Designs of Empirical Studies in ecological perspective: National Labor Unions

It may seem ironic that we studied unions at a time when labor historians have largely abandoned the subject, preferring to study noninstitution- alized forms of labor action (see Gutman 1976, for example). The American labor movement has lost its momentum. In each year since 1957, the fraction of the civilian labor force affiliated with labor unions has declined. The history of the union movement in the United States is now being rewritten as a history of failure—the history of a movement that has fallen victim to its own limited earlier success.

This state of affairs does not, however, diminish the value of the roughly 150-year history of this movement as a source of data about organizational ecology. It may be true that national labor unions in the United States adapted narrowly to peculiar economic and social conditions that are disappearing and may not be repeated. In particular, the fact that labor unions can almost never organize across national boundaries (except in the case of the United States and Canada) while capital and capitalist production move freely across such boundaries is a notable problem of fit for the union movement. The current form of union organization may be an evolutionary dead end. Evolutionary biologists and paleontologists continue to be fascinated with dinosaurs and the evolutionary situation they typify; perhaps we can profit from the study of what may become sociological dinosaurs.

Analyzing the ecology of labor unions allows us to deal with two common criticisms discussed in Chapter 2: that organizational ecology cannot deal effectively with large and powerful organizations, and that it cannot deal with “non-market” organizations. Some unions have grown large and powerful; the largest unions have had more than a million members. Individually and collectively, labor unions have had a major impact on American politics and on industrial structure. So it is hard to argue that they have been especially small and powerless organizations. At the same time, unions are not profit- seeking participants in markets as conventionally defined. So analysis of the dynamics of populations of unions should help to clarify the boundaries of application for organizational ecology.

Another advantage is a tactical one. As we have noted, much organiza- tional research is limited by the fact that all but the largest and longest- lived organizations are ignored. We have already noted the importance of studying the full range of variation on these dimensions. In the case of national labor unions, the record contains information on numerous small unions with short lifetimes (often less than a year). The richness of the record reflects the adversary and often radical nature of many new unions. The formation of national labor unions, no matter how small, has apparently seldom escaped notice and mention in the press. In addition, labor historians and the older schools of institutional economists (especially the Wisconsin school under John R. Commons) have made a sustained effort to reconstruct the historical record, painting a broad picture of the evolution of unions and pointing to many of the relevant primary sources. For these reasons, the study of national labor unions can avoid the problem of bias resulting from endogenous sampling (sampling on the dependent variables, size and longevity).

The available data are also rich in the time dimension: the record extends back 150 years. Since the length of the period is long relative to founding rates and disbanding rates of individual unions, components of selection processes can be studied directly.

1. Diversity of Forms

Collective action directed at affecting levels of compensation, conditions of work, control over jobs, and other more fundamental changes in the organization of work has taken many forms. Sometimes it consisted of bursts of spontaneous collective action by workers and their families. Some episodes of collective action were directed narrowly at employers at the work site, as in strikes and boycotts. Other episodes were directed at other corporate actors such as legislative and judicial bodies, as in general strikes and political demonstrations.

At various times each kind of collective action by workers became em- bodied in permanent formal organizations. Organizations of workers that are directed at controlling the work relation are usually called labor unions. Those taking on broader political agenda are usually called worker political movements or political parties. The distinctions between these more or less organized forms of collective action are sometimes blurred in historical reality. Strikes sometimes begot unions; unions sometimes transformed themselves into political movements; political parties sometimes created unions, demonstrations, and strikes. Although analyzing the ecology of the full diversity of forms of worker collective action would be a fascinating study, we did not have sufficient resources to attempt this. Instead, we decided to study only national labor unions for the reasons mentioned earlier.

A labor union is defined here as a permanent organization of workers with the ostensible goal of affecting the conditions of work by threatening the collective withholding of labor. The restriction to an organization that is intended to be permanent distinguishes a union from a prolonged strike or boycott. The restriction to the goal of affecting conditions of work distinguishes a union from broader political movements of workers (such as socialist parties) and from utopian movements (such as Owenite community movements of the 1870s). The restriction to the device of threatening to withhold labor distinguishes a union from employee associations (such as company-sponsored unions), mutual-benefit or benevolent associations of workers, and organizations of workers that try to affect working conditions by other means (such as terrorist organizations).

