Economic effects of unions

Given some recent discourse on unions in our comments area, I decided to reach back to a fairly comprehensive literature survey on the empirical evidence about unions. The report comes not from the CCPA, CAW or CLC, nor does it come from the ILO, but from the World Bank. It is a rather weighty tome, published in 2002, and can be downloaded here. Below is an excerpt from the Introduction and Summary with a long list of the key empirical findings.

Unions and Collective Bargaining: Economic Effects in a Global Environment

Toke Aidt and Zafiris Tzannato

This book examines the economic effects of unions and collective bargaining. Although the study is based on a large and detailed literature survey (covering more than 1,000 primary and secondary studies), it cannot claim to be conclusive. The main contribution is to take stock of existing knowledge, thereby paving the way for more innovative future research on the link between labor standards, collective bargaining, and economic performance.

…[T]he book focuses on two more specific questions: what is the impact on the economic well-being of individual workers and the performance of firms of basing industrial relations on collective bargaining between unions and employers rather than relying on individual contracting (question A) and what is the impact on the macroeconomy of adopting different institutional approaches to collective bargaining (question B).

… We begin with three general observations:

1. Comparative studies reveal little systematic difference in economic performance between countries that enforce the two relevant labor standards and countries that do not. This is partly a reflection of the difficulties of isolating the effects of labor standards from other determinants of economic performance, and suggests that the impact of labor standards perhaps best can be analyzed on a case-by-case basis.

2. The microeconomic consequences of collective bargaining are context-specific, and although unions in both industrial and developing countries are successful in securing a wage markup for their members and other workers covered by collective agreements, no general conclusions about the net costs (or benefits) of unions can be reached. Depending on the economic, institutional, and political environment in which unions and employers interact, collective bargaining as opposed to individual contracting can contribute negatively or positively to the economic performance of firms and to the well-being of workers.

3. The macroeconomic impact of collective bargaining is hard to disentangle from other determinants of economic performance. While the available evidence from comparative studies of the OECD countries is fragile, two general features should be emphasized. First, the impact of collective bargaining on various aspects of macroeconomic performance depends on the economic, legal, and political environment in which collective bargaining takes place and can vary over time. Second, important complementarities exist between key aspects of the bargaining system. Therefore, the impact of individual aspects such as union density or centralization of bargaining cannot be assessed in isolation. It is the package of institutions that matters.

We elaborate on these themes in the following two sections where we attempt to summarize in more detail the specific findings related to questions A and B.

Findings Regarding Question A (Microeconomic Efects)

The human rights argument in support of workers’ rights is compelling. But from an economic point of view the key questions are: What are the costs and benefits of unions? Is collective bargaining a useful institution that contributes to the achievement of desirable economic outcomes at the firm and/or sector level, or is it another labor market distortion that prevents the market from doing its job?

The existence of unions arises from the asymmetry in contracting between individual workers and employers, the concern for basic workers’ rights, and the different perceptions about the merits of employment relations governed by individual contracts or collective agreements. Textbook reasoning suggests that the alternative to a unionized labor market is one characterized by the atomistic, perfectly competitive structure that ensures that individual workers choose whether or not to work by comparing the given perfectly competitive wage with the marginal utility of nonmarket activity. However, the reality facing policymakers is far less clear-cut than this suggests. First, the “removal” of unions may not reveal an underlying perfectly competitive situation in the labor market; instead, it may expose market imperfections on the labor demand side in the form of monopsony, that is, a situation in which there is only one buyer of the relevant labor services. Alternatively, firms in some industries may voluntarily pay workers more than the going market rate to motivate existing workers or to attract new ones. Hence, policy decisions whose central objective is the “return” to a perfectly competitive labor market (with all its well-known potential benefits) can succeed only if they are accompanied by policies designed to free up the demand side of the market. Indeed, the presence of unions in such circumstances may offer a second-best alternative to free competition. In this case, the countervailing influence of unions may result in a set of outcomes closer to the competitive equilibrium than those that would result from competition on the supply side of the labor market and monopsony on the demand side. The removal of unions may also reveal imperfections on the supply side of the labor market unrelated to unionism. For example, workers with specific skills or those protected by high turnover costs can gain “insider power,” which can be used to raise wages above the competitive level. Moreover, the potential benefits (referred to as participatory benefits) associated with the presence of unions in the form of “voice” (empowerment) as opposed to “exit” (separation) effects should be counted against the costs (in the form of welfare losses due to misallocation effects). We see that theoretical reasoning does not allow us to reach unambiguous conclusions about the net benefits (or cost) of unions. Whether collective bargaining contributes to the achievement of desirable economic outcomes or it prevents the market from doing its job is, at the end of the day, an empirical question.

