Elsevier

European Economic Review

Volume 101, January 2018, Pages 605-624
European Economic Review

Inequality, financial development and economic growth in the OECD, 1870–2011

https://doi.org/10.1016/j.euroecorev.2017.11.004Get rights and content

Abstract

Inequality's effect on growth remains elusive, largely due to endogeneity, complex interactions, and lead–lag relationships. We revisit this issue by examining the four main channels through which inequality transmits to growth: savings, investment, education, and knowledge production. We construct new panel data for 21 OECD countries spanning 142 years. External communist influence is used as a new time-varying instrument for inequality and the effects of inequality on the outcome variables are made conditional on the stage of financial development. Our results show that inequality hampers growth at low to moderate levels of financial development but has little effect on growth at advanced levels of financial development.

Introduction

One of the long-standing questions in macroeconomics is whether income inequality impairs or promotes growth. Despite a burgeoning literature on the growth effects of inequality, the empirical evidence remains inconclusive and a definitive assessment of the effects of inequality on growth remains elusive.1 This impasse is not surprising given that the relationship between inequality and growth is masked by various channels that transmit to growth with markedly varying lags, inequality may have growth as well as level effects, and the growth effects of inequality are likely to depend on moderating factors, such as the stage of financial development.

In this paper we extend previous research by examining school enrollment, ideas production, savings, and investment, as potential channels through which inequality impacts growth. These are the key drivers of economic growth in standard economic growth models. All four transmission channels relate to the savings–investment decision and are, henceforth, generically referred to as ‘savings’. We pay particular attention to the interaction between financial development and inequality, since credit is arguably the most important channel through which inequality affects growth (see Aghion and Bolton, 1997, Banerjee and Newman, 1993, Galor and Zeira, 1993, Piketty, 1997). Inequality is potentially harmful to growth in financially underdeveloped economies because investment in education, R&D, and fixed capital are often indivisible with high sunk costs, as discussed in detail below. Numerous empirical studies have found financial development to be a major driver of growth and development (e.g., Levine et al., 2000, Bordo and Meissner, 2015, Madsen and Ang, 2016), suggesting that financial development is likely to be a key channel which compounds the growth-effects of inequality. Moreover, our empirical estimates show that the coefficient on inequality is implausible unless the interaction between inequality and financial development is included in the estimated models.

This paper makes three contributions to the literature on growth and income inequality. First, the paper seeks to overcome the inherent problems of regressing growth directly on inequality by instead exploring the effects of inequality on the key drivers of growth (savings, investment, ideas production, and school enrollment at the primary, secondary and tertiary levels). This approach has several advantages over the reduced form approach where productivity is regressed on inequality: it provides a much deeper understanding of how inequality transmits to growth, it enables inequality to have growth or level effects depending on the specific channel, and it allows education to affect growth with a considerable time-lag that cannot easily be captured from reduced-form growth regressions. While some studies have investigated the investment and savings channels, there has been relatively little attention to the innovation and the education channels. Moreover, prior studies have largely ignored the moderating role of financial development. In contrast, we specifically examine the interaction between inequality and financial development to test the hypothesis that inequality is potentially more harmful for growth in financially underdeveloped economies than in economies with fully developed financial systems.

Our second contribution is to assemble a new macroeconomic panel for 21 OECD countries spanning 142 years from 1870 to 2011. These data have several salient benefits. OECD economies provide a more nuanced sample, diminishing the unobserved heterogeneity that would arise from larger country samples with a short time-dimension (Alesina et al., 2016). Since the available inequality data for the poorest countries often consist of only two or three data points, little is gained from a larger panel of countries and the associated point estimates are likely to be estimated with low precision. As Johnson et al. (2013, p. 273) note: “in general, annual data from non-OECD countries should be treated with caution”. Using the long historical data, we are able to trace transitions from significant financial underdevelopment in the earlier years through to the highly sophisticated financial systems that today characterize OECD countries. We are then able to identify the effects on growth through the interaction between inequality and financial development. Econometrically, the benefits of using long panel data are that the fixed effects estimator becomes more consistent as the sample grows, and the instrumental variable parameter estimates can be severely biased in small samples (Davidson and MacKinnon, 2006).

