Introduction
Concertation and conflict have long been alternate options in the relationship between labor and capital. What leads the labor movement to choose the option of concertation rather than that of conflict? Historically, the labor movement has tended to choose the option of concertation when the labor market suffered from high rates of unemployment. In other words, unemployment has been a catalyst inducing the labor movement to seek a compromise with capital [Scharpf Reference Scharpf Fritz1991; Korpi Reference Korpi2002]. Notwithstanding the intrinsic relationship between unemployment and the politics of tripartite relations, the issue of unemployment has not been sufficiently highlighted in studies that focus on the formative periods of the modern, democratic, and welfare state.
This study is concerned with the Swedish case during the interwar period. Our purpose is to answer the question of why and how the most conflict-driven Swedish society in terms of industrial relations [Korpi Reference Korpi1983: 172] eventually came to pursue the institutionalization of industrial peace. Theoretical interpretations regarding this issue abound. Power resources theory (PRT) and the theory of cross-class alliance (TCA), two major theories in this field, provide opposite interpretations regarding the role of labor and capital. Despite on-going debates in comparative political economy [Ibsen and Thelen Reference Ibsen and Thelen2017; Iversen and Sokice Reference Iversen and Soskice2015; Iversen and Stephens Reference Iversen and Stephens2008; Korpi Reference Korpi2006; Swenson Reference Swenson2004; Thelen Reference Thelen2012], these theories have in common that they employ society-centric approaches and explain policymaking in terms of politics between labor and capital. In contrast, inspired by the theoretical perspective that “new policies create a new politics” [Schattschneider Reference Schattschneider1935: 288], this study explores how the implementation of unemployment policies by the state affected the relationship between labor and capital.
In Sweden, unemployment became entrenched in the labor market after the economic depression of 1920, reaching more than 10% until the mid-1930s. The Swedish state launched large-scale unemployment policies throughout the 1920s. After 1920, strike activities took a downturn. Strike activities, a primary cause of social unrest at the time, were positively related to unemployment before the economic depression of 1920. Thereafter, however, the strength of the relationship decreased substantially. The LO, in fact, began to probe for a sphere of concertation with the Swedish Employers’ Association (SAF). Why did the labor movement change its strategy of collective action after the economic depression of 1920? Was this transformation in the relationship affected by state intervention in the labor market? If so, in what ways did state intervention bring about the change in the labor movement? And what theoretical implications can we derive from analysis of these events?
We conduct statistical analysis and then show through historical interpretations that the results of statistical analysis represent actual relations. As Huber and Stephens suggest, we seek to “identify robust patterns of association” by statistical analysis and then to “examine historical evidence to establish causal sequences” [Huber and Stephens Reference Huber and Stephens2001: 8]. We employ a panel data method to analyze the strike activities of 21 union federations (förbund) covering the period 1916-1938. It was after 1916, the year in which the First World War almost came to an end, that industrial conflicts were once again aggravated, concluding the short period of labor’s quiescence after the general strike of 1909. However, industrial conflicts began to decrease after the economic depression of 1920. First, we test the existing theories that have claimed to explain the fluctuation of labor strikes. Second, we analyze the historical “turning point” after which the turbulence of labor strikes began to subside [Abbott Reference Abbott1997], the extent to which the downturn of labor strikes was related to unemployment, and the types of workers who exited from the arena of labor strikes as unemployment went up, and thus the state intervened in the labor market through unemployment policies.
The historical portion of this study explores the feedback effects that the implementation of unemployment policies entailed. We explicate the extent to which unemployment policies contributed to reinforcing the capacities of the state in the sphere of industrial relations. We also study how unemployment policies brought about the split of the union movement, plunging it into internal conflicts. In particular, our study focuses on the impacts that the state’s provision of relief work had on the union movement. The LO supported the low-wage policy of relief works after 1921. Thus, it was faced with strong opposition not only from the unemployed participating in relief work, but also from labor unions whose labor market competed with relief work. We argue that these new developments pushed the LO to look for an alternative to the regime of high unemployment in the labor market. After the economic depression of 1920, in fact, the LO began to positively evaluate the issue of the rationalization of production, an issue to which it had vehemently objected since the 19th century.
We draw the following theoretical conclusions: it was neither the empowerment of the Social Democratic Workers’ Party (SAP) in the 1930s nor the launching of aggressive lockouts by employers in the 1920s that led the LO to have an interest in the establishment of a concertative relation with the SAF. Rather, the turnaround was provoked by state intervention in the labor market and the subsequent internal conflicts of the union movement in the 1920s. The empowerment of the SAP after the 1932 election played the role of reinforcing the concertative efforts that the LO had attempted in the 1920s. The Swedish case may well be a case in point to trace the processes in which the implementation of a policy provokes a reactive sequence endogenously within a historical path, thus changing the contexts of politics that had presided over the making of the very policy.
Policy and Politics: an Alternative Theoretical Perspective
Depending on which agent of class conflict the analysis is focused on, existing studies of class compromise can be broken down into PRT and TCA. The PRT adopts the organizational and/or political power of labor as a crucial explanatory variable affecting strike activities, whereas the TCA emphasizes the coordinating role of capital. Thus, from opposite points of view, the two theories interpret the development of Swedish industrial relations. According to the PRT, after the victory of the SAP in the 1932 elections, the union movement had an interest in avoiding costly industrial conflicts and employing “political instruments to influence the distributive processes” [Korpi Reference Korpi1983: 173; Korpi Reference Korpi1978; Korpi Reference Korpi2006]. In contrast, the TCA focuses on the following historical facts: employers in exposed industries took the lead not only in institutionalizing centralized wage bargaining but also in launching lockouts to subdue labor strikes. The SAF even adopted a strategy of threatening its “ally” LO with multi-industry lockouts in order to press the latter to “intervene against” labor unions in sheltered industries [Swenson Reference Swenson1991; Swenson Reference Swenson2002: 90-92, 101-102]. Consequently, the TCA argues, what made the turbulence of labor strikes subside was not the empowerment of the labor movement, but employers’ aggressive lockout strategies.
