econ his 1 hour

Economic History Association

The Law and Labor Strife in the United States, 1881-1894 Author(s): Janet Currie and Joseph Ferrie Source: The Journal of Economic History, Vol. 60, No. 1 (Mar., 2000), pp. 42-66 Published by: Cambridge University Press on behalf of the Economic History Association Stable URL: http://www.jstor.org/stable/2566796 Accessed: 02-11-2016 23:02 UTC

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted

digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about

JSTOR, please contact [email protected].

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at

http://about.jstor.org/terms

Economic History Association, Cambridge University Press are collaborating with JSTOR to digitize, preserve and extend access to The Journal of Economic History

This content downloaded from 139.182.97.89 on Wed, 02 Nov 2016 23:02:52 UTC All use subject to http://about.jstor.org/terms

The Law and Labor Strife in the United States, 1881-1894

JANET CURRIE AND JOSEPH FERRIE

This article examines the effects of state-level legal innovations governing labor disputes in the late 1 800s. This was a period of legal ferment in which worker orga- nizations and employers actively lobbied state governments for changes in the rules governing labordisputes. Cross-state heterogeneity inthe legal environment provides an unusual opportunity to investigate the effects of these laws. We use a unique data set with information on 12,965 strikes to show that most of these law changes had surprisingly little effect on strike incidence or outcomes. Important exceptions were maximum hours laws and the use of injunctions.

Evaluating the role of the legal envlronment in shaping economic outcomes is often difficult, because of either the lack of microlevel data

or the absence of sufficient heterogeneity in laws across jurisdictions. These problems are most severe in recent data, because confidentiality rules limit the availability of data and the imposition of federal law standardizes the legal environment across states. Studies using historical data have overcome these difficulties in examinations of the impact of maximum hours laws, compulsory school attendance laws, banking regulations, mortgage fore- closure moratoria, and regulation of natural resources.'

Surprisingly, the same historical approach has not been applied to one of the most contentious areas of legislative intervention in the economy: the regulation of disputes between workers and employers. In this article, we

7he Journal of Economic History, Vol. 60, no. 1 (March 2000). ? The Economic History Association. All rights reserved. ISSN 0022-0507.

Janet Currie is Professor, Department of Economics, University of California, 405 Hilgard Avenue, LosAngeles, CA 90024, telephone: (310) 206-8380,e-mail: [email protected]; andResearch Associate, NBER. Joseph Ferrie is Associate Professor, Department of Economics, Northwestem University, 2003 Sheridan Road, Evanston, IL 60208-2600, telephone (847) 491-8210, e-mail [email protected]; and Research Associate, NBER.

We thank David Card for generously providing some of the data and Ludwig Chincarini and Chris Kim for excellent research assistance. Armen Alchian, Richard Freeman, Sandy Jacoby, Bentley MacLeod, Robert Margo, Joseph Tracy, and participants at the NBER Development of the American Economy Program Meeting in March 1994 provided excellent comments. Janet Currie thanks the Alfred P. Sloan Foundation for fimancial support. Joseph Ferrie thanks the Olin Foundation and the Institute for Policy Research at Northwestern University for financial support. All opinions are those of the authors and do not necessarily reflect the views of the Sloan Foundation, the Olin Foundation, or the Institute for Policy Research.

On maximum hours laws, see Landes, "Effect"; Goldin, "Maximum Hours Legislation"; and Atack and Bateman, "Whom Did Protective Legislation." On compulsoxy school attendance laws, see Angrist and Krueger, "Does Compulsory Schooling"; and Margo and Finegan, "Compulsory Schooling Legislation." On banking regulations, see Rockoff, "Free Banking Era." On mortgage foreclosure moratoria, see Alston, "Farm Foreclosures." On regulation of natural resources, see Libecap, "Economic Variables" and "Bureaucratic Opposition."

42

This content downloaded from 139.182.97.89 on Wed, 02 Nov 2016 23:02:52 UTC All use subject to http://about.jstor.org/terms

The Law and Labor Strife 43

explore the relationships among laws goverming collective bargaining disputes, strike incidence, and strike outcomes. For this purpose, we have assembled information about the outcomes and characteristics of 12,965 labor disputes that took place in the United States between 1881 and 1894, and about the legal environment in each state and year.

This was a period of legal ferment in which worker organizations and employers actively lobbied state governments to have the rules of the bargaining game changed in their favor. Indeed, legal developments sometimes followed pitched battles involving state and federal troops, loss of life, and property damage. Organized labor lobbied for the legalization of unions, abolition ofthe blacklist, and passage of maximum hours laws, while employers applauded the use of the injunction against striking workers and the passage of laws restricting the use of intimidation and boycotts.

We do not aim to test a particular theory of strikes, but we do find strike theory to be useful in interpreting our results. In particular, strike theory suggests that we should not necessarily expect collective bargaining laws favored by labor to have prolabor effects on strike outcomes or those favored by employers to have pro-employer effects. Instead, we need to look at how each law is likely to have affected the cost of any given strike or uncertainty about strike outcomes. In terms of strike theory, our results suggest that maximum hours laws reduced the uncertainty surrounding possible strike outcomes, whereas the use of the injunction increased it.

LABOR LAW IN THE LATE NINETEENTH CENTURY AND ITS LIKELY EFFECTS ON STRIKE ACTIVITY

A nascent trade union movement had appeared in the United States as early as the 1 830s, embracing roughly 26,000 workers.2 But these organiza- tions were locally isolated and focused mainly on craftsmen-proprietors: as John Commons et al. note, "it was only during the sixties that labor organ- izations began to think and act on a lasting national basis."3 By the middle of the 1880s, more than 700,000 workers, both skilled and unskilled, were members of national labor organizations, with most ofthis growth occurring over just the previous ten years.4 With the appearance of large organizations of workers, state legislatures and courts were increasingly called upon to intervene in labor disputes and to develop new legal means to do so.5

2Lebergott, "American Labor Force," p. 220. 3 Commons et al., History (1918), p. 43. 4 Lebergott, "American Labor Force."

5 Thoughthe federal ShermanAnti-TrustAct of 1890 was later an important antiunion and antistrike tool, the federal government played only a small role in shaping labor law during the period that we examine. The ShermanActwas first employed against labor in 1894 in theprosecutionofEugene Debs in the Pullman case, but the case was eventually decided on other grounds (In re Debs, 158 U.S. 564, 15 Sup. Ct. 900).

This content downloaded from 139.182.97.89 on Wed, 02 Nov 2016 23:02:52 UTC All use subject to http://about.jstor.org/terms

44 Currie and Ferrie

The first aspect of the employment relation addressed was the length of the workday. States passed legislation to regulate the hours of women and children throughout the nineteenth century, but they were more cautious in regulating the hours of men.6 No argument could be made for regulating the hours of men on purely social grounds as had been made for women and children. And courts had held that the freedom of contract prevented states from dictating the maximum number of hours employees could work each

week.7 Three strategies emerged in response to these constraints. The first was regulation of employees' hours when the state was a direct

party to the labor contract, as when states employed workers themselves or hired contractors who in turn hired workers. New York passed such legislation in 1870.8 The second was regulation of specific industries, particularly those in which worker fatigue could result in injury to the workers themselves or to others because the work involved heavy machin- ery. New York's maximum hours legislation for railroad workers in 1888 and 1892 fits into this category.9 The third approach was prescribing a maximum number of hours that employees could be forced to work "in the absence of any agreement to the contrary." Since most labor contracts specified both a wage and a work schedule, this last provision was relatively toothless, but may have signaled at least some concem on the part of the state for workers' interests. Five of the states we examine below had at least

one of these types of legislation on the books before 1880 (Connecticut, Illinois, Maine, New Hampshire, and New York), and all of the states we examine except Delaware had such laws by 1890.

The explicit legalization of unions by state legislatures came relatively late in the nineteenth century. Throughout much of the period, the court's approach to organized labor was based on the doctrine of conspiracy in English common law."1 In England, the Journeyman Tailors case (8 Mod., 11) in 1721 established that attempts to raise wages by forming "combina- tions" were criminal conspiracies. Initially, in two cases in Pennsylvania and two in New York, U.S. state courts followed this precedent and held that trade unions were illegal combinations, criminal conspiracies designed to accomplish an illegal end-raising the wages of their members.1

fGoldin, "Maximum Hours Legislation." 7Stimson, Handbook, p. 43. Friedman, History, pp. 493-94.

