SEARCH

SEARCH BY CITATION

Abstract

  1. Top of page
  2. Abstract
  3. Introduction
  4. Union decline
  5. Levels of unionization, earnings inequality and redistribution
  6. Electoral mobilization
  7. Comparative analysis of within-country change
  8. Unionization by income
  9. Union membership and preferences for redistribution
  10. Final remarks
  11. Acknowledgements
  12. References

This article explores the implications of the OECD-wide decline of union density for earnings inequality and income redistribution by looking at aggregate (country-level) data. Over the period 1975–95, countries that experienced relatively large declines in union density also experienced relatively large increases in earnings inequality. In addition, governments apparently became less willing to engage in ‘compensatory redistribution’ in these countries. Since the early 1990s, however, union decline has become less closely associated with rising earnings inequality and redistributive policy changes. I argue that the declining relevance of unionization has to do with changes in the position of union members in the income distribution. In most OECD countries, the average union member has become relatively better off as union density has declined and union members have probably become less supportive of wage solidarity and redistributive government policies.

Introduction

  1. Top of page
  2. Abstract
  3. Introduction
  4. Union decline
  5. Levels of unionization, earnings inequality and redistribution
  6. Electoral mobilization
  7. Comparative analysis of within-country change
  8. Unionization by income
  9. Union membership and preferences for redistribution
  10. Final remarks
  11. Acknowledgements
  12. References

This article explores the implications of union decline for the rise of inequality and the politics of redistribution in the OECD countries over the last three decades. Relying on aggregate (country-level) data, my discussion of this topic is not meant to be definitive: rather, my goal is to take stock of what such data tell us and to identify questions for further research and analysis.

Recent literature in comparative political economy identifies, and seeks to resolve, what Lindert (2004: 15) and Iversen and Soskice (2009: 440) refer to as the ‘Robin Hood paradox’.1 Although higher inequality would seem to imply that more citizens stand to gain from redistribution, governments in countries with a more egalitarian distribution of market income, measured based on gross earnings from employment, typically engage in more redistribution through taxes and transfers than governments in more unequal countries. Looking within countries over the last two or three decades, the same puzzle appears, with government policies in many countries apparently becoming more ‘pro-rich’ as inequality increases. Why have democratically elected governments so rarely responded to rising inequality by engaging in new redistributive initiatives?

The observation that unionization has declined dramatically at the same time as inequality has risen will not come as a surprise to anyone, but it is fair to say that the new literature on the political economy of redistribution does not pay much attention to trade unions or other ‘intermediary organizations’. Most of this literature has a distinctively individualist orientation, focusing on individual preferences for redistribution and, secondarily, on the aggregation of individual preferences by political parties. There is, however, an important theoretical tradition in comparative welfare-state research, commonly known as ‘power resources theory’ (PRT), which assigns a key explanatory role to the collective organization of workers. As articulated most clearly by Bradley et al. (2003), the PRT solution to the Robin Hood paradox holds that earnings compression and redistribution are both to be attributed to ‘working-class mobilization’ or, in other words, that the negative association between earnings inequality and redistribution is essentially spurious. Representing the interests of the median union member, the wage demands articulated by unions typically favour low-wage workers relative to high-wage workers, and, as a result, union strength is associated with earnings compression. At the same time, unions mobilize workers to vote for parties that promise to redistribute income. This mobilization may lead to left government, as typically emphasized by PRT scholars (see also Korpi 2006), but it could also lead all parties to engage in more redistribution.

In what follows, I will show that, across OECD countries, levels of earnings inequality and redistribution are indeed associated with levels of unionization in the manner posited by PRT. I shall also present evidence in support of the key PRT claim that unions play an important role as agents of electoral mobilization. Across countries, the decline of union density correlates quite closely with the decline of voter turnout over the last two or three decades. However, cross-national comparison of changes over time poses a challenge for PRT. Over the period from the late 1970s to the mid-1990s, union decline is clearly correlated with changes in earnings inequality and redistribution in the manner predicted by PRT, but this is not the case for the period from 1995 to 2010. Union decline seems to have become less important as a driver of rising inequality and also seems to have become less relevant for explaining the lack of compensatory redistribution.

My tentative explanation of the attenuation of the egalitarian effect of unionization has to do with the position of union members in the income distribution. In the PRT-inspired literature, the interests of workers and union members are always the same: what varies, across countries and over time, is the extent to which workers are organized in unions. By contrast, I want to bring to the fore the question of who unions organize. As we shall see, union members are relatively well paid today in most OECD countries. Thus is by no means obvious that the material interests of the average union member are served by earnings compression or redistribution. Although I do not have any historical data to support this claim, it seems likely that union decline has been accompanied by significant shifts in the composition of union membership, rendering union members less supportive of redistributive wage policies and fiscal policies than they were in the 1970s or 1980s.

In a sense, my article seeks to improve on PRT by incorporating considerations of individual self-interest from the literature on the political economy of redistribution. However, unions cannot be understood simply to represent the self-interest of their members. As I will briefly report, comparative analyses of recent survey data indicate that, controlling for income, union members are more supportive of egalitarianism and redistribution than other citizens. This important observation must also be taken into account if our goal is to assess the implications of union decline for redistributive politics.

Some readers will be troubled by my inattention to institutional differences among OECD countries. In particular, they may object to the way I equate union strength with union density. In many countries, collective bargaining coverage rates may well be a better measure of union influence over distributive labour market outcomes. I focus on unionization because this variable is more directly relevant to the role of unions in the electoral domain. It also deserves to be noted that changes in union density and bargaining coverage since 1990 are quite closely correlated.2 More broadly, let me state at the outset that I do not mean to imply that industrial relations institutions and political institutions — for example, electoral rules, party systems and federalism — are inconsequential. In a short article such as this one, choices obviously have to be made: I have chosen to focus on the (changing) effects of unionization across different institutional contexts.

My discussion unfolds in seven steps. First, I present cross-national data on union density and briefly discuss reasons for the pervasive decline of union density in OECD countries. Secondly, in support of the PRT solution to the Robin Hood paradox, I present scatterplots showing level relationships between, on the one hand, unionization and, on the other hand, earnings inequality and redistribution. Thirdly, I elaborate on the electoral mobilization mechanism in PRT and present evidence indicating that declining voter turnout can partly be attributed to union decline. Fourthly, I demonstrate that union decline has become less relevant to changes in earnings inequality and redistribution since the early 1990s. Fifthly, I present recent data on unionization by income and some reasons why it seems plausible that union members have moved up in the income distribution as unionization has fallen. Sixthly, I discuss the relationship between union membership and preferences for redistribution. Finally, I conclude by explaining how my argumentation differs from that of right-wing union critics.

Union decline

  1. Top of page
  2. Abstract
  3. Introduction
  4. Union decline
  5. Levels of unionization, earnings inequality and redistribution
  6. Electoral mobilization
  7. Comparative analysis of within-country change
  8. Unionization by income
  9. Union membership and preferences for redistribution
  10. Final remarks
  11. Acknowledgements
  12. References

Jelle Visser's comprehensive and much-cited dataset provides annual observations of net union density (employed union members in percent of the employed labour force) since 1960. Based on this source, Table 1 identifies the peak year of union density for 22 OECD countries and reports the percentage change in union density (defined as employed union members in percent of the employed labour force) from the peak to 2011. I have sorted the countries into three groups, based on the timing of trend changes. The first group consists of countries in which union density peaked during or prior to the 1960s.3 This group includes the United States, Austria, the Netherlands, Switzerland, Norway, Japan and France. With the exception of Norway, which is really a case of union stability rather than decline, these countries have experienced a very long, more or less continuous, decline in union density. By 2010, their unionization rates had fallen by at least half, and in two cases, France and the US, by nearly two-thirds. The second group consists of countries in which the onset of union decline began between 1975 and 1989. This group includes three of the four countries that have experienced the most dramatic cumulative decline in union density — New Zealand, Australia and Portugal — but it also includes several countries characterized by relatively modest cumulative decreases: Spain, Canada and Denmark. Finally, Belgium, Finland, Sweden and Germany make up a third group of countries, the ‘late comers’. In these countries, union density peaked in the first half of the 1990s, and, with the very notably exception of Germany, the cumulative decline of union density has so far been quite limited.

