Redistributive efficiency of fiscal policy: The role of decentralization and good governance

ABSTRACT

In this paper, we empirically analyse the redistributive efficiency of fiscal policy considering its degree of decentralization for a sample of thirty-five countries over the 2000–2016 period. To do this, we follow a two-stage process where we first employ a bootstrap Data Envelopment Analysis (DEA) method to estimate the redistributive efficiency of cash transfers and direct taxes at the central and subnational levels. We obtain evidence that redistributive efficiency has diminished after the Great Recession (2007–2008), and is lower in the Southern European countries, and higher in almost the rest of Europe and the Latin American countries. Second, a bootstrap truncated regression analysis is used to identify the contextual factors that might explain redistributive efficiency variation across countries and time, supporting the argument that good governance is a requirement for decentralization to be efficient in promoting policy outcomes.

KEYWORDS:

• Redistributive efficiency
• fiscal policy
• decentralization
• governance
• DEA
• truncated regression analysis

JEL CODES:

• C14
• C34
• H30
• H50
• H51

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes

1 See, for example, Piketty and Saez (2006), Piketty (2014), and Alvaredo et al. (2018).

2 The increase in income inequality can reduce economic growth by reducing the investment in human capital (Galor and Zeira 1993), increasing the fertility rate of poor people (de la Croix and Doepke 2003), engaging socio-political instability which discourages private investment (Alesina and Perotti 1996) and increasing ethnic tensions and social polarization that reduce the security of property and contract rights (Keefer and Knack 2002).

3 This is found for instance by the following authors: Gupta and Verhoeven (2001) for 37 countries in Africa from 1894 to 1995; Afonso and Aubyn (2005) in OECD countries for 2000; Afonso and Aubyn (2011) in OECD countries for 2000–2003 period; Hauner and Kyobe (2010) for a large panel of economies from 1980 to 2004; Adam, Delis, and Kammas (2014) for 21 OECD countries during 1970–2000.

4 See Solt (2020) for a complete description of the SWIID; we employ version 8.3 (update in May 2020).

5 This measure is extensively used by several authors (Korpi and Palme 1998; Bradley et al. 2003; Mahler and Jesuit 2006; Huber and Stephens 2014; Kyriacou, Muinelo-Gallo, and Roca-Sagalés 2018).

6 See Table A3 in the Appendix for descriptive statistics of input-output variables.

7 The definitions and sources of all variables are presented in Table A2 and descriptive statistics in Table A4, both in the Appendix.

8 After analysing 25 decentralization indexes, Harguindéguy, Cole, and Pasquier (2021) conclude that the RAI indexes are the ones of the highest quality. Unfortunately, the TERRGO indexes from Trinn and Schulte (2020) were not included in their analysis.

9 To obtain the DEA inefficiency score, we use the ‘Benchmarking’ package (function DEA.boot) in R software and the bootstrap truncated regression analysis was performed in STATA. All codes are available from the authors upon request.

10 Despite the fact that in our context we may find multiple governments having fiscal responsibilities simultaneously (at central, regional, and local levels), we also check for the influence of the ‘party orientation’ concerning economic policy using a variable from the Database of Political Institutions (DPI) that classifies the chief-executive’s party from right to centre to left (Cruz, Keefer, and Scartascini 2018). Moreover, we control for the influence of ‘political polarization’ through a DPI variable that measures the maximum polarization between the executive party and the four principal parties of the legislature. We do not find any of these variables to be associated with redistributive efficiency, nor does their inclusion alter our findings (results not reported but available upon request).