BACKGROUND

During the first half of the 20th century, collective bargaining by public workers was largely illegal throughout the country, and few government employees belonged to unions. Unions fared much better in the private sector, where the membership and political involvement of unions grew dramatically following the adoption in 1935 of the National Labor Relations Act (NLRA): a hallmark of the Democratic New Deal. By the late 1940s, the labor-Democrat alliance stood at the center of the American party system, with the Democrats benefiting from a newly empowered constituency they had played a key role in creating. Public-sector workers were on the sidelines. They were not included in the NLRA, and they remained unorganized and politically unimportant (Slater Reference Slater2004; Walker Reference Walker2014).

Change, however, was just around the corner. In 1959, Wisconsin became the first state to adopt a collective bargaining law for public-sector workers. In 1962, President Kennedy issued an executive order authorizing collective bargaining at the federal level. And soon the nation experienced a wave of new labor legislation—modeled after the NLRA—that quickly brought collective bargaining to virtually all states outside the South. Along with it came a surge in union organizing that increased membership tenfold between 1960 and 1976 (Goldfield Reference Goldfield1989–90; also DiSalvo Reference DiSalvo2015; Freeman Reference Freeman1986; Moe Reference Moe2011).

By the early 1980s, union density had risen to 37 percent of the public workforce (and much higher in many states and cities), where it stabilized. Meantime, private-sector unions—beset by rising competition, globalization, and structural change in the economy—fell into decline. Today, less than 7 percent of private-sector workers are unionized, and public-sector unions are the leaders of the union movement.

These developments have had large impacts on American politics. Public-sector unions enroll more than eight million members, are top contributors to political campaigns, field vast numbers of campaign workers, and marshal many other political resources—almost all of them in support of Democrats and in opposition to Republicans (DiSalvo Reference DiSalvo2015). They are core members of the Democratic Party coalition. They are also interest groups whose resources, electoral activity, and lobbying organization put them in the top tier of all groups that seek political influence. Their rise has been consequential at the national level, but they have been especially consequential at the state and local levels—where most of the nation's public employees work and most public money is spent.

The states’ adoption of public-sector labor laws is a prime example of policies making their own politics. The winners and losers, moreover, are clear: these developments were good for the Democratic Party and bad for the Republican Party. So were Democrats the driving force behind these laws, and did Republicans try to stop them?

Scholars have barely explored these matters. Political scientists have paid little attention to public-sector unions, and even less to the policy changes that fueled their growth. The same is true for historians (as noted by McCartin Reference McCartin2006; Shaffer Reference Shaffer2002). There is a literature on labor history, but it focuses on private-sector unions as the embodiment of the union movement, and a common theme is that the movement is in decline. In those rare cases when political scientists have written about unions, they too have focused on private-sector unions, particularly the relationship between those unions and the Democrats; and they too have highlighted the political consequences of union decline (e.g., Dark Reference Dark1999; Schlozman, Verba, and Brady Reference Schlozman, Verba and Brady2012).

Although the questions concerning us here have yet to be studied, there are well-known lines of evidence that point in a common direction. Pre-1960 collective bargaining breakthroughs in New York City and Philadelphia were due to supportive Democratic mayors. The Wisconsin breakthrough in 1959 came about under the leadership of Gaylord Nelson, the state's first Democratic governor in more than 20 years. Efforts to organize federal workers expanded when Kennedy's executive order signaled a more union-friendly era in government (e.g., McCartin Reference McCartin2006; Slater Reference Slater2004). The importance of Democratic control also shines through in case studies of labor law adoption in particular states (e.g., Saltzman Reference Saltzman, Richard and Ichniowski1988).

