Traditional welfare theory has employed the production function ¹ or production possibility set ² as its central concept for analyzing "the production process." Production is visualized as an instantaneous transformation of one set of material "inputs" into a different set of material "outputs" through the mediation of specific "quantities" of human labor, interestingly enough, measured in units of time. The production possibility set, or the less general production function, are treated as exogenous data describing all possible transformations available at the moment of analysis to production units of an economy. It is a matter of choosing from the set of possible transformations according to some criterion.

In this conception the production process itself disappears inside what has been critically referred to as the "black box" of the production possibility set. For every combination of "inputs" entering the black box a determinate combination of "outputs" emerges. The whys and wherefores are left to other disciplines to consider. The traditional welfare theorist takes the "mappings" as exogenous. ³

Outside the economic mainstream a dramatically different conception of production that stresses the social aspects of the "labor process" has long existed. In this conception the social coordination and direction of human labor occurring in the production processes of any economy is a critical subject for economic analysis. And the logic of these social activities, that is, the dynamics within the neoclassical "black box," cannot be ignored without debilitating consequences.

Marxist economists have long argued that the distinction between "labor" (the work actually performed), and "labor power" (the capacity to do work) is critical to an understanding of different economies. Institutional economists have developed an analysis of segmented labor markets, internal labor markets, and labor-management relations by viewing the firm as an elaborate social organization. Recently, neoclassical theorists have begun to treat the information and incentive complexities of production relations in the theory of contracts and principal agent models.

Adherents of the view that production must be analyzed within a particular social organization of the labor process have argued:

But what is relatively new is the idea that the nontraditional conceptualization has important welfare-theoretic implications-that one or more of the above points implies the traditional conclusion that profitmaximizing behavior of owners will coincide with socially efficient production decisions holds only under highly restrictive and unrealistic assumptions. A host of modern, radical economists have contributed to developing a theory that has come to be known as the "conflict theory of the firm." Harry Braverman, Stephen Marglin, William Lazonick, Kathy Stone, Herb Gintis, Richard Edwards, David Gordon, Michael Reich, James Devine, Michael Burawoy, Andy Zimbalist, David Noble, Michelle Naples, and Sam Bowles, ⁴ among others, have played variations on Marx's theme that there is a fundamental difference between the capacity to work, which the employer pays for, and the actual work that is done. They have greatly refined the argument that the implications of neglecting the conditions under which the former is transformed into the latter are far from trivial.

Besides criticism inspired by a rereading of Marx's analysis of "alienated labor," a body of work referred to as "dual or segmented labor market theory" has emerged from empirical work by institutionalist and labor economists challenging the traditional notion of competitive labor markets. Building on the work of J. R. Commons, Peter Doeringer and Michael Piore ⁵ sparked a new wave of interest pursued by others such as Barry Bluestone, Samuel Rosenberg, Paul Osterman, Martin Carnoy Russell Rumberger, Robert Buchele, Francine Blau, and Lawrence Kahn. Closer to the mainstream, Armen Alchian and Harold Demsetz initiated a literature inspired by Ronald Coase's conception of the firm as a consequence of high market transaction costs. ⁶

Finally, very recent efforts have attempted to apply the analytical tools of principal-agent theory to the firm. Since the principal-agent model was designed for situations in which "principals" contract with "agents" to engage in activities that have different utilities for "principals" and "agents" and about which "principals" and "agents" have different information, the potential for sharpening treatment of the employment relation is obvious. We review these different bodies of literature and leave the task of elaborating their weaknesses and building upon their insights to chapter 8 where we reconsider the institution of private enterprise in light of our new welfare paradigm and theory.

2.1 The Labor Exchange

2.1.1 Karl Marx

Marx himself considered recognition of the difference between "labor" and "labor power" his single most important contribution to economic theory because he believed it solved the mystery of the origins of capitalist profits. By distinguishing between "labor" and "labor power" Marx believed he had perfected the Labor Theory of Value inherited from David Ricardo and had located the origins of profits "in the production process," thereby explaining how positive profits resulted in the context of "equal" (competitive) exchange. This is not the place to dwell on our conviction that by choosing to focus on a presumed "quantitative" difference between the "exchange value" of labor power and labor performed, Marx (and most of his economic disciples since) proceeded down one of the longest dead ends in the history of economic thought. ⁷ However, a number of radical economists have recently returned to exploration of the implications of the qualitative difference between labor and labor power that Marx originally emphasized in his treatment of alienated labor in the 1844 Philosophical Manuscripts. It is the development and relevance of this qualitative distinction that we treat here.

In addition to pointing out the qualitative distinction between the capacity to work, or what he called "labor power," and actual work performed, which he called "labor," in the 1844 Philosophical Manuscripts Marx criticized production as it is carried out under the conditions of private enterprise independently of its "quantitative" outcome, or what he called the degree of "exploitation." According to Marx, regardless of how poorly or well workers were paid, laborers were necessarily "alienated" from the work process, from the products of their work, from each other, and from their "species being" under a capitalist organization of production. The reason was that under capitalism workers do not conceptualize and organize their own laboring activities. The conceptualization and direction of workers' capacities is done by another, either their employer or a manager representing the employer. For this reason, Marx argued, workers under capitalism are inevitably "alienated" from the work process. He further concluded that if they are "alienated" from the work process they must be "alienated" from the "products" of that process as well. Moreover, if they do not control their work efforts individually, they are certainly in no position to coordinate collaborative efforts with other workers, which "alienates" them from their fellow workers. Most generally, Marx argued that the ability to conceive, evaluate, plan, and coordinate the effects of one's laboring capacities with others was the defining characteristic of the human species. Therefore, any economic system that denied people the opportunity to exercise this capacity necessarily "alienated" them from their "species being."

