Working Class Power, Capitalist Class Interests, and Class Compromise
Erik Olin Wright
Department of Sociology
University of Wisconsin
Madison, Wisconsin 53706
April 1999
draft 7.1

Acknowledgment note: I would like to thank Samuel Bowles, Andrew Mitchner, Michael Burawoy, Joel Rogers and Robert Boyer for their extremely constructive criticisms of earlier drafts of this paper.

The concept of "class compromise" invokes three quite distinct images. In the first, class compromise is an illusion. Leaders of working class organizations -- especially unions and parties -- strike opportunistic deals with the capitalist class which promise general benefits for workers but which, in the end, are largely empty. Class compromises are, at their core, one-sided capitulations rather than reciprocal bargains embodying mutual concessions.

In the second image, class compromises are like stalemates on a battlefield. Two armies of roughly similar strength are locked in battle. Each is sufficiently strong to impose severe costs on the other; neither is strong enough to definitively vanquish the opponent. In such a situation of stalemate the contending forces may agree to a "compromise": to refrain from mutual damage in exchange for concessions on both sides. The concessions are real, not phoney, even if they are asymmetrical. Still, they don't constitute a process of real cooperation between opposing class forces. This outcome can be referred to as a "negative class compromise".

The third image sees class compromise as a form of mutual cooperation between opposing classes. This is not simply a situation of a balance of power in which the outcome of conflict falls somewhere between a complete victory or a complete defeat for either party. Rather, here there is a possibility of a non-zero-sum game between workers and capitalists, a game in which both parties can improve their position through various forms of active, mutual cooperation. This outcome can be called a "positive class compromise."

The basic objective of this paper is to explore the theoretical logic of positive class compromises and propose a general model of the conditions conducive to them in developed capitalist societies. The paper will not attempt a systematic empirical investigation, although empirical illustrations will be used to clarify elements of the model. The premise of the analysis is that so long as capitalism in one form or another is the only historically available way of organizing an economy, a positive class compromise -- if it is achievable -- will generally constitute the most advantageous context for the improvement of the material interests and life circumstances of ordinary people. If one is interested in advancing such interests, therefore, it is important to understand the conditions which facilitate or hinder the prospects for positive class compromise

The central argument I will make is that the possibilities for stable, positive class compromise generally hinge on the relationship between the associational power of the working class and the material interests of capitalists. The conventional wisdom among both neoclassical economists and traditional Marxists is that in general there is an inverse relationship between these two variables: increases in the power of workers adversely affect the interests of capitalists (see Figure 1). The rationale for this view is straightforward for Marxist scholars: since the profits of capitalists are closely tied to the exploitation of workers, the material interests of workers and capitalists are inherently antagonistic. Anything which strengthens the capacity of workers to struggle for and realize their interests, therefore, negatively affects the interests of capitalists. The conventional argument by neoclassical economists is somewhat less straightforward, for they deny that in a competitive equilibrium workers are exploited by capitalists. Nevertheless, working class associational power is seen as interfering with the efficient operation of labor markets by making wages harder to adjust downward when needed and by making it harder for employers to fire workers. Unions and other forms of working class power are seen as forms of monopolistic power within markets, and like all such practices generate monopoly rents and inefficient allocations. As a result, unionized workers are able to extort a monopoly rent in the form of higher wages at the expense of both capitalists and nonunionized workers.

-- Figure 1 about here --

This paper explores an alternative understanding of the relationship between workers' power and capitalists' interests: instead of an inverse relationship, this alternative postulates a curvilinear reverse-J relationship (see Figure 2).(1) As in the conventional wisdom, capitalist class interests are best satisfied when the working class is highly disorganized, when workers compete with each other in an atomized way and lack significant forms of associational power. As working class power increases, capitalist class interests are initially adversely affected. However, once working class power crosses some threshold, working class associational power begins to have positive effects on capitalist interests. As we shall see in more detail below, these conditions allow for significant gains in productivity and rates of profit due to such things as high levels of bargained cooperation between workers and capitalists, rationalized systems of skill upgrading and job training, enhanced capacity for solving macro-economic problems, and a greater willingness of workers to accept technological change given the relative job security they achieve because of union protections. The upward-bending part of the curve, where increases in working class power have positive effects on capitalist class interests, generates conditions for positive class compromise. The goal of this paper, then, is to elaborate a general theoretical model of the causal processes underlying the relation presented in figure 2.

-- Figure 2 about here --

Section 1 briefly defines the core concepts used in the analysis and discusses a number of methodological issues. Section 2 situates the problem of positive class compromise within a broader literature on inter-class cooperation, labor relations and economic governance. Section 3 then frames the problem of class compromise in terms of various possible game theory models of the interactions of workers and capitalists. With this game theoretic background, section 4 elaborates a general theoretical model and underlying mechanisms for the reverse-J model of positive class compromise. The paper concludes in section 5 with a somewhat speculative discussion of the impact of globalization on the prospects of class compromise.

1. Core Concepts and methodological issues

None of the concepts used in this paper have transparent meanings. In particular, the concepts of "class", "interests", and "power" are all highly contested. I will not attempt to elaborate analytically precise definitions of any of these concepts here, but some brief clarifications are necessary.(2)


Class and its related concepts -- class structure, class struggle, class formation, class compromise -- can be analyzed at various levels of abstraction. For some purposes it is important to deploy a highly differentiated class concept which elaborates a complex set of concrete locations within class structures. My work on the problem of the "middle class" and "contradictory locations within class relations" would be an example of such an analysis (Wright, 1985, 1997). For some problems the causal processes cannot be properly studied without specifying a range of fine-grained differentiations and divisions within classes on the basis of such things as sector, status, gender, and race. For other purposes, however, it is appropriate to use a much more abstract, simplified class concept, revolving around the central polarized class relation of capitalism: capitalists and workers. This is the class concept we will use in most of this paper.

In a stylized Marxian manner, I will define capitalists as those people who own and control the capital used in production and workers as all employees excluded from such ownership and control. In this abstract analysis of class structure I will assume that these are mutually exclusive categories. There is no middle class as such. No workers own any stock. Executives, managers and professionals in firms are either amalgamated into the capitalist class by virtue of their ownership of stock and command of production, or they are simply part of the "working class" as employees. Of course this is unrealistic. My claim, however, is that this abstract, polarized description of class relations in capitalism can still be useful in clarifying real mechanisms that actual actors face and is thus a useful point of departure for developing a theory of class compromise.(3) This is not to prejudge the question of the extent to which our conclusions will hold when a more complex class concept is introduced at lower levels of abstraction. What I do assume, however, is that it will be easier to pose the question of the problem of class compromise for more complex class models once we have clarified the simpler problem, which, in any case, is itself quite complex. Most of this paper, therefore, will work from a very simple class analysis in which there are only two classes -- capitalists who own capital assets and control access to employment opportunities, and workers, who own labor power assets and sell their labor power to capitalists. Initially we will treat these two classes as internally homogeneous. When we discuss the effects of globalization at the end of the paper, the problem of heterogeneity within classes will be introduced.


Throughout this paper, our attention will be restricted to what can be narrowly termed the "material interests" of people by virtue of their class location, or what I will refer to in shorthand as "class interests".(4) In general, I will make two radical simplifying assumptions about the nature of these interests: first, that class interests can be reduced to a single quantitative dimension so that one can talk about the extent to which the interests of the members of a class are realized; and second, that all people in a given class location share the same class interests. Both of these assumptions are problematic when we study concrete capitalist societies.

In terms of the first assumption, for both workers and capitalists, there are quite heterogeneous interests linked to their class locations: workers have class interests with respect to such things as working conditions (including the enjoyment and meaning of work), earnings, leisure, and security, and all of these interests have multiple time-horizons -- short term interests, interests in long-term prospects, interests in the material welfare of one's children, etc. Similarly for capitalists, there are interests in the rate of profit, interests in growth, interests in stability, and each of these have both short term and long term aspects.(5) Nevertheless, in this paper I will not attempt to deal systematically with the complexity posed by the multi-dimensionality of interests of either capitalists or workers. When I refer generically to "capitalist class interests" and "working class interests", therefore, I mean some unspecified amalgam of these various quantitative considerations, weighted towards the middle-term.(6)

In terms of the second assumption, both workers and capitalists are internally heterogeneous as social categories, and this heterogeneity also impacts on their material interests. Capitalists are differentiated by sector, geographical scope of transactions (local, national, multinational capital), and many other factors; workers are differentiated by such things as sectors, skills, and the nature of the firm in which they are employed, not to mention race, gender and other non-class forms of social division. The material interests of real workers and capitalists are certainly affected by these dimensions of internal differentiations. Nevertheless, for the analytical purposes of the models we will explore in this paper, most of the analyses will treat these class categories as internally undifferentiated and thus ignore the problem of internally heterogeneous interests.


Like "interests", "power" is used in many different ways in social theory. In the context of class analysis, power can be thought of as the capacity of individuals and organizations to realize class interests. Insofar as the interests of people in different classes -- say workers and capitalists -- are opposed to each other, this implies that the capacity of workers to realize their class interests depends in part on their capacity to counter the power of capitalists. Power, in this context, is thus a relational concept.

In this paper our concern is mainly with what I will term working class "associational" power -- the various forms of power that result from the formation of collective organizations of workers. This includes such things as unions and parties, but may also include a variety of other forms, such as works-councils or forms of institutional representation of workers on boards of directors in schemes of worker co-determination, or even, in certain circumstances, community organizations. Associational power is to be contrasted with what can be termed "structural power" -- power that results simply from the location of workers within the economic system. The power of workers as individuals that results directly from tight labor markets or from the strategic location of a particular group of workers within a key industrial sector would constitute instances of structural power.(7) While such structural power may itself influence associational power, it is on associational power as such that we will focus.

The models we will examine do not directly concern the role of associational power of capitalists in the formation of class compromise. As the literature on neocorporatism has pointed out, there are certain institutional settings of class compromise in which the associational power of employers plays a pivotal role (Streeck and Schmitter, 1985; Streeck 1992; Pontusson, 1997), and there will be places in the discussion where reference will be made to the role of such associations. Our concern, however, is not with capitalist associational power as such, but with the ways in which working class associational power impacts on the interests of capitalists.

Sites of Class Compromise

Class struggle and compromise do not occur within an amorphous "society", but within specific institutional contexts -- firms, markets, states. The real mechanisms which generate the reverse-J curve in figure 2 are embedded in such institutional contexts. Three institutional spheres(8) within which class struggles occur and class compromises are forged are particularly important:

There is a rough correspondence between different kinds of working class collective organizations and each of these institutional spheres of class conflict and class compromise: labor unions are the characteristic associational form for conflict/compromise in the sphere of exchange; works councils and related associations are the characteristic form within the sphere of production; and political parties are the characteristic form within the sphere of politics. This correspondence, however, is only an "elective affinity". In particular, trade unions can be deeply and directly involved in both production and politics. When unions mobilize members to vote in electoral campaigns or when they use their resources to lobby politicians, they constitute a form of working class associational power in the political sphere, whether or not they are closely aligned with a political party. Similarly, when unions bargain at the firm level over conditions within the labor process and play an active role in shop-floor governance through things like grievance procedures and health and safety committees, they constitute a form of associational power within production. For simplicity of exposition I will generally assume the rough correspondence of types of association with spheres of social interaction, but the important theoretical issue is the arena within which associational power is formed and operates, not the specific form that such organizations take.

The central task of our analysis, then, is to examine the mechanisms which enable these different forms of working class associational power -- unions, works councils, parties -- to forge positive class compromises within the spheres of exchange, production, and politics.

Strategy of Analysis

I will not attempt in this paper to develop formal mathematical models of the relationships we will be discussing. As an alternative, the relationships will be presented as less formalized graphical representations as in figure 2.

