How does trade openness influence budget deficits in developing countries?

Abstract

This paper analyses the effects of trade openness on budget balances by distinguishing the effects of natural openness from those of trade policy. Revealed indicators of natural openness and trade policy are computed. Using GMM-system estimator, the econometric analysis focuses on 66 developing countries over 1974–98. The results show that trade openness increases a country's exposure to external shocks. This enforces the negative impact on budget balances of terms of trade instability. Additionally, trade openness influences budget balances through several other channels: corruption, income inequalities, etc. Then natural openness and trade policy have opposite effects: the former deteriorates budget balances whereas the latter enhances them.

JEL classifications:

• H62
• F02
• F13

Acknowledgements

The authors would like to thank Jha Raghbendran and two anonymous referees for their helpful comments; also the participants to the 58th Congress of International Institute of Public Finance (IIPF) in Helsinki, August 2002; and to the 51st Congress of the Association Française des Sciences Economiques (AFSE) in Paris, September 2002. The usual disclaimers apply.

Notes

1. Nevertheless, Barro (1985 and 1986) views the tax smoothing theory not only as a normative, but also as a positive theory. Thus, Barro has tested the tax smoothing model on 200 years of American and British data. The two experiences are, broadly speaking, consistent with the basic principles of tax smoothing: the debt to GNP ratios increase during wars, decrease in peacetime, and fluctuate with the business cycles. The problem with this theory is that ‘any fiscal policy can be rationalised from a tax smoothing perspective, if expectations are a “free variable” (Alesina and Perotti, 1994, p. 5). The tax smoothing model could explain even the high deficits in the 1980s in the US. Indeed, suppose that in the early 1980s, it became known that, with a temporary increase in military spending, the ‘cold war’ could have been won and, by the 1990s military spending could be cut below the initial level in 1980. The optimal policy is to cut taxes and to increase military spending in 1980s as well as to run deficits in the 1980s and surpluses in the 1990s.

2. The uncertainty on the duration of shocks has no impacts, on average, on budget deficits if the costs of errors are symmetric.

3. There is a widespread practice in volatile economies to systematically underestimate fiscal revenues in order to prevent overspending (Talvi and Végh, 2000).

4. Procyclical fiscal policies could be also explained by borrowing constraints. Potentially budget deficits out of negative shocks might be lower, in absolute value, than budget surpluses out of positive shocks of the same duration because of the asymmetry created by borrowing constraints, particularly in developing countries. The imperfections of the credit market imply that during a positive shock saving rates are lower than those predicted by optimal policies and during a negative shock dissaving rates are lower too. This could explain why during a positive (negative) shock both public spending increases (decreases) and tax rates decrease (increase).

5. The case of Nigeria in the 1970s is a good example. Nigeria provides what is almost a natural experiment for hypothesis that rents cause corruption. After the oil shock, observers have noted that Nigeria's oil income created extraordinary opportunities for corruption (Ades and Di Tella, 1999).

6. We are grateful to an anonymous referee for shedding light on this deficit creating nature of aid flows.

7. Nevertheless, replacing non tariffs barriers by tariffs can mitigate the negative impact of trade liberalisation on government revenue.

8. The data for terms of trade of goods and services (1995 = 100) come from the World Bank Global Development Network (GDN) web site: www.worldbank.org/research/growth/GDNdata.html

9. We thank an anonymous referee for suggesting us this idea.

10. If we change the sample, a relatively more open country in the first sample could become relatively more closed in the second and vice versa.

11. The data are not available for all countries on the whole period, so we used an unbalanced panel. Each country retained in the sample has at least two sub-periods.

12. See Baltagi (1995, chapter 8).

13. The first difference estimator exacerbates the bias due to errors in variables (Griliches and Hausman, 1986).

14. However some variables could be strictly exogenous.

15. However, the coefficient for trade policy ∗ instability in Column 5 is about 100 times lower than in Column 2. But, the comparison of the pseudo R2 suggests that the results of Column 2 are better.