form 1(worst) to (6) best)

Finally, I also use three objective variables which measure the form of the political system and political instability. The first variable is a variable from the Polity III data set presented by Jaggers and Gurr (1996) which rates the degree of democracy in the election process. This variable is rated from 0 (not democratic) to 10 (fully democratic) for the year 1990. The second is a dummy variable war which takes the value 1 if there was a war or a civil war in the country and the last variable, ethnic is a index of ethnolinguistic fractionalization, for 1960. It measures the probability that two randomly selected people from a given country will not belong to the same ethnolinguistic group. Both variables are obtained from Easterly and Levine (1997).

The classical trade theories try to explain the pattern of trade between countries rather than the overall export performance of a particular country. Accordingly there is a substantial empirical literature which tests the theoretical predictions of these models. The traditional variables which drive trade models are differences in factor endowments, namely capital and labor. However there is a large number of additional factors which have been shown to be important in explaining trade patterns. They include, variables of human capital, variables which capture transportation costs (country size, country distance to trading partners) variables which measure transaction costs (language, tariffs, quotas) and distortions in the economy (real exchange rate distortion, black market premium, duties on imports and exports). One way of approaching the question of export performance is to estimate predicted trade shares from based on a trade model and then compare these to the actual trade shares. In this paper I will follow a more direct approach because I am interested in explaining differences in overall export growth, rather than explaining the structure of trade.

In the empirical section I will estimate regressions for the growth in the volume of exports over the period 1980 to 1995 and test if different measures of institutional quality can help explaining differences in external performance across countries. As a minimal specification I will control for the initial level of GDP per capita and average inflation during the period. The level of GDP per capita is a summary measure of the stage of development, which captures a number of factors relevant for trade performance: At higher levels of income trade tends to be is more specialized and therefore grows at a faster pace than traditional exports. Higher levels of income go hand in hand with higher levels of education, therefore income is also a proxy for the level of human capital. I tested specifications which include the level of secondary school enrollment instead of the level of income and this did not alter the results. Inflation, the second control variable, in the base specification attempts to measure overall distortions in the economy. Instead of inflation I also tested more direct measures of distortions, such as an index of protectionism constructed by Leamer (1988), the black market premium, and the level of export and import duties and I introduced the size of the country as additional variable. However, the explanatory power of all these variables is quite low. For the entire period the R2 are in the region of 10-20 percent. For the period 1980-1990 the explanatory power is better, especially in the case of the African sample. Here a regression with the base specification obtains a R2 of 35 percent.

There is a question of causality in this approach, however, I would argue that it is not particularly strong. I postulate that good institutions (a predictable rule making process, property rights security, a stable political system etc.) lead to a better export performance by providing enterprise with a fertile business environment. The reverse argument would be that higher exports and a larger share of exporting firms create political pressure which leads to better institutions. This argument may in part hold for institutions such as the participation between government and business associations or even for corruption. It is much more difficult to make for such fundamental things as the security of property rights.

The aim of the next section is to explain differences in external performance both across countries and within Africa. As has been noted above, there are substantial differences in external performance within Africa and I will test whether differences in institutional variables can explain these. The set of countries is shown in Appendix I.

This section presents the empirical results for the two sets of countries. The first subsection test institutional variables in a set of 49 developing and developed countries. The second subsection does the same for the African sample only. Within the African sample I find evidence that the most important variable for explaining differences in export performance across African countries is the security of property rights. The third subsection explores this issue in more detail.

Table 1 shows the results of multivariate regressions. Every row shows one regression. The dependent variable is always the average growth of export volumes over the period 1980-1995. The independent variables are the initial GDP per capita and the average inflation rate 1980-1992 and a institutional variable INST VAR. The first column presents the coefficient of the respective institutional variable. T-Statistics are in parenthesis.

The first set of institutional variables all measure the credibility and accountability of the rule making process. It has been suggested that one of the features of successful export promotion in East Asia was that private enterprise was involved in decision making through business groups and deliberation councils. Therefore the private exporting sector was not only informed about changes in rules and regulations but could actively participate in the process. The results was that there were few negative policy surprises and government announcements were generally credible, allowing the export business to plan their investment strategy and expand. In the cross country regressions three out of the four variables, policy surprises (regression (1)), information, (regression (3)) and participation (regression (4)) are significantly related to export growth. Out of the significant ones, information , regression (3) has the best fit and the highest coefficient.