Two comments on the shape of the curves we will be examining are needed. First, throughout this paper I will draw relationships like the one in Figure 2 as smooth curves that are more or less symmetrical in the regions of their troughs. Because the metrics on the axes of these graphs have no natural units, there is no particular reason to assume that the curves actually have precisely this shape. A reverse-J curve as in Figure 2, for example, could plummet much more steeply to the left of the trough and rise much more slowly to the right. The two sides of the curve could even be straight lines, meeting at a sharp disjuncture at the trough, although it seems less plausible that the real causal processes in the world would have such a disjoint "tipping point". The specific representation of these relationships as smooth, symmetrical curves, therefore, is theoretically arbitrary.

Secondly, I have drawn Figure 2 as a reverse-J rather than a U-shaped curve. This reflects a theoretical stance towards the underlying nature of capitalism which conceptualizes capitalist class relations in terms of domination and exploitation. If one accepts this characterization of capitalist class relations, then class compromises are compromises between inherently antagonistic interests rather than simply different or complementary interests. There is thus always at least a latent threat to the class interests of capitalists posed by high levels of working class associational power. This suggests that while capitalists may benefit from a highly organized working class relative to a moderately organized working class, in general capitalists would still prefer an atomized and disorganized working class to a highly organized one . This implies that the curve rises to a higher level on the left of the trough than on the right and thus has the reverse-J shape.(9) While most of the arguments which I will advance in this paper do not hinge on the assumption of the reverse-J curve asymmetry, I will represent the relations in this way because of the general theory of class relations within which the analysis is embedded.

2. Situating the concept of Class Compromise(10)

In the most abstract and general terms, class compromise -- whether positive or negative -- can be defined as a situation in which some kind of quid pro quo is established between conflicting classes in which, in one way or another, people in each class make "concessions" in favor of the interests of people in the opposing class. The "compromise" in class compromise is a compromise of class-based interests -- members of each class give up something of value. Class compromise is thus always defined against a counterfactual in which such concessions are not made. Typically this is a situation in which the use of threats, force and resistance plays a more prominent active role in class interactions.

Defined in this way, the idea of class compromise is closely linked to Gramsci's (1971) concept of "hegemony". Gramsci uses the concept of hegemony to distinguish two general conditions of capitalist society. In a nonhegemonic system, capitalist class relations are reproduced primarily through the direct, despotic use of coercion. In a hegemonic system, in contrast, class relations are sustained in significant ways through the active consent of people in the subordinate classes. Coercion is still present as a background condition -- hegemony is "protected by the armor of coercion" in Gramsci's famous phrase -- but it is not continually deployed actively to control people's actions. To quote Przeworski (1985: 136), "A hegemonic system is, for Gramsci, a capitalist society in which capitalists exploit with consent of the exploited." For hegemony to be sustained over time, there must be, in Przeworski's (1985: 133-169) apt expression, "material bases of consent". This, in turn, requires some sort of class compromise: "....the fact of hegemony presupposes that account be taken of the interests and the tendencies of the groups over which hegemony is to be exercised, and that a certain compromise equilibrium should be formed - in other words, that the leading group should make sacrifices of an economic-corporate kind." (Gramsci, 1971: 161).

Gramsci developed the concept of class compromise in only a sketchy and fragmented form. The scholar who has most systematically and rigorously elaborated this concept is Adam Przeworski. Przeworski makes the central quid pro quo of class compromise explicit:

Przeworski's formulation here is close to what I have called "negative class compromise" in so far as he emphasizes the abstention of workers from levels of militancy which would interfere with the production of profits in exchange for material concessions by capitalists. Elsewhere, he explores the positive face of class compromise in his analysis of how Keynesianism, backed by organized labor and social democratic parties in the advanced capitalist countries in the post-World War II period, expanded aggregate demand in ways which ultimately benefitted capital as well as labor (Przeworski, 1985: 205-211). The model of class compromise which I develop in this paper can be viewed as an extension and reformulation of Przeworski's core idea through the elaboration of this positive side of class compromise.

Before examining the details of this model, it will useful to situate it within a broader array of alternative treatments of the problem of cooperation and compromise among class actors. Figure 3 organizes this conceptual space along two dimensions: first, whether the strategic basis of class compromise is primarily individual strategies or associational power, and second, whether the form of class compromise is primarily negative or positive. The four categories generated by these two dimensions constitute distinctive ways in which interclass cooperation and compromise can be generated.

-- Figure 3 about here--

The paradigmatic case of negative class compromise grounded in individual strategies is so-called "efficiency wages".(11) An efficiency wage is a wage premium - a wage above the equilibrium "market-clearing wage" - paid by an employer as part of a strategy to reduce shirking on the part of employees. As elaborated by Bowles (1985) and Bowles and Gintis (1990, 1998: 36-9), building on the earlier work of Alcian and Demsetz (1972), Shapiro and Stiglitz (1984), Ackerloff and Yellen (1986), Oliver Williamson (1985) and others, employers face a problem of the "extraction of labor effort" from workers -- getting workers to work harder than they want to do spontaneously -- since the labor contract is neither transparent nor costlessly enforceable. The level of effort expended by workers in a capitalist firm is a function of the costs to the worker of losing a job and the effectiveness of the firm's ability to monitor labor effort (and thus detect shirking). Employment rents (the difference between an efficiency wage and the market clearing wage) save on monitoring costs because they increase the cost of job loss to workers and thus reduce the amount of shirking for any given level of monitoring. Employers thus face a trade-off between spending more money on improving the effectiveness of monitoring or paying higher employment rents. The magnitude of this trade-off will vary depending on a variety of factors: the nature of the technology, the complexity of interdependencies among workers in the division of labor, the visibility of effort levels. Such efficiency wages are a form of negative class compromise insofar as the higher wages are an alternative to more purely coercive strategies by employers in the face of strategies of resistence (shirking) by individual workers.

Investigations of institutional arrangements like efficiency wages, especially by economists, generally do not explicitly analyze them in class terms. From the point of view of standard neoclassical economics, such arrangements are simply profit-maximizing strategies of employers designed to minimize the transactions costs associated with the inherent human tendency, in Williamson's (1985 :47) expression, for people to be "self-interest seeking with guile". These arrangements are treated as top-down, employer-initiated solutions to problems of securing optimal conditions for the performance of individual contracts, not for solving a problem in class relations.(12)

Bowles and Gintis, in contrast, firmly locate the problem of efficiency wages in the class relations of capitalist production. They argue that the extent to which efficiency wages are a solution to incentive problems depends not only on technical properties of the division of labor, but also, crucially, upon the nature of the class relations within which that division of labor is organized. In firms in which workers were also owners, employment rents would be less necessary to elicit cooperation (i.e. labor effort) both because mutual monitoring of labor effort by workers would be more effective and because the normative expectations of workers would underwrite higher effort levels.(13) The employment rent in capitalist firms should therefore be viewed as a response to the specific class dynamics of monitoring labor effort, not simply universal individual propensities to shirk. In these terms, efficiency wages constitute concessions by employers to enhance interclass cooperation, albeit mainly in response to the actions of individuals within those relations rather than the associational power of workers.

Positive class compromises can also emerge out individual strategic interactions between employers and workers. Perhaps the best example is internal labor markets.(14) Although, as in the case of efficiency wages, internal labor markets may increase the effectiveness of negative sanctions by employers since workers within internal labor markets have more to lose if they are disciplined or fired (Bartlett, 1989: 135-7), most analyses of internal labor markets emphasize the ways they are designed to elicit active cooperation rooted in loyalty and commitment of the individual to the interests of the organization. This is one of the central themes in the extensive literature on Japanese work organization (Dore, 1973; Ouchi, 1981 Aoki 1988), but has also figured in the broader analysis of the internal organization of capitalist firms (Williamson, 1985; Foulkes, 1980, Sorenson, 1994) and even some more abstract analyses of class relations as such.(15) Again, as in efficiency wage theory, the point of departure is a problem faced by employers. Here the most critical issues are typically the problem of training and retaining employees with portable skills and of eliciting responsible, creative contributions of skilled workers and managers to the firm. By linking a person's career to the fate of a firm over a long time horizon, internal labor markets help generate these kinds of loyalties. Insofar as these institutional arrangements enhance interclass cooperation between individuals and involve concessions by capitalists in the form of more rigid employment practices (job security, seniority arrangements, etc) than they would ideally like, this can be considered a form of positive class compromise.

Most discussions of class conflict and class compromise pay relatively little attention to these forms of interclass compromise, positive or negative, generated by the strategies of individuals. Rather, they focus on the ways class compromises are forged through class struggles rooted in class-based associational power. Analyses of negative class compromise emerging from class struggle are particularly prominent in the Marxist tradition. If the interests of workers and capitalists are inherently antagonistic and polarized, then it would appear that whatever compromises emerge from class struggle simply reflect balances of power between contending forces. A typical example is given by David Kotz (1994:55), writing within a strand of Marxist theory known as the "social structures of accumulation perspective":

Such views, however, are not restricted to Marxists. John R. Commons' conception of collective bargaining, for example, is essentially a conception of negative class compromise. Commons (1970 [1950]) felt that in developed capitalist economies dominated by large corporations, union recognition and collective bargain were essential for working people to be able to engage in a process of fair "equal exchange" with capitalists.(16) He also felt that it was in the interests of capitalists to accept collective bargaining for the alternative was class war and disruption which would, in the end, only lead to despotism. Ultimately, Commons felt, collective bargaining was the only viable alternative to military dictatorship: Unions might be in the general "public interest," but Commons does not claim that unionization and collective bargaining as such are directly beneficial to capitalists: "The unions and administrative commissions were organized to restrain corporations, also in the public interest, from abuse of their corporate power over individuals." (Commons, 1970 [1950]; 132)

Negative class compromise is unquestionably an important phenomenon, but it is only part of the story. The central idea n the concept of positive class compromise is that unions and other forms of workers' associational power can, under appropriate circumstances, confer positive benefits on capital as well as workers. Although the expression "positive class compromise" is not used in the literature, this idea is at the core of the large literature on social democracy and neo-corporatism (e.g. Korpi, 1983; Soskice, 1990; Esping-Anderson, 1990) and considerable recent work in economic sociology that focuses on the problem of the economic performance of different capitalist economies (e.g. Streeck and Schmitter, 1985; Kenworthy, 1995; Gordon, 1996; Crouch and Streeck, 1997). As Rogers and Streeck (1994:130) put it: "The democratic left makes progress under capitalism when it improves the material well-being of workers, solves a problem for capitalists that capitalists cannot solve for themselves, and in doing both wins sufficient political cachet to contest capitalist monopoly on articulating the 'general interest.'"

The classic form of this argument is rooted in the Keynesian strand of macroeconomic theory. The argument is familiar:

Full employment, insofar as it implies high levels of capacity-utilization and higher aggregate demand for the products of capitalist firms, potentially serves the interests of capitalists. But it also risks a profit squeeze from rapidly rising wages and spiraling levels of inflation. Keynes himself recognized this as a serious problem: "I do not doubt that a serious problem will arise as to how wages are to be restrained when we have a combination of collective bargaining and full employment" (cited in Glynn, 1995: 37). Kalecki (1943:356), in a famous article entitled "The Political Aspects of Full Employment" felt that these pressures would require innovative institutional solutions: "Full employment capitalism will, of course, have to develop new social and political institutions which will reflect the increased power of the working class. If capitalism can adjust to full employment, a fundamental reform will have been incorporated in it."

The emergence and consolidation in a number of countries of strong, centralized unions capable of imposing wage restraint on workers (and on employers) was perhaps the most successful solution to this problem. In this sense, a powerful labor movement need not simply constitute the basis for a negative class compromise, extracting benefits for workers through threats to capital. If a labor movement is sufficiently disciplined, particularly when it is articulated to a sympathetic state, it can positively contribute to the realization of capitalist interests by helping to solve macroeconomic problems.

