Abstract
This paper empirically re‐examines the long‐run co‐movements and the causal relationships between GDP and defence expenditures in a multivariate model with real defence expenditure per capita (ME), real GDP per capita (GDP), and real capital stock per capita (K). We apply the view of the aggregate production function to construct the empirical model. Using up‐to‐date data for 27 OECD countries and 62 non‐OECD countries for the 1988–2003 period, we combine cross‐sectional and time series data to re‐investigate the relationship between GDP and ME. Previous studies using time series data may have yielded misleading results on account of the short time span of typical datasets. By contrast, we use recently developed panel unit root tests and heterogeneous panel cointegration tests, and conclude that there is fairly strong evidence in favour of the hypothesis of a long‐run equilibrium relationship between GDP and ME. The long‐run panel regression parameter results, such as the fully modified OLS, indicate that a positive relationship between GDP and ME only holds for OECD countries, whereas a negative relationship from ME to GDP only exists in non‐OECD countries under examination and in the panel as a whole. Furthermore, by implementing the dynamic panel‐based error correction model, we determine that GDP and ME lack short‐run causalities, but do show long‐run bidirectional causalities in both OECD and non‐OECD countries.
ACKNOWLEDGEMENTS
We would like to thank the reviewers and editor for the useful comments and suggestions. All remaining errors are our own.
Notes
1According to Shieh et al. (2002), the positive relationship between defence expenditure and economic growth is now often called the Benoit Hypothesis. Ever since the publication of Benoit’s work, many empirical studies have subsequently focused on addressing the validity of Benoit’s assertions.
2Expansion of the time horizon might cause the regime‐shift problem.
3Many researchers suggest that panel‐based tests have higher power than tests based on individual series.
4Harris and Tzavalis (Citation1999) determine that these panel tests allow for both parameter and dynamic heterogeneity across groups and that they are considerably more powerful than conventional tests. We then use the cointegration tests for a panel of countries instead of following a time‐series or traditional panel data approach.
5Panel cointegration techniques provide new evidence in many areas of economic research such as economic growth, international finance, and macroeconomics issue (e.g. Kao et al. Citation1999 and Pedroni, Citation2004). The objective of this paper is to determine if these powerful tools are applicable to defence economic research involving pooled cross‐section and time‐series datasets.
6The null of the Hadri (Citation2000) test examines for I(0), while the null of the remaining two tests do so for I(1).