Employment Segregation in the Early Twentieth Century

Cliometric Sessions at 1990 ASSA Meetings--December 28, 10:15 AM



William A. Sundstrom
Department of Economics
Santa Clara University

1. Introduction

The historical relationship between black workers and the labor union movement in the United States has been anything but smooth. Racial exclusion was practiced by many of the craft unions that emerged during the late nineteenth century and was tolerated by the AFL. Not all black activists or intellectuals could share Booker T. Washington's rejection of trade unionism and apparent embrace of strikebreaking as a road to black economic progress; but even a socialist like W.E.B. DuBois had to admit that the treatment of blacks by the white unions had "convince[d] the American Negro that his greatest enemy is not the employer who robs him, but his fellow white workingman". For their part, white unionists were incensed by the employers' use of black workers to break strikes.

On the other hand, some unions, such as the Knights of Labor and the Mine Workers, adopted a policy of integration from an early date, as did the CIO later on. In recent years, despite occasional conflicts between affirmative action goals and union seniority rules, there is no consistent pattern of union exclusion of minorities. In fact, controlling for industry, black males are somewhat more likely to be represented by unions than are other workers (Ashenfelter 1973; Leonard 1985).

In this paper I analyze the economic incentives facing white unionists when they choose between policies of integration and exclusion of black workers. For any union, racial exclusion involves a fundamental trade-off: if minorities can effectively be excluded from some or all of the occupations represented by the union, the total labor supply is diminished and there may be upward pressure on white unionists' wages; at the same time, whites may sacrifice bargaining power if minority workers will not join in actual or potential strikes. If whites have a marginal preference to organize other whites first, the exclusion decision will depend on the union's optimal membership size and the number of black workers already in the market. This is not to deny the importance of stronger racist motives or "tastes for discrimination," but to emphasize how strategic interests may or may not reinforce them in the behavior of unions and employers. The results would generalize to the case of any sub-group in the labor market, such as women or recent immigrants.

Economists and industrial relations researchers who have studied the racial policies of American trade unions have posited several hypotheses regarding the differences between unions, in particular between the craft unions of the AFL and the industrial unions of the CIO. It is frequently argued that racial exclusion was more consistent with the craft-union strategy of restricting and controlling labor supply to raise wages than it was with the industrial-union strategy of mass organization at all skill levels (Marshall 1965; Ashenfelter 1973). The logic of this argument is not self-evident, however, since the basic costs and benefits of organizing more workers must have affected both types of unions. An exclusive craft union must worry about leaving so many potential replacements unorganized that strikes or strike threats become ineffective, even if actual strikes rarely occur. And industrial union members would still have to be concerned about the adverse effect of organizing and employing too many additional members on the wellbeing of each current member. Although the historical record does confirm that exclusion of minorities from union membership was more prevalent in the skilled crafts, there were important exceptions, and closer attention to the incentives involved seems warranted. Moreover, although industrial unions tended to be more inclusive in terms of membership, it is possible that they were discriminatory in job allocation among members. Whereas white craft unions kept blacks out of the skilled positions by excluding them from union membership and apprenticeship programs, some industrial unions accomplished a similar end by establishing separate promotion ladders and job categories by race under the labor contract (see Dewey 1955, Hill 1973).

A second and clearly relevant factor affecting the racial policies of white-dominated unions was simply the availability of minority workers in the relevant labor force. A union in a labor market with large numbers of black workers might have felt compelled to organize them or face almost certain defeat in a strike; how strongly compelled depends on the union's desired membership from a strategic point of view.

2. Exclusion, integration, and strikebreaking: historical overview

Until World War I, the large majority of black workers in the United States were located in the South. National unions that attempted to organize in markets with significant numbers of black workers generally followed a pragmatic organizing policy: the decision to include or exclude was largely left to locals, which possessed considerable autonomy. In a number of cases, where white locals would not admit black members, the national unions encouraged the formation of segregated black locals (Spero and Harris 1931). A notable exception to this laissez faire attitude toward race relations was the Knights of Labor, which adopted an explicit policy of equal treatment for black and white members. The policy of open membership would survive the Knights in some of its successor unions, especially the United Mine Workers.

