1. Address delivered on March 21–22, 1907, pp. 269, 272. From an address reprinted in The Open Shop publication. Allen, Andrew J., “The Benefits of Organization,” The Open Shop (1907): 269–77Google Scholar.

2. Employers’ Association of Detroit, “Advertisement,” Detroit Saturday Night, 1926. This advertisement is cited in Dunn, Robert W., The Americanization of Labor: The Employers’ Offensives Against the Trade Unions (New York: International Publishers, 1927), 87–88Google Scholar.

3. For recent comparative political economy accounts that theorize employer associations as largely cooperative with organized labor, see Hall, Peter A. and Soskice, David G. E., ed., Varieties of Capitalism: The Institutional Foundations of Comparative Advantage (Oxford: Oxford University Press, 2001)Google Scholar; Jo Martin, Cathie and Swank, Duane, “Does the Organization of Capital Matter? Employers and Active Labor Market Policy at the National and Firm Levels,” American Political Science Review 98, no. 4 (2004): 593–611Google Scholar; Cusack, Thomas R., Iversen, Torben, and Soskice, David, “Economic Interests and the Origins of Electoral Systems,” American Political Science Review 101, no. 03 (2007): 373–91Google Scholar; Crouch, Colin, Industrial Relations and European State Traditions (Oxford: Clarendon Press, 1993)Google Scholar.

4. Foner, Philip S., History of the Labor Movement in the United States. Vol. 8: Postwar Struggles, 1918–1920 (New York: International Publishers, 1987)Google Scholar.

5. Martin argues that the NAM was perceived to be too linked to the Republican Party, preventing it from being a true corporatist institution. Jo Martin, Cathie, “Sectional Parties, Divided Business,” Studies in American Political Development 20 (Fall 2006): 160–84Google Scholar. The most relevant discussion of the electoral systems argument is found in Martin, Cathie Jo and Swank, Duane, “The Political Origins of Coordinated Capitalism: Business Organizations, Party Systems, and State Structure in the Age of Innocence,” American Political Science Review 102, no. 02 (2008): 181–98Google Scholar; Martin, Cathie Jo and Swank, Duane, The Political Construction of Corporate Interests: Coordination, Growth, and Equality (New York: Cambridge University Press, 2012)CrossRefGoogle Scholar; Martin, Cathie Jo and Swank, Duane, “Gonna Party Like It's 1899: Party Systems and the Origins of Varieties of Coordination,” World Politics 63, no. 1 (2011): 78–114Google Scholar.

6. Putting the U.S. outcome in comparative context, Martin and Swank argue that proportional electoral systems promote and solidify corporatist employer associations while majoritarian electoral systems inhibit them; thus initial employer associations in the United States did not become more encompassing due to its party system. Cusack, Iversen, and Soskice propose a similar causal logic but advocate the reverse direction: in their account, different kinds of pre-existing economic coordination affect the choice of electoral system. Cusack et al., “Economic Interests and the Origins of Electoral Systems.” Both sets of explanations highlight the importance of electoral systems in the development of coordination among firms, as well as that of long-term industrial institutions that create incentives for firms to establish cooperative institutions with workers. Neither theoretical view dwells on the development of repressive employer associations. However, Swenson discusses the importance of the lockout strategy to employer associations in Sweden. Swenson, Peter, Capitalists against Markets: The Making of Labor Markets and Welfare States in the United States and Sweden (Oxford: Oxford University Press, 2002)Google Scholar.

7. The term comes from Harris, Howell John, “Employers’ Collective Action in the Open-Shop Era: The Metal Manufacturers’ Association of Philadelphia, c. 1903–1933,” in The Power to Manage? Employers and Industrial Relations in Comparative-Historical Perspective, ed. Tolliday, Steven and Zeitlin, Jonathan (London: Routledge, 1991), 118–46Google Scholar. For other accounts of sectors of firms that sought cooperation with unions (with evidence from largely before WWI), see Bowman, John R., Capitalist Collective Action: Competition, Cooperation and Conflict in the Coal Industry (New York: Cambridge University Press, 1989)Google Scholar; Carpenter, Jesse Thomas, Competition and Collective Bargaining in the Needle Trades, 1910–1967 (New York: State School of Industrial and Labor Relations, 1972)Google Scholar.

8. Such an organization could also have rules to constrain member firms from recognizing unions, though of course organizations varied in the degree of enforcement of such rules.