2.1.2 Cornelius Castoriadis

Cornelius Castoriadis was the first to effectively reemphasize the profound difference between what an employer hopes he will get and what he actually buys when he purchases the chance to try and direct someone's productive capabilities for an agreed period. Castoriadis argued not only that the outcome of the capitalist labor exchange remains to be determined by antagonistic struggle between employees and employer over who will direct the formers' productive activities toward what ends, but that, ironically, the very reproduction of the capitalist economic system hinges on neither protagonist in this fundamental contest emerging totally victorious.

Obviously if employees directed their efforts only toward their own ends after receiving their wages-doing whatever suits their inclinations and preferences without regard to orders levied by capitalists-there would be no more buyers of "labor power" because "owning" labor power would give capitalists no advantages. Capitalism would cease to exist.

Less obviously, Castoriadis pointed out that "in real life, capitalism is obliged to base itself on people's capacity for self-organization, on the individual and collective creativity of the producers without which it could not survive a day even though the whole 'official organization' of modern society ignores and seeks to suppress these abilities to the utmost." ⁸

Certainly this observation provides a poignant reminder that much of what occurs during the capitalist labor process does not result from explicit directives from owners and their managerial personnel, but emanates instead from employees' initiatives and responses to unforeseen eventualities arising during the labor process within the context of a specific system of organizational control and motivation. Put differently, for Castoriadis the assumption that employers will get exactly what they hoped they had purchased in the labor market is at odds with an implicit assumption of any analysis of a capitalist economy that the basic institutions of private enterprise market economies will be reproduced.

Regrettably, Castoriadis' work in this area was largely unavailable to Englishspeaking audiences until 1988. As a result, rekindling interest in the qualitative distinction between "labor" and "labor power" among U.S. radical economists fell to Stephen Marglin, who, in an unpublished essay in 1969 titled "What Do Bosses Do?," ⁹ stimulated a veritable flood of research about the historical transformation of the labor process.

2.1.3 Stephen Marglin

Stephen Marglin observed that (1) capitalist production defeated noncapitalist artisan and putting-out production in the rise of English capitalism, and (2) capitalist production utilized a hierarchical system of control. Yet Marglin denied these facts necessarily implied that hierarchical systems of control were technically more efficient than nonhierarchical systems-a conclusion that had long gone unchallenged even by most who argued against hierarchical organization of production on other grounds.

Even should Marglin's historical hypothesis eventually prove insubstantial, it served the important function of undermining the long unchallenged assumption that "the weight of history" had closed the door on any rational questioning of the technological superiority of hierarchical systems of control of the labor process.

If hierarchical control was not (or might not have been) initiated because it permitted an increasing division of labor and greater technological efficiency and thereby allowed capitalists to lower costs and out-compete noncapitalist producers, why was it implemented? Marglin offered a twofold answer:

Without discussing either the validity of Marglin's historical evidence, the weight of contrary historical evidence he did not consider at the time of his article, or the logical consistency of the arguments as he presented them, it is sufficient for us to note that Marglin reopened a debate that had long been closed for a new generation of radical economists by

(1) challenging the conclusion that history proved that new technology made possible by hierarchical control of production explained the victory of early capitalist producers over their noncapitalist rivals, and

(2) suggesting that hierarchical control served other functions besides promoting increased technological efficiency in a capitalist production process

Not surprisingly, members of the new generation of radical economic historians quickly emerged to examine the validity of Marglin's hypothesis in other historical contexts. Kathy Stone studied the transformation of the postCivil War American steel industry, ¹⁰ and a number of other contributors to the journal Radical America examined the function of changing job structures in different historical contexts in a debate that came to be referred to (unfortunately) as "profit versus control."

2.1.4 Harry Braverman

In 1974 the field was restimulated by publication of Harry Braverman's major work, Labor and Monopoly Capital. Unlike Marglin's essay, which was suggestive and highly provisional, Braverman's work was the culmination of a lifetime of research applying Marx's theory of the capitalist labor process to the economic transformations undergone during the era of 'monopoly capitalism" in the U.S. In addition to reiterating the qualitative uniqueness of human labor power and the inherent struggle over the division of "conception" and "execution" that the capitalist form of production implies, which Marx and, more recently, Castoriadis had elaborated, Braverman argued that the system of Taylorism was developed not only to increase the accountability of workers, thereby permitting extraction of a greater quantity of "labor" from the "labor power" purchased, but also to "deskill" employees, thereby permitting capitalists to negotiate lower wage bills for the labor power they purchased.

Moreover, to his great credit Braverman traced recognition of the phenomenon of "de-skilling" back to the eighteenth-century British economist Charles Babbage after whom Braverman dubbed the employer's motivation to de-skill the "Babbage Principle."

