Forecasting military expenditure:

Research & Politics

Tobias Böhmelt¹ 

关键词: Bayesian model averaging;    forecasting;    in-sample prediction;    military spending;    out-of-sample prediction;   

DOI  :  10.1177/2053168014535909

学科分类：社会科学、人文和艺术（综合）

来源: Sage Journals

【 摘 要 】

To what extent do frequently cited determinants of military spending allow us to predict and forecast future levels of expenditure? The authors draw on the data and specifications of a recent model on military expenditure and assess the predictive power of its variables using in-sample predictions, out-of-sample forecasts and Bayesian model averaging. To this end, this paper provides guidelines for prediction exercises in general using these three techniques. More substantially, however, the findings emphasize that previous levels of military spending as well as a country’s institutional and economic characteristics particularly improve our ability to predict future levels of investment in the military. Variables pertaining to the international security environment also matter, but seem less important. In addition, the results highlight that the updated model, which drops weak predictors, is not only more parsimonious, but also slightly more accurate than the original specification.

【 授权许可】

CC BY-NC-ND