We studied only national labor unions, those that organized in more than one state. However, we note briefly the other kinds of union organization that have existed in this country. The point of this overview is to show that there has been considerable diversity of form in the union movement in the United States and that much of this diversity occurs within the category of national unions.

The most obvious dimension of diversity among American labor unions is scope of organizing. Unions have varied widely in the breadth of types of workers (crafts and occupations) and types of industries they tried to organize. The forms defined on this dimension can be described as follows.

Single-craft, single-city unions. The first permanent labor organizations in the United States, called trade societies, organized journeymen in a single craft in a single city. The first was apparently the Federal Society of Journeymen Cordwainers, founded in Philadelphia in 1794. The trade society form was dominant until the Panic of 1837, which destroyed most unions.

Multicraft, single-city unions. Next federations of trade societies within cities were formed. The first, the Mechanics Union of Trade Associations, established in Philadelphia in 1827, initially included carpenters, painters, bricklayers, and glaziers. The New York Trades’ Society included 52 crafts by 1836 (before being wiped out by the results of the Panic in the next year).

National craft unions. The period just before the Panic of 1837 saw the creation of national unions of craftsmen. Apparently the first was the Society of Cordwainers, founded in 1836 by a convention of representatives of trade societies of cordwainers (shoemakers) from 16 cities. In the same year, the Society of Journeymen House Carpenters and the National Typographic Society were also begun. Mittleman (1927, pp. 452-453) reports that national unions of comb makers and of hand loom weavers were also created in 1836.

Multicraft unions. As industrialization increased in scope, it became apparent that many craft unions had organized too narrowly. A number of mergers took place in the late nineteenth century among unions whose members had a relatively high degree of shared fate by virtue of working at the same site (meaning that a strike by one union would stop work for the others), working with substitutable processes or materials (for example, painters and wallpaperers), or working along a vertical flow of work (for example, warehousemen and stevedores). For instance, the Bricklayers’ International Union (1865-1985) absorbed the Stonemasons’ Union in 1883.

More commonly, multicraft unions were created by expansion of the jurisdiction of a union, with no formal merger. The most spectacular example is the United Brotherhood of Carpenters. The UBC began with its membership consisting only of carpenters and joiners. In 1901 it expanded its jurisdiction to cover “all that’s made of wood,’’ later extended to “or that was ever made of wood.’’ Using this jurisdictional claim, the UBC eventually organized millwrights, pile drivers, bridge, deck, and wharf carpenters, underpinners, timbermen, boat builders, caulkers, cabinet makers, millmen, floor layers, shinglers, insulators, acoustic and dry wall applicators, house movers, loggers, lumber and sawmill workers, furniture workers, casket and coffin makers, and others.

National industrial unions in a single industry. Craft unionism did not always match well with the changing nature of industrial organization in later periods. Semiskilled and unskilled workers played an increasingly important role as technical innovations replaced some kinds of skilled workers, the scale of enterprise grew, and technical change kept reshuffling the roster of identifiable jobs. What came to be known as industrial unionism used a different organizing principle. Instead of defining its jurisdiction in terms of a set of crafts or occupations, an industrial union attempted to organize all production workers in a set of industries, regardless of job title or skill level.

The first unambiguous industrial union in the United States was the American Miners’ Association (1861-1867), which attempted to organize all workers “in and around the mines.” The Knights of St. Crispin (1867-1878), which tried to organize all workers in the shoe industry, became the first really large union, with membership estimated at half a million.