Table 1. Union Decline, 1960–2010
 Peak year (or earlier)Peak2010% Change since peakRank
Source: Visser (2013).
United States (US)196030.911.4−63.15
Austria (AT)196067.928.4−56.86
The Netherlands (NL)196040.019.3−50.87
Switzerland (CH)196036.117.2−52.48
Norway (NO)196160.854.8−9.921
Japan (JP)196435.518.4−48.29
France (FR)196922.27.9−64.42
Australia (AU)197650.218.0−64.14
Italy (IT)197650.535.5−29.715
United Kingdom (UK)197851.827.1−47.711
Ireland (IE)197864.036.6−42.813
New Zealand (NZ)198069.120.8−69.91
Portugal (PT)198054.819.3−64.83
Greece (GR)198039.025.4−34.914
Spain (ES)198018.715.6−16.619
Denmark (DK)198380.868.5−24.816
Canada (CN)198435.930.0−16.418
South Korea (SK)198918.69.7−47.812
Germany (DE)199136.018.6−48.310
Finland (FI)199380.770.0−13.320
Sweden (SE)199487.468.9−21.217
Belgium (BE)199555.750.6−9.222

Sweden, Denmark and Finland stand out as the three countries with the highest peaks of union density (above 80 per cent). Belgium and Norway are the two countries that have experienced the smallest declines of union density (below 10 per cent). As union density has fallen dramatically in other countries, these five countries have become an increasingly distinctive high-density group. While there is clearly a common de-unionization trend across the OECD countries, it is not the case that unionization rates have converged (see also Pinto and Beckfield 2011).

It is commonplace to attribute high union density in Sweden, Denmark, Finland and Belgium to the role of unions in the administration of unemployment insurance (e.g. Ebbinghaus et al. 2011; Scruggs 2002; Western 1999). Selective incentives created by the ‘Ghent system’ served to sustain unionization in these countries as unemployment rose in the 1980s and 1990s, and recent reforms of unemployment insurance have had a negative impact on union density in Finland (Böckerman and Uusitalo 2006) as well as Sweden (Kjellberg 2011). More broadly, political and institutional factors are surely important for understanding cross-national variation around the OECD-wide trend for union density to decline. Yet the main challenge, it seems to me, is to explain the general trend. What accounts for the OECD-wide decline of union density over the last 20–30 years?

Very usefully summarized by Schnabel (2012), the existing literature provides very little evidence of any direct impact of globalization on union density.4 It seems more plausible to conceive deindustrialization as the ‘driver’ behind the OECD-wide trend, although this line of argument is complicated by the fact that public-sector workers are typically much better organized than private-sector workers. In many countries, the expansion of public-sector employment in the 1970s and 1980s meant that declining private-sector unionization did not translate into declining aggregate union density.

Visser's dataset contains information on union density by sector as well as aggregate union density, but this part of the dataset leaves much to be desired: the sector-specific data pertain to different time periods for different countries, and there are only single-year observations, often quite dated, for some countries. Table 2 summarizes the available data on union density in three sectors: the public sector, manufacturing and ‘commerce’.5

Table 2. Union Density by Sector, Various Years
 Public sectorManufacturingCommerce
EarliestRecent% ChangeEarliestRecent% ChangeEarliestRecent% Change
Source: Visser (2013).
Australia (1982–2009)73.046.3−36.654.021.0−61.122.012.7−42.3
Austria (1980–98)67.560.6−10.267.656.5−16.426.414.7−44.3
Belgium (1980–92)   79.892.9+16.432.237.9+17.7
Canada (1984–2008)71.870.6+1.745.026.4−41.311.012.5+13.6
Denmark (1980–2008)69.283.4+20.598.078.4−20.053.055.0+3.8
Finland (1991–2008)85.086.0+1.280.072.0−10.062.053.0−14.5
Ireland (2001–09)56.366.6+18.341.929.3−30.121.215.5−26.9
Italy (1980–97)59.6  55.939.2−29.922.123.1+4.5
The Netherlands (1980–2007)59.728.0−53.142.825.0−41.69.612.0+25.5
Norway (1980–2008)74.380.0+7.181.057.0−29.615.025.0+66.7
Sweden (1980–2008)81.184.0+3.695.079.0−16.846.057.0+23.9
United Kingdom (1995–2009)61.356.6−7.732.320.8−35.611.012.4+12.7
United States (1980–2009)34.837.4+8.635.010.9−68.98.74.8−44.8
France (2003) 12.5  7.5  2.5 
Germany (1985) 50.4  50.0  14.0 
Greece (2003) 63.8       
Japan (1988) 23.3  32.0  9.0 
New Zealand (2008)    37.5  3.7 
Portugal (1989) 45.0       
South Korea (2006) 34.3       
Switzerland (1985) 70.6  33.0  11.0 

The consistency and extent of union decline in the manufacturing sector is the most outstanding feature of Table 2. In all but one of the countries for which we have measures of change over time, we observe significant declines in manufacturing union density. (The one exception is Belgium, and this exception may well be due to the fact that the Belgian time series ends in 1992.) By comparison, the public-sector story is much more mixed. In just about half the countries covered by these data, unionization of the public sector has declined; in the other half, it has held steady or increased. That public-sector unions have fared better than manufacturing unions will not come as a surprise to anyone. What is more surprising is that unionization trends in low-end private services (commerce) are also very mixed. Over varying time periods, we observe big declines in the unionization of commercial workers in Australia, Austria, Finland, Ireland and the United States, but quite significant increases in Belgium, Canada, the Netherlands, Norway, Sweden and the United Kingdom. Austria and Finland stand out as the only two cases in which unionization of commerce has declined more than unionization of manufacturing.

Clearly, the OECD-wide decline of aggregate union density cannot be explained simply by the failure of unions to keep up with the shift of employment to private services, but it remains plausible to argue that the shift of employment to private services has been an important source of de-unionization. As Table 2 shows, union density in low-end private services (commerce) has always been, and remains, much lower than union density in either manufacturing or the public sector. This observation holds for every one of the 18 countries for which Visser provides at least one observation of union density in commerce. Holding unionization rates constant, we should expect a shift in employment from manufacturing or public services to private services to be associated with lower aggregate union density.6

The existing literature demonstrates that establishment size and atypical employment are strongly associated with levels of unionization within and across sectors (see Schnabel 2012). Workers in large production units are more likely to perceive employers as adversaries and to join unions. In addition, it is commonly argued that economies of scale make it easier (cheaper) for unions to organize workers in larger production units.7 Regarding atypical employment, the standard argument is that part-time and temporary workers are less willing to bear the costs of joining a union and also more vulnerable to employer pressure (Ebbinghaus et al. 2011). The expansion of temporary employment is primarily a continental European phenomenon, but the point about employer pressure applies equally to employees with indefinite employment contracts when such contracts are not backed by legal protections (King and Rueda 2008). Critically, decentralization of production and the spread of atypical employment provide a plausible explanation of the decline of union density in manufacturing as well as low levels of union density in private services (relative to manufacturing and public services).8

Levels of unionization, earnings inequality and redistribution

  1. Top of page
  2. Abstract
  3. Introduction
  4. Union decline
  5. Levels of unionization, earnings inequality and redistribution
  6. Electoral mobilization
  7. Comparative analysis of within-country change
  8. Unionization by income
  9. Union membership and preferences for redistribution
  10. Final remarks
  11. Acknowledgements
  12. References

As noted at the outset, PRT resolves the ‘Robin Hood paradox’ by arguing that working-class mobilization, measured by unionization and electoral support for left parties, promotes compression of market earnings as well as redistributive government policies. With union density as the measure of labour power, comparing OECD countries at any given point in time yields considerable support for these claims.