In addition to Braverman's own application of this theoretical framework to highlight the contradictory interests of employers and employees over the effects of new technology and organizational systems on both the level and dispersal of skills and on the ability of employers to extract work in a number of industries in the twentieth-century U.S. economy, Braverman's book stimulated others including Michael Burawoy to explore similar matters. ¹¹

The important question for welfare theory is whether application of the Babbage Principle serves the social interest as fully as the private interest of the employer. We can assume Adam Smith's invisible hand of competition compels employers to adopt any technological possibilities that minimize the use of expensive skills. But Braverman argued that if these expensive (and, therefore, presumably productive) skills already exist in the labor force to some extent, underemploying them serves no social purpose. As we shall see in chapter 8, the extent to which private and public interest converge or diverge here must be treated more carefully in a dynamic model that includes education and training of the labor force. Even if Braverman misformulated parts of his critique of capitalist efficiency, he certainly deserves credit for raising the possibility of a divergence between public and private interests in "de-skilling."

2.1.5 Herb Gintis

Herb Gintis most clearly formulated radicals' alternative view of the labor process as a direct challenge to conclusions of traditional welfare theory. In his early work, ¹² Gintis outlined the logic of traditional welfare theory's analysis of the capitalist firm and cataloged a series of empirical anomalies. Gintis pointed out that precisely by modeling the production process as a black box and treating labor as conceptually indistinguishable from nonhuman inputs, traditional welfare theory was able to deduce that profit maximizing production decisions would be Pareto optimal ; wage differentials would reflect differences in the marginal productivity of workers; and job structures would reflect the trade-off in workers' desires for wages and job satisfaction. But contrary to these expectations Gintis cited: numerous experimental and policy-oriented attempts to introduce varying degrees of worker control in production that tended to exhibit increases in both productivity and worker satisfaction; strong evidence that wage differentials within firms correlate weakly with skill differentials among workers and conditions of market supply; and the empirical importance of nontechnical considerations such as sex, race, and demeanor in determining wages.

While one could try to explain these anomalies as results of imperfect information and special circumstances, Gintis argued that a faulty analysis of the labor exchange on the part of traditional welfare theory was responsible for the divergence between theoretical predictions and reality. Because the employer must be concerned with guaranteeing the "integrity" of the labor exchange during production since the exchange cannot be taken as a simple quid pro quo, Gintis reasoned that choice of technology and organization of work would be influenced by concerns other than choosing the technology and organization of work that would maximize output if employees could always be counted on to perform according to employers' desires. Specifically employers would have to take into account to what degree a particular technology would elicit the performance from workers they hoped they had purchased, not just how much output and how agreeable work roles (and, therefore, how little wages) a technology would yield if workers could always be counted on to perform as desired. He summarized the problem as follows:

In later work Gintis and Donald Katzner ¹⁴ constructed a formal model of internal decision making in the firm. As a special case, their model included a treatment of the firm originally envisioned by Coase ¹⁵ and Simon. ¹⁶ Both Coase and Simon recognized authoritative rather than market allocation as the hallmark of the firm as an economic organization. As Coase pointed out as early as 1937:

However, Coase and Simon argued that the existence of these "authoritative allocations" within the firm was no reason to fear that private enterprise, market economies would make socially inefficient decisions. Coase argued that certain allocations are more efficiently executed by authoritative decree than by market transaction. And precisely because the profitmaximizing entrepreneur can be relied on to determine which these are, Coase reasoned that the combination of private enterprise and competitive market institutions generate socially efficient allocations.

Simon, in turn, contributed to a sophisticated defense of the neoclassical treatment of the production process as a black box by making the assumptions of authoritative allocation totally explicit. ¹⁸ He argued that if every worker is capable of performing a set of activities, A, then "work" or "production" can occur in either of two ways. If the worker contracts to perform a specific activity, from A for a price, then he or she is working as an independent agent and the "problem" of authoritative allocation disappears. But the worker may instead contract a fixed wage, w, in return for which he or she turns over to the employer the right to choose any b in B equal to some subset of A defined as part of the labor contract. Simon concluded that utilitymaximizing employees and profit-maximizing employers facing each other in competitive labor markets could be relied on to fix w and B so that the structure of work would be characterized by "producer sovereignty," and the selection of b from B would maximize output for a given set of inputs.

Coase and Simon essentially assumed that "once the contract between entrepreneur and worker, [w,B] is settled, the former tells the latter exactly what to do, [and] the worker either does it or finds another job." ¹⁹ In contrast, Gintis and Katzner argued "contracts are not explicit enough, nor are capitalists' powers sufficiently encompassing to dictate a worker's attitudes, values, relations with superiors, subordinates and co-workers, the manner in which he or she discharges his or her responsibilities, and so on." ²⁰ In other words, Gintis and Katzner claimed it cannot be assumed that the capitalist will be able to choose any b in B. "The worker may allow the capitalist to do this, or the capitalist may somehow induce the worker to perform b, but this is certainly not guaranteed by the contract." ²¹

