Labor unions are an important institution in U.S. labor markets today. Nearly 16.5 million workers, or 13.9 percent of all wage and salary workers, are members of unions, and more than 18 million, or 15.3 percent of workers, are represented by unions. Although the share of union workers in the economy has fallen from its peak in 1948, unions still affect almost every aspect of the employment relationship, and the presence of unions has effects on employment contracts for nonunionized as well as for unionized workers (Freeman and Medoff 1984). The tables presented here document changes in the size and character of unionization in the U.S. labor force since the beginnings of the modern labor movement in the 1880s, along with measures of selected effects of unionization. These include the impact of unionization on wages, the incidence of strikes, and public attitudes toward organized labor.

Back To Top

The first reliable measures of union membership date from 1880, near the beginning of what may be called the modern era of organized labor. Although union progress was halting and irregular during most of the nineteenth century, there was a long history of organized labor prior to the 1880s. Skilled workers in a number of urban crafts formed organizations to pursue their interests as early as the 1780s. In addition to seeking higher wages and shorter hours, these organizations often provided fraternal benefits for sickness and burial. Among the challenges confronting these early labor organizations was their ambiguous legal status. Until the 1842 case of Commonwealth v. Hunt, in which Massachusetts Chief Justice Lemuel Shaw for the first time recognized the legality of unions, state courts had generally interpreted strikes as illegal conspiracies. Even after this ruling, however, employers continued to use the conspiracy argument with some success, and the passage of legislation repealing the conspiracy doctrine emerged as an important priority of organized labor in the late nineteenth century. Despite growing legislative support for labor in the late nineteenth century, striking workers continued to face the threat of injunctions issued to prevent injury to property.¹

During the first half of the nineteenth century, the number of local craft unions proliferated, but most labor organizations remained transitory – forming during economic upswings and collapsing during business downturns. Growth in the number of labor organizations led to increased efforts at federation. In 1827, unions representing an array of different crafts in Philadelphia formed a trade council to pursue issues of common concern. Similar organizations soon appeared in most other major cities, and in 1834, the first national federation of labor organizations, the National Trades Union (NTU), was established in New York. This organization proved short-lived, however, vanishing, like many other labor organizations, in the depression that followed the panic of 1837.

By midcentury, the economy's growing geographic integration prompted the founding of the first national trade union, the National Typographical Union, in 1852 (Ulman 1955). It was followed by organizations of hat finishers (1854), journeyman stonecutters (1855), cigarmakers (1856), machinists, blacksmiths, and iron molders (1859). In 1866, another effort was made to form a national labor organization, the National Union, but like the NTU this effort soon collapsed. The depression that began in 1873 contributed to a substantial decline in union strength: membership declined sharply and fewer than ten of the thirty national unions in existence at the beginning of the decade survived into the second half of the 1870s.

As the economy rebounded at the end of the 1870s, labor organizations began to expand. Particularly successful was the Noble Order of the Knights of Labor. Founded in Philadelphia in 1869 by a group of garment cutters, the Knights of Labor remained a small organization until a series of successful railroad strikes in 1876–1877 attracted widespread attention to the group. In contrast to the craft-based organization of most previous unions, the Knights were organized into local and district assemblies, most of which welcomed workingmen regardless of craft or skill level. In addition to advocating the eight-hour day, the Knights promoted a variety of other social reforms, such as the graduated income tax and the prohibition of contract labor, and also sponsored consumer and producer cooperatives. In the first half of the 1880s, membership grew rapidly. By 1885, the organization had nearly 700,000 members nationwide. In the wake of several unsuccessful strikes and the backlash of public opinion following the 1886 Haymarket riot, however, the Knights were unable to sustain their previous momentum, and their membership collapsed as suddenly as it had expanded.