9 Ibid.

I0 This doctrinehaditsroots in the StatuteofLaborersof 1349 (22ndEdwardlM) whichspecificallyforbade

groups ofworkers from sikingto raise theirwages, and made any attempt t do so aeiminal conspiracy. The

statute, designed in the wake of the Black Death to set wages and prevent laborers from raising their wages

by refusing to work, was later embodied in the Elizabehan Statut of Artificers (5th Elizabeth, Ch. 4). "These cases were the Philadelphia Cordwainers case (1806), People v. Melvin (2 Wheeler Crimi-

nal Cases, NY, 262, 1809), the Journeyman Cordwainers of Pittsburgh case (1811), and People v. Fisher (14 Wendell, NY, 1, 1835).

This content downloaded from 139.182.97.89 on Wed, 02 Nov 2016 23:02:52 UTC All use subject to http://about.jstor.org/terms

The Law and Labor Strife 45

These early American decisions were made in inferior courts. When the first superior court decision was rendered, in 1842 in Commonwealth v. Hunt

(15, 4 Met., 111), the Massachusetts supreme court discarded the Journey- man Tailors precedent and established for the first time the per se legality of labor combinations and strikes."2 The result was a sharp reduction in the number of union conspiracy trials in the 1 850s and early 1860s.13 However, after the Civil War labor unrest increased, and many strikers again faced prosecution under anticonspiracy laws. Unions made the repeal of the conspiracy doctrine one of their highest priorities.14 Before 1880 the only states in our sample that had passed laws recognizing the right of unions to exist were New Jersey, New York, and Pennsylvania; by 1893 five other states had done so.

However, even injurisdictions where the interpretation of anticonspiracy laws left workers free to strike, workers accused of intimidating other workers or organizing boycotts continued to be prosecuted under anti- conspiracy laws. Many states formalized this practice into statutes outlawing intimidation and boycotts. Between 1887 and 1897 six states also limited the behavior of employers by banning blacklisting of workers who joined unions or went on strike."5

Finally, the late 1880s and early 1890s saw a sharp rise in the use of the injunction against striking workers. Both federal and state equity courts had issued injunctions to prevent injury to property during labor disputes throughout the nineteenth century. But it was not until courts recognized that the right of employers to conduct business was a form of "property at risk of injury" in a work stoppage that injunctions were routinely issued to bar work stoppages altogether, to prevent certain forms of behavior by striking workers (such as boycotting or aggressive picketing), or to end stoppages by requiring that workers return to work. The injunction was seen as a powerful weapon against labor stoppages: unlike conspiracy prosecutions which took time and required at least a modicum of evidence, injunctions could be granted after a brief hearing and a mere assertion that harm to a fim's commerce was imminent. Victoria Hattam notes that "The AFL and other

12 Stimson, Handbook, pp. 20304. 13 Fredman, History, pp. 486-87. 14 Hatan, Labor Visions, pp. 20, 72, and 14041. "For example, New York's conspiracy law (in 1887) was actually a change to allow prosecution

of employers who blacklisted workers joining unions: "Any person or persons, employer or employers of labor, and any person or persons of any corporation or corporations on behalf of such corporation or corporations, who shall hereafter coerce or compel any person or persons, employe or employes [sic], laborer or mechanic, to enter into an agreement, either written or verbal, from such person, persons, laborer, or mechanic, not to join or become a member of any labor organization, as a condition of such

person or persons securing employment or continuing in the employment of any such person or persons,

employer or employers, corporation or corporations, shall be deemed guilty of a misdemeanor" (Laws of the State ofNew York, Chap. 688, p. 897, 24 June 1887). The New York Workingmen's Assembly strongly supported passage of this legislation.

This content downloaded from 139.182.97.89 on Wed, 02 Nov 2016 23:02:52 UTC All use subject to http://about.jstor.org/terms

46 Currie and Ferrie

union leaders understood all too well the demoralizing impact of the injunc- tion and renewed their efforts to check the courts' power.",16 She describes a 20-year campaign by organized labor to neutralize the injunction as a legal weapon.

This overview has emphasized statutes that established maximum hours, legalized unions, outlawed intimidation or boycotts on the part of unions, and outlawed the use of blacklists on the part of employers. In addition, courts acted by issuing injunctions against striking workers. What effects would we expect these actions to have on labor disputes?

The occurrence of strikes has long been a puzzle to economists. Factors that change the balance of power in a negotiation are predicted to change the outcome of negotiations, but not to lead to a breakdown in negotiations. The reason is that strikes reduce the size of the pie to be divided between the two parties, and thus any resolution arrived at after a strike could be dominated by a division that occurred without a strike. Faced with this dilemma, Sir John Hicks concluded that strikes were simply "mistakes.""7 However, this does not imply that strikes will occur randomly-simple economics suggests that even mistakes should be less frequent when they are more costly.

More recent theories of strikes emphasize asymmetric information.18 The idea is that the strike can actually play a productive role by transmitting information that could not be credibly transmitted in any other way. Suppose for example that the firm knows the true state of profits, but workers do not. Then, if wages are contingent on profits, the firm has an incentive to report a bad state of the world. Workers can force the firm to report honestly by striking when the firm announces a bad state. If profits really are bad, then the firm will be willing to take the strike. If profits are really good, then it will be costly for the firm to lose production and it will not announce a bad state. Once again, factors that reduce the cost of any given strike are likely to increase the probability that the strike-as-truth-elicitation-device is used. Although these models are often cast in terms of union ignorance about profits, imperfect knowledge about the firm's more general bargaining position yields the same predictions.

These models imply that laws that reduce the cost of any given strike are likely to increase strike incidence, whereas those that reduce the uncertainty surrounding the state of the world (and hence likely strike outcomes) should reduce strike incidence. If there are many states of the world, and it is the

workers who lack information, then a model of this type will generate a "concession curve" in which the gains to the union fall with the length ofthe

16 H m, Labor Visions, p. 163. 17 Hicks, Theory.

18 For a review of this literature, see Kennan, "Economics."

This content downloaded from 139.182.97.89 on Wed, 02 Nov 2016 23:02:52 UTC All use subject to http://about.jstor.org/terms

The Law and Labor Strife 47

strike. Thus, factors that increase uncertainty will also increase the length of strikes, and result in more negative outcomes for labor.

These simple models yield a set of predictions for the effects of the laws we consider. First, maximum hours laws will increase the cost of a given strike under certain assumptions. Suppose, for example, that some employers follow the example set by these laws even when they are not legally bound to do so and others do not. Then employees in low-hours jobs will be afraid of losing those jobs and thus be less likely to strike. If there are fixed costs of employment, then employers honoring maximum hours laws might choose to hire fewer, better quality workers, who will be more costly to replace in the event of a strike. Maximum hours laws might also reduce the uncertainty surrounding strike outcomes (at least when hours of work were the main area of contention) leading to fewer strikes, or strikes of shorter duration, with better outcomes for labor.

The legalization of unions would be expected to reduce the costs of strikes to workers (by eliminating the possibility of being charged with conspiracy), but might increase the cost of strikes to employers if employers respond to higher union wages by hiring better quality workers. Currie and Sheena McConnell argue that the legalization of public sector employee unions after 1960 reduced strike activity by reducing the uncertainty surrounding the collective bargaining rights of these employees.19 One might expect the legalization of unions in the 1 880s to have had similar effects. On the other hand, if some employers did not recognize the legitimacy of the union's position, then uncertainty about the employer's bargaining position could be increased, thereby increasing strike activity.

Finally, laws outlawing intimidation, boycotts, and blacklisting were directly aimed at reducing the costs of strikes to employers and employ- ees, while injunctions were used to limit the cost of any given strike by forcing employees back to work. Thus, cost arguments suggest that these laws ought to have been associated with increased strike activity. One caveat is that the imposition of injunctions was often accompanied by violence, which presumably increased the costs of striking to both workers and firms.

The information model, however, suggests that laws outlawing intimidation, boycotts, and blacklisting ought to have reduced the un- certainty surrounding strikes outcomes, while the occasional use of the injunction would have increased it. Thus, the former laws should be associated with reductions in strike activity or improved labor outcomes, and injunctions should be associated with increased strike activity or worse outcomes for labor.

