How Labor Markets Work: Reflections on Theory and Practice.
This volume contains papers presented at a 1986 conference. One purpose of the conference was to honor the work of four distinguished labor economists, John Dunlop, Clark Kerr, Richard Lester, and Lloyd Reynolds, each of whom contributed a paper. In addition, there is an overview of their combined contributions by the editor (Bruce Kaufman), and a commentary by a representative of a later generation of labor economists, Richard Freeman.
A second purpose of the conference was to survey in retrospect a style of research in labor economics that all of the participants agree is now much less in vogue than during the 1940s and 1950s. Although this research style is familiar, its characterization and even its name presented problems, both to its practitioners and to the editor (pp. 187-90). In particular, Kerr rejects the connotations of "neo-institutionalist" in favor of "neo-classical revisionist" (pp. 10ff).
Clearly, more than nomenclature is involved in this search for a title. At least some of the honorees are anxious to stress their connection with mainstream economic theory; Kerr, in particular, takes great pains to set forth the neo-classical lineage of the Revisionists' ideas. The emphasis of the work being celebrated, however, was on revision and not on further development. What comes through in all the essays is that the Revisionists were, first and foremost, keen participant-observers of the operations of real world economies who were anxious to account for what they saw.
Given that the birthdates of the Revisionists range from 1908 to 1914, it is not surprising that events during their graduate school years and the early part of their professional careers (roughly 1933-45) should have had a far greater impact on their pre-analytic vision than what might have been learned from the history of earlier periods. As the essays make clear, the massive unemployment of the 1930s, combined with the impact of Keynes's General Theory, gave them a strong feeling that full employment was not the normal state of affairs, and made them impatient with any theory that seemed to imply the contrary.
Before the lessons of a decade of very high unemployment could be fully absorbed, the wartime boom of the early 1940s was upon them, and the Revisionists actively participated in setting wages and other terms of employment in a controlled economy where the allocative function of prices and wages was strongly repressed. It is fair to say that during these two decades market forces had as little impact on price-quantity behavior as at any time in the past century. In short, observation, the spirit of rising new theories (The General Theory, Imperfect and Monopolistic Competition), and the Revisionists' own early practical experience as administrators of wage and price controls all combined to generate a research style antithetical to emphasis on the role of market forces.
To term this style "neo-classical revisionism" is to suggest greater receptivity to neo-classical ideas than the work of the Revisionists displayed. With few exceptions, when discussing neo-classical theory the Revisionists accentuated the negative;(1) when not irrelevant to the analysis of price-quantity interaction, the main thrust of their papers was to point out errors and limitations of the neo-classical paradigm. Points of agreement, or possibilities of reconciliation of neo-classical and Revisionist views, were generally downplayed or ignored, and
(1) Noteworthy among the exceptions are: Clark Kerr, "Balkanization of Labor Markets," in Labor Mobility and Economic Opportunity (Cambridge, Mass.: Technology Press, 1954); Lloyd H. Fisher, The Harvest Labor Market in California (Cambridge, Mass.: Harvard University Press, 1953); C. Kerr and L. H. Fisher, "Effects of Environment and Administration on Job Evaluation," Harvard Business Review, Vol. 28 (July 1950), pp. 77-96; and J. T. Dunlop, Wage Determination Under Trade Unions (New York: Macmillan, 1944), Chapters 1 and 2. antinomies were emphasized. (To be sure, their neo-classical opponents exhibited an equal lack of interest in reconciliation and, for that matter, in any form of intellectual interaction.)
The confrontational posture of the Revisionists is not confined to the past. In varying degree, the honorees are unreconciled to the current domination of labor economics by adherents of the neo-classical approach. By a wide margin, Dunlop is the most hostile of the four, but doubts and reservations about "recent developments" are expressed in all of the papers. Kerr's essay shows much greater interest in and appreciation of post-1960 research than the others, but his list of "tasks for labor economics" (pp. 35-37) and "general rules for the realistic study of labor economics" (pp. 37-39) indicate considerable dissatisfaction with many facets of recent work.
The irrepressible antagonism of the Revisionists to neo-classical research is due only in part to misunderstanding, although there is plenty of that. It stems also from the Revisionists' strong desire to be realistic.(2) In this context, realism means (inter alia) that the assumptions of theoretical models should be as descriptively accurate as possible, and (accordingly) that it is undesirable for a theory's premises to be counterfactual.(3) The neo-classical rebuttal to such contentions that emerged in the 1950s, stemming largely from Friedman's 1953 paper on methodology ("The Methodology of Positive Economics," in Essays in Positive Economics), is that realistic assumptions are neither necessary nor possible.
