Unions, Industrial Relations and Market Income Inequality in Canada’s Provinces

Scrimger, Phillippe

doi:https://doi.org/10.7202/1070351ar

Introduction

The last few decades of scholarship have produced broad consensus that unions compress the distribution of wages and incomes (Ahlquist, 2017). While recent cross-national studies continue to support this conclusion (Tridico, 2018), growing evidence suggests that unions’ within-country egalitarian effect is dwindling, as unionization declines and membership composition changes (Baccaro, 2011; Farber et al., 2018; Pontusson, 2013).

Declining economic growth, globalization, financialization, fiscal austerity and the rise of pro-capital right-wing coalitions are increasingly limiting unions’ effectiveness as a countervailing power to the forces that shape inequality (Baccaro, 2011; Baccaro and Howell, 2011; Jacobs and Myers, 2014; Peters, 2011; 2012; Pontusson, 2013; Streeck, 2014a; 2014b; 2016). These forces have produced more precarious and insecure forms of work, contributing to the segmentation of labour markets and ensuing division within the working class (Cranford, Vosko and Zukewich, 2003; Fudge, 2017; Stone and Arthurs, 2014). Meanwhile, within the new political economy, unions everywhere are often expected to do more to fight inequality by rebuilding the broad solidarities of old (Crouch, 2017; Kelly, 2015a).

This article examines these debates within the Canadian provincial context by exploring the relationship between rising market income inequality (Heisz, 2016) and falling union density rates (Legree, Schirle and Skuterud, 2017). Considering industrial relations (IR) systems more broadly, the article also looks at how provincial variations in work stoppages and collective labour rights are tied to market income inequality. The study seeks to answer the question: do unions and industrial relations still matter for reducing market income inequality? The answer is provided through interpreting a set of quantitative analyses of macro-level provincial data from the mid-1980s to the early 2010s.

The role unions and IR play in shaping market income inequality have been under-investigated at the Canadian provincial level. Studies have showed how wage solidarity and standardisation resulting from collective bargaining compress the distribution of wages within Canadian labour markets (Freeman and Medoff, 1984; Card, Lemieux and Riddell, 2004; Fortin, Green and Lemieux, 2012). Beyond this within-sector wage equalizing effect of collective bargaining, however, the broader impact of unions and IR institutions on provincial inequality is less well understood. Unions can affect the distribution of wages and income more broadly by influencing the rules that govern labour, commodity and financial markets (Kelly, 2008; Stiglitz, 2015). They can also produce egalitarian outcomes by institutionalizing norms of equity throughout the labour market (Western and Rosenfeld, 2011). Studies that have looked at unions’ broader distributive impact in the provinces have produced evidence of a broader inequality-reducing union effect (Breau, 2007; Cousineau and Merizzi, 2015). However, these studies often undertheorize and under-examine this relationship. Studies that have put unions at the centre of the analysis have focused on organized labour’s role in the politics of redistribution (Haddow, 2013; 2014). Attempting to fill this gap in the literature, this article aims to contribute to the body of knowledge on unions’ distributive effect and to a better understanding of the forces that shape inequality in Canada’s provinces.

The article is divided as follows. It starts by exploring provincial-level data on union density, IR systems and market income inequality. This is followed by theoretical discussion of how these trends may relate. After refining theoretical expectations, the article then provides an evaluation of key relationships through a multivariate analysis of provincial panel data from 1984 to 2012. Results are then discussed and summarized in the conclusion.

The weakening of unions and industrial relations

An extensive literature documents the causes of union decline. Analysts who emphasize the exogenous causes typically cite contributing factors such as deindustrialization (Kochan, 2012); open employer opposition (Freeman, 2005); anti-union political rhetoric and institutional deregulation (Baccaro and Howell, 2011; Jacobs and Myers, 2014); and changes to the distribution of power between workers and employers resulting from financialization, globalization and fiscal austerity (Darcillon, 2015; Kollmeyer and Peters, 2018; Peters, 2011; 2012; Vachon et al., 2016). Those who focus on the endogenous sources of decline generally highlight the misapprehension and under-exploitation of unions’ own power resources (Lévesque and Murray, 2010); unions’ inability to represent the interests of a diverse workforce and to construct a shared identity (Culpepper and Regan, 2014; Dufour and Hege, 2010); their incapacity to promote a convincing progressive political alternative to neoliberalism (Gumbrell- McCormick and Hyman, 2013); and their own integration of neoliberal strategies (Macdonald, 2014). Whatever the true story behind decline may be, the malaise likely combines multiple factors lying both within and outside unions. Looking at three dimensions of provincial industrial relations systems—union density, strike activity and collective labour statutes—this section explores whether the decline narrative fits provincial-level data.