After the economic depression of 1920, the Swedish state intervened massively in the labor market with its unemployment policies. The Swedish labor market was hit unprecedentedly hard. The unemployment rate skyrocketed to 26.5% in 1921, reached 34% in January 1922, and did not drop below 10% during most of the period between 1921 and the mid-1930s [Kungl. Arbetsmarknadsstyrelsen Reference Kungl1974: 5]. We may posit an alternative hypothesis: the implementation by the state of unemployment policies had impacts on the relationship between the LO and the SAF and thus on strike activities.
We approach this issue on the basis of Schattschneider’s [1935: 288] thesis that “new policies create a new politics” [Pierson Reference Pierson1993]. Theoretically, Schattschneider’s thesis paves the way for employing public policy as an independent variable of politics. Public policy has traditionally been treated as a dependent variable to be explained in terms of politics [e.g., Heclo Reference Heclo1974]. However, the state can “create” or “destroy” interests through policies [Schattschneider Reference Schattschneider1935: 288].
Policy and politics interact with each other dynamically. Public policies affect the formation and development of the state. This is because the capacities of the state are transformed or expanded while new public policies are implemented [Skocpol Reference Skocpol1992: 58]. However, the implementation of the public policy entails feedback effects which, in turn, may bring about changes in the preferences, interests, and meanings of actors as well as in the capacities of the organizations involved. Then, these changes conversely affect the contexts of politics in which the public policy had been made [Pierson Reference Pierson1993; Pierson Reference Pierson2000]. A public policy does not always produce positive feedback. It also inflicts negative feedback on some actors, provoking backlashes and thus setting “a reactive sequence” in motion [Mahoney Reference Mahoney2000: 526].
As will be explored in the historical portion of our study, while implementing unemployment policies, the state centralized its system of intervention in the labor market, enlarged its bureaucratic realm of domination, and tightened its regulations against labor strikes. Furthermore, the state’s provision of relief work provoked intra-organizational conflicts within the union movement. Due to its support for the low-wage policy of relief works after 1921, the LO was faced with strong oppositions not only from the unemployed who were participating in relief work, but also from labor unions whose labor market competed with relief work. Consequently, we argue, these unprecedented developments pushed the LO to probe for a sphere of concertation with the SAF.
Changed patterns of workers’ collective action
Theorists of power resources argue that labor strikes substantially reduced after the formation of the social democratic government in 1932 and the “red-green” coalition in 1933. However, empirical evidences strongly indicate that the pattern of collective action by Swedish workers had already undergone a transformation after 1920. If we do not take into account the period 1910-1915, as Figure 1 shows, the trend of strike activities took a downturn after 1920. The period 1910-1915 was exceptional in that the union movement had yet to recover from the aftermath of the general strike of 1909 and had to be defensive due to the impacts of the First World War. By 1916, the Swedish union movement had recuperated organizational strength and thus resumed its offensive strategies against employers [LO 1943: 409; Kjellberg Reference Kjellberg1983]. The numbers of participants in labor strikes fluctuated significantly after 1920. However, if we deduct lockout-provoked labor strikes, the numbers of participants in labor strikes decreased gradually after 1920 (Figure 2) [Fulcher Reference Fulcher1987; Hamark Reference Hamark2018; Molinder, Karlsson and Enflo Reference Molinder, Karlsson and Enflo2019; Swenson Reference Swenson1991].
Figure 2 Participants in labor strikes and lockouts, 1916–1938
Source: Appendix 1.
Based on Zivot-Andrews unit root test, we can also estimate a structural break-year in the trend of strike activities. Since we seek to identify a structural break year in the trend of strike activities, we employ “the model B” among Zivot and Andrews’s three models. It allows for one endogenously determined structural break in the slope of the trend function [Zivot and Andrews Reference Zivot and Andrews1992]. Based on t-statistic at the year 1921, we reject the null of unit root for the variable, i.e., labor strikes excluding lockouts at the 1% significance level. The t-statistic at 1921 is –6.384 whereas the critical value for 1% for Zivot-Andrew test is –4.94. Figure 3 shows graphically that the largest t-statistic in absolute value occurs at the year 1921 following the economic depression of 1920.
Figure 3 Zivot-Andrews test for labor strikes excluding lockouts, 1919–1935
Note: Zivot-Andrews test avoids testing for breaks near the beginning and the end of a time-series, i.e., 1916-1918 and 1936-1938, because standard errors for breaks are comparatively large. See Baumann and Schap, Reference Baumann and Schap2015.
The fact that Swedish workers’ collective action underwent a transformation after 1920 becomes more apparent when we trace the trajectory of labor strikes in combination with workers’ electoral turnout. Universal male suffrage was legislated in 1908 and applied to the 1911 election. Between the 1914 and the 1920 elections, as Figure 4 shows, workers’ electoral turnout decreased successively whereas the numbers of participants in labor strikes increased continually. However, this trend was reversed after the 1920 elections, and continued until 1936.
Figure 4 Workers’ participation in labor strikes and electoral turnout, 1893–1940
Note: Workers’ (class III) election turnouts are estimated ones (Lewin, Jansson and Sörbom Reference Lewin, Jansson and Sörbom1972: 66-68, 142–143). The numbers of participants in labor strikes are annual averages between elections to the Second Chamber of the Parliament. For the sources of data on labor strikes, see Appendix 1.
The ideological positions of the governing parties were not likely to affect this trend reversal. The SAP formed a coalition government with the Liberal Party in 1917 and a single-party government in 1932. Between 1917 and 1932, there emerged such diverse governments as the conservative government (April 1923-October 1924 and October 1928-June 1930), the liberal coalition government (June 1926-October 1928), the liberal government (June 1930-September 1932), and the social democratic minority government (March 1920-October 1920, October 1921-April 1923, and October 1924-June 1926) [Hadenius, Molin and Wieslander Reference Hadenius, Molin and Wieslander1988: 350-354]. It is hardly plausible that the ideological position of any particular government affected the continuity in the changed pattern of workers’ collective action. Rather, Swedish workers gradually and successively shifted the locus of their involvement in collective action from the economic sphere to the political sphere.