Positive class compromise is not limited to the Keynesian preoccupations of full employment and wage restraint. Increasingly attention is being paid to the importance of class compromise within production itself. A good example is Streeck's (1992) research on "diversified quality production" (DQP), the high value-added production generally identified with "high road" capitalist development. DQP is a demanding form of capitalist production and requires high levels of cooperation within production, effective mechanisms for skill formation and continual skill upgrading of workers, flexibility in reorganizing labor processes as technology changes, and so on. All of these conditions, Streeck argues, depend upon fairly high levels of social peace and trust, which in turn depend upon stable, long-term employment relations.

The question, then, is how these underlying conditions can be produced and maintained. There are two main possibilities: through voluntary actions by employers or through constraints imposed on them by unions and the state. Of these, Streeck (1992:21) argues, the latter are likely in general to be more effective:

In setting these constraints, the associational power of workers plays a critical role, for it underwrites institutions that: David Soskice (1990: 190-91, 198-9) makes a similar point about the crucial role of unions in fostering sustainable labor-management workplace cooperation: Again, workers associational power can help capitalists solve certain problems which they find it difficult to solve themselves, in this case within the organization of production itself.(17)

Most discussions of positive class compromise are built on a relatively binary contrast between the economic problems in situations without such a class compromise and the economic gains in situations where positive class compromise occurs. The general claim is that moving from the former to the latter is a "win-win" situation in which all classes benefit. As Glyn (1995:33) puts it, the Keynesian class compromise "was a rare example of a true Pareto improvement." Relatively few studies explicitly posit the kind of curvilinear relationship between workers power and capitalist interests illustrated in Figure 2.

The best known empirical study to explore the curvilinear relationship between workers power and capitalist interests is Calmfors and Driffill's (1988) study of the effects of union centralization on economic performance(18) (see also Pohjola, 1992; Freeman, 1988; Calmfors and Driffill, 1993; Garrett, 1998: 26-50; Rowthorn, 1992). Following Mancur Olson's (1982) suggestion that "organized interests may be most harmful when they are strong enough to cause major disruptions but not sufficiently encompassing to bear any significant fraction of the costs for society of their actions in their own interests" (Calmfors and Driffill, 1988: 15), they demonstrate that among 18 OECD countries, during the period 1963-1985, economic performance measured in a variety of ways was best among those countries with either highly centralized or highly decentralized wage bargaining structures, and worst in the intermediary countries.(19) A similar result, using different kinds of indicators, is found in Hicks and Kenworthy's (1998) study of the impact of various forms of cooperative institutions on economic performance. They observe a strong curvilinear relationship between union density and real per capita GDP growth for the period 1960-1989 in 18 OECD countries, indicating that countries with either low or high union density had higher growth rates during these three decades than countries with middling levels of union density.(20)

The rest of this paper will attempt to elaborate theoretically this curvilinear model of positive class compromise. We will begin, in the first part of next section, by framing the problem through a game theoretic perspective on strategic conflicts between workers and capitalists and then turn the problem of the mechanisms which generate the curvilinear relationship between workers power and capitalist interests..

3. Strategic Games and Class Compromise

In order to analyze the relationship of working class associational power to capitalist class interests and class compromise, we must first more rigorously understand the strategic contexts for the conflicts of interests of workers and capitalists. We will do this by exploring a series of stripped-down game theory models based on a highly simplified picture of class conflict in which workers and capitalists each face a binary strategic choice: to cooperate with the other class or to actively oppose it's interests. Because the actors in this game have qualitatively different roles in the system of production, the meaning of "cooperate" and "oppose" are different for each. As summarized in Figure 4, for workers to cooperate with capitalists means that they work hard and diligently in order to maximize the capitalists' rate of profit. Workers rely primarily on market mechanisms (changing jobs) as a way of expressing dissatisfaction with pay or working conditions; while they may have collective associations (unions), they do not engage in active struggles to collectively pressure capitalists for improvements; nor do they engage in political struggle to advance workers interests against those of capitalists. To oppose capitalists is to struggle against them, individually and collectively, in order to raise worker incomes and enhance the extent to which workers control their own labor effort, and thus to minimize the extent to which capitalists exploit and control workers. This includes political struggles to expand worker rights and their capacity to organize collective associations. For capitalists, cooperation with workers means paying workers as much as is possible compatible with maintaining a rate of profit sufficient to reproduce the firm; accepting workers' organizations (unions and parties) and responding to worker demands over working conditions; and moderating their own consumption in favor of employment-generating investment. To oppose workers' interests means paying them as little as possible, given market and technological constraints; getting as much labor as possible out of workers; and resisting worker organizations. As in the case of workers, such opposition includes political action such as opposing unemployment benefits and welfare safety nets that raise the reservation wage and supporting restrictive labor laws that impede unionization. Taking these two alternatives for each class yields the four possible configurations of class conflict presented in Figure 4. In terms of these alternatives, "class compromise" constitutes the situation in which both classes agree to cooperation (C,C).(21)

- Figures 4 and 5 about here --

Figure 5 presents a variety of alternative ways in which the interests of workers and capitalists (their pay-offs to alternative strategic combinations) might be affected by these four combinations of cooperation and opposition. Model I can be called a Unilateral Capitalist Domination Game. Here, the best outcome for capitalists is (C,O): workers cooperate with capitalists (working hard, not organizing, etc.), and capitalists oppose workers (pay them only what the market dictates, oppose collective organization, etc.). The second best outcome for capitalists is mutual opposition, (O,O). In this game, capitalists are sufficiently powerful relative to workers that they can punish workers at relatively little cost to themselves when workers organize against them. Workers are thus worse off under (O,O) than under unilateral workers cooperation (C,O). Struggle doesn't pay. In this game, therefore, (C,O) will be the equilibrium outcome: capitalists are always better off opposing workers, and given that capitalists oppose workers, workers are better off cooperating with capitalists.

Model II represents the standard Marxist view of class conflict in which the interests of workers and capitalists are treated in a purely inverse relation as a zero-sum Pure Conflict Game. The optimal situation for capitalists is that they oppose the interests of workers while workers cooperate with them (C,O). The second best situation for capitalists is mutual cooperation (C,C). This, however, is less advantageous for workers than is mutual opposition (O,O). In the traditional Marxist view, because the interests of workers and capitalists are strictly polarized, it is always better for workers to struggle against capitalists -- to actively oppose capitalist interests -- then to willingly cooperate. The C,C solution, in effect, is an illusion: "cooperative" capitalists, the argument goes, treat workers only marginally better than capitalists who actively oppose workers, but cooperative workers are much less able to force their employers to make concessions than are oppositional workers. Above all, when working class associations actively cooperate with capitalists they weaken their capacity for mobilization and, ultimately this invites capitalists to oppose workers interests, thus leading C,C to degenerate into C,O. The O,O option, therefore, generally promises a better long term payoff for workers than does the C,C option. As a result of such struggles there will be moments when capitalists indeed do make concessions to workers as a result of these struggles -- grant them pay raises, improve working conditions, etc. These concessions are at best a negative class compromise -- concessions in the face of struggle. For both classes in this game, opposition is better than cooperation regardless of what the other class does, and thus the equilibrium will be mutual opposition (O,O) -- active forms of class struggle.(22) The class struggle is much more like trench warfare with occasional victories and defeats for each combatant, and perhaps periods of relatively stable balances of forces underwriting a negative class compromise.

Model III is the standard Prisoner's Dilemma Game. This is a game with symmetrical payoffs for the two classes:(C,C) is the second best outcome for each class and (O,O) is the third best outcome. Unlike in Model I, mutual opposition is now costly to capitalists. This implies that workers have sufficient power to be able to punish capitalists within class struggles. Unlike in Model II, however, workers are better off in mutual cooperation than mutual opposition. Both classes are thus better off if they cooperate with each other than if they mutually oppose each other. Still, if this were a one-shot game, in standard PD fashion the equilibrium outcome would be (O,O) since both classes could improve their payoffs by defecting from the mutual cooperation outcome. If this is a repeated game, as it would be in the real world of class interactions, then the outcome is less determinate. As Axelrod (1984) and many others have shown, in an iterated prisoner's dilemma, mutual cooperation can be a stable solution depending upon the ways opposition in future rounds of the game is used to punish players for noncooperation in earlier rounds. As the possibility of a stable C,C solution occurs, then positive class compromise also becomes possible.

Model IV is a standard Assurance Game: for both classes the optimal solution is mutual cooperation and unilateral cooperation is worse than mutual opposition. Unless there is reasonable confidence that the other class will cooperate, therefore, mutual cooperation is unlikely to occur. If class conflict was an assurance game, the failures of cooperation would primarily reflect a lack of enlightenment on the part of actors -- they simply don't know what's good for them. This corresponds to the views of a certain kind of naive liberalism, where conflict is always seen as reflecting misunderstanding among parties and "win-win" solutions are always assumed to be possible. A strict Assurance Game of this form is unlikely in capitalist economies since if capitalists can get full cooperation from workers without having to make any concessions - the (C,O) outcome - they would seem to have little incentive to cooperate in return. Nevertheless, there may be situations in which the C,C pay-off moves in the direction of an Assurance Game, and certainly situations in which the gap for both classes between C,C and O,O becomes very large.

Finally, Model V, the Unilateral Workers Domination Game, is the symmetrical model to Model I. Here workers are sufficiently strong and capitalists sufficiently weak, that workers can force capitalists to unilaterally cooperate, including forcing them to invest in ways that enhance future earnings of workers (thus making O,C preferable to C,C). This corresponds to the theoretical idea of democratic socialism: an economy within which workers effectively dominate capitalists.(23)

Lurking in the background of the models in Figure 5 is the problem of power: the balance of power between workers and capitalists can be thought of as determining which of these strategic games is being played. As illustrated in Figure 6, as working class power increases from extremely low levels (and thus as the ability of workers to impose sanctions on capitalists increases), the (O,O) alternative in Model I shifts downward and then to the right. This shifts the configuration in the direction of Model II in which working class militancy becomes sustainable and the possibility of negative class compromise -- a class compromise based on the balance of force -- emerges. Further increases in working class power begin to move the (C,C) option to the right creating the prisoner's dilemma of Model III. This sets the stage for the possibility of positive class compromise. If working class associational power can push (C,C) in Model III in an upward direction towards the north-east quadrant approaching the Assurance Game in Model IV, the possibility for a positive class compromise increases: the gains from stable, mutual cooperation increase. This is the central game-theoretic logic underlying the argument developed in this paper: as working class power increases, the unilateral capitalist domination game is initially shifted to a pure conflict game making negative class compromise possible; with further increases in working class associational strength the strategic environment can shift towards an iterated prisoners dilemma opening the prospect for positive class compromise. The more the game shifts towards an assurance game -- even though it is unlikely to actually become one - the more stable the possibility of positive class compromise will become. Underlying this double shift is thus the problem of the relation of working class associational power to the interests of capitalists to which we now turn.

Figure 6 about here

4. A curvilinear model of positive class compromise

To the extent that increases in working class power can contribute not merely to the realization of working class material interests, but also to the realization of some capitalist class interests, then class compromises are likely to be more stable and beneficial for workers; to the extent that every increase in working class power poses an increasing threat to capitalist class interests, capitalist resistance is likely to be more intense, and class compromises, even if achieved, are likely to be less stable. The intensity of class struggle, therefore, is not simply a function of the relative balance of power of different classes, but also of the intensity of the threat posed to dominant interests by subordinate class power.

If the relationship between workers' power and capitalists' interests were the simple inverse relationship of Figure 1, then class compromises would always be relatively fragile and vulnerable to attack, for capitalist interests would always be served by taking advantage of opportunities to undermine workers power. Negative class compromise -- the class compromise of a prisoners dilemma -- would be the most one could achieve. If the shape of the relationship is as pictured in Figure 2, on the other hand, then class compromise can potentially become a relatively durable feature of a set of institutional arrangements. In general, when class conflict is located in the upward sloping region of this curve, class compromises are likely to be both more stable and more favorable for the working class. If the shape of this curve assumes the form of a more U-shaped version of a reverse-J (i.e. if the upward sloping section becomes more symmetrical), then conditions for class compromise can be said to be more favorable; if the reverse-J degenerates into a strictly downward sloping curve, then the conditions for class compromise become less favorable.