19 Cuffie and McConnell, "Impact."

This content downloaded from 139.182.97.89 on Wed, 02 Nov 2016 23:02:52 UTC All use subject to http://about.jstor.org/terms

48 Currie and Ferrie

TABLE 1

DATE OF PASSAGE OF STATE LAWS DEALING WITH LABOR AND STRIKES, THROUGH 1894

Unions Maximum Intimidation Boycotts Blacklists Injunction

State Legala Hoursb Illegalc Illegald Illegal Usede

Connecticut no law 1 Jan. 1867 29 Mar. 1878 29 Mar. 1878 no law no

Delaware no law no law no law no law no law no

Illinois no law 1 May 1867 13 Feb. 1863 1 Jul. 1887 1 Jul. 1887 1886 Indiana 25 Feb. 1893 10 May 1889 14 Apr. 1881f no law 9 Mar. 1889 1893 Maine no law 1 Jan. 1871 13 Mar. 1889 13 Mar. 1889 no law no Maryland 8 Apr. 1884 1 Apr. 1886 no law no law no law no Massachusetts 14 Mar. 1888 3 May 1890 14 May 1875 no law 31 May 1892 1888 Michigan 6 Jun. 1883 5 Jun. 1885 27 Mar. 1867 no law no law no NewHamp- no law 3 Jul. 1847 29Sep.1887 29 Sep. 1887 no law no

shire

New Jersey 9 Mar. 1877 8 Apr. 1887 no law no law no law 1894 NewYork 17 Feb. 1870 26Apr. 1870 30 Jun. 1882 30 Jun. 1882 24 Jun. 1887 no Ohio 14 Apr. 1892 1 May 1886 no law no law no law 1887 Pennsylvania 8 May 1869 24 Mar. 1887 no law no law no law 1888 Number of 5 7 4 4 4 6 changess a Laws declaring iat collective action to raise wages was not an actionable conspiracy, or laws prevent- ing the discharge of workers because of union membership, or laws establishing procedures for unions to incorporate. b Laws specifying the maximum length of the workday either in the absence of any specific agreement to the contrary, or in specific industries, or in the employment of the state.

I Laws preventing the use of threats or force to prevent workers from practicing their trade. dAnti-intimidation laws that were written to include the intimidation ofcustomers, anticonspiracy laws that were written to prevent collective action that was directed at preventing trade or business, or laws

that specifically outlawed boycotts.

e The earliest year for which a citation could be found in Frankfurter and Greene (Labor Injunction), Oakes (Law), Petro ("Assumptions"), or Witte ("Early American Labor Cases") to indicate that an injunction had been issued and sustained by a federal or state court in a labor dispute. f Repealed 9 March 1889. g Number of these laws passed in the period we examine (1881 through 1894). Source: See note 20 in the text.

THE DATA

We have drawn information about the labor law in effect in each state between 1881 and 1894 from published state statutes, proceedings of state legislatures, and compilations ofjudicial decisions in labor disputes brought before the courts.20 Table 1 summarizes the legal environment in each state in terms of these categories. A date indicates that the relevant law was passed at that time, while "no law" indicates that there was no law in effect as of 1894. The table illustrates both the tremendous heterogeneity in state legal environments and the difficulty in identifying "packages" of laws that tended to go together. For example, several states both recognized unions

20 These sources are described in detail in a data appendix available from the authors or on-line at http://www.econ.nwu.edu/faculty/ferrie/papers/strikelawl.pdf .

This content downloaded from 139.182.97.89 on Wed, 02 Nov 2016 23:02:52 UTC All use subject to http://about.jstor.org/terms

The Law and Labor Strife 49

and outlawed the intimidation of strikebreakers. Table 1 also shows that in several large states, laws banning intimidation and boycotts were passed simultaneously, which makes it difficult to identify the separate effect of these laws. Since both laws were intended to place limits on workers' right to organize, we have grouped them together in our empirical work. Finally, the table illustrates some changes in the legal environment that occurred after the Haymarket Riot of 1886. Between 1887 and 1894 four states adopted maximum hours laws for at least some groups of workers. Eight states had such a law prior to 1886. Illinois, one of the states hardest hit by the strike wave of 1886, passed an innovative law outlawing both boycotts and blacklisting in 1887.

The data that will allow us to assess the effects of these laws are drawn from two sources: the Third Report and Tenth Report of the U.S. Commis- sioner of Labor. The reports describe strikes in the years 1881 to 1886 and in 1887 to 1894, respectively. Investigators combed through newspaper reports and other contemporary sources in order to compile an initial list of strikes. They then conducted interviews m each location where a strike was reported to obtain detailed information about each strike, as well as information about other strikes.2" For each strike, the reports include the beginning and ending dates of the strike, the industry, the location, the number of male and female workers in the firm before the strike, the number

of workers involved, the hours of work before and after the strike, whether the strike was authorized by a union, and whether replacement workers were used. The Third Report also recorded workers' wages before and after strikes and firms' employment levels after strikes. Compared to recent strike data sets, this is a very rich source of information about industrial disputes.

We coded information about 13,302 strikes in 13 states drawn from three broad regions: the Midwest (Illinois, Indiana, Michigan, and Ohio); New

21 By way of comparison, recent strike data are compiled from newspaper reports, and only include

strikes involving over 1,000 workers, althoughpriorto 1981, the BureauofLabor Statistics used survey data and attempted to collect information about all strikes. Edwards (Strikes) and Griffm (Strikes) both contain extensive discussions of the reliability of the Commissioner of Labor's data collection proce- dures. A recent re-examination of the data from the Third Report and the Tenth Report for Terre Haute,

Indiana, however, fimds that only half of the strikes for which there exists a record were included (Bailey, "Commissioner"). The strikes that were omitted appear no different (in size, industry, or duration) from those that were included. A second problem with the reports is that the Third Report used the enterprise as the unit of observation (that is, related strikes at different plants were counted as

separate strikes). The Tenth Report used a broader definition of a strike that counts strikes that began at roughly the same time over similar issues as a single dispute. It is not clear how religiously the new definition was applied in the Tenth Report. We find many instances in which apparently related strikes were nonetheless coded as separate strikes. Hence, we have chosento treat the datafrom the two reports in the same way, and to rely on the inclusion of year effects to capture systematic differences in strike

prevalence associatedwith changes in reporting conventions over time. See Card and Olson ("Bargain- ing Power") for additional discussion of these issues. These data have also been used by Rosenbloom ("Strikebreaking") to study the use of strikebreakers, and by Friedman ("Politics") to study the impact of unions.

This content downloaded from 139.182.97.89 on Wed, 02 Nov 2016 23:02:52 UTC All use subject to http://about.jstor.org/terms

50 Currie and Ferrie

England (Connecticut, Maine, Massachusetts, and New Hampshire); and the Middle Atlantic states (New Jersey, New York, Pennsylvania, Delaware, and Maryland). These states were chosen because they experienced almost 90 percent of all reported strikes and because they exhibit great variation in labor law. We excluded fims with under six prestrike employees because we felt that strikes in fimns this small were likely to be less accurately reported than strikes in larger firms. We also excluded one firm that reported a prestrike wage of greater than $1 0 per hour. These exclusions leave us with a sample of 12,965 strikes. We focus on six measures of strike outcomes: the percentage changes in

wages, employment, and hours; whether strike replacements were used; the fraction of workers replaced conditional on the use of replacements; and the unconditional mean fraction of the workforce replaced. We also examine two measures of strike cost: strike duration and the number of working days lost. In general, increases in wages, smaller poststrike employment losses, reductions in hours, reduced use of strike replacements, and shorter strikes could all be regarded as positive outcomes for labor. Reduced strike lengths represent a Pareto improvement in that they are also a benefit to employers. Note that wage and employment changes are only available for the 1881 to 1886 period while antiblacklist laws and injunctions generally came into effect only after 1886.22 Therefore, it is not possible to assess the effects of these two types of laws on post-strike wage and employment outcomes. In addition to these measures, we examine the extent to which strikes were authorized by unions as one indicator of how the legal environment might have influenced the composition of our sample. Table 2 describes the strike-level data. Our measure ofthe extent to which

strike activity was union-authorized appears in column 1. Strikes were most likely to be union-authorized in New York, and in the building trades, tobacco, and food processing and brewery industries. Columns 2 and 6 of Table 2 show the distribution of prestrike employment and the changes in employment that occurred following strikes. Columns 3 to 5 give the fraction of strikes in which strike replacements were employed, the fraction replaced conditional on strikebreakers being used, and the overall mean percentage replaced. The use of strikebreakers was very common, especially in Delaware and

in some industries (printing, publishing, and telegraph; food processing and brewing; transportation; and the residual category) in which strikebreakers

22 Table 2 shows that where measures of strike outcomes are available in both reports, there is continuity between the information contained in the Third Report (1881 to 1886) and the Tenth Report (1887 to 1894). Nevertheless, possible changes in reporting conventions between the two reports provide a fiurther justification for the inclusion of year dummies in our regression analysis as discussed later. We use nominal wages since we focus on percentage wage changes following strikes, so that the relevant time interval (the duration of the strike) is usually quite short.