9. On the NMTA, see Haydu, Jeffrey, “Employers, Unions, and American Exceptionalism: Pre-World War I Open Shops in the Machine Trades in Comparative Perspective,” International Review of Social History 33, no. 1 (1988): 25–41Google Scholar; Haydu, Jeffrey “Two Logics of Class Formation? Collective Identities Among Proprietary Employers, 1880–1900,” Politics & Society 27, no. 4 (1999): 505–25Google Scholar. See the excellent account on select open shop cities by Pearson, Chad, Reform or Repression: Organizing America's Anti-Union Movement (Philadelphia: University of Pennsylvania Press, 2016)CrossRefGoogle ScholarPubMed. For the most detailed evidence of the MMA, see magisterial accounts of Harris, Howell John, Bloodless Victories: The Rise and Fall of the Open Shop in the Philadelphia Metal Trades, 1890–1940 (Cambridge: Cambridge University Press, 2000)CrossRefGoogle Scholar; Harris, “Employers’ Collective Action in the Open-Shop Era”; Harris, “Getting It Together: The Metal Manufacturers’ Association of Philadelphia, c. 1900–1930,” in Masters to Managers: Historical and Comparative Perspectives on American Employers, ed. Jacoby, Sanford (New York: Columbia University Press, 1991), 111–32Google Scholar. For a summary of the response of the MMA to local IMU strikes, see the excerpt in Brenner, Aaron, Day, Benjamin, and Ness, Immanuel, The Encyclopedia of Strikes in American History (Armonk, NY: M.E. Sharpe, 2009)Google Scholar.

10. Fine, Sidney, Without Blare of Trumpets: Walter Drew, the National Erectors’ Association, and the Open Shop Movement, 1903–57 (Ann Arbor: University of Michigan Press, 1995)CrossRefGoogle Scholar.

11. Harris, “Getting It Together”; Harris, Bloodless Victories. For a description of the open shop ideology at work in Worcester, Massachusetts, see Pearson, Chad, “Making the ‘City of Prosperity’: Engineers, Open-Shoppers, Americanizers, and Propagandists in Worcester, Massachusetts, 1900–1925,” Labor History 45, no. 1 (February 2004): 9–36Google Scholar.

12. Voss also finds that craft conservatism, ethnic diversity, and industrial diversity were less important factors explaining union successes and survival, and that local employer associations explains the reduction in KL presence in New Jersey between 1889 and 1895. Voss, K., The Making of American Exceptionalism: The Knights of Labor and Class Formation in the Nineteenth Century (Ithaca, NY: Cornell University Press, 1993)Google Scholar. This emphasis on employer repression as a cause of weakness in the Knights of Labor contrasts with the previous emphases on variables related to the influx of unskilled workers and craft consciousness. The classic arguments are in Foner, Philip Sheldon, History of the Labor Movement in the United States. Vol 1: From Colonial Times to the Founding of the American Federation of Labor (New York: International Publishers, 1947)Google Scholar; Perlman, Selig and Taft, Philip, History of Labor in the United States, 1896–1932 (New York: Macmillan, 1955)Google Scholar.

13. For other influential scholarship on the consequences of such organizations, see Jacoby, Sanford, “American Exceptionalism Revised: The Importance of Management,” in Masters to Managers: Historical and Comparative Perspectives on American Employers, ed. Jacoby, Sanford (New York: Columbia University Press, 1991), 173–200Google Scholar; Friedman, Gerald, State-Making and Labor Movements (Ithaca, NY: Cornell University Press, 1998)Google Scholar; Tolliday and Zeitlin, “Employers and Industrial Relations between Theory and History”; Zeitlin, Jonathan, “The Triumph of Adversarial Bargaining: Industrial Relations in British Engineering, 1889–1939,” Politics & Society 18, no. 3 (1990): 405–26Google Scholar. This literature on the importance of different institutions and instruments of repression, including the limited research on collective employer repression, is part of a larger canonical historiography that explains why the United States lacked a relatively strong labor movement. See Sombart, Werner, Why Is There No Socialism in the United States? (Armonk, NY: M. E. Sharpe, 1976)Google Scholar; Archer, R., Why Is There No Labor Party in the United States? (Princeton, NJ: Princeton University Press, 2007)Google Scholar. An abbreviated list of reasons include political culture, ethnic divisions and immigration, relative prosperity, timing of enfranchisement, high religiosity, craft versus industrial sector divisions within the working force, geographical size inhibiting collective action, and size of labor force. Addressing the relative role of employer repression to these other explanations is beyond the scope of this article, though many scholars argue that such instruments and institutions of repression undoubtedly played a strong role in the stunting of U.S. labor mobilization.

14. Forbath and Orren focus on the hostile legal environment. Forbath, William E., Law and the Shaping of the American Labor Movement (Cambridge, MA: Harvard University Press, 1991)CrossRefGoogle Scholar; Ernst, Daniel R., Lawyers against Labor: From Individual Rights to Corporate Liberalism (Champaign: University of Illinois Press, 1995)Google Scholar; Orren, Karen, Belated Feudalism: Labor, the Law, and Liberal Development in the United States (New York: Cambridge University Press, 1992)Google Scholar. On the role of state and federal government and their role in the high levels of violence resulting from labor repression, see the classic accounts by Taft, Philip and Ross, Philip, “American Labor Violence: Its Causes, Character, and Outcome,” in The History of Violence in America: Historical and Comparative Perspectives, ed. Davis, Hugh Graham and Ted Robert Gurr (New York: Praeger, 1969), 281–396Google Scholar; Gitelman, H. M., “Perspectives on American Industrial Violence,” The Business History Review 47, no. 1 (1973): 1–23Google Scholar; Adams, Graham Jr., Age of Industrial Violence 1910–15: The Activities and Findings of the United States Commission on Industrial Relations (New York: Columbia University Press, 1966)Google Scholar. An enormous literature links such repression to long term union inhibition. For a sample of the most important recent works, see Archer, Why Is There No Labor Party in the United States?; Mann, Michael, The Sources of Political Power (Cambridge: Cambridge University Press, 1993)Google Scholar; Friedman, State-Making and Labor Movements; Jacoby, “American Exceptionalism Revised”; Montgomery, David, The Fall of the House of Labor: The Workplace, the State, and American Labor Activism, 1865–1925 (New York: Cambridge University Press, 1989)Google Scholar.