In order to more deeply understand the social processes reflected in the reverse-J hypothesis of Figure 2, we need to elaborate and extend the model in various ways. First we will examine more closely the underlying causal mechanisms which generate this curve. Second, we will extend the range of the figure by examining what happens at extreme values of working class associational power. Finally, we will examine various ways in which the institutional environment of class conflict determines which regions of this curve are historically accessible as strategic objectives.

4. A curvilinear model of positive class compromise

To the extent that increases in working class power can contribute not merely to the realization of working class material interests, but also to the realization of some capitalist class interests, then class compromises are likely to be more stable and beneficial for workers; to the extent that every increase in working class power poses an increasing threat to capitalist class interests, capitalist resistance is likely to be more intense, and class compromises, even if achieved, are likely to be less stable. The intensity of class struggle, therefore, is not simply a function of the relative balance of power of different classes, but also of the intensity of the threat posed to dominant interests by subordinate class power.

If the relationship between workers' power and capitalists' interests were the simple inverse relationship of Figure 1, then class compromises would always be relatively fragile and vulnerable to attack, for capitalist interests would always be served by taking advantage of opportunities to undermine workers power. Negative class compromise -- the class compromise of a prisoners dilemma -- would be the most one could achieve. If the shape of the relationship is as pictured in Figure 2, on the other hand, then class compromise can potentially become a relatively durable feature of a set of institutional arrangements. In general, when class conflict is located in the upward sloping region of this curve, class compromises are likely to be both more stable and more favorable for the working class. If the shape of this curve assumes the form of a more U-shaped version of a reverse-J (i.e. if the upward sloping section becomes more symmetrical), then conditions for class compromise can be said to be more favorable; if the reverse-J degenerates into a strictly downward sloping curve, then the conditions for class compromise become less favorable.

In order to more deeply understand the social processes reflected in the reverse-J hypothesis of Figure 2, we need to elaborate and extend the model in various ways. First we will examine more closely the underlying causal mechanisms which generate this curve. Second, we will extend the range of the figure by examining what happens at extreme values of working class associational power. Finally, we will examine various ways in which the institutional environment of class conflict determines which regions of this curve are historically accessible as strategic objectives.

Mechanisms underlying the reverse-J relation

The reverse-J curve presented in Figure 2 can be understood as the outcome of two kinds of causal processes -- one in which the interests of capitalists are increasingly undermined as the power of workers increases, and a second in which the interests of capitalists are enhanced by the increasing power of workers. These are illustrated in Figure 7. In broad terms, the downward sloping curve reflects the ways in which increasing power of workers undermines the capacity of capitalists to unilaterally make decisions and control resources of various sorts, while the upward sloping curve reflects ways in which the associational power of workers may help capitalists solve certain kinds of collective action and coordination problems. The specific mechanisms which generate the component curves in Figure 7 can be differentiated across the three institutional spheres within which class compromises are forged: exchange, production, and politics. These mechanisms are summarized in Figure 8.

-- Figures 7 and Figure 8 about here -

The sphere of exchange

Capitalists have a range of material interests within the sphere of exchange that bear on their relationship with the working class: minimizing labor costs; having an unfettered capacity to hire and fire without interference; selling all of the commodities they produce; having a labor force with a particular mix of skills in a labor market that provides predictable and adequate supplies of labor. As has often been argued by both Marxists and nonMarxist political economists, some of these interests contradict each other. Most notably, the interest of capitalists in selling commodities means that it is desirable for workers-as-consumers to have a lot of disposable income, whereas capitalists' interest in minimizing their own wage bill implies an interest in paying workers-as-employees as little as possible.

Increases in working class associational power generally undermine the capacity of individual capitalists to unilaterally make decisions and allocate resources within labor markets. In the absence of unions, capitalists can hire and fire at will and set wages at whatever level they feel is most profitable. Of course, this does not mean that employers set wages without any constraints whatsoever. Their wage offers will be constrained by the tightness or looseness of the labor market, the reservation wages of workers, and, as discussed earlier, the need to pay workers a sufficiently high wage to motivate individual workers to work diligently. Capitalists' decisions are thus always constrained by the actions of individual workers and by general economic conditions. The issue here, however, is the extent of constraint on capitalists imposed by the collective action of workers reflecting their associational power in various forms. Such associational power reduces capitalists' individual capacity to make profit-maximizing decisions on labor markets and thus hurts their material interests.

If capitalists' interests within the sphere of exchange consisted entirely of interests in their individual ability to buy and sell with minimal constraint, then something close to the inverse relation portrayed in Figure 1 would probably hold. But this is not the case. The material interests of capitalists -- their ability to sustain a high and stable rate of profit -- depends upon the provision of various aggregate conditions within the sphere of exchange, and these require coordination and collective action. The solution to at least some of these coordination problems can be facilitated by relatively high levels of working class associational power.(24)

The classic example of this is the problem of inadequate aggregate demand for the consumer goods produced by capitalists. This is the traditional Keynesian problem of how raising wages and social spending can underwrite higher levels of aggregate demand and thus help solve "underconsumption" problems in the economy. Inadequate consumer demand represents a collective action problem for capitalists: capitalists simultaneously want to pay their own employees as low wages as possible and want other capitalists to pay as high wages as possible in order generate adequate consumer demand for products. High levels of unionization, in effect, prevent individual firms from "defecting" from the cooperative solution to this dilemma. Working class strength can also contribute to more predictable and stable labor markets. Under conditions of tight labor markets where competition for labor among capitalists would normally push wages up, perhaps at rates higher than the rate of increase of productivity thus stimulating inflation, high levels of working class associational power can also contribute to wage restraint (see Calmfors and Driffill, 1988; Glynn 1995; Pontusson, 1997). Wage restraint is an especially complex collective action problem: individual capitalists need to be prevented from defecting from the wage restraint agreement (i.e. they must be prevented from bidding up wages to workers in an effort to lure workers away from other employers given the unavailability of workers in the labor market), and individual workers (and unions) need to be prevented from defecting from the agreement by trying to maximize wages under tight labor market conditions. Wage restraint in tight labor markets, which are important for longer term growth and contained inflation, is generally easier where the working class is very well organized, particularly in centralized unions, than where it is not.

These positive effects of workers strength on capitalist interests in the sphere of exchange need not imply that capitalists themselves are equally well organized in strong employers associations, although as the history of Northern European neo-corporatism suggests, strongly organized working class movements tend to stimulate the development of complementary organization on the part of employers. In any case the ability of workers power to constructively help solve macro-economic problems is enhanced when capitalists are also organized.

Assuming that the positive Keynesian and labor market effects of working class power are generally weaker than the negative wage-cost and firing discretion effects, the combination of these processes yields the reverse-J relationship for the sphere of exchange in Figure 8.

The sphere of production

A similar contradictory quality of the interests of capitalists with respect to workers occurs within the sphere of production: on the one hand, capitalists have interests in being able to unilaterally control the labor process (choosing and changing technology, assigning labor to different tasks, changing the pace of work, etc.), and on the other hand, they have interests in being able to reliably elicit cooperation, initiative and responsibility from employees.

As working class associational power within production increases, capitalists' unilateral control over the labor process declines. This does not mean that capitalists are necessarily faced with rigid, unalterable work rules, job classifications, and the like, but it does mean that changes in the labor process need to be negotiated and bargained with representatives of workers rather than unilaterally imposed. Particularly in conditions of rapid technical change, this may hurt capitalist interests.

On the other hand, at least under certain social and technical conditions of production, working class associational strength within production may enhance the possibilities for more complex and stable forms of cooperation between labor and management. To the extent that working class strength increases job security and reduces arbitrariness in managerial treatment of workers, then workers' time horizons for their jobs are likely to increase and along with this their sense that their future prospects are linked to the welfare of the firm. This in turn may contribute to a sense of loyalty and greater willingness to cooperate in various ways.

The German case of strong workplace-based worker organization built around works councils and co-determination is perhaps the best example. Streeck describes how codetermination and works councils positively helps capitalists solve certain problems:

This tighter coupling of interests of labor and capital with the resulting heightened forms of interclass cooperation helps employers solve a range of concrete coordination problems in workplaces: more efficient information flows within production (since workers have more access to managerial information and have less incentive to withhold information as part of a job-protection strategy); more efficient adjustments of the labor process in periods of rapid technological change (since workers are involved in the decisionmaking and are thus less worried that technological change will cost them their jobs, they are more likely to actively cooperate with the introduction of new technologies ); more effective strategies of skill formation (since workers, with the most intimate knowledge of skill bottlenecks and requirements, are involved in designing training programs). Most broadly, strong workplace associational power of workers creates the possibility of more effective involvement of workers in various forms of creative problem-solving.(25)

With so many positive advantages of such cooperative institutions, it might seem surprising that strong workplace associational power is so rare in developed capitalist countries. The reason, as I have argued throughout this paper, is that such cooperative advantages come at a cost to capital. Streeck recognizes this even in the German case:

Above all, codetermination carries with it considerable costs in managerial discretion and managerial prerogatives.....Integration cuts both ways, and if it is to be effective with regards to labor it must bind capital as well. This is why codetermination, for all its advantages, is seen by capital as a thoroughly mixed blessing.....Both the short-term economic costs and the long-term costs in authority and status make the advantages of codetermination expensive for the capitalist class, and thus explains the otherwise incomprehensible resistance of business to any extension of codetermination rights. (Streeck 1992: 1965)

Because of these costs, capitalists in general will prefer a system of production in which they do not have to contend with strong associational power of workers in production. Thus, the reverse-J shape of the functional relation between workers' power and capitalists' interests within production.

The sphere of politics

The two components of the reverse-J relationship between working class associational power and capitalist interests are perhaps most obvious in the sphere of politics. As a great deal of comparative historical research has indicated, as working class political power increases, the capitalist state tends to become more redistributive: the social wage increases and thus the reservation wage of workers is higher; taxation and transfer policies reduce income inequality; and in various ways labor power is partially decommodified.(26) All of these policies have negative effects on the material interests of high-income people in general and capitalists in particular. Working class political power also tends to underwrite institutional arrangements which increase working class power within the sphere exchange and often within the sphere of production as well. Working class associational power in the political sphere, therefore, may also contribute to the downward sloping curves in the spheres of exchange and production.

The upward sloping class compromise curve in the sphere of politics is the central preoccupation of social democracy. The large literature on tripartite state-centered corporatism is, in effect, a literature on how the interests of capitalists can flourish in the context of a highly organized working class (Esping-Anderson, 1990; Schmitter and Lembruch, 1979; Schmitter, 1988). Sweden, until the mid-1980s, is usually taken as the paradigm case: the social democratic party's control of the Swedish state facilitated a set of corporatist arrangements between centralized trade unions and centralized employers' associations that made possible a long, stable period of cooperation and growth. The organizational links between the labor movement and the social democratic party were critical for this stability, since it added legitimacy to the deals that were struck and increased the confidence of workers that the terms of the agreement would be upheld in the future.(27) This made it possible over a long period of time for Swedish capitalism to sustain high capacity utilization, very low levels of unemployment, and relatively high productivity growth. State-mediated corporatist anchored in working class associational strength in the political sphere played a significant role in these outcomes.

The inventory of mechanisms in Figure 8 provides a preliminary set of variables for characterizing the conditions of class compromise within different units of analysis across time and space. Class compromises within the sphere of exchange can occur in local, regional, national labor markets, or within labor markets linked to particular sectors. Production level compromises typically occur within firms, but they may also be organized within sectors.(28) Class compromises in the sphere of politics are especially important within the nation state, but local and regional political class compromises are also possible. The emergence of various forms of meso-corporatism involving local and regional levels of government may indicate the development of political class compromises within subnational units. The reverse-J curves that map the terrain of class compromise, therefore, can be relevant to the analysis of class compromises in any unit of analysis, not simply entire countries.