Dependent Variables INST VAR Constant DP Inflation Adj. R2

(1) Policy surprises 3.75 -8.4 -0.0001 -1.26 0.14

The second set of variables relate to the predictability of law enforcement. The first, regression (5), the predictability of the judiciary, is highly significant in the export regression. Rule of law and bureaucratic efficiency are two variables from the expert survey. Rule of law in regression (6) is clearly significant whereas bureaucratic efficiency (regression (7)) is marginally not (P- Value of 0.11).

The next set of variables measures the security of property rights. The first variable is based on a question which asked entrepreneurs to rate a list of potential obstacles to their business operations. The measure of crime and theft as a business obstacle (regression (8)) is highly significant in the export regression. Almost a third of the variation of export performance can be explained by this specification. The same is true for the second variable, the security of property rights (regression (9)).

The fourth set of variables capture the impact of corruption from different angles. The first is the mean of answers to a question in the WDR+ survey which asks for the frequency of corruption payments, the second variable is the standard deviation of responses to the same question. The idea is that the larger this standard deviation the higher the uncertainty of corruption in the respective country. For instance, a county may on average have a relatively low corruption, however, there are large and unpredictable differences in treatment of private firms and therefore the standard deviation of responses to this question would be high. It has been argued that uncertainty of corruption may be at least as damaging to economic performance as a high corruption. If the same argument is applied to external performance we would expect a negative correlation with export growth. The last variables is derived from a question which asks comparatively whether corruption is considered an important obstacle for doing business. The correlation between this variable and the frequency of corruption need not be perfect. In other words, even in a high corruption country, the local business community may be so accustomed to the situation and there may be such well established channels for bribing that corruption is no longer perceived as a mayor business obstacle. In fact, this last indicator of corruption is the only one which is significant at conventional level (regression (12)). The other two have the expected sign (i.e. positive for regression (10) and negative for regression (11)) but they are not significant indicating that the relationship between export performance and corruption is complex. This result is consistent with the East Asian story where, high levels of corruption in some countries have not prevented them from increasing exports at a rapid pace.

It should be noted that the variables tested so far are all subjective indicators. In other words they do not necessarily reflect the "true" scale of crime and theft or of corruption. They reflect the perceived scale of the problem from the point of view of the private sector. Given that these perceptions guide entrepreneurs’ decisions these subjective feelings --rather than objective measures of institutional problems -- should ultimately be relevant in determining economic performance. Nevertheless, there are a number of objective variables on political instability which can be thought of as proxies for perceived uncertainties. The last set of institutional variables test three of these measures which are commonly used in the empirical growth literature.

This last set of variables included measures of the form and the stability of the political system. Regression (13) includes a democracy variable. This variable is significant in the cross country export regression. The second, regression (14) includes a dummy which takes the value 1 if there was a war in the country. The war dummy has the expect sign but is not significant. The last regression (15) includes a measure of ethnic diversity. This variable is significant but has the expected negative sign. The idea is that ethnic diversity is a cause of uncertainty and of political instability and that countries which are more ethnically diverse will have lower economic performance. It seems that this hypothesis holds also for the export performance.

To conclude, out of fifteen variables tested 10 were significance and all had the expected sign, indicating that institutional and external performance are indeed closely associated. The variables which have the highest power in explaining differences in external performance across countries are the security of property rights and crime and theft. They are followed by variables which measure the credibility and accountability of rule making and the predictability of rule enforcement. Variables on corruption and on the political instability have mixed results.

In this sub-section I study differences in export performance across African countries. As noted above, export growth has been varied within African countries and the question is whether differences in institutional performance can help explain these differences in external performance.

The WDR+ data set includes 21 African countries (see Appendix I for a list of countries). Table 2 shows regression results for export growth in these 21 countries with the same right-hand variables as in the previous section.

The results from this exercise are quite striking. There is only one set of variables which contributes significantly to explaining differences in export growth across Africa, namely the variables related to security of property rights. Almost all other institutional variables are not significantly related to differences in export growth and some of them have the "wrong" sing.