In the 1890s, a new, more narrowly focused organization, the American Federation of Labor (AFL), achieved considerable organizing success. Established in 1886 at a meeting of delegates from a number of national trade unions in Columbus, Ohio, the AFL was organized along craft lines, and its members limited their objectives to work-related issues, such as shorter hours, higher wages, and improved working conditions. With the emergence of the AFL, the modern era of labor organization may be said to have begun. Drawing on a per capita tax on all members to fund its activities, the AFL was able to successfully consolidate its position and expand its jurisdiction. Since the early twentieth century, several rival organizations have emerged to challenge the AFL's leadership of the labor movement. The most serious of these challenges occurred in the 1930s, when advocates of industrial unionism broke with the AFL leadership to establish the Committee for Industrial Organization (later renamed Congress of Industrial Organizations) (CIO). Eventually, however, differences with this group were bridged, and the two organizations merged to form the AFL–CIO, which remains the most important national labor organization in the United States today.

Back To Top

The growth of labor union membership from 1880 to the present is traced in Table Ba4783–4791. Several different series are available, reflecting different approaches to measuring union membership. Membership levels calculated from household interviews conducted as part of the Current Population Survey (CPS) (series Ba4788) are consistently below those in the series from Leo Troy and Neil Sheflin and the U.S. Bureau of Labor Statistics (BLS), which are based on union-reported data (series Ba4783–4786). The union-reported data may be higher because of double counting as a result of individuals holding membership in more than one union. However, it may also reflect a desire by these unions to inflate membership statistics for strategic reasons. Most scholars accept the individual survey data collected by the CPS to be a more accurate measure of union membership than earlier approaches based on union-reported data.

For 1897–1978, the BLS and Troy-Sheflin series provide alternative approaches to calculating union membership based on union-reported data. Whereas the Troy–Sheflin series relies consistently on the concept of dues-paying membership and makes use of union financial data, the BLS series is based on union-reported membership figures. Over the entire period, the two series reflect broadly the same patterns, but they differ more significantly over shorter subperiods. The divergence is especially great during the decade of the 1930s, with the BLS series showing a much sharper drop in membership in the early 1930s and a more rapid expansion of membership after 1933. On the face of it there is little basis for selecting one series over the other. Researchers must decide which concept of union membership is closest to the one they seek to measure. For 1897–1914, Gerald Friedman's recent revisions (series Ba4789) are clearly an improvement over both the Troy–Sheflin and BLS series, incorporating data from a broader and more consistent range of unions, as well as adjusting for Canadian membership in U.S. unions.

Despite the differences between the alternative series, all suggest a broadly similar chronology of union growth and decline. In absolute terms, union membership increased substantially between 1880 and the mid-1940s, when union density – the ratio of union membership to the nonagricultural labor force – reached a peak of around 35 percent (see Figure Ba-U). After 1945, the absolute number of union members continued to increase until 1976, when it began to decline. The growth of union membership after 1945 failed, however, to keep pace with the growth of the labor force, resulting in a decline in the relative number of union members beginning sometime in the early 1950s and continuing to the present.

An interesting feature of the membership data in Table Ba4783–4791 is that rather than increasing continuously, union membership and density have grown mainly during a small number of episodes of intense growth (see Freeman 1998, pp. 267–9). The first of these spurts occurred in the early 1880s, and it reflects the organizational successes scored in that decade by the Knights of Labor. Between 1883 and 1886, union membership more than tripled. The second spurt in union growth occurred from 1897 to 1904, when the AFL achieved significant gains in organizing skilled workers. During this period, membership more than doubled so that by 1904, more than one of every ten nonagricultural workers in the nation was a union member. After 1904, employer resistance stiffened, and union membership stagnated until World War I. Between 1916 and 1920, tight labor markets, together with government pressure to avoid interruptions of vital war production, combined to facilitate expanded organization, pushing the number of organized workers to new heights.

