Macroeconomic conditions, inequality shocks and the politics of redistribution, 1990–2013

ABSTRACT

This contribution explores common trends in inequality and redistribution across Organization for Economic Co-operation and Development (OECD) countries from the late 1980s to 2013. Low-end inequality rises during economic downturns while rising top-end inequality is associated with economic growth. Most countries retreated from redistribution from the mid-1990s until the onset of the Great Recession, and compensatory redistribution in response to rising unemployment was weaker in 2008–2013 than in the first half of the 1990s. As unemployment and poverty risk have become increasingly concentrated among workers with low education, middle-income opinion has become more permissive of cuts in unemployment insurance generosity and income assistance to the poor. At constant generosity, the expansion of more precarious forms of employment reduces compensatory redistribution during downturns because temporary employees do not have the same access to unemployment benefits as permanent employees.

KEYWORDS:

• Comparative political economy
• income inequality
• poverty risk
• redistribution
• unemployment

Acknowledgements

Earlier versions of this contribution were presented at the European University Institute (1 July 2016), Sciences Po Paris (19 October 2016) and the London School of Economics (24 January 2017). We received many useful comments and suggestions in response to these presentations. In particular, we wish to acknowledge detailed comments provided by Silja Häusermann, Bruno Palier, John Stephens and two anonymous reviewers.

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes on contributors

Jonas Pontusson is professor at the Department of Political Science and International Relations, Université de Genève.

David Weisstanner is a PhD student at the Institute of Political Science, University of Bern.

Notes

1 Emphasizing negative externalities of inequality for the affluent (in the first instance crime), Rueda and Stegmueller (2016a) stake out a middle ground between the two camps. For our purposes, the problem with their analysis is that it does not seem to shed any light on why it is that rising inequality has not been accompanied by more redistribution.

2 Many studies show that cyclical unemployment disproportionally affects low-skilled workers (e.g., McIntosh 2008; Nickell and Bell 1995; Pollmann-Schult 2005). The standard explanation in labour economics is that the costs of dismissing and re-hiring more skilled workers, especially workers with firm-specific skills, are higher than the costs of dismissing and re-hiring low-skilled workers. See OECD (2013: ch. 1) for a detailed study of the incidence of unemployment by skill levels during the Great Recession.

3 See Jenkins et al. (2013) and OECD (2015: ch. 3) on the immediate impact of the Great Recession on post-fisc inequality among all households; and OECD (2011) on trends in income inequality among all households over the 20 years preceding the Great Recession. Eurofound (2017) provides a data-rich, EU-wide assessment of changes in different forms of income inequality over the period 2005–2014.

4 In the terminology of LIS, the former measure pertains to ‘market income’ and the latter to ‘disposable income.’ We use the terms ‘pre-fisc’ and ‘post-fisc’ for convenience, but also to signal that the distribution of income before taxes and transfers is not simply a ‘market phenomenon.’ The inequality and poverty estimates behind the figures presented in Tables 1–2 were calculated based on LIS (2016) and EU-SILC microdata. Household income data have been adjusted using the square root of the number of household members as the equivalence scale, top-coded at 10 times the median non-equivalized income and bottom-coded at 1 per cent of equivalized mean income. The aggregate indicators based on equivalized household income are restricted to household members aged between 18 and 64 using adult weights. Our LIS-based estimates of Gini coefficients correspond very closely to the Gini coefficients recorded in the ‘Comparative Welfare States Data Set’ (forthcoming version, calculated in July 2016). For 32 overlapping country-years, the correlation between our LIS-based and SILC-based estimates of pre-fisc Gini coefficients is 0.94 (p = 0.000) while the correlation between LIS-based and SILC-based estimates of post-fisc Gini coefficients is 0.95 (p = 0.000).

5 While LIS is the source of all our data for the first and second periods, our data for the third period come from LIS in three instances (Australia, Canada and the US), otherwise from SILC. As indicated in the first panel of Table 1, the exact time periods to which the LIS data refer vary by country.