Therefore, Gintis and Katzner propose a model of the enterprise as an institution in which not only is the employer maximizing profits, but workers are maximizing their own objective functions after signing their contracts subject to indirect constraints employers can place on them in the form of rule, incentive, and information/guidance structures. In their own words, what emerges from a more rigorous treatment than traditional neoclassical analysis is the "fundamental observation ... that except under highly restrictive conditions, the pursuit of profit maximization by the firm may not allocate internal human activity Pareto optimally ." ²²

More specifically, Gintis and Katzner rigorously establish a series of theorems stipulating sufficient conditions for profit-maximizing decisions by employers to yield Pareto optimal outcomes. Although a number of technical complications must be guaranteed as well, any of the following conditions is essentially sufficient:

It is worth noting that Gintis and Katzner do not establish necessary conditions for the equivalence of profit-maximizing decisions with decisions that generate a Pareto optimal organization of the workplace, whose restrictiveness could then be explicitly evaluated. On the other hand, the highly restrictive sufficient conditions they do establish is certainly suggestive. If one must assume complete employer control over worker behavior, or worker preferences that coincide with employer preferences concerning what workers do for their wages, or workers whose only on-thejob desire is to please the immediate supervisor, or workers who will do anything their employer asks for the tiniest monetary bonus, in order for profit maximization to yield Pareto optimal organization of production, how likely is this to occur?

In 1985, the American Economic Review ended more than ten years of "benign neglect" of the radical "conflict school" when it published an article by Samuel Bowles that succinctly explained the logic of many conflict theorists. ²³ Bowles presented the argument in terms familiar to traditional theorists. He demonstrated that if an employee's "objective function" includes some positively valued on-the-job activities (or inactivity) that are associated with a positive opportunity cost in terms of working," ²⁴ individual employer profit-maximizing behavior sometimes requires (1) employing supervisory workers even though they produce nothing themselves, and (2) rejecting technologies that generate more output for ones that produce less.

The logic is quite simple, profit-maximizing employers must hire supervisory staff who produce nothing, as long as their wages are less than the revenues gained from the additional output the supervisors extract from other employees. When choosing between alternative technologies, profit maximizers must forgo technologies yielding greater output in favor of technologies that require smaller supervisory staffs and thereby reduce supervisory costs by more than the loss in revenues. Bowles argued that his simple model demonstrated the essential logic of the conflict theory as he made clear that the conclusions were independent of asymmetric knowledge on the part of employer and employee.

However, these conclusions should come as no surprise to traditional theorists who cannot have been unaware that capitalists employ and pay supervisors who produce nothing themselves. Once it is conceded that employing supervisors capable of producing something, but who, in fact, produce nothing is necessary for profit maximization, all of Bowles' formal modeling and the first two-thirds of his argument become unnecessary because the conclusion of social inefficiency is immediate.

If it could be shown that in an environment which workers perceived to be more fair, or more consistent with their self-respect, for example, they would choose to expend more effort for any given employer strategy, then it is a simple matter to demonstrate that the initial outputs could be produced with unchanged levels of labor effort in production and using less surveillance labor. ²⁶

And Bowles makes explicit the challenge of the conflict school:

In the end, the entire issue is remarkably simple, if there is an alternative social organization of production in which workers have less impulse to shirk, or in which that impulse can be checked at less social cost, private enterprise is demonstrably inefficient. The waste might take the form of employing people who could have been producing output to merely watch over others or using technologies that produce less but reduce surveillance costs. In either case, more waste occurs under conditions of private enterprise production than under some alternative arrangement in which work effort is more readily forthcoming or coerced at lower social cost.

Regrettably, Bowles does not explain further in his American Economic Review article exactly what alternative social organizations of production might reduce the problem of malfeasance or why we should expect them to do so. But from this and other writings it is clear he believes more participatory and egalitarian arrangements would be likely to induce greater work effort at less social cost, ²⁸ and we certainly agree. However, as we explain in chapter 8, a "hard boiled" Neo-Hobbesian might feel uncompelled to accept Bowles' conclusion without further elaboration.

2.1.7 Gil Skillman

The task of applying principal-agent theory to the conflict theory of the firm has only recently begun. In 1985, Gil Skillman ²⁹ reviewed previous contributions and paved the way for further applications. In the principalagent relationship agents are contracted by a principal to perform actions that affect the utility of the latter under conditions where (1) the principal and agents have different preferences regarding decisions such as effort, compensation, or risk; and, (2) the principal and agents have nonidentical information sets concerning agents' potentials or actions.

The central thrust of this literature has analyzed the implications for allocative efficiency and choice of contracts of moral hazard, in which the agents' unseen action affects the probability distribution of the outcome, and adverse selection, in which the agents' actions can be observed, but not evaluated properly. ³⁰ But if we relax traditional assumptions of perfect knowledge of the production possibility set by employers and employees alike, potential for using the theory to gaze inside the "black box" erected by traditional welfare theory around the production process is obvious.