These accounts have given rise to a conventional wisdom that seems to make good sense. As Joseph A. McCartin (Reference McCartin2006, 79) expresses it, “the success of public sector unions was almost always dependent upon an alliance between those unions and Democratic politicians. . . . The record could not be clearer on this point: without the close collaboration that emerged between public sector trade unionists and Democratic leaders at all levels of government, the public sector movement would not have grown as quickly as it did” (see also Walker Reference Walker2014, 190). Quantitative analysis affirms these conclusions. Notably, Saltzman's (Reference Saltzman1985) study of state labor laws for teachers shows that their adoption was more likely when Democrats had more political control.Footnote ²

But empirical support for this conventional wisdom is thin. Even Saltzman's (Reference Saltzman1985) study—the most comprehensive to date—only scratches the surface. His model measures Democratic strength as the number of state government institutions controlled by Democrats, which assumes that all unit increases in Democratic strength have the same effect on adoption. But if the conventional wisdom is correct, what should really matter is whether Republicans are in a position to block—which holds under both divided government and unified Republican control. Marginal cases aside, these laws should only pass under Democratic unified government.

Is that the case? To answer this question, we assembled data on when the states enacted collective bargaining for five categories of government workers: teachers, firefighters, police, other local government workers, and state government workers.Footnote ³ We then grouped the states by party control at the time of enactment.

The results, shown in Table 1, are revealing: most states that adopted these bargaining laws did so under divided government, not Democratic unified government. Equally important, several bargaining laws were actually adopted under Republican unified government. This pattern is hard to reconcile with conventional wisdom—as well as with the Schattschneider-based expectation that Republicans, with so much at stake for their party's political future, would have opposed these new laws. It may well be that Democrats were the strongest supporters of state collective bargaining laws, but it is not true that Republicans strongly opposed them. Far from it, for most bargaining laws were actually adopted in political contexts where Republicans could have blocked their enactment—but didn't.

TABLE 1.

Party Control of State Government at the Time of Collective Bargaining Adoption

THEORY

If policies make their own politics, then it seems straightforward that politicians would want to use policy to shape politics to their advantage. But we need to look more closely at their actual incentives for doing that.

Consider first the policy feedback effects highlighted in the literature, where the focus is on how policies can change politics—by providing benefits to mass constituencies, for example—in ways that bolster their own political support. Feedback effects occur whether they are intended or not, but they will differ depending on how the policies are designed. Would their advocates have incentives, then, to favor designs that make the feedback effects as politically potent as possible (Patashnik Reference Patashnik2008; Patashnik and Zelizer Reference Patashnik and Zelizer2013)?

There is good reason to think the answer is yes. Politicians are concerned with pleasing the constituencies and interest groups that can get them reelected; and if a politician supports a policy to benefit those interests, she could easily have incentives—in furthering the same interests—to favor designs that best promote the policy's future political support. There are qualifications, however. Politicians may be unaware in any instance that they can use policy to shape politics, or of what the consequences might be; they are also notoriously myopic, thinking mostly of their next election rather than the long-term effects of policy (Arnold Reference Arnold1990); and as Mayhew (Reference Mayhew1974) reminds us, their embrace of policy may be due to position-taking or credit-claiming, not to genuine concern for the policy's success. All these factors can weaken the strategic use of policy for politics. Even granting these qualifications, though, the saving grace is that there may be interest groups that are aware, do look ahead, and are genuinely concerned about the policy—and that pressure politicians to “make politics” in ways that bolster the policy's durability.

All things considered, then, it is reasonable to suggest that politicians will often have incentives to take politics-shaping actions that are policy specific. This does not mean that they will get the outcomes they desire—for as policy scholars have argued (Pierson Reference Pierson2004) and as Oberlander and Weaver (Reference Oberlander and Weaver2015) have recently shown with application to the Affordable Care Act (ACA), policies and their environments can be complex and their feedback effects difficult to anticipate, leading to outcomes that depart from the designers’ intentions (see also Howard Reference Howard1997). Our point here, however, is that they will often have the incentives to try—which held true, we should note, even in the case of the complex ACA.

But will they have similar incentives when it comes to using policy to shape the larger structure of partisan politics? Here the situation is different. Public-sector labor laws, for example, have far-reaching effects that favor Democrats and disfavor Republicans—and these effects are collective goods (or bads). If the Democratic Party were a unitary actor (and sufficiently aware and forward-looking), it would support these laws, and if the Republican Party were a comparable unitary actor, it would oppose them. But they are not unitary actors. The decision-makers are individual politicians; and each of them, concerned about reelection and constituency, has incentives to do what is best for herself—not what is best for the party.