6 It is important to keep in mind that the second period is longer than the first. In countries that continued to experience growing inequality, the growth of inequality slowed down in the second period. On average, the German pre-fisc income Gini coefficient increased by 0.62 per year from 1989 to 1994 and by 0.33 per year from 1994 to 2007.

7 Note that the third period in Table 1 encompasses the fiscal stimulus phase of 2008–2009 as well as the early stages of the fiscal consolidation undertaken by most OECD countries from 2010. The retreat from compensatory redistribution would be more pronounced if the analysis were restricted to 2010–2013 (OECD 2015: ch. 3).

8 The correlation coefficients for changes in pre-fisc Gini coefficients and changes in pre-fisc poverty rates are 0.90 (N = 87, p = 0.000) based on LIS estimates and 0.61 (N = 140, p = 0.000) based on SILC estimates. For levels of the two variables, the correlation coefficients are 0.89 (N = 106, p = 0.000) based on LIS and 0.82 (N = 156, p = 0.000) based on SILC.

9 In addition to the 11 countries included in Tables 1–2, the 19-country analysis includes Austria, Belgium, Greece, Ireland, Italy, Portugal, Spain and Switzerland. For both samples, the dependent variable is the change in pre-fisc inequality/poverty from one LIS survey wave to the next. We have added SILC-based estimates for 2004, 2007, 2010 and 2013 when a country is not part of the corresponding LIS wave. The regression models include country dummies as well as the level inequality/poverty in the initial year (relative to which change is measured). As these fixed-effects models capture changes within countries, the negative effect of initial of inequality/poverty does not necessarily imply convergence of inequality trends across countries.

10 Out of the 11 countries, Australia alone did not experience a contraction of GDP in 2009. For the other countries, GDP contractions in 2009 ranged between a low of −2.9 per cent (Norway) and a high of −8.7 per cent (Finland). In the first half of the 1990s, there are only four instance of annual GDP contraction greater than 2 per cent: Canada in 1991 (−3.3), Finland in 1991 (−6.4) and 1992 (−3.9) and Sweden in 1993 (−2.6).

11 Income transfers account for the lion's share of overall redistribution in most OECD countries (see Pontusson 2005: Table 7.4). Over the period 1990–2013, the correlation between changes in transfer redistribution and changes in total redistribution for our eleven countries is 0.95 (N = 75, p = 0.000).

12 Scruggs’s index takes into account the coverage of unemployment insurance and duration of insurance benefits as well as net replacement rates of unemployment insurance, but appears to be weighted towards the latter, and misses important changes in access to unemployment benefits. To illustrate, the coverage rate of unemployment insurance in Sweden fell from 84 per cent in 2004 to 68 per cent in 2008 and has since held steady according to Scruggs’s data. As reported by the Swedish employment agency, however, the percentage of unemployed receiving insurance benefits fell from 77 in 2004 to 58 in 2008 and 44 in 2012 (Arbetsförmedlingen 2013: 19). Measuring generosity as public spending on passive labour market programmes in percentage of GDP divided by the rate of unemployment, Rueda (2014, 2015) documents a strong OECD-wide tendency for generosity to decline. Substituting Rueda’s measure for Scruggs’s yields regression results that are very similar to the ones we report in Table 6 (available upon request). Since it does not include the rate of unemployment, the Scruggs measure is preferable for our purposes.

13 Based on Model 7, the effect of a one percentage point increase in unemployment are as follows: 0.820*** at low concentration (first decile); 0.449*** at median concentration; and −0.154 at high concentration (ninth decile). With only 61 observations, we run into serious collinearity issues if we add interactions with sub-period dummies alongside the interaction with concentration. Still, the point estimates for unemployment change generated by this exercise are noteworthy: 2.973 for 1990–1995; 2.911 for 1996–2007; and 3.134 for 2008–2013.

14 According to the World Wealth and Income Database, the US top 1 per cent share fell from an all-time high of 18.3 per cent in 2007 to 16.7 per cent in 2009, but surpassed the 2007 figure in 2012, and stood at 17.9 per cent in 2014.

Additional information

Funding

Jonas Pontusson acknowledges the support of the National Research Foundation of Korea (2014S1A3A2044032).