Union density

Figure 1 shows provincial trends and variations in union density (union members as a proportion of all employees) from 1984 to 2012. The aggregate time trend (Figure 1.1) indicates a general decline in unionization starting in the 1990s[1], which is consistent with national- and provincial-level evidence produced elsewhere (Galarneau and Sohn, 2013; Legree, Schirle and Skuterud, 2017). However, as Figure 1.2 shows, there is a significant amount of variation across provinces. Newfoundland and Labrador, British Columbia, New Brunswick and Alberta experienced sharp declines over the period covered. In contrast, unionization slightly increased in Prince Edward Island and only slightly declined in Quebec, Saskatchewan and Manitoba. As for Nova Scotia and Ontario, they can be best described as having experienced average deunionization. Overall, data suggest that the decline narrative fits the majority of cases.

Citing the importance of union incorporation in governing institutions and the crucial role of strikes and political protest, some have criticized the use of union density as a primary proxy of union power (Sullivan, 2010; Crouch, 2017). However, much of these arguments are based on observations from Western Europe, where unions arguably play a much bigger role in the political economy, regardless of effective membership numbers. For example, low density in France masks unions’ great capacity to mobilize the broader workforce for support when necessary (Crouch, 2017: 147). However, when comparing subnational units (the provinces) that are absent of high-level deliberative institutions and large coalitional dynamics, changes and differences in union density are likely to better reflect varying levels in union power.

Strike activity

Figure 2 presents trends in strikes and lockout.[2] Estimates of the preferred measure in comparative work on industrial conflict—person-days lost due to work stoppages per 1000 employees—are plotted against time and compared across provinces. The days-lost measure is a relative indicator sensitive to both the size (number of workers involved) and duration of strikes and lockouts. Figure 2.1 indicates an average downward trend in work stoppages. As the large spikes in its evolution suggests, the days-lost measure is moderately responsive to singular large conflicts such as the 1986 International Woodworkers of America (IWA) strike in British Colombia, and the 2004 public sector strike in Newfoundland and Labrador by the Newfoundland and Labrador Association of Public and Private Employees (NAPE) and the Canadian Union of Public Employees (CUPE). While Figure 2.2 suggests some variation between the provinces, especially earlier in the period covered, the days-lost measure marks a stark downward convergence of conflict levels over time. Newfoundland and Labrador’s persistent distinctiveness in the later part of the period is largely attributable to the 2004 public-sector strike mentioned above.

Figure 2

**Trends and variations in provincial strikes and lockouts, 1984-2012**

The decline in provincial work stoppages mirrors international trends (Godard, 2011; Kelly, 2015b; Vandeale, 2016). However, strikes remain an important tool in organized labour’s repertoire of action; even if environmental forces have constrained its usage and unions have become more cautious and calculative in its exercise (Hennebert and Faulkner, 2017). Further, vigorous union campaigns that include strike action can have a positive impact on the recruitment and retention of members (Hodder et al., 2017), suggesting that militancy remains a source of union power.

At the very least, estimates in Figure 2 show a fundamental change in provincial IR systems and a major shift in the disruptive tactics of unions. They also suggest a displacement in the expression of economic conflict towards less observable manifestations (Godard, 2011).

Collective labour statutes

The liberalization of collective bargaining institutions through deregulation and decentralization in Western Europe is well documented. While the erosion of industrial relations systems takes varying forms across regions, the main dimension of change has been a push towards greater employer discretion in the establishment of working conditions (Baccaro and Howell, 2011; Marginson, 2015). With predominantly uncoordinated and decentralized firm-level collective bargaining models, there was much less room for a corrosion of IR institutions in Canada’s provinces.