To evaluate the costs and benefits of unions empirically, we need, in principle, to know how the labor market would work in their absence. The counterfactual is, of course, never observable in reality, nor can it, as argued above, be deduced from theory. Therefore, evaluations of the costs and benefits of unions must necessarily be based on a comparison of economic outcomes in those sectors of the economy that are unionized with those that are not, rather than comparing outcomes in currently unionized sectors with the likely outcomes if those sectors had not been unionized. In practice, this is done by estimating a union/nonunion differential or markup from individual worker or establishment data. The union/ nonunion differential is the difference between the target variable (wages, employment growth, productivity, and so on) in an average unionized firm (for a unionized worker) and an average nonunionized firm (for a nonunionized worker). Much of the empirical evidence on union/nonunion markups comes from the United States and the United Kingdom, but studies from other industrial countries as well as from some developing countries do exist, and are included in our survey.

We summarize the findings in 23 separate points. To aid exposition, we have grouped the findings into several related subject areas, preceded by an explanatory statement and a judgment about the robustness of the particular findings. The first group of findings relates to the wage markup; these results are very robust.

1. Union members and other workers covered by collective agreements in industrial as well as in developing countries do, on average, get a wage markup over their nonunionized (or uncovered) counterparts.

2. The markup is somewhat larger in the United States (15 percent) than in most other industrial countries (5 to 10 percent). In developing and middle-income countries, the markup can be higher or lower. For example, it appears high in Ghana, Malaysia, Mexico, and South Africa but relatively low in the Republic of Korea (in 1988, before the expansion of unionism).

3. Unions compress the wage distribution. In particular, the wage differentials between skilled and unskilled workers and the private return to education are reduced when unions are present.

4. One, albeit incomplete, way to assess the adverse effects of unions is to evaluate the welfare loss that the wage markup creates through the misallocation of resources in the whole economy. In general, these effects have been found to be small and of comparable magnitude to the deadweight loss arising from monopolies in product markets — no more than 0.2 to 0.5 percent of gross domestic product (GDP). However, even these low estimates may overstate the allocative loss of unions because they do not take full account of the unions’ potential effects on the productivity of their members (see points 21-23). On the other hand, they do not include all the potential costs of unions, such as the adverse impact that unions may have on firms’ investment behavior, so the estimates may understate the allocative loss of unionism (see point 20).

The size of the wage markup depends on a variety of worker and workplace characteristics. These include the following:

5. There is no significant difference between the wage markup for female workers and that for male workers in the United States and Australia. In some other countries such as Germany, Japan, Mexico, South Africa, and, perhaps, the United Kingdom, however, unionized women workers have a greater pay advantage over their nonunionized counterparts than unionized men.

6. There is some evidence from Canada and Malaysia to suggest that unions contribute to a reduction in the overall gender pay gap. British studies on the subject are inconclusive.

7. In the United States and the United Kingdom, unionized nonwhite workers tend to get a higher wage markup than white workers, although the U.S. evidence is mixed. In South Africa, “black unions are associated with a smaller markup than “white” unions. In Mexico and Canada, unions have been found to reduce the discrimination against indigenous people.

8. The wage markup tends to be higher in the private sector than in the public sector in industrial countries. The markup also depends on the economic environment in which unions and firms operate.

9. The impact of competitive conditions at the product market on the wage markup is not clear-cut and depends on how the competitiveness of the product market is measured. When firmspecific indicators of the competitive environment are used, unions are more successful in establishing a high wage markup if the relevant firms operate under less competitive conditions in the product market. This is not the case if industry concentration ratios are used as an indicator of product market competition. Arguably firm-specific indicators of competition are preferable to industry-wide indicators and so, on balance, product market competition seems to prevent unions from establishing a high wage markup.

Finally, the size of the wage markup also depends on the specific aspects of how collective bargaining is organized, and from the evidence it is possible to identify particular aspects of industrial relations that add to the markup.

10. Industries with high overall union density tend to have a higher wage markup.

11. Although in some countries, such as Malaysia and the United States, industry-level collective bargaining is associated with a higher markup than firm-level bargaining, this is not so in other countries. For example, recent studies from the United Kingdom fail to find a difference.

12. Multiunionisin at the firm level (when different unions compete to recruit or organize the same workers) does not lead to a higher markup per se. However, evidence from the United Kingdom shows that the markup is high in multiunion firms that negotiate separately with each union.

13. Pre-entry closed shops (union membership is a prerequisite to obtain employment) but not post-entry closed shops (union membership is required after hiring) are associated with an additional wage markup. Again, this evidence comes from the United Kingdom only.

The union impact on aspects of economic performance other than wages is less well understood. Although a number of conclusions can be drawn with some certainty, most should be treated as tentative. The most robust results relate to hours worked, job mobility, and profitability.

14. Voluntary job turnover is lower and job tenure longer in unionized firms. The evidence on this finding from Australia, Japan, Malaysia, the United Kingdom, and the United States seems quite robust. On the other hand, layoffs, particularly temporary layoffs, are more frequent in unioszed firms than in nonunionized ones.

