Corps de l’article

This book by Hristos Doucouliagos, Richard Freeman and Patrice Laroche provides an in-depth and impressively executed meta-analytic investigation and synthesis of the econometric evidence produced on the economic effects of trade unions since publication of the landmark volume What Do Unions Do? (WDUD) in 1984 by Freeman and Medoff. For Freeman, the book is a fitting tribute to one of the most famous and enduring works ever written on unions and, similarly, for Doucouliagos, Laroche and their colleague Stanley, it is a fitting tribute to their pioneering work on use of meta-analysis in labour economics.

One of the standout innovations of WDUD (1984) was to broaden econometric analysis of unions beyond the traditional topic of wage effects to include a variety of non-wage outcomes, such as productivity level and growth, capital investment, employee turnover, job satisfaction and firm profitability. This broader palette of union effects was inspired, in turn, by Freeman and Medoff’s equally innovative expansion of the traditional monopoly model of union wage effects to include collective voice and institutional response effects. They presented the model as capturing the “two faces of unionism”—a negative monopoly wage effect on economic efficiency/performance and positive voice effect from lower turnover, improved working conditions, reduced wage inequality and more effective workplace cooperation—and argued the net effect of unions on the economy can thus be positive, negative, or neutral depending on the quantitative size of the various monopoly and voice outcomes. The most eye-catching and controversial econometric finding in WDUD was that unions have a positive effect on firm-level productivity, a finding which, if substantiated, shifts the economic verdict on union economic effects toward a more neutral or even positive direction.

Publication of WDUD (1984) spurred a very large follow-up empirical literature on the various non-wage effects of unions with analysis expanded beyond the U.S. to include a number of other countries and also disaggregated by industry/sector. The diversity among non-wage outcome variables and substantial heterogeneity in findings across studies has made it difficult in the three decades since WDUD to form a reliable estimate of the degree to which Freeman and Medoff’s original results and conclusions have withstood the test of time. The purpose of the present volume by Doucouliagos, Freeman and Laroche (henceforth DFL) is to use advanced techniques of meta-analysis to synthesize from this extensive empirical literature “best estimate” values of the mean union effect on five outcome variables related directly or indirectly to the overall union effect on firm productivity and performance. They are: productivity level, productivity growth, investment in physical capital and research/development (innovation), employee behavioral responses (turnover, job satisfaction, organizational commitment) and firm profitability.

For the meta-analysis, DFL comb the literature and assemble 301 studies that report 2,257 effect estimates. Chapter two of the book provides overview and context with a statistical bibliographic analysis of the empirical literature on union non-wage effects since publication of WDUD. Chapters three through seven report the meta-analytic results for the five productivity/performance-related outcomes. Chapter eight summarizes the findings, the implications for WDUD three decades on, and implications for union research and policy writ large.

Table 8.1 in the book handily summarizes the main results of the chapter-by-chapter meta-analysis. For productivity level, unions in manufacturing have no effect in the US, a negative effect in the UK, and positive effect in developing countries, while for non-manufacturing (e.g., construction and education), the union effect on productivity is positive. For productivity growth, no union effect is found; for profitability, a negative effect is found; for physical capital investment, a negative effect is found; and for job satisfaction, no effect is found.

DFL also test for a time trend in each of these union effects, with results reported in Table 8.3. The union effect on productivity level in manufacturing has become more negative over time but with no change in other sectors. No time trend is found in the union effect on productivity growth; the respective effects on physical capital investment and intangible capital investment are less negative over time; the union effect on job satisfaction cycles over time; and the union effect on profitability is less negative over time. Although the book does not include a chapter on meta-analysis of union wage effects, based on other studies DFL report in Table 8.3 a negative time trend for this outcome variable.

What does all of this evidence add up to in terms of state of knowledge and evaluation of unions? I judge this the most important question and, unfortunately, where the book most falls short. In particular, the shortfall comes in the last chapter where DFL don’t adequately take their meta-analytic tree results in preceding chapters and fit them into a larger so-what forest picture viz. WDUD and union effects thirty years later.

Organizing the book around the WDUD (1984) volume and its main findings is both useful and appropriate given it has served as a foundational work for the last three decades of union research. However, the subtitle of the book (emphasis added), “A Study of a Research Field and its Findings”, promises a broader perspective that extends beyond WDUD, and informs and integrates with a larger body of literature on union effects. Instead, in the summary chapter the discussion is narrowly focused on WDUD and, then, within WDUD, the discussion is further telescoped to include only the five productivity/performance-related outcomes with no summary of implications for theory, policy, or the broader literature.

The authors do provide a very succinct and helpful one-paragraph summation of the book’s principle findings and conclusions viz. the earlier WDUD volume. The overall conclusion is first stated (p. 149, second italics added), “The bottom line of our investigation is simple. Hundreds of econometric studies and two thousand or so estimates after What Do Unions Do? the fundamental claims of the book stand up well, although not perfectly.”