Multiple-industry industrial unions. Some industrial unions organized workers in more than one (broadly defined) industry. The most interesting example is the Industrial Workers of the World (1905-1921). The IWW began as a mixed syndicalist political movement and labor organization. However, it directed most of its attention to organizing semiskilled and unskilled workers in mills, agriculture, mining, and lumbering. It had its greatest success in the textile industry (following its support of the famous strike in Lynn, Massachusetts), in the timber industry of the Pacific Northwest, and among tenant farmers and sharecroppers in the Southwest. For a brief period the IWW mobilized thousands of workers in a truly radical union movement. The IWW was crushed by the state for its radical politics during the First World War and had ceased to operate as a national union by 1921.

This brief sketch of the evolution of labor union forms suggests that the composition of the population of labor unions has changed markedly over time. But there is no simple trend toward increasing breadth at the level of the national union, as is often asserted (see Brooks, 1937, for example). Many of the most successful nationals have retained a narrow craft orientation. In at least one well-known case, that of the International Typographical Union (1850-1985), the direction at the organization level was toward more narrow organization. The ITU, though initially dominated by compositors, included all of the crafts in the printing trade as well as journalists. However, each of the other crafts eventually split off and began independent unions, making the ITU a singlecraft union.

We think that the dynamics of diversity in union populations have been more contingent than a simple trend hypothesis implies. Unions merged and then dissolved mergers; they admitted new classes of members and then expelled them; they lost whole classes of members as production processes became obsolete. More important to our perspective, the life chances of narrow craft and broad industrial unions seem to have depended on the nature of the industrial and political environments.

2. Research Design

We have tried to collect information about every national labor union that has existed, however briefly, in the United States. The first step was to compile the lists of names (with starting dates) contained in reports pub lished in various years that claimed exhaustive coverage of the population of unions. We found listings for the following years:

1887 (New Jersey Bureau of Labor Statistics 1898)

1893 (Finance 1894)

1901 (Industrial Commission Report, vol. 17, 1901)

1926 (U.S. Bureau of Labor Statistics 1926)

1936 (Stewart 1936)

1944 (Peterson 1944)

1956 (National Industrial Conference Board 1956)

1962 (Troy 1965)

1975 (Fink 1977)

1985 (Gifford 1985)

These publications, supplemented by annual reports of the Department of Labor from 1932 through 1985, yielded an initial master list of unions. We extended the master list by consulting standard histories of the labor move- ment, which were especially useful for the period 1830-1870. We relied mainly on reports in Commons et al. (1927), Foner (1947), and Fink (1977).

Each union on the master list was given an identification number, and a file was begun. As the search for information progressed, we found that records of the existence of marginal and short-lived unions which had been ignored previously were sometimes noted in historical accounts of unions on the master list. Each newly identified union was added to the master list, and the process was repeated. We eventually identified 633 unions on which some usable data could be coded.24 Although there is no way to tell exactly what fraction of all national unions this number represents, we believe that we have identified most members of the population.

In collecting data on individual unions, we relied heavily on published histories of unions, of the union movement, and of industries. When it was feasible, we also used union periodicals and proceedings. We did not try to code data from general newspapers because the relevant events occurred in many cities over a long time period.

3. Unit of Observation

A national labor union stands midway in a hierarchy of organizational forms. In almost all cases, it is composed of local unions. In some ways national unions are federations of locals. At the same time, many national unions are embedded in federations of nationals, such as the American Federation of Labor (AFL) and the Congress of Industrial Organizations (CIO). Each level is a meaningful actor for some kinds of processes.

One way of conceptualizing the relationship between nationals and locals is to think of the national as a population of local unions. From this perspective, the fate of a national reflects the dynamics of the population of locals. In particular, the demise of a national would be equivalent to the extinction of a population. This reasoning makes it appropriate to apply population logic to the histories of individual national unions.

The main problem with this approach is that national unions frequently disappear through merger with other nationals. Although merger requires approval of the majority of the membership, it is an event that pertains to the national per se. This kind of event cannot be reduced to events at the level of the locals. Therefore, we consider individual national unions to be the appropriate units of observation, acknowledging that their subunits, the locals, are partially autonomous.

