Political budget cycles in military expenditures: A meta-analysis

Abstract

In this study, a meta-regression analysis is performed of studies that have examined whether military expenditures are affected by election cycles. Theoretically, the direction of these cycles is not immediately clear. This ambiguity stems mainly from the trade-off governments face when elections are upcoming. On the one hand, the incumbent government may try to improve the odds of being re-elected by boosting the performance of the national economy. One way is by increasing the procurement from the domestic defense industry using additional defense spending. On the other hand, it might be more favorable for the ruling cabinet to cut defense spending in an election year to finance expansions in other public spending categories that are preferred more by voters. The empirical studies considered in this meta-analysis are very different in nature concerning the election measure constructed, the sample of countries used, the time periods covered, model specification, estimation method, and publication outlet. Based on more than two hundred estimates, the main results indicate that there is only a weak negative genuine effect of elections on military expenditures. This finding suggests that governments follow a more contractionary policy in defense spending when elections are approaching. However, it also appears that this election effect substantially differs across country samples and relies to a great extent on certain country-specific characteristics. In particular, it turns out that the relationship between elections and military spending is likely to be positive in countries where the defense industry has some considerable political influence and economic power. Also, in countries that have to deal with serious security risks, this positive relationship is likely to prevail. Finally, I do not find any evidence that a publication bias inflates the election effect found among the considered studies. This is partly explained by the fact that both positive, as well as, negative results are theoretically plausible and, therefore, apparently equally likely to be published.