Research Design
Panel construction and dependent variables
In our statistical analyses, we take union federation as the unit of analysis. Union federations prevailed in the process of deciding whether to go on strike until the LO Congress decided to reinforce the role of the LO in 1941 [Höglund Reference Höglund1979].Footnote 1 About 85 union federations appeared in the 20th century. However, many of these were merged or integrated with other union federations. In our analysis, we include 21 union federations affiliated with the LO during the period 1916-1938 (see Appendix 2).Footnote 2
We adopt relative involvement in labor strikes (RILS)—number of participants in labor strikes/number of union members x 1,000—as our dependent variable. In contrast to frequency and volume (the number of idle worker-days) of strikes, the number of workers involved in labor strikes reflects “the degree to which workers have been drawn into conflicts”, thereby encompassing “political and social importance” [Franzosi Reference Franzosi1995: 13; Korpi Reference Korpi1983: 161-162]. Since we seek to discuss changes in the social and political relations among the state, labor and capital, RILS may well be a valid variable. While the number of participants in labor strikes and lockouts fluctuated considerably (see figure 2), the RILS trend displays a gradual and continuous downturn after 1920 (figure 5).
Figure 5 RILS, 1916–1938
Sources: Appendix 1.
We break our dependent variable RILS into four sub-categories according to 1) whether to include lockout-provoked strikes, and 2) whether to include labor strikes by non-union workers (Table 1; for time-series plots of dependent variables by union federations, see Appendix 3). We seek to analyze the extent to which the relationships between independent variables and RILS change across the four sub-categorized models. We include strike activities by non-union workers in our statistical analysis with a caveat. In calculating RILS, we divide the number of non-union workers participating in labor strikes by the number of union members. For the case of non-union workers, there is no relevant denominator that could be used to compare the relative size of participation across union federations. However, the adoption of the denominator can be theoretically justified: we are theoretically interested in the extent to which union federations mobilized non-union workers into labor strikes.
Table 1 Categorization of dependent variables
Hypotheses and the operationalization of independent variables
Based on the theoretical discussions in the previous section, we posit two hypotheses to test for the applicability of PRT:
1) RILS declined after 1932, at which time the SAP formed a single-party government for the first time.
2) RILS is positively related to the organizational power of union federations.
We test the first hypothesis by employing SAP in the government dummy variable. To test the second hypothesis, we operationalize the organizational power of union federations in terms of costs of mobilization per member and capital assets per member. In so doing, we count capital assets at the end of the previous year because they were slated to be used to compensate participants in labor strikes for the loss of wages in the coming year (Table 2).
Table 2 Independent variables and operationalization
For TCA, we posit a hypothesis: RILS declined more in exposed industries than in sheltered industries. Regarding the role of lockouts in subduing strike activities, the TCA is composed of a twofold argument: on the one hand, it was employers’ lockouts that brought about a decrease in strike activities in the 1920s. Following this logic, strike activities should have subsided more in exposed industries than in sheltered industries. This is because employers in exposed industries, VF (the Engineering Employers’ Association) in particular, took the lead in launching lockouts. On the other hand, SAF’s strategy of threatening its ally the LO with multi-industry lockouts pressed the latter to “intervene against” labor unions in sheltered industries [Swenson Reference Swenson1991; Swenson Reference Swenson2002]. However, the SAF adopted this strategy only twice: for paper pulp trades in 1932, and for building and construction trades in 1933-1934 [Swenson Reference Swenson2002]. We employ a cross-class alliance dummy variable as a variable to test the hypothesis (Table 2; for classification of industries, see Appendix 2). First, we test whether the active role of employers in terms of lockouts did indeed affect decreases in strike activities for the entire period 1916-1938. Second, we first divide the entire period 1916-1938 into three sub-periods: 1) 1916-1920, 2) 1921-1932, and 3) 1933-1938. We then examine whether there were differences among the three periods in the extent to which employers’ lockouts affected strike activities.
For the theory of dynamics of policy and politics, we posit two hypotheses:
1) Starting in the year 1921, a downward shift took place in the level of RILS.
2) The relationship between the rate of unemployment and RILS was positive before the economic depression of 1920, whereas the strength of the relationship decreased when the state intervened in the labor market with its unemployment policies.
The first hypothesis implies that the year 1920 represents a historical turning point after which unemployment policies played a role as an institution, producing the feedback effects of a downward shift in strike activities. We employ a state intervention dummy variable to see whether there was a statistically significant break in strike activities starting in the year 1921 (see Table 2).
To test the second hypothesis, we compare the effects of unemployment on strike activities among the three sub-periods. First, we assume that before the economic depression of 1920, strike activities were positively related to unemployment. Swedish scholars provide detailed historical studies regarding the formation of the working class in Sweden in the 19th and early 20th centuries.Footnote 3 Furthermore, in the aftermath of the Bolshevik Revolution of 1917, Swedish society was engulfed in socio-political turmoil characterized by a fully-fledged revolutionary atmosphere [Tingsten Reference Tingsten1941]. Following Polanyi [Reference Polanyi1944], we assume that, in these social contexts, unemployment was not only an individual but also a social concern, thus provoking a collective reaction against employers—i.e., labor strikes. The Swedish state had yet to become sufficiently modern and bureaucratic to discipline working-class individuals.
Second, we assume that after the economic depression of 1920, unemployment policies pressed workers not to join labor strikes. On the one hand, the state furnished a loophole, that is, relief work, for unemployed people to exit from the union movement. On the other hand, by limiting the eligibility for unemployment relief strictly to the unemployed who had not been involved in labor strikes, the state sought to exert influence on those who remained in the labor market.