15. Net company profits (price-cost ratios, Tobin’s q, subjective profitability assessments, and the like) tend to be lower in unionized firms than in similar nonunionized firms (in Japan, the United Kingdom, and the United States). There seems to be a relatively large negative impact on profitability in firms that have product market power.

16. Hours worked is lower among unionized than nonunionized workers. This is true for both total and normal hours. In addition, unionized workers are more likely to get paid for the overtime work that they do.

The evidence concerning employment-related benefits, spending on research and development (R&D) and physical investment, and employment growth is less robust, but the following could be noted.

17. Fringe benefits are more commonly found among unionized workers than among nonunionized ones (in Australia, Japan, Malaysia, the United Kingdom, and the United States ). Benefits can include severance pay, paid holidays, paid sick leave, pension plans, and so on. At the same time, there is evidence that part of the wage markup is compensation for an inflexible and structured work environment.

18. Employment growth can be slower in unionized than in nonunionized firms (as suggested by evidence from Canada, Jamaica, Malaysia, the United Kingdom, and the United States), but the evidence is not particularly strong, and the observed differences most likely represent situations of disequilibrium.

19. Although spending on R&D tends to be lower in unionized than in nonunionized firms, unionized firms seem to adopt new technology as fast as nonunionized ones (in Canada, Malaysia, the United Kingdom, and the United States).

20. The investment rate (physical capital) tends to be lower in unionized than in nonunionized firms with otherwise similar characteristics (in the United Kingdom and the United States). The adverse impact seems to be relatively larger when firms operate in competitive product markets, although. only one study (from the United Kingdom) has addressed this issue directly.

The least robust results relate to productivity, training, and pay systems.

21. The impact of unions on productivity levels (in terms of both labor productivity and total factor productivity) is empirically indeterminate. Some studies suggest a positive impact, but.others imply a negative impact or no impact at all. For example, unions appear to have a negative impact on productivity levels in the United Kingdom but a positive impact in Malaysia. In the United States, there is no discernible impact, on average, but there is considerable variation across industries. Industries operating in competitive product markets and firms with “high quality” industrial relations (as measured by grievances among workers, strikes, and the like) have, on average, high productivity.

22. The relationship between unions and productivity growth is not clear either. In the United States, the union/nonunion differential is found to be negative or insigruficant. In the United Kingdom, some studies suggest that the weakening of British unions is one factor explaining the high productivity growth in the United Kingdom in the 1980s.

23. Unionized workers tend to receive more training than their nonunionized counterparts, especially company-related training. Overall, these findings show that the extent to which particular costs prevail or particular benefits materialize depends on the economic environment in which unions and employers operate, as well as the way in which collective bargaining is organized. It is of primary interest to note that specific aspects of the economic and institutional environment, such as product market competition, absence of pre-entry closed shops and so on, can help to minimize the net costs or maximize the net benefits of unions. In devising union regulations, policymakers must recognize this fact and seek to remove the costs of unions while at the same time retaining their benefits.

Findings Regarding Question B (Macroeconomic Effects)

The impact of collective bargaining on macroeconomic performance can best be assessed through comparative studies where the performance of countries with (very) different bargaining systems is systematically compared. Most studies look at the economic performance of the OECD countries during the period from 1960 to 1998, and ask how the framework of collective bargaining affects a large number of macroeconomic performance indicators (such as unemployment and inflation) and labor market flexibility indicators (such as real wage flexibility) in an environment in which workers’ rights can be taken as granted. The importance of collective bargaining as opposed to other ways of organizing contracting in the labor market can be measured by union density (the proportion of workers who are union members) and bargaining coverage (the proportion of the work force that is covered by a collective agreement). With respect to these indicators of collective bargaining, we find:

1. Union density per se has a very weak association, or perhaps no association, with economic performance indicators such as the unemployment rate, inflation, the employment rate, real compensation growth, labor supply, adjustment speed to wage shocks, real wage flexibility, and labor and total factor productivity. There is, however, one significant exception: union density correlates negatively with labor earnings inequality and wage dispersion.

2. Bargaining coverage tends to be associated with higher real wage growth (with no impact on productivity growth), lower employment rates, higher unemployment rates, and higher inflation. As with union density, bargaining coverage correlates negatively with labor earnings inequality and wage dispersion.

