Cross-country evidence on determinants of fiscal policy effectiveness: the role of trade and capital flows

ABSTRACT

This article studies the determinants of size differentials between fiscal multipliers in countries around the world, both advanced and developing economies. We introduce variables not considered before for explaining multiplier size differentials, such as capital flows and the openness of capital markets, while controlling for domestic conditions and exchange rate regimes. We also disaggregate GDP into its main components in order to identify the channels through which external and internal factors can influence GDP after a change in fiscal policy. Our results point to the existence of a new channel through which fiscal policy effectiveness is affected. Capital flows, especially FDI flows, play an important role in determining the sizes of fiscal multipliers, and a country’s external conditions largely explain GDP changes after fiscal expenditure shocks. Our results also point towards a strong link between a country’s international position and its real economy.

KEYWORDS:

• Fiscal policy
• government spending multiplier
• fiscal transmission mechanism
• cross-country heterogeneity
• international linkage

JEL CLASSIFICATION:

• E62
• F41

Acknowledgements

The authors are grateful to Seiichi Fujita, Kenneth N. Kuttner, Shugo Yamamoto, participants of September 2014 Modern Monetary Economics Summer Institute at Kobe University, and two anonymous Referees for helpful comments and suggestions. Shibamoto acknowledges the financial support from a Grant-in-Aid from the Japanese Ministry of Education.

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

¹ An alternative approach bases on the narrative evidence and focuses on the changes in fiscal policy expenditure at the specific events not related directly with a country’s economic conditions. That is, it takes unanticipated changes in fiscal policy as a proxy of the fiscal expenditure shocks. For instance, Ramey and Shapiro (1998) use narrative evidence on military spending to construct fiscal shock variables. Ramey (2011a) compares results based on VAR shocks with the results achieved with shocks based on narrative evidence. While both approaches intend to control for economic situation in the process of fiscal shocks’ identification, the results might differ depending on the approach used, resulting in the continued discussion in the literature on the appropriate identification methodology.

² Ramey (2011b), while briefly reviewing both theoretical and empirical literature on fiscal policy effects, mentions mostly identification methodology and the effects of fiscal policy shocks.

³ We also extend the model to a four-variable version by adding the ratio of the current account balance to the GDP (CA_t) and real effective exchange rate (REER_t), then estimating the model with two lags for each variable. The results are available in the Appendix.

⁴ The figures of IRFs based on 2-variable VAR models are available in the Appendix.

⁵ The results change little, however, if no discounting takes place.

⁶ As fixed exchange rate regime countries we include countries with no separate legal tender, pre announced pegs, currency board arrangements, crawling pegs, or de facto or pre-announced bands or crawling bands with margins no larger than ±2%.

⁷ The Appendix reports the results of robustness checks: use of 2000 values of all explanatory variables as well as different periods of dummy variables.

⁸ Table A.IV.1 in Appendix contains the correlations of the main cross-country explanatory variables. Few of our explanatory variables show high correlations with each other. The biggest exception is the developed country dummy, which shows high correlations with most of the other variables. Yet the high correlations demonstrate the characteristics of the advanced economies quite well and thus pose no problems for the regressions. Specifically, developed countries have more open capital accounts and lower net inflows than developing economies, as the developed countries with high gross capital inflows are also major capital exporters. The high correlation coefficient with the monetary union dummy derives from the fact that all euro area members are classified as advanced economies. Developed countries are usually highly indebted relative to their GDPs which may derive from the typically higher credibility and ability to borrow more money on the market at lower cost. We can observe a similar relation for the monetary union countries. The high correlation coefficient between the MU dummy and exchange rate dummy can be explained by our treatment of euro area countries as countries with fixed exchange rate regimes.

Meanwhile, we find no high correlations between KAOPEN and net inflows, no high correlations between KAOPEN and the other variables, and no high correlations between net inflows and the other variables. One noteworthy finding is that lower capital account openness is associated with higher net capital inflows, and vice versa. This may point at the fact that capital account restrictions are not always symmetrical and might often restrict capital outflows more than capital inflows. We also see in our correlation matrix, reassuringly, that capital account openness and trade openness are almost wholly uncorrelated. The introduction of the KAOPEN index into the regression clearly brings in new information not included in the trade openness variable.

⁹ In the robustness section of the Appendix we show the results of the analysis for 4^th period cumulative multipliers.

¹⁰ Table A.IV.2 in the Appendix reports correlation coefficients for the gross capital flows. We first notice a very high correlation between gross capital inflows and gross capital outflows. Meanwhile, both of the new explanatory variables correlate weakly with the developed country dummy, which may be attributable to the tendency of developing countries to have higher gross flows as a ratio of GDP, partly explainable by their lower GDPs. The higher correlation with outflows versus inflows in the developed country dummy shows that these countries often export more capital than they import. Gross capital inflows and gross capital outflows both have low positive correlations with capital account openness. The higher openness stimulates capital flows in both ways, albeit not to a considerable degree.

¹¹ Correlations for the detailed capital flows are presented in Table A.IV.3 in the Appendix. Few of the capital flow variables show high correlations with the other explanatory variables. The biggest exception is the high negative correlation between net FDI flows and the developed country dummy. This, again, points to the basic characteristics connected to a country’s development level and implies that net FDI flows are lower in advanced economies and higher in developing ones. Additionally, we can observe high correlations between outflows and inflows of the same category.

We also check correlations for all types of inflows and outflows with each other. In doing so, we find high correlation coefficients between portfolio outflows and other outflows. The values are less pronounced for the rest of the categories, so we assume there are no problems with carrying out regression that include all categories of inflows and outflows at the same time.

¹² There are competing views regarding the efficacy of fiscal policy; an increase in government expenditure has a positive effect on GDP/private consumption (standard Keynesian effect), no effect (Ricardian equivalence), and a negative effect (non-Keynesian effect). Using a nonlinear fiscal policy model and a cross-country data of OECD countries, Perotti (1999) finds that the non-Keynesian effect occurred only when the government debt to GDP ratio was high. Blanchard (1990) and Sutherland (1997) have theoretically demonstrated this nonlinear effect and its dependence on the economic environment.

¹³ There is usually a lag between a decision on fiscal policy expenditure and its implementation. Most fiscal policy changes are therefore believed to be anticipated by the private sector and to influence the private sector’s behaviour even before the fiscal variables change. As a consequence, an analysis of the effects of fiscal policy expenditure based on the VAR methodology, i.e. an analysis focused only on the effects of the innovations of fiscal expenditure, might be inconsistent with the one of the effect of ‘unanticipated’ fiscal shocks. However, as Ilzetzki, Mendoza, and Végh (2013) argue, we believe that this problem is mitigated especially in developing countries.

Additional information

Funding

This work was supported by the Grant-in-Aid for Scientific Research from Japanese Ministry of Education, Culture, Sports, Science, and Technology [No 24243044 and No 15H05729].