Third, however, we assume that, as the extent to which the state intervened in the labor market with unemployment policies reduced after 1933, the strength of the relationship between unemployment and strike activities once again increased. The regime of unemployment policies underwent a transformation after 1933 because the first single social democratic government revamped it fundamentally. Major resources were increasingly earmarked for public work rather than for relief work. Public work was fundamentally different from relief work in that the former was designed to prevent economic depression. Furthermore, the SAP succeeded in abolishing the conflict directive in 1933. In 1934, the SAP also accomplished the enactment of the Ghent system of unemployment insurance which the LO had regarded as an alternative to the existing unemployment policies. In a nutshell, the regime of unemployment policies no longer worked as in the past.
Finally, we control for the change of real wages. The economic models of strike activities, which are beyond our theoretical interests, argue that changes in real wages in the previous year affect strike activities [Ashenfelter and Johnson Reference Ashenfelter and Johnson1969]. Unfortunately, data on real wages by union federation are not available. However, we have the index of annual real wages (1913=100) covering all industries. We give the same value—i.e., the percentage rate of change in the previous year (Δwt-1)—to each of the union federations for each year (see Table 2). Therefore, Δwt-1 does not reflect cross-sectional variations.
Data description
Categorization of RILS into four sub-categories is made possible by LO’s provision of diverse and detailed data. The LO published its annual report based on the data submitted by union federations. These data cover the types of labor strikes, the types and numbers of workers involved, and the unemployment rates of union members. The reports by union federations also detailed the capital assets of union federations and their costs of organizational activities spent on agitation, union newspapers and education. The LO aggregated those data on the organizational activities of mobilization and published them in a report in 1932 and in 1942 (see Appendix 1).
Unemployment rate data are in need of a more detailed description. After 1911, union federations submitted a monthly report to the LO regarding the rates of unemployment based on surveys of their members. Data on the unemployment rates of non-union workers are not available. In-depth research sponsored by the Swedish government, however, argues that, although there was a slight tendency to exaggerate the unemployment of union workers relative to that of non-union workers, the data reported by each union federation came to represent the related industry as a whole as the surveys covered large numbers of union workers and a wide scope of the industry [SOU 1931 (20): 44-61].
Methods
We employ the random effects model (REM) among panel data methods. Except for the period 1933-1938, Hausman’s test results suggest that the random effects estimators are consistent. REM also makes it possible to estimate the observable, time-constant variable(s), i.e., the cross-class alliance dummy variable (dm3) in our model.
$$ \begin{array}{l}{Y}_{i,t}={\beta}_{\mathbf{1}}+{\beta}_{\mathbf{2}} dm{1}_t+{\beta}_{\mathbf{3}} dm{2}_t+{\beta}_{\mathbf{4}} dm{3}_i+{\beta}_{\mathbf{5}}\mathit{\ln}({unemp}_{i,t})+{\beta}_{\mathbf{6}}{cm}_{i,t}\\ {}+{\beta}_{\mathbf{7}}{ca}_{i,t-1}+{\beta}_{\mathbf{8}}{\boldsymbol{\Delta} \mathbf{w}}_{t-1}+{v}_{i,t}\end{array} $$
where vi,t = ci + ui,t. vi,t is composite error term, ci is unobserved effects, and ui,t is idiosyncratic errors. i denotes cross section observation, i.e., union confederations, and t denotes time.
For the period 1933-1938, we employ the fixed effects model (FEM) because the p-values of Hausman’s test are close to 0, which is “interpreted as evidence against” the assumption of REM [Wooldridge Reference Wooldridge2002: 288]. With FEM, however, it is not possible to estimate the observable, time-constant variable(s) such as gender or race, because they get eliminated in the process of fixed effects transformation [Ibid. 2002: 266-267]. In addition, we employ a hybrid model in which we can estimate the observable, time-constant variable. The hybrid model decomposes variables into “between variation and within variation,” thereby estimating both within-effects (i.e., fixed effects) and between-effects in REM [Schunck Reference Schunck2013: 66].
Our analyses are conducted in two stages. First, we include two year dummy variables—state intervention dummy (dm1) and SAP in the government (dm2)—to estimate their effects on RILS during the entire period 1916-1938. We take the log of the variable unemployment rates (ln(unempi,t)) to improve the linear model. To take the problem of panel heteroscedasticity into account, we use robust standard errors in addition to standard errors. Second, in order to estimate the effects of unemployment and lockouts on RILS in different socio-political contexts, we divide the entire period into three sub-periods: 1) 1916-1920, 2) 1921-1932, and 3) 1933-1938. Thus, the two year dummy variables, i.e., dm1 and dm2, are eliminated.
Empirical Results
Our panel data analysis supports what we hypothesize based on the theory of dynamics of policy and politics (Table 3). The coefficients on the state intervention dummy (dm1) are negative and large, and all of them are statistically significant at the 1% level. Use of the robust standard error does not change the statistical significance. It implies that there was a downward shift in the level of strike activities starting in 1921.
Table 3 Regression results for RILS (REM and Hybrid Model), 1916–1938
Note 1): Between–effects are omitted because of collinearity.
Note 2): OLS standard errors are in parentheses, and robust standard errors are in brackets. Lockouts include sympathy lockouts. Hausman test scores are based on classical standard errors. Since the fixed effect model eliminates the cross–class dummy variable (dm3), we do not include it when conducting Hausman test. Number of obs = 481; Number of groups = 21. *** p <0.01; **p <0.05; *p <0.1.
Unemployment was positively associated with RILS before the economic depression of 1920, but the strength of association changed thereafter. For the entire period 1916-1938, as Table 3 shows, the coefficients of (logged) unemployment rates (ln(unemp)) are positive and statistically significant at the 1% level, and at the 5% level if we use robust standard errors. However, the coefficients of (logged) unemployment rates differed not only among the three sub-periods but also between groups. We may therefore need a threefold comparison. First, as Table 4–1 shows, during the period 1916-1920, the coefficients are conspicuously different depending upon whether non-union workers are included—(1-1) and (2-1) vs. (3-1) and (4-1). When non-union workers are included, the estimated coefficients are relatively larger. However, if we use robust standard errors, the levels of significance are slightly larger than the conventional level of significance. If we compare the coefficient in the model (1-1) with that in the model (3-1), the difference is approximately 58, which means that a 1% increase in the rate of unemployment rate is associated with a 0.58 increase in RILS. This implies that, before the state intervened in the labor market, unemployment was a “social concern” in the Polanyian sense, provoking collective reaction against employers. Furthermore, as the rate of unemployment rose, non-union workers participated more actively in labor strikes than union workers did.