15. See the following classic works: Montgomery, The Fall of the House of Labor; Perlman and Taft, History of Labor in the United States, 1896–1932. Letwin describes the judicial repression unions faced in William Letwin, Law and Economic Policy in America (New York: Random House, 1965)Google Scholar.

16. On the actions of major firms such as U.S. Steel, many accounts discuss their role in victories in major strikes and the resulting setbacks for unions in the late nineteenth and early twentieth century. See Brody, David, Labor in Crisis: The Steel Strike of 1919 (Champaign: University of Illinois Press, 1987)Google Scholar; Bemis, Edward W., “The Homestead Strike,” The Journal of Political Economy 2, no. 3 (1894): 369–96Google Scholar; Krause, Paul, The Battle for Homestead, 1880–1892: Politics, Culture, and Steel (Pittsburgh, PA: University of Pittsburgh Press, 1992)CrossRefGoogle Scholar. The role of such firms in strikebreaking and strike prevention was quite large in the late nineteenth century; one study estimates that between 1881 and 1900, firms hired more than 500,000 replacement workers of the 4.8 million who went on strike; strikebreakers were used in 40 percent of strikes in this period and substantively increased firms’ chances of defeating strikes. Rosenbloom, Joshua L., “Strikebreaking and the Labor Markets in the United States, 1881–1894,” Journal of Economic History 58, no. 1 (1998): 183–205Google Scholar. During this period with few sector-level agreements between firms and workers, unions struck to attain bargaining rights, and both large and small firms responded with repressive strategies. These strategies of firms and political actors could also act in tandem. For example, the judicial apparatus and contested legal status of unions after 1890 also greatly helped firms inhibit unionization. For an overview of this process, see Bok, Derek C., “Reflections on the Distinctive Character of American Labor Laws,” Harvard Law Review 84, no. 6 (April 1971): 1394–463Google Scholar. Following the passage of the Sherman Antitrust Act in 1890, a long legal battle persisted in state and federal courts between firms and unions on whether unions themselves were illegal agreements, and whether strikes were legal; firms could legally (and of course did) file injunctions against unions in anticipation of a strike, resulting in costly battles and delays for unions. This logic is explicated with many examples in Fitch and Lewin. Fitch, John A., The Causes of Industrial Unrest (New York: Harper & Brothers, 1924)Google Scholar; Letwin, Law and Economic Policy in America.

17. Martin discusses the turn of the NAM towards a more neo-liberal ideology but does not explicitly address the rise of other antilabor movements or repression as a strategy before this transition. Her excellent account focuses on the link between the Republican Party and the NAM at the national level. Martin, “Sectional Parties, Divided Business.” Martin and Swank discuss the rise of the open shop movement at the local level outside of the industrial northeast, but do not focus on the relevant local-level variation in the movement. Martin and Swank, The Political Construction of Corporate Interests, 106.

18. Bernstein, Irving, The Lean Years: A History of the American Worker (Boston: Houghton Mifflin, 1960)Google Scholar.

19. Bernstein notes that this period saw the advent company-level unions, to blunt the organizational impact and appeal of other unions.

20. The role of antilabor firm activity at the local level is particularly relevant for firms that could not for example easily relocate, particularly for small and medium-sized enterprises. Gordon analyzes the growth of cities as a function of capitalist development and accumulation of capital power, but does not discuss the role of organization among such firms. Gordon, David, “Capitalist Development and the History of American Cities,” Marxism and the Metropolis (1978): 25–63Google Scholar.

21. For a different application of how the federated structure of interest groups can play developmentally into the politics of state formation, see Skocpol, Theda, Protecting Soldiers and Mothers (Cambridge, MA: Harvard University Press, 1995)Google Scholar.

22. Gordon provides a theoretical account of how capitalist development and urban growth are intrinsically linked, but does not discuss the specific role of firm repression of workers or of collective repression. Gordon, “Capitalist Development and the History of American Cities.”

23. A key theoretical monograph on the collective action of firms is by Schmitter and Streeck, which makes similar assumptions about employer associations as those works in the previous note (largely ignoring the activities of historical repressive associations). I draw on insights from this work later in the theoretical discussion. Philippe C. Schmitter and Wolfgang Streeck, “The Organization of Business Interests: Studying the Associative Action of Business in Advanced Industrial Societies” (Max-Planck-Institute für Gesellschaftsforschung [MPIfG] Discussion Paper, Cologne, Germany, 1991).