This content downloaded from 139.182.97.89 on Wed, 02 Nov 2016 23:02:52 UTC All use subject to http://about.jstor.org/terms

The Law and Labor Strife 5 1

TABLE 2

STRIKE CHARACTERISTICS IN EACH STATE, YEAR, AND INDUSTRY

Fraction Fraction Mean Mean Mean Mean Mean

Percentage Mean Using Replaced if Mean Percentage Hours Percentage Wage Percentage Median Median Number of Strikes Employment Strike Replacements Fraction Change in before Change in before Change in Strike Days of Authorized before Strike Breakers Used Replaced Employment Strike Hours Strike Wage Duration Lost Strikes (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) All strikes 12,965 0.650 387 0.417 0.246 0.103 -1.60 58.3 -0.611 1.91 2.27 7 765

By state

Connecticut 517 0.342 283 0.426 0.195 0.083 -0.051 60.4 -0.442 1.51 1.27 5 735 Delaware 39 0.615 268 0.615 0.307 0.109 -20.7 56.5 -2.46 1.61 -8.03 17 2,635 Illinois 1,587 0.681 743 0.400 0.279 0.112 -0.396 58.9 -1.46 2.00 1.60 7 990 Indiana 308 0.597 408 0.328 0.280 0.092 -2.89 58.7 -0.705 1.78 1.01 7 888 Massachusetts 2,392 0.565 336 0.470 0.204 0.096 -2.90 58.5 -0.271 1.74 2.90 7 852 Maryland 164 0.732 308 0.494 0.247 0.122 -3.02 58.5 -0.349 1.67 1.52 10 1,175 Maine 119 0.580 360 0.462 0.189 0.087 -5.48 59.8 -0.667 1.61 2.58 8 1,575 Michigan 273 0.553 297 0.487 0.300 0.146 -2.51 58.9 -0.740 1.71 1.82 10 1,000 New Hamp- 85 0.459 314 0.335 0.204 0.072 -4.74 59.1 -0.772 1.66 1.24 9 1,470

shire

New Jersey 490 0.682 304 0.429 0.264 0.113 -5.39 58.4 -0.710 1.91 2.23 7 873 New York 3,962 0.825 190 0.384 0.281 0.108 -0.656 57.1 -0.642 2.21 5.62 6 256 Ohio 1,075 0.572 327 0.392 0.220 0.086 -4.14 59.3 -0.440 1.77 0.723 12 1,314 Pennsylvania 1,954 0.517 664 0.434 0.225 0.098 -0.760 58.7 -0.333 1.76 0.593 11 2,000

By year

1881 503 0.531 268 0.384 0.280 0.108 -0.652 62.5 -1.04 1.89 5.40 7 990 1882 434 0.551 371 0.364 0.266 0.097 -0.556 59.9 -0.190 2.01 3.37 7 1,137 1883 468 0.607 308 0.429 0.289 0.124 -2.35 58.1 -0.015 1.92 2.16 10 1,000 1884 424 0.597 347 0.417 0.268 0.118 -3.52 59.6 -1.16 2.04 -1.65 11 1,014 1885 622 0.566 392 0.391 0.277 0.108 -2.87 59.4 -0.138 1.83 1.58 14 1,694 1886 1,860 0.662 310 0.420 0.278 0.117 -1.05 59.9 -1.69 1.88 2.29 11 1,035 1887 1,355 0.633 420 0.445 0.235 0.105 - 58.9 -0.426 - - 6 700

This content downloaded from 139.182.97.89 on Wed, 02 Nov 2016 23:02:52 UTC All use subject to http://about.jstor.org/terms

52 Currie and Ferrie

TABLE 2-continued

(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) 1888 813 0.679 376 0.439 0.246 0.108 - 57.8 -0.260 - - 7 720 1889 922 0.672 454 0.425 0.225 0.096 - 57.7 -0.493 - 6 644 1890 904 0.612 457 0.402 0.217 0.087 - 58.5 -0.923 7 900 1891 1,515 0.731 287 0.393 0.253 0.099 - 56.9 -0.497 - - 6 483 1892 1,182 0.690 295 0.409 0.226 0.092 56.1 -0.296 6 525 1893 1,150 0.681 403 0.462 0.226 0.104 57.4 -0.305 7 560 1894 813 0.627 805 0.400 0.199 0.080 - 57.1 -0.063 - - 7 725

By industry

1 2,901 0.627 330 0.404 0.180 0.08 -2.70 59.9 -0.129 1.60 2.81 7 980 2 965 0.456 374 0.512 0.259 0.13 -1.95 60.9 -0.636 1.63 2.35 7 780 3 803 0.783 206 0.435 0.265 0.12 -1.16 58.5 -1.37 1.92 1.31 10 630 4 238 0.857 586 0.567 0.305 0.18 -0.860 72.1a -4.00 1.96 7.50 4 190 5 1,100 0.497 822 0.307 0.212 0.07 -2.91 57.9 -0.364 1.78 -1.96 14 3,150 6 275 0.635 385 0.465 0.182 0.09 -1.60 59.1 -0.570 2.16 2.23 12 1,350 7 1,379 0.546 458 0.442 0.195 0.09 -0.542 59.8 -0.338 1.94 1.37 10 1,600 8 334 0.763 167 0.664 0.322 0.23 -3.95 58.1 -0.218 2.16 3.94 6 300 9 255 0.188 266 0.486 0.367 0.19 7.29 59.4 -0.693 1.51 2.30 3 325 10 627 0.349 1,088 0.555 0.257 0.14 0.666 65.1 -0.424 1.96 5.98 3 588 11 2,567 0.836 250 0.335 0.284 0.11 -0.938 53.9 -0.940 2.64 4.89 5 228 12 257 0.618 315 0.377 0.167 0.06 -0.829 55.7 -0.106 2.11 -0.798 14 2,390 13 502 0.731 153 0.420 0.280 0.12 0.215 55.1 -1.09 2.28 2.09 7 366 14 762 0.870 189 0.412 0.362 0.15 -3.50 54.4 -0.467 1.67 4.11 14 600

a The median was also 72.

Notes: The fraction replaced is conditional on strike replacements being used. Poststrike employment and wage data (columns 6, 9, and 10) are available only for 1881 to 1886. The industry codes are as follows: I = clothing, textiles, and shoes, 2 = miscellaneous, 3 = cooperage, wooden goods, and furniture, 4 = food preparation and brewing, 5 = mining industry, 6 = machines and machinery industries, 7 = metals and metallic goods, 8 = printing, publishing, and telegraph, 9 = public way or works constructions, 10 = transportation industry, 11 = building trades including construction of carriages and transportation equipment, 12 = glass and pottery, 13 = stone

quarrying and cutting, and 14 = tobacco.

Sources: U.S. Commissioner of Labor, Third Annual Report and Tenth Annual Report.

This content downloaded from 139.182.97.89 on Wed, 02 Nov 2016 23:02:52 UTC All use subject to http://about.jstor.org/terms

The Law and Labor Strife 53

were used in over half of all strikes. When strike replacements were used,

typically about a quarter of the prestrike workers were replaced. Overall,

approximately 11 percent of striking workers were replaced, although this

proportion was as high as 23 percent in the printing, publishing, and

telegraph industry. Column 6 shows the difference between poststrike and

prestrike employment. As noted previously, poststrike employment appears only in the ThirdReport, so it is unavailable after 1886. Except in Delaware, the overall employment effect of strikes was small. Hence, the main threat to striking workers was that they would be replaced, not that their establish- ments would be "down-sized" or shut down.

Columns 7 through 1 0 of Table 2 show the mean hours worked and wages before the strike, and the mean percentage change in hours and wages after the strike. The average strike was accompanied by a very small change in hours. A 0.6 percent reduction in hours at an establishment with a 58 hour week is a reduction of less than half an hour. The strike wave of 1886 was associated with somewhat larger average percentage reductions in hours-at the mean of 60 hours per week, a 1.7 percent reduction amounts to an average reduction of 1 hour per week. Columns 9 and 10 indicate that strikes in the early years of the sample were also associated with modest increases in wages (except in Delaware). The largest percentage wage increases were

in New York (5 .6 percent) and in the food processing and breweries industry (7.5 percent).