Table 4–1 Regression results for RILS (REM), 1916-1932
Note 1): If we employ hybrid model and estimate dm3, there is no noticeable change in the case of model (3-1). In the case of model (3-2), however, the coefficient is reduced to 24.924, and p-value increases to 0.110.
Note 2): See Note 2) at Table 3. Number of observations is 104 for the period 1916-20, and 251 for the period 1921-32. Number of groups is 21 for the two periods, respectively. *** p <0.01; **p <0.05; *p <0.1.
Second, in the period 1921-1932, the coefficients (ln(unemp)) are positive and statistically significant, but the differences between groups are not large relative to the previous period. In comparison with the previous period, the coefficients are considerably reduced when RILS includes non-union workers—(1-1) and (2-1) vs. (1-2) and (2-2) at Table 4–1. It implies that, as the rate of unemployment rose, non-union workers were more likely to exit from the labor market and thus the sphere of labor conflicts. Whether lock-out provoked strikes are included in the dependent variables does not make a conspicuous difference in the coefficients.
Finally, in the period 1933-1938, the social democratic dominance in government brought about a transformation in the regime of unemployment policies. The state could not intervene in the labor market as before, because relief work was increasingly replaced by public work. Statistically, based on the p-values of Hausman’s test, we can infer that FEM estimators are consistent (Table 4–2). If RILS excludes lockout-provoked labor strikes, the coefficients (ln(unemp)) are statistically significant not only in FEM but also in REM. Furthermore, the estimated coefficients increase to a large extent relative to the previous period 1921-1932, as the model (2-3) and the model (4-3) show. This empirical finding supports our historical interpretation in the next section: after the economic depression of 1931, the LO raised the issue of unemployment strongly, thus ceasing the concertation with the SAF for industrial peace that the SAF and the LO had promoted in collaboration after 1928.
Table 4–2 Regression results for RILS (FEM, REM and Hybrid Model), 1933-1938
Note 1): In the estimation of percentage change in real wage (Δwt-1), between-effects are omitted because of collinearity.
Note 2): See Note 2) at Table 3. Number of observations is 126. Number of groups is 21
According to TCA, strike activities should have decreased in exposed industries. However, the decline in strike activities after 1920 proceeded to the contrary. In the period 1921-1932, if RILS includes lockout-provoked labor strikes, the coefficients of the cross-class dummy (dm3) are positive rather than negative—model (1-2) and model (3-2) at Table 4-1. If we bring the previous period into analysis, this relationship is valid only in union workers’ labor strikes—model (3-1) and model (3-2) (see also Table 3). It implies that employers in exposed industries launched lockouts more frequently, and union workers fought against the lockouts. Furthermore, in the period 1933-1938, all the coefficients are not statistically significant regardless of whether we employ the REM or the hybrid model (Table 4–2). At best, we argue, Swenson’s case study on the building and construction trades cannot be generalized. Our interpretation is also corroborated by time-series plots of RILS. If we include lockout-provoked labor strikes, there is a noticeable difference between domestic industries and exposed industries (figure 6-1). However, when they are excluded, this apparent difference disappears particularly after the mid-1920s (figure 6-2).
Figure 6 Changes in RILS: union workers
By and large, the regression results support PRT. The coefficients on the SAP in the government dummy variable (dm2) are negative and statistically significant (Table 3). As far as the variable capital assets (cai,t-1) is concerned, the coefficients are positive and statistically significant (Table 3). However, depending on the sub-periods and on the sub-categories, the level of statistical significance varies. It was only in the period 1933-1938 when the SAP was in government that the coefficients turn out to be positive and statistically significant at the 1% level across the sub-categories. As for the cost of mobilization (cm), it is noteworthy that the coefficients are negative and statistically significant at the 1% level in the period 1933-1938 when lockout-provoked labor strikes are included—model (1-3) and model (3-3) at Table 4–2. That is, union federations succeeded in persuading workers not to react against lockouts through labor strikes when the SAP was in government. Consequently, we may argue that PRT is valid. Nonetheless, PRT suffers from a theoretical shortcoming: it does not contain a theoretical construct explaining why workers’ involvement in labor strikes began to decline after 1920.
Historical Interpretations
The economic depression of 1920 and the empowerment of state capacities
Beginning with the economic depression of 1920, the National Unemployment Commission (AK: Statens Arbetslöshetskommission) implemented large-scale unemployment policies, which lasted throughout the 1920s.Footnote 4 Implementation of unemployment policies entailed feedback effects that substantially reinforced the capacities of the state in its relationship with the union movement.
First, the state intervened more directly in the labor market. That is, the AK increasingly carried out projects of relief work rather than providing cash relief. In 1921, the state founded two large-scale organizations for relief work, which were under the direct control of the AK: the State Work of Southern Sweden (Södra Sveriges Statsarbeten) and the State Work of Norrland (Norrlands Statsarbeten). Between 1921 and 1924, the two organizations provided 712 and 972 relief-work projects, respectively. Out of the total number of unemployed (80,271) who reported to the AK in 1921, 31.8% were employed in relief work projects, and 31.3% received cash relief. In 1922, the share of relief work rose to 46.1%, whereas the share of cash relief dropped to 9.7%. After 1923, the state no longer earmarked funds for cash relief [AK 1929: 356-358, 368-369; SOU 1936 (32): 52].