4. Records of Vital Events

Each observation begins with a starting time and a starting event:

  1. Founding: A national union is formed by the joint decision of several locals or by some unorganized group of workers; of the 633 unions, 479 (76%) began with a founding.
  2. Secession: A union is formed by a faction that secedes from a national union—for example, the National Woolsorters’ and Graders Union (1909-1912) was created by a split in the United Textile Workers; 95 unions (15%) began this way, including 7 that had previously been independent unions.
  3. Two or more unions merge to form a new national union— for example, the Fur Workers’ International Union (1913-1939) merged with the National Leather Workers’ Association (1933-1939) to form the International Fur Workers and Leather Workers’ Union (1939- 1955); 55 unions (9%) began with mergers.
  1. Transformation: An employee association adopts the goal of collec-tive bargaining and becomes classified as a labor union (for example, the National Education Association announced that it would engage in collective bargaining in 1947); 4 unions were created in this way.

The record for each union also includes an ending date and an ending event. The main complication in evaluating ending events concerns mergers. We have characterized each merger as one of three types for each partner: (1) the organization is absorbed without changing the strategy and structure of the dominant party to the merger; (2) the organization absorbs another; and (3) the organization merges with one or more roughly “equal- status” partners and a new organization results. We use information on name changes, distributions of executive positions, and the sizes of the partners to make these distinctions.

Thus we distinguished four ways in which a union record might end:

  1. Right-censoring: The data record ends before the union has ended— the data are censored on the right (for example, the Coopers’ Interna- tional Union, founded in 1890, was still in existence when our coding ended in 1985); of the 633 unions, 160 (25%) were right-censored.
  2. Disbanding: A union is dissolved (for example, the National Forge of the United Sons of Vulcan, formed in 1862, disbanded in 1876). A total of 203 (32%) unions disbanded. However, 12 of these events occurred on the same day to members of the Trade Union Unity League at the order of the Comintern. Since these were obviously not independent events, we exclude these 12 unions in our analysis of disbandings. Therefore, the effective count of disbandings is 191.
  1. Absorption: A union is absorbed by merger to a dominant union (for example, the Tunnel and Subway Constructors’ Union, founded in 1910, was folded into the International Laborers’ Union in 1929); 140 (22%) unions were absorbed in mergers.
  1. Equal-status A union merges with one or more “equal-status partners,” in which case the records for all partners end and a new record begins—a type 3 starting event (for example, the Oil Workers’ International Union, founded in 1937, merged with the Union of Gas, Coke, and Chemical Workers of America, founded in 1943, to form the Oil, Chemical, and Atomic Workers of America in 1955); 130 (21%) unions ended in such mergers.

Range of Occupations Organized

We have attempted to code each union’s initial organizing strategy together with the dates of all subsequent major changes and the new strategies adopted. In order to do so reliably for a broad range of unions over a long period, we defined organizing strategies in simple and relatively conventional ways.

Labor historians appear to agree that specialism in terms of the range of jobs and occupations to be organized is the key distinguishing element among different kinds of unions. Thus we paid most attention to coding variation on this dimension, using the following categories:

  1. Single narrow occupation: The union organizes only members of one narrowly defined craft or occupation, usually skilled.
  2. Composite occupation: A union organizes members of several crafts or jobs which involve some differentiated skills and training but which share a common set of skills, training, and experience.
  1. Several occupations: A union organizes two or more separate work roles. Often these work roles are skilled crafts or occupations distin- guished by training, experience, and sometimes credentials. Although there may be common skills (for example, the ability to read blueprints), the variations in training and work tasks are more salient than the commonalities.
  1. All production occupations: A union organizes all of the production workers at the work site.

We simplified this coding scheme in the analyses reported in later chapters by collapsing the first three into one “craft union” category. We call the other category an “industrial union.”

Skill Levels

Unions also specialized on the skill dimension. Some unions organized a narrow range of skills; others organized mixes of skills. Although no single skill hierarchy applies across the entire work force, each industrial sector recognizes some crude skill hierarchy. As we pointed out earlier, many unions were begun mainly by skilled craft workers. The question of admitting less skilled workers usually generated intense political conflict within unions. Changes in the mix of skill levels provide a potentially interesting case of adaptive change. As technical changes made narrow craft organizations increasingly unfit, the gains from broadening the mix of skills increased. Yet doing so threatened the privileged position of skilled workers. Struggles for political control within the union sometimes threatened its existence as much or more than continuing with an obsolete craft form. We can learn about both the rates of such adaptive changes and their consequences for survival chances by coding event histories of changes in skill levels.

