Why States Preempt City Ordinances: The Case of Workers’ Rights Laws

Abstract

Despite being popular with the public and preventing racial and economic inequality, states often preempt their local governments’ ability to adopt workers’ rights laws. We test several competing theories of preemption (ideology, political institutions, interest group involvement, demographics, and policy diffusion) using a time-series, cross-sectional approach. Using data on state legislative activity from 1993 to 2018, we find that increasing legislative conservatism, and more unified political control of the state government, regardless of party, are associated with a higher risk of preempting local workers’ rights laws, all else equal. Our focus on legislative ideology, a more precise measure than party control at the subnational level, as the nexus of preemption activity helps clarify prior contradictory results in the literature. For those looking to prevent or overturn workers’ rights preemptions, the most direct approach appears to be to change the ideology of state legislatures.

In 2015, Michigan passed the Local Government Labor Regulatory Limitation Act, a sweeping bill prohibiting the state’s local governments from regulating workers’ rights, including setting a minimum wage higher than the state minimum and mandating paid leave, prevailing wages, or fair scheduling. Groups such as the National Federation of Independent Business (2015) and the Michigan Chamber of Commerce celebrated the law, while opponents such as Progress Michigan nicknamed the legislation the “Death Star Bill,” as a symbol of “tyrannical power” (Progress Michigan 2015, para. 4). Such legislation limiting local government authority, called state preemption, has been increasing in recent years, both generally (DuPuis et al. 2018; Goodman et al. 2021; Riverstone-Newell 2017) and for workers’ rights specifically (Economic Policy Institute 2019).

Many labor policies are controlled by state governments (Hertel-Fernandez 2019), yet workers’ rights reform movements increasingly focus on cities (Johnson 2021). In the face of slow or nonaction by the state, cities have exercised their real or perceived policymaking powers in this area, setting the stage for state preemption, which ensures state priorities take precedence over local policy preferences. Five workers’ rights policies that are often preempted are minimum wages (floors on hourly wages), fair scheduling (requiring employers to provide predictable work schedules with advance notice), project labor agreements (contracts setting basic safety conditions and pay), prevailing wage requirements (mandating pay of at least the local median wage), and paid leave (for family and medical reasons).

Workers’ rights preemptions are particularly salient for two reasons. First, these preemptions may not reflect public preferences. For example, the minimum wage is often less than the public wants, with the difference between the public’s ideal minimum wage and the existing one is greatest for states with preemption compared to those without it (Simonovits and Payson 2020). Second, these preemptions can lead to inequities. Labor market regulations have a direct impact on income inequality (Hatch and Rigby 2015), with the rise in income inequality for the last 40 plus years being at least partially a function of the declining real value of the minimum wage (Dalmat 2005). People of color are most likely to benefit from minimum wage increases (PowerSwitch Action 2017), raising racial equity concerns. “Preempting wage, leave, and anti-discrimination regulation places particular harms and burdens on groups that have faced historic discrimination—such as women, people of color, and LGBTQ populations” (Johnson 2021, 1193). In addition, variation in policies such as paid leave leads to inequities across jurisdictions (Pomeranz et al. 2022). According to Johnson (2021), challenging workers’ rights preemptions in court is risky, so it is preferable for advocates to address them politically. However, in order to do that, advocates need to know what increases the likelihood that a state will adopt a workers’ rights preemption.

Several studies have started to examine the reasons for workers’ rights preemptions. Narrative case studies (Riverstone-Newell 2017; Bunch 2021) and cross-sectional analyses (Flavin and Shufeldt 2020) find that these preemptions are most likely in Republican-controlled states. While suggestive, these studies do not explain changes in preemption laws over time. Most closely related to this study, Bucci and Jansa (2021) explore the reasons for restrictive labor policies (including some workers’ rights preemptions). We build on their study by focusing on the reasons for workers’ rights preemptions specifically because preemptions, which limit local policymaking, have different policy objectives than general state labor laws. Unlike general state-level laws, preemption alters relationships within the federalist system and, therefore, may have varying antecedents.