Different countries, then, will be characterized by different combinations of values on these three pairs of class compromise curves.(29) In Germany, for example, working class associational power is especially strong within the sphere of production, somewhat less strong in the sphere of exchange, and rather weaker in the sphere of politics. In Sweden -- at least in the heyday of social democracy -- it has been very strong in the spheres of exchange and politics, and perhaps a bit weaker in the sphere of production. In the United States, working class associational power as dwindled within all three spheres, but is strongest in the sphere of exchange within certain limited sectors. The overall reverse-J curve for class compromise within a society, therefore, is the result of a complex amalgamation of the component curves within each of these spheres.

Making the model more complex: extending the theoretical domain of variation

The range of variation in Figures 2 and 8 can be considered the typical spectrum of possibilities in contemporary, developed capitalist societies. It will be helpful for our subsequent analysis to consider what happens when working class power increases towards the limiting case of society-wide working class organization and solidarity simultaneously in all three spheres of class compromise. This corresponds to what might be termed "democratic socialism," where socialism is not defined as centralized state ownership of the means of production but as working class collective control over capital.

What happens to capitalist class interests as working class associational power approaches this theoretical maximum? Figure 9 presents the relationship between one crucial aspect of capitalists' interests -- their control over investments and accumulation (allocation of capital) -- and working class power. The control over investments is perhaps the most fundamental dimension of "private" ownership of the means of production within capitalism.(30) In most capitalist societies even as working class power increases, this particular power of capital is not seriously eroded. Even with strong unions and social democratic parties, capitalists still have the broad power to disinvest, to choose their individual rate of savings, to turn their profits into consumption or allocate them to new investments, etc. Of course, all capitalist states have capacities to create incentives and disincentives for particular allocations of capital (through taxes, subsidies, tariffs, etc.). And in special circumstances "disincentives" can have a significant coercive character, effectively constraining capitalists' capacity to allocate capital. Still, this fundamental aspect of capitalist property rights is not generally threatened within the normal range of variation of working class power. When working class associational power approaches its theoretical maximum, however, the right of capitalists to control the allocation of capital is called into question. Indeed, this is the heart of the definition of democratic socialism -- popular, democratic control over the allocation of capital. This suggests the shape of the curve in Figure 9: a relatively weak negative effect of working class power on capitalist interests with respect to the control over the basic allocation of capital until working class power reaches a very high level, at which point those interests become seriously threatened.(31)

-- Figure 9 about here --

When Figure 9 is added to Figure 2, we get the roller-coaster curve in Figure 10.(32) There are two maxima in this theoretical model: the capitalist utopia, in which the working class is sufficiently atomized and disorganized to give capitalists a free hand in organizing production and appropriating the gains from increased productivity without fear of much collective resistance; and the social democratic utopia, in which working class associational power is sufficiently strong to generate high levels of corporatist cooperation between labor and capital without being so strong as to threaten basic capitalist property rights. These two maxima, however, constitute quite different strategic environments for workers and capitalists. Statically, capitalists should only care about where they sit on the vertical axis of this figure: if you draw a horizontal line through the figure that intersects the curve at three places, capitalists should be statically indifferent among these three possibilities. Understood dynamically, however, capitalists in general will prefer points in the left hand region of the curve.

--Figure 10 about here--

It is at least in part because of this threat of a society-wide shift in the balance of class power that capitalists might prefer for working class associational power to remain to the left of the social democratic "peak" of this curve even though this peak might be theoretically advantageous to capitalist interests. Arriving at the peak looks too much like a Trojan Horse: small additional changes in associational power could precipitate a decisive challenge to capitalists interests and power. The local maximum of the "social democratic utopia" in Figure 10 may thus be a kind of tipping point which is seen by capitalists as too risky a zone to inhabit. This is one interpretation of the strident opposition by Swedish capitalists to the initial formulation of the "wage-earners fund" proposal in Sweden in the 1970s. The wage earners fund, as initially conceived, was a proposal through which Swedish unions would gain increasing control over the Swedish economy via the use of union pension funds to purchase controlling interests in Swedish firms. From the point of view of economic performance and even the middle-run profit interests of Swedish firms, it was arguable that this might be beneficial for Swedish capital, but it raised the possibility of a long-term slide towards democratic socialism by significantly enhancing the power of Swedish labor. The result was a militant attack by Swedish capital against the Social Democratic party. As Glynn (1995:53-4) writes: "The policies which the Social Democrats were proposing impinged on the authority and freedom of action of business which was supposed to be guaranteed in return for full employment and the welfare state. This seems to lie at the root of the employers repudiation of the Swedish model, of which full employment was a central part."
The different regions of this curve correspond to the different game theory models in Figure 5. The "Capitalist Utopia" corresponds to the unilateral capitalist domination game in which (C,O) is the equilibrium solution. The downward sloping region in the center of the figure is the pure conflict game where, at best, negative class compromise is possible. The upper sloping part of the curve is the iterated prisoner's dilemma, where a stable C,C solution, a positive class compromise, can emerge. The apex of this region of the curve, the "Social Democratic Utopia", is the point which is closest to an assurance game. If in fact it actually became a proper assurance game (i.e. the C,C payoff in figure 5 moved into the NE quadrant of the pay-off matrix) then the curve in Figure 10 would become a J-curve rather than a reverse-J; the "social democratic utopia" would be higher than the "capitalist utopia" and become a kind of social democratic nirvana in which mutual cooperation between classes was self-reinforcing, no longer resting on a background condition of potential working class opposition. Finally, democratic socialism corresponds to the uilaterial working class domination game in which O,C is the equilibrium solution.

Working Class interests and the class compromise curve

The models in Figure 5 contain both working class interests and capitalist class interests. Figure 11 adds working class interests to the class compromise curve in Figure 10. The different regions of these curves can be thought of as specific hypotheses about the effects of marginal changes of working class power on the relationship between workers interests and capitalists interests:

-- Figure 11 about here --

1. The gap between workers interests and capitalist interests is greatest at the ends of the spectrum: when working class associational power is weakest (the fully atomized working class) or at the maximum strength (democratic socialism).

2. Increases in working class associational power steadily increase the realization of working class material interests up to relatively high levels of associational power. Of course, in actual historical processes of increasing working class power it may well happen that there will be episodes in which the resistence of capitalists results in declines in the realization of working class interests. Nevertheless, in general, increasing workers power is expected to improve the realization of working class interests.

3. The region of the curve around the "liberal democratic trap" is the region corresponding to the mutual opposition (O,O) pay-off in Models II and III , Figure 5: workers effectively oppose capitalist interests and capitalists effectively oppose workers interests.

4. There is one region of the curve where the functional relation between workers power and class interests has the same general shape for both workers and capitalists: the upward sloping section to the right of the liberal democratic trough. This is the region of maximally stable class compromise.

5. As working class power extends beyond corporatist associative practices, the immediate realization of working class interests again decline. This region of the curve defines the "transition trough" between capitalism and socialism discussed by Adam Przeworski (1985). Capitalists respond to the threat of losing control over the allocation of capital by disinvesting, shifting investments to other places, or by more organized forms of a "capital strike". This has the effect of provoking an economic decline which hurts workers' material interests. It is only when workers associational power increases to the point at which investments can be democratically allocated (in the sense of democratically-imposed direction on allocation) that the working class interest curve once again turns upward. Once there is a full realization of hypothetical democratic socialism, the interests of workers and capitalists are once again maximally divergent.

One more complexity: zones of unattainability

In the practical world of real capitalist societies, not all values within this theoretically defined range are historically accessible. There are two different kinds of exclusion-mechanisms which have the effect of narrowing the range of real possibilities. These can be termed systemic exclusions and institutional exclusions.

Systemic exclusions define parts of the curve that are outside the limits of possibility because of the fundamental structural features of the social system. Specifically, the presence of a constitutionally secure democracy removes the fully repressed and atomized working class part of the curve from the historical stage, and the presence of legally secure capitalist property rights removes the democratic socialism part of the curve. This does not mean that there are no historical circumstances in which these zones of the curve might become strategically accessible, but to get there would require a fundamental transformation of the underlying social structural principles of the society.

Institutional exclusions refer to various kinds of historically variable institutional arrangements, formed within the limits determined by the systemic exclusions, which make it difficult or impossible to move to specific regions of the curve. For example, restrictive labor law can make it difficult to extend working class associational power towards the corporatist associative practices part of the curve (Rogers, 1990). On the other hand, generous welfare state provisions which render workers less dependent on capital, and strong associational rights which facilitate unionization may make it difficult to move towards the right-wing managerialist region. Such institutional exclusions, of course, are themselves the outcomes of historical conflicts and should not be viewed as eternally fixed. But once in place, they help to define the range of feasible strategy immediately open to actors, at least until the time when actors can effectively challenge these institutional exclusions themselves.

These two forms of exclusion are illustrated in Figure 12. The central region of the curve defines the space that is immediately accessible strategically. To use a game theory metaphor adopted by Alford and Friedland (1985), this is the domain of ordinary politics, of liberal vs conservative struggles over "plays" within a well-defined set of institutional "rules of the game". The other regions of the curve become the objects of politics only episodically. Reformist vs reactionary politics are struggles over the rules of the game that define institutional exclusions; revolutionary vs counter-revolutionary politics are struggles over the systemic constraints that define what game is being played.(33)

-- Figure 12 about here --

In Figure 12, the "zones of unattainability" defined by the systemic and institutional exclusions symmetrically span the tails of the theoretical curve of possibilities. There is no reason, of course, to believe that the real world is this neat. Indeed, one of the reasons for introducing this complexity is precisely to provide tools for understanding forms of variation across time and place in these exclusions. This historical variability is illustrated in Figure 13 which compares the United States and Sweden in the periods of most stable Swedish social democracy and American liberal democracy.

-- Figure 13 about here --

Systemic exclusions in the United States and Sweden are roughly comparable: both have structurally secure democratic states and capitalist property relations. Where they differ substantially is in the nature of the historically variable institutional exclusions which confront their respective working classes.

In the United States, a variety of institutional rules create a fairly broad band of institutional exclusions to the right of the central trough of the curve. Electoral rules which solidify a two-party system of centrist politics and anti-union rules which create deep impediments to labor organizing all push the boundary of this zone of institutional exclusion to the left.(34) On the other hand, such things as the weak welfare state, the very limited job protections afforded workers, and laws which guarantee managerial autonomy all have the effect of narrowing the institutional exclusions centered around right-wing managerialist anti-associational practices. The band of accessible strategy in the United States, therefore, affords very little room to manoeuver for labor and keeps working class associational practices permanently lodged on the downward sloping segment of the curve to the left of the trough.

Swedish institutional exclusions, particularly during the most stable period of Social democracy, all work towards facilitating working class associational power. Labor law is permissive, making it quite easy to form and expand union membership, and the generous welfare state and job protections significantly reduce the scope of right-wing managerialist strategies. The result has been that the Swedish labor movement has for a long time been located on the upward sloping section of the curve to the right of the trough.

Actors living within these systems, of course, do not directly see this entire picture. To the extent that the institutional exclusion mechanisms have been securely in place and unchallenged for an extended period of time, they may become entirely invisible and the parts of the curve which they subsume may become virtually unimaginable. From the vantage point of actors within the system, therefore, the range of "realistic" possibilities may look like those portrayed in Figure 14 rather than Figure 13. The American labor movement faces a terrain of possibilities which places it chronically on the defensive. Every marginal increase of workers strength is experienced by capitalists as against their interests, so whenever the opportunity arises, capitalists attempt to undermine labor's strength. Anti-union campaigns are common and decertification elections a regular occurrence. In Sweden, at least until recently, the institutionally delimited strategic environment is much more benign for workers. The central pressure on capitalists has been to forge ways of effectively cooperating with organized labor, of creating institutional spaces in which the entrenched forms of associational power of workers can be harnessed for enhanced productivity. This need not imply that employers actively encourage enhanced working class associational power, but it does suggest less sustained effort to undermine it.