The final two columns of Table 2 show median completed strike durations and numbers of working days lost (calculated as the product of employment

before the strike and the number of strike days). We show the medians because the distributions are skewed to the right by a few particularly long or large strikes. Because ofthe skewed distribution of strike lengths or sizes,

we will use the logarithms of these variables as the dependent variables in our regression analysis.23 The median strike lasted seven days and involved 765 working days lost. The median duration reached a peak of 14 days in 1885, and then declined back to six or seven days after 1886. The median number of working days lost shows a similar temporal pattern. In New York, where most strikes were union-authorized, the median strike was short and involved relatively few days lost.

Table 3 shows how the same measures of strike outcomes and costs vary with the legal environment. Since we know the date each strike began, we

can group strikes according to whether a particular type of law was in effect on the day the strike began. Strikes were more likely to be authorized by unions in jurisdictions where unions were legal, maximum hours legislation

23 Recall that Ordinary Least Squares regression assumes that errors are normally distributed. If instead, errors follow a distribution that is skewed to the right, taking logs can make them appear more "normal."

This content downloaded from 139.182.97.89 on Wed, 02 Nov 2016 23:02:52 UTC All use subject to http://about.jstor.org/terms

54 Currie and Ferrie

TABLE 3

MEAN STRIKE CHARACTERISTICS AND OUTCOMES BY LEGAL ENVIRONMENT

Fraction Fraction Mean Mean Mean Mean Mean

Percentage Mean Using Replaced if Mean Percentage Hours Percentage Wage Percentage Median Median Number of Strikes Employment Strike Replacements Fraction Change in before Change in before Change in Strike Days of Authorized before Strike Breakers Used Replaced Employment Strike Hours Strike Wage Duration Lost Strikes (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) All strikes 12,965 0.650 387 0.417 0.246 0.103 -1.60 58.3 -0.611 1.91 2.27 7 765

Unions legal

Yes 8,775 0.683 343 0.425 0.245 0.104 -1.15 57.8 -0.477 1.99 3.10 7 630 No 4,190 0.580 478 0.399 0.247 0.099 -2.10 59.3 -0.895 1.82 1.38 7 1,098

Maximum hours legislation

Yes 8,800 0.703 407 0.406 0.259 0.105 -0.821 58.0 -0.732 2.02 3.44 7 616 No 4,165 0.538 344 0.439 0.219 0.096 -2.59 58.9 -0.356 1.77 0.822 7 1,190

Intimidation or boycotts illegal

Yes 9,007 0.683 341 0.412 0.251 0.103 -1.07 58.0 -0.672 1.99 2.80 7 566 No 3,958 0.574 490 0.427 0.234 0.100 -2.33 59.1 -0.475 1.80 1.56 10 1,500

Blacklists illegal

Yes 3,516 0.768 254 0.391 0.246 0.096 - 56.2 -0.322 - - 5 270 No 9,449 0.606 436 0.427 0.245 0.105 - 59.1 -0.719 - - 7 1,050

Injunction used

Yes 3,358 0.580 478 0.474 0.212 0.100 - 58.2 -0.318 - - 7 1,050 No 9,607 0.674 355 0.397 0.260 0.103 - 58.3 -0.713 - - 7 686

Note: Columns 6, 9, and 10 use 1881 through 1886 only.

Source: See Table 2.

This content downloaded from 139.182.97.89 on Wed, 02 Nov 2016 23:02:52 UTC All use subject to http://about.jstor.org/terms

The Law and Labor Strife 55

existed, intimidation of strikebreakers or the use of boycotts were illegal,

blacklists were banned, and the injunction had not yet been used against labor (column 1).

Similar patterns hold for the size of striking firms (column 2): smaller firms were more likely to have strikes injurisdictions with legal unions, laws outlawing boycotts, intimidation, and blacklists, and no history of antilabor injunctions. Maximum hours laws, however, are an exception to this pattern: they were associated with somewhat larger prestrike employment levels.

The third column of Table 3 shows that employers were more likely to use strikebreakers in jurisdictions in which the injunction had been used. However, conditional on strikebreakers being used, a smaller fraction of workers were replaced in these jurisdictions. Overall, as column 5 shows, liberal labor laws (legal unions, maximum hours laws, and illegal blacklists, and no recent use of the injunction against labor) were associated with a slightly higher probability of being replaced, as were laws outlawing intimidation and boycotts.

Columns 6, 8, and 10 of Table 3 indicate that although wage gains were higher in jurisdictions with liberal labor laws and in those that outlawed intimidation and boycotts, there was little variation across legal environ- ments in poststrike employment losses in the 1881 to 1886 period, or in hours changes over the whole period. The prestrike wage differential between striking firms in union-legal jurisdictions and striking firms in other jurisdictions was approximately 10 percent ($1.99 versus $1.82) as shown in column 9. Wages were also higher for striking firms in jurisdictions with maximum hours laws and injurisdictions in which intimidation and boycotts had been outlawed. However, they were lower in jurisdictions in which the injunction had been used. These data indicate that most strikes resulted in 1 percent to 3 percent increases in wages. Wage increases were highest in jurisdictions with legal unions and maximum hours legislation, and where boycotts and intimidation of strikebreakers were illegal.

Turning to strike length and days lost in columns 11 and 12, the outlawing of intimidation, boycotts, and blacklists was associated with reduced strike duration. These laws were also associated with reductions in the number of strike days lost, as were legal unions and maximum hours laws.

In summary, Table 2 suggests that there were large variations in all our measures of strike length, days lost, and outcomes by state, year, and industry. Table 3 indicates that there was also a great deal of variation in these measures across different legal environments. On the whole, liberal labor laws were associated with lower strike lengths and days lost, higher poststrike increases in wages, and a greater probability of workers being replaced.

This content downloaded from 139.182.97.89 on Wed, 02 Nov 2016 23:02:52 UTC All use subject to http://about.jstor.org/terms

56 Currie and Ferrie

EFFECTS ON STRIKE INCIDENCE

In addition to strike outcomes conditional on a strike having taken place, it is possible that law changes had an impact on strike incidence, and on the composition of striking finns. In order to examine incidence, we construct state-year-level measures of strike activity by calculating state-year measures of the number of strikes, the number of strikers, and the number of days lost due to strikes from our strike-level data. Table 4 examines aggregate strike activity using OLS models of the form

STRIKEst=a+bLAWSst + cSTATEs+dYEARt+ eTRENDst +ut (1)

where STRIKE is one of our three measures of strike activity (the number of strikes, the number of workers involved, and the number of strike days lost), LAWS is a vector referring to the five law variables discussed previously, STATE is a vector of state dummies intended to control for fixed characteris- tics of states that might be related to strike activity (such as the average population over our period, the average level of economic activity over the time period, or the industrial composition of the state in a base year), YEAR is a vector of year dummies to control for factors that are common across states in a given year (such as the effect of a national recession), and TREND refers to a state-specific time trend. These state-specific trends allow, for example, for population growth or industrialization that proceeded at different rates across states. The subscript s refers to the state, while the subscript t refers to the year.

Table 4 suggests that the legal environment had little effect on aggregate- level strike activity.24 In Table 5 we adopt a different approach and look at how the prestrike characteristics of individual striking manufacturing firms in 1881 and 1891 differed from the characteristics of the average firm in the Censuses of Manufacturers in 1880 and 1890. We use data from the 1880 and 1890 Censuses of Manufacturers in order to compare our striking firms to the average manufacturing firm in each state. The characteristics we examine are prestrike employment, the fraction female, and the number of prestrike hours worked. We use information on striking firms from 1881 and 1891 because data for striking fmns in 1880 are not available. For compara- bility with the census data, the models examining employment and the

24 We also graphically examined the relationship between aggregate strike incidence and the imposi- tion of specific laws for each state, with similar results. The figures are available on-line at http://www.econ.nwu.edu/faculty/ferrie/papers/strikelaw2.pdf and from the authors. The legalization of unions was preceded by a fall in our measures of strike activity and followed by a temporary rise,

and the imposition of a maximum hours law was preceded by a rise in strike activity and followed by a temporary fall in strike activity. There were no clear patterns for laws banning boycotts, intimidation,

or blacklisting, or for the use of the injunction. The states that experienced no change in their laws during the 1881 to 1894 period (Connecticut and Delaware) had cycles in strike activity at least as pro- nounced as those associated with legal changes in the other states.