Perhaps the most unusual spurt in union membership is the one that began in 1933, at the height of the Great Depression. The rapid growth of union membership in an era of peacetime high unemployment reflects the massive shift in the legal environment of organized labor. In 1932, the Norris–La Guardia Act curtailed the federal courts' power to issue injunctions against striking unions, and in 1933, workers' right to organize and bargain collectively through representatives of their own choosing was given legal recognition in section 7(a) of the National Industrial Recovery Act (NIRA). The National Labor Board, chaired by Senator Robert F. Wagner, was created to settle disputes arising under this act and was authorized to hold elections of employees to determine their bargaining representation. Although the NIRA was invalidated by the Supreme Court in 1935, the labor provisions of the act were reinstated in the same year as part of the Wagner–Connery National Labor Relations Act. Under this act, the National Labor Board became the National Labor Relations Board (NLRB).

The final spurt in union growth occurred during World War II. Following on the heels of labor's advances in the 1930s, union successes during the war meant that by 1945, more than one third of nonagricultural workers were represented by unions. At this time, it appeared that organized labor was an established force in the American labor market. Throughout the 1950s, unions won well over half of all NLRB representation elections (see series Ba4946–4947), and the public's views toward organized labor remained strongly favorable, with two thirds to three quarters of those polled expressing approval of organized labor (series Ba4995).

Beginning in the 1960s, however, union successes began to erode and membership declined, first in relative terms, and then in absolute numbers. One immediate cause of declining membership was declining success rates in representation elections. Table Ba4946–4949 reveals that by the mid-1970s, unions were winning less than half of all representation elections, and fewer than half of the workers were eligible for representation in each election.

Part of the explanation for this decline is the growth of employer opposition to unionization. Employers can attempt to block unionization efforts legally by offering unorganized workers greater benefits and attempting to convince workers that voting against unions is in their best interest. Or they can use illegal means, such as identifying and firing leading pro-union workers (Freeman 1985, p. 52). The series in Table Ba4950–4953 provide several indexes of the extent of such illegal employer opposition to unions. Complaints to the NLRB of unfair management practices (series Ba4950) more than doubled between the mid-1950s and the early 1960s, continuing to rise to a peak around 1980, and complaints about workers being fired for union activity followed a similar path. The number of workers awarded back pay (series Ba4952) and reinstated (series Ba4953) grew in tandem with complaints, providing confirmation of the validity of these union charges. Increased management opposition was paralleled by the erosion of public support for organized labor. Table Ba4995–4998 shows a decline in public approval of and confidence in organized labor after the mid-1960s.

Back To Top

The earliest spurts in union membership in the 1880s and at the turn of the century were associated with innovations in union organization. The growth of unionization in the 1880s was tied to the emergence of the Knights of Labor, whereas growth in the 1897–1904 period was tied to the success of the craft-oriented and narrowly focused goals advocated by the AFL. The success of the AFL's strategy of organizing craft workers and concentrating on improving wages and working conditions is mirrored in the growth of unions and union members affiliated with the Federation. Series Ba4792 shows that between 1897 and 1904, the number of AFL-affiliated unions more than doubled, rising from 58 to 120. At the same time, the AFL's share of all union members increased from around 60 percent to close to 80 percent (series Ba4793).

The AFL's opposition to widespread efforts to organize unskilled workers along industrial lines was a continuing source of tension within the labor movement in the early twentieth century. In 1905, this discontent led to the formation of the Industrial Workers of the World (IWW) as an alternative vehicle for labor organization along industrial lines. Despite some initial successes, however, the IWW did little to erode the AFL's dominant position. The issue again arose in 1935, when advocates of organization through industrial unions within the AFL formed the Committee for Industrial Organization, later renamed the Congress of Industrial Organizations (CIO). Following this split, both the AFL and the rival CIO staged massive organizing campaigns, which contributed to the expansion of unionization at this time.

Two alternative sets of membership estimates are reported in Table Ba4792–4804. The BLS series Ba4793 and series Ba4795 attribute a much greater share of union membership to the CIO in the late 1930s than do the series constructed by Troy and Sheflin (series Ba4797–4798). The large discrepancies between these series illustrate some of the hazards in attempting to measure union membership in an era of substantial organizational changes. Even the head of the CIO, Philip Murray, was uncertain about the membership of the unions affiliated with his organization. In 1939, he indicated that per capita payments from affiliated unions implied a membership of 1.7 million, but he estimated that the actual membership figure may well have been as high as 3 million. Meanwhile, the BLS series Ba4795, which is based on reports from individual unions, indicates a figure of 4 million.