Using time-series cross-sectional models and data on five workers’ rights preemptions adopted between 1993 and 2018, we test the various hypotheses in the literature surrounding the factors associated with preemption adoption: ideology, political institutions, interest groups, demographic factors, and policy diffusion. Our research makes three contributions to the literature. First, our focus on workers’ rights preemptions rather than state-level labor laws or preemptions generally highlights the unique mechanisms associated with limiting local governments’ authority over labor conditions within their jurisdictions. Second, our time-series data allow us to examine factors explaining changes in labor preemptions over time, improving on previous studies that, by their cross-sectional nature, could only study correlations between existing policies and context. Such cross-sectional studies explain all existing policies as a function of the state context at one point in time, while time-series analyses examine the adoption of a preemption within the state context at that time, acknowledging the political landscape shifts over time. Third, our use of ideology rather than partisanship increases the precision of our analysis. As Shor and McCarty (2011) note, political parties at the subnational level are heterogeneous. A Republican-dominated legislature in one state may be more liberal than a Democrat-dominated legislature in another, thus muddying the usage of political parties to predict preemption. We find that more ideologically conservative state legislatures are likelier to adopt state preemptions of workers’ rights, while less politically unified governments will preempt less often. Contrary to expectations and previous studies on preemption more generally, legislative professionalism, interest groups, demographic factors, and policy diffusion do not significantly impact the adoption of workers’ rights preemptions.

In the next section, we discuss the relevant literature on factors associated with state preemption and workers’ rights laws, paying particular attention to existing studies on workers’ rights preemptions, and develop our hypotheses. Next, we introduce the data and methodology. After presenting our results, we discuss the theoretical and methodological contributions of our study to the preemption literature, as well as the implications of our study for workers’ rights groups.

Background

State Preemption

While the federal government also preempts its states (SoRelle and Walker 2016), our focus here is on state preemption of its local governments. Such preemption is defined as “the use of coercive methods to substitute state priorities for local policymaking” (Goodman et al. 2021, 147). Each branch of government—judicial (Swanson and Barrilleaux 2020), executive (McDonald et al. 2020; Weissert et al. 2021), and legislative (Riverstone-Newell 2017)—can initiate a preemption, but the workers’ rights preemptions we study here are legislative.

There are four main types of preemptions (Wagner et al. 2019). Ceiling preemptions set a maximum on a policy, such as stating that no local government can enact a minimum wage over the state rate. Floor preemptions are the opposite, setting a minimum on a policy, such as stating that the state has a minimum wage, but local governments can enact higher minimum wages if they want. Vacuum preemptions occur when the state prohibits their local governments from regulating a policy area, but the state does not enact additional regulations in that area, such as by stating that local governments cannot adopt a minimum wage without setting a state minimum wage. Punitive preemptions have consequences, such as fines or withholding of state funds, if local governments act contrary to the preemption. Workers’ rights preemptions tend to fall into the first and third types.

The literature identifies five frequent explanations for why states preempt their cities. The most common justification is ideology or partisanship (Bunch 2021; Flavin and Shufeldt 2020; Fowler and Witt 2019; Goodman and Hatch 2023; Riverstone-Newell 2017). These studies note that Republican legislatures and states with more conservative citizenry are more likely to preempt their Democratic and more liberal cities. Konisky and Nolette (2022) predict preemptions are likely to continue because of entrenched political dynamics, with state legislatures being more conservative and cities more liberal. Ideology or partisanship may explain workers’ rights preemptions specifically, either as an independent factor or in association with interest group involvement. Bucci and Jansa (2021) observe Republican control increases the likelihood of states adopting restrictive labor policies, although Kim et al. (2021) detect no such effect. Both studies identify that unionization rates change the influence of Republican control, though in opposite directions. As Shor and McCarty (2011) explain, subnational political parties are heterogeneous—a Republican in one state is not necessarily comparable to a Republican in another state. This distinction may be at the root of the prior contradictory findings. We remedy this by focusing on ideology and hypothesize that more ideologically conservative legislatures will be more likely to adopt workers’ rights preemptions.