This content downloaded from 139.182.97.89 on Wed, 02 Nov 2016 23:02:52 UTC All use subject to http://about.jstor.org/terms

The Law and Labor Strife 57

TABLE 4

THE EFFECTS OF THE LAW ON AGGREGATE STRIKE ACTIVITY

Number of Number of Strikers Number of Strike Days

Strikes (OOOs) (OOOs)

Independent variable (1) (2) (3)

Intercept -1.92 -10.3 549

(0.032) (0.389) (0.553) Unions legal 24.1 6.78 397

(0.973) (0.619) (0.967) Maximum hours legislation -31.4 -4.22 249

(1.24) (0.377) (0.593) Intimidation or boycotts illegal 10.43 9.06 204

(0.290) (0.568) (0.342) Blacklist illegal 15.9 23.8 39.3

(0.515) (1.74) (0.077) Injunction used -26.1 0.613 -562

(1.02) (.054) (1.33)

Observations 182 182 182 R-squared 0.784 0.754 0.688 Mean of dependent variable 71.2 27.6 769

Notes: t-statistics are in parentheses. All regressions also included indicators for each state and year as well as state-specific time trends.

fraction female exclude strikes in mining, printing, publishing, telegraph, public ways or works construction, transportation industries, building trades, and stone quarries. In 1890 hours are available for a somewhat different set of industries (agricultural implements; boots and shoes; carriages and wagons; cheese, butter, and condensed milk; flouring and grist mill; leather; paper; slaughtering and meat packing; and wholesale slaughtering excluding meat packing).25

25 The number of fms and the numbers of male and female wage earners by state for 1880 and 1890

are given in the 1900 Census of Manufacturers (UJ.S. Bureau of the Census, Twelfth Census, table 1, pp. 982-88). The average employment per establishment and fraction female in the labor force were calculated for each state directly from these figures. Average daily hours of labor were not reported in the published volumes ofthe 1880 Census ofManufactures. These datawere, however, collected as part ofthe census, and Atack and Bateman have retrieved them from the manuscript schedules ofthe census

for a sample of firns (Atack and Bateman, State Sample). Comparable figures for daily hours in 1890 were obtained from the published returns of the 1890 Census of Manufacturing (U.S. Census Office, Eleventh Census, table 8, pp. 654-738). In 1890 hours were reported only for a subset of industries, however. From this subset, nine industries were selected that covered most of employment inmanufac-

turing (agricultural implements; boots and shoes; carriages and wagons; cheese, butter, and condensed milk; flouring and grist mill; leather; paper; slaughtering and meat packing; and wholesale slaughtering excluding meat packing). Average daily hours were calculated by taking the establishment-weighted average of the reported figures for these nine industries (similar results were produced using employ-

ment as weights). The set of industries drawn from the Atack and Bateman sample was restrictedto the

nine industries used in 1890 for this calculation. For both 1880 and 1890, the average daily hours figure

used was the figure for May to November if the strike occurred in that period and the figure for Novem-

ber to May if the strike occurred then.

This content downloaded from 139.182.97.89 on Wed, 02 Nov 2016 23:02:52 UTC All use subject to http://about.jstor.org/terms

58 Currie and Ferrie

TABLE 5

DIFFERENCES BETWEEN STRIKING FIRMS AND THE AVERAGE MANUFACTURING FIRM IN 1881 AND 1891

Employment Fraction Female Hours

Independent variable (1) (2) (3)

Intercept 140 -0.118 0.916

(0.885) (2.05) (0.396)

Unions legal -165 0.013 -2.94 (1.40) (0.306) (1.81)

Maximum hours legislation -37.5 -0.039 4.69 (0.346) (0.979) (3.12)

Intimidation or boycotts illegal -306 -0.019 - 1.82 (0.992) (0.168) (0.475)

Blacklist illegal 50.2 0.135 -6.85 (0.157) (1.16) (1.71)

Injunction used -18.1 0.065 -0.730 (0.154) (1.52) (0.450)

Observations 1041 1041 1192 R2 0.163 0.413 0.569 Mean of dependent variable 244 -0.026 -3.01

Notes: t-statistics are in parentheses. All regressions also included state, year, city, and industry dummies. Columns 1 and 2 include strikes only from industries 1, 2, 3, 4, 6, 7, 12, and 14. Column 3 includes strikes only from industries 1, 4, 6, 7, and 11.

These regressions are estimated using the firm-level data and are of the form

CHARstj = a + bLAWSst + cSTATEs + dYEARt + eMSAst4 + (2) fINDUSTRYsti + usti

where CHAR is the difference between a firm's characteristic (such as the number of employees) and the average characteristic of finms in the same state and year, MSA is a vector of 18 controls for the firm's city, and INDUSTRYis a vector of 13 industry-level dummy variables. When we use the firm-level data, it is important to control for characteristics of cities and industries, because there may be systematic differences in the characteristics of striking firms in different cities and industries which have little to do with the legal environment. We cannot include state time trends in this model since there are only two years of data (1 881 and 1891). The subscript i refers to the individual firm.26

Table 5 shows that the typical striking finn was about 244 workers larger than the typical manufacturing finn in the same state and year. It also had a slightly lower fraction of female employees and had lower weekly hours. Columns 1 and 2 show that the legal environment appears to have had little

26 The major cities are: Boston-Cambridge, Brooklyn, Baltimore, Chicago, Cincinnati, Cleveland, Detroit, New York, Springfield (Illinois), Springfield (Massachusetts), Worcester, Fall River (Massa- chusetts), Indianapolis, Lynn, Newark, New Haven, Philadelphia, and Pittsburgh.

This content downloaded from 139.182.97.89 on Wed, 02 Nov 2016 23:02:52 UTC All use subject to http://about.jstor.org/terms

The Law and Labor Strife 59

effect on the prestrike size of striking firms or on the fraction of female employees at striking firms relative to the average. The largest changes in the composition of striking firms relative to other firms appear to have been in the prestrike hours of striking firms. Laws permitting unions and anti- blacklist laws were associated with increases in strike activity in firms with relatively low hours, whereas maximum hours laws increased strike activity at firms with relatively long hours. Finally, in column 1 of Table 6 we use the firm-level data to examine the

effect of the legal environment on the probability that a union authorized a particular strike. The estimating equation is

OUTCOMESti = a + bLA WSSt + cSTA TEs + d YEAR, + eTRENDst + (3) fMSASti + gINDUSTRYsti + hXsti + iNUMSTRsti + usti

where OUTCOME is union authorization of a strike, and the vector X includes all of the observable characteristics of the firm such as the log of prestrike employment in the firm, prestrike hours, and the fraction female in the firm, and NUMSTR is the total number of strikes in the same state-year- industry cell. This latter variable is included to control for any omitted variables that might be correlated with strike waves in a particular state, year, and industry. As we will show, these variables had important and interesting effects on strike outcomes. Unless otherwise noted, however, their exclusion does not affect the estimated coefficients on the law variables. We do not include the wage at the beginning of the strike since it is not available for the entire period.27 Note that we have used linear probability models in cases for which the

dependent variable was a zero-one indicator. Logistic regression models produced very similar estimates, and we feel that the linear probability models are easier for readers to interpret and compare to models with continuous dependent variables estimated using Ordinary Least Squares. Table 6 shows that although unions were much more likely to authorize

strikes in firms with low fractions of female workers, and they were more likely to have authorized a given strike in years with many strikes (evidence of union-sponsored strike "waves"), we find relatively little effect of the legal environment. Legal unions made union authorization only marginally more likely, whereas previous use of the injunction in a state made umnon authorization of strikes less likely. In summary, once we control for differences between states, years, or

industries, there is little evidence of any systematic effect of the legal

27 Results forthe 1881 to 1886 subsamplewere similarwhetherornotwages were included, although including the wage caused the coefficients on size, hours, and the fraction female to fall in absolute value. This is to be expected, since firm size and wages are strongly positively correlated, while hours and fraction female are strongly negatively correlated with wages in these data.