In 1955, the AFL and CIO agreed to merge, forming the American Federation of Labor–Congress of Industrial Organizations (AFL–CIO). Together, their membership accounted for close to 90 percent of union membership at the time. Table Ba4805–4813 shows that subsequently there has been some resurgence in independent or unaffiliated unions and a decline in the AFL–CIO share of union membership. Unfortunately, with the shift to reliance on individual survey data to measure union membership, the government ceased to collect data on membership of individual unions or their affiliations, and so these series end in the early 1980s.

Back To Top

Additional insight into the growth and decline of organized labor can be gained by decomposing membership along industrial and demographic lines. Tables Ba4814-4883 report a number of series showing the breakdown of union membership along industrial lines for different subperiods. Because the sources for each table are different, and the industry definitions are not entirely consistent, it is not possible to construct a single continuous set of industry-level membership series.

At the turn of the twentieth century, the major centers of union strength were in the building trades (series Ba4816), mining (series Ba4815), and transportation and communications (series Ba4825), especially the railroads. Together, these three groups of unions accounted for over half of all unionized workers from 1900 through 1934. Although it is more difficult to calculate union density by industry, the proportion of workers belonging to unions in these industries was well above the national average (Wolman 1936, Chapter 8).

In the mid-1950s, when industry-level statistics are again available, the relative importance of manufacturing workers in the union movement had increased considerably. According to series Ba4832, manufacturing workers made up close to half of all union members in 1956. The share of unionized workers employed in manufacturing has fallen more or less continuously since 1956, however, so that by 1999, manufacturing workers accounted for less than 20 percent of the unionized workforce (series Ba4870). The decline of the relative importance of manufacturing has been paralleled by the increasing role of government employees in the union sector (series Ba4842 and series Ba4880). This growth was facilitated by a series of actions greatly expanding the scope for organized labor in the public sector. In 1962, President John Kennedy granted federal employees the right to organize and bargain collectively. Similar actions by many state legislatures granted similar rights to state and municipal employees.

With the exception of data on union membership by sex, which was collected as early as 1954, there is little information about the demographic characteristics of union membership prior to 1983. Table Ba4906–4907 shows a rising proportion of women among organized workers. For the most part, this increase mirrors the rising participation of women in the paid labor force over this period. In 1954, women workers were substantially less likely than men to be members of a labor organization: 14.9 percent of women versus 31 percent of men were union members. Although the gap has narrowed, this imbalance still persists. In 1999, union density was 12.8 percent among women and 16.7 percent among men.

Beginning in 1983, data on union membership collected as part of the CPS provide a much richer picture of the characteristics of the unionized workforce. With the exception of the rise in women union members and some decline in the proportion of younger workers (series Ba4908), there are few pronounced trends in these series. These data are most interesting when set against the proportions of each group in the labor force as a whole. The greater rate of unionization among men has already been noted. It turns out as well that blacks are more likely to be union members than whites (17.2 percent of blacks versus 13.5 percent of whites in 1999), that full-time workers are more likely to belong to unions than part-time workers (15.4 percent versus 7.0 percent), and that younger workers, those under 25, are less likely than older workers to belong to unions (5.7 percent versus 15.5 percent).

Back To Top

The presence of unions affects almost every facet of the employment relationship. Unions not only affect wages but also alter the compensation package, increasing the share of fringe benefits. They also affect the distribution of wages, the likelihood of quits and layoffs, the response of firms to cyclical fluctuations in demand, promotion practices, productivity, and firm profitability.² Disentangling the effects of unions per se from other characteristics in each of these cases is not simple, however, and most examinations of these impacts have relied on cross-sectional data and sophisticated econometric techniques to measure these impacts.