The OECD's employment database provides annual observations of decile earnings ratios for full-time employees. For a handful of countries, these observations extend back to the 1970s. The rest of the OECD countries enter the dataset in later years, some as late as the 2000s. For all countries for which the OECD provides comparable 1990 data, Figure 1 plots 90-10 ratios against the level of union density.9 Figure 2 repeats this exercise with data for 2009. In both figures, we see clearly that countries in which unions are strong tend to have a more equal distribution of earnings. The slope of the best-fit line is almost exactly the same in the two figures, but we observe more dispersion around the best-fit line in Figure 2 (R2 = 0.443) than in Figure 1 (R2 = 0.575), suggesting that union density has become a less consistent predictor of earnings inequality over time.10

figure

Figure 1. Earnings Inequality (Full-Time Employees) and Union Density, 1990.

Coefficient = −0.024, t = −3.49, R2 = 0.575.

Note: German data refer to West Germany.

Sources: OECD (2013), Visser (2013).

Download figure to PowerPoint

figure

Figure 2. Earnings Inequality (Full-Time Employees) and Union Density, 2008.

Coefficient = −0.026, t = −3.73, R2 = 0.423.

Note: 2005 data for the Netherlands, 2007 data for Belgium.

Sources: Same as Figure 1.

Download figure to PowerPoint

Figures 3 and 4 in turn speak to the relationship between union density and redistribution. Following conventional practice, redistribution is here defined as the percentage change in the Gini coefficient produced by taxation and income transfers, that is, the percentage change from inequality measured in terms of market (‘pre-fisc’) household income to inequality measured in terms of disposable (‘post-fisc’) household income.11 With 2010 data (Figure 4) as well as 1990 data (Figure 3), it is clearly the case that, across countries, unionization and redistribution are positively associated with each other. Again, however, this association appears to have become less consistent over time (indicated by a smaller t-statistic and R2 for Figure 4 relative to Figure 3). In particular, France, Germany, the Netherlands and Portugal seem to have emerged as an increasingly distinctive group of countries characterized by relatively low levels of union density and relatively high levels of redistribution.

figure

Figure 3. Redistribution and Union Density, 1990.

Coefficient = 0.348, t = 4.17, R2 = 0.478.

Sources: Solt (2011) and Visser (2013).

Download figure to PowerPoint

figure

Figure 4. Redistribution and Union Density, 2010.

Coefficient = 0.335, t = 3.06, R2 = 0.330.

Sources: Same as Figure 3.

Download figure to PowerPoint

It is immediately evident from visual inspection of Figures 1-4 that the cross-national association of union density with redistribution as well as earnings compression depends crucially on the Nordic countries. Dropping the Nordic countries from the analysis, the core PRT claims could no longer be sustained in this manner. But if we are interested in exploring the effects of union density, we surely ought not restrict our sample by eliminating all cases with high values on this variable.

Needless to say, the evidence presented in Figures 1-4 does not allow us to infer that unionization is a ‘cause’ of wage compression or redistribution. The empirical associations shown in these graphs could be spurious. The possibility of reverse causality must also be considered. I shall return to the latter issue in due course. For the time being, suffice it is to note that a number of multivariate regression analyses of OECD data on relative earnings, pooling cross-section time-series observations and controlling for potentially confounding variables, such as wage bargaining centralization and public-sector employment, report significant negative effects of union density or the percentage of the labour force covered by union-negotiated wage agreements (e.g. OECD 2011; Pontusson et al. 2002; Rueda and Pontusson 2000).12 Further empirical evidence supporting the idea that unions tend to compress earnings differentials is provided by Scheve and Stasavage (2009), who analyse the evolution of top income shares in 12 countries over the period from 1916 to 2000. Although Scheve and Stasavage warn against a causal interpretation of this finding, union density turns out to be the only ‘institutional variable’ that is consistently associated with (less) income inequality in their long-run analysis. It is also noteworthy that Iversen and Soskice (2006) find that unionization has a significant positive effect on redistribution while controlling for government partisanship and electoral rules.

Electoral mobilization

  1. Top of page
  2. Abstract
  3. Introduction
  4. Union decline
  5. Levels of unionization, earnings inequality and redistribution
  6. Electoral mobilization
  7. Comparative analysis of within-country change
  8. Unionization by income
  9. Union membership and preferences for redistribution
  10. Final remarks
  11. Acknowledgements
  12. References

Electoral mobilization featured prominently in early PRT formulations linking working-class mobilization to welfare-state development (see Korpi 1983 in particular). Building on PRT insights, I want to suggest that the role of unions as agents of electoral mobilization is a theme that should be resuscitated and brought to bear on the puzzles identified in the recent literature on the politics of inequality and redistribution.

The argumentation that I have in mind rests on two premises: first, relative income is an important determinant of support for redistribution (poor citizens are more likely to support redistribution) and, secondly, preferences for redistribution in turn matter for vote choice (citizens who support redistribution are more likely to vote for parties that promise to redistribute). The growing literature on preferences for redistribution provides ample evidence in support of the first proposition, but has yet to tackle the question of what difference support for redistribution makes for voting behaviour. Starting with Finseeras (2009), every single study that I have seen finds a statistically significant negative association between relative income, measured by self-reported household income, and support for redistribution, measured by agreement with the statement that ‘the government should take measures to reduce differences in income levels’. Importantly, Rueda (2012) demonstrates that the effect of relative income still holds when we control for social class (based on Goldthorpe's class schema).

Analysing data from the Luxembourg Income Study, Mahler (2008) shows that a number of different measures of redistribution of income through taxation and transfers are strongly associated with voter turnout (cf. Kenworthy and Pontusson 2005: 459–62). Mahler also demonstrates that income-based disparities in voter turnout tend to decline as aggregate voter turnout rises, an empirical observation that is reinforced by Schäfer (2011) and Schäfer and Streeck (2013: 13–16). Given that relative income is a strong (negative) predictor of support for redistribution at the individual level, it seems highly plausible to suppose that the cross-national association between voter turnout and redistribution stems from the fact that low-income citizens represent a more important electoral constituency in countries with higher voter turnout and that the same logic applies to temporal variation within countries.

Stronger representation of low-income citizens in the electorate ought to present left parties with opportunities to gain support and may encourage these parties to prioritize redistribution, but standard models of electoral competition imply that centrist and centre-right parties would seek to preempt leftist gains by adopting more redistributive platforms. For the purposes of this short article, it is not necessary to resolve the thorny question of whether — or under what circumstances — parties cater to core constituencies or behave according to median voter logic.13

Turning to the effects of unionization on voter turnout, many studies of electoral participation in the United States have shown that union membership is associated with electoral turnout, controlling for other individual characteristics that are also associated with turnout. Based on 1978–1990 data from the General Social Survey, Kerrissey and Schofer (2013) estimate that the probability of voting is 18 per cent higher among union members than survey respondents who are eligible to vote but do not belong to a union. Pooling European Social Survey (ESS) data for 19 countries in 2008–2010, Mosimann and Pontusson (2013) report similar results: survey respondents who identify themselves as union members are more likely to vote than other survey respondents. This association remains significant when we control for self-declared in interest in politics, suggesting that it is not simply the case that politically engaged individuals are more likely to join unions. Unions may stimulate political interest, but being a union member also seems to have a direct ‘mobilizational effect’. We might think of this effect either in terms of deliberate union efforts to lower the costs of voting for individuals (most obviously relevant in the US case, distinguished by voter registration requirements and weekday elections) or as a social-network effect, premised on the idea that individuals are more likely to vote to the extent that they interact with others who vote (or at least say that they vote).14

Consistent with the results of these individual-level analyses, Radcliff and Davis' (2000) multivariate analysis of macro-level data shows that union density is a powerful predictor of variation in voter turnout across US states over the period 1964–1982 as well as a powerful predictor of variation in voter turnout in 19 OECD countries over the period 1960–1988. Plotting average turnout in national elections between 2000 and 2009 against union density in 2005, Figure 5 indicates that the cross-national association between union density and voter turnout identified by Radcliff and Davis still holds.15

figure

Figure 5. Average Voter Turnout in 2000–09 and Union Density in 2005.

