Abstract

This research combines information from the AFL-CIO Union Cities program with national survey data to examine whether politically active labor councils affected voter turnout in the 2000 national election. Results indicate that congressional districts with a Union City were associated with approximately 5 percent higher voter turnout. This effect, however, is mediated by preexisting political activity. Further analysis indicates that districts with a Union City have relatively higher rates of voting among minorities and the working class. These results support the general theory that organized labor is socializing labor-capital conflict.

In The Semisovereign People by E. E. Schattschneider (1960) proposes a theory on interest group behavior that applies to the contemporary union movement. The strategic response to intergroup conflict, according to Schattschneider, is essentially a choice over the scope of conflict. In most circumstances, when two or more groups are engaged in a confrontation the advantaged party will try to contain, or "privatize," intergroup conflict. Keeping the conflict private preserves power relations and, by extension, the hegemonic status of the advantaged. Weaker parties, on the other hand, will attempt to alter the balance of power by recruiting sympathetic outsiders in order to build a unified alliance against the dominant: a process Schattschneider refers to as "socializing" conflict. This dynamic of opposing tendencies toward the privatization and socialization of conflict can operate in reverse: formerly weak parties that become powerful tend to shed alliances to pursue an independent path.

In most local, national, and international contexts, labor has lost power relative to capital, and consistent with Schattschneider, the tactics of labor reflect a shift toward socializing conflict. The evolving rhetoric behind organizing drives, with the right to bargain collectively increasingly advocated for in universal terms, such as "justice" or "respect," is purposely framed to appeal to non-union organizations. Some unions are circumventing the legal conventions for achieving bargaining rights, rejecting NLRB supervised elections in favor of a card-check recognition process that often leverages local community support. Once workers are formally represented, pressuring employers to agree to contract terms is less a function of strike capacity and more dependent on corporate campaigns: a broad array of tactics that exert diffuse and multidirectional forms of pressure on industry leaders. As the recent phrase "social movement unionism" implies, these tactics are symptomatic of a strategic shift toward socializing conflict between labor and capital.

A Return to Political Militancy

One implication of the socialization of labor-capital conflict is that organized labor will become more politically active. This happens, in part, because the legal rules and economic policy that encumber new organizing and threaten existing members can be modified only through a political process. Pressure also comes from newly invited allies. Working coalitions entail interorganizational compacts that constrain the ability for one member to pursue policy that conflicts with interests of others in the coalition. Because the reform agenda of traditional and prospective non-union coalition partners, such as civil rights groups, interfaith councils, students, and environmentalists, are nearly always framed in political terms,¹ it follows that any sustained strategy to socialize conflict requires labor to become politically engaged and respect, if not adopt, the issues of non-union coalition partners. Without genuine reciprocity, coalitions are either short-lived or they fail to grow beyond mere words.

Union political expenditures do indicate a shift toward politics. Figure 1 provides inflation-adjusted election-cycle donations (in 2002 dollars) to political candidates for the 1990 through 2002 national elections.

According to these aggregate data, organized labor nearly doubled the magnitude of political contributions to candidates over the twelve-year period.

Source: Center for Responsive Politics. See: http://www.opensecrets.org.

This has occurred, by and large, without deviating from the "reward our friends, punish our enemies" formula of Samuel Gompers. The distribution of political contributions across the two major political parties was stable over the 1990 to 2002 time period, with anywhere from 93 percent to 96 percent going to Democrats. And as a group, the Democrats have been far more responsive to labor than Republicans. In 2000, for example, the average pro-labor voting record among Democrats was 87.6 percent, compared with an 8.6 percent average among Republicans.

This trend in resource appropriation is consistent with historical evidence of an inverse relationship between organized labor's ability to negotiate tangible gains at the bargaining table and labor's role as an agent for political insurgency. Perlman (1922) describes labor's embrace of egalitarian political reform during the nineteenth century—achieving suffrage for wage earners and public education—during a time when criminal conspiracy doctrines suppressed the growth of unions as economic organizations. Greene (1998) documents Samuel Gompers's reluctant immersion into partisan politics during the first decades of the twentieth century to counter the assault on AFL membership.

