Trade liberalization and import demand: The Central American experience

Abstract

We investigate the effect of trade liberalization on the price and income elasticities of demand for imports in five Central American economies. In contrast to recent studies, we find that liberalization has had little impact on import demand's sensitivity to price. However, the income elasticity is found to change significantly in three countries; doubling in two and halving in the third. Despite the similar characteristics and trade policies of the five economies, none of the explanations for structural change in import elasticities suggested in the literature can account for more than a minority of our results.

Keywords:

• imports
• trade liberalization
• structural change
• income elasticity
• price elasticity

JEL Classification:

• F14

Notes

1. As the average tariff rate (which is naturally quite highly correlated with LIB) is also included in the model (in the rpm variable), part of the change in the level of imports directly attributable to the shift to a more liberal regime will be accounted for by the resulting fall in rpm. Hence, strictly speaking, the coefficient on LIB reflects the impact of liberalization on imports that is not attributable to the change in the average tariff rate. Consequently, its sign is expected to be positive for the countries that removed significant non-tariff barriers as well. It is also conceivable that the coefficient on LIB could turn out to be negative if non-tariff barriers were unimportant and a significant proportion of tariffs had been set at prohibitive levels, as the change in rpm due to the tariff cuts multiplied by the estimated price elasticity would overstate the effect of liberalization on import demand.

2. We define t* as 1986 for Costa Rica, 1990 for El Salvador, 1987 for Guatemala and 1991 for Honduras and Nicaragua.

3. Following liberalization, each variable is measured relative to its value in period t* to make the two parts of the estimated relationship meet (ceteris paribus) at the switch point (see, for example, Kmenta 1986).

4. Nicaragua's rpm series comes close to rejecting the null at the 5% test level. However, as this is not a particularly strong result and as the consequences of treating a stationary series as if it were non-stationary are far milder than those associated with the alternative, rpm will be treated as a non-stationary series.

5. Note that the proposed cointegrating relationship excludes fxc on both statistical and theoretical grounds – i.e. it is stationary and changes in fxc are expected to have only a transitory effect on imports (see Patterson 2000).

6. A test of the null hypothesis that α₃ + α₁₀ = 0 for El Salvador generates a t-ratio of 0.69.

7. A test of the null hypothesis that α₁ + α₆ = 0 for Guatemala generates a t-ratio of 0.40.

8. Between 1981 and 1987, yr fell by less than 3%, rpm was unchanged and fxc rose by 7%.