Coefficient = 0.267, t = 2.23, R2 = 0.238 (without Finland: coefficient = 0.351, t = 2.68, R2 = 0.325).

Sources: Armingeon et al. (2012), complemented by turnout data for Korea from International Institute for Democracy and Electoral Assistance (http://www.idea.int/vt/); Visser (2013).

Download figure to PowerPoint

For our present purposes, the crucial question concerns change over time: has union decline been a source of declining voter turnout in OECD countries? While other variables must surely be taken into account as well, Figure 6 suggests that union decline deserves more attention than it has received in the literature on this topic (notably Franklin 2004). On the vertical axis, Figure 6 reports the percentage change in voter turnout from the election immediately following the peak of union density (or the peak year itself if there was an election that year) to the most recent pre-2008 election. On the horizontal axis, it reports the percentage change in union density from the peak year to the year of the most recent pre-2008 election. While restricting the analysis to the time period since the onset of union decline allows us to focus on the component of the decline of voter turnout that could potentially be explained by union decline, I have excluded elections since 2008 on the grounds that the Great Recession seems to have precipitated ‘extraordinary’ voter mobilization in many countries, apparently unrelated to unionization.16 With Canada and New Zealand as notable outliers, we observe a quite strong association between the extent of union decline and the extent of turnout decline.17

figure

Figure 6. Percentage Change in Voter Turnout and Union Density since Union Density Peak.

Coefficient = 0.314, t = 5.21, R2 = 0.630.

Note: See text for explanation.

Sources: Same as Figure 5.

Download figure to PowerPoint

Comparative analysis of within-country change

  1. Top of page
  2. Abstract
  3. Introduction
  4. Union decline
  5. Levels of unionization, earnings inequality and redistribution
  6. Electoral mobilization
  7. Comparative analysis of within-country change
  8. Unionization by income
  9. Union membership and preferences for redistribution
  10. Final remarks
  11. Acknowledgements
  12. References

Level relationships are broadly supportive of core PRT claims: across countries, union density has been and remains associated with (a) less earnings inequality, (b) higher voter turnout, and (c) more redistributive government. In addition, change in union density appears to be associated with change in voter turnout (Figure 6). In this section, I want to probe the question of change over time further. Looking across countries, has union decline also been associated with rising earnings inequality and declining redistribution? With regard to redistribution, a more appropriate formulation might be: has union decline been associated with a decline in the responsiveness of redistribution to increases in market inequality?

The existing literature addresses these questions indirectly by pooling cross-section time-series data and estimating regression models with country fixed effects. The standard set-up (e.g. OECD 2011) focuses on the determinants of within-country change in the dependent variable, but assumes that the effects of independent variables, such as union density, have been constant over time. The fact that union density seems to be a less consistent predictor of earnings inequality and redistribution with 2008–2010 data than with 1990 data raises questions about this assumption. In a preliminary fashion, I propose to explore changes in the effects of unionization by comparing patterns in pre-1995 data with patterns in post-1995 data. The 1995 cut-off point is arbitrary, but convenient in the sense that it allows for comparing time periods of more or less equal duration.

Let us begin by considering the relationship between union density and earnings inequality. For 10 countries, Figure 7 plots changes in 90-10 earnings ratios against changes in union density over the period 1975–1995, and Figure 8 does the same for the period since 1995. In Figure 9, I have added six additional countries for which we have observations of change in earnings inequality over the period 1995–2009. As collective bargaining agreements often last for two or three years, changes in union density have been lagged by three years to generate these graphs. It should also be noted that change in both variables is measured relative to initial levels or, in other words, in percentages. The intuition that justifies this choice is simple. On the one hand, a decline in union density by, say, five percentage points surely represents a bigger decline in union influence over the distribution of earnings if union density is 15 per cent (the United States) than if union density is 80 per cent (Sweden). On the other hand, any given increase in the 90-10 earnings ratio, measured in absolute terms, represents a larger increase of earnings inequality if the initial ratio is smaller.

figure

Figure 7. Percentage Change in 90-10 Earnings Ratios and Union Density, 1975–95.

Coefficient = −0.309, t = −2.92, R2 = 0.517 (without Japan: coefficient = −0.340, t = −4.10, R2 = 0.706).

Note: The change in wage inequality is measured over the following time periods: 1984–1995 for Germany, 1984–1996 for New Zealand, 1980–1995 for Denmark, 1977–1995 for the Netherlands and 1975–1995 for the remaining countries. Changes in union density have been lagged by three years.

Sources: Same as Figure 1.

Download figure to PowerPoint

figure

Figure 8. Percentage Changes in 90-10 Earnings Ratios and Union Density (10 Countries), 1995–2009.

Coefficient = −0.160, t = −1.44, R2 = 0.205.

Note: See Figure 9 for time periods covered.

Sources: Same as Figure 1.

Download figure to PowerPoint

figure

Figure 9. Percentage Changes in 90-10 Earnings Ratios and Union Density (16 Countries), 1995–2009.

Coefficient = −0.113, t = −0.65, R2 = 0.030. Time periods covered: 1994–2008 for Ireland; 1995–2009 for Australia, Japan, South Korea, New Zealand, the United Kingdom and United States; 1995–2008 for Denmark, Finland, Germany and Sweden; 1995–2005 for the Netherlands; 1996–2008 for Switzerland; 1997–2009 for Norway; and 1999–2007 for Belgium.

Sources: Same as Figure 1.

Download figure to PowerPoint

As noted earlier, the OECD database on relative earnings extends back to the 1970s for only a handful of countries. The distribution of earnings is a sticky, slow-moving variable, and we would not expect union decline (or growth) to have discernible effects on the distribution of earnings unless it persists over some reasonably extended period of time. At the same time, meaningful cross-national comparisons presuppose a reasonably large number of countries. For half of the 10 countries included, Figure 7 reports percentage changes in the 90-10 earnings ratio from 1975 to 1995, matched with percentage changes in union density from 1972 to 1992. For the other half, the time span is shorter (inside the 1975–1995 window), with union density lagged accordingly. The data summarized in Figure 8 cover more or less the same time period for all countries while Figure 9 again encompasses some countries for which the data on changes in earnings inequality are more limited.

With Japan as a notable outlier, Figure 7 shows a strong negative association between changes in union density over 1972–1992 and changes in earning inequality of over 1975–1995. As suggested by Cecchi, Visser and van de Werfhorst (2010), we should be cautious in interpreting such evidence to imply that union decline has been a cause of rising earnings inequality (see also Scheve and Stasavage 2009). Observing that an individual's probability being a union member falls as his or her earnings rise above median earnings, these authors argue that workers who are farther above the median stand to lose more from solidaristic wage policies and will therefore shun unions. Thus union decline might, in part, be a consequence of increasing earnings inequality. Assuming that unions typically pursue solidaristic wage policies, however, this argument complements rather substitutes for the argument that union decline has been a source rising inequality. It should also be noted that lagging changes in union density by three years lends plausibility to the standard causal interpretation.

For our present purposes, the more important point is that we do not observe any consistent cross-national association between changes in union density and earnings inequality for the period since 1995. Restricting the analysis to the 10 countries included in Figure 7, the slope of the best-fit line is still negative, but the association between the two variables fails to satisfy even the most lax criteria of statistical significance (Figure 8). With other countries added (Figure 9), we find no indication whatsoever that countries experiencing larger declines in union density have also experienced larger increases of earnings inequality. On the one hand, Japan and Ireland stand out as cases that have combined wage compression with large decreases in union density. On the other hand, strong and relatively stable unions have apparently accommodated large increases in earnings inequality, albeit from low initial levels, in Norway, Denmark and Finland.18

Replicating Figures 7-9 with changes in redistribution on the vertical axis, we do not observe any association between changes in union density and redistribution over the period 1980–1995, let alone 1995–2010. From a dynamic perspective, the relationship between union density and redistribution is more complicated than the relationship between union density and earnings inequality. As Kenworthy and Pontusson (2005) demonstrate, most contemporary welfare states are organized such that redistribution increases more or less automatically in response to increases in household income inequality. To the extent that union decline has been a source of rising inequality before taxes and transfers, it would thus be associated with more redistribution.