Figure 3 plots the Labour Relations Index (LRI) constructed by Legree, Schirle and Skuterud (2014; 2017). Reflecting three broad dimension of collective labour rights—union recognition, first-contract negotiations and strikes—the LRI indicates how favourable legal statutes are to union activity.[3] As Figure 3.1 suggests, the LRI index increased earlier in the period observed, then generally declined from the mid-1990s onward. The sudden drop observed between 1995 and 1997 was mostly driven by the enactment of statutes forcing secret-ballot certification and impeding the use of strikes in Ontario and Manitoba. Nonetheless, it appears that legal statutes have remained relatively stable in the provinces over the full period. In fact, comparing 1984 to 2012 values in Figure 3.2 indicates that laws governing labour relations have become slightly more favourable to unions in half the provinces, and have remained stable in Nova Scotia. However, the index did fall moderately in Quebec, British Columbia, Ontario, and declined markedly in Alberta.

Figure 3

**Trends and variations in provincial labour statutes, 1984-2012**

On the whole, this statistical portrait of the state of unions and industrial relations in Canada’s provinces broadly supports the decline narrative, especially when it comes to unionization rates and strike activity. The LRI indicates institutional stability in some regions, but this should not necessarily be taken as reflecting union strength. While the legislative context has been relatively stable, private sector unions are nonetheless making concessions in the form of wage reductions, heightened scheduling variability, less generous benefits, and the growing acceptance of non-union workers within their establishments (O’Brady, 2018). As for the public sector, changing managerial strategies and fiscal austerity have resulted in cuts to employment and compensation (Peters, 2012). Moreover, the stability of provincial IR legal statutes conveyed by the LRI index masks continuing issues with the general effectiveness of Wagnerism as an organizing model, especially for precarious workers (McCrystal, 2014; Rolland, 2017; Vosko, 2014).

The rise of market income inequality

This section focuses on describing trends and variations in inequality in the provinces over the same period unions and IR systems were examined above. Four measures of inequality are used to track changes in different areas of the distribution: the Gini coefficient, two deciles ratios (D9:D5 and D5:D2) and the share of income held by the top 1% of earners. The Gini coefficient is a good overall measure of inequality, but it is known to be especially sensitive to change in the middle of the distribution (Atkinson, 1970). The D9:D5 ratio measures how well the top of the distribution does relative to the middle. The D5:D2 ratio measures how well the middle of the distribution does relative to the bottom.[4] Finally, the share of total income held by the top 1% measures changes at the tail end of the distribution. Estimates for each measure are taken directly or compiled from Statistics Canada’s online database formerly known as the Canadian Socioeconomic database (CANSIM). Each of these measures are calculated using working age population adjusted market incomes.[5]

Figures 4 through 7 show the broad evolution of market income inequality in the provinces. To locate distinctive individual patterns, each figure includes yearly data points from each province. Generally, the figures show an overall increase in inequality, corroborating results in Heisz (2016). The only exception is the D5:D2 measure presented in Figure 5, which increased in the early 1990s, but then slowly returned to levels comparable to those of the mid-1980s. The figures also show that most of the sustained growth in average market income inequality, as measured by the indicators, happened prior to the 2000s.

Figure 4

**Gini coefficient, 1984-2012**

Figure 7

**Top 1% income share, 1984-2012**

As for notable individual cases, Newfoundland and Labrador (NL) is distinctive, showing much higher levels of inequality as measured by the Gini coefficient, D5:D2 ratio, and the D9:D5 ratio. This likely has much to do with NL’s comparatively high unemployment rate and seasonal economy. Often considered Canada’s most liberal province, it is perhaps surprising to find that Alberta shows relatively low levels of market income inequality, with the exception of income concentration at the very top (Figure 7), for where it competes with Ontario for highest levels among all provinces. At least in part, Alberta’s low levels of inequality is well explained by booms in the extractive resources sector, which compress the distribution from the bottom by having a positive effect on the wages of younger, less-educated workers (Fortin and Lemieux, 2016). However, as market income does not take into account taxes and social transfers, this statistical portrait overestimates Alberta’s performance. If one looks at income inequality after state redistribution, Alberta performs substantially differently (see Haddow, 2013; 2014).

The general decline in unionization described earlier and the broad rise in inequality portrayed above provide coincidental evidence of a substantive negative relationship between the two developments. Before proceeding to a more robust evaluation of this relationship, relevant theory is discussed to set formal expectations on its nature.

Theory: linking unions and IR to inequality

The following discussion explores how unions function to abate inequality. It starts by considering channels through which unions reduce inequality and then examines mechanisms by which they sometimes increase it. In preparation for the multivariate analysis to come, other drivers of inequality are also briefly discussed.