-- Figure 14 about here --

The immediately accessible strategic environments of workers struggles for associational power as illustrated in Figure 14 should not be viewed as fixed by an unalterable historical trajectory. The range of attainable possibilities can change, both as the result of conscious political projects to change institutional exclusions and as the result of dynamic social and economic forces working "behind the backs" of actors. Institutional exclusions are created by victories and defeats in historically specific struggles; they can potentially be changed in a similar fashion. But equally, dynamic changes within economic structures can potentially change the shape of the curve itself. It is to that issue which we now turn in a more speculative manner.

5. Transformations of the terrain of class compromise

If the general model of class compromise we have been exploring is reasonably on target, then this suggests that the prospects of positive class compromise can be altered through three different routes:

The first of these involves the functional relations depicted in Figures 7 and 8, the second involves the institutional "zones of unattainability" in Figure 12, and the third concerns the specific location within a strategic space as in Figure 14. A full-blown theory of class compromise, then, would provide an account of the causal processes which generate these three kinds of effects.

I cannot offer such an elaborated theory. What I will do in what follows is propose some relatively speculative hypotheses about the way certain developments in contemporary capitalism are affecting these three elements in the model of class compromise. Specifically I will focus on the possible impact of increasing international competition and globalization of capital on the shape of the curve, and the impact of technological change on the associational power of workers.

Effects on the shape of the curves

For purposes of understanding the changing conditions for positive class compromise, the critical part of the curve in Figure 7 is the upward sloping segment in which working class associational power positively helps capitalists solve various kinds of collective action and coordination problems. Figure 8 listed a range of interests of capitalists that are facilitated by increasing working class associational power within the spheres of exchange, production and politics. The question I will examine, then, is how globalization and increasing international competition might affect the relationship between workers' power and capitalist interests within each of these three spheres. Figure 15 presents a set of tentative hypotheses concerning these effects. All of these hypothesized effects are highly speculative. My purpose here is less to present a well-worked out argument for the patterns of change, then to offer some preliminary suggestions which may stimulate constructive discussion of the issues.

Consider first the effects of globalization on the sphere of exchange. One of the standard arguments in discussions of globalization is that the increasing mobility of financial capital and globalization of markets has undermined the "Keynesian" solutions to macro-economic problems in advanced capitalism. To the extent that the market for the commodities of capitalist firms are increasingly global, the realization of the economic value of those commodities depends less upon the purchasing power of workers in the countries within which those firms are located. Furthermore, heightened international competition and the constant threats by employers to move production abroad has served to reduce wage pressures thus reducing the positive effect of strong unions on the problem of wage restraint. The positive effects on capitalist interests of strong, centralized labor unions has thus probably been reduced by globalization. While there may still be some positive value for capitalists of a strong labor movement in terms of collective action problems of predictable, well ordered labor markets, especially with respect to the problem of skill formation, nevertheless it seems that, on balance, globalization is likely to depress the positive slope of the first curve of Figure 15.(35)

-- Figure 15 about here --

Globalization may have quite different effects in the sphere of production. The characteristic form of working class associational power within production are works councils and other forms of organized workers representation within the process of production. As already noted, strong works councils may serve employer interests in a variety of ways: they may increase productivity through greater worker loyalty; they may help spot problems and improve quality control; they may increase the willingness of workers to accept flexible job classifications and work assignments. Under conditions of the intensified competition that comes from increased globalization, the positive impact of each of these effects could increase. If, therefore, there are significant untapped sources of increased productivity obtainable through enhanced cooperation at the point of production, and if working class associational power within production facilitates such cooperation, then the upward sloping part of the class compromise curve within the sphere of production may rise more steeply as a result of increased competitive pressures.(36)

The extent to which these positive effects on the prospects of positive class compromise in the sphere of production occur depends upon the nature of the labor process and the organization of work. As Streeck (1991, 1992), Aoki (1988) and others have argued, the maximally productive use of advanced technologies often requires higher levels of information-coordination, problem-solving and adaptability than in traditional mass production. To the extent that strong working class associational power within the sphere of production enhances the levels of trust between employees and managers, and such trust is necessary for such new forms of work organization, the positive effect of workers' power on capitalist interests may be strengthened. On the other hand, if the technological conditions of production foster weak interdependencies among workers within highly atomized labor processes, increased globalization and competitive pressures would probably not enhance the positive effects of workers' associations within production. This suggests that there are probably strong interactive effects (rather than merely additive effects) of globalization and technological change on the conditions for class compromise.

Perhaps the most commonly told story about the negative effects of globalization on the prospects of positive class compromise concerns the sphere of politics. Because of the heightened international mobility of capital, especially financial capital, the argument goes, the capacity of states to engage in deficit spending and other reflationary policies has eroded (Stewart, 1984). One of the key ways in which a politically well-organized working class positively benefitted capitalists in the past was by creating the conditions for expansive state spending programs which bolstered aggregate demand. The reduced fiscal autonomy of the state resulting from increased globalization both reduces the benefits from such policies and the capacity of the state to sustain them, and thus reduces the positive slope of the class compromise in the sphere of politics.

Institutional exclusions

If it is true that the globalization of capital, especially the high levels of easy international mobility of finance capital, has undermined the fiscal capacities of states to fund a generous social wage, then the institutional exclusions to the left of the trough in Figure 10 are likely to become narrower. This, in turn increases the incentive for the capitalist class to launch offensives to undermine the associational power of workers, since there is more room to move towards the "peak" of the "capitalist utopia". The prospects of class compromise, therefore, may increasingly depend upon the way the institutional rules of the game facilitate or impede access to the upward sloping region to the right of the trough. Specifically, in the American case, the legal rules which significantly impede union organizing and, especially, make it difficult to forge strong working class associational practices within the sphere of production, and the political rules of the game which impede the formation of parties linked to working class interests can be seen as restricting the scope of potential class compromise.

The question then becomes whether increasing globalization of capitalism makes it easier or harder to change the institutional restrictions on associational practices. I do not think there is a clear answer to this question. At the level of political debate, the forces of international competition, the global economy and the new technical requirements of production can be enlisted both to support anti-associational laws as well as to support labor law reform which would enhance the capacity for working class organization. On the one hand, neo-liberal arguments about the need to eliminate government regulations to allow for a relatively unfettered market in order to compete internationally reinforces anti-associational biases. On the other hand, the arguments that "high road", high performance economies require longer time horizons with durable forms of productivist cooperation suggests the need to open up more space for collective organization of employees in all spheres of social interaction. Opening up such space by changing the institutional exclusions, of course, does not in and of itself guarantee a move in the direction of more associational strength of workers, but it would improve the strategic setting in which struggles to accomplish this would take place.

The prospects for working class associational power within the strategically available limits

In contemporary American discussions of the choice between "high road" and "low road" capitalism -- a capitalism centered on high wages, high skills, high innovation capacity and high productivity versus a capitalism based on lower wages, declining skills, and stagnant productivity -- many commentators seem to suggest that the main obstacle to the high road is capitalist myopia and ideological rigidity. Enlightenment thus seems to be the critical solution: if only capitalists would see the upward sloping part of the curve, they would embrace it. If it were in fact the case that the class compromise curve was a full fledged J-curve rather than a reverse-J, then perhaps this diagnosis would be correct: in an imaginary J-curve capitalism, capitalists' own interests are best served by high levels of working class associational power. In such a situation, when capitalists are on the downward-sloping part of the curve -- the part to the left of the trough -- they may not believe that the upward sloping part of the curve even exists. Because of their limited vision, intensified by short time horizons of capital markets and pervasive intellectual confusion, capitalists may simply extrapolate in a linear manner from their immediate situation and assume that increasing working class strength will only make things worse. If the true relationship between working class power and capitalist interests were a strong J-curve, therefore, enlightenment of capitalists might be the key to movement in the direction of the high road.

On the other hand, if the relationship is a reverse-J curve as I have argued in this paper, then capitalist enlightenment is unlikely to do the trick, at least when the balance of power places working class associational power in the downward sloping part of the curve. In such a situation, even if globalization means that capitalist interests would be furthered by the enhanced productivity of a production-centered class compromise, capitalists are generally likely to be hesitant to accept the level of working class associational power needed to make such compromises durable. Since such associational power among workers within production inherently reduces capitalist autonomy, most capitalists most of the time will not voluntarily move to that part of the class compromise curve. The result is that, just as in the earlier era of the Keynesian class compromise within the sphere of exchange, a class compromise favorable for workers within the sphere of production is only likely as the outcome of struggle. It is not enough for capitalists to have an enlightened view of the trade-offs they face in which they understand that the class compromise curve slopes upward with working class associational power beyond a certain level. Enlightenment may help, but it needs to be backed by constraints, and this requires struggle and power.

How, then, do globalization and technological change affect the prospects for working class struggles over associational power within the strategically accessible space of possibilities? It is hard to be optimistic. Broadly speaking, there are three effects of current economic transformations which potentially undermine the capacity of workers to struggle for enhanced associational power.

First, globalization is one aspect of a more general process of increasing marketization of social life which tends to increase heterogeneity within the labor force. The crucial dynamic here is not so much the global character of markets as such, but the increasingly pervasive and unfettered role of markets in organizing economic activities.(37) Markets are machines for accentuating inequalities. As the publicly imposed institutional constraints on the functioning of all sorts of markets -- labor markets, capital markets, commodity markets, consumer markets, etc. -- has declined, economies have moved in the direction of winner-take-all lotteries, where small differences among competitors in a given market can produce huge differences in outcomes. As such differences accumulate over time, heterogeneity within markets increases. Within the labor market this is reflected in sharp rises in the levels of earnings inequalities both between occupational and educational strata and within those strata. Such heterogeneity and inequality, in turn, undermine the economic conditions for solidarity, thus making it harder to move to higher levels of working class associational power, especially when this involves challenging institutional rules of the game.

Second, in combination with advanced forms of technological change, the increasing heterogeneity generated by intensified marketization has intensified tendencies towards a specific kind of dualism within developed economies in which some sectors of the labor force are in a position to forge productivist class compromises while others are not. Dualistic tendencies, of course, are not a new phenomenon. In the 1960s and 1970s there was much discussion of dual labor markets and the division between the "monopoly" and "competitive" sectors of the economy. The argument here is that the weakening of the possibilities for centralized class compromises through the state has lead to an intensification of the divisions connected to the new, technologically-grounded dualism and reduced the scope for forging broader class alliances that would serve the material interests of workers in general.

Third, it is often argued that globalization increases the vulnerability of an increasing proportion workers to job loss, either through the effects of intensified international competition or through the threat of capital mobility to low wage areas. Even if this threat is exaggerated and most job loss accompanying deindustrialization is the result of technological change rather than capital mobility, still it is the case that the belief by workers that their jobs are vulnerable because of the threat by capital to move jobs to low wage regions undercuts their willingness to struggle for enhanced associational power.

Take together, these arguments suggest a more hostile setting for durable, positive class compromises in the developed capitalist countries: the shape of the class compromise curve itself is somewhat less conducive to positive class compromise in the spheres of exchange and politics, leading to more narrowly production-centered compromises; institutional exclusions in many countries seem to be opening more space for an attack on working class associational power; and the combination of increasing heterogeneity, deepening dualism, and decreasing job security reduces the capacity of workers to defend, let alone extend, associational power even if the institutional exclusions were to open up new possibilities.

If the arguments of this paper are correct, the most plausible strategy for reconstructing the conditions favorable to positive class compromise would be to build outward from compromises within the sphere of production. Macro-corporatism was at the heart of advanced forms of traditional Keynesian politics of "effective demand", a politics which helped capitalists solve collective action problems around the macro-economic demand for the products they produced. The development of new forms of meso-corporatist institutions constitutes what Rogers and Streeck (1994) have called the politics of "effective supply", a politics which enable capitalists to solve collective action problems in the provision of certain critical inputs into production. Technological changes have enhanced the potential benefits of such politics of effective supply, and at least within the sphere of production, the intensified competition generated by globalization may have also enhanced the potential willingness of capitalists to move towards the kinds of class compromise which make such solutions durable.