The CIO facilitated both economic agitation and widespread grassroots political mobilization during the 1930s, just prior to and during the greatest surge in U.S. union membership growth (Foster 1975).

Yet in 1947, with private sector unions nearing their peak postwar economic strength, labor failed to muster enough political support to prevent a congressional override of Truman's veto of the Taft-Hartley Act. In the Cold War era that followed, labor infamously purged leftist leaders from their ranks, opposed rank-and-file support of third-party candidates, was slow to join the Civil Rights movement, and alienated students by supporting the Vietnam War (Brody 1980; Meyer 1992; Rosswurm 1992; Zeiger 1986). Although there are notable exceptions to this pattern, a broad read of history indicates that when economic progress is achievable through bilateral bargaining, progressive alliances are disregarded and political militancy is suppressed. Coalition activity, progressive posturing, and political militancy ascend when bargaining fails.

Union Cities and Voter Turnout

Attempts to socialize labor-capital conflict are evident in "Union Cities," the AFL-CIO program to revitalize labor councils (LCs) as centers of regional political activity (Moberg 2000).² Union Cities was initiated after the Republican takeover of Congress in 1994, when it became doubtful that a block of reliable Democrats and a handful of labor Republicans could obstruct the passage of anti-labor legislation, let alone enact pro-labor measures (Dark 2000; Gerber 1999). Labor's political influence was waning, due in part to inattention to regional capacity building. To augment labor's pol itical presence in areas of high union density, the Union Cities program instructs LC leaders to participate in community coalitions, perform political outreach in working-class neighborhoods, diversify and train labor-friendly political leaders, and expand labor's voice through the media. The aim is to socialize the role of LCs by encouraging relations with labor-friendly constituencies, particularly targeting groups sympathetic to the interests of minorities and the working class.

We explore the socialization hypothesis by examining the association between Union Cities and voter turnout for the general population in the 2000 election. Although researchers have examined voter turnout of union members (Delaney, Masters, and Schwochau 1988; Sousa 1993; Zullo 2004), few have tested whether unions facilitate voter turnout for the general population. After controlling for state-level rates of urbanization, education, and income, Radcliff and Davis (2000, Tables 3 and 4) estimate that a 1 percent increase in union density is associated with approximately 0.20 to 0.25 percent higher vote turnout. In a complementary analysis, Radcliff (2001) models the propensity for U.S. citizens to vote as a function of national union density from 1952 to 1992. Factoring out whether a respondent is from a union household, and controlling for demographic factors, results indicate that union density is positively correlated with the probability of voting, leading to the conclusion that unions mobilize both their members and nonunion citizens.

Our analysis draws a more precise bead on this topic by evaluating voter turnout at the congressional district level. Consistent with prior research on unions and voting (Delaney et al. 1988; Radcliff 2001; Sousa 1993), voter turnout is analyzed using National Election Study (NES) data. Turnout is modeled as a function of the existence of a regional Union City LC, labor PACs, district turnout in 1992, and relevant controls. Results indicate a positive association between Union Cities and voter turnout, although this effect is mediated by preexisting regional political activity. To further explore the socialization hypothesis, we compare voting with respect to respondent characteristics in districts with a Union City and those without, focusing on race, class, and 1992 turnout.

Data and Variables

Data were compiled from four sources. The list of LCs designated as Union Cities were recognized as "Central labor councils committed to becoming a Union City" during the 2001 AFL-CIO convention. Using geographic information systems (GIS) software, the zip code for each Union City LC was matched against the boundaries for the 106th Congress. If a congressional district boundary crossed the zip-code boundary of a Union City LC, then it was assumed that the LC was politically active in that district during the 2000 election. Congressional districts with a Union City LC were coded 1; otherwise they were coded zero.

The Union City variable was merged with the 2000 NES data. The NES is a comprehensive biennial election survey conducted by the University of Michigan Center for Political Studies (Burns et al. 2001). The dependent variable (Voted) was derived from a question asking respondents whether they voted (v001241). Those responding "I'm sure I voted" were coded 1; otherwise they were coded zero.