How unions reduce inequality

Most comparative studies of the economic distributive effect of unions are grounded in power resources theory (PRT). PRT contends that the balance of power between labour and capital determines the allocation of material resources in capitalist democracies (Kelly, 2008; Korpi, 1998; 2006). Typically using cross-national comparative research designs, analysts often test and find supportive evidence of PRT’s basic narrative: that when and where broad coalitions exist between powerful unions and labour-friendly social-democratic political parties, one usually finds more equal societies. Unions play two key roles in creating and sustaining these broad egalitarian coalitions. First, through collective bargaining, they act as a vehicle for the combination and actualization of individual labour power resources, enabling workers to extract bigger rents from employers. Second, outside the employment relationship, unions act as working-class representatives more generally and, through political mobilization, strive to reshape power structures by influencing the rules that govern labour, commodity and financial markets.

Economic theory (Freeman, 1980; Freeman and Medoff, 1984; Card, Fortin and Riddell, 2004; Fortin, Green and Lemieux, 2012) provides comprehensive explanations about how collective bargaining reduces inequality. First, as uniformity takes wages out of competition, unions strive to standardize wage rates of comparable unionized workers across establishments of a given industry. Second, within establishments, unions tend to raise wages disproportionally at the bottom of the income distribution. As unions are democratic organizations, it is expected that the majority of members would not allow wages to become concentrated in the hands of a few unionists and that those located below the mean wage would favour union wage policies guaranteeing greater gains for lower-earners (Freeman, 1980). Moreover, great wage disparities between members of a single union would likely harm the union’s cohesion and thus its strength. Beyond standardizing and compressing wages within the unionized sector, high levels of unionization condition employers to raise non-union wages so as to avoid unionization altogether, a phenomenon known as the threat effect.

As for the role of economic strikes, they represent an important source of union power which can be harnessed internally (Sullivan, 2010). Along with union presence, strikes are particularly effective in reining-in top-income earners through their impact on market values and executive pay (Gomez and Tzioumis, 2006; DiNardo and Hallock 2002, Lee and Mas, 2012). However, some evidence suggests that union decline has made strikes increasingly rare (as shown earlier) and less effective (Rosenfeld, 2006; 2014).

Unions also play a broader role in the political economy as working-class representatives. Sociologists have argued and provided evidence that unions reduce inequality by contributing to a moral economy that institutionalizes norms of fairness and equity (Western and Rosenfeld, 2011). Beyond facilitating organizing and recruiting activities, strong institutional underpinnings allow unions to play this wider role. If institutions are defined as more or less stable compromises reflecting coalitional power dynamics (Mahoney and Thelen, 2010), it follows that the quality of collective labour statutes may reflect the porosity of civil society and government to union political rhetoric. Institutional underpinnings act as a platform for unions’ social and political influence, providing them with the legitimacy to play a wider role in civil society (Rigby and Garcia Calavia, 2018), which includes supporting broader social movements aimed at the betterment of minimum wages and other universal basic labour standards.

How unions increase inequality

Some genres of, mostly neoclassical, economic theory propose that union monopolies increase inequality through a spillover effect. As certified unions gain monopolistic control over the labour supply, they raise wages in unionized firms. As wages go up in the unionized sector, demand for labour falls. This creates a spillover effect putting downward pressure on non-union wages. If this between-sector effect is larger than the within-sector standardization and compression effect described above, unionization can lead to increased levels of overall inequality.

Evidence from Canada and the United States from the last few decades suggests that unions, on balance, reduce inequality (Card, Fortin and Riddell, 2004; Farber et al., 2018; Freeman and Medoff, 1984; Western and Rosenfeld, 2011). However, as unions decline and membership diminishes, the inequality-reducing effect of unionism appears to be dwindling. Farber et al. (2018) show that most of unions’ egalitarian impacts in the US were felt in the middle decades of the 20th century when unionization was at its peak and unions represented much higher proportions of low-wage unskilled workers. As union density fell and the skill level of membership increased, unions’ equalizing effect substantially diminished. International studies are increasingly pointing in the same direction; that unionization is no longer significantly associated with within-country variations in inequality (Baccaro, 2011; Pontusson, 2013). While Baccaro (2011) argues that the increasingly stringent structural constraints under which unions operate has made them ineffectual, Pontusson (2013) points to the possibility that increasingly smaller and better-off unions have become less supportive of broad wage solidarity. As for the provinces, Legree, Schirle and Skuterud (2016) argued that increasing unionization under the current regime of industrial relations would likely do little to reduce inequality as membership gains would be limited to union-saturated well-off segments of the distribution.