This collective action problem is fundamental to the theoretical perspective on parties that, over the last two decades, has been most central to the field of American politics, particularly when the focus has been on legislators. The theory argues that legislators within the same party carry a common brand and recognize their shared electoral fate, and to achieve mutual gains they delegate authority to party leaders—who have incentives to promote the larger interests of the party by orchestrating member votes and controlling the agenda (Cox and McCubbins Reference Cox and McCubbins1993, Reference Cox and McCubbins2005; Jenkins and Monroe Reference Jenkins and Monroe2012). This solution to the collective action problem, however, is only partial at best, and is most effective—with members supporting stronger party constraints on their behavior—when they are already in substantial agreement on policy (Aldrich and Rohde Reference Aldrich, Rohde, Lawrence and Oppenheimer2001; Rohde Reference Rohde1991). The more diverse the members are, the weaker the enforced constraints of party, and the more members can follow their individual incentives.Footnote ⁴

One implication is that, to the extent parties are internally homogeneous, they are better able to promote the parties’ interests. The modern era of polarization, then, is a best-case setting for politicians to use policy to shape the structure of partisan politics. In earlier decades, however, these conditions did not prevail (Mayhew Reference Mayhew1974; Rohde Reference Rohde1991). The parties were more internally diverse and thus, theory would suggest, less capable of getting their members to shape politics to the parties’ advantage. The labor laws we are concerned with here, of course, were enacted during this earlier era. So our expectation is that, while the Democratic Party stood to be advantaged by these laws and the Republican Party disadvantaged, the legislators in each party would not weigh this heavily in deciding how to vote. They would be mainly motivated by their own individual-level concerns, not by what was best for the party.

What would this mean, more concretely, for how legislators would vote? Here, the mainstream theory needs elaboration. As recent critics have pointed out, the theory arises out of a Downsian framework of politicians and voters, and pays no direct attention to the specifics of policies, the constituencies and interest groups associated with them, and the latter's influence on what politicians do (Hacker and Pierson Reference Hacker and Pierson2014). Indeed, it has recently been challenged by an alternative perspective, which argues that parties should be treated as coalitions of interest groups, with groups as the key decision-makers (Bawn et al. Reference Bawn, Cohen, Karol, Masket, Noel and Zaller2012; Cohen et al. Reference Cohen, Karol, Noel and Zaller2008). We cannot, of course, adjudicate this debate here. What we can do is extend the theory we have already developed, taking account of the specific nature of the policies, constituencies, and interest groups of relevance to legislators’ votes.

For Democrats, all of these factors worked to their advantage. Outside the South, they were the party of labor, allied with unions—interest groups at the core of the party—and for virtually all nonsouthern Democrats, there were constituency and ideology-based reasons, at the individual level, for supporting these new laws. In following their own personal incentives as politicians, then, individual Democrats would automatically be making politics to the party's larger advantage—even though, theory suggests, that was not what motivated them.

The Republican Party was less fortunate. While today's party can accurately be characterized as antiunion and homogeneously conservative, in earlier decades it had an appreciable number of moderates (anchored in politically moderate constituencies)—and they were more favorable to labor than conservatives were (Shafer Reference Shafer2003). Thus, there was a contingent of potential defectors—moderates—who had constituency and ideology-based incentives to vote in favor of these labor laws despite their future impacts on the party.

A second problem deepened the party's troubles. During these earlier times, Republican conservatives were known as antilabor because they and their business allies were often opposed to the private-sector unions that dominated the union movement—and those private-sector unions (often in manufacturing) were not a forceful presence in conservative constituencies. But the workers now demanding bargaining rights were the employees of governments, not business. Business groups therefore had less incentive to oppose public-sector labor laws, freeing many Republicans from group pressure to vote “no.” Workers in the public sector, moreover, were present in every district, including conservative districts. Even for conservative Republicans, then, a vote for public-sector labor laws may often have been the smart thing to do politically—regardless of its impact on the party.