In just the past few years "principal-agent theory" has been used to increase the theoretical rigor of both "radical/conflict" and "institutionalist/malfeasance" analysis of the firm. Sappington analyzed contracts under the condition that agents cannot be forced to work "too hard" in debilitating conditions, exploring implications of the "limited liability condition" for employment contracts. ³¹ Stiglitz analyzed the trade-off between piece rates and wages in a principal-agent relationship. ³² Holmstrom, Sappington, Harris, Raviv, Christensen, and Skillman have analyzed the implications of assuming agents are better informed than the principal about productive parameters. ³³ There have been numerous applications with multiple agents, and Radner and Skillman have initiated models in which the principal-agent relation is analyzed over time. ³⁴

2.2 Economic Discrimination

The question is not whether economic discrimination exists in the U.S., because empirical evidence is overwhelming. ³⁵ The question is not even what the causes may be, since there are clearly various reasons why individuals of different races and sexes, but with comparable initial potentials, receive disparate forms of employment. The question is whether private employment, under competitive conditions, can be expected to ameliorate or aggravate economic discrimination, or if there are reasons to believe it may do both.

Once again, the debate-or more accurately, the lack of productive debate between mainstream and nonmainstream theories of economic discrimination is highly instructive about the influence of the traditional paradigm. First we review traditional neoclassical treatment and then "divide and conquer" theory.

2.2.1 Kenneth Arrow

Kenneth Arrow has provided a beautiful presentation of the logic and coherence of the neoclassical view, and since he goes to considerable lengths to consciously consider the methodological biases of the approach, his treatment is of particular interest. ³⁷

Arrow begins by warning that his discussion is "a programmatic and methodological rather than a confident analysis," ³⁸ and that his "intention is to present the deficiencies of neoclassical analysis" as well as its strengths. He continues:

But after noting that income differentials are obviously linked with other social dimensions, and there is no reason to impose the burden of a full explanation upon economic theory, Arrow elaborates upon the traditional neoclassical theory of economic discrimination. While Arrow accepts the fact that only about half the 35 percent differential that persists between black and white mean earnings in the U.S. can be explained by differences in unemployment rates and supply-side factors, he argues that economic discrimination can still best be explained by the neoclassical theory in which white employees' "taste for discrimination" plays the critical role. His own contribution is to elaborate the concepts of "personnel investment" and "employer information costs" to eliminate important anomalies in previous versions of this neoclassical model.

In the process, Arrow specifically exonerates employers from the list of suspects who might be responsible for economic discrimination. According to Arrow the forces of competition both eliminate employers who would discriminate out of a personal "taste" to do so, and prevent employers from colluding to enhance profits through a "divide and conquer" strategy. We quote Arrow at great length to illustrate how the blind spot in the traditional paradigm for the production process can affect even the most critically self-conscious and brilliant theoretician.

Arrow reasons: "Since it appears that supply considerations can explain only part of the black-white income differential, it is advisable to turn to the demand side." "If we assume away productivity differences between black and white employees, the simplest explanation of the existence of wage differences is the taste of the employer." He then recapitulates the results of modeling an "employer taste for discrimination" hypothesis. If all employers discriminate equally, and if employers' utility depends only on the ratio of the two kinds of workers, it can be shown that black workers lose, white workers gain an equal amount of wages, and employers neither gain nor lose as compared with a nondiscriminatory situation.

At this point Arrow observes, "We have a coherent and by no means implausible account of the economic implications of racial discrimination. In the grossest sense, it accounts for the known facts." Yet he concludes, "Still, I do not find this is satisfactory."

He explains it is not "the excessive generality of utility hypotheses about economic behavior" that troubles him, since this "seems intrinsic in the nature of the case." Nor is the charge that "we have offered no explanation of racial discrimination but simply referred the problem to an unanalyzed realmlike Moliere's intellectual who explained that opium produces sleep because it contains a great deal of the dormitive principle--what leads him to reject "tastes for discrimination" explanations. To this charge he answers, "In a sense, all scientific explanation involves the same process of musical chairs; all we ask is that the explanatory principles have some degree of generality and parsimony."

Instead, Arrow rejects "employer taste" explanations of persistent wage differentials unrelated to productivity differentials because they neglect the influence of "those vast forces of greed and aggressiveness that we are assured and assure students are the mainsprings of economic activity in a private enterprise economy; not the best but the strongest motives of humanity, as Marshall had said." Assuming the trade-off between discrimination and profits is less for some employers than others, "presumably they will take advantage of the gap between black and white wages by demanding the black labor. In the long run, the less discriminatory will either drive the more discriminatory out of business or, if not, will cause the wage difference to fall. If we suppose that there are some actual or potential employers who do not discriminate at all, then the wage difference should, in the long run, fall to zero." In other words, if employer tastes for discrimination were the cause of wage discrimination, competition for profits and investment funds in goods and capital markets should erode these differentials in the long run.

Arrow qualifies this conclusion by observing that even with forces of competition in the capital market to make up for whatever degree of lack of competition might exist in goods markets providing "leeway" for the survival of discriminating employers, it is not necessary to assume that "they are driven out as sharply as might be supposed." It is more that "the price they have to pay for their tastes will depend on the tastes of others in the market" as well as the degree of indeterminacy permitted by the incompleteness of competitive forces. But regardless of their degree of success, Arrow reiterates the key conclusion of neoclassical theory concerning discrimination:

The fundamental point is that the competitive pressures, to the extent that they are decisive, work toward the elimination of racial differences in income, under the usual assumptions of economic theory [emphasis added].

