Defense spending and economic growth: evidence from China, 1952–2012

Abstract

This paper examines whether defense expenditures contributed to economic growth in China for the 1952–2012 period. We examine the contribution of defense to economic growth using recently published official data on economic activity, defense, and government expenditures. We employ the Feder-Ram and augmented Solow models of economic growth to explore the defense-growth relationship. The Feder-Ram model appears to poorly explain economic growth in China. The augmented Solow model suggests, however, that a 1% increase in defense expenditures raises the economic growth rate by approximately 0.15–0.19%.

Keywords:

• China
• Defense expenditures
• Economic growth
• Defense-growth relationship

Acknowledgments

The views expressed in this article are the opinions of the authors and do not reflect the official views of the Naval Postgraduate School, Department of Defense, or the United States Government. We are grateful for comments and perspective from David Epstein, Andrew Marshall, and George Shambaugh. The authors are grateful to the Potomac Foundation for their support in this research. The usual disclaimers apply.

Notes

¹ The first translated version of these data (that we are aware of) was presented in Karber and Cross (2011).

² Anti-Access/Area Denial includes the development of new anti-ship missiles (Anti-Access) and the use of soft power to prohibit access to an area (Area Denial) through basing agreements, forward bases, and the denial of logistical support.

³ There are other perspectives too such as those developed by Deger (1986) and Barro and Sal-i-Martin (Barro 1990; Barro and Sala-i-Martin 1992); but we focus on the most commonly used below.

⁴ Both perspectives originate in neoclassical economics and each have their own build in strengths and weaknesses. For the purposes of comparing the models’ applicability to our data, the limitations of the models may be less relevant; but for any applicability to understanding the empirical dynamics of the defense-growth relationship, one should take those limitations into account. They include: specifications of which and how many sectors to include in the growth equation in the Feder-Ram model (Ram 1986, 1995) The augmented Solow model, while providing a more realistic model of economic growth (Heo 2010), still holds build in (often) equilibrating assumptions about e.g. the influence of defense spending on national factor productivity and does not for instance account for a possible influence of technology on growth which most modern macro approaches have failed to incorporate (Solow 2008). Thus, a future possible avenue for scholars interested in the topic would be to follow the leads of evolutionary theorists interested in growth (Dosi, Fagiolo, and Roventini 2010; Nelson 1998).

⁵ Employing a Vector Autogressive model, we fail to reject the null hypothesis that growth in real government expenditures does not Granger cause growth in real GDP. Similarly, we fail to reject the null hypothesis that the growth in real military expenditures does not Granger cause growth in real GDP. The χ² test statistic for real government expenditures is 2.43 with 2 degrees of freedom. For the growth in real military expenditures, the χ² test statistic is equal to 3.21 with 2 degrees of freedom. These results suggest that neither of the variables of interest Granger cause GDP growth. While not a test for strict exogeneity, it does suggest that we can proceed with our analysis.

⁶ The Durbin-Watson test statistic is 1.90. For the Breusch-Godfrey test with 3 lags, the χ² test statistics is 5.90 with 4 degrees of freedom. For the ARCH Lagrange Multiplier (LM) test, the χ² test statistics are 0.003, 0.195, 4.243, and 6.673 for the 1, 2, 3, and 4th lags, respectively. We fail to reject the null of no serial correlation with each test. On the other hand, the χ² test statistic for the Breusch-Pagan test of the null hypothesis of homoscedasticity is 6.31 with 1 degree of freedom, rejecting the null at the 1% level of significance.

⁷ We also fail to reject the null hypothesis of a unit-root with the Phillips-Perron test. We do, however, reject the null with the inclusion of a quadratic term for the Dickey-Fuller test. The literature notes that the presence of a quadratic trend in levels results in a stationarity series with a trend in first differences (cite Baltagi 2008).

⁸ For the respecified model, The Durbin-Watson statistic with 53 degrees of freedom is 1.68. The Bruesch-Godfrey test statistic with 3 lags is equal to 3.32. The ARCH LM statistic is 0.976 with 2 degrees of freedom. In all three cases, we fail to reject the null of no serial correlation. We reject the null hypothesis of homoscedasticity at the 5% level of significance with the Breusch-Pagan test of 4.94 and 1 degree of freedom.

⁹ For the Feder-Ram and Solow models, we fail to reject the null hypothesis of no serial correlation and reject the null hypothesis of homoscedasticity. Test statistics are available upon request.

¹⁰ He also notes that as there are both honorable and less honorable manifestations of a society build on relationships, ‘the test of maturity’ of China as a modern society will be ‘the ability of the Chinese people to arrive at a shared understanding as to what kinds of guanxi behavior should be seen as honorable, decent qualities of human relations, and what practices of guanxi should remain shameful and dishonorable’ (Pye 1995, 45). Regardless of possible ethical issues, a better understanding of real organizing behavior in China – honorable or not – is part of understanding the reporting and organization of data as well.