Although the AFL initially adopted an official policy prohibiting exclusion of blacks from affiliated unions, by 1895 it was clear that the federation would not force affiliates to practice integration or even admission of separate black locals. The first test case was the International Association of Machinists, which was admitted to the AFL in 1895 after it dropped the exclusion of blacks from its constitution. Blacks remained effectively excluded, however, by a pledge in the machinists' membership ritual. In 1900 the AFL constitution was revised to provide for direct organization of separate black locals by the federation when affiliated unions refused to admit blacks. By 1909, unions whose constitutions explicitly barred black members were admitted to the AFL (Spero and Harris 1931).

In spite of the AFL's tolerance of discrimination, a number of trade unions at the turn of the century gave black members full membership rights. In 1902, DuBois estimated that 20,000 of the United Mine Workers' 224,000 members were black. The Longshoremen and the Cigar Makers also, in DuBois's view, had virtually open membership policies for blacks. Overall, he estimated, there were about 41,000 black unionists among the 1.2 million total union members in the United States. Unions that excluded black members entirely made up about 500,000 of the total union membership. DuBois's analysis of what distinguished integrated from exclusionary unions would be echoed by many others:

The rule of admission of Negroes to unions throughout the country is the sheer necessity of guarding work and wages. In those trades where large numbers of Negroes are skilled they find easy admittance in the parts of the country where their competition is felt. In all other trades they are barred from the unions, save in exceptional cases, either by open or silent color discrimination (DuBois 1979, p. 346).

In markets with considerable competition from black workers, white union locals sometimes took the lead in organizing black workers, even in some supposedly exclusionary craft unions (see Gutman 1968; Worthman 1969).

Still, the presence of large numbers of qualified blacks was not a guarantee of a more open racial policy, as the case of the brotherhoods of railroad firemen and trainmen demonstrates. In many southern states, blacks were a large minority or even a majority of the workers in the semiskilled jobs of fireman and brakeman, which were occupations blacks had occupied since slave times. Yet the unions representing these trades not only had exclusively white memberships but adopted programs to drive black workers out altogether. Spero and Harris (1931) attribute the virulent racism of the brotherhoods to their origins as fraternal social organizations. In the South, admission of blacks to a "pure-and-simple" union might have been an economic necessity, but admission to a fraternal organization would confer social equality, which southern racial ideology would not allow. Below I argue that the contracts between the white unions and the railroads permitted the employment of blacks as a cheaper labor "reserve" that was gradually replaced by white unionists as the unions grew stronger in the South.

Initially, the CIO and its affiliated industrial unions pushed strongly for racial equality in the labor movement. By 1946, when both the AFL and CIO launched organizing drives in the South, both organizations sought black members. But even CIO-affiliated internationals had limited control over locals in the South. Some locals barred black members, while in other cases blacks were organized into separate or auxiliary locals (Marshall 1965). After the AFL-CIO merger in 1955, opposition to racial discrimination remained the official federation policy, although problems of occupational segregation, separate promotion lines, seniority provisions, and segregated locals led to criticisms by civil rights organizations such as the NAACP (Marshall 1965).

If it is true that the admittance of black members into unions has generally occurred where large numbers of blacks were already present in the market (possessing the requisite skills to hold jobs in the trade or industry), then the fear of losing bargaining power must have constituted the major economic incentive operating on the unions. This loss of bargaining power manifested itself most dramatically in cases of actual strikebreaking by black workers. Whatley's (1989) search of the secondary literature for evidence of black strikebreaking turned up 141 incidents between 1847 and 1929, a "negligible" number compared with the total number of strikes that occurred during the period. Most of the incidents occurred in northern cities, where black strikebreakers were recruited from local labor markets or imported from the southern branch of the industry engaged in the strike. Of course, most strikes that were broken were probably broken by whites; Whatley speculates that recent immigrants may have played an important role in strike failures. But the threat of using black or other minority workers as replacements during a strike was surely there even when a strike never occurred. And black strikebreaking did play a central role in some historic losses for organized labor, such as the 1919 steel strike.

Although employers may have exploited the special animosities between white and black workers to thwart union solidarity, these observations on union exclusion and strikebreaking make it clear that the basic strategic decision from the point of view of a white-dominated union was quite general: how many workers to organize given the greater likelihood of nonmembers to break a strike.