Our third contribution relates to the identification of the effect of inequality and its interaction with financial development on the outcome variables. Prior studies rarely use time-varying external instruments; instead they use OLS or internal instruments that exploit orthogonality conditions of the independent variables Forbes, 2000, Banerjee and Duflo, 2003. A common external instrument for inequality in cross-country regressions is the time-invariant ratio of land suitable for wheat and sugarcane production (Easterly, 2007). We here propose an identification strategy that uses the strength of communist influence in culturally similar countries as a new time-varying instrument for inequality. In addition, contract-intensive money is employed as an instrument for financial development in the robustness section.

Culturally weighted foreign communist influence is a strong predictor of income distribution as workers’ wage aspirations are more willingly accommodated by elites, governments, and employers, when there is an emerging or eminent communist threat. Revolutions often result from general discontent that is kept dormant until a significant event triggers an outburst that can potentially spread internationally and result in regime change, unless the ruling elites take voluntary pre-emptive measures (Weyland, 2010). The historical record contains numerous examples of waves of pro-labor movements across the globe inspired by events in culturally neighboring countries, as discussed in detail in Section 3 below.

Why would elites accept increasing taxation and real wages in periods of communist upheaval in cultural neighboring countries when this reduces their after-tax income? The answer is that they may be forced to accept income redistribution concessions because of emerging threats of communist revolution. Communist threat contributes to polarization and class conflict, e.g., communist influence grows in parts of the trade union movement and other organizations. Labor movements and capitalists alike observe the ease with which regimes are overturned. For example, when Tsar Nicholas’ II downfall in 1917 made other regimes suddenly look weak and vulnerable, labor movements across the industrializing countries gained confidence to act on these weaknesses. Furthermore, governments and elites considered a communist revolution to be a real threat and consequently redistributive reforms may be a lower price to pay in exchange for a peaceful labor movement. Weyland (2010), for example, contends that fear of Bolshevism induced pre-emptive suffrage reforms in Britain, Sweden, Germany, and Finland in the period 1917–1919. Similar preemptive measures have been identified in the democratization and expansion of franchise across the world by Acemoglu and Robinson (2000) and Aidt and Jensen (2014).

The approach taken in this study overcomes some of the problems faced by many previous studies in which per capita growth is regressed directly on measures of income inequality without paying sufficient attention to possible transmission channels, endogeneity, unobserved country heterogeneity, the interaction between inequality and financial development, and lead–lag relationships. The nexus between inequality and growth is particularly masked with regard to education. The influence of inequality on education takes several years to affect production because it can take up to two decades before students enter the labor force, and workers of retirement age, cohorts whose education was also affected by inequality when they acquired their education, exit the labor force. Thus, the effect of inequality on income through education is highly unlikely to be captured by standard growth regressions. Furthermore, if the savings-effect of inequality is facilitated by financial development, then unconditional savings regressions are unlikely to reveal much about the nexus between inequality and savings.

The results show that income inequality hampers growth through savings, investment, education, and ideas production at the levels of financial development that prevailed in the OECD countries approximately before 2000 and today for most countries in the world, particularly low income countries. Counterfactual simulations show that the declining inequality experienced over the period 1870–1965 in the 21 advanced OECD economies included in our sample resulted in a 90% increase in labor productivity.

The rest of the paper is organized as follows. Section 2 discusses the transmission channels and model specifications. Section 3 discusses in detail the identification strategy and exclusion restriction, while the data are discussed in Section 4. The empirical results and robustness checks are presented in Sections 5 and 6. Simulations of the historical evolution of elasticities conditional on financial development are carried out in Section 7 and the impact of inequality on growth is assessed in Section 8. The final section concludes and discusses the implications of our findings.