24. Snyder documents the relevance of exploiting subnational variation within a polity to test general theories, as many national-level factors can be held fixed. Snyder, Richard, “Scaling Down: The Subnational Comparative Method,” Studies in Comparative International Development 36, no. 1 (2001): 93–110Google Scholar.

25. See the accounts provided by Hurvitz and Wakstein: Wakstein., Allen M. “The Origins of the Open-Shop Movement, 1919–1920,” Journal of American History 51, no. 3 (1964): 460–75Google Scholar; Wakstein, Allen M., The Open-Shop Movement, 1919–1933 (Champaign: University of Illinois Press, 1961)Google Scholar. A more recent discussion of select open shop movements is found in Chad Pearson, “‘Organize and Fight’: Communities, Employers, and Open-Shop Movements, 1890–1920” (PhD diss., University at Albany, State University of New York, 2008. ProQuest 3327492). Wakstein, Allen M., “The National Association of Manufacturers and Labor Relations in the 1920s,” Labor History 10, no. 2 (1969): 163–76Google Scholar. For discussion of the industrial conference, see Hurvitz, Haggai, “Ideology and Industrial Conflict: President Wilson's First Industrial Conference of October 1919,” Labor History 18, no. 4 (1977): 509–24Google Scholar. Hurvitz's evidence on the positions of different representatives comes from the Proceedings of the First Industrial Conference, October 6 to 23 1919 (Washington, DC: Government Printing Office, 1920)Google Scholar.

26. Representatives from firms included Bethlehem Steel, DuPont, General Motors, International Harvester, General Electric, and Standard Oil of New Jersey. See Hurvitz, “Ideology and Industrial Conflict.”

27. Hurvitz, “Ideology and Industrial Conflict,” 519.

28. Ibid., 520. The original source of the quotation is in the Proceedings of the First Industrial Conference, October 6 to 23 1919, 196.

29. Hurvitz, “Ideology and Industrial Conflict,” 520. There were other third-party representatives appointed by President Wilson, but the main issues of contention were discussed by the organizations representing unions and employers. Firm representatives also advocated another strategy to undermine unions, by using “shop councils” which would be forums for workers to air grievances, but they would not be able to call strikes or make decisions without managerial consent. Employer associations also argued that firms could solve labor problems locally; they opposed the AFL's vision of large scale labor unions bargaining on an equal footing with employer associations. Firms also expressed anti-union preferences through their role in the National Civic Federation, which had minimal impact on actual employer association policy. Green, Marguerite, The National Civic Federation and the American Labor Movement (Washington, DC: The Catholic University of America Press, 1956)Google Scholar.

30. Dunn, The Americanization of Labor, 22.

31. Foner, History of the Labor Movement in the United States, vol. 8.

32. Pearson, Reform or Repression.

33. “Wages Working to Normal Basis,” Iron Trades Review (January 6, 1921): 7–10Google Scholar.

34. Dunn, The Americanization of Labor, 22.

35. Foner, History of the Labor Movement in the United States, vol. 8, p. 170.

36. Swenson discusses the logic and use of these lockouts, and mixed results in collective action success in the United States. Swenson, Capitalists against Markets. Swenson also discusses the conflicts between employers over union recognition, such as between manufacturing and building trades, although there were many instances of shared anti-union sentiment across industries, within OSAs. Swenson, Capitalists against Markets, 175–77. Harris describes the lockout use in Philadelphia. See Harris, Bloodless Victories; Harris, “Getting It Together,” 111–32. The Senate committee hearings chaired by Senator Robert La Follette provide much detail about the use of lockouts, labor spies, and blacklists. See U.S. Congress, Senate, Subcommittee of the Committee on Education and Labor, Hearings Pursuant to S. Res. 266, Violations of Free Speech and Rights of Labor, 74th–76th Cong. (1936–1940). See also Weiss, Robert, “The Emergence and Transformation of Private Detective Industrial Policing in the United States, 1850–1940,” Crime and Social Justice, no. 9 (1978): 35–48Google Scholar.

37. Bernstein quotes the historian Bonnett's summary of their actions, as “the belligerent associations may fight the union in actual battles with machine guns; it may oppose the union in legislative and political matters; it may combat the union strikes; it may carry on a continual propaganda against the union in every particular or only against certain practices….” Bernstein, The Lean Years, 153–54.

38. Wakstein, “The Origins of the Open-Shop Movement, 1919–1920”; Brenner et al., The Encyclopedia of Strikes.

39. Wakstein, “The Origins of the Open-Shop Movement, 1919–1920,” 462.

40. Ibid., 463–64. The San Antonio association was incorporated by the state in June 1919, and represented all interests except for retail merchants. W. S. Mosher, “Open Shop in the Southwest,” The Open Shop Review (1921): 116. See also Foner, History of the Labor Movement in the United States, vol. 8, p. 287.

41. Bernstein, The Lean Years, 154.

42. Bernstein also describes employer association and NMTA victories against IAM locals in Cincinnati during this period (Bernstein, The Lean Years, 165).