3. Membership and bargaining power: a model

Setting aside for the moment "racial preferences" among the union's leadership or rank and file, the costs and benefits of racial exclusion are essentially the costs and benefits of restricting the membership of the union to the white fraction of the potential workforce in the occupation or industry. By recruiting minority members and treating them on an equal basis, white union members could benefit from greater solidarity in the event of a strike, reducing the pool of strikebreakers. The cost to the union is a "dilution" of the wage and employment benefits to each white member. How these costs and benefits of membership size balance out and depend on the size of the overall workforce is explored in the models in this section. Mathematical details have been left out of this version of the paper but are included in another version available from the author.

I assume that the union negotiates with a single employer or employer association over both wages and employment levels and that the contract agreement satisfies the Zeuthen-Nash cooperative bargaining solution, in which the two parties "split the difference" between their welfare under the contract and their welfare at the threat points. The threat points consist of the profits and union welfare in the event of a strike. The assumption that the union negotiates over employment as well as the wage implies that the bargain will be efficient, but not necessarily on the firm's labor demand curve. This assumption conflicts with the conventional wisdom that real-world unions seldom bargain directly over employment levels, but recent evidence suggests that union objectives do place considerable weight on employment, and that union contracts, whether or not they are efficient, do not seem to lie along the demand curve. Work rules under the contract may serve implicitly to regulate employment levels.

The profits of the employer are --�= R(e) - we where w and e are the wage and employment respectively, and R(e) is a revenue function. One could think of R as being equal to pf(e), where p is the output price and f the production function. I assume that R is strictly concave, so R''< 0, and that R' and R'' are finite at e=0.

The union seeks to maximize the average utility of its members, where each member's utility depends on the wage. That is, the union's objective is v(e,w) =(e/m)u(w) + [1-(e/m)]U, where m is the union membership,U is the alternative utility for members not employed under the contract, and u(w) is concave (Carruth and Oswald 1987 discuss the properties of this objective function). This formulation assumes that employment is assigned to union members randomly; that is, when e < m, each union member has an equal probability of being employed at the union rate. I assume that the union is able to hold onto members only if it offers them equal treatment under the contract, so it makes sense to look at the average member's chances. This of course rules out allocation of employment in order of seniority. Since usually v >U, self-interested nonmembers would like to join the union, but it is not necessarily in the interest of the average union member that the membership expand without limit. There may be an optimal membership size from a strategic point of view, which I call m*.

If the union has a marginal preference for white members over blacks, it will recruit white members until there are no whites left; at that point, if the union wants to expand its membership further, it must recruit black members. Suppose the labor pool consists of a total of L workers, of whom some number b are black. Then if m* is the optimal membership of the union, the union can achieve optimal membership and remain exclusively white as longmm*� L-b. From a strategic point of view, then, the question of exclusion versus integration depends on both m* and b. The importance of b has been stressed in the literature, while the determination of m* in this context has been largely ignored.

If a strike were to occur, each union member would receive the strike utility V. Union members are assumed to remain loyal to the union during a strike in return for the promise of equal treatment under the contract. The employer's profits during a strike depend on how many union nonmembers she can employ as strikebreakers. For simplicity, I assume that the total supply of workers to the industry or occupation is perfectly elastic at wage wr up to the maximum level L, at which point it becomes perfectly inelastic. One could think of the labor pool L as the short-run supply of workers with the relevant skills. Thus L-m is the total number of potential strikebreakers available to the firm. During a strike, all nonmembers must be paid the market-clearing wage, which is simply wr if the firm employs fewer than L-m nonmembers and the marginal product R'(L-m) if the employer decides to hire all L-m nonmembers during a strike (see Figure 1). Of course, if the employer's optimal employment decision were less than L-m, she could maximize profits without having to deal with the union in the first place; so the only interesting case is that in which the union has enough members to make negotiating a contract more profitable to the employer than hiring nonmembers. For this case, the employer's profits during a strike can be written as = R(L-m) - (L-m)R'(L-m). Note that the employer's strike profits are decreasing in m: ��/�m = (L-m)R"(L-m) < 0.

