World Bank Trade Adjustment Loans and Export Policy Distortions

Abstract

This work investigates whether World Bank loans fostering trade liberalization are associated with less distorted export policies, by employing some gravity model‐based measures of anti‐export bias, and a Herfindhal index of export revenues concentration. When accounting for non‐random selection in a sample of 88 developing countries over the period 1980‐2000, the receipt of trade adjustment loans seems to have reduced the policy distortion under scrutiny. Such a beneficial influence, however, vanishes when a longer time horizon is considered, casting doubts on the country ownership of waves of liberalizations supported by the Bank.

Key Words:

• Trade adjustment loans
• special interest groups
• anti‐export bias
• gravity models
• non‐random selection

Jel Codes:

• C2
• F13
• F35

Acknowledgements

The author wishes to thank Francesco Aiello, Mario Quagliariello, Damiano Silipo, and the editor Michael Connolly, for their helpful comments and suggestions on earlier versions of this paper. I am also grateful to Andrew Rose for sharing a software routine.

Notes

1. This literature takes into explicit account the political circumstances, which determine the final level of protection in a country. Grossman and Helpman (1994), in particular, establish that special interest groups are the main reason for the lack of free trade in a small, competitive economy, where the government and organized sectors undertake a two‐stage non‐cooperative game.

2. It should be noticed that the conditional equilibrium is characterized by a lower distortion, but the government is worse off than when unconditional assistance is granted, and this may constitute a “source of resistance to strict enforcement of conditionality”. Besides, the public welfare may be lower, therefore the government might attack conditionality by putting forward the worsening of public conditions.

3. More precisely, a dummy variable is included for each importing country in each year, so that its coefficient expresses “the distorting effects of each importing country’s policies in each year when compared to the mean for the entire sample”. To obtain the deviation of each country policy from a free trade benchmark, the difference between each country‐year effect and the maximum sample intercept is computed. The final policy measure is obtained by dividing these (positive) deviations for the predicted imports, when all variables are set to their mean and the intercept is set to its maximum.

4. As Subramanian and Wei (2003) highlight, this also allows the coefficients of the other control variables to vary over time, at the expenses of a loss of efficiency, which is negligible given the large dataset employed.

5. The same is done for the Hiscox and Kastner‐type measure, in order to avoid the country‐year fixed effects to subsume the effect of i’s income.

6. The observations are clustered at the pair level so that the error terms may be correlated within each couple of countries. Indeed, dyad‐specific factors, such as cultural and historical ties or particular preferences in one country for products of the other are likely to characterize the trade relationship between two countries.

7. The selection equation variables present some missing values, and this reduces the sample available for the Heckman approach in comparison to that available when estimating the substantial equation only.

8. The set of geographical dummy variables employed individuates six different areas: Sub Saharan Africa, Europe and Central Asia, East Asia and Pacific, Latin America, Middle East and North Africa, South Asia. This classification is drawn from the World Bank New Projects Database. The dummy variable indicating a low‐income country is drawn from WDI 2003.