This content downloaded from 139.182.97.89 on Wed, 02 Nov 2016 23:02:52 UTC All use subject to http://about.jstor.org/terms

60 Currie and Ferrie

TABLE 6

UNION AUTHORIZATION, STRIKE LENGTH, DAYS LOST, AND THE LEGAL ENVIRONMENT

Union Authorized Log(Strike Length) Log(Strike Days Lost) Independent variable (1) (2) (3)

Intercept 0.641 1.91 7.60 (7.44) (7.57) (19.7)

Unions legal 0.045 0.076 -0.155 (1.78) (1.02) (1.35)

Maximum hours legislation 0.011 -0.451 -0.369 (0.457) (6.12) (3.24)

Intimidation or boycotts illegal 0.019 0.139 0.385 (0.505) (1.23) (2.21)

Blacklist illegal 0.008 0.138 -0.056 (0.019) (2.72) (0.666)

Injunction used -0.073 0.211 0.250 (2.81) (2.72) (2.09)

Log(prestrike employment) -0.002 0.119 (0.636) (14.1)

Prestrike hours -0.112 -0.837 -0.733 (1.74) (4.43) (2.51)

Fraction female -0.475 -0.067 1.64 (17.53) (0.841) (13.4)

Number of strikes in state-year- 0.021 -0.130 -0.210 industry cells (00s) (2.51) (5.24) (5.49)

Observations 12,829 12,829 12,829 RI 0.229 0.144 0.228 Mean of dependent variable 0.651 2.04 6.69 F-test for 5 laws 2.07 9.77 9.77

(p-value) (0.066) (0.0001) (0.0001)

Notes: t-statistics are in parentheses. All regressions also included state, year, city, and industry

dummies as well as state-specific time trends.

environment on aggregate strike activity. Moreover, striking firms do not appear to be selected differently in different legal environments in terms of size or the fraction female. The legal environment did, however, exert some impact on the selection of striking firms in terms of hours: laws legalizing unions and eliminating blacklists encouraged relatively low-hours firms to strike, whereas maximum hours laws encouraged relatively high-hours fmns to strike. The law also affected unions' propensity to authorize a strike: union recognition increased the odds of union authorization whereas the use of the injunction decreased it.

In terms of the theory, these findings suggest that changes in the laws governing collective bargaining either had ambiguous effects on strike lengths and days lost, or were unsuccessful in reducing the uncertainty of the bargaining parties. We will be able to shed more light on these questions after examining the effects of the legal environment on strike outcomes.

This content downloaded from 139.182.97.89 on Wed, 02 Nov 2016 23:02:52 UTC All use subject to http://about.jstor.org/terms

The Law and Labor Strife 61

EFFECTS OF THE LEGAL ENVIRONMENT ON STRIKE OUTCOMES

In this section we turn to the strike-level data and examine the effects of the

legal environment on strike lengths, days lost, and strike outcomes, given that a strike occurred. The models are of the form of equation 3. Columns 2 and 3 of Table 6 show models of strike length and the number of working days lost. Prestrike employment is not included in the model of the number of working days lost, since there is a mechanical relationship between the two

variables-the larger the establishment, the more working days can be lost.

The estimates in columns 2 and 3 indicate that maximum hours laws were

associated with both significantly shorter strikes and fewer days lost. In terms of the theory, these findings suggest that maximum hours laws re-

duced the union's level of uncertainty about employer intentions or likely strike outcomes, given that a strike had already occurred. Both strike length and the number of worker days lost in strikes were significantly higher in states where an injunction had previously been issued. The use ofthe injunc-

tion increased strike length and days lost by slightly more than 20 percent. This result may seem counter-intuitive, since the injunction was often used notjustto prevent striking workers from engaging in specific activities (such

as aggressive picketing or the distribution of leaflets) but also to force workers to return to work, with union leaders subject to contempt penalties for failure to comply. Strikes that were actually enjoined are likely to have been shorter than they would have been otherwise. However, by creating additional uncertainty about the likely outcome of the strike, the threat of an injunction apparently made it more difficult for parties to a dispute to settle.

Finally, evidence regarding the effects of banning intimidation, boycotts, and blacklists is mixed. Banning intimidation and boycotts increased the number of days lost, without increasing strike length. Banning blacklists increased strike length without seeming to affect the number of days lost. In both cases, the elimination of familiar strike weapons seems to have been associated with increasing uncertainty about the likely outcome ofthe strike, and thus with an increase in strike length or days lost.

Table 7 explores further the effect of legislation on strike outcomes. These models use the same variables as those in Table 6. The one exception is that in models of the percentage change in employment, we exclude prestrike

employment, and similarly we exclude prestrike wages and prestrike hours in regressions on the percentage changes in wages and hours, respectively.

Our purpose was to avoid incorporating a purely mechanical relationship between changes and levels. Regressions including these prestrike variables produced estimates similar to those shown here.28 No states had antiblacklist

28 The only exception is that the fraction female has a positive effect on the percentage change in employment when prestrike employment is included.

This content downloaded from 139.182.97.89 on Wed, 02 Nov 2016 23:02:52 UTC All use subject to http://about.jstor.org/terms

62 Currie and Ferrie

TABLE 7

STRIKE OUTCOMES AND THE LEGAL ENVIRONMENT

Percentage Percentage Percentage Change Strike Fraction Replaced given Mean Fraction

Change in Hours Change in Wage in Employment Replacements Used Replacements Used Replaced

Independent variable (1) (2) (3) (4) (5) (6) Intercept 0.271 -1.40 3.73 0.136 0.581 0.186 (0.324) (0.491) (0.936) (1.39) (7.66) (4.56) Unions legal -0.146 4.05 -2.31 0.056 -0.037 0.002 (0.278) (1.46) (0.572) (1.96) (1.75) (0.143)

Maximum hours legislation -0.178 5.82 0.767 -0.083 0.025 -0.016 (0.645) (4.32) (0.393) (2.91) (1.18) (1.36)

Intimidation or boycotts illegal -0.345 -0.970 -1.99 0.006 0.007 0.000

(0.817) (0.706) (1.03) (0.147) (0.218) (0.024)

Blacklist illegal -0.145 - 0.062 0.010 0.022 (0.705) (2.92) (0.629) (2.47) Injunction used 0.202 0.022 0.037 0.018 (0.696) (0.718) (1.60) (1.46) Log(prestrike employment) 0.019 -0.313 0.016 -0.078 -0.030 (0.600) (2.59) (4.90) (32.99) (21.94) Prestrike hours 10.78 -5.14 0.101 0.007 0.020 (4.17) (1.42) (1.38) (0.142) (0.657)

Fraction female -0.272 -3.43 -0.577 -0.112 -0.113 -0.069 (0.908) (2.75) (0.327) (3.62) (4.99) (5.36) Number of strikes in state- 0.023 1.98 -1.64 -0.063 -0.026 -0.030 year-industry cell (00s) (0.249) (2.13) (1.24) (6.51) (3.33) (7.60) Observations 12,829 4,147 4,269 12,829 5,352 12,829 R?2 0.037 0.123 0.048 0.057 0.260 0.090 Mean of dependent variable 0.596 2.32 -1.58 0.417 0.245 0.102 F for laws 0.504 6.69 0.486 4.05 1.56 1.60 (p-value) (0.774) (0.0002) (0.692) (0.001) (0.169) (0.157) Notes: t-statistics are in parentheses. All regressions also include state, year, city, and industry dummies as well as state-specific time trends. Columns 2 and 3 use 1881

through 1886 only. The dependent variable in column 4 is binary (1 = replacements used, 0 = no replacements used).

This content downloaded from 139.182.97.89 on Wed, 02 Nov 2016 23:02:52 UTC All use subject to http://about.jstor.org/terms

The Law and Labor Strife 63

laws between 1881 and 1886, and the injunction was not used in this period, so we excluded these law variables from the analysis of changes in wages and changes in hours, which we observe only between 1881 and 1886.

Recall that strike theories that rely on asymmetric information predict that factors that raise the cost of any given strike will also be associated with poorer outcomes for workers, if it is primarily the workers who lack information. Conversely, factors that reduce the costs of any given strike should be associated with better worker outcomes. The results in column 2 of Table 7 support the theory: we find that maximum hours laws, which reduced strike lengths and days lost, were associated with a wage increase of 6 percent after strikes. Maximum hours laws were also linked to reductions in the probability that strikebreakers were used, without any change in the fraction replaced.

On the other hand, laws banning blacklists, which increased strike lengths, were associated with increases in the use of strikebreakers of 6 percent without any change in the fraction replaced conditional on re- placement. The net effect was a 2 percent increase in the probability of being replaced. Laws banning intimidation and boycotts also had a negative impact on labor outcomes following strikes: these laws were associated with a marginally significant increase in the mean fraction of workers replaced by strikebreakers.

Laws legalizing unions had no significant effect on strike lengths or days lost, and are estimated to have had only weak effects on strike outcomes. For example, laws legalizing unions increased the probability that strikebreakers were used, but reduced the fraction replaced conditional on strikebreakers being used. Hence, on balance, as column 6 shows, the legalization of unions had no significant effect on a striking worker's overall probability of being replaced.