43. Wakstein, The Open-Shop Movement, 1919–1933; “Public Sponsors Open Shop Associations,” Iron Trades Review (November 11, 1920): 1339–48; “Declare for Open Shop in 78 Cities,” Iron Trades Review (August 12, 1920): 440. The NMTA, NEA, and NAM all supported the open shop. Associations varied in whether they were officially sponsored by the state; for example, the OSA of Jefferson County was chartered by state.

44. Wakstein, The Open-Shop Movement, 1919–1933.

45. Only after the reaction at the local level did the NAM take on a leadership role. Elbert Gary of U.S. Steel supported the open shop after the 1919 steel strike. Wakstein and Foner both note some preliminary patterns in the movement's initial diffusion, but the data presented are not systematic, nor do they address the question of why some localities had such movements. Wakstein, “The Origins of the Open-Shop Movement, 1919–1920.”

46. Using county-level OSAs as a proxy of the degree of repressive employer coordination likely underestimates such coordination, because there is evidence that such employer associations formed in counties but were not reached by the AEI survey discussed below. This type of local-level employer coordination differs from other important trade associations that formed in the early twentieth century to share information and ultimately innovate in areas such as cost accounting. Berk and Schneiberg in a number of innovative articles examine the development of such associations, particularly in the printers’ industry. They discuss the importance of these firms learning from previous mistakes and theorize them as alternative forms of employer coordination. In this study, however, I focus on the determinants of one specific, albeit consequential, type of anti-union association that adopted the above repressive strategies; the question is what would influence firms to decide to form an association that advocated the open shop and employ strategies to maintain it. Berk, Gerald and Schneiberg, Marc, “Varieties in Capitalism, Varieties of Association: Collaborative Learning in American Industry, 1900 to 1925,” Politics & Society 33, no. 1 (2005): 46–87Google Scholar; Berk, Gerald and Schneiberg, Marc, “From Categorical Imperative to Learning by Categories: Cost Accounting and New Categorical Practices in American Manufacturing, 1900–1930,” Research in the Sociology of Organizations 31 (2010): 255–92Google Scholar.

47. Because the question of interest is in local-level variation, theorizing is partly constrained by what characteristics should be relevant at the local or country level.

48. Schmitter and Streeck's influential essay on business associations posits numerous “logics of membership” that detail different considerations for the formation and influence of sector-level considerations; some of the theoretical variables discussed in this section follow from their architecture, although they do not focus as closely on explaining cross-sectional variation. That monograph presents a complicated theoretical architecture but is oriented towards contemporary economies and does not focus on the particular issue of repressive associations, nor does it present specific historical evidence for the many theoretical claims.

49. These families are not mutually exclusive, of course, particularly as the political context may shape firm incentives by posing a distinct redistributive threat to firms via government intervention. There are other variables that cannot be easily tested using locality-level data, such as theories related to the importance of external interlocutors and internal organization issues. Future work should more carefully distinguishing among organizational issues, external interlocutors, and political contexts; doing so is difficult with locality-level data.

50. The hypotheses that follow may not apply to other types of associations, as discussed for example in Berk and Schneiberg. Berk and Schneiberg, “Varieties in Capitalism, Varieties of Association.”

51. Conell and Voss discuss how previous organizational experiences of Knights of Labor locals affected subsequent types of organization; similar hypotheses have not been tested for employer associations. Conell, Carol and Voss, Kim, “Formal Organization and the Fate of Social Movements: Craft Association and Class Alliance in the Knights of Labor,” American Sociological Review 55, no. 2 (1990): 255–69CrossRefGoogle Scholar.

52. Schmitter and Streeck, “The Organization of Business Interests”; Olson, Mancur, The Logic of Collective Action: Public Goods and the Theory of Groups (Cambridge, MA: Harvard University Press, 1968)Google Scholar. Note that this hypothesis ignores the relative size of firms, as larger firms could bear additional costs provide public goods of associations.

53. This is also emphasized in Schmitter and Streeck's consideration of the logic of membership, although they do not frame this in the context of historical coordination.

54. Within individual sectors, Derber documents the importance of homogeneity across firms. Derber, Milton, “Employers Associations in the United States,” in Employers Associations and Industrial Relations, ed. Windmuller, John P. and Gladstone, Alan (Oxford: Clarendon Press, 1984), 79–114Google Scholar. The evidence presented by Schneiberg and Berk, although focusing on a different kind of business association, also points to the importance of similarity in product and labor markets across firms as facilitating the formation of trade associations. Berk and Schneiberg, “Varieties in Capitalism, Varieties of Association.” Schmitter and Streeck also pose the theoretical possibility that “too much” homogeneity across firms can also inhibit collective action, as that would be linked to increased intensity in price competition, reducing incentives to cooperate. This observation was not linked to the probability of observing such cooperation within a given locality, however. Schmitter and Streeck, “The Organization of Business Interests.” See also McCaffree for discussion of this point: Kenneth M. McCaffree, “A Theory of the Origin and Development of Employer Associations” (paper presented at the Proceedings of the 15th Annual Meeting of the Industrial Relations Research Association Pittsburgh, PA, December 27–28 1962).