Second, political institutions, particularly competition (both within the government and electoral) and professionalism, may explain why states preempt their cities. While preemption is often seen as purely ideological, some researchers argue it is unified political control of the state government (a lack of competition) that actually matters. Preemption is most likely when Republicans control both houses of the legislature and the governorship (Flavin and Shufeldt 2020; Fowler and Witt, 2019); however, Swindell, Svara, and Stenberg (2021) note that while Republican trifectas most often preempt, Democratic trifectas engage in similar behavior. Barber and Dynes (2023) find evidence of political control, regardless of party, associated with preemption activity. These findings suggest that the level of competition within the state government for political control is associated with preemption; thus, we hypothesize that more unified political control of the state government leads to a higher likelihood of preemption, regardless of party.

In addition to competition between parties for control of the state government, electoral competition influences preemption activity. More electoral competition makes legislators more likely to be punished for controversial bills (Rogers 2017). Likewise, less electoral competition reduces the consequences of unpopular votes. There is little empirical evidence to test this theory related specifically to workers' rights preemption, although Goodman and Hatch (2023) find no evidence that electoral competition predicts affordable housing preemption. However, given that workers’ rights preemptions may be politically unpopular (Simonovits and Payson 2020), we hypothesize that electoral competition will be negatively associated with adopting workers’ rights preemptions.

Jansa, Hansen, and Gray (2019) provide evidence that less professional legislatures are more likely to copy from other adopters. At the same time, groups such as the American Legislative Exchange Council (ALEC) often produce model legislation for its members. These two forces suggest that less professionalized legislatures may be more likely to preempt. However, the evidence of this is somewhat mixed. In their study of preemption across seventeen policy areas, Fowler and Witt (2019) find legislative professionalism positively correlates with preemption, while Goodman and Hatch (2023) observe a negative correlation between legislative professionalism and preemption of affordable housing policies. Other studies do not detect a relationship between legislative professionalism and preemption generally (Flavin and Shufeldt 2020) or labor preemption specifically (Kim et al. 2021). This suggests that legislative professionalism may be a factor that matters for preemptions in some policy areas but not others. Based on the Kim et al. (2021) study, we expect legislative professionalism will not impact the adoption of workers’ rights preemptions.

Third, research identifies interest groups as having a positive effect on the likelihood a state adopts a related preemption law (Givel and Glantz 2001; Goodman et al. 2021; Pomeranz and Pertschuk 2017). Conservative groups increase the likelihood that a state adopts preemption laws generally (Flavin and Shufeldt 2020) and labor preemptions specifically (Hertel-Fernandez 2019). States with low minimum wages have more labor preemptions, which Kim et al. (2021) argue is an indicator of the influence of corporations on these policies. For instance, prominent business associations such as the National Restaurant Association (2014) and the National Retail Federation (2014) oppose local, state, and federal minimum wage increases. Free market policies are often packaged and distributed to more conservative state legislators via groups such as ALEC and the U.S. Chamber of Commerce (Hertel-Fernandez 2016). These organizations have successfully influenced legislators to introduce and pass legislation consistent with their pro-business policy preferences (Hertel-Fernandez 2014; Riverstone-Newell 2017). If this is the case, states with more pro-business interest groups might actively preempt workers’ rights-related local ordinances.¹

On the other side of most workers’ rights policies are unions, which can constrain the effect of business group mobilization on the state business climate (Witko and Newmark 2005). Evidence on the relationship between unionization and workers’ rights legislation is somewhat nuanced. Bucci and Jansa (2021) find that unified Republican governments are less likely to adopt restrictive labor laws when there are strong unions, but when the authors add minimum wage and paid leave preemptions to their list of restrictive labor laws, total unionization levels become insignificant. Looking only at workers’ rights preemptions, Kim et al. (2021) observe that unionization decreases the likelihood of these preemptions, except in Republican-controlled states where higher unionization rates are positively associated with more labor preemptions. Grossmann (2013) argues labor policies are a sparse network with high centralization around unions, providing further reason to believe unions will be a significant factor in explaining workers’ rights preemptions. Thus, we expect a relationship between unionization and workers’ rights preemption, although we do not hypothesize in which direction.