These problems for the Republican Party were compounded by a third political force, this one arising from public-sector strikes. During the decades when these laws were being considered, the nation was swept by strikes as government workers and their nascent unions sought to bring pressure on policymakers. These new developments were quite troubling to the public, which feared a loss of government services, and many policymakers felt a pressing need to deal with the immediate problem (or impending threat). Collective bargaining was presented by academics, labor-law experts, and union advocates as a solution: a way to substitute negotiation for conflict, and thus to bring “labor peace” (McCartin Reference McCartin2008; Shaffer Reference Shaffer2002). This, then, was yet another force potentially pushing Republicans to support new labor laws—for if they did not, and if they stood in the way of collective bargaining, they risked being blamed by voters for future disruption of government services. There is good reason to think, therefore, that if collective bargaining was viewed as a way to bring labor peace (a contingency we will revisit in our analysis), strikes were likely to succeed in getting Republicans to support it.

In sum, while the collective action problem was relevant to both parties, the individual-level incentives of Democrats (outside the South) on these labor laws lined up with what was best for their party's political future—but the opposite was true for Republicans. Our theory suggests that Republican moderates were especially likely to defect, but also that many conservatives had political grounds for defecting too—and that, for both, the incentives for doing so were heightened by the political pressures arising from strikes. For Republicans, the individual incentives of politicians were simply not aligned with promoting the party's collective good—and what they would get, as a result, was a collective bad.

PARTISAN SUPPORT FOR COLLECTIVE BARGAINING LAWS

To explore these matters, we compiled a new dataset. For each duty-to-bargain law passed between 1959 and 1990, we collected the final votes of as many state legislators as possible.Footnote ⁵ The completed dataset includes 5,439 votes on 42 bills in 31 states, as well as the party and district number of each legislator.

We begin with two simple questions: Were Democrats unified, or nearly unified, in support of collective bargaining? And was Republican support limited to a few crossover votes, or was it substantial? To answer these questions, we display in Table 2 the percentages of Democrats and Republicans who voted for each bill, sorted by year.

TABLE 2.

Party Support for Collective Bargaining Laws

As expected, Democratic support was extremely high in almost all states. In three quarters of the bills, more than 90% of Democrats voted in favor, and in many their support was unanimous. We find five exceptional cases of Democratic support below 80%, but two were in the South (where Democrats were overwhelmingly conservative), and the remaining three were passed under unusual political circumstances—with some Democrats voting “no” because the bills were not sufficiently prounion. Aside from these unusual cases, the clear pattern among Democrats is near-universal support for public-sector bargaining laws.

For Republicans, our findings are especially instructive. Earlier, we showed that Republicans failed to block collective bargaining laws when they could have done so. Here we see, in exploring the votes, that the percentage of Republicans supporting enactment was actually quite large. In some states, such as Pennsylvania in 1968, almost all Republicans voted “yes.” In others, such as South Dakota, Republican support was unanimous. There was substantial variation in Republican support across states, however. In some, such as Iowa, these laws passed with smaller percentages of Republican votes; and in a few, such as Ohio, not a single Republican voted “yes.” But what is most striking about the findings in Table 2 is that so many Republicans voted to grant bargaining rights to government employees: in our dataset, 63% of the Republican votes are “yes” votes. Such extensive support is hard to square with the idea that they were using policy to promote the best interests of their party.

Why did so many Republicans support collective bargaining? Our first expectation is that, if Republicans were acting on their individual-level incentives, moderates would be more likely to vote “yes” than conservatives. To test this expectation, we constructed a measure of legislative district ideology using county data on Democratic presidential vote share. Because this vote share can vary greatly across elections, we chose two elections in which the national popular vote was closely divided between Republicans and Democrats: 1960 and 1976. Then, for all years between 1960 and 1976, as well as years before and after, we linearly interpolated the Democratic vote within each county. Next, we used historical maps and descriptions of district boundaries for relevant years to identify the county or counties contained in each district, and by computing a weighted average of the Democratic presidential vote across these counties, we arrived at our measure of each district's ideology.

We also expect that Republicans with high concentrations of government employees in their districts should be more likely to vote “yes” than Republicans with low concentrations. To test this, we again turned to county data—this time U.S. Census data on government employment in 1970 and 1980. As before, we linearly interpolated the percentage of workers in each county who were employed by government, and we matched those counties to state legislative districts using the same procedure as before. (See the Online Appendix for more detail.)