From these discussions and given the general scholarly consensus that unions reduce inequality (Ahlquist, 2017), we derive a basic hypothesis positing that higher levels of unionization, strike activity and more union-friendly collective labour statutes are associated with lower levels of income inequality. However, hypotheses presented and tested in this study will explore the possibility that the nature of these relationships is changing.

Other drivers of inequality

A literature on the rise of inequality indicates that there are drivers of income disparities. Three of the most oft-cited determinants—globalization, financialization and technological change—are briefly discussed here. Beyond their impact on inequality, these forces also affect unions directly and indirectly, as was discussed in the overview of the decline literature previously discussed.

First, international trade theory suggests that globalization increases inequality within advanced economies by reducing the demand and wages of unskilled labour whilst having the opposite effect on skilled wage earners (Freeman, 2009). Moreover, the increased mobility of capital in the digital age favours employer bargaining power as relocation threats become more plausible. This trend is inclined to make governments more responsive to the policy and regulatory preferences of capital (Berger, 2000). The consequences of international competition for investments have been the transformation of employment regimes towards non-standard, precarious and insecure forms of work (Cranford, Vosko and Zukewich, 2003; Fudge, 2017; Stone and Arthurs, 2014). At the same time, some argue that tax cuts aimed at increasing investments and growth reduced state revenue and produced public deficits, for which the response has mostly been fiscal austerity (Peters, 2012; Streeck, 2014a).

Second, financialization is said to increase inequality. Financialization detaches the interests of management from those of shareholders through emphasizing short-term performance rather than longer-term sustainability (Palley, 2007; Peters, 2011). Rather than investing profits in research and development or fixed capital—the basis of long-term profitability, real growth and higher employment and wages—firms increasingly prefer to act upon short-term strategies measured using a narrow range of financial indicators. Financialization also allows non-financial firms to decouple the generation of surplus from production as increasing amounts of earnings are generated through auxiliary financial institutions, strengthening employers’ power relative to non-financial workers (Lin and Tomaskovic-Devey, 2013). Finally, the growth of the high rent-seeking financial industry may contribute to increasing inequality on its own. Indeed, the rise in financial elites’ compensation puts downward pressure on wages in non-financial sectors and increases the demand for low-wage service industries catering to the needs of those in the financial sector (Hyde, Vachon and Wallace, 2017).

Third, the theory of skill-biased technological change suggests that the direction of technological change in the production of goods and services, such as the introduction of new information and communication technology, increases the demand for and wages of skilled workers, leaving unskilled workers behind (Acumoglu, 2002; Violante, 2008). However, technological change should not be seen as a purely exogenous process. States can influence the direction, nature and distributive consequences of technological change through public funding of research initiatives and the procurement objectives of government (Atkinson, 2015). Moreover, Green and Townsend (2013) argue that states can endogenously dictate the direction and distributive impact of technological change by favouring the development of certain skills. They show that policy aimed at increasing the human capital of unskilled workers in Canada have resulted in a decrease of both skilled and unskilled workers’ wages.

Results

Before turning to the multivariate analysis, variables, data sources and the modeling strategy are presented. This is followed by a presentation of results and a discussion of their meaning.

Variables, data, and modeling strategy

Variables

The outcome variable, market income inequality, is operationalized into the same four measures used in the earlier trend analysis: the Gini coefficient, two deciles ratios (D9:D5 and D5:D2) and the share of market income held by the top 1% of earners. These measures, sensitive to changes in different areas of the distribution, are calculated using adjusted market income estimates of working-age persons. Changes in union power and industrial relations systems are measured by the same indicators as before: union density, person-days lost due to strikes and lockouts per 1000 employees, and the labour relations index.