Fourth, demographic factors may increase the likelihood of preemption. Preemptions are often responses to city actions (Riverstone-Newell 2017). While cities appear to play a role in preemption, Flavin and Shufeldt (2020) and Fowler and Witt (2019) do not detect a significant relationship between the percent of the state that is urban and the number of state preemptions. We, therefore, expect the urban population not to affect workers’ rights preemptions. Another demographic factor potentially explaining preemption is race. “Preemption legislation is often passed by predominantly white legislatures blocking laws benefiting and supported by majority communities of color” (PowerSwitch Action 2017, para. 1). Flavin and Shufeldt (2020) find support for this contention: the number of preemptions increases when a state has more African Americans. Therefore, we expect states with a larger non-white population will preempt workers’ rights more often.

A final potential explanation for preemption is policy diffusion. Policy diffusion occurs when policy innovations spread across jurisdictions due to factors internal and external to those jurisdictions (Berry and Berry 1990; Boushey 2010). Few studies have specifically examined preemption diffusion. However, Shipan and Volden (2008) observe that state preemption reduces the diffusion of city-level policies, and Goodman and Hatch (2023) detect no evidence of affordable housing preemption diffusion. Mallinson (2021) argues that states may preempt in an attempt to quarantine their state from local policies diffusing from nearby states. However, there is evidence of policy diffusion in labor policies: states adopt their own right-to-work laws when their neighboring states do as a way of trying to attract business (Pugh 2012). We, therefore, predict workers’ rights preemptions will similarly diffuse across states.

Workers’ Rights Laws

Broadly, workers’ rights law “governs the wages, benefits, rights of workers and responsibilities of employers” (Bucci and Jansa 2021, 410). While the federal government provides a floor for many such policies, significant policy variation exists among states and localities. The Economic Policy Institute (EPI, 2018) collects information on five workers’ rights preemptions²: minimum wage, fair scheduling, project labor agreement, prevailing wage, and paid leave. Minimum wage preemptions prevent local governments from setting minimum wages above the state minimum wage. Riverstone-Newell (2017) argues minimum wage preemptions are a response by state legislators to specific city actions, yet Langan and McFarland (2017) observe state preemption has no statistical effect on the likelihood of a city having a minimum wage policy. Paid leave preemptions prohibit local governments from requiring employers to provide paid medical or family leave. The number of these policies, as well as their preemptions, has been increasing since 2009 (Pomeranz et al. 2022). Fair scheduling preemptions prevent local governments from requiring that employers provide stable and predictable work schedules with advance notice. Project labor agreement preemptions prohibit local governments from requiring municipal contractors to abide by contracts that include basic conditions for pay and safety. Prevailing wage preemptions are also aimed at municipal contractors—these preemptions prohibit local governments from requiring contractors pay workers at least the local median wage for that type of work. To the best of our knowledge, no studies explicitly examine preemption of these latter three policies alone.

There are three key studies that examine workers’ rights preemptions and upon which this research builds. In a cross-sectional analysis, Flavin and Shufeldt (2020) study the probability of each of the five workers’ rights preemptions, finding Republican control is a significant factor for four out of the five policies, race is a significant factor for two of the policies, and percent of the state that is urban is a significant factor for one policy. They do not observe citizen liberalism, legislative professionalism, conservative group networks, initiative use, or Dillon’s rule to have a significant association with preemption of these policies. In another cross-sectional analysis, Kim et al. (2021) explore factors associated with five workers’ rights preemptions (substituting right to work for project labor), detecting more labor preemptions in states with low minimum wage wages (a proxy for state alignment with large corporations) and Republican control combined with higher unionization rates. In other words, they contend the existing labor environment and politics matter for the likelihood of having a workers’ rights preemption.