Second, for those participating in relief work projects, the AK constructed increasing numbers of “fixed workplaces” all over the country. In contrast to those workers in the so-called “day colonies” who could commute from their places of residence, those employed in the fixed workplaces were forced to be separated from their families and live in very humble conditions. Those who had to move to northern parts of Sweden underwent particular hardships [AK 1929: 394]. The share of the unemployed involved in relief via fixed workplaces out of the total number of unemployed in relief work increased from 21% in 1921 to 47% in 1922 and to 52% in 1923. The share rose to as high as 70% in 1924 [Ibid.: 362-363, 391]. Whether the unemployed were assigned to fixed workplaces or day colonies was at the discretion of public officials. To avoid being assigned to fixed workplaces, the unemployed tried to convince local unemployment committees and official doctors by any means possible that they were physically weak [Ibid.: 1929: 389].
Finally, in an effort to bring industrial conflicts under its control, the state took advantage of the system of public labor exchange. The numbers of unemployed registered at the offices of labor exchange increased substantially after the economic depression of 1920, from 278,826 in 1920 to 838,599 in 1922, and the figure did not drop below 450,000 throughout the 1920s [SOS 1960: 130]. The AK limited the applicants for unemployment relief to those who did not obtain an “appropriate job” at the office of labor exchange. Furthermore, in order to thwart labor strikes, the AK excluded from the program of unemployment relief the members of labor unions waging labor strikes and those who refused to accept the assigned relief work under conflict. In 1922, the AK even issued the so-called “conflict directive,” indicating that it would exclude from unemployment relief “all the workers in a trade or an industry” where a “general conflict”—a labor strike producing a nation-wide impact on a trade or an industry—occurred [Lindeberg Reference Lindeberg1968: 58; Unga Reference Unga1976: 91-92; Rothstein Reference Rothstein1992: 163; SOU 1936: 27-31]. The conflict directive exerted a profound impact on the union movement. This was because the AK tended to classify “partial conflicts” as general conflicts, thereby excluding large numbers of the unemployed from unemployment relief. The social democratic minority government resigned in 1923 and again in 1926 when it failed to relieve the regulations of the conflict directive in Parliament. The conflict directive remained valid until 1933 [SOU 1936 (32): 30; Unga Reference Unga1976: 93-95].
In terms of the theoretical interpretation of our historical analyses, the Swedish state grew to be a modern, bureaucratic state while implementing unemployment policies. The state enlarged its bureaucratic realm of domination in the labor market and tightened its regulations against labor strikes. Consequently, the state power network of, to use Foucault’s terminology, “surveillance” and “observation” penetrated deep into the lives of working-class individuals [Foucault Reference Foucault1979: 170-194].
Internal conflicts within the union movement
After 1921, Swedish social democrats withdrew their previous position that there should be no difference in wages between relief work and the open labor market. The SAP accepted AK’s policy of settling relief work wages at lower levels than the level of unskilled workers’ wages in the open labor market. The social democrats’ new position was in line with the policy of deflation that the social democratic minority government adopted in 1921 [Lewin Reference Lewin1970: 50-53; Lindeberg Reference Lindeberg1968: 18-24; Öhman Reference Öhman1970: 76; Unga Reference Unga1976: 55-58]. The social democratic government pursued a twofold goal. First, it sought to induce a reduction in wages in the open labor market. Second, it tried to provide relief work for as many unemployed as possible. The social democratic government placed great importance on workers’ solidarity but, at the same time, did not have sufficient funds to earmark for unemployment relief due to its policy of stabilization. Thus, it seemed reasonable for the SAP to accept the low wage policy of the AK [Lewin Reference Lewin1970: 52; Lindeberg Reference Lindeberg1968: 21; Unga Reference Unga1976: 83, 86-87].
The LO also supported AK’s low-wage policy, but differed from the SAP in its rationale. In a strict sense, the LO and the SAP had different organizational goals: whereas the SAP pursued the solidarity of the entire working class, the LO’s primary concern lay in protecting union members’ interests. What union leaders worried most during and after the economic depression of 1920 was that the many unemployed who remained in the labor market might accept increasingly lower wages, thereby aggravating “the competition of underbidding” (underbudskonkurrens). Thus, the LO wanted relief work projects to absorb as many unemployed as possible [Unga Reference Unga1976: 85-86, 126).
The low relief work wage policy of the LO and the SAP plunged the union movement into internal conflicts throughout the 1920s. Union federations whose labor markets were constricted due to the expansion of relief work protested strongly against the LO and the SAP. Among them were the Municipal Workers’ Federation, the Road and Water Construction Workers’ Federation, and the Unskilled and Factory Workers’ Federation. They were opposed to the joint relief work of the state and the municipalities (statskommunala arbete) in particular, which the SAP enacted in Parliament in 1922. After 1927, the three union federations lost their patience. The Municipal Workers’ Federation, in particular, intensified its protests against the policy of the LO and the SAP at various conferences [Lindeberg Reference Lindeberg1968: 311-318; SKF 1934: 14-20, 30-34; Unga Reference Unga1976: 127-130]. This was due to the fact that, in 1927, Parliament turned down the social democrats’ proposal for unemployment insurance based on the Ghent system. The union movement had regarded this as an alternative to the AK regime of unemployment [Edebalk Reference Edebalk Per1975: 125-135]. In 1928, to make matters worse, a joint survey by the LO and the three union federations revealed that relief work encroached greatly on the labor market. According to the survey, 21% of the joint relief work during the period 1923-1926 was work that should have belonged to the labor market [Unga Reference Unga1976: 140]. In fact, the LO persuaded union leaders not to blockade the locations of relief work until the outcomes of the survey were released [LO 1928a].
Search for concertation: the rationalization of production
Before the economic depression of 1920, the Swedish labor movement was adamant in its opposition to the rationalization of production. In contrast, employers had pursued such a rationalization since the 1890s, when Sweden underwent a rapid growth in industrialization. The rationalization of production was aimed at reducing the costs of production and enhancing productivity. From the labor movement’s perspective, however, this would not compensate hard labor with corresponding increases in wages but would give rise to unemployment, thereby aggravating capitalist exploitation of the working classes [Johansson Reference Johansson1989: 93-95].