Three sets of control variables are included in the analysis. First, following PRT, political partisanship in the provinces is accounted for by the inclusion of two dummy variables for left-party and centre-party incumbency, a common method used in interprovincial studies (Haddow, 2013; 2014; 2016; Noel and Deault Picard, 2015; Roy and Boychuk, 2016). No dummy variable is included for right-party incumbency to avoid the dummy variable trap, meaning that right-wing parties act as the reference category. Following Haddow (2014), the New Democratic Party and the Parti Québécois are coded as the political left, the Liberal Party as the centre, and the Progressive Conservatives as the political right. The British Columbia Social Credit Party and the Saskatchewan Party are both classified as the political right. While some have coded the British Columbia Liberal Party as the political right (Noel and Deault Picard, 2015; Roy and Boychuk, 2016), tests with both options here showed no great difference in outcomes.

Second, the models below take into account other well-known drivers of inequality. A measure of international trade (exports + imports share of provincial GDP) is used to control for globalization. As in Van Arnum et al. (2013), the share of provincial GDP generated by the finance, insurance and real estate (FIRE) sector is used to control for financialization. Estimates of investments in software, research and development, and computer and electronic products as a share of total provincial non-residential investments are used to measure technological change.

Third, a set of standard control variables is added to account for the economic context of each province. These variables include estimates of provincial GDP per capita, employment rates and the share of GDP generated by the extractive sector.

Data

Most variables are constructed using survey estimates from Statistics Canada’s publicly accessible Canadian socioeconomic database (CANSIM).[6] CANSIM estimates rely on data from various government surveys such as the Labour Force Survey (LFS) and the Survey of Labour and Income Dynamics (SLID). Many studies similar to this one have used the CANSIM database (see Breau, 2007; Cousineau and Merizzi, 2015; Haddow, 2013; 2014; 2016). Data for the political partisanship variables are taken from the Canadian Parliamentary Guide and estimates of the LRI were graciously provided by Legree, Schirle, and Skuterud (2014).

Collaging all variable estimates allows for the construction of a perfectly balanced panel data set spreading from 1984 to 2012. With data from 10 provinces over 29 years, the total sample reaches 290 observations. While data for some variables are available for longer time frames, key variables limit the sample study at both time extremities.

Modeling strategy

Multiple tests are applied to select the preferred modeling strategy. Hausman tests show that a fixed-effects (FE) modeling strategy is more efficient than a random-effects approach. FE models provide net within estimates describing the relationship between predictors and outcome variables as they partial-out all time-invariant unobserved differences between provinces (Torres-Reyna, 2013). As tests suggest that unexpected yearly variations or special events are signifi-cantly affecting outcome variables, time fixed-effects are also added to each model. Moreover, the modified Wald tests suggests the presence of groupwise heteroscedasticity; the Breusch-Pagan Lagrange multiplier test of independence indicates that residuals across provincial panels are correlated; and the Lagrange multiplier test additionally detects the presence of serial correlation in all models. To deal with these issues, fixed-effects models are estimated with panel-corrected standard errors (PCSEs) and first-order autocorrelation (AR1).

Results

Table 1 presents regression results including all predictors—union and IR variables, other drivers of inequality and economic controls. Each model uses a different measure of inequality. The presentation of results starts with providing an answer to whether unions and IR still matter for market income inequality. Significant estimates suggest that they do, but the evidence is not overwhelming.

Table 1

**Regression of market income inequality on union and industrial relations variables**

Out of the twelve union and IR estimates, only two are statistically significant. The first is in Model 1, and suggests that, in a given province, over time increases in union density levels are associated with a decline in market income inequality as measured by the Gini coefficient. The second significant estimate is in Model 4, and it suggests that over time increases in the union favourableness of collective labour statutes (LRI) are associated with lower levels of income concentration at the very top of the distribution. Compared to density rates and the quality of union institution, none of the estimates for person-days lost are significantly associated with inequality, suggesting that striking may no longer matter for the distribution of market income, which is consistent with findings elsewhere (Rosenfeld, 2006). Moreover, none of the union or IR estimates are significantly associated with over time changes in inequality as measured by the two inter-decile ratios (Model 2 and 3).

Table 1 also provides other noteworthy findings. Given the central role accorded to political parties and partisanships in the coalitional power struggles that are part and parcel of power resources theory, it is surprising that none of the political variables reach conventional levels of statistical significance. Estimates indicate that left- and centre-party incumbency do not affect market income inequality significantly differently than right-part incumbency. As for the other control variables, the most compelling results are those from employment rates estimates, which combine to suggest that employment levels play an important role in reducing inequality in many areas of the distribution.