Those two studies are an important foundation for understanding why states adopt workers’ rights preemptions. However, as cross-sectional studies, they only observe correlations, and cannot explain changes in laws. In contrast, Bucci and Jansa (2021) take a time-series, cross-sectional approach to identify factors associated with state-level restrictive labor laws. Their main specification explains a labor policy restrictiveness factor score, which includes prevailing wage preemption. They find unified Republican governments are more likely to adopt restrictive labor laws, although these governments are less likely to adopt restrictive labor laws when there is a strong union presence.³ In a secondary analysis where minimum wage and paid leave preemptions are added to the labor policy restrictiveness factor score, unified Republican government remains statistically significant, as does the interaction between unified Republican government and unionization rates, but unionization alone is not a significant predictor of labor policy restrictiveness. Our analysis differs from that of Bucci and Jansa by focusing on workers’ rights preemptions, which replace local policymaking with state priorities, rather than state-level policies which do not alter the state-local balance of power.

We build on these studies by asking what factors affect the likelihood of states adopting workers’ rights preemptions. In contrast to cross-sectional studies that can only identify factors associated with the existence of these laws, we take a time-series, cross-sectional approach that allows us to test several competing theories (ideology, political institutions, interest group involvement, demographics, and policy diffusion) for explaining the adoption of these preemptions. In the following section, we discuss our data and methodology.

Data and Empirical Strategy

Preemption of Workers’ Rights Policies

Data on state preemption of local ordinances are difficult to track. As of yet, there is no centralized repository of information on state preemption activity. We rely on the work of the Economic Policy Institute (EPI) to begin our analysis. EPI provides legislative citations for five workers’ rights-related preemptions (outlined above). We further expand on these data by collecting the date of enactment to form the final dataset. The primary dependent variable for this analysis is a preemption of any kind from 1993 to 2018. As such, this is a dichotomous variable indicating zero if there are no preemptions in a given year and one if there are any in a given year.⁴

Figure 1B shows the geographic distribution of these data. Numerous states have yet to preempt any of the five policies legislatively; however, there is variation, with three states (Kansas, Michigan, and Tennessee) preempting all five over the 25-year time period. As seen in figure 1A and table 1, preemption of local minimum wage ordinances is the most prevalent and consistent over time; however, the other four types have become common in recent years. Preemption of local paid leave ordinances leads this group of recent legislation.⁵

Figure 1

Workers’ rights preemptions across time and space.

Table 1

Summary statistics

Variable	Source	Units	Mean	p25	p50	p75	Std. Dev.
Any workers’ rights related preemption	EPI	Indicator, 0 or 1	0.045	0.000	0.000	0.000	0.208
Minimum wage preemption	EPI	Indicator, 0 or 1	0.020	0.000	0.000	0.000	0.139
Fair scheduling preemption	EPI	Indicator, 0 or 1	0.007	0.000	0.000	0.000	0.082
Project labor agreement preemption	EPI	Indicator, 0 or 1	0.016	0.000	0.000	0.000	0.125
Prevailing wage preemption	EPI	Indicator, 0 or 1	0.007	0.000	0.000	0.000	0.082
Paid leave preemption	EPI	Indicator, 0 or 1	0.016	0.000	0.000	0.000	0.125
Average state legislative majority ideology	ShMc	Common Space	0.033	−0.708	0.043	0.740	0.780
Average state legislative chamber ideology	ShMc	Common Space	0.062	−0.414	0.187	0.577	0.591
Legislative professionalism	Squire	Fraction	0.196	0.116	0.169	0.232	0.121
Folded Ranney index	BH	Fraction	0.842	0.789	0.838	0.902	0.083
Holbrook and Van Dunk index	SLER	Percentage	39.913	31.527	40.500	48.564	11.590
Union membership density	CPS	Fraction	0.117	0.069	0.109	0.158	0.057
Population	Census	1000s	5802.545	1570.746	4000.591	6695.497	6501.792
Personal income, per capita	REIS	Ratio	42.945	37.258	41.400	47.385	8.331
Population density	Census	Ratio	379.607	42.018	97.359	219.250	1374.946
Urban population	Census	Fraction	0.719	0.588	0.738	0.874	0.196
Population aged 65 & over	SEER	Fraction	0.134	0.121	0.133	0.146	0.022
Population aged 19 & under	SEER	Fraction	0.273	0.259	0.273	0.287	0.025
College degree (%)	CPS	Fraction	0.236	0.195	0.226	0.268	0.060
Ethnic fractionalization index	SEER	Fraction	0.281	0.164	0.278	0.407	0.139
% of neighboring states with any preemption	EPI	Fraction	0.246	0.000	0.167	0.400	0.289