The one area that has been relatively well documented concerns unions' effects on wages. Even here, data are only available for a relatively short period.³ Table Ba4928–4945 compares average hourly wages of union and nonunion workers within a variety of demographic groups. It is important to recognize that the comparisons possible with these data reflect a variety of different factors. First, because employers may raise wages of nonunion workers to deflect organizing drives – so-called union threat effects – wage differentials may not fully capture the impact that unionization has on wage levels. Second, because workers are heterogeneous, unionized workers may differ systematically in characteristics from nonunion workers. Third, the industrial distribution of unionized and nonunionized workers may differ. As a result, these differentials may not accurately reflect the change in wages that would be achieved by an individual worker moving between the union and nonunion sectors. Bearing these caveats in mind, it is apparent that the presence of unions is associated with substantially higher wages, and that this effect is most pronounced for blacks and women. Indeed, whereas white male union members earn only a 10–15 percent premium, the union wage premium for black men is from 20 to 30 percent, and that for black women is from 30 to 40 percent.

Back To Top

Strikes are one of the principal tools that workers can use to exert pressure on employers. In the early history of the union movement, strikes were an important vehicle used to gain employer recognition of organized labor's right to bargain. Today, they are used primarily as a means to enforce worker demands for better contract terms. By their nature, strikes are highly visible, and the apparent growth in labor conflict manifested in strikes during the 1880s was a major public policy concern. Consequently, data on strikes were one of the first kinds of economic statistics collected by the Federal Labor Statistics Bureau after its establishment in 1884. Table Ba4954–4964 summarizes a number of aggregate measures of strike activity beginning in 1881. Data on all strikes are available until 1981, but after that date, the U.S. Department of Labor began to collect data only for strikes involving 1,000 or more workers. Although only about 5–10 percent of strikes are so large, this subset of strikes accounts for two thirds to three quarters of all striking workers, and during the years when the two series overlap, their movements are highly correlated.

Figure Ba-V depicts two indexes of the number of workers involved in strikes, constructed by using changes in the number involved in large strikes to extrapolate the series for all strikes after 1981. The first index is simply a measure of the relative number of workers on strike in each year, whereas the second index shows changes in the ratio of workers on strike to nonagricultural employment in each year. Both indexes are characterized by large cyclical fluctuations in magnitude. Prior to World War II, episodes of rapid growth in union membership – the early 1880s, the late 1890s, World War I, and the late 1930s – were also typically characterized by sharp increases in strike volume. In the 1950s, both the number of strikes and the number of workers on strike were at historically unprecedented levels. Strike activity declined from its peak in the 1950s through the late 1960s, before experiencing another upsurge. Since the mid-1970s, the number of striking workers has fallen. In terms of absolute numbers of workers, strike activity follows something of an inverted-U shape, rising from relatively low levels in the 1880s to a peak immediately after World War II, and then trending downward. Relative to nonagricultural employment, however, there appears to be no long-term trend in worker involvement in strikes from 1881 through 1980. However, since 1980, strike activity has fallen off sharply.

Data on the industrial distribution of strikes from 1937 to 1981 are reported in Table Ba4971–4994. In the 1950s, the majority of striking workers were employed in manufacturing industries, but the share of striking workers employed in nonmanufacturing industries has grown consistently since 1955, with the result that by the mid-1960s, nonmanufacturing workers made up a majority of strikers. Information on the industrial distribution of union membership is incomplete for this period, but it is clear from the available data (Table Ba4814–4831, Table Ba4832–4844) that prior to 1968, the share of striking workers drawn from the manufacturing industry far exceeded manufacturing's share of unionized workers. From 1968 through 1978, the share of striking workers in manufacturing was more closely aligned with manufacturing's contributions to the unionized workforce.