Table 3 reports the results of estimating a simple multivariate regression model that separates the ‘welfare-state compensation effect’ from the direct effect of changes in union density and also takes into account potential convergence in levels of redistribution across OECD countries. Based on data for 20 countries, the results are quite interesting.19 For the period 1980–1995, we find that redistribution increased more (or declined less) in countries that experienced more rapid growth of inequality measured in terms of market income. This remains true for the period 1995–2010, but redistribution appears to have become less responsive to inequality over time. For 1980–1995, we also observe some convergence among OECD countries: over these 15 years, redistribution increased more in countries that, on average, redistributed less in 1980. This effect no longer appears when the model is estimated with data for 1995–2010. Most importantly for our present purposes, the model yields a significant positive effect of change in union density for the period 1980–1995 and no such effect for 1995–2010. Controlling for catch-up and inequality-induced redistribution, increasing unionization appears to have been a source of pressure on governments to redistribute more — or, conversely, union decline removed pressure on governments to redistribute — in the 1980s and early 1990s. For 1995–2010, the sign of the (insignificant) coefficient for change in union density is actually negative.

Table 3. Determinants of Change in Redistribution
 1980–19951995–2010
  1. Note: Standard errors in parentheses.

  2. Sources: Visser (2013), Solt (2011).

  3. * p < 0.05; ** p < 0.01.

Change in Gini for market income0.950** (0.288)0.430* (0.180)
Initial level of redistribution–0.354* (0.132)–0.088 (0.086)
Change in union density0.347** (0.119)–0.121 (0.178)
Constant13.632* (5.487)1.219 (3.340)
N2020
R20.7020.307

The evidence presented above suggests that the egalitarian effects of unionization, for government redistribution as well as the distribution of earnings from employment, have diminished, possibly disappeared altogether, over the last two decades or so. In other words, union decline seems to have become a less powerful explanation of inegalitarian labour market trends and retreat from redistribution over time. It goes without saying that this observation needs be verified through more rigorous analyses based on pooling cross-section time-series data. For the time being, let me simply note that the data patterns shown in Figures 7-9 are consistent multivariate regression results reported by Baccaro (2011) and Golden and Wallerstein (2011).

Unionization by income

  1. Top of page
  2. Abstract
  3. Introduction
  4. Union decline
  5. Levels of unionization, earnings inequality and redistribution
  6. Electoral mobilization
  7. Comparative analysis of within-country change
  8. Unionization by income
  9. Union membership and preferences for redistribution
  10. Final remarks
  11. Acknowledgements
  12. References

One plausible explanation for the diminishing effects of unionization on inequality and redistribution is that the position of union members in the income distribution has changed over time. This relates to what I consider to be the main weakness of PRT, namely the assumption that workers constitute a more or less homogenous constituency with a common interest in social protection and redistribution. The critical variable in PRT-inspired literature is the extent to which this constituency is organized, or mobilized, and thus able to exert influence over distributive outcomes, with ‘working-class mobilization’ varying across countries and over time. To my mind, the PRT framework needs to incorporate the idea that the interests represented by unions also vary across countries and over time. In particular, I want to suggest that the dramatic decline of unions that has occurred in many OECD countries since the 1970s has been accompanied by important shifts in the position of union members in the income distribution.

The recent literature on ‘dualization’ frequently argues (or implies) that unions organize and represent labour market insiders or, in other words, middle-aged male workers with protected jobs and relatively high wages (e.g. Rueda 2007 and Emmenegger et al. 2012). Sidestepping conceptual and empirical problems involved in distinguishing between ‘insiders’ and ‘outsiders’, my discussion here will be restricted to unionization by income. Table 4 presents the main findings of Becher and Pontusson's (2011) analysis of ESS data from 2006 and 2008. The analysis is based on a survey question asking respondents to identify the disposable income of the household to which they belong. Having adjusted for household size, we estimated the distribution of household income for all survey respondents belonging to the dependent labour force (i.e. either employed by someone else or registered as unemployed), assigned each respondent to an income quintile and then calculated the unionization rate for each quintile.20 The last column reports our estimates of the percentage of union members who belong to households below the median disposable household income of the dependent labour force.

Table 4. Unionization by (Household) Income in Western Europe, 2006–2008
 Unionization by quintile% Union members with income below median
BottomSecondMiddleFourthTop
  1. Note: The figures for Austria, Belgium and Ireland are based on ESS 2006 while the figures for the other countries are based on ESS 2008. Austria and Ireland were not part of ESS 2008. For Belgium, our estimation method produces an exceptionally low rate of unionization in the middle quintile based on 2008 data (41 per cent); otherwise, the results are virtually identical to those presented above (see Becher and Pontusson 2011: 194).

Source: European Social Survey, calculations by Michael Becher.
Denmark768792899247.8
Finland708475787750.8
Sweden607672697449.8
Norway455661596646.9
Belgium555861463951.0
Ireland243652453841.9
Austria303637404742.7
United Kingdom132533352839.9
The Netherlands192424302045.0
Germany122017242440.6
Switzerland151917151651.4
Greece5810133022.9
Spain11914121646.5
France81010101442.0
Portugal341121316.8

The most striking feature of Table 4 is that for every country but Belgium and Portugal, union density in the top income quintile is higher than union density in the bottom quintile. In a majority of countries, union density peaks in either the fourth or the fifth quintile. The exceptions to the latter generalization are Finland, Sweden and Switzerland, where unionization is highest in the second quintile, and Belgium and Ireland, where unionization is highest in the third quintile. Denmark is a special case, with the third and fifth quintiles being tied for first place. Setting Switzerland aside, the countries with the highest aggregate rates of unionization are also the countries with the highest percentage of union members with household incomes below the median. By mathematical necessity, the proportion of union members with incomes below the median will converge on 50 per cent as aggregate union density approaches 100 per cent. What is striking is that Switzerland seems to be the only case in which this logic would reduce (rather than increase) the percentage of low-wage workers among union members.

As noted above, much of the literature on implications of unionization for the earnings distribution proceeds from the idea that the wage of the median union member is lower than the average wage, and, as a result, the median union member favours solidaristic wage policies (see, most notably, Freeman 1980). Similarly, my earlier discussion of the implications of unionization for redistributive politics assumes that unions primarily mobilize citizens with market incomes below the average and thus shift the centre of political gravity — or, in other words, ‘the preferred policy of the median voter’ — in favour of redistribution. According to Becher and Pontusson (2011: 197), however, the ratio of the household income of the median union member to the household income of the median voter ranges between a low of 1.00 for Switzerland and a high of 1.37 for Greece. Assuming that preferences for redistribution are strictly self-interested, we would not expect higher rates of voter turnout among union members to be a source of higher support for redistribution in the electorate.

Although I do not have any comparable estimates of unionization by income for earlier time periods, I find it hard to believe that a similar analysis for, say, the 1970s would yield results akin to those presented in Table 4.21 In many OECD countries, organized labour has probably become more representative of high-wage workers over the last 20 or 30 years. Holding wage differentials constant, the relative earnings of the median or average union member will increase if it is the case that union density among high-wage workers increases more or declines less than union density among low-wage workers. It is tempting to suppose that union decline derives first and foremost from the failure of unions to organize low-wage sectors whose share of total employment has increased, but the data on unionization by sector presented in Table 2 raise questions about this line of argument. In general, workers in manufacturing are better paid than workers in commerce, but it does not appear to be the case that union density has held up better in manufacturing than in commerce.