Notes: EPI, Economic Policy Institute; ShMc, Shor and McCarty; Census, Census Bureau; REIS, Regional Economic Information System; SEER, Surveillance, Epidemiology, and End Results program; Squire, Squire, multiple years; BH, Bibby and Holbrook; SLER, Klarner, State Legislative Election Returns; CPS-ASEC, Current Population Survey—Annual Social and Economic Supplement. Data are for forty-nine states from 1993 to 2018 excluding Nebraska.

Other Data

Until relatively recently, data on state legislative ideology was severely lacking. The work of Shor and McCarty (2011) remedy this issue. They provide state legislator ideal points based on state legislative roll call voting and the Project Vote Smart National Political Awareness Test (NPAT). The data are available for the years 1993 to 2018.⁶ Our analysis aggregates the individual ideal points to the chamber level, and the median value is calculated. Following Shor and McCarty (2011), we average each state’s two legislative chambers’ (except Nebraska) median ideal points for the majority party of the chamber and the overall chamber⁷ and use these two variables as measures of state legislative ideology. Shor and McCarty’s (2011) ideology measure is centered on zero, with scores below zero signifying more liberal majorities and scores above zero indicating more conservative majorities. As table 1 shows, the average legislative majority has a slight conservative tilt (slightly more conservative for the average chamber ideology), but there is much variation around the mean.

We include several variables measuring political institutions. First, we operationalize state legislative professionalism using Squire’s (2017) index of legislative professionalism. The index equally weights legislator salary, number of legislative staff per legislator, and total days in session, and it is interpreted relative to congressional professionalism (Squire 1992, 2000, 2007, 2012, 2017). On average, professionalism is low at approximately 20 percent of congressional professionalism. Second, we include the competition between parties for control of the state government using a folded Ranney index as calculated in Bibby and Holbrook (2004) with a four-year moving average.⁸ A value of 0.50 indicates complete one-party dominance; single-party control decreases as the index approaches one. On average, the index takes on a value of 0.84, indicating far less than perfect single-party control. Lastly, we include the Holbrook and Van Dunk (1993) index of legislative electoral competition. The index is a composite of four factors: the winning percentage of the popular vote, the winning candidate’s margin of victory, an indicator of whether the district is “safe,” and an indicator of whether the district is contested. The index is calculated at the legislative district level and aggregated to the state. A zero value indicates no competition and increases to a theoretical maximum of 100. As seen in table 1, the average state-year is 39.9, suggesting a moderate level of electoral competitiveness.

Union membership density is an important proxy for the power of relevant interest groups. Many of the potentially preempted policy issues are important to union members, and unions may exert lobbying pressure to prevent their preemption. The variable is supplied by Hirsch and Macpherson (2003) and is extracted from the Current Population Survey, Annual Social and Economic Supplement.

Our fourth category of independent variables is demographic factors. We include racial heterogeneity of the population, operationalized following Alesina, Baqir, and Easterly (1999), as the probability that two randomly drawn individuals belong to two different ethnic groups.⁹ We include two variables for the size and/or prevalence of urban populations. The first is population density. The second is the percentage of the state population living in urban areas (as defined as living in a metropolitan county). These two variables are similar but measure different concepts. Density measures the prevalence of urban areas in a state. As average density increases, more of the physical landscape of the state is urban. Urban population measures how much of the state population lives in these areas. Rural states typically have low average density but a high percentage of urban residents. Given the discussion above about the importance (or lack thereof) of urban legislative delegations, these two variables provide important controls for the ability of state legislators representing cities to fend off preemption.