Then come revisions, per their statement (p. 149), “We identify three areas where succeeding work has revised the picture of unions: 1- on productivity, where ensuing research replaces the generalization of positive effect with a central tendency of negligible effects…; 2- on job satisfaction, where the causal route of the negative union association with satisfaction lies in selectivity of workplaces and members into unions rather that the collective voice stressed in Freeman and Medoff; and 3- on the variation of union effects with institutional settings, found by extending studies outside the United States.”

Nonetheless, evaluated as a thirty-year assessment/revision of WDUD and, more generally, the literature and knowledge base on union effects, DFL’s book has its own “two faces” quality. To give the authors their due credit, the book is an impressive tour de force of meta-statistical analysis, well illustrates the tremendous advances made in empirical labour economics over the last three decades, is likely to provide the definitive statement for years to come on the profession’s best estimate of the union effect on the five productivity/performance-related outcomes examined in the book, and helpfully narrows the confidence intervals and adjusts several of the estimated means on WDUD’s major empirical propositions and findings. Certainly, I would be proud to have my name on the front cover.

There are, however, also several minus aspects to the book contained in the concluding chapter. As noted above, for example, the book covers only a slice of the topics and analysis in WDUD. For this reason, just as positioning the book as “a study of a research field” is an overstatement, so too is the assertion (above quote) that, “the fundamental claims of the book stand up well, although not perfectly.” I see two problems here, one inconvenient and the other substantive.

Regarding the former, DFL tell readers in the last paragraph of the book (p. 163), “Finally, union activities almost invariably trade off some economic efficiency for the greater justice at workplaces and reduced inequalities valued by persons on the left and almost invariably do so in ways that constrain the personal liberties and economic freedoms valued by libertarians on the right.” Considered in the book, however, are only a select group of economic-related aspects of unions and the efficiency side of the two-faces debate while omitted are non-economic aspects and the social justice/inequality side. Logic suggests, therefore, that either part of the case underlying WDUD’s “fundamental claims” is in fact missing from the book or, alternatively, the social justice/inequality part of WDUD is secondary and non-fundamental for evaluating the positive voice case for unions. Given that the three authors come from the mainstream of economics where social welfare effects are typically evaluated on the singular criterion of efficiency (the smaller the Harberger triangle the better), and the focus of the book is on economic effects of unions, the latter interpretation seems favoured. Economists will like this choice because it keeps the analysis focused on what they consider the positive/objective side of the union debate but for non-economists, including most people in industrial relations, it represents an artificial truncation of what unions do for society with an accompanying bias toward market/individualist ordering.

Even if evaluated primarily/solely on efficiency grounds, DFL seem to overstate the degree to which their findings leave intact the fundamental propositions of WDUD. On this subject, DFL (p. 10, emphasis added) state, “Freeman and Medoff’s work challenges the traditional economics view of unionism. They argue that while unions have a monopoly face, they also have a collective voice face. By comparing the effects of these two faces on productivity, income distribution, and other economic outcomes, Freeman and Medoff argue that, in net terms, unions are socially desirable institutions.”

For this fundamental proposition to hold, the positive collective voice/institutional response face of unions has to increase the size of the joint surplus in the firm more (or not less) than the negative monopoly wage face reduces it. This outcome is seemingly affirmed in WDUD for the US case because Freeman and Medoff find a large positive union effect on productivity, which they can set against the negative union wage effect. However, thirty years later the balance on the efficiency scale seems to tip toward a “guilty” verdict on the social utility of unions. One reason is that the union productivity effect has gone from positive to zero while the union wage effect, though smaller, remains negative (chapter 3); a second reason (chapter 6) is that the evidence presented by DFL on a positive union voice effect on productivity-related employee behaviours (turnover, job satisfaction, commitment) is weaker relative to WDUD.

Looking back, a significant part of WDUD’s appeal and claim to fame originated in reversing, or at least neutering, the economists’ traditional negative indictment of unions on labour monopoly grounds by adding the positive collective voice face in the theoretical model and then presenting considerable new empirical evidence that seemed to confirm unions’ counterbalancing positive effect on non-wage outcomes. The new meta-analytic empirical evidence presented by DFL is an order of magnitude more sophisticated and persuasive than that in WDUD, but also less supportive of the two-faces thesis that made it famous.

One thing a reader notices in DFL’s summary chapter is they assiduously keep the discussion empirically focused and do not address the implications of the book’s findings for the exist-voice model or union effect models more generally. Establishing causality and motivation is difficult but one cannot help but wonder if part of the explanation is that the authors realize silence is preferable to voice on this subject for the net effect of their meta-statistical evidence seems to be to reduce the two faces of unions model back toward the traditional one face model Freeman and Medoff wrote WDUD to challenge.

An irony of this book, therefore, is that it is positioned as a thirty-year commemoration and celebration of WDUD, but seems, on balance, to undercut its central theoretical/empirical foundation. Evaluated on its own merits, however, DFL’s book represents a definite advance in both empirical analysis and knowledge.