Figure 10 provides the basis for a different story as to why organized labour has become more representative of high-wage workers over the last two decades.22 Pooling data from 16 OECD countries, this figure shows the evolution of average real hourly earnings in export-oriented manufacturing, public services and low-end private services (hotels and restaurants as well as wholesale and retail trade) from 1990 to 2007. On average, earnings in manufacturing increased by more than 31 per cent while earnings in public services increased by just about 25 per cent, and earnings in low-end private services increased by 17–18 per cent over this period.23 In other words, earnings have grown much faster in sectors where unions have been and remain relatively strong, and, as a result, the relative earnings of union members have increased. Given the very large initial differences in union density between manufacturing and commerce, it seems reasonable to suppose that the effects of changing earnings relativities have trumped the effects of disproportionate declines in union density among manufacturing workers.

figure

Figure 10. Trajectory of Average Real Hourly Wages by Sector in 16 OECD Countries, 1990–2007.

‘High-end manufacturing’ includes sectors with NACE codes D21 through D37: paper and pulp, petroleum products, chemicals, rubber and plastics, non-metallic mineral products, metal products, machinery, electrical and optical equipment, transportation equipment and miscellaneous manufacturing. ‘Low-end services’ include NACE codes G52 (retail trade) and H (hotels and restaurants). Average wages for each sector take into account the employment of sub-sectors. The countries included are Australia, Austria, Belgium, Canada, Denmark, Finland, France, Germany, Ireland, Italy, Japan, the Netherlands, Spain, Sweden, the United Kingdom and the United States. PA, public administration.

Source: EU KLEMS dataset (http://www.euklems.net), calculations by Lucio Baccaro.

Download figure to PowerPoint

My working hypothesis concerning the diminishing egalitarian effects of unionization can be summarized as follows. In many countries, national union movements resisted market pressures for inter-sectoral wage differentials to rise from the 1970s onwards. Manufacturing unions initially went along with national union policies, but in response to membership losses they became increasingly supportive of employer efforts decouple of wage developments in manufacturing and services. As the relative earnings of union members have increased, the ‘pro-redistribution effect’ of electoral mobilization by unions has in turn diminished.

Union membership and preferences for redistribution

  1. Top of page
  2. Abstract
  3. Introduction
  4. Union decline
  5. Levels of unionization, earnings inequality and redistribution
  6. Electoral mobilization
  7. Comparative analysis of within-country change
  8. Unionization by income
  9. Union membership and preferences for redistribution
  10. Final remarks
  11. Acknowledgements
  12. References

Based on self-interest, the account of the implications of (de-)unionization for redistributive politics proposed above needs to be qualified. Recent comparative analyses of individual preferences for redistribution that include union membership as an independent variable find that, controlling for income, survey respondents who identify themselves as union members are significantly more likely to support redistribution than respondents who are not union members (Cecchi et al. 2010; Mosimann and Pontusson 2013; Rueda 2012). Analysing support for redistribution with 2008–2010 data from the ESS, Mosimann and Pontusson (2013) also find a robust positive effect of interacting union membership and income: in other words, the difference in support for redistribution between union members and non-members is biggest among survey respondents with high (household) incomes.

Cecchi et al. (2010) interpret the finding that union member favour egalitarianism and redistribution as a self-selection effect. Simply put, their argument is that individuals with a more egalitarian ideological disposition are more likely to join unions. The obvious alternative interpretation is that the experience of being a union member generates — or sustains — egalitarian attitudes. These interpretations need not be conceived as mutually exclusive. Assuming that individuals tend to conform to the attitudes of other members of social networks or organizations to which they belong, one might plausibly argue that unions nurture egalitarianism precisely because they attract individuals with egalitarian attitudes. To the extent that unions encompass a wide range of the earnings distribution, it might also be argued (following Lupu and Pontusson 2011) that they constitute social networks that generate social affinity with the poor among middle and high-income individuals.

The challenge of parsing between self-selection and socialization lies well beyond the scope of this article, but it is noteworthy that the association between union membership and support for redistribution still holds when the analysis is restricted to countries with Ghent systems of unemployment insurance. According to conventional wisdom, selective incentives to join unions explain high levels of unionization in these countries. By implication, ideological motivations for joining unions should be less prominent, yet it appears that union members are still more supportive of redistribution. To my mind, this constitutes quite compelling evidence that the association between union membership and support for redistribution is not entirely due to self-selection.

Recall the opening question of this article: why do we observe so little effort by democratically elected governments to offset rising inequality through new redistributive initiatives? Union decline deserves to be taken into account in addressing this question. The most obvious way that union decline matters for redistributive politics is that it has contributed to declining voter turnout, especially among poorly paid workers with strong self-interested reasons for supporting redistribution. But there is a second potential mechanism involved here that also deserves to be noted: as unionization among well-paid workers has fallen, other-regarding support for redistribution among well-paid workers has arguably fallen as well. Related to this second mechanism, an important question that I wish to pursue in future research concerns the effects of different types of unions for individual attitudes and preferences. My working hypothesis is that union decline implies not only that well-paid workers (as well as poorly paid workers) are less likely to belong to unions, but also that they are more likely to belong to unions that encompass a relatively narrow range of occupations and incomes (cf. Hassel 2013).

Final remarks

  1. Top of page
  2. Abstract
  3. Introduction
  4. Union decline
  5. Levels of unionization, earnings inequality and redistribution
  6. Electoral mobilization
  7. Comparative analysis of within-country change
  8. Unionization by income
  9. Union membership and preferences for redistribution
  10. Final remarks
  11. Acknowledgements
  12. References

Some readers have responded to earlier drafts of this article by asking me to elaborate on how my interpretation of the contemporary role of trade unions differs from that of ‘right-wing commentators’ who allege that unions represent a privileged labour aristocracy. Let me briefly comply with this request and, in so doing, restate the main points of the preceding discussion.

In contrast to many right-wing union critics, I consider contemporary unions to be, broadly speaking, democratic organizations that represent the preferences of their members. To a large extent, the preferences of union members are self-interested, but this is not entirely the case. We have at least some evidence indicating that being a union member tends to encourage people to be more other-regarding or to conceive of their own interests in broader terms. I do not believe that unions are more ‘selfish’ than other economic interest groups; quite the contrary, my view is that unions have always been and remain associated with solidaristic values of a distinctively universalistic cast.

The aggregate, cross-national evidence presented above indicates that the egalitarian effects of unionization have become weaker over time, but it remains the case that more unionized countries are, on average, characterized by less earnings inequality and more redistribution than less unionized countries. None of the evidence presented above indicates that strong unions have become associated with more inequality.

In the end, right-wing critiques of unions invariably seem to boil down to the claim that these organizations deliver benefits to their members (or officials) at the expense of the common good — or at the expense of unorganized and less privileged workers. I have not made any argument along these lines or presented any evidence supporting such an interpretation. The fundamental point of this article is that any political project aspiring to redress the growth of inequality in the OECD countries ought to promote unionization and, in particular, the unionization of low-wage and precariously employed workers. Existing union organizations have neither the incentives nor the resources to tackle this challenge on their own.

Acknowledgements

  1. Top of page
  2. Abstract
  3. Introduction
  4. Union decline
  5. Levels of unionization, earnings inequality and redistribution
  6. Electoral mobilization
  7. Comparative analysis of within-country change
  8. Unionization by income
  9. Union membership and preferences for redistribution
  10. Final remarks
  11. Acknowledgements
  12. References

This article draws extensively on previous collaborative work with Michael Becher and David Rueda, and ongoing collaborative work with Lucio Baccaro and Nadja Mosimann. It has benefitted greatly from seminar discussions at the University of Geneva, the University of St. Gallen and the Juan March Institute, as well as comments by Michael Donnelly, John Godard, Noam Lupu, Daniel Oesch, Mary O'Sullivan, Line Rennwald, John Stephens and the aforementioned collaborators.

Notes
  1. 1

    In Lindert's words, ‘redistribution from rich to poor is least present when and where it seems to be most needed’.

  2. 2

    For 21 of the 22 countries included in Table 1 (South Korea missing for lack of coverage data), the correlation between percentage changes in union density and bargaining coverage from 1990 to 2007 is 0.712 (p = 0.000). Without Austria, the single biggest outlier (and missing from my analysis of change in earnings inequality), the correlation is 0.776 (p = 0.000). Data source: Visser (2013).