55. Another possibility is that unions that have more “radical” demands could be viewed to be more of a threat to firms, and thus more likely to induce cooperation among firms. They are only more extreme in the sense that they supported greater redistribution for workers and plausibly posed more of a threat to firms. I discuss this in the evidence in the notes below.

56. Major sector-level associations such as the NEA supported labor policies in the Republican Party platform beginning in 1908. As Martin notes the NAM became closely aligned with the Republican Party and remained so after its turn towards hostile policies against organized labor. Martin, “Sectional Parties, Divided Business.” Such associations opposed legislation on the eight-hour day, opposed legalization of boycotts, supported legislation that would allow firms to pursue injunctions against unions in anticipation of strikes, and did not want unions exempted from the Sherman Antitrust Act, all goals more supported by the Republican Party at the time. See documentation from the NEA and local-level associations from the Drew Collection at Bentley Library, roll 20. The NEA also opposed the Clayton Act and related perceived antibusiness legislation, mentioning Democratic support for it and related legislation. Walter Drew, “Letter to Members of the National Erectors’ Association,” 1912, pp. 1–3, Drew Collection, roll 20, Bentley Library, University of Michigan, Ann Arbor. While there was local-level variation in employer associations’ explicit opposition to state-level Democratic candidates, such associations generally opposed legislation advocated by the party during this time period and were of course hostile to union-endorsed candidates. For examples of both, see electoral discussions by the NMTA. Fred W. Job, “The Mythical Labor Vote,” The Bulletin of the National Metal Trades Association (1904); and discussion of legislation by the AEI. Andrew J. Allen, “Minutes of Board Meeting: Associated Employers of Indianapolis, Inc.” (Indiana Historical Society, 1930), 150–53.

57. Of course, there was much variation in the relationship between the Democratic Party and local-level union positions, even though after 1908 the Democratic Party platform became much more pro-union. Greene describes the different goals, practices, and ideologies of local-level affiliates of the AFL and the national office; her account shows how the AFL national office slowly came to support the Democratic Party after 1908, and how such participation in in national-level politics and occasional support for Democratic candidates alienated or divided many local-level unions because some unions allied with Socialist or other more left-wing political parties. Greene, Julie, Pure and Simple Politics: The American Federation of Labor and Political Activism, 1881–1917 (Cambridge: Cambridge University Press, 1998)Google Scholar.

58. Korpi, W., The Working Class in Welfare Capitalism: Work, Unions and Politics in Sweden, International Library Society Series (Boston: Routledge & Kegan Paul, 1978)Google Scholar.

59. Greene, Pure and Simple Politics.

60. A reasonable alternative hypothesis is that a less left-wing electoral climate could mean that the local population or their elected officials would be willing to support actions by employer organizations (such as lockouts or importation of substitute workers during a labor conflict). Therefore, left-wing electoral support could be negatively correlated with more employer organizations. Political support for employer organizations could take the form of pro-firm or anti-union legislation, or executive intervention in industrial disputes. This support for employer organizations could take place an electoral environment that is more hostile to the goals of organized labor. The empirical test below helps arbitrate between these competing hypotheses.

61. The AEI's survey data are discussed in several editions of the Iron Trades Review. The existence of an OSA in a county does not indicate that the city government or the electorate chose to have an OSA; it only indicates the collective decision of firms within a given locality to form an OSA.

62. This results in a reduced number of observations because some counties have more than one city which had an active OSA). Unfortunately the sampling methodology used by the AEI to form their list of active OSAs is not available.

63. The data are from all the associations described in Bonnett, Clarence E., History of Employers’ Associations in the United States (New York: Vantage Press, 1956)Google Scholar. To my knowledge, these data have not been used in this type of analysis before.

64. Unless otherwise noted, all county-level economic or structural variables come from U.S. Census Data. See Michael R. Haines and the Inter-University Consortium for Political and Social Research, “Historical, Demography, Economic, and Bibliographic Citation: Social Data: The United States, 1790–2000” (data set; Ann Arbor: MI, Inter-University Consortium for Political and Social Research, 2004).

65. Miriam King, Steven Ruggles, J. Trent Alexander, Sarah Flood, Katie Genadek, Matthew B. Schroeder, Brandon Trampe, and Rebecca Vick, “Integrated Public Use Microdata Series, Current Population Survey: Version 3.0. [Machine-Readable Database],” ed. Minnesota Population Center (Minneapolis, MN: Minnesota Population Center, 2010). The results below are similar regardless of the construction of the concentration measure.

66. Garlock, Jonathan, Knights of Labor Assemblies, 1879–1889 (Ann Arbor, MI: Inter-University Consortium for Political and Social Research, 2009)Google Scholar.