In effect, the neo-classical counter-argument is as follows: all descriptions of the "real world" are necessarily incomplete. Wise choice among differing descriptions of the same phenomena depends on the importance of what is included and of what is left out. "Importance" varies with purpose; for the economic theorist, the purpose is to infer the price-quantity consequences of changes in exogenous variables. For this purpose, advocates of the (neo-classical) competitive hypothesis abstract from departures from competitive equilibrium (such as profit maximization and ignorance), contending that the effect of such abstraction (upon price-quantity interrelations) is--to a "first approximation"--negligible. Put differently, the neo-classical approach is to impound the effects of departures from the competitive model in the disturbance terms of price and quantity equations, and to justify that impoundment by arguing that it has but a negligible effect on estimates of the relevant structural parameters.
(2) In an earlier paper ("The Intellectual Role of the Neorealists in Labor Economics," Industrial Relations, Vol. 22, pp. 298-318), Kerr actually termed "Neorealists" those whom he here calls "Neo-classical Revisionists."
(3) Dunlop's essay is replete with examples of assertions to this effect; Lester's famous attack on "marginalism" ("Shortcomings of Marginal Analysis for Wage-Employment Problems," American Economic Review, Vol. 36, pp. 63-82, March 1946) is a locus classicus of this type of argument.
The validity of such contentions is a matter of current debate, and the Revisionist side is not without contemporary supporters--Freeman, for example. The Revisionists themselves, however, do not attempt to present detailed critiques of post-1960 neo-classical labor market research, but are content to offer overviews that are generally unsympathetic to the methodology used.
Freeman's essay compares the findings of modern (that is, post-1960) labor market research with those of the Revisionists; the comparisons are generally favorable to the latter. Although these comparisons are interesting, the judgments offered are of uncertain validity because they are based on the findings of other researchers, which are utilized without an explicit appraisal either of data or of research methodology. Although contrary to the valedictory spirit of the conference, the study of "how labor markets work" would have been better served had there been a live neo-classicist to present the other side of the argument.
The need for representation of the other side is especially keen because, as its name indicates, Neo-Classical Revisionism was focused on exposing the errors and limitations of neo-classical labor economics. As will be seen, I think the Revisionists had--and have--a case. But the presentation of that case was flawed by an inadequate perception of what they were critizing. Specifically, they failed to appreciate the many qualifications with which neo-classical propositions are surrounded.
Thus, both Dunlop and Lester make a point of noting the existence and persistence of wage differentials among seemingly comparable workers and jobs, alleging that this is contrary to neo-classical theory.(4) But in so arguing, they ignore the role of disturbances that incorporate
(4) Dunlop, pp. 56-57; Lester, pp. 92-95. Although I cannot find explicit statements on this point by Kerr and Reynolds in their essays in the present volume, their previous writings and the general tone of their essays lead me to conjecture that they also would subscribe, more or less, to this view. Also see Kaufman's essay, pp. 149-62. the many "other factors" that generate dispersions of wage rates among seemingly comparable workers, jobs, localities, industries, and so forth at every moment of time. Provided there is no temporal pattern among such disturbances, the mere persistence of dispersion provides no evidence against--or for--the neo-classical theory. But if, as is often the case, there is a temporal pattern among the disturbances such that some workers (or jobs) are paid consistently more than others over a substantial period of time, then there is an anomaly that requires explanation.
To provide such an explanation, a search must be made for differential characteristics associated with relative wage levels, and adjustments should be made (in one way or another) for the wage differences attributable to differences in these characteristics. The neo-classical theory would predict the absence (among comparable workers or jobs) of persisting differences in wages so adjusted. Nevertheless, such differences have sometimes been found to persist over substantial periods of time. In such cases, the adjusted wage data should be studied to see if the persisting wage differences can be interpreted as tending to zero with the passage of time. If so, the dispersions might be considered to be neo-classically compatible for the long run, though nor\t for the short run; if not, the data should be interpreted as rejecting the neo-classical theory.
Manifestly, the Revisionists did not envisage such a painstaking research program. Partly, this was because their aversion of that theory.(5) In particular, they never faced up to the task of interpreting the theory when it was augmented with disturbances. But, even more important, their tropism for "realism" prevented them from appreciating how abstract, and (therefore) how difficult to test, neo-classical wage theory was.
