Does trade liberalization reduce child mortality in low- and middle-income countries? A synthetic control analysis of 36 policy experiments, 1963-2005

Abstract

Scholars have long argued that trade liberalization leads to lower rates of child mortality in developing countries. Yet current scholarship precludes definitive conclusions about the magnitude and direction of this relationship. Here I analyze the impact of trade liberalization on child mortality in 36 low- and middle-income countries, 1963-2005, using the synthetic control method. I test the hypothesis that trade liberalization leads to lower rates of child mortality, examine whether this association varies between countries and over time, and explore the potentially modifying role of democratic politics, historical context, and geographic location on the magnitude and direction of this relationship. My analysis shows that, on average, trade liberalization had no impact on child mortality in low- and middle-income countries between 1963 and 2005 (Average effect (AE): -0.15%; 95% CI: -2.04%-2.18%). Yet the scale, direction and statistical significance of this association varied markedly, ranging from a ∼20% reduction in child mortality in Uruguay to a ∼20% increase in the Philippines compared with synthetic controls. Trade liberalization was also followed by the largest declines in child mortality in democracies (AE 10-years post reform (AE₁₀): -3.28%), in Latin America (AE₁₀: -4.15%) and in the 1970s (AE₁₀: -6.85%). My findings show that trade liberalization can create an opportunity for reducing rates of child mortality, but its effects cannot be guaranteed. Inclusive and pro-growth contextual factors appear to influence whether trade liberalization actually yields beneficial consequences in developing societies.

Keywords: Child mortality; Developing countries; Sustainable Development Goals; Synthetic control; Trade liberalization.

Fig. 1

Impact of trade liberalization on child mortality: estimates and 95% confidence intervals for post-reform average effect and 5- and 10-years post-reform effect. Notes: 95% confidence intervals are estimated by calculating the mean effect in 5000 placebo samples of 32 ‘fake’ liberalization experiments. Like the average effect estimates, the means of these placebo samples effects were weighted so that weights correspond to each model's prediction error. See Appendix 1 for further detail.

Fig. 2

Effect of trade liberalization on child mortality by democratic status. Notes: Figure shows normalized average effects where the difference between child mortality in treated countries and synthetic controls was first normalized so that the year of liberalization = 1. These estimates were then averaged using the same RMSPE-weighting procedure as my main analysis (see Appendix 1).

Fig. 3

Effect of trade liberalization on child mortality by region. Notes: Figure shows normalized average effects where the difference between child mortality in treated countries and synthetic controls was first normalized so that the year of liberalization = 1. These estimates were then averaged using the same RMSPE-weighting procedure as my main analysis (see Appendix 1).

Fig. 4

Effect of trade liberalization on child mortality by decade. Notes: Figure shows normalized average effects where the difference between child mortality in treated countries and synthetic controls was first normalized so that the year of liberalization = 1. These estimates were then averaged using the same RMSPE-weighting procedure as my main analysis (see Appendix 1).

Fig. 5

Intra-regional comparisons. Notes: The RMSPE Ratio is the ratio of the pre-intervention RMSPE in the model using the specified comparison group to the RMSPE in my original specification. Higher ratios (lighter blue to light grey) indicate better model fit. (For interpretation of the references to colour in this figure legend, the reader is referred to the Web version of this article.)