Having concluded that "employer taste" explanations are unpersuasive, Arrow proceeds to apply the same utility analysis to "other members of the productive team." If white workers in higher ranks in the firm dislike working with blacks under them, and if it is assumed that these discriminatory tastes "are determined by the ratio of blacks to whites under them rather than the amounts, it can be shown that in equilibrium the black workers lose, the white workers in lower ranks gain an equal amount, and neither the white higher ranking workers nor the employers gain or lose money income." Similarly, if white workers in lower ranks dislike working with blacks above them, "neutral" employer behavior will yield a black/white wage disparity in the higher echelons.

Moreover Arrow points out: "Even if the wage compensation needed to work with blacks is the same in the two situations, lower working with higher and vice versa, the cost to the employer is much greater in one case than the other simply because there are so many more lower level employees ... [which is] especially interesting, because it explains why more highly educated blacks are more heavily discriminated against."

But the critical reason Arrow finds "employee taste" explanations of discrimination more plausible than "employer taste" models is that he finds competitive pressures more restrictive of employer tastes than white worker tastes.

A model in which white employers and employees were motivated by a dislike of association with blacks as well as more narrowly economic motives would give a satisfactory qualitative account of observed racial discrimination in wages but, at least as far as employers are concerned, it is hard to understand how discriminatory behavior could persist in the long run in the face of competitive pressures [emphasis added].

Seeing no disciplining forces of competition to render "employee taste" explanations implausible, Arrow discusses two such models in detail: the first due to Gary Becker with white and black employees perfect substitutes, and the second an extension of Becker's model to the case of complementary" kinds of labor.

Suppose for a simple model that there is only one kind of utility function [of the white employees] expressing a tradeoff between wages and the proportion of white workers in the labor force of the firm. Any employer can purchase black labor at a fixed price, but for white labor he must choose some point on an indifference curve between wages and the white proportion .... If the wages required by whites for an all-white labor force are lower than black wages, total segregation for whites is optimum for the firm, while in the contrary case an all-black labor force is cheapest .... a general equilibrium with full employment of both types of labor, some firms must be segregated in one direction and some in the other. It would never pay a firm to have a mixed labor force, since they would have to raise the wages of their white workers above the level for the all-white option. The firms would also have to find the two types of segregation equally profitable; otherwise, they would all switch to one or the other. This requires that ... there would be no wage differentials.

The analysis of discriminatory feelings by perfect substitutes has lessons for discrimination by complementary types of labor too. If we suppose that there are black workers available at both higher and lower levels the employer can exploit any racial wage differentials by hiring a labor force that is black at all levels. If the proportions of the different skills in the black labor force are different from those desired, the resulting equilibrium will not equate wages at each level, but there will be a tendency to equate wages on the average.

However, Arrow quickly admits obvious problems with these models. While they "explain" the significantly improbable degree of segregation we find in individual firms, they also predict (1) in the long run discriminatory wage differentials should disappear; and (2) slight variations in black/white wage differentials should stampede firms from one kind of segregation to another. Arrow agrees that neither of these "predictions" correspond to reality. These were the "anomalies" that plagued the traditional neoclassical explanation of economic discrimination. Here we can do no better than let Arrow speak for himself.

One might search for other and more stable explanatory structures, but I know of none that have been proposed or that seem at all credible. I propose that we look more closely at the long run adjustment process.

In particular, as I have already suggested, when dealing with nonconvexities [the fact that firms prefer extreme alternatives to compromises is technically a "failure of convexity"], the adjustment process may have to be very rapid indeed. You must recall that in these circumstances marginal adjustments are punished, not rewarded. If the firm is to gain by a change, it has to go all the way. Intuitively, we are not surprised that a firm will hesitate to scrap its entire labor force and replace it with another. The problem is to give an acceptable formalization of this intuition.

While we seldom find Arrow engaged in such practices, this appears to be attaching an epicycle to an old theory to rationalize an anomaly that would otherwise fester like an open sore. Arrow argues:

This concludes the argument, but the full rescue of neoclassical theory is spelled out as follows:

Having gone to these great lengths to rescue the traditional explanation of economic discrimination as deriving essentially from the prejudices of white employees, Arrow finally explains why he finds a theory based on the "proximate determinant of the demand for labor, the [profit-maximizing] employer's decisions," unacceptable. Again we offer extensive citation to illustrate the pervasive effects of the traditional paradigm.

Thus Arrow rejects what he admits at first appears a more reasonable form of explanation, one in which the perpetrators of the discriminatory acts do so because it is in their own interest.

Our purpose in dwelling at such length on the neoclassical treatment of economic discrimination was to illustrate the detrimental impact of the traditional welfare paradigm on where and how one looks for explanations. In this particular case we find a peerless theorist, far more sensitive than most to the limitations and biases of his chosen approach, struggling to make his theory fit the facts. Our point is not that he does not, in the end, succeed in explaining the anomalies. Ironically our point is precisely that he does succeed!

Arrow provides the best treatment of economic discrimination possible within the traditional paradigm. ⁴² With a black box surrounding the production process it is literally impossible to make competitive employer behavior consistent with purposeful economic discrimination on the part of employers. Hence the necessity of what we find a far-fetched, convoluted explanation. Indeed, it appears Arrow finds it so as well.