Section snippets

Transmission channels and model specifications

We focus on the accumulation channels – physical, human, and knowledge capital – that theory has identified as the key determinants of long term economic growth. These channels play a central role in the neoclassical growth model (e.g. Solow, 1956), extensions to this model (e.g. Mankiw et al., 1992), endogenous growth models (e.g. Romer, 1990), and Schumpeterian growth theory (Aghion and Howitt, 2006). For example, human capital accumulation plays a fundamental role in these theories. Human

Identification strategy

It is necessary to instrument inequality because there are several channels through which the outcome variables (broadly savings) affect inequality. For example, Krusell et al. (2000) argue that investment in machinery and equipment is a strong indicator of skill-biased technological progress and, consequently, these investments increase inequality. Patenting is likely to lead to increasing inequality because it is complementary to fixed capital stock. Savings also widens inequality, through

Data

The income inequality data used in this paper are the post-tax, post-transfer Gini coefficient (Gini) and pre-tax top 10% income shares (Top10) for 21 high income OECD countries over the period 1870–2011. The 21 OECD countries are: Australia, Austria, Belgium, Canada, Denmark, Finland, France, Germany, Greece, Ireland, Italy, Japan, Netherlands, New Zealand, Norway, Portugal, Spain, Sweden, Switzerland, the United Kingdom, and the United States. The data are backdated and interpolated using

Unconditional estimates

Unconditional IV regressions are presented in Table 2. Communist influence has the expected negative sign in the first-stage regressions and the F-tests for excluded instruments exceed 39, suggesting that the maximum bias in the IV estimators in these regressions is well below 5%. The coefficients of inequality and financial development are highly statistically significantly negative in all cases regardless of the measurement of inequality and outcome variables. Furthermore, the coefficients of

Instrumenting financial development

Thus far we have only instrumented inequality as our main focus variable. However, the coefficients of financial development and its interaction with inequality may be biased because of feedback effects from the dependent variables. In this section we address this problem by using contract-intensive money as an instrument for financial development. Contract-intensive-money is measured as (M2−M0)/M2, where M2 is broad money and M0 is high-powered money.

Contract-intensive money is suggested by

Time-variation of elasticities

In this section, we assess the economic significance of inequality on the four outcome variables by computing the outcome semi-elasticities of inequality. Average credit-income ratios are used to trace how the semi-elasticities have evolved since 1870. The semi-elasticities are computed as follows, where the coefficient estimates are taken from the conditional IV regressions in columns (5)–(8) in Table 3: ηts=sΦ=α^1+α^3Ψt=1.303+0.011ΨtηtI/Y=(I/Y)Φ=(β^1+β^3Ψt)(IY)¯=0.760+0.011ΨtηtGER=GERΦ

Growth and inequality: counterfactual analysis

In this section we seek to estimate the long-run labor productivity effects of changes in inequality conditional on a given level of financial development. We do this in two steps. First we regress labor productivity on knowledge stock, capital, and educational attainment. Thereafter, we derive the effects of inequality on labor productivity through each of the outcome variables. We do not regress labor productivity directly on inequality because, as argued in the introduction, inequality

Conclusions

The consequences of inequality on growth and the transmission channels through which inequality affects growth remain unclear. In this paper we have sought to gain greater clarity on these issues by empirically examining the extent to which income inequality transmits to growth through savings, investment, education, and knowledge production. We have compiled a panel data set for 21 OECD countries over the period 1870–2011. External communist influence is used as a new time-varying instrument

References (77)

  • R. Levine et al.

    Financial intermediation and growth: causality and causes

    J. Monet. Econ.

    (2000)
  • U. Panizza et al.

    Public debt and economic growth: is there a causal effect?

    J. Macroecon.

    (2014)
  • P.F. Peretto et al.