It is remarkable that the existence of maximum hours laws for some groups of male workers had such significant effects on strike outcomes, since, as discussed above, many of these laws appear to have been relatively toothless. However, Claudia Goldin found that the passage of maximum hours laws for women also reduced hours for men.29 She attributes some of this effect to a general sentiment among workers in favor of shorter hours. It is possible as James Heckman and Brooks Paynor have suggested, that the law affects labor markets not so much through enforcement as through the establishment of new social norms.30

Finally, in view of the importance that has been attached to the use of the in-

junction by authors such as Hattam, it is surprising that we do not find a sig-

29 Goldin, "Maximum Hours Legislation." 30 Heckman and Paynor ("Determiing the Impact") make this argument with regard to the effects

of Civil Rights legislation on the employment of blacks in southern textile mills.

This content downloaded from 139.182.97.89 on Wed, 02 Nov 2016 23:02:52 UTC All use subject to http://about.jstor.org/terms

64 Currie and Ferrie

nificant injunction effect on strike outcomes.3" Obviously, injunctions were used to alter strike outcomes in specific cases. But our results suggest that the mere threat that an injunction could be imposed had little

effect on strike outcomes, though it increased strike lengths and the number of days lost.

CONCLUSIONS

One might expect collective bargaining laws favored by unions to provide unions an advantage in labor disputes, and those favored by employers to

give the advantage to employers. However, strike theory suggests that we need to look more closely at whether laws reduce either the cost of conducting a given strike, or uncertainty about the "state of the world" and likely strike outcomes. Legal changes that affect one side's bargaining strength but not the cost of conducting strikes or uncertainty may well have important effects on collective bargaining outcomes more generally, but there is no reason to expect them to affect either strike incidence or strike outcomes. Therefore, an important caveat to our work is that we cannot tell whether changes such as the legalization of unions improved labor's general collective bargaining position using our sample of strikes.

This article is intended not as a test of a specific strike theory (such as that conducted by David Card and Craig Olson) but as an exploration of the effects of legal structure on strike activity.32 Nevertheless, strike theory provides a useful lens for interpreting our findings. Most legal changes at this time, including the legalization of unions, had little effect on strike

incidence, strike lengths, days lost, or strike outcomes. This suggests that they also had little effect on uncertainty. Our most remarkable result is that maximum hours laws reduced strike lengths and days lost and improved strike outcomes for labor. This result suggests that they reduced uncer- tainty about the likely outcome of strikes once they had occurred. For example, the passage of maximum hours laws for some groups of workers may have made it more likely that other workers would be successful in striking for lower hours. A second interesting finding is that the use of the injunction increased strike lengths and days lost without giving either side an advantage in terms of strike outcomes. An obvious interpretation is that the possibility of legal intervention raised the uncertainty associated with

strike outcomes, making it more difficult for parties to a dispute to settle their differences.

31 Hattam, Labor Visions.

32 Card and Olson, "Bargaining Power."

This content downloaded from 139.182.97.89 on Wed, 02 Nov 2016 23:02:52 UTC All use subject to http://about.jstor.org/terms

The Law and Labor Strife 65

REFERENCES

Alston, Lee J. "Farm Foreclosures in the United States During the Interwar Period." This

JOURNAL 53, no. 4 (1983): 885-903. Angrist, Joshua, and Alan Krueger. "Does Compulsory School Attendance Affect

Schooling and Earnings?" Quarterly Journal of Economics 106, no. 4 (1991): 979-1014.

Atack, Jeremy, and Fred Bateman. State Samplefrom the 1880 Census of Manufacturing. ICPSR 9384. Ann Arbor, MI: ICPSR study no. 1990.

. "Whom Did Protective Legislation Protect? Evidence From 1880." NBER Working Papers on Historical Factors in Long Run Growth No.33 (December 1991).

Bailey, Gary L. "The Commissioner of Labor's Strikes andLockouts: A Cautionary Note." Labor History 32, no. 3 (1991): 432-40.

Card, David, and Craig Olson. "Bargaining Power, Strike Duration, and Wage Outcomes: An Analysis of Strikes in the 1880s." Journal ofLabor Economics 13, no. 1 (January 1995): 32-61.

Commons, John R., David J. Saposs, Helen L. Sumner, E. B. Mittelman, H. E. Hoagland, John B. Andrews, and Selig Perlman. History of Labour in the United States. New York: Macmillan, 1918-1935.

Currie, Janet, and Sheena McConnell. "The Impact of Collective Bargaining Legislation on Disputes in the U.S. Public Sector: No Policy May Be the Worst Policy." National Bureau of Economic Research Working Paper No. 3978, (1992).

Edwards, P. K. Strikes in the United States, 1881-1974. Oxford: Basil Blackwell, 1981. Frankfurter, Felix, and Nathan Greene. The LaborInjunction. NewYork: Macmillan, 1930. Freeman, Richard. "Contraction and Expansion: The Divergence of Private Sector and

Public Sector Unionism in the United States." Journal of Economic Perspectives 2, no. 2 (1988): 63-88.

Friedman, Gerald. "Politics and Unions." Ph.D. thesis, Harvard University, 1986. Friedman, Lawrence M.A History ofAmerican Law. New York: Simon and Schuster, 1973. Goldin, Claudia D. "Maximum Hours Legislation and Female Employment in the 1920s:

A Reassessment." Journal of Political Economy 96, no. 1 (1988): 189-205. Griffin, J. I. Strikes: A Study in Quantitative Economics. New York: Columbia University

Press, 1939.

Hattam, Victoria. Labor Visions and State Power. Princeton, NJ: Princeton University Press, 1993.

Heckman, James, and Brooks Paynor. "Determining the Impact ofFederal Anti-discrimina- tion Policy on the Economic Status of Blacks: A Study of South Carolina." American Economic Review 79, no. 1 (1989): 138-77.

Hicks, John. The Theory of Wages. Gloucester, MA: P. Smith, 1957. Kennan, John. "The Economics of Strikes." In Handbook of Labor Economics, Vol. 2,

edited by Orley Ashenfelter and Richard Layard, pp. 1091-137. New York: North- Holland, 1986.

Landes, Elisabeth M. "The Effect of State Maximum-Hours Laws on the Employment of Women in 1920." Journal of Political Economy 88, no. 3 (1980): 476-94.

Lebergott, Stanley. "The American Labor Force." In American Economic Growth, edited by Lance Davis et al., pp. 184-229. New York: Harper and Row, 1972.

Libecap, Gary D. "Economic Variables and the Development of the Law: The Case of Western Mineral Rights." This JOURNAL 38, no. 2 (1978): 338-62.

. "Bureaucratic Opposition to the Assignment of Property Rights: Overgrazing on the Western Range." This JOURNAL 41, no. 1 (1981): 151-58.

Margo, Robert, and T. Aldrich Finegan. "Compulsory Schooling Legislation and School

This content downloaded from 139.182.97.89 on Wed, 02 Nov 2016 23:02:52 UTC All use subject to http://about.jstor.org/terms

66 Currie and Ferrie

Attendance in Turn of the Century America." NBER Working Papers on Historical Factors in Long Run Growth No. 89 (July 1996).

Oakes, Edwin Stacey. 7he Law of Organized Labor and Industrial Conflicts. Rochester, NY: The Lawyers Co-operative Publishing Company, 1927.

Petro, Sylvester. "Assumptions and Premises of National Labor Policy: 1,032 Points of Light on the Subject." Wake Forest Law Review 26, no. 4 (1991): 965-1184.

Rockoff, Hugh. "The Free Banking Era: A Re-Examination." Journal of Money, Credit, and Banking 6, no.2 (1974): 141-67.

Rosenbloom, Joshua L. "Strikebreaking and the Labor Market in the United States, 1881-1894." This JOURNAL 58, no. 1 (1998): 183-205.

Stimson, F. J. Handbook to the Labor Law of the United States. New York: Scribners, 1896.

U.S. Census Office. Eleventh Census of the United States, 1890: Vols. 11-13, Report on Manufacturing Industries in the United States at the Eleventh Census. Washington, DC: GPO, 1892-1897.

. Twelfth Census of the United States, 1900: Vols. 7-10, Manufacturers. Washing- ton, DC: GPO, 1902.

U.S. Commissioner of Labor. ThirdAnnual Report. Washington, DC: GPO, 1888. . Tenth Annual Report. Washington, DC: GPO, 1896.

Witte, Edwin E. "Early American Labor Cases." Yale Law Journal 35, no. 7 (1926): 825-37.

This content downloaded from 139.182.97.89 on Wed, 02 Nov 2016 23:02:52 UTC All use subject to http://about.jstor.org/terms