Meta-analysis, military expenditures and growth

Abstract

This article conducts a meta-analysis of the effect of military expenditures on growth within a structured analytic framework. We extend the pioneering study of Aynur Alptekin and Paul Levine, by using a much larger sample of studies. Like them we confine our attention to studies that use the share of military expenditure in GDP, the military burden, as the independent variable, but unlike them we include not just those that use the military burden directly, what we call the core sample, but also those that use other functions of it, such as logarithms, differences, etc., which we call the remaining sample. We also consider an overall sample which pools all results. The t-statistic on the coefficient of military burden is used as the dependent variable. Our null hypothesis is that military expenditure has no significant effect on growth and we explain why this is plausible. The estimates are sensitive to the sample and type of data used, estimation method adopted, and the controls included. Overall, the results are consistent with the hypothesis of no effect: the average effect across all studies is close to zero. Certain study characteristics appear significant determinants of the effect of military expenditure on growth, but there does not appear to be a simple pattern and different characteristics were significant in the three samples. This might be a result of data mining to produce a significant result. However, there does not appear to be strong evidence of publication bias towards positive or negative results, perhaps because there is no strong a priori belief in the direction of the effect.