Take for instance the variable on ethnic diversity. It is has a positive correlation with export growth across African countries, meaning the that countries that have been more ethnically diverse have actually managed to export more. The variable is not significant at conventional levels, i.e. there is a 15 percent chance that the true coefficient is zero. The fit of the regression impressively low. The war dummy, does similarly badly. It has the expected sign, i.e. negative but it is nowhere near significance. In other words, the fact that some African countries had a war and others did not does not help explaining differences in export performance. Neither can the form of the political system, corruption, or differences in accountability help explains such differences.

(1) Policy surprises -0.40 2.97 0.0005 -6.17 0.04

The only variable, in addition to security of property rights ones that has some explanatory power is the rule of law variable. At the same time this is a variables which is perhaps the most intimately related to the security of property rights. Regression (6) shows that differences in rule of law (plus the control variables) explain about 20 percent of differences in export performance. Regression (9) shows a similar result. The fit is best in regression (8). Countries where crime and theft was perceived as a large business obstacle where also the ones that had the lowest export performance. This results quite are robust to specification. I tested a number of other control variables, including other institutional variable. The results continued to hold. Given the strength of this result on property rights I will explore this issue in more detail in the next section.

I now take the analysis one step further down by asking if there are any institutional variables which would help explain the differences in security of property rights across African countries. For instance, it would seem obvious that in a country with a autocratic government, with large ethnic division or even with a war the security of property rights would be lower. Before testing this proposition I present some descriptive statistics on the variable which was most strongly associated with export performance within Africa as well as in the large country set. Table 3 shows the average answer to the questions how large crime and theft were perceived as obstacles for doing business in the respective country.

Source: WDR+ dataset

This table presents the country ratings sorted from the highest to the lowest obstacle. A rating of 1 means on average entrepreneurs in that country responded that crime and theft represented a major obstacle for their business operation. A rating of 6 means that this is not considered a problem at all. In other words the country were crime and theft are perceived as the largest business obstacle is South Africa and the one where crime and theft are perceived as the smallest obstacle is Mauritius. As noted above, this does not mean that the absolute level of crime and theft is highest in South Africa. But it indicates that the local private sector perceives it as being very high and indeed a serious obstacle to their business operations. In the specific case of South Africa it is worth mentioning --what will not come as a surprise-- that the perceived security of property has significantly deteriorated over the past ten years. But even then entrepreneurs seem to have taken into consideration the potential future conflicts, their rating from ten years earlier is 3.6.

Table 4 presents three regressions with three variables of property rights security as dependent variables. The independent variables are in the rows. The first regression tries to explain the perceived security of property rights with three variables, war, ethnic division and democracy. Non of the variables is significantly associated with perceived property rights security. The second regression uses the variables "crime and theft as a business obstacle" as a dependent variable and the same right hand variables. And obtains the same result. War, ethnic division and the form of the political system cannot explain differences in how large crime and theft are perceived as obstacles to doing business. Finally, the last regression does the same thing for the variable rule of law, the third variable which was significant in the African export regressions. Here the result is somewhat different. War and ethnic division are, again, not significant (war does not even have the expected sign) but democracy is. This indicates that in more democratic countries the rule of law tends to be better respected. However, this relationship is not very robust to changes in measurement as was shown with the previous two regressions.

It seems then that security of property rights in Africa is not systematically related to "fundamental" factors such as the propensity for political instability or by ethnic divisions. This could be interpreted as good news because it would imply that property rights security may be improved with institutional reform which it is much more under the control of governments than for instance the ethnic composition of a society.

The evidence presented in this paper suggests that a good strategy for promoting external performance in Sub-Saharan Africa would be institutional reforms. They should primarily be directed at improving the rule of law and property rights security.

(Source: The World Bank)

Comparison between Sub-Saharan Africa and three South East Asian Counties

(Source: The World Bank)

Benin

Cameroon

Chad

Congo

Cote d’Ivoire

Ghana

Guinea

Guinea-Bissau

Kenya

Madagascar

Malawi

Mali

Mauritius

Mozambique

Nigeria

Senegal

South Africa

Tanzania

Togo

Uganda

Zambia

Zimbabwe

Hong Kong

Malaysia

Singapore

South Korea

Thailand

Bolivia

Colombia

Costa Rica

Ecuador

Jamaica

Mexico

Venezuela

Austria

Canada

Fiji

France

Germany

India

Ireland

Italy

Jordan

Morocco

Portugal

Spain

Switzerland

Turkey

United Kingdom

United States

Return to the previous page

Return to the UNU Homepage