For most of the period that data on strikes have been collected, government officials have attempted to identify the primary "causes” of each strike. Of course, strikes occur for a wide variety of reasons, and in many instances, there may be multiple factors involved. Nonetheless, some additional insights can be gained by classifying strikes into a small number of relatively broad categories on the basis of what appears to be their primary cause. Table Ba4965–4970 shows both the number of strikes and the number of workers involved, classified by three major categories of strike issues: (1) wages and hours, (2) union recognition, and (3) other issues. In the late nineteenth century, wages and hours issues were the predominant cause of strikes, but the proportion of strikes over union organization rose rapidly while wage and hours issues were declining in importance from 1880 through the 1910s. During the 1930s, when overall strike activity was relatively high, union organization was the chief cause of strikes. But between 1940 and 1981, when the data end, wage and hours issues once again rose in importance, while strikes over union organization diminished in importance.

Back To Top

Union membership: 1890–1914, series Ba4789; 1915–1976, series Ba4783; thereafter, series Ba4788.

Nonagricultural employment: 1890–1937, series Ba471 minus series Ba472; thereafter, series Ba481 minus series Ba482. For the years before 1890, figures are extrapolations based on data from Robert Margo, "The Labor Force in the Nineteenth Century,” in The Cambridge Economic History of the United States, volume 2, The Long Nineteenth Century, edited by Stanley L. Engerman and Robert E. Gallman (Cambridge University Press, 2000), pp. 209, 213.

For the years before 1890, the nonagricultural labor force was extrapolated on the basis of changes in population, adjusted for shifts in the sectoral distribution of the labor force and changes in the labor force participation rate. The Margo source reports unpublished figures from Thomas Weiss showing that the labor force as a share of total population increased from 34.7 percent in 1880 to 37.4 percent in 1890, while agriculture's share of the labor force fell from 47.7 percent in 1880 to 40.1 percent in 1890. Linear interpolations were used to create annual series for labor force participation and for the nonagricultural share of the labor force. Indexed versions of these series were multiplied together and then multiplied by an indexed version of the annual U.S. population figures from series Aa7 (with all indexes set to 1890 = 1). The result was a composite index that could be used to extrapolate the available 1890 value for the nonagricultural labor force (series Ba471 minus series Ba472) back to 1880.

Back To Top

Based on series Ba4955 and series Ba4962, as well as the nonfarm employment data used in Figure Ba-U.

The index for workers involved in strikes was computed by first indexing series Ba4955 and series Ba4962 to 1960 = 100. Then the indexed series were linked at 1981–1982.

The index for workers involved in strikes as a share of nonfarm employment was computed by first expressing series Ba4955 and series Ba4962 as a share of the nonfarm employment data used in Figure Ba-U. These ratio series were then indexed to 1960 = 100. Finally, the indexed series were linked at 1981–1982.

In both cases, the link was made so that the values based on series Ba4955 cover the years through 1981 and those based on series Ba4962 cover later years.

Back To Top

Currie, Janet, and Joseph Ferrie. 2000. "The Law and Labor Strife in the United States, 1881–1894.” Journal of Economic History 60 (1): 42–66.

Freeman, Richard B. 1985. "Why Are Unions Faring Poorly in NLRB Representation Elections?” In Thomas A. Kochan, editor. Challenges and Choices Facing American Labor. MIT Press.

Freeman, Richard B. 1998. "Spurts in Union Growth: Defining Moments and Social Processes.” In Michael D. Bordo, Claudia Goldin, and Eugene White, editors. The Defining Moment: The Great Depression and the American Economy in the Twentieth Century. University of Chicago Press.

Freeman, Richard B., and James L. Medoff. 1984. What Do Unions Do? Basic Books.

Lewis, H. Gregg. 1963. Unionism and Relative Wages in the United States: An Empirical Inquiry. University of Chicago Press.

Pencavel, John, and Catherine E. Hartsog. 1984. "A Reconsideration of the Effects of Unionism on Relative Wages and Employment in the United States, 1920–1980.” Journal of Labor Economics 2 (2): 193–232.

Ulman, Lloyd. 1955. The Rise of the National Trade Union. Harvard University Press.

Wolman, Leo. 1936. Ebb and Flow in Trade Unionism. National Bureau of Economic Research, number 30.

Back To Top