Two consistent predictors of voting from the NES were used as control variables. First, age has a strong curvilinear association with voting, with voting rates lowest for the young, peaking at around seventy years, and then declining afterward (Miller and Shanks 1996; Rosenstone and Hansen 1993; Wolfinger and Rosenstone 1980). Age in years and age-squared are in all equations. Second, those who have a strong psychological attachment to a political party vote at higher rates than independents (Miller and Shanks 1996; Rosenstone and Hansen 1993). To capture the turnout effect related to partisan attachment, responses to the question "Do you think there are any important differences in what the Republicans and Democrats stand for?" are included (v001435). Affirmative responses are coded 1; otherwise they are coded zero.

Several researchers report lower turnout among working-class citizens (Shields and Goidel 1997; Teixeira 1987; Wolfinger and Rosenstone 1980). To explore the association between class and voting, a variable was operationalized from the NES based on a series of questions on class (v000998 to v001004). Respondents who self-identify as "average working class," "upper working class," or "working class" were coded 1; otherwise they were coded zero. To examine the relationship between race and voting, an indicator variable was included for respondents who describe their race as white (v001006).

Two final control variables are used in the analysis. First, to control for the intensity of the AFL-CIO political campaign by congressional district, total labor PAC donations to the candidates for the 2000 election cycle are included. These data originate from the Federal Election Commission and are compiled by the Center for Responsive Politics. Second, to control for the level of political activity by congressional district prior to the Union City Program, voter turnout was calculated for each congressional district in the 1992 election. Voter turnout by congressional district was estimated by dividing the sum of the votes for all 1992 presidential candidates (Barone and Ujifusa 1993) by the age-eligible population (over eighteen years old) in each district from the 1990 census. Variables and descriptive statistics are provided in Table 1.

Results

Union City and Voter Turnout

Table 2 provides the probit regression estimates for voting in the 2000 election. Model (1) includes control variables for age and party difference along with the primary variable Union City. The positive and statistically significant coefficient for Union City (β= 0.153; p < 0.05) indicates that respondents located in congressional districts with a Union City LC were approximately 5.4 percent more likely to vote in the 2000 election. However, the Union City coefficient becomes insignificant when the equation includes a control for the district turnout rates in 1992. Model (2) indicates that the 1992 turnout is a positive predictor of the likelihood that a NES respondent voted (β= 1.395; p < 0.001). Although the coefficient for Union City in Model

(2) remains positive, it fails to reach conventional levels of statistical significance (β= 0.088; p = n.s.), and the point estimate for the Union City effect is reduced to 3 percent. These results imply that Union City LCs are more likely to exist in districts with historically above-average political activity.

Model (3), which includes labor PACs, our proxy for the intensity of labor's effort by congressional district, offers evidence that the Union City phenomena is local rather than national. If PAC expenditures were positively correlated with activist LCs, then the Union City coefficient should decline in Model (3). Instead, the stable Union City coefficient across Models (1) and (3) indicates independence between LC activity and the strategic allocation of national PAC resources. Consistent with prior research, these findings suggest that national unions do not allocate PAC funds on the basis of regional membership strength (Gopoian 1984). Model (4), which includes all the independent variables, affirms this conclusion.

Union City, Race, Class, and 1992 Turnout

Do Union Cities reflect the socialization of political conflict across race and class? To explore this question, Table 3 compares voter turnout in congressional districts with a Union City and those without with respect to race, class, and 1992 turnout. Models (1) and (2) test whether there is a difference in turnout across race. The positive and statistically significant coefficient for white voters (β= 0.248; p < 0.05) in Model (1) indicates that the probability of whites voting was 9.3 percent higher than non-whites in districts without a Union City. Model (2) offers the same equation for districts with a Union City. Here the coefficient for white voters is positive yet statistically indistinguishable from zero (β= 0.110; p = n.s.), implying that whites and nonwhites voted at comparable rates in districts with a Union City.