After the economic depression of 1920, however, union leaders began to justify the rationalization of production. The discursive logic went as follows: the rationalization of production could enhance both the conditions of wages and employment, provided that it would ensure economic growth. In 1921, the LO’s meeting of representatives demanded that, at the initiative of the state, the union movement and employers take joint responsibility for rationalizing the systems of production [Johansson, Reference Johansson1989: 54]. The LO Congress of 1926 substantiated this issue in more concrete terms: labor’s living conditions could be improved only when the efforts of rationalizing production led to a growth of production. Edvard Johanson, chairman of the LO, made it clear that the LO would not adopt the strategy of intensifying the struggle against employers simply to raise wages (LO 1926: 65-68; Johansson Reference Johansson1989: 56).
Union leaders did not support the rationalization of production merely to tackle the antinomy inherent in unemployment and wages. They also considered this as a way to break the deadlock between labor and capital. While debating the rationalization of production at a meeting of LO representatives in 1928, for example, Sigfrid Hansson, editor of Fackföreningsrörelsen (The Union Movement), urged representatives to “find the conditions of mutual understanding and to further the cooperation between employers and workers,” emphasizing that there must be “a common interest” between them [LO 1928b: 21-27]. In December 1928, the LO entered official talks with the SAF for the first time since the general strike of 1909. The conference of industrial peace confirmed that labor and capital had common interests in the rationalization of production and industrial peace. The conference also agreed to organize the Joint Delegation for Industrial Peace [Arbetsfredsdelegationen 1929: 4-5, 20-21; Johansson Reference Johansson1989: 85-86; Westerståhl Reference Westerståhl1945: 189-195].
The efforts toward cooperation for industrial peace were aborted when the economic depression of 1931 once again hit the labor market. Unexpectedly high rates of unemployment reaching at 16.8% in 1931 and 23.3% in 1932 hindered the rationalization of production from making headway. Talks between the LO and the SAF also ceased. Union leaders attributed unemployment to the rationalization of production, whereas employers revived the argument of classical economics, namely that union workers’ wages had been raised so high that economic depression was bound to occur [Lewin Reference Lewin1970: 46-47; Johansson Reference Johansson1989: 94-99; Öhman Reference Öhman1970: 56-61, 86]. To make matters worse, the Ådahlen accident in 1931, in which four workers and a girl were shot to death, aggravated and escalated the labor strikes, most of which were led by communists [Kennerström Reference Kennerström1971]. The LO Congress of 1931 decided to withdraw its representatives from the Joint Delegation for Industrial Peace, thus ceasing to collaborate with the SAF for the institutionalization of industrial peace [LO 1931: 215-216, 349-361; Johansson Reference Johansson1989: 103-116].
This decision by the LO is a counterfactual. It shows that, after the economic depression of 1920, the LO opted for the rationalization of production in order to end the stalemate in which the LO and the union movement were caught due to the AK’s unemployment regime. After 1920, as mentioned above, Swedish society continually suffered from unprecedentedly high and enduring rates of unemployment. The subsequent implementation by the state of unemployment policies entailed the following feedback effects: on the one hand, it empowered the state enough to discipline labor not to join labor strikes. On the other hand, the entire union movement was plunged into internal conflicts. The LO had to confront backlashes by the unemployed participating in relief work as well as by the employed union workers whose labor markets competed with relief work. In 1928, the LO sat across the table from the SAF and was in fact ready to accommodate the rationalization of production and industrial peace. However, in 1931, convinced that the rationalization of production had aggravated unemployment, the LO resolutely discarded the concept. Consequently, we may infer that without being caught in the AK unemployment regime after 1920, the LO would not have opted for the rationalization of production and looked for a sphere of concertation with the SAF.
After holding political power in 1932, the SAP revamped its unemployment policies entirely. In 1933, the social democratic state adopted ground-breaking policies including the fiscal policy of expansion, the abolition of the conflict directive, and the provision of public work. These policies were made possible by the “red-green coalition” between the SAP and the Farmers’ Party [SOU 1936 (32): 60-74; Söderpalm Reference Söderpalm and Koblik1975]. The social democratic government failed to abolish the AK due to opposition by the conservatives, placing emphasis on the provision of public work rather than work relief. The aim of public work resided in preventing economic depression in advance [AK 1937: 143; Clark Reference Clark1941: 95; Gustafsson Reference Gustafsson, Nilsson, Hildebrand and Öhngren1974:126-127]. Finally, in 1934, the SAP succeeded in enacting the Ghent system of unemployment insurance which the SAP and the LO had regarded as an alternative to the AK unemployment regime.
As unemployment rates dropped, the social democratic government again showed an interest in the rationalization of production and industrial peace. The social democratic state made aggressive efforts to bring turbulent industrial conflicts under control. In 1934, for example, the social democratic government threatened the Construction Wood Workers’ Federation, the Masons’ Federation, and the Unskilled and Factory Workers’ Federation: if they did not stop waging strikes, the state would have recourse to legislative methods, something that the union movement had long striven to avoid. Eventually, these labor strikes ended with LO’s mediation [Westerståhl Reference Westerståhl1945: 282-382]. In 1934, the state founded the Nothin Committee to investigate the conditions of industrial peace, in which the LO and SAF participated. In 1935, the Nothin Committee issued its report, emphasizing that the rationalization of production was a prerequisite for economic growth and thus people’s welfare [Johansson Reference Johansson1989: 132-135].