It would be difficult to find a better example of a brilliant theorist at disadvantage precisely because of his knowledge. As Arrow states with justified confidence, "Especially when dealing with problems central to economics, the difference in approach between trained economists and others, however able, is enormous." But not only does mastery of theory make possible clarification of truly insightful intuitions, it "selects" what intuitions will receive attention and "channels" their development. For this reason it is possible from time to time for noneconomists to see an economic problem more clearly than "trained economists." Who but "trained economists" I for instance, would have believed in the validity of Say's Law, which asserts the theoretical impossibility of "a general glut" during the century between the writings of David Ricardo and J. M. Keynes? And who but "trained economists" would find their explanation of racial discrimination in a rationalization of what long-run adjustment really means to permit nonconvexities imposed by discriminating tastes of employees to coincide with the observed fact that profit-maximizing employers continue to find it profitable to pay blacks less than whites?

Our argument is not with Kenneth Arrow, who is constantly at pains to avoid the pitfalls of his theoretical training (read, "traditional paradigm"), and who almost always explicitly warns of the dangers of hidden neoclassical presumptions, and to whom we owe much. Our argument is with the paradigm itself, which in this case prevents analysts from dropping the abstractions that create a black box around the production process, for inserting a small quantity of "realism" (read, "one ray of light") inside the black box permits formulation of a direct explanation of continuing discrimination on the grounds that it is profitable for employers even in competitive conditions, that is, an explanation that avoids the pitfalls of assuming a successful conspiracy. It is only the traditional paradigm, with its blind spot inside the black box, that makes the far more "natural starting point, the proximate determinant of the demand for labor, [profit-maximizing] employer's decisions," in Arrow's words, impossible to accept because it is impossible to reconcile with the traditional framework.

2.2.2 Michael Reich

While Michael Reich is not the only one to present a "divide and conquer" explanation of economic discrimination in terms of the "conflict theory of the firm," ⁴³ his is the easiest analysis for us to build upon.

Reich presents a theoretical model ⁴⁴ in which output depends only on "labor done," and in which black and white labor are technically perfect substitutes. However, Reich stipulates that "labor done" is a function not only of the quantity of "labor power" hired, but of the "bargaining power" of employees within the firm after they have signed the labor contract. He further stipulates that the "bargaining power" of employees is a function of the white/black wage ratio and the white/black employment ratio within the firm. In particular, since black and white labor are perfect substitutes and all labor is homogeneous, Reich reasons that by paying whites more than blacks, employers can aggravate racial antagonisms among the work force, thereby weakening workers' bargaining power and enhancing employers' ability to extract "labor done" from "labor power hired." Reich also stipulates that by raising the white/black employment ratio within the firm above the white/black ratio in the work force as a whole, employers can aggravate racial antagonisms, thereby weakening their employees' solidarity.

The employer must choose how many whites and how many blacks to hire and what to pay employees of each race. Assuming that the black wage rate is set in the marketplace, but that individual employers can raise white employees' wage rate above the market rate for blacks, the employer's choice variables can be modeled as: how many blacks to hire, what white/black employment ratio to establish, and what white/black wage ratio to set.

Reich demonstrates that in this model necessary conditions for profit maximization are that individual employers make the white/black wage ratio in their firms greater than one and the white/black employment ratio greater than the whitelblack ratio in the labor force at large. In other words, Reich demonstrates that under his assumptions employers who are competitors in both input and output markets must engage in wage and employment discrimination precisely in order to maximize profits.

In this very simple but elegant "conflict" model of the firm we arrive at exactly the opposite conclusions from Arrow regarding who benefits from discrimination and the relationship of competitive pressures to maximize profits and economic discrimination. For while white employees do better than blacks in Reich's model as well, the competitive employer is a direct beneficiary of his discriminatory actions, and it is impossible to say whether or not white employees would have a lower or higher real wage in the absence of discrimination. The reasons for these results are obvious:

1. It would be profit reducing for employers in Reich's model not to aggravate racial antagonisms as long as doing so raises revenues by increasing the actual labor they extract from the labor power they hire more than it raises costs from the differential paid their white workers.

2. Whether or not white employees receive a lower or higher real wage due to employer discrimination depends on whether the differential money wage they receive is outweighed by the greater amount of work they must perform because of lower bargaining power.

The model abstracts from all extraneous matters: Homogeneous. labor is the only input. Black and white labor are perfect technical substitutes. The model stipulates competitive conditions; no collusion or cartel is necessary to explain how employers benefit from discrimination. Consequently, there is no need to explain, in turn, how the cartel is preserved from competitive pressures to cheat. The only racial condition assumed is the historical legacy of racial mistrust and antagonism between whites and blacks; there is no presumption of "tastes" for discrimination on the parts of employers nor even any assumption that white workers prefer not to work under or over blacks. The model fulfills all of Arrow's wishes for a more direct explanation, based on the self-interest of those who make decisions as those interests are defined by forces of competition for profits.