Although abstract, this basic framework can be altered in various ways to capture features of specific historical situations. Of particular interest is whether the contract allows the employer to hire nonunion as well as union labor, and at what wage. Below I consider two alternative specifications. In the first, only union members are employed; I call this the "closed shop" case. In the second specification the employer hires e0 union members at the negotiated union wage w0 and e1 nonmembers at the reservation wage wr. The employed nonmembers constitute a reserve or secondary labor force. I argue that black workers formed something very much like such a reserve labor force in the market for firemen and brakemen in the southern railroad industry earlier in the twentieth century.

a. Only union members employed

In this case the contract is determined by the Nash bargaining solution (e*, w*), which maximizes the product of the differences of each party's objective from its threat position: maxe,w (�-�)(v-V) , where �,�, and v are given above and V=U, the utility of a member not employed under the contract.1 Examining the first-order conditions for a maximum, one finds that in general both e* and w* are functions of all the exogenous parameters of the model, and that it is not possible to find closed-form solutions. As it is difficult to say much about the effect of m on the union's welfare for the general case, I proceed with a simplification of the union's objective to maximization of rent per worker; this involves assuming that u(w) = w, and U=W, where W is the income flow of an unemployed union member. Solving the first-order conditions one then obtains closed-form solutions for e* and w*.

What is the optimal number of members for this union? I assume that the union has a good idea what the outcome of bargaining will be, based on its circumstances and what it knows about the employer, and attempts to recruit the number of members that puts the average member in the best position. Under the Nash bargaining solution, changes in m affect the union in two ways. First, an increase in m reduces the employer's threat profits � by diminishing the pool of potential strikebreakers. In this model, the decrease in � has no effect on e* but does increase the union wage w*. I refer to the effect of a change in membership on union welfare by way of the threat as the "solidarity effect" of increased membership. Second, for any given level of union employment, an increase in m decreases the probability of any given member being employed at the union wage. I refer to this as the "dilution effect" of increased membership. Plugging the Nash-solution values e* and w* into the union's objective function one obtains the average union member's utility under the Nash contract, v*, which is a function of the membership size m: v* = W + [1/(2m)][rent(W) - �]. Rent(w) is defined as the total rent that would be earned by the firm above the unemployment income W if employment were set where W equals marginal product.

If one assumes that the revenue function satisfies the condition R"+ eR'''< 0, then the optimal membership m* can be found where dv*/dm = 0. The comparative statics for m* show the conditions under which current union members would want the union to be more or less inclusive (have larger or smaller memberships). In this model, it can be shown that dm*/dW> 0, and dm*/dL > 1. Optimal membership is an increasing function of worker income when unemployed (W), because higher unemployment income reduces the dilution effect.2 It is hardly surprising that optimal membership should also grow with the size of the labor pool L, but furthermore dm*/dL > 1, so the desired size of the union grows at a faster absolute rate than L. In fact, the ratio of desired members to total labor force (m*/L) increases in L. Thus a union wants to organize a larger percentage of the labor pool the larger that labor pool is, ceteris paribus. For any given percentage black, this implies a greater tendency to organize black workers in a market with a larger labor pool. One interpretation of this result is that the exclusivity of craft unions vis-a-vis industrial unions had something to do with the relatively small supply of workers with the requisite industry or occupation-specific skills. Industrial unions were more likely to organize industries in which semiskilled jobs predominate and the pool of qualified workers was quite large. Of course, a complete model would have to consider the long-run process of skill acquisition in either type of industry.

Adding more structure to the revenue function by supposing that R(e) = pf(e,K), where K is a nonlabor input, one can derive the additional results: dm*/dp < 0, and dm*/dK is definitely positive (negative) if both fek and feek are negative (positive). If the marginal product of labor increases with increases in K, so at least fek is positive, perhaps we can expect dm*/dK < 0: productivity-enhancing inputs reduce desired union size. With regard to the price p, one can show that the elasticity of m* with respect to changes in p is exactly the opposite of the elasticity of m* with respect to changes in W. Thus m* exhibits "money neutrality": if both p and W increase by the same percentage, the desired m* is unchanged.

