Cross-country evidence on the distributional impact of fiscal policy

ABSTRACT

This article provides new evidence on the distributional effects of fiscal policy using data on a panel of OECD economies over the last four decades. We study how four measures of income inequality and poverty respond to several stock and flow variables accounting for fiscal actions. We find that increases in government debt and expenditure promote a less unequal distribution of income. We detect a significant distributional impact of education and social spending as well as of government consumption expenditure. We also investigate potential redistributive implications of large fiscal expansion and consolidation episodes finding no evidence of additional effects beyond those associated with conventional fiscal variables.

Abbreviations: OECD: Organisation for Economic Co-operation and Development; GDP: Gross Domestic Product; G20: Group of 20 economies (forum of 19 dvanced and emerging countries plus the European Union); CGE: Computational General Equilibrium models; DSGE: Dynamic Stochastic General Equilibrium models; UN-WIDER: United Nations World Institute for Development Economics Research; SWIID: Standardized World Income Inequality Database; WDI: World Development Indicators; PPP: Purchasing-Power Parity; LIS: Luxembourg Income Database; GMM: Generalized Method of Moments; FE: Fixed Effects; RE: Random Effects; SE: Standard Errors; CPI: Consumer Price Index

KEYWORDS:

• Fiscal policy
• income inequality
• poverty
• distributional effects

JEL CLASSIFICATION:

• E62
• D63
• D31

Acknowledgments

We wish to thank Salvatore Morelli for kindly sharing the inequality data he assembled with Tony Atkinson (we used those data on a previous version of this article before they eventually became publicly available). The usual disclaimer applies.

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

¹ Wealth tends to be more concentrated than income. In this article we only deal with the distribution of the latter.

² On the other hand investigating the determinants of gross-income distribution involves dealing with factors such as technological change and globalization as well as structural policies such as education and labour and product market regulation (Koske Fournier and Wanner 2012). Government spending and taxes do not appear in those analyses although it has been suggested that labour demand may be affected by the tax wedge in non-perfectly competitive markets (Bassanini and Duval 2006).

³ On the other hand Duncan and Sabirianova Peter (2008) use data for more than 100 countries to conclude that progressivity only reduces observed inequality in reported gross and net income but has a smaller impact on what they label as ‘true’ inequality approximated by consumption-based measures of Gini. Their policy implications call for the adoption of flat tax policies.

⁴ Some authors have instead focused on the consequences for economic growth (Alesina and Ardagna 2010; International Monetary Fund 2010; Perotti 2012).

⁵ This literature is predominantly empirical with only limited theoretical attempts to model the distributional consequences of consolidation (see e.g. Jensen and Rutheford 2002).

⁶ Other measures of income inequality used in the literature are for instance the shares of income of the top 10% to the bottom 10% but as those measures are predominantly based on gross income we do not employ them in our analysis.

⁷ Note that we use WDI data and not the Penn World Tables (Heston Summers and Aten 2012). The latter is a frequently used source for this measure which we avoid given the shortcoming properly noted by Knowles (2001) namely the fact that the latter source calculates the series by imposing a set of PPPs prices that may yield imprecise values.

⁸ The SWIID data are based on the Luxembourg Income Study (LIS) dataset.

⁹ On the theoretical side one of the fundamental contributions on the relationship between redistribution and inequality is that of Meltzer and Richard (1981) whose model predicts that when mean income rises relative to the median income the lower income agents will form a coalition supporting higher taxes presumably direct taxes as they are more progressive than indirect taxes. More recent contributions by Benabou (2000) and Kerr (2014) conclude instead that higher income inequality is associated with less redistributive government spending. This latter result stems from the existence of capital market imperfections implying different investment opportunities among individuals leading to income inequalities that persist over time. Therefore according to those contributions fiscal policy responds to income distribution rather than vice versa.

¹⁰ The fact that the p-value of the latter is in all cases equal to 1.00 may suggest over-fitting although we see this as a result of the dimensions of the sample which are different from the optimal ones required for the GMM estimator.

¹¹ We only report the FE estimates for the sake of brevity but the RE estimates obtained with 5-year periods’ data (not reported but available upon request) also support findings of the annual estimates.

¹² Similar estimates with poverty_AM as the dependent variable lead to the same conclusions.