Abstract
There is considerable disagreement in the literature as to whether the unilateral and nonreciprocal preferences have succeeded in promoting developing countries’ exports. One strand of the literature suggests that multilateral removal of tariffs is more effective than unilateral trade preferences. For example, Francois et al (2006) find that developing countries underuse preferences to an average of 4 percent of the goods traded due to burdens directly associated with administration of preferences. Unilateral preferences are criticised for limited product coverage, stringent eligibility rules and selectiveness of inclusion and exclusion of countries. The World Trade Organisation (WTO) offers no formal redress mechanism for ironing out disagreements on country eligibility and product coverage for trade preference agreements. The preference-giving countries and potential beneficiaries have such obligations at own levels. Another strand of the literature suggests that the preferences are good for developing countries, arguing that multilateral removal of tariffs without discrimination erodes the gains for preferences receiving countries. Whereas there is no disagreement that increased openness is better for countries; variants of the literature for example Ng and Yeats (1996) suggest that the underperformance can be attributable to developing countries’ failure to effectively seize the opportunities granted under the preferences. The preferences are often initiated by the preference-giving countries with a view to realizing better outcomes for the beneficiary countries. However, empirical evidence on the outcomes achieved is limited.
This study evaluates the effectiveness of the AGOA (AGOA), a case of a developed country’s (USA) unilateral and nonreciprocal trade policy preference towards developing (Sub-Saharan African) countries to shed light on the issue. The empirical results show that the policy intervention had a dismal effect on U.S. real imports from beneficiary countries. The estimated average marginal effect of the AGOA intervention on U.S. imports of textiles is found to be between 1.7 percent and 1.9percent which are between 22 to 25 times smaller than a typical estimate of 42 percent found by Frazer and Biesebroeck (2010). The results suggest that the gains from the duty-free preferences could be much smaller than the previous literature has shown.
In the current study, weighted least squares coarsened exact matching and propensity score matching estimation techniques are applied on U.S. real imports of 8-digit products from the AGOA treatment-group and the comparable control-group during the 1997-2008 periods. The average treatment marginal effect estimate of 1.7 percent for U.S. imports of apparel products corresponds to the difference in the predicted U.S. imports of apparel of 10.5 percent from the AGOA treatment group and 8.8 percent of the same from the comparable matched control group with propensity scores as matching weights. The estimate suggests that the U.S. imports of apparel from the AGOA treatment group were 1.7 percent higher on average than it could have been if the AGOA intervention was never implemented.
Similarly, the 1.9 percent corresponds to the difference in the predicted U.S. imports of apparel of 7.1 percent from the treatment group and 5.2 percent of the same from the comparable matched control group with the coarsened exact matching weights. This estimate also suggests that U.S. imports of apparel were about 2 percent higher on average than could have been the case without the intervention. The average treatment marginal effects of 1.7 percent and 1.9 percent are statistically significant but are quantitatively a small positive effect, much smaller than the previous empirical literature has shown.
The above results can be interpreted in US$ real imports values. With the propensity scores matching as weights, the predicted U.S. imports response of apparel from the AGOA treatment group and from the comparable matched control group is estimated to be US$ 458,000 and US$ 359,000 on average. These predictive margins suggest that U.S. imports of apparel from the AGOA treatment group were on average US$ 99,000 higher than if the AGOA intervention was never implemented measured at the year 2000 constant prices. This marginal effect mirrors the 1.7 percent apparel marginal effect.
Similarly, based on the coarsened exact matching weights, the predictive margins corresponding to the 1.9percent average marginal effect are estimated. Based on the exact matching weights, the results show that U.S. real imports from the treatment and control groups are predicted to be US$ 312,000 and US$ 211,000 respectively. This suggests that U.S. imports of apparel from the treatment group were on average US$ 101,000 higher than if the AGOA intervention was never present, measured at constant 2000 prices. Therefore apparel marginal effects of 1.7 percent and 1.9 percent mirrors the estimated average marginal effects of US$ 99,000 and US$ 101,000 on average. The two matching approaches yield similar results.
It is observed that the U.S. real imports average responses from the treatment and control group of US$ 312,000 and US$ 211,000 based on exact matching are lower than the US$458,000 and US$ 359,000 from the same groups based the propensity scores weighting. These differences are expected because the sample size differs. The propensity scores sample has 40 African countries and the coarsened exact matched sample has 51 African countries. But the average marginal effects are unaffected by the sample size differences. Even though the sample size differs by 11 countries, the estimated marginal effects are close: the US$ 99,000 from CEM sample with 51 countries and US$ 101,000 from the propensity score sample with 40 countries differs only slightly. The GMM estimation with the dependent variable in levels of US$ and besides the groups being matched controls for two factors that are unrelated to the policy intervention but which affect both groups’ outcomes in a similar way. These are U.S. imports history by including lagged dependent variable and the U.S. market size change over time, by including a time effect. By controlling for these factors, the estimated average marginal effect of U.S. of apparel imports from the treatment group become insignificant. This suggests after controlling for time effects and U.S. import history, the effect AGOA intervention on imports of apparel disappears. These results suggest that the empirical evidence linking the AGOA intervention to increased U.S. imports of apparel dissipates and is therefore weak.
The AGOA, 2000 is part of the U.S. Trade and Development Act of 2000. The policy seeks to promote increased Sub-Saharan African country exports to the USA, by allowing for duty-free entry of eligible products for a period of 15 years. Using disaggregated 8- digit Harmonised System panel data on U.S. imports from the policy beneficiaries and the excluded African countries before and after the intervention, the effectiveness of the policy has been estimated.
The sample used has 51 countries consisting of 41 AGOA-eligible and 10 AGOAineligible with multivariate matched characteristics. The policy beneficiary countries served as the ‘treatment’, the excluded matched group served as the ‘control’. The period 1997-2000 is the ‘before’, and 2001-2008 is the ‘after’, intervention periods. Differences between the treatment and control groups are addressed by exact matching and estimated propensity scores to control for bias in the estimation of the treatment effect. The study applied dynamic panel estimation and weighted least squares estimation on the matched data. The dynamic panel data estimation applies the differences-in-differences as an alternative identification strategy to the matched data. The results are comparable. Overall, the results suggest smaller effects than the existing literature has shown.