Defense–Growth Causality: Considerations of Regime-Switching and Time- and Country-Varying Effects

Abstract

This paper develops a panel smooth transition vector autoregressive model to investigate the economic growth–defense causality. This model simultaneously resolves the estimation problems of endogeneity, heterogeneity, and nonlinearity. Empirical results support that the causality is bidirectional, nonlinear, time- and country-varying. Economic growth has a negative impact on military spending and vice versa. The larger the HDI, the smaller the negative causality. Evidently, the increase in the level of country development can reduce the negative impact of military outlays on economic growth. Reducing the ratio of military spending to GDP is beneficial for countries with low HDI scores; however, moderately increasing the share of military expenditure is favorable for countries with extremely high HDI scores. Policy authority needs to set optimal education, health, and economic development shares of GDP for purchasing a maximum economic growth rate.

Keywords:

• Regime switching
• Time- and country-varying causality
• Panel smooth transition vector autoregressive model (PST-VARM)
• Human development index

JEL Classification:

• C23
• O40
• H56
• O15

Notes

¹ Ali and Dimitraki (2014) find empirical evidence that the relationship between military spending changes and economic growth is state-dependent.

² In fact, these variables are crucial factors that influence the economic growth of a country. In addition, in the linear VECM or panel VECM, these lagged variables have been responded in the lagged income (or lagged economic growth). Hence, in evaluating the defense–growth link, this study does not consider these variables.

³ In empirical study, economic freedom is another candidate transition variable. However, several previous studies have shown that there exists a high nexus between economic freedom and HDI. For example, in the work of Chodak and Kowal (2011), they find evidence that there exists a statistically significant relationship between economic freedom and HDI. Georgiou (2015) also finds a bidirectional causality between economic freedom and HDI. More importantly, the correlation coefficient between these two variables in the sample period of this paper is 0.81. Thus, the similar result can be expected when we replace HDI with economic freedom as the transition variable.

⁴ Table A1 in Appendix 1 provides the estimation results by replacing HDI with economic freedom index (EFI) as the transition variable. The characteristics of the estimated defense-–growth causality in these two PST-VARMs are remarkably similar. For example, both verify that there is a bidirectional, nonlinear, negative, and time-varying causality between defense and growth. In addition, the larger the EFI, the smaller the negative causality. However, Table A1 shows that only one-fourth of the estimated coefficients (=4/16) associated with the causality are statistically significant, and has higher values of AIC and BIC than those in Table 6.

⁵ We appreciate this helpful suggestion by one of the reviews.

⁶ Another interesting finding can be observed from the volatility of HDI scores. During the entire sample period, the countries with the top three largest volatilities in HDI scores are Cambodia, China, and Turkey. In 2000, their HDI scores were 0.466, 0.591, and 0.653, and in 2013, their HDI scores were 0.584, 0.719, and 0.759, respectively. The rapid development of China and Turkey has driven more different policies for enhancing economic growth.