Collective bargaining is potentially a powerful means to facilitate bargaining coordination; that is, the extent of coordination between unions and employers’ organizations in wage setting and other aspects of industrial relations (for example, working conditions, holidays and leave provisions and so on). Six different aspects of bargaining coordination can be identified: union centralization, union concentration, employer centralization, level of collective bargaining, informal coordination, and corporatism. Bargaining coordination is increasingly seen as an influential determinant of labor market and macroeconomic performance. For example, the Japanese system of wage setting is decentralized (firm-based) but coordinated in the sense that it follows company rules based on seniority (hence, they are transparent) rather than individual contracting. In this system, workers are not paid wages equal to their individual reservation wage (that is, the wage level below which the worker will not supply his or her labor), as would have been the case under individual contracting, but this difference does not adversely affect efficiency. The Netherlands and Germany also have coordinated systems through strong employer organizations, coordination among giant companies or across industries, and coordination among unions. In France the government provides coordination in the form of public services, utilities, and large nationalized industries. In Italy, there is informal employer coordination (via the big firms and regional employers’ associations) and between some union confederations. Finally, Sweden has a centralized employers’ organization as well as centralized union confederations.

The comparative literature focuses on two hypothesizes about the relationship between bargaining coordination and economic performance:

Hypothesis 1. Coordinated collective bargaining leads to better economic outcomes compared to semicoordinated collective bargaining, which, in turn, performs better than uncoordinated collective bargaining.

Hypothesis 2. (The hump hypothesis) Semicoordinated collective bargaining leads to worse economic outcomes than both coordinated and uncoordinated collective bargaining.

The evidence suggests that bargaining coordination did have a beneficial impact on macroeconomic performance in the 1970s and 1980s, but the evidence is fragile and in the 1990s the impact seemed to disappear for most indicators. More specifically, we find:

3. Countries with highly coordinated collective bargaining tend to be associated with lower and less persistent unemployment, less earnings inequality and wage dispersion, and fewer and shorter strikes compared to countries with semicoordinated (for example, industry-level bargaining) or uncoordinated (for example, firm level bargaining or individual contracting) collective bargaining. In terms of productivity growth and real wage flexibility, countries with highly coordinated collective bargaining tend to perform slightly better than countries with semicoordinated collective bargaining but may not perform differently than countries with uncoordinated collective bargaining. This lends some support to hypothesis 1, but only for the 1970s and 1980s. For most economic indicators, the differences disappear in the 1990s. Two exceptions are earnings inequality and wage dispersion. These indicators are comparatively low in countries with highly coordinated collective bargaining throughout the whole period.

4. Although countries with either uncoordinated or coordinated collective bargaining tend to be associated with lower and less persistent unemployment and higher productivity growth than semicoordinakd collective bargaining during the period 1960 to 1990, the evidence in favor of the hump hypothesis is, in general, very weak, particularly for the 1990s.

5. In terms of Inflation and the employment rate, there seems to be little difference between coordinated, semicoordinated, and uncoordinated collective bargaining.

These conclusions refer to one dimension of industrial relations and take other dimensions as given (either by controlling for them or by inappropriately ignoring them). This ignores the possibility of complementarities between union density/bargaining coverage and bargaining coordination. Such complementarities are important for the impact of collective bargaining on economic performance, and it can therefore be misleading to focus on one particular aspect in isolation. In particular, the following fact should be emphasized:

6. High union density and bargaining coverage do not contribute to poor unemployment performance so long as they are complemented by high bargaining coordination (particularly among employers).

Bargaining coordination is related to a number of different aspects of industrial relations, such as the centralization of collective bargaining, corporatism, informal coordination between employers or unions, and so on. As far as different types of coordination are concerned, the following points can be emphasized:

7. Informal coordination of wage bargaining (informal consultations between firms and/or unions or pattern bargaining) tends to mitigate the potential disadvantage (in terms of relative high unemployment) associated with semicoordinated (such as industry-level) wage bargaining, and can arise in countries with relatively low union density and bargaining coverage.

8. Coordination among employers tends to be more important in producing low unemployment than coordination among employees. This suggests that employers’ organizations are more effective in controlling wage drift than union confederations.

9. Countries that have competing unions (multiunionism) and many different union confederations tend to perform worse (in terms of unemployment and inflation) than other countries.

10. The effects of coordination can be compromised or accentuated depending on the political orientation of the government. “Good” economic outcomes (in terms of economic growth) can arise either when strong, centralized unions are paired with a strong leftwing government or when weak, decentralized unions are paired with a right-wing government. A mismatch (weak unions paired with a strong left-wing government or strong unions paired with a right-wing government) can lead to poor economic outcomes.

This concludes the long list of specific macroeconomic findings. Although some patterns emerge, we feel that the evidence is too weak and fragile to warrant grand generalizations about the performance of specific labor market institutions. Instead, we want to stress that the relationship between collective bargaining and economic performance cannot be fully understood unless the general economic and political environment in which bargaining takes place is taken into account. One should therefore be careful not to infer that institutional forms that work well in one environment would also work well in other often very different-environments. With this caveat in mind, the synthesis of the literature embodied in our list of findings can provide a useful starting point for more specific studies of labor market reform.