67. The industries or crafts where union locals were coded are the following: Boiler Makers, Boot and Shoe makers, Bricklayers, Carpenters and Joiners, Cigar makers, Coopers, Flint glass workers, Gold beaters, Hod Carriers, Machinists (IAM) , Electrical Workers (IBEW), Ladies Garment Workers (ILGWU), Molders (IMU), Typographers (ITU), Leather workers, Metal polishers, Painters, Postal workers, Pattern makers, Sheet metal workers, Stove mounters, Amalgamated Wood Workers. The states for which county-level data on union local presence were available are Alabama, Arkansas, Florida, Georgia, Illinois, Kentucky, Louisiana, Michigan, Minnesota, Mississippi, Missouri, North Carolina, South Carolina, Tennessee, Texas, Virginia, and Wisconsin. These data were shared by Gerald Friedman.

68. Unfortunately, annual data do not exist. Nor do data on union density by county level exist for this period. But the data at the time-period level capture much of the threat posed by unions to firms by documenting the number of locals in different industries.

69. Similarly, if a county had one union local in the same industry across all three time periods, its value on the Union local variable would be 3.

70. Note that the effect of this measure could still be imprecisely estimated beyond a certain range because not many counties have large numbers of different industries represented, as most counties have a few industries represented and almost one union local per industry.

71. I use these data on conflict due to its geographic precision and many characteristics of strikes noted. Report of the Secretary of the Interior Being Part of the Message and Documents Communicated to the Two Houses of Congress at the Beginning of the First Session of the Fiftieth Congress (Washington, DC: Government Printing Office, 1887)Google Scholar; Commissioner of Labor, ed., Tenth Annual Report of the Commissioner of Labor: Strikes and Lockouts, Vol. 1 (Washington, DC: Government Printing Office, 1896Google Scholar.

72. The union locals with greater voting weight in the AFL and in the data set are Boiler Makers (2.7%), Carpenters and Joiners (8%), Machinists (IAM) (8.7%), Electrical Workers (IBEW) (3.6%), Ladies Garment Workers (ILGWU) (2.7%), and painters (2.7%). Results do not change if a lower threshold of votes is chosen.

73. Standard errors are clustered by state. One methodological concern could be that given that fewer data are available regarding county-level union presence, there could be selection bias when estimating the effect of union presence on OSA adoption. It could be the case that for the counties for which data on union presence were available are also counties in which OSAs are adopted, but it may not be the case that the union presence has an independent causal effect on OSA adoption. To address the possible selection issues, I also estimate a Heckman probit model with sample selection. This is the equivalent of a Heckman selection model, except the model consists of a probit model in the selection equation, and a probit model in the outcome equation. In this model, the selection equation has the binary dependent variable indicating if data were available on union presence or not for a specific county; the main dependent variable is still OSA presence in a county. The intuition is that some of the economic independent variables might explain why union data are more available in certain counties. In the Heckman probit model, the statistical significance of the parameter estimate of ρ is the correlation of errors between the two outcomes. If ρ is zero, then the unobserved variables that affect whether union data are available for a county are independent of the unobserved variables which might affect adoption of an OSA. In all specifications of the selection model, ρ was statistically insignificant indistinguishable from zero, and the coefficient on both the union threat variables was the same. I therefore report results of the standard probit model. Interested readers should consult Dubin and Rivers for details on the model. Dubin, Jeffrey A. and Rivers, Douglas, “Selection Bias in Linear Regression, Logit and Probit Models,” Sociological Methods & Research 18, no. 2–3 (1989): 360–90Google Scholar. See Berinsky for an application of the model: Berinsky, Adam J., “The Two Faces of Public Opinion,” American Journal of Political Science (1999): 1209–30Google Scholar. Note 73 also discusses a simple method of dealing with missing data on the independent variables without listwise deletion of data, which the following analyses conduct.

74. Overall, about half of all counties, 52 percent, had at least one KL assembly, and the average number of KL locals was 3.2.

75. 24 percent of counties experienced at least one strike in the lagged time period.

76. Due to missing data for some of the independent variables, I present models that include binary indicators of missing data and recode counties missing on each of the control variables as 0 on those variables. This prevents listwise deletion of data, while allowing for an intercept shift for counties with missing data. Omission of counties with missing data does not substantively change the coefficients. The results are substantively similar using probit estimations. In all the tables that follow, all models control for both missing data binary indicators. All models also control for a dummy indicating county presence in a state in the geographical South (Alabama, Arkansas, Florida, Georgia, Louisiana, Mississippi, North Carolina, South Carolina, Tennessee, Texas, and Virginia); the binary variable coefficient is not precisely estimated. The results are the same if the estimations control for each economic indicator separately.

77. This noneffect of the number of firms could have several interpretations. A higher number of firms in a county would make it more difficult for a collective employer association to form. But it is possible that such a relationship could be nonlinear: a small number of firms may decide they are capable of repressing workers without the need for a formal OSA, and thus not form an OSA. Conversely, if there are a large number of large firms, they may be better able to overcome collective action problems. The number of firms does not capture the relative sizes of firms, which would be a preferable but unavailable county-level proxy for difficulty of collective action as a predictor for formation of an employer organization.