Despite the active role of the social democratic state, the LO and the SAF agreed in 1938 to build an industrial relation of “centralized self-regulation” and thus not to allow state intervention [Kjellberg Reference Kjellberg, Ferner and Hyman1998: 79]. Before March 1936, the LO seemed to expect that the state would take the initiative of promoting industrial peace. In fact, in 1935, the Nothin Committee released a report stating that state intervention in industrial conflicts might be unavoidable, although it would be desirable for partners in the labor market to promote industrial peace. At a secretarial meeting of the LO held in January 1936, Per Albin Hansson and Gustav Möller, Minister of Health and Social Affairs, expressed a similar view [LO 1936: § 4; Johansson Reference Johansson1989: 132-136]. However, the LO changed its position. On March 9, 1936, the SAF demanded that the LO accept the state’s exclusion from the negotiations for industrial peace. One week later, an extraordinary meeting of LO representatives concluded that the union movement should initiate an unconditional discussion with the SAF, and that the state should remain outside of those discussions. According to the LO’s Chairman, Albert Forslund, the LO needed to cooperate with the SAF because it was not yet able to “exert a decisive influence on the processes of legislation,” which implied that the LO could not but concede “inimically disposed decisions” in many cases [LO 1936: § 7; Johansson Reference Johansson1989: 136-137; Westerståhl Reference Westerståhl1945: 202-203]. In May 1936, the LO and the SAF founded the Committee of the Labor Market. The following process of negotiations lasted two years, ending with the formation of the Saltsjöbaden Basic Agreement in 1938.
Conclusion
Studies of political economy have yet to put an end to the debates on why and how concertative industrial relations and a certain type of welfare policy regime came to be institutionalized. Existing studies are divided between labor-centric approaches and capital-centric approaches. However, neither labor-centric nor capital centric-approaches focus on the influence that state intervention in the labor market brings to bear on the labor movement and thus industrial relations. In contrast, based on Schattschneider’s thesis, we analyzed the way in which the unemployment policies of the state affected the formation and development of the modern state, on the one hand, and the politics of tripartite relations, on the other, both of which were crucial for the democratic transition to the welfare state.
The year 1920 was a historical turning point in the development of industrial relations in Sweden. Starting in 1921, strike activities began to display a downward trend. In addition, the strength of the positive relationship between unemployment and labor strikes decreased substantially. These changes took place in the processes by which unemployment policies produced feedback effects, eventually transforming the socio-political contexts in which the union movement was embedded. On the one hand, the state power network penetrated deep into the lives of working-class individuals and gained firm control over industrial conflicts. On the other hand, unemployment policies plunged the entire union movement into internal conflicts. Confronted with these unprecedented situations, the LO looked for a sphere of cooperation with the SAF as a way of coping with the AK unemployment regime. It was in the 1920s—that is, a period much earlier than the 1930s—that the LO chose to enact such a turnaround.
Our interpretations are corroborated by the fact that, after the economic depression of 1920, the LO began to evaluate the rationalization of production positively, paving the way for the LO and the SAF to agree on industrial peace. Our interpretations are also attested by a counterfactual situation: the LO resolutely discarded the rationalization of production together with the issue of industrial peace in 1931 when it believed the rationalization of production had aggravated unemployment. Resumption of talks with the SAF had to wait until 1934 when the social democratic government successfully controlled unemployment through its fiscal policies. Furthermore, once the AK unemployment regime no longer worked effectively, the LO reacted aggressively to unemployment: if lockout-provoked labor strikes are excluded, the estimated coefficients of unemployment rates turn out to be large relative to those in the previous 1920s—–(2-3) and (4-3) at Table 4-2. From the vantage point of the mid-1930s, the Swedish social democratic government was to approach the issue of unemployment anew.
In Sweden, the state had already incorporated non-union workers in the realm of its political power in the 1920s while implementing unemployment policies. This occurred long before the SAP took hold of political power in 1932. The entire labor movement was not empowered during the interwar period. Under the AK unemployment regime, the union movement suffered from internal conflicts, while the state reinforced its capacities to control strike activities. Even after 1932, the LO had not yet shed its suspicion that state intervention in industrial relations might do the union movement harm. Thus, it made sense for the LO to strike a compromise with the SAF on “centralized self-regulation” in 1938, although this ended with an agreement that was more advantageous to the latter. In Sweden, this compromise was historical in the sense that the path of “centralized self-regulation” was not to be ruptured until the beginning of the 1970s [Johansson and Magnusson Reference Johansson and Magnusson1998: 18].
Appendix 1 Data sources
Labor strikes
LO, Berättelse över Landsorganisationens verksamhet (Stockholm: Arbetarnes Tryckeri), each year.
Union federations: lists and numbers of union members
D’Agostino H, 1987. Arbetarförbundens Medlemsutveckling i Sverige 1900–1985. Report for the Institutet för Social Forskning of Stockholm University, Meddelande 9.
Rates of unemployment
1916-1920: LO, Berättelse över Landsorganisationens Verksamhet (Stockholm: Arbetarnes Tryckeri), XXXVa-b, each year. Annual unemployment rates calculated on the basis of monthly data.
1921-1930: SOS, 1931. Arbetslösheten inom Fackförbunden (Stockholm: Norstedt & Sönner): 2.
1931-1938: K. Socialstyrelsen, Sociala Meddelanden (Stockholm: Norstedt & Sönner), Nr. 3, each year.
Mobilization costs and assets
LO, 1932. Sifferuppgifter och Grafiska Framställningar över Landsorganisationens och de Svenska Fackförbundens Ekonomiska Verksamhet Åren 1913-1930 (Stockholm: Tiden).
LO, 1942. Sifferuppgifter och Grafiska Framställningar över Landsorganisationens och de Svenska Fackförbundens Ekonomiska Verksamhet Åren 1931-1940 (Stockholm: Tiden).
Index of annual real wages
SOS, 1948. Lönestatistisk årsbok för Sverige (Stockholm: Allmänna Förlaget).
Appendix 2 Names of union federation and types of industry
Note: The trades belonging to commerce, construction, communication, and transportation were typical cases of the domestic industries, whereas industries such as engineering, transport equipment, pulp, paper, iron and steel, and wood products were Swedish export industries. It is, however, difficult to classify the textile, leather, and stonecutting industries. We classified them based on the extent to which exports contributed to the respective industry [See, Fridlizius Reference Fridlizius1963: 54-63; SOU 1931: 20, 212-222; SOS 1972: 260-293].
APPENDIX 3 Union workers’ RILS by union federation
Note: For the names of union federations, see Appendix 2.