  3. 3

    In the United States, unionization of non-agricultural workers peaked at 35.5 per cent in 1945, and union membership in percent of the total labour force peaked at 25.4 per cent in 1954 according to the Bureau of Labor Statistics (BLS) (cited by Goldfield 1987: 10). The 1960 BLS figure for non-agricultural workers (31.4 per cent) is very close to Visser's figure (30.9 per cent). For lack of comparable data for other countries, my discussion is restricted to the period since 1960. It should also be noted that Table 1 ignores very high figures immediately following the transition to democracy in Portugal and Spain. According to Visser's dataset, union density peaked at 60.8 per cent in Portugal, in 1978, and at 45.1 per cent in Spain, also in 1978. My decision to use 1980 as the baseline for these countries is entirely arbitrary and potentially problematic. In the Spanish case, union density stabilized around 10 per cent in the mid-1980s and then rose to a new peak of 16.7 per cent in 2000. In the Portuguese case, by contrast, we observe a continuous decline of union density from the late 1970s onwards. Note that none of the empirical generalizations made in this article depend on Portuguese or Spanish data.

  4. 4

    Lange and Scruggs (2002) convincingly demonstrate the absence of any direct linkage between globalization and union decline.

  5. 5

    Visser's dataset also includes some observations of union density in other service sectors, such transportation, finance and hotel and restaurants (as well as construction). Provided that we knew the employment share of each sub-sector, we could potentially use these data to generate overall estimates of union density in private services, but the data are too fragmentary, and Visser's documentation (for the sector-specific part of his dataset) is not sufficiently detailed to make such an exercise worthwhile. Note that the figures on union density in hotels and restaurants reported by Visser are generally quite close to his figures for union density in commerce: thus, it seems reasonable to conceive union density in commerce as a rough proxy for union density in low-end private services.

  6. 6

    This expectation is borne out by Lee's (2005) analysis of pooled cross-section time-series data.

  7. 7

    The argument about economies of scale is deployed by Wallerstein (1989) to explain why union density tends to be higher in smaller economies. See Pontusson (1995) and Bartolini (2000) for further discussion of the implications of plant size for union strength and working-class support for Left parties.

  8. 8

    The argument about atypical employment invites concerns about reverse causality, for the expansion atypical employment might, of course, be attributed to labour market reforms made possible by the declining power of unions. The same holds for reforms of unemployment insurance.

  9. 9

    The 90-10 ratio is the ratio of the earnings of someone in the 90th percentile of the earnings distribution to the earnings of someone in the 10th percentile. The OECD data for France refer to net rather than gross earnings (i.e. after-tax earnings): therefore, I do not include France in this analysis.

  10. 10

    Because Figure 2 includes more countries than Figure 1, the bivariate regression results cannot be directly compared, but the point about increased dispersion holds if the analysis of 2009 data is restricted to the 10 countries included in Figure 1 (coefficient = −0.020, t = −2.52, R2 = 0.443).

  11. 11

    In contrast to some of the existing literature (including Kenworthy and Pontusson 2005), the measure used here includes the elderly population. This allows me to include more countries and look at data for a single year.

  12. 12

    Bradley et al. (2003) demonstrate that unionization is a significant determinant of ‘pre-fisc’ income inequality among working-age households.

  13. 13

    Elsewhere, I have argued that unionization and voter turnout condition the extent to which left parties differentiate themselves from other parties (Kwon and Pontusson 2010; Rueda and Pontusson 2010). More important for present purposes, I want to acknowledge that Barnes' (2013) careful comparative analysis of state-level voter turnout and welfare spending in the United States runs counter to the argument suggested here. Barnes finds no consistent effect of turnout on welfare spending and also shows that the distance between the median voter's income and the mean income is not very different from the distance between the median and the mean income. There are two main reasons why her results do not necessarily call into question the cross-national argument. First, federal transfers finance a large part of welfare spending by US states, which arguably weakens the link between turnout and spending. Second, voter turnout across US states ranges between 33 and 56 per cent (2010 midterm elections) as compared with the cross-national range of 46–86 per cent shown in Figure 5. As Barnes notes, the correlation between aggregate voter turnout and income bias in turnout may be weaker at lower levels of aggregate turnout.

  14. 14

    While Mosimann and Pontusson (2013) fail to find a significant effect of interacting union membership and income, Kerrissey and Schofer's (2013) report a significant effect of interacting union membership with education: it is primarily among US survey respondents with low levels of education that union membership is associated with higher turnout in elections. On union membership as a variable that conditions class voting, see Oesch (2006) and Rennwald (2013).

  15. 15

    Australia, Belgium and Greece have been dropped from Figure 5 on account of mandatory voting (in the Greek case, until the early 2000s). The French and US turnout figures include both legislative and presidential elections.

  16. 16

    Note that for Japan, Figure 6 measures the change in voter turnout to the 2003 election rather than the 2005 election (the one and only exception to the ‘most recent pre-2008 election’ rule). According to the Armingeon dataset, Japanese voter turnout was 59 per cent in 1996, 62.5 per cent in 2000 and 59.9 per cent in 2003, then jumped to 67.5 per cent in 2005. The 2003 election appears to be more representative of ‘pre-crisis conditions’ than the 2005 election.

  17. 17

    Franklin's (2004) account of declining voter turnout emphasizes changing dynamics of electoral competition while downplaying changing characteristics of voters. Franklin repeatedly asserts that unionization does not add much, if any, explanatory power to his empirical models, but the analysis behind these assertions is not entirely transparent. Curiously, Franklin states (p. 18, footnote 10) that ‘outside the United States, union density has declined less — by under 2 per cent on average across eighteen advanced democracies … so this is not likely to be a variable that could account for widespread turnout decline’.

  18. 18

    Again, France is not included in Figures 7-9 because the OECD data for France refers to net earnings. If France were included, it would be a major outlier in all three figures, characterized, much like Japan, by a dramatic decline in union density combined with stability or decline of earnings inequality. All existing studies seem to agree that minimum-wage legislation is the main reason for the significant compression French wage differentials that occurred between 1980 and 2000 (see, e.g., Charnoz et al. 2011).

  19. 19

    The analysis includes 20 of the 22 countries listed in Table 1: Greece and South Korea are missing for lack of data on redistribution in the 1980s.

  20. 20

    In contrast to Visser's, our measure of union density includes the unemployed (in the numerator as well as the denominator), but this makes very little differences for purposes of cross-national comparison: the correlation between our ESS-based estimates of aggregate union density and Visser's net density figures for 2008 is 0.971 (p = 0.000).

  21. 21

    Together with Lucio Baccaro and Ruya Kocer, I have begun to explore using data from World Values Surveys and the International Social Survey Program to track changes in the composition of union membership since the mid-1980s.

  22. 22

    Lucio Baccaro kindly helped me produce Figure 10. We are currently working on a paper that seeks to explain the evolution of inter-sectoral earnings differentials since 1970 (see Baccaro and Pontusson 2013).

  23. 23

    Needless to say, the 16-country averages presented in Figure 10 hide a good deal of cross-national variation, but this variation primarily pertains to the ability of public-sector workers to keep up with manufacturing wages. Ireland is the only case in which low-end services did not fall behind in a big way between 1990 and 2007.