78. The estimated marginal effect of a new industry in a county is greater than the estimated impact of a new union local. Each additional local raises the probability of OSA adoption by about 1 percent, whereas each additional industry does so by nearly 1.5 percent.

79. However, the marginal impact of this variable is much smaller; complete previous success in strikes only raises the probability of OSA adoption to .02, though this is double the probability of a county with no successful strikes.

80. As an alternative consideration of union threat, I also coded sums of unions differentiating between more versus less extremist unions. Because systematic data on county-level union variation in union-local characteristics are not available, I calculate a rough proxy based on the union's voting positions in important industrial-relations resolutions raised at the annual AFL conventions. I use the resolutions highlighted by Marks as being particularly important to the AFL. Marks, Gary, Unions in Politics: Britain, Germany, and the United States in the Nineteenth and Early Twentieth Centuries (Princeton, NJ: Princeton University Press, 1989)CrossRefGoogle Scholar.These propositions were whether the AFL should: engage in more political activity (1894); organize to attain more political and economic power (1902); support legislation on the eight-hour working day (1915); be opposed to the National Civic Federation and labor's connection with it (1911). I code unions that had at least 50 percent of their delegates at the AFL convention vote for at least one of these measures and did not vote against any of the measures, as being “more extremist” unions. They are only more extreme in the sense that they supported greater redistribution for workers and plausibly posed more of a threat to firms. I code unions that had at least 50 percent of their delegates at the AFL convention vote against one of the measures as “more moderate” unions. Seven unions had recorded votes on these issues that fall into neither category; they were not coded as more extreme or moderate. Of course extremism is relative, as these measures were far less extremist for example than labor parties’ resolutions to engage in mass strikes, for example. Marks labels some of these resolutions as supporting “more radical” behavior. The set of more extremist unions include: goldbeaters, pattern makers, carpenters, brewers, garment workers, sheet metal workers, bricklayers, leather makers, international molders union, metal workers, and stove mounters. The set of more moderate unions include: boot makers; machinists; cigar makers; international typographers union. The following unions were classified as neither, either because of lack of data on voting or because of a mixed voting record on the four resolutions: boilermakers; coopers; flint glass workers; gold beaters; hod carriers; postal workers; amalgamated wood workers. Of course extremism is relative, as these measures were far less extremist for example than labor parties.’ The results show that the aggregate threat posed by more extreme unions is positively correlated with OSA adoption, versus almost no effect for the threat posed by more moderate unions, regardless if we measure the aggregate union threat by number of locals or number of industries. The estimated marginal effect for an additional “extreme” union is nearly 2 percentage points, though the marginal effect for a similar union local is small (less than 1 percentage point, though still larger than zero). This is suggestive evidence that a type of union—a union whose delegates are more likely to support pro-worker redistribution measures—is correlated with an increased probability of OSA adoption. These effects remain positive but become less precise once we consider the marginal impact of more than four unions. These results are available upon request.

81. We do not observe the same results when we consider number of union locals that are representing stronger versus weaker industries. This might be because there are fewer “stronger” unions in this data set, as localities with union locals that have lower voting weight within the AFL in fact have more such locals; this is indicated by the fact that overall union local presence is correlated with the total sum of locals representing “weaker” unions. Another possibility is that the marginal impact of a new industry represented by a stronger union matters more, not just the presence of an additional local in such an industry.

82. Similar marginal effects are estimated if we examine the impact of more recent union industries versus the threat of industries in the previous period.

83. I also test hypotheses about the impact of changes in aggregate union threat, measured by both local and industrial sums, between the period 1890–1910 and the 1910–1914. I code a variable that captures the change in number of union locals between the first two time periods and third time period (between 1890–1910 and 1910–1914). For each industry for which there are data in these two time periods, I calculate the difference in number of union locals (most counties either experienced zero growth or some positive growth in the number of new locals). The sum of the growth in union locals across all industries represents the change in aggregate union threat. Using this change in union threat variable, I construct simple binary variable that indicates whether the county experienced any increase in union growth between the period 1890–1910 and 1910–1914. The variable is 0 if there was no growth and 1 if there was at least one more local in the sum. Neither the continuous nor binary indicators of union growth are correlated with OSA adoption. Even if counties experienced a positive change in union presence between these two periods, the cross-sectional differences in amount of union locals or different industries matters more for adoption of an OSA.

84. See Pearson, “‘Organize and Fight.’” I also estimate all models in Table 1 with the economic variables normalized by number of workers in the county (as opposed to per capita); the coefficients on both the union variables do not substantively change and remain statistically significant. The inclusion of other economic controls—percent of owned homes, amount of manufacturing value added—does not affect the results.

85. These results also challenge the alternative hypothesis that an increased number of industries represented by union locals does not reduce the threat because of difficulty in cooperation among unions. One might expect that this increased number would lead to higher fractionalization of the labor force and thus a lower threat from workers, but this positive correlation does not support that view.

86. While the specific findings about the role of union threats are limited to a subset of U.S. states, these states (covering the Midwest, for example), were industrially important and were home to many of the important battlegrounds of organized labor.

87. Voss, The Making of American Exceptionalism.