Still, it remains the case, as in earlier eras of capitalism, that forging class compromises within production and extending them outward will depend in significant ways on power and struggle. The relationship between working class power and capitalist class interests remains a reverse-J relation, and capitalists will thus always be tempted, when historical opportunities occur, to try to move to toward the "capitalist utopia" of weak working class associational power. This option needs to be foreclosed in order for the new possibilities of productivist class compromises to become attractive to capitalists. Closing off this option requires state intervention, and this implies that removing the "low road" path of an atomized working class with relatively low wages and low skills ultimately depends upon the revitalization of a mobilized political forces supporting an effective economic program of "associational productivism".

6. Conclusion

This paper has attempted to chart out a general, abstract model of class compromise in capitalist society. The core intuition builds on Gramsci's insight that in democratic capitalist societies the capitalist class is often hegemonic, not merely dominant, and this implies that class conflict is contained through real compromises involving real concessions, rather than brute force. The bottom line of the argument is that the stability and desirability of such compromises depends upon the specific configurations of power and interests that characterize the relationship between the capitalist class and the working class: when it is the case that working class associational power positively contributes to solving problems faced by capitalists, then such compromises will be much more durable than when they emerge simply from capacity of workers to impose costs on capital.

The theory of class compromise proposed here is thus in keeping with the traditional core of Marxist theory in arguing that the power and struggle are fundamental determinants of distributional outcomes in capitalist societies. But contrary to traditional Marxist ideas on the subject, the model also argues that the configuration of capitalist and worker interests within the "game of class struggle" is not simply determined by capitalism itself, but depends upon a wide variety of economic, institutional and political factors. Above all, the model argues that class power not only can affect the outcome of class conflict, but the nature of the game itself: whether or not the confrontation of capital and labor takes the form of a sharply polarized zero-sum conflict or an iterated prisoners' dilemma or, perhaps, a strategic context with significant assurance game features conducive to positive class compromise.


Ackerloff, George. 1982. "Labor Contractsd as Partial Gift Exchange." Quarterly Journal of economics 97 (2): 543-69

Ackerloff, George and Janet Yellen. 1986. Efficiency Wage Models of the Labor Market. New York: Cambridge University Press.

Alcian, Armen and Harold Demsetz.1972. "Production, Information Costs and Economic Organization." American Economic Review 52 (December): 777-95

Aoki, Masahiko. 1988. Information, Incentives and Bargaining in the Japanese Firm (Cambridge University Press, 1988

Axelrod, Robert. 1984. The Evolution of Cooperation (New York: Basic Books)

Bartlett, Randall. 1989. Economics and Power (Cambridge: Cambridge University Press)

Bowles, Samuel. 1985. "The Production Process in a Competitive Economy: Walrasian, Neo-Hobbesian and Marxian Models." American Economic Review 75: 1 (March)

Bowles, Samuel and Herbert Gintis. 1990. "Contested Exchange", Politics & Society, June.

Bowles, Samuel and Herbert Gintis. 1998. Recasting Egalitarianism: new rules for communities, states and markets, vol III in the Real Utopias Project, edited by Erik Olin Wright (London: Verso)

Calmfors, L. and J. Driffill.1988. "Bargaining Structure, Corporatism and Macro-economic Performance," Economic Policy, 6: April, 13-61

Cameron, David R. 1978."The Expansion of the Public Economy: a Comparative Analysis", The American Political Science Review, 72:4

Christman, John. 1994a. "Distributive Justice and the Complex Structure of Ownership," Philosophy & Public Affairs summer 23:3, pp. 225-250,

Christman, John. 1994b. The Myth of Property (Oxford University Press, 1994)

Crouch, Colin and Wolfgang Streeck (eds). Political Economy of Modern Capitalism: map[ping convergence and diversity. (Thousand Oaks, CA: Sage)

Doeringer and Piore, 1985. Internal Labor Markets and Manpower Analysis (Lexington MA: D.C. Heath & Co, 1985 (second edition)

Dore, Ronald. 1973. British Factory, Japanese Factory: the Origins of National Diversity in Industrial Relations (Berkeley: University of California Press).

Edwards, Richard . 1979 Contested Terrain: The Transformation of the Workplace in the Twentieth Century, Basic Books: NY 1979

Esping-Anderson, 1990. Three Worlds of Welfare Capitalism (Princeton, N.J.: Princeton University Press)

Foulkes, F.K. 1980. Personnel Policies in Large nonUnion Companies. (Englewood Cliffes: Prentice Hall)

Freeman, Richard. 1988. "Labour Market Institutions and Economic Performance," Economic Policy, No. 6: 64-80

Garrett, Geoffrey. 1998. Partisan Politics in the Global Economy (Cambridge: Cambridge University Press).

Goldthorpe, John. 1982. "On the Service Class", in Anthony Giddens and Gavin Mackenzie (eds), Social Class and the Division of Labor (Cambridge University Press, 1982).,pp.162-85

Gordon, David M. 1976. "Capitalist efficiency and Socialist efficiency", Monthly Review1976, 29:19-39

Gramsci, Antonio. 1971. Selections from the Prison Notebooks, edited and translated by Quintin Hoare and Geoffrey Nowell Smith. (New York: International Publishers)

Greenwald, Bruce, Adverse selection in the Labor Market (NY: Garland Press, 1979)

Grey, Thomas. 1980. "The Disintegration of property," Nomos XXII, edited by J. Roland Pennock and John W. Chapman (New York: NYU Press), pp.69-85.

Kalecki, Michael. 1943. "The Political Aspects of Full Employment", in J. Osiatynski (ed). Collected Works of Michael Kalecki, vol. I, (Oxford: Oxford University Press, 1990)

Kenworthy, Lane. 1995. In Search of National Economic Success: balancing competition and cooperation. (Thousand Oaks, CA: Sage)

Korpi, Walter. 1983. The Democratic Class Struggle. (London: Routledge and Kegan Paul);

Kotz, David. 1994. "Interpreting the social structure of accumulation theory" in David Kotz, Terrance McDonough and Michael Reich (eds), Social Structures of Accumulation (Cambridge: Cambridge University Press). pp 51-70

Gordon, David. 1996. Fat and Mean (New York: The Free Press)

Nakane, Chie. 1984 Japanese Society (Tutle Paperback)

Offe, Claus and Helmut Weisenthal. 1980. "Two Logics of Collective Action" in Political Power and Social Theory, vol. 1, edited by Maurice Zeitlin, (Greenwich, CT: JAI Press), pp.67-116.

Olson, Mancur. 1982. The Rise and Decline of Nations (New Haven, Connecticut: Yale University Press)

Ouchi, W. G. 1981. Theory Z: How American Business can Meet the Japanese Challenge. Reading, MA: Addison-Wesley

Perrone, Luca. 1984. "Positional Power, Strikes and Wages", American Sociological Review, June

Pohjola, Matti. 1992. "Corporatism and Wage Bargaining." in Jukka Pekkarinen, Matti Pohjol and Bob Rowthorn (eds), Social Corporatism: a Superior Economic System?(Oxford: The Clendon Press) pp. 44-81

Pontusson, Jonas. 1997. "Between Neo-liberalism and the German Model: Swedish capitalism in tradition", in Colin Crouch and Wolfgang Sstreeck (eds). Political Edconomy of Modern Capitalism: mappoing convergence and diversity. (Thousand Oaks: Sage Publications)

Przeworski, Adam. 1985. Capitalism and Social Democracy (Cambridge: Cambridge University Press)

Roemer, John. 1994. A Future for Socialism (Cambridge: Harvard University Press)

Roemer, John. 1996. Equal Shares: making market socialism work. The Real Utopoias Project, vol. II, edited by Erik Olin Wright (London: Verso)

Rogers, Joel and Wolfgang Streeck (1994). "Productive Solidarities: economic strategy and Left Politics," in David Miliband, ed. Reinventiung the Left (Cambridge: Polity Press).

Rogers, Joel. 1990. "Divide and Conquer: further 'Reflections on the Distinctive Character of American Labor Law'", University of Wisconsin Law Review (1990), pp. 1-147.

Rowthorn, Bob. 1992. "Corporatism and Labour Market Performance." in Jukka Pekkarinen, Matti Pohjol and Bob Rowthorn (eds), Social Corporatism: a Superior Economic System?(Oxford: The Clendon Press) pp.82-131

Schmitter, Philippe. 1988. "Corporatism is Dead! Long Live Corporatism! Reflections on Andrew Schonfield's Modern Capitalism"Government and Opposition 24: 54-73.

Shapiro, Carl and Joseph Stiglitz. 1984. Equilibrium Unemployment as a worker Discipline Device," American Economic Review, 74: 203-224

Sorenson, Aage,. 1994. "Forms, Wages and Incentives," in Neil Smelser and and Richard Swedberg (eds), Handbook of Economic Sociology (Princeton: Princeton University Press), pp.504-528)

Soskice, David. 1990. "Wage Determination: The Changing Role of Institutions in Advanced Industrialized Countries." Oxford Review of Economic Policy 6: 36-61

Stewart, Michael. 1984. The Age of Interdependence (Cambridge, M.I.T. Press)

Streeck, Wolfgang. 1991. "On the Institutional Conditions for Diversified Quality Production," in E. Matzner and W. Streeck, eds. Beyond Keynesianism: the Socio-Economics of Production and Employment (London: Edward Elgar).

Streeck, Wolfgang. 1992. Social Institutions and Economic Performance: studies of industrial relations in advanced capitalist economies (Newbury Park: Sage Publication).

Streeck, Wolfgang and Philippe Schmitter. 1985. "Community, Market, State - and Associations? The Prospective Contribition of Interest Governance to Social Order", European Sociological Review, 1: 119-38.

Waldman, Micahel, "Job Assignments, Signaling and Efficiency," Rand J of Economics, XV (1984) 255-67. Workers stay with firms longer than they would have otherwise stayed because of private info of firms.

Williamson, Oliver. 1985. The Economic Institutions of Capitalism (New York: The Free Press)

Wright, Erik Olin. 1985. Classes (London: Verso)

Wright, Erik Olin. 1997 Class Counts: comparative studies in class analysis (Cambridge, UK: Cambridge University Press)

Wright, Erik Olin. 1989. The Debate on Classes (London: Verso)

Wright, Erik Olin. 1994. Interrogating Inequality (London: Verso)


1. The reverse-J shaped relationship between working class power and capitalist interests was first suggested to me in a paper by Joel Rogers (1990).

2. I have written extensively elsewhere on the concepts of class and interests, and to a lesser extent, on the concept of power. For a discussion of the concept of class which underlies the present analysis, see Wright (1997: chapter 1); for interests and class, see Wright (1989:278-288); for power, see Wright (1994: chapters 4 and 5)

3. The claim that this abstract polarized concept of class is analytically useful may be controversial. Given that actual capitalist firms engaging in complex strategies and bargaining with their employees encounter considerable heterogeneity, and this heterogeneity in fact does matter for the optimal profit maximizing strategy of the firm, it may seem illegitimate to bracket such complexity in favor of a simple model of capital and labor. As in all attempts a elaborating theoretical models, the appropriate level of abstraction is a matter of contestation.

4. When I speak of "class interests" I will always mean the interests of people determined by their location in the class structure. I do not believe that classes as collective entities have "interests" in a literal sense. Of course, individuals may have interests in the strength of the collective organizations of classes, and the class interests of individuals may be contingent upon the security of the interests of other members of the same class, but this still does not mean that classes-qua-collectivities have "interests".

