What do panel data say on inequality and GDP? New evidence at US state-level
Introduction
The empirical literature concerning the effects of per capita GDP on inequality (Gini) is remarkable. The results are mixed, depending on the functional form and the methodology used, so that no unambiguous conclusions have been drawn upon the relationship. In particular, this seems to be true for those analyses related to the US, which have shown the existence of U-shaped, inverted U-shaped, S-shaped and N-shaped relationship between Gini and per capita GDP (for a brief survey, see Table A1 in online Appendix).
The aim of this paper is to revisit the empirical relationship between inequality and per capita GDP for the US using panel data for 50 states over the period 1960–2015. The advantages of using panel data at state level have been emphasized by Kim et al. (2011). This paper not only applies panel data at state-level, but it also conducts a preliminary inspection of the data using a nonparametric approach to determine the functional form of the relationship between inequality and GDP.
The paper attempts to contribute to the literature on the US in some respects. First, the paper uses new recently released data for inequality. Second, the relationship between Gini and per capita GDP is estimated using a recently developed penalized principal component (PPC) estimator by Li et al. (2016), which accounts for cross-sectional dependence and multiple structural breaks.
The empirical findings show the existence of a stable S-shaped relationship between Gini and per capita GDP when using the PPC estimator.
The rest of the paper is organized as follows. Section 2 illustrates the estimator by Li et al. (2016). Section 3 describes data and discusses empirical results. Section 4 concludes.
Section snippets
Li et al. (2016)’s PPC estimation method
Li et al. (2016) consider the following panel data model with interactive fixed effects where is a vector of explanatory variables, is a vector of unknowns coefficients that may vary over time, and are vectors of unobservable factor loadings and common factors, respectively, which can be correlated with . denotes the idiosyncratic error term.1
Data and empirical results
Data on real per capita GDP (), expenditure on health care () and welfare () are taken from http://www.usgovernmentspending.com/download_multi_year. Gini index data (Gini) are from Frank (2009) (updated from the author up to 2015) available at http://www.shsu.edu/eco_mwf/inequality.html. In the spirit of Mushinski (2001) and Duncan (2016), we use a nonparametric estimation technique to determine the functional form of the empirical relationship between inequality (Gini) and real per
Conclusions
This paper revisits the empirical relationship between inequality and real per capita GDP for the US using panel data for 50 states over the period 1960–2015. A preliminary nonparametric inspection of the data and the usage of a novel panel estimator with interactive fixed effects and multiple structural breaks seem to suggest a S-shaped curve for the empirical relationship, which may be interpreted as an extension of a U-shaped relationship. Further research may benefit from the empirical
References (12)
- et al.
Refitting the Kuznets curve
Econom. Lett.
(2012) Does the US current account show a symmetric behavior over the business cycle?
Int. Rev. Econ. Finance
(2016)- et al.
Economic growth and income inequality in the Asia-Pacific region: A comparative study of China, Japan, South Korea, and the United States
J. Asian Econ.
(2017) - et al.
Determining the number of factors in approximate factor models
Econometrica
(2002) - et al.
Series estimation of partially linear panel data models with fixed effects
Ann. Econ. Finance
(2002) - et al.
Revisiting the Relationship Between Welfare Spending and Income Inequality in OECD Countries. Tech. Rep. 72020
(2016)