    Resources, innovation and growth in the global economy

    J. Monet. Econ.

    (2011)
  • F. Venturini

    Product variety, product quality, and evidence of endogenous growth

    Econ. Lett.

    (2012)
  • F. Venturini

    Looking into the black box of Schumpeterian growth theories: an empirical assessment of R&D races

    Eur. Econ. Rev.

    (2012)
  • D. Acemoglu et al.

    Why did the West extend the franchise? democracy, inequality, and growth in historical perspective

    Q. J. Econ.

    (2000)
  • P. Aghion et al.

    A theory of trickle-down growth and development

    Rev. Econ. Stud.

    (1997)
  • P. Aghion et al.

    Inequality and economic growth: the perspective of the new growth theories

    J. Econ. Lit.

    (1999)
  • P. Aghion et al.

    Endogenous Growth Theory

    (1998)
  • P. Aghion et al.

    Joseph Schumpeter lecture appropriate growth policy: a unifying framework

    J. Eur. Econ. Assoc.

    (2006)
  • A. Alesina et al.

    Ethnic inequality

    J. Polit. Econ.

    (2016)
  • A.B. Atkinson

    Bringing income distribution in from the cold

    Econ. J.

    (1997)
  • A.B. Atkinson et al.

    Top incomes in the long run of history

    J. Econ. Lit.

    (2011)
  • A.V. Banerjee et al.

    Inequality and growth: What can the data say?

    J. Econ. Growth

    (2003)
  • A.V. Banerjee et al.

    Occupational choice and the process of development

    J. Polit. Econ.

    (1993)
  • R.J. Barro

    The stock market and investment

    Rev. Financ. Stud.

    (1990)
  • R.J. Barro

    Inequality and growth in a panel of countries

    J. Econ. Growth

    (2000)
  • R.J. Barro

    Inequality and Growth Revisited

    (2008)
  • J.M. Barry

    The Great Influenza: The Epic Story of the Greatest Plague in History

    (2004)
  • Bentolila, S., Saint-Paul, G., 2003. Explaining movements in the labor share. Contributions in Macroeconomics 3,...
  • O. Blanchard et al.

    Macroeconomic effects of regulation and deregulation in goods and labor markets

    Q. J. Econ.

    (2003)
  • M.D. Bordo et al.

    Growing up to Stability? Financial Globalization, Financial Development and Financial Crises

    (2015)
  • M. Bruno et al.

    Economics of Worldwide Stagflation

    (1985)
  • D. Checchi et al.

    Labour market institutions and the personal distribution of income in the OECD

    Economica

    (2010)
  • C. Clague et al.

    Contract-intensive money: contract enforcement, property rights, and economic performance

    J. Econ. Growth

    (1999)
  • R.B Collier

    Paths Toward Democracy

    (1999)
  • T.G. Conley et al.

    Plausibly exogenous

    Rev. Econ. Stat.

    (2012)
  • L.P. De La Escosura

    Inequality, poverty and the Kuznets curve in Spain, 1850–2000.

    Eur. Rev. Econ. Hist.

    (2008)
  • Cited by (0)

    We gratefully acknowledge helpful comments and suggestions from participants at the Australasian Public Choice Conference 2015, the Australasian Development Workshop 2015, the European Public Choice Society Annual Meeting 2017, seminar participants at the University of Southern Denmark, and the University of Sydney. We are particularly indebted to Cecilia García-Peñalosa and Theo Eicher (editors), two anonymous referees, and Toke Aidt, Axel Dreher, Erich Gundlach, and Roman Horvath. Aart Kraay kindly provided the GAUSS code to implement weak exclusion restriction tests. Stoja Andric, Thandi Ndhlela, Paula Madsen, and Cong Wang provided excellent research assistance. Financial support from the Australian Research Council, grants DP150100061 and DP170100339 (J.B. Madsen) is also gratefully acknowledged.

    View full text