A similar pattern emerges with class. Models (3) and (4) compare the voter turnout rates across class for districts with and without a Union City. In districts without a Union City, NES respondents who self-identify as working class vote at significantly lower rates than others (β= – 0.281; p < 0.01). In such districts, the probability of voting by working-class respondents was 10.4 percent lower than for respondents who did not identify as working class. By comparison, in districts with a Union City, coefficient estimates indicate no statistical difference between the voting rates of working-class and non-working-class respondents (β= -0.140; p = n.s.). The coefficient for working-class voters in Model (4) yields an estimated voter turnout differential of 4.7 percent.

The final equations examine the extent that district turnout in 1992 predicts voting in 2000. Turnout in 1992 is a strong positive predictor of voting by NES respondents for districts without a Union City but less so for districts with a Union City. The coefficient for 1992 turnout in Model (5) (β= 1.661; p < 0.001) indicates that for every percentage point increase in 1992 turnout, the probability of a respondent's voting increased by 0.61 percent. By comparison, the coefficient for 1992 turnout in Model (6) (β= 0.921; p = n.s.), while positive, does not reach conventional levels of statistical significance. Thus, voting rates in districts without a Union City were relatively undisturbed by events taking place between 1992 and 2000. Conversely, this finding supports the claim that the 1994 election results shocked LC leaders and other allies into a more activist mode.

Discussion and Limitations

The theoretical work of Schattschneider (1960), combined with the historic inverse between successful bilateral bargaining and union political activism, explain recent efforts by organized labor to reach out to progressive allies on the organizing front and the renewed emphasis on political militancy. These are mutually reinforcing tactics that enlarge the scope of labor-capital conflict. The question before labor leaders, then, is not whether unions should abstain from politics—for they cannot without undermining the progressive partnerships they need to rebuild collective representation in the private workplace. Instead, labor's declining bargaining power should push unions toward political mobilization tactics that complement a general strategy for enlarging the scope of labor-capital conflict.

Labor councils, as regional coalitions of local unions, are strategically positioned to expand alliances to include non-union organizations (Ness and Eimer 2001). If LCs are indeed shifting toward a strategy of socializing labor-capital conflict, there should be a positive association between the most active labor councils and the political participation of demographic groups traditionally aligned with organized labor. Our results do imply that LCs designated as "Union City" play a role in increasing political participation among the general population. Respondents in congressional districts with a Union City voted at rates that were approximately 5 percent higher than in districts without a Union City. This above-average turnout, however, was mediated by district turnout in 1992, suggesting necessary preconditions among the general population for the formation of Union City LCs.³

The supposition that LC strategy and capacity are partially a function of regional factors is reinforced when we examine other district characteristics with respect to Union City status. In districts without a Union City, nonwhites and working-class citizens voted at rates that were significantly lower than whites and non-working-class citizens. In districts with a Union City, nonwhites and working-class respondents voted at rates that were comparable to others. We tentatively conclude that relatively high levels of political participation by minorities and the working class help enable the formation of active LCs. This is not to imply that LC outreach has no effect on the voting rates of minorities and the working class. Indeed, outreach to working-class voters is presently a directive by the national AFL-CIO. Rather, our findings suggest that active LCs tend to arise in contexts where it is possible to form coalitions with progressive organizations working to expand the political voice of minorities and the working class.⁴

Although this research improves upon prior work in this area, there are still limitations related to data precision. The "Union City" designation is a crude indicator of whether an LC has a functional political mobilization program. Certainly many LCs that have not earned Union City status are involved in politics, and a dichotomous indicator fails to capture variation in political tactics and effort. Subsequent research on this topic would benefit from more comprehensive data on LC activities and, in particular, on measures of coalition activity between LCs and non-union organizations. Finally, the spatial match can be improved by examining the LC effect on smaller geographic regions. Congressional districts are large and often oddly shaped regions. In all likelihood, communities within the immediate proximity of LCs are most affected by union political mobilization. Both of these limitations, however, would tend to understate our point estimate for the Union City effect.