We begin the analysis by using logistic regression to estimate the effects of district ideology and government employment on state legislators’ votes (“yes” = 1, “no” = 0). Because we expect these variables to matter little for Democrats but a great deal for Republicans, we also include legislators’ party (Republican = 1, Democrat = 0) and its interaction with ideology and government employment. We cluster the standard errors by state.

The results are set out in the first column of Table 3. As expected, the coefficient on Democratic presidential vote is insignificant, suggesting that the ideology of Democrats’ districts had no significant relationship to their votes on collective bargaining. However, we find that district ideology did matter for Republicans. At the bottom of column 1, we show that the sum of Democratic presidential vote and its interaction with Republican is positive and significant at the 1% level. Republicans were more likely to favor collective bargaining, then, the more moderate their constituencies.

TABLE 3.

State Legislators’ Collective Bargaining Votes

Notes: Standard errors clustered by state in parentheses. *p < 0.10; **p < 0.05; ***p < 0.01.

Similarly, we find that Democrats’ votes did not vary with the amount of government employment in their districts, but Republicans’ votes did. As we show at the bottom of column 1, the sum of Government employment and its interaction with Republican is positive and statistically significant. This finding aligns nicely with our expectations: the greater the concentration of government employment in the district, the greater the likelihood that a Republican voted “yes” on public-sector collective bargaining.

Table 4 presents some key predicted probabilities from this model that convey the magnitudes of these effects. We predict the probability of a “yes” vote for two types of Republicans: “Conservative Republicans” are those from districts where 34% of the presidential vote went to the Democratic candidate (the 5th percentile) and “Moderate Republicans” are those from districts where 55% of the presidential vote went to the Democratic candidate (the 95th percentile). We start by examining the probabilities for these two types of Republicans in districts with low government employment—which we define as those where 10% of workers are employed by government (the 5th percentile). In row 1 of Table 4, we find that the predicted probability of a “yes” vote was 46% among conservative Republicans. By contrast it was 70% for moderate Republicans—a 24 point difference. Clearly, then, in districts with low government employment, moderate Republicans felt greater pressure to support collective bargaining than conservative Republicans.

TABLE 4.

Predicted Probabilities

In row 2 of Table 4, we calculate the same probabilities but for Republicans in districts where 29% of workers were employed by government (the 95th percentile). Here, too, we find a marked difference in labor law support between conservative and moderate Republicans (20 percentage points). But we also see, by comparing rows 1 and 2, a large effect of government employment on Republicans’ votes. For conservatives, the predicted probability of a “yes” vote was 61% in districts with high government employment, but only 46% in districts with low government employment. Moderate Republicans from districts with high government employment voted “yes” on collective bargaining 81% of the time—11 points more often than moderates in districts with low government employment. The evidence indicates, then, that many Republicans—conservatives and moderates alike—supported collective bargaining because government employees were important local constituencies for them.

Next we add state fixed effects to account for the possibility that state characteristics—such as private-sector union strength and the professionalization of the legislature—affected Republican support. This modeling strategy allows us to ask whether within states, and thus controlling for time-invariant state characteristics, the basic relationships we found in the original model continue to hold up.Footnote ⁶

The answer is yes. As column 2 of Table 3 indicates, moderate Republicans were significantly more likely to vote for duty-to-bargain laws than conservative Republicans in the same state, and Republicans from districts with high government employment were more supportive than those from districts with low government employment. In rows 3 and 4 of Table 4, we present predicted probabilities from this model.Footnote ⁷ Moderate Republicans were 14 to 20 percentage points more likely to vote “yes” than conservative Republicans. And for Republicans with high concentrations of government employees in their districts, the probability of support was 14 to 20 points higher than for Republicans with low concentrations.Footnote ⁸

It is important to emphasize that all bills in our dataset are the same at their core: they all established formal collective bargaining rights for government workers, modeled after the NLRA. Yet there are some differences between them at the margins—for example, in how they deal with strike penalties or what occupations they target. To partial out the effects of these bill characteristics on votes, we add bill fixed effects to the model in column 3 of Table 3. Our results are substantively unchanged. Again, at the bottom of Table 3, we find that the effects of district ideology and government employment are positive and statistically significant for Republicans. (See Table 4 for the magnitudes.)