5. While there is no simple "bottom line" to these various dimensions of interests, the problem is somewhat simpler for capitalists than for workers. For workers, the interests directly shaped by their class location concern central issues in the quality of life, not simply financial remuneration for work. For capitalists, on the other hand, the various components of material interests all revolve around their ability to secure and maintain profits. While concrete interest heterogeneity among capitalists may still lead to significant conflicts among capitalists, nevertheless in general their material interests are more transparent than is the case for workers. For a provocative discussion of the qualitative heterogeneity of working class interests and its implications for class power and class struggle, see Offe and Weisenthal,(1980). Streeck (1992: chapter 3) challenges Offe and Weisenthal's analysis, arguing that in certain pivotal respects capitalist interests are at least as heterogeneous as workers.

6. The issue of the time horizons of capitalist interests is quite important in the analysis of class compromise. The time horizon within which interests are framed is not simply an exogenous datum to the process of class compromise (or any other issue in class analysis) reflecting cultural standards or psychological discount rates. The existence and stability of class compromises contributes to stretching time horizons for both workers and capitalists.

7. For an interesting discussion of how strategic location of employment within the input-output matrix of an economy confers structural power on groups of workers, see Luca Perrone (1984).

8. I am using the general term "institutional sphere" here to suggest a cluster of interconnected institutional sites rather than well-bounded "organizations".

9. There is a gap in the reasoning here, for even if it is the case that capitalists' interests are realized through the exploitation of workers, it would not inherently follow from this that they are better off with a completely atomized working class than a highly organized one. This would all depend, on the on the hand, upon precisely how much productivity gain and other advantages for capitalists are generated by the cooperation accompanying high worker organization and, on the other, how much the social control apparatus (supervisors, police, etc.) costs to sustain high levels of atomization. These costs and benefits will depend upon technological factors, the nature of international markets, and many other considerations. Nevertheless, while I cannot establish the reverse-J relation as a tight logical deduction, it corresponds to my theoretical intuitions about the typical situation of capital-labor relations, and thus I will adopt this form of the relationship throughout this paper. Readers who are skeptical of the asymmetry of the reverse-J can substitute a symmetrical U-shaped curve throughout most of the paper.

10. Although the actual term "class compromise" appears mainly within the Marxian tradition of social theory, the substantive idea has much broader currency. I will not limit the discussion here to instances where the term is explicitly deployed.

11. There are some treatments of efficiency wages which treat them more as a form of positive than of negative class compromise. For example Akerloff (1982) sees such arrangements less as a concession in response to a form of individual resistence -- shirking -- and more as a way of improving generalized morale through a kind of normatively-grounded "gift exchange". The implication is that efficiency wages underwrite active cooperation.

12. Botwinick (1993:58) provides a critique of standard efficiency wage theory for ignoring the class dimensions of the problem: "EW [efficiency wage] theories suggest that the primary conflict within the labor process does not stem from the exploitation of labor, but rather from the innate tendency for individuals to shirk (and otherwise misbehave) when their activities cannot be accurately 'metered'. The conflict therefore arises not from any fundamental contradictions between the interests of capital and labor as economic classes, but from the supposedly eternal problem of self-serving 'human nature' and conflicting individual preference structures. Moreover, this conflict over work effort is essentially benign for it is easily muted by simply paying workers higher wages. Thus contrary to Marx's claims, the conflict of interest between capitalists and workers is not irreconcilable and collective worker organization is not inherently necessary to protect the interests of labor."

13. Bowles and Gintis do not emphasize the normative component of the problem of labor effort extraction. For a discussion of this issue, see Burawoy and Wright (1990).

14. Some discussions of internal labor markets treat them primarily as examples of negative class compromise, emphasizing the ways in which ILMs are instigated to divide the working class, weaken unions and in other ways enhance capitalist control over labor. See for example, Gordon (1976), Edwards (1979). Others treat ILMs strictly as a problem of solving problems of internal efficiency, typically linked to information costs and the problem of retaining skilled employees, without systematic reference to the class character of these efficiency considerations. See for example Doeringer and Piore, (1985), Greenwald, (1979); Waldman, (1984).

15. Goldthorpe's (1982) concept of the "service class" revolves around the problems employers face when their employees sell a "service" rather than simply "labor". The creation of career ladders built around prospective rewards which create a longer time horizon of commitment for such employees is at the core of his analysis of the specificity of this employment relation. My analysis of "loyalty rents" for managerial class locations (Wright, 1998: 21) also emphasizes the problem of creating deeper commitments for certain categories of employees by anchoring their jobs in career ladders. In both of these treatments of class relations, internal labor markets are created as responses to the strategies of individuals within firms.

16. "If a corporation with 10,000 employees loses one employee," Commons (1970 [1950): 269) writes, "it loses only one ten-thousandth part of its labor force, but the employee loses 100 per cent of his job. No wonder the laborers should reason that if the 10,000 should quit or were discharged as a unit, then the corporation also would lose 100 per cent of its working force, while each laborer was losing 100 per cent of his job"

17. Not everyone who discusses the importance of trust and cooperation within production places such importance on the associational power of workers. Piore and Sabel (1984) in their writing on "flexible specialization" cite many of the same requirements for successful production as does Streeck, but contrary to Streeck believe that such trust and cooperation can be sustained primarily through cultural mechanisms, social networks and voluntary interactions. In effect they adopt a stance closer to the upper right hand cell in figure 3.

18. Strictly speaking, Calmfors and Driffill study the relationship between workers power and various measures of general economic "performance" rather than capitalists' interests as such, but in the context of their arguments this can reasonably be taken as an indicator of capitalists' interests.

19. Wage restraint, they argued, was the critical factor here: In the countries with highly decentralized bargaining structures, labor market pressures tended to restrain wages, whereas in the countries within highly centralized bargaining, the unions accomplished the same thing. In the intermediary countries, unions were strong enough to push wage demands, but were not encompassing enough to internalize the negative effects of excessive wage demands.

20. The standardized coefficients for the union density and union density squared variables in Hicks and Kenworthy's (1998: 1657) analysis are -.85 and +.77 respectively (thus indicating a parabolic shaped curve), both of which were highly significant.

21. Obviously in the real world, the options are much more complex than this stark contrast - not only are there various degrees of opposition and cooperation, but a variety of qualitatively distinct forms of both. Nevertheless, for purposes of developing a general inventory of strategic contexts for class compromise in which mutual cooperation occurs it will be useful to abstract from such complexity and examine games in which members of each class (considered either as individuals or as members of associations) make such simple, dichotomous choices

22. Even though mutual opposition is the equilibrium solution, Model I is not a prisoners dilemma for workers since in a prisoner's dilemma actors prefer mutual cooperation to mutual opposition.

23. In this theoretical conception of socialism, capitalists, albeit with curtailed property rights, can exist within a socialist economy just as they existed centuries earlier within a feudal society. It is another question how stable and reproducible such a structure of class relations would be. For a formal model of a sustainable socialist society within which capitalists still have some economic space, see Roemer (1994, 1996).

24. This does not mean that working class associational power is a necessary condition for the solution to such coordination problems. There may be other devices which may constitute alternative strategies for accomplishing this. All that is being claimed is that working class associational power can constitute a mechanism which makes it easier to solve such problems.

25. It is possible, under certain social and cultural conditions, for some of these forms of cooperation to emerge and be sustained without strong workplace associational power of workers. This is often the way the relatively cooperative system of employment relations in Japan is described (e.g. Nakane, 1984), although others have criticized such culturalist views (e.g. Aoki, 1988: 304ff). In any event, under many conditions high levels of worker cooperation within production are likely to be difficult to sustain if they are not backed by some form of significant associational power.

26. The degree of commodification of labor power refers to the extent to which the costs of its production and reproduction is secured strictly through exchange and the market (see Esping-Anderson, 1990). When various components of the costs of producing and reproducing labor power are provided directly in the form of usevalues, then labor power is accordingly partially decommodified. Universal free health care, free public education, housing subsidies, child allowances, etc. are all instances of state programs which partially decommodify labor power.

27. There is no suggestion here that a strong, working-class based social democratic party is a necessary condition for corporatist institutions that generate these positive effects for the interests of capitalists. There may be other political paths to this outcome. The claim here is simply that high levels of working class political power can produce these positive effects thus generating the upward sloping curve within the sphere of politics in figure 8.

28. In the spheres of production and exchange, there may be considerable heterogeneity in the shape of the class compromise curves and the degree of working class associational power across firms and sectors. The result is that within a given country the conditions for class compromise may be much more favorable in some firms and sectors than in others. Within the sphere of production, it is easy enough to see how the upward sloping curve can be restricted to a particular sector or even firm, since most of the gains from cooperation are contained within firms. In the sphere of exchange, while many of the positive effects of high levels of unionization for capitalists come from aggregate, macro-economic effects, some of the positive effects -- such as stabilization of labor markets, rationalized skill formation, and wage restraint in tight labor markets -- may be concentrated in specific sectors or localities. The reverse-J curve characterizing a given sphere, therefore, is itself an amalgamation of the distribution of such curves across firms, sectors and other less aggregated units of analysis.

29. The actual variation across time and place is, of course, much more complicated than is being portrayed here. Countries will vary not simply in where they are located on each of these curves, but also on: 1) the relative weights of the various curves in defining the overall configuration for the society; 2) the units of analysis within countries within which class compromises are most rooted; 3) the specific shapes of the component curves themselves. In some times and places, for example, the upward-sloping segments of some of the curves might be relatively flat, in other cases, quite steep. My theoretical understanding of these relations is insufficient to say anything very systematic about either of these two sources of variation.

30. A number of contemporary commentators on the concept of "private property" or "private ownership" have stressed the multidimensionality of the rights and powers implied by these terms. These include such things as rights to the residual income generated by the use of the property, rights to decide among alternative uses of the property, and rights to alienate the property (including selling, giving away or even destroying the property). See John Christman (1994a, 1994b) and Thomas Grey (1980).

31. The x-axis in figure 9 is working class associational power undifferentiated into the spheres of production, exchange, and politics. It thus represents an under-theorized amalgam of the associational power within the three spheres (which are themselves amalgams of associational power across the various units of analysis that make up a sphere). The underlying intuition is that viable democratic socialism requires high levels of workers associational power within all three spheres, and that a sustainable threat to fundamental capitalist property rights under democratic conditions can only occur when such unified associational power occurs. This does not imply, however, that the three spheres are of equal weight in this theoretical gestalt. Traditionally Marxists have argued that working class power at the level of the state is most decisive for challenging capitalist property rights, whereas syndicalists have argued that the pivot is workers power within production.

32. As in Figure 9 the x-axis in this figure is an unspecified weighted average of working class associational power within exchange, production, and politics.

33. For a discussion of Alford and Friedland's three-level game theory metaphor as a way of thinking about the class character of politics, see Wright (1994: chapter 5)

34. For a detailed discussion of some of these legal impediments to extensions of working class associational power, see Rogers (1990).

35. The negative effects on capitalist interests of a strong labor movement (the downward sloping curve in Figure 7) may also have become more significant in the context of intensified competitive pressures. In particular, many economists argue that the rigidity in employment relations generated by strong labor movements which make it more difficult for employers to fire workers have also become more costly to employers in the context of heightened competition. The constant calls for increased "flexibility" in labor markets and employment relations reflect this (supposedly) increasingly negative effect of working class associational power in the sphere of exchange on the interests of capitalists.

36. Again, just to reiterate the central argument behind the reverse-J curve: works councils, like all forms of working class associational power, also have negative effects on capitalist interests. Works councils impose various kinds of rigidities on employers which interfere with their capacity to unilaterally reorder production in the face of competitive pressures. The downward sloping curve -- not shown in Figure 15 -- might therefore descend more precipitously under conditions of intensified global competition. The claim here, then, is not that the net effect of globalization necessarily enhances the value of institutions like works councils, but simply that the positive effects become stronger.

37. Increasing marketization is partially driven by the erosion of various forms of decommodification of labor in the form of social wages, job protections, and other forms of regulation, and partially by the intensification of competition for desirable jobs. Increasing marketization is not simply an economic phenomenon driven by economic dynamics; there is a crucial political component in the transformation of public policy around the regulation of various markets.