Discussion

Linking back to the theoretical framework, this discussion provides a nuanced interpretation of results. It focuses on two objects of analytic interest: 1- interpreting the significant results and insignificant results for union density and the labour relations index, and 2- understanding why economic strikes no longer matter for inequality.

Consistent with theory, results suggest that higher unionization reduces levels of inequality as measured by the Gini coefficient, an overall measure of inequality with a particular sensitivity to changes across the middle of the distribution. As predicted by certain genres of economic theory, this happens as unions’ wage compression and standardization effects increase with density levels. It also happens as the union threat effect in a given sector increases and as unions have more resources and legitimacy in their promotion norms of equity (Western and Rosenfeld, 2011).

The reason density is not significantly associated with lower levels of inequality in the lower and upper halves of the distribution (D5:D2 and D9:D5 ratios) may have a lot to do with where unions are located in the overall income spectrum. Figure 8 offers a portrait of the distribution of union members by income in Canada. It shows that very few unionized persons are located in the lower two deciles (D1 and D2). This means that the within-sector inequality-reducing effect of unions likely does little to compress the distribution from the very bottom. Conversely, the lack of an effect in the upper half of the distribution may be rationalized by a saturation effect. Specifically, increases in membership may have a marginal effect on inequality in an area of the distribution where union presence is already strong.

Figure 8

**Proportion of union members by market income deciles in Canada, mean values (1996-2011)**

Teasing out a coherent narrative concerning the labour relations index estimates is difficult. The non-significant estimates may be explained by the quality of union legal institutions being secondary to unions’ power-in-numbers (union density). This means that much of the LRI’s relationship with inequality may be washed away by union density, which may mediate the LRI-inequality relationship. However, this does not explain why the LRI is significantly associated with lower levels of income concentration at the top. It may be that, over membership size, the quality of union institutions better reflects unions’ ability to operate as a fairness factor that directly affects top income earners (Gomez and Tzioumis, 2006) or that the LRI is a better proxy of unions’ capacity to institutionalize norms of equity in the moral economy (Western and Rosenfeld, 2011). While this is a first attempt, understanding how different dimensions of union power interact is flagged here as a promising area for future research.

As for why worker strikes no longer matter, other than their likely increasing ineffectiveness (Rosenfeld, 2006), the issue may lie in this study’s measurement strategy, which does not account for broader union militancy. Some have argued that strike action has moved outside the workplace and into civil society through general strikes and political protest involving not just unions, but also grass-root militant and community organizations (Kelly, 2015b; Vandeale, 2016). Such activities are not captured by the person-days lost measure.

Conclusion

Motivated by the substantive decline literature, the increasingly difficult environment in which unions operate and a growing scholarship suggesting unions’ dwindling distributive effect, this article aimed to answer a straightforward question: do unions and industrial relations still matter for inequality? The answer is a cautious yes. Rare significant results from multivariate provincial-level panel data regressions suggest that higher union density and more union-friendly collective labour statutes promote a more equal distribution of market income. Findings indicate that unionization’s inequality-reducing effect is limited to changes in the middle of the distribution, as measured by the Gini coefficient. As for the quality of collective labour statutes, such measures are directly and significantly linked to less income concentration only at the very top of the distribution. No significant results are found regarding the distributive effect of strike action, pointing to the growing ineffectiveness of such measures and to a potential displacement of union militancy, from the workplace to broader civil society.

This article opens up avenues for future research. First, a better understanding of the interaction between macro-level measures of union power (density, quality of institutional underpinnings, strike action) is needed. A particularly promising area of investigation would be focus on establishing the causal relationship between unionization rates and the quality of union legal institutions. Second, looking at new ways to quantitatively measure macro-level union militancy, outside traditional strike action, would likely provide a better understanding of what unions do aside from collective bargaining.

Finally, while this study’s results indicate that unions have retained some capacity to reduce inequality over the last few decades, the lack of overwhelming evidence concerning this phenonenon suggests that it may only be a matter of time before unions’ distributive effect dissipates in the provinces. If unions’ equalizing effect does further abate, the construction of new forms of worker countervailing power within capitalist market economies will become increasingly urgent.

Unions, Industrial Relations and Market Income Inequality in Canada’s Provinces Syndicats, relations de travail et réduction des inégalités dans les diverses provinces au Canada

Abstract

Résumé

Introduction