Trade openness and child health: a heterogeneous panel cointegration analysis

ABSTRACT

In this paper, we examine the impact of trade openness on child health, based on a cross-country panel data set covering 66 countries for the period 1960–2013. To account for the time-series properties of the data and potential cross-country heterogeneity in the impact of trade openness, the study employs heterogeneous panel cointegration techniques that are robust to omitted variables and endogeneity problems. The results reveal that trade openness and child health are cointegrated, and that trade works to reduce the child mortality rate significantly in the long-run. The results are robust to the methodology and trade openness and child health indicators employed, as well as to the presence of cross-sectional dependence and changes in the sample composition. The findings also suggest that the impact of trade on child health tends to be stronger in countries with better institutional quality, lower corruption, good governance, political stability, and sound policies that promote private sector development.

KEYWORDS:

• Trade openness
• child health
• cross-country heterogeneity
• panel cointegration

JEL CLASSIFICATION:

• C23
• F02
• P45
• I12

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

¹ Note that our discussion of potential links is only exemplarily and is not meant to be comprehensive. For a complete conceptualization and discussion of the potential linkages between trade and food security, see FAO (2000, 2003).

² See, for example, cross-country studies of Edwards (1998), and Wacziarg and Welch (2008).

³ Data and further descriptions of the estimation method for child mortality can be found at http://www.childmortality.org.

⁴ Among others, see: (Dollar and Kraay 2004; Chang, Kaltani, and Loayza 2009). Nevertheless, in the robustness section of the paper, we assess the sensitivity of the results with regard to alternative trade openness measures.

⁵ For summary statistics of the variables employed, see Table A1 in the appendix.

⁶ See Pesaran and Smith (1995) for a detailed description of this estimator. We use the MG estimator as it is specifically designed for non-stationary variables and allows for heterogeneous slope coefficients across group members, such that it is particularly useful for estimating our equation of interest. Note, however, that applying the CD test to the residuals from a FE model (see e.g. De Hoyos and Sarafidis 2006), which does not allow for heterogeneous slope coefficients, led to the same conclusions.

⁷ For more details on the bootstrap method see Westerlund (2007).

⁸ To confirm the result of cointegration between trade and child health for the panel countries, we additionally test for cointegration between the trade-to-GDP ratio and the under-five mortality rate using the residual-based heterogeneous panel cointegration tests advanced by Pedroni (1999, 2004). These additional results can be found in Table A2 in the appendix.

⁹ See Pedroni (2001) for a more detailed representation of the group-mean method to calculate the averaged coefficients and associated t-statistics.

¹⁰ For a more detailed representation of the globalization index and its different components, see Dreher, Gaston, and Martens (2008).

¹¹ Note that we tested all variables for unit roots and cointegration before applying the PDOLS estimator, confirming that all series are integrated of order 1, I(1), and cointegrated.

¹² The data and a detailed description of the methodology underlying the construction of each of the governance indicators is available at www.govindicators.org.

¹³ Please note that the conclusions remain valid when employing other democracy indicators that are available for a long period of time, e.g. the democracy measure drawn from the POLITY project (see Marshall, Gurr, and Jaggers 2016).