But to be accepted, the model also requires analysts to treat labor as a qualitatively distinct input. It requires analysts to distinguish between what is paid for in the labor market and the actual result in the labor process. It requires looking inside the black box of the production possibility set and seeing at least the rudimentary forms of a labor process carried out under the social organization of private enterprise. In sum, it requires analysts to do the one thing that is most difficult for them-to step outside the traditional welfare paradigm.

2.3 Segmented Labor Markets

In the late 1960s and early 1970s interest and empirical research rekindled along institutional lines originally pioneered by J. R. Commons concerning the structure of internal and external labor markets. What is most interesting from our perspective was the new light this work shed on possible reasons for the appearance of "segmented labor markets." Joan Robinson had developed a coherent analysis of segmented labor markets in terms of differential elasticities of supply of different laboring groups as early as the 1930s. ⁴⁵ But the work of a new generation of institutionalists investigating segmented labor markets implied an alternative explanation: development of internal job ladders (or what was confusingly termed internal labor markets) within large firms might contribute to segmentation of external markets for labor. On the one hand, an explanation was sought on the supply side, and the traditional paradigm was no obstacle to developing the analysis. On the other hand, analysts looked to the demand side, and a detailed vision of the dynamics of the production process inside the traditional black box was required.

2.3.1 Peter Doeringer and Michael Piore

Doeringer and Piore were credited for sparking a body of work that came to be known as the "dual" or "segmented" labor market school. ⁴⁶ Inherent in their approach was the idea that segmented labor markets were due more to employers'behavior than objective conditions of labor supply. Difficulty in even conceiving the demand-side explanation within the traditional paradigm is sufficient to demonstrate our point. But it is worth pointing out empirical anomalies of the supply-side explanation. If empirical evidence were totally consistent with a supply-side explanation of segmented labor markets, the theoretical failure of the traditional paradigm would, perhaps, be of little practical consequence. But explanation in terms of differential supply elasticities of workers implies that the group with the smaller elasticity of supply would receive the lower wage. While there may be reason to think that econometric studies have misspecified what are, in fact, demand-side influences as differential supply elasticities of black and female labor, the case remains that a wide variety of careful empirical studies conclude that the labor supply of blacks and women is more, not less, elastic than that of white men. ⁴⁷

In any case, the logic of the demand-side explanation that was invisible to those immersed in the traditional paradigm is as follows: Employers have an incentive to develop internal job ladders with pay rates that vary from marginal revenue products and pay rates for equivalent labor categories in external markets. The incentive derives from the internal ladder's usefulness in extracting effort and diminishing employee solidarity-nonproblems for the traditional paradigm but very real problems for employers. This entails two consequences.

1. If much worker mobility occurs within the job ladders of large firms rather than between firms, ⁴⁸ employers can create a degree of monopsony power for themselves with internal job ladders even if external labor markets are competitive.

2. If participation in the economy of large firms with internal job ladders is significant, over time internal structures will necessarily shape demand structures of external labor markets since the latter are ultimately nothing more than aggregates of the internal structures of all firms. Hence arises a demand-side explanation of segmented external labor markets.

We will use our new welfare theory to further explore the logic of such a demand-side explanation of segmented external labor markets and their relation to internal job ladders, more carefully specifying the conditions under which such phenomenon would presumably exist, in chapter 8.

2.4 Influence of the Traditional Paradigm

Without doing them too great an injustice, we can organize nontraditional treatments of the labor process into "schools" that variously combine four types of insight or "themes":

Different previous treatments of the labor process emphasize different combinations of these themes. And certainly all four themes have important consequences that deserve to be explored. However, as we will explain in chapter 8, traditional theory can pose appealing rebuttal to these criticisms as they have been presented. On the other hand, we believe the criticisms we have reviewed in this chapter are well taken and apply even in the most competitive conditions.

We will find the implications of the four themes taken alone are limited, whereas the criticisms can be better defended, and the consequences are more far reaching when presented within the new welfare paradigm and theory we develop in part 2. Specifically, combining the above themes in a multiperiod model with a fifth theme that has been little emphasized in nontraditional treatments to date yields a more theoretically solid critique of private enterprise production. The fifth theme, which emerges directly from the alternative paradigm we develop in chapter 5, can be stated as follows:

While the full argument must await future chapters, it is important to note here the debilitating influence of the traditional paradigm on a careful analysis of the production process under private enterprise. The traditional paradigm ignores the problem of getting employees to "honor" the wage contract, offers no vision of human development, defines no concepts identifying "human characteristics" and "human outputs" of production, and offers no treatment of social relations among people in different and similar roles in production. All these omissions blind traditional theorists to important aspects of the production process necessary for advancing our understanding of discrimination, labor market structure, and the efficiency of private enterprise. Instead of a welfare theoretic analysis of the production process we have had a vacuous production possibility set, that is, a nontheory of production. ⁴⁹

As Abba Lerner once said, "An economic transaction is a solved political problem. Economics has gained the title of queen of the social sciences by choosing solved political problems as its domain." ⁵⁰ Which is perfectly fine if we are analyzing an economic arena in which all the political problems have truly been solved. But after signing the labor contract, important "political" problems still remain. And particularly in a multiperiod analysis of production, among the important remaining problems is determination of what individual and group characteristics employees will have when they face their employer in future negotiations over the wage rate and over how much "integrity" the labor contract will have.