References

  1. Top of page
  2. Abstract
  3. Introduction
  4. Union decline
  5. Levels of unionization, earnings inequality and redistribution
  6. Electoral mobilization
  7. Comparative analysis of within-country change
  8. Unionization by income
  9. Union membership and preferences for redistribution
  10. Final remarks
  11. Acknowledgements
  12. References
  • Armingeon, K., Weisstanner, D., Engler, S., Potolidis, P. and Gerber, M. (2012). ‘Comparative Political Data Set I 1960–2010’. http://www.ipw.unibe.ch/content/team/klaus_armingeon/comparative_political_data_sets/index_ger.html
  • Baccaro, L. (2011). ‘Labor, globalization and inequality’. In D. Brady (ed.), Research in the Sociology of Work, vol. 22B. London: Emerald Press, pp. 213285.
  • Baccaro, L. and Pontusson, J. (2013). ‘Coordinated Wage Bargaining and Inter-Sectoral Differentials in a Baumolian World’. Paper presented to the annual conference of the Society for the Advancement of Socio-Economics, Milan, 27–29 June.
  • Barnes, L. (2013). ‘Does median voter income matter?’. Political Studies, 61 (1): 82100.
  • Bartolini, S. (2000). The Political Mobilization of the European Left, 1860–1980. Cambridge: Cambridge University Press.
  • Becher, M. and Pontusson, J. (2011). ‘Whose interests do unions represent?’. In D. Brady (ed.), Research in the Sociology of Work, vol. 22B. London: Emerald Press, pp. 181211.
  • Böckerman, P. and Uusitalo, R. (2006). ‘Erosion of the Ghent System and union membership decline’. British Journal of Industrial Relations, 44 (2): 283303.
  • Bradley, D., Huber, E., Moller, S., Nielsen, F. and Stephens, J. (2003). ‘Distribution and redistribution in postindustrial democracies’. World Politics, 55: 193228.
  • Cecchi, D., Visser, J. and van de Werfhorst, H. G. (2010). ‘Inequality and union membership’. British Journal of Industrial Relations, 48: 84108.
  • Charnoz, P., Coudin, E. and Gaini, M. (2011). ‘Decreasing Wage Inequality in France 1976–2004’. INSEE working paper.
  • Ebbinghaus, B., Göbel, C. and Koos, S. (2011). ‘Social capital, “Ghent” and workplace context matter’. European Journal of Industrial Relations, 17: 107124.
  • Emmenegger, P. , Häusermann, S. , Palier, B. and Seeleib-Kaiser, M. (eds.) (2012). The Age of Dualization. Oxford: Oxford University Press.
  • Finseeras, H. (2009). ‘Income inequality and demand for redistribution’. Scandinavian Political Studies, 32 (1): 94119.
  • Franklin, M. (2004). Voter Turnout and the Dynamics of Electoral Competition in Established Democracies since 1945. Cambridge: Cambridge University Press.
  • Freeman, R. (1980). ‘Unionism and the dispersion of wages’. Industrial and Labor Relations Review, 34: 323.
  • Golden, M. and Wallerstein, M. (2011). ‘Domestic and international causes of the rise of pay inequality in OECD nations between 1980 and 2000’. In D. Brady (ed.), Research in the Sociology of Work, vol. 22A. London: Emerald Press, pp. 209249.
  • Goldfield, M. (1987). The Decline of Organized Labor in the United States. Chicago, IL: University of Chicago Press.
  • Hassel, A. (2013). ‘Trade Unions and the Future of Democratic Capitalism’. Paper presented to the 20th International Conference of Europeanists, Amsterdam, 25–27 June.
  • Iversen, T. and Soskice, D. (2006). ‘Electoral institutions and politics of coalitions’. American Political Science Review, 100 (2): 165181.
  • Iversen, T. and Soskice, D. (2009). ‘Distribution and redistribution’. World Politics, 61 (3): 438486.
  • Kenworthy, L. and Pontusson, J. (2005). ‘Rising inequality and the politics of redistribution in affluent countries’. Perspectives on Politics, 3: 449471.
  • Kerrissey, J. and Schofer, E. (2013). ‘Union membership and political participation in the US’. Social Forces, 91 (3): 895928.
  • King, D. and Rueda, D. (2008). ‘Cheap labor’. Perspectives on Politics, 6: 279297.
  • Kjellberg, A. (2011). ‘The decline in Swedish union density since 2007’. Nordic Journal of Working Life Studies, 1: 6793.
  • Korpi, W. (1983). The Democratic Class Struggle. London: Routledge and Kegan Paul.
  • Korpi, W. (2006). ‘Power resources and employer-centered approaches in explanations of welfare states and varieties of capitalism’. World Politics, 58: 167206.
  • Kwon, H. Y. and Pontusson, J. (2010). ‘Globalization, labour power and partisan politics revisited’. Socio-Economic Review, 8 (2): 251281.
  • Lange, P. and Scruggs, L. (2002). ‘Where have all the members gone?’. Journal of Politics, 64: 126153.
  • Lee, C.-S. (2005). ‘International migration, deindustrialization and union decline in 16 affluent countries’. Social Forces, 84 (1): 7188.
  • Lindert, P. (2004). Growing Public. Cambridge: Cambridge University Press.
  • Lupu, N. and Pontusson, J. (2011). ‘The structure of inequality and the politics of redistribution’. American Political Science Review, 105 (2): 316336.
  • Mahler, V. (2008). ‘Electoral turnout and income redistribution by the state’. European Journal of Political Research, 47: 161183.
  • Mosimann, N. and Pontusson, J. (2013). ‘Union Membership: Electoral Mobilization and Preferences for Redistribution’. Paper presented to the 20th International Conference of Europeanists, Amsterdam, 25–27 June.
  • OECD (2011). Divided We Stand. Paris: OECD Publishing.
  • OECD (2013). ‘Distribution of Gross Earnings of Full-Time Employees’. http://www.oecd.org/employment/employmentpoliciesanddata/onlineoecdemploymentdatabase.htm
  • Oesch, D. (2006). Redrawing the Class Map. New York: Palgrave Macmillan.
  • Pinto, S. and Beckfield, J. (2011). ‘Organized labor in European countries, 1960–2006’. In D. Brady (ed.), Research in the Sociology of Work, vol. 22B. London: Emerald Press, pp. 153179.
  • Pontusson, J. (1995). ‘Explaining the decline of European social democracy’. World Politics, 47 (4): 495533.
  • Pontusson, J., Rueda, D. and Way, C. (2002). ‘Comparative political economy of wage distribution’. British Journal of Political Science, 32: 281303.
  • Radcliff, B. and Davis, P. (2000). ‘Labor organization and electoral participation in industrial democracies’. American Journal of Political Science, 44 (1): 132141.
  • Rennwald, L. (2013). ‘Partis socialistes et classe ouvrière’. PhD thesis, Department of Political Science, University of Geneva.
  • Rueda, D. (2007). Social Democracy Inside Out. Oxford: Oxford University Press.
  • Rueda, D. (2012). ‘Food Comes First, Then Morals?’ Paper presented at conference on ‘The Political Economy of Skills and Inequality,’ University of Konstanz, 26–27 July.
  • Rueda, D. and Pontusson, J. (2000). ‘Wage inequality and varieties of capitalism’. World Politics, 52 (April): 350383.
  • Rueda, D. and Pontusson, J. (2010). ‘The politics of inequality’. Comparative Political Studies, 43 (6): 675705.
  • Schäfer, A. (2011). ‘Republican Liberty and Compulsory Voting’. Max-Planck Institute for the Study of Societies, Cologne, Discussion Paper 11/17.
  • Schäfer, A. and Streeck, W. (2013). ‘Introduction: politics in the age of austerity’. In A. Schäfer and W. Streeck (eds.), Politics in the Age of Austerity. London: Polity Press, pp. 125.
  • Scheve, K. and Stasavage, D. (2009). ‘Institutions, partisanship and inequality in the long run’. World Politics, 61 (2): 215253.
  • Schnabel, C. (2012). ‘Union Membership and Density’. Institute for the Study of Labor (IZA), Bonn, Discussion Paper No. 6792.
  • Scruggs, L. (2002). ‘The Ghent system and union membership in Europe, 1970–1996’. Political Studies Quarterly, 55: 275297.
  • Solt, F. (2011). ‘Standardized World Income Inequality Database’. version 3.1, http://myweb.uiowa.edu/fsolt/swiid/swiid.html
  • Visser, J. (2013). ‘Database on Institutional Characteristics of Trade Unions, Wage Setting, State Intervention and Social Pacts’, version 4.0, http://www.uva-aias.net/207
  • Wallerstein, M. (1989). ‘Union organization in advanced industrial democracies’. American Political Science Review, 83: 481501.
  • Western, B. (1999). Between Class and Market. Princeton, NJ: Princeton University Press.