We do not pretend that we can code skill-level composition uniformly across industries. The distinctions among skill levels that unions have made and that labor economists and sociologists recognize are at least partly specific to the history of each industry. Industries that began earlier are more likely to contain clearly defined crafts and skill hierarchies than more recent industries. Moreover, skill levels imputed to positions change over time. A review of census classifications over the period from 1890 to 1920 shows that many jobs once considered skilled became less skilled (that is, semiskilled) when the production process became mechanized. This was the case for cigarmakers, chainmakers, tool makers, and others. The semiskilled category is especially problematic. In many ways it has been used as a residual category for cases about which it was hard to decide whether the skills needed for the job required much training. Therefore, it includes such jobs as sailor, brakeman, meatcutter, weaver, sawyer, and welder.

Instead of trying to code absolute distinctions between skill levels, we identified times at which unions widened or narrowed the mix of skills represented. This always meant organizing some new class of workers or dropping some class of workers who were considered to have a different skill level. In particular, we did not try to code changes over time in the skill levels of the same class of workers. For example, when cigarmaking was demoted in census classifications to a semiskilled job, we did not record a change for cigarmakers’ unions on skill composition. Thus our measure of changes in skill composition refers to changes in the kinds of workers a union tried to organize rather than to changes in the real skills possessed by its members or to changes in the processes of production.

Our histories of changes in skill composition record a union’s initial position: which skill levels it organized, and the dates of all subsequent changes with the new mix. The basic coding categories include:

  1. Professional workers: Examples include the American Federation of Physicians and Dentists (1973-1985) and the White Rats Actors Union (1900-1919).
  2. Proprietors, managers, and officials: For example, the National League of District Postmasters (1894-1976) organized managers and the Journeymen Barbers, Hairdressers, and Cosmetologists (1887—1984) represented proprietors.
  1. Clerks and kindred workers: Examples are the Insurance Workers of America (1950-1959) and the United National Association of Post Office Clerks (1899-1961).
  1. Skilled workers: Examples of unions of skilled workers are the Jour- neymen Tailors’ National Trades Union (1865-1876), the Diamond Workers’ Protective League (1902-1955), and the Brotherhood of Commercial Telegraphers (1898-1900).
  2. Semiskilled workers: Examples include the Seafarers’ Union (1891—1985) and the International Ladies Garment Workers’ Union (1900- 1985).
  1. Unskilled workers: Examples are the United Farm Workers’ Union(1962-1985) and the Brotherhood of Railroad Freighthandlers (1901— 1915).

There has been considerable diversity in coverage of skill levels by American unions. The commonest strategy was to organize only skilled workers (N 155). Yet 149 unions organized unskilled and semiskilled workers. We simplified the conventional census classifications to provide a binary variable, which we call LOWSKILL. It receives a value of one if all of the workers organized by the union are in occupations classified as “semiskilled” or “unskilled”; otherwise its value is zero.


We also collected data on the dates of affiliation with federations. Of course, the main federations over the recent period have been the AFL and the CIO. Other affiliations that show up in our data are the Knights of Labor, the (officially Communist) Trade Union Unity League, the IWW, and the Alliance for Labor.

In addition to these event histories, we collected whatever data were available on membership. Such data are available after 1930 but are scanty before then for all but AFL unions. We can, however, obtain information about membership size around the time of founding for 519 unions.

Finally, we collected diverse information about the environments facing unions. We discuss these data in Chapters 9 and 11 in the context of presenting analyses of ecological processes.

Source: Hannan Michael T., Freeman John (1993), Organizational Ecology, Harvard University Press; Reprint edition.

Leave a Reply

Your email address will not be published. Required fields are marked *