In the final column of Table 3, we test our expectations about how strikes influenced Republican votes. As we discussed earlier, the sudden upsurge in strikes by public workers was a political concern in the 1960s, and many experts argued that granting them collective bargaining would bring labor peace. Until the late 1960s and early 1970s, there was little direct evidence to counter that claim, and Republicans had good reason to bank on it—and embrace collective bargaining as a way of slowing government strikes.

As time went on, however, more and more states had actual experiences with collective bargaining, and legislators could look to them for evidence. If it appeared that strike activity declined after states adopted duty-to-bargain laws, we would expect Republicans to continue their support. But if strike activity stayed the same or increased after adoption, we would expect their support to decline.

What, then, would state legislators have observed had they looked to the experiences of other states? In Figure 1, we plot government strikes over time for 22 states—the states that enacted duty-to-bargain laws for three out of the five categories of government employees at the same time. (We exclude states that extended bargaining rights to only one or two categories of workers, because strikes may have continued on behalf of the excluded workers.) The vertical line in each graph depicts the year that the state adopted its law.

FIGURE 1.

Strikes by state and local government employees

The trend immediately apparent in Figure 1 is that, in almost all states, strike activity tended to increase after they adopted duty-to-bargain laws. In some, the increase was dramatic. In Michigan, which adopted a comprehensive law in 1965, there were only two public strikes in the years leading up to adoption, but four years after adoption there were 70 strikes. Ten years later, there were nearly 100 strikes. The pattern was similar in many other states. Given this experience, it seems likely that state legislators would question the earlier claim that collective bargaining brought labor peace—and we would expect Republican support for collective bargaining to decline as a result.

To test for this, we created a measure of how Republicans might have thought about the relationship between collective bargaining and strikes at the time of their votes. For each state in Figure 1, we calculated the average number of strikes during the three years prior to the law's adoption. Then, for all years after adoption, we subtracted this prebargaining baseline from the number of strikes in that particular year. We then summed these differences across all 22 states, by year. This aggregate measure, which we call Postbargaining strikes, captures the extent to which strike activity changed after the adoption of duty-to-bargain laws. We link this measure to the votes in our dataset using a one-year lag.

In column 4 of Table 3, we include this variable and its interaction with party. As expected, we find that as legislators could witness more and more strikes in states that already had collective bargaining, Republican support declined. We can see this from the hypothesis test at the bottom of column 4: when we add the coefficients of Postbargaining strikes and its interaction with Republican, we estimate a negative effect, significant at the 1% level.Footnote ⁹

The predicted probabilities from this model (see Table 4) show that the effect of Postbargaining strikes was quite large. For conservative Republicans in districts with low government employment, shifting from an environment of no strikes in bargaining states (and therefore no reason to doubt that collective bargaining would reduce strikes) to an environment of 206 strikes per year in bargaining states (the maximum in our dataset) reduced the probability of voting “yes” from 81% to 27%. For moderate Republicans, the effect was a change from 94% support to 58%. We see a similar pattern in districts with high government employment, where Republican support tended to be higher across the board. Specifically, our model predicts that the vast majority of these Republicans—moderates and conservatives—voted “yes” on collective bargaining when postbargaining strike incidence was zero (89% for conservatives and 97% for moderates). At the maximum level of Postbargaining strikes, however, conservative support dropped to 42%, and moderate support decreased to 73%. These results support our theoretical expectation: Republicans’ support dropped as they observed what was happening in states that had already adopted collective bargaining.

Taken together, our findings show that, on the key policy issue of collective bargaining, most Republicans did not act in the interest of their party. Although they were usually in a position to block, nearly two-thirds of the Republicans in our dataset voted “yes.” Those “yes” votes were disproportionately likely to come from moderate Republicans, Republicans from districts with high concentrations of government employees, and Republicans who thought they could prevent government strikes. Thus, an outcome that appears collectively irrational—Republicans supporting prounion labor laws—is sensible in terms of the self-interest of individual legislators. By thinking about their own districts and constituencies, Republicans played a pivotal role in enacting legislation that disadvantaged their party.