In sum, this model suggests that unions desire large memberships (and therefore may be more receptive to organizing minority workers) if they are in a market with a relatively large pool of available workers, a relatively low capital intensity, or a relatively high ratio of unemployment income to output price. Although I have suggested that the effect of L may help explain the relative inclusiveness of industrial unions, further empirical study is necessary to determine the importance of these factors.

b. Black workers as a labor reserve

It seems that in some labor markets, unions were able to negotiate a union wage, but nonunion employees -- sometimes blacks -- were also hired at a lower nonunion wage. In this model I assume that some nonunion labor is hired at the reservation wage wr , and the labor contract sets the union wage and both union and nonunion employment levels. The nonunion reserve workers may include both blacks and whites, but any increases in union membership are assumed to come from the white reservists first. The union objective is the same as above. For the employer, profits are now given by � = R(e0+e1) - w0e0 - wre1. The Nash bargaining solution (e0*, e1*, w0*) again maximizes the product of the differences of each party's objective from its threat position.3

The first-order conditions for the Nash bargaining solution imply that black workers are hired up to the point where their marginal revenue product R' is equal to their wage. White workers, on the other hand, earn a wage "mark-up" over blacks, which is a function of U, wr , and the parameters of the utility function. In this case the union wage is invariant to changes in output price, production parameters, or employer threat point �. Essentially, under such a contract the union sets the desired mark-up over the reserve labor's wage, and any changes in bargaining power, production conditions, or output price take the form of employment adjustments. Since, ceteris paribus, changes in union membership affect the strike threat for employers (�) but not w0*, it is evident that as this union grows larger and more powerful it is able to insist on the substitution of union members for nonunion reserve workers at the fixed wage mark-up. 4

The first-order condition for the optimal number of members yields the same formula for dv*/dm as the model in section (a) above, except wr replaces . Therefore all the same comparative-static results hold, with the exceptions that now dm*/d = 0 and dm*/dwr > 0. In particular, it is still the case that m*/L increases in L.

This model, although highly stylized, bears considerable resemblance to the conditions that existed on the railroads of the American South early in the twentieth century. The unions of the firemen and the brakemen followed a policy of strict exclusion of blacks from membership, although blacks were widely employed by the railroads as firemen and brakemen. Black workers seem to have constituted an almost officially sanctioned reserve labor force. Contracts with the white unions often specified a racial wage differential, with white workers earning on the order of 15 to 30 percent more than blacks on identical runs. Some of the most bitter labor disputes were waged over the level of employment of blacks in these positions. The growing numerical strength of the white unions in the South after 1900 was reflected in a declining percentage of black workers employed in these positions (see Sundstrom 1990).

The model with reserve labor might also be reinterpreted to capture some aspects of the role of minority workers in some industries organized by industrial unions. If minority workers of similar qualifications to whites are given job classifications that pay less than the better union jobs reserved for whites, then the contract functions in a manner similar to that explored here, although the reserve workers may technically belong to the union. The division of labor by race within some industries is discussed by Marshall (1965) and Dewey (1955). In his case studies of black employment in southern industry during the 1950s, Dewey finds that it was very rare for a black worker and a white worker to perform the same task but hold different job classifications; but the fact that many higher- paying job classifications were usually reserved for whites suggests the use of black workers as a cheap labor reserve.

4. Conclusions

The models explored here, although abstract, suggest that the strategic interest of a white union in its strike threat should have systematic implications for the willingness of the union to expand its membership, possibly to include minority members. In essence, expanding the membership helps reduce the pool of strikebreakers. In markets where a white union was highly vulnerable to replacement by black or other ethnic minority strikebreakers, a more inclusive membership policy might have benefited the average white member, at least economically. Where feasible, white unions may have accepted the employment of black workers as a labor reserve that could be replaced with union members as the union's power grew.


Reference list available from the author.

1 The assumption that V=U means that worker utility during a strike is the same as worker utility when unemployed. The consequences of relaxing this assumption are discussed below.

2 If one allows U (unemployment income) and V (strike income) to differ, dm*/dU>0, but dm*/dV=0. So the optimal membership is independent of strike benefits, except insofar as they are correlated with unemployment income.

3 The results in this section do not require the simplifying assumption that u(w)=w.

Reference list available from the author.