Let me elaborate: (1) The argument of the preceding paragraphs implies that no cross-section of observations referring to a single date can ever of itself provide a test of statements about relative wages (or prices), unless some a priori restrictions are imposed on the disturbances.(6) Such restrictions are not implied by neo-classical assumptions, but must be made and justified ad hoc. In consequence, barring unusually strict specification of the characteristics of a sample, a time series of cross-sections (such as a panel) is required in order to test neo-classical propositions about differences among relative wage earners.
(2) To test neo-classical wage theory, it is necessary to adjust observed wages for compensating differentials including those reflecting anticipated future payoffs for current investment in human capital. Obviously, making such adjustments is difficult; it often involves controversy over procedures, and it is rarely defended as more than "the best possible adjustment of available data in the present state of the art."
(3) Finally, if, as often happens, the adjusted wage data appear to reject the neo-classical theory for a given time interval, it is necessary to consider whether the data for a suitably longer time interval would be interpreted as being supportive of the theory, with the anomalous data rationalized as reflecting a short-run distrurbance.
(5) See Reynolds's candid remarks (p. 138) on the influence of his emotional bias against "accepted theory."
(6) For example, the disturbances are uncorrelated with any fixed characteristic that is correlated with equilibrium wage levels.
Together, the foregoing considerations suggest that the implications of neo-classical wage theory are compatible with a wide variety of observed patterns of relative wage rates and labor quantities, and that therefore the theory is difficult to reject. According to this interpretation, the attacks of the Revisionists were misguided: much of what they had to say about "how labor markets work" was quite compatible with acceptance of neo-classical theory. Many of the ad hoc regularities they described could be readily accommodated within the neo-classical framework as short-run phenomena. But these regularities (for instance, wage rounds, job clusters, and orbits of coercive comparison) are specific to time and place, and difficult to distinguish from the possible realizations of one or another stochastic process on the disturbances. Thus, they are not proper "laws of economics."
Nevertheless, the Revisionist type of research can continue to make a very real contribution by providing insightful descriptions of the process by which wages, labor quantities, and characteristics of particular labor market institutions are generated at particular times and places. These descriptions are of analytical value because they identify the characteristics of situations in which particular processes operate; such identification is rarely, if ever, provided by a formal model. Some examples of insightful Revisionist work are the Kerr-Fisher articles "Multiple Employer Bargaining" and "Effects of Environment and Administration on Job Evaluation"; Kerr's "Balkanization of Labor Markets"; Ross's Trade Union Wage Policy; and Lipset, Trow, and Coleman's Union Democracy: The ITU.(7)
Although I see my remarks in this review as amounting to a highly favorable assessment of the Revisionists' achievements, I doubt that many of the Revisionists will accept them as such. They would like their work to be considered a competitor of neo-classical wage theory. Rather than pointing out how little long-run equilibrium propositions limit the patterns of wage rates and labor quantities that might be observed, they insist on interpreting neo-classical theory as implying strong but empirically incorrect statements. This unfortunate disposition is related to the "two cultures" attitude that runs through the Revisionist essays in this volume, and Freeman's as well.
The labor economics of the 1980s is heavily econometric, and too often its practitioners focus on a given data set (usually prepared by others) rather than on the phenomena to which the data refer. Very often these investigators ignore evidence (especially non-quantitative evidence) not contained in the data set and, worse yet, restrict research to topics for which data sets are ready to hand. The result is a technique-oriented subculture generating an esoteric literature that is often irrelevant to policy and insensitive to institutional developments.
That situation is both unfortunate and unnecessary. The Revisionists contributed much to our knowledge of "how labor markets work," and the present generation of technique-dominated labor economists would do well to incorporate certain elements of their research style. In particular, recent labor economics has been woefully deficient in providing analytically descriptive vignettes of interesting cases, a task at which the Revisionists excelled.
But to urge that this deficiency be corrected, and condemn the obsession with methodological refinement that is responsible for it, it is not necessary to attack either economic theory or carefully used econometric tools. In their neglect of technical refinement, both in their own work and especially in that of their thesis advisees, the Revisionists connived at their own decline in influence. The restoration of a proper share of influence will not be promoted by anti-intellectual rejection of the technical developments of the past quarter-century. Those developments must be used where appropriate, and explicitly but respectfully set aside where irrelevant, with emphasis placed on the positive contribution offered by the research findings being presented.
(7) It is no accident that all of these examples are products of the California (Berkeley) Institute of Industrial Relations. I would also draw attention to Lloyd Fisher's "The Harvest Labor Market in California," which elucidates the limitations of a spot market model of wage determination by providing a specific example of such a market at work and describing its very atypical characteristics.