Our final key independent variable is the number of neighboring states with a workers’ rights preemption. Specifically, this variable is measured as the percentage of geographically neighboring states having any workers’ rights preemption. Following Berry and Berry (1990), we use this to measure policy diffusion.

In addition to the measures for our five hypothesized explanations for workers’ rights preemptions, we include several control variables. Population is included to control for the size of state. Personal income per capita is included to control for potential wealth effects. We include both the percentage of residents who are under the age of 19 and over the age of 65. The prevalence of such residents may shift legislators’ time away from issues of workers and toward issues of youth or the elderly. We include the percentage of the state’s population with a college degree to control for educational endowments. Lastly, following Boehmke (2009), we condition the regressions on the number of workers’ rights-related preemptions a state has already adopted. It is unclear whether prior experience with preemption increases the likelihood of preemption in the future or reduces it. We report data sources and summary statistics for all the variables in table 1.

Empirical Strategy

We analyze the probability of adoption of any workers’ rights-related preemption. To do so, we construct a dichotomous variable [0,1] that indicates if a state has adopted any workers’ rights-related preemption in a particular year. Given the cross-sectional, time-series nature of the data and our interest in estimating within-state effects, unit-specific fixed effects are necessary to control for any time-invariant, unobserved heterogeneity (Green et al. 2001). In doing so, the baseline rate of adoption is allowed to vary by state (Boehmke 2009), something that is likely important given that some local policies (minimum wage or paid leave) are preempted by a great many states while others (fair scheduling) are not.¹⁰

The methods typically used with data such as ours are thought to suffer from an incidental parameters problem (Neyman and Scott 1948). Rather than rely on a single estimation technique, we take a varied approach. Following the advice of Beck (2018) for the estimation of binary-dependent variables with fixed effects, we use a linear probability model with state- and year-fixed effects (Angrist 2001), conditional logit (Chamberlain 1980), and logit with state-fixed effects.¹¹ We use the results from the three estimation techniques to corroborate the findings across models. To aid in comparing the three techniques, we constrain the OLS sample to those states that exhibit some time variation. As Beck (2018) explains, not doing so likely biases the $\beta$ coefficient toward zero as the marginal effects of this “ALL0” group are zero. This procedure eliminates twenty states from the dataset.¹² The OLS and logit techniques allow for the recovery of marginal effects; however, the conditional logit does not.

In further analyses, we explicitly model the adoption of multiple workers’ rights-related preemptions in any year. The count nature of the dependent variable suggests using negative binomial regression; however, including conditional fixed effects is difficult.¹³  Allison and Waterman (2002) suggest that including a $j-1$ series of state dummy variables and using outer product of gradient standard errors is a good approximation of conditional fixed effects.¹⁴ Given the structure of the data (table 1), one might expect a zero-inflated model to be appropriate. We conducted a Vuong test for each model in table 3, and there is little evidence to suggest that a zero-inflated model performs better than a normal negative binomial regression.

Table 2

Any workers’ rights preemption, 1993–2018

	(1)	(2)	(3)	(4)	(5)	(6)
Variable	OLS	CLOGIT	FELOGIT	OLS	CLOGIT	FELOGIT
Average state legislative majority ideology	0.1044**	3.3299**	0.1265**	—	—	—
	(4.10)	(3.93)	(2.64)
Average state legislative chamber ideology	—	—	—	0.1768**	6.3974**	0.2336**
				(4.66)	(4.10)	(2.62)
State legislative professionalism	−0.2727	−5.3843	−0.1881	−0.2898	−7.0349	−0.2387
	(−1.36)	(−1.09)	(0.90)	(−1.50)	(−1.37)	(1.22)
Folded Ranney index	−0.4148*	−12.7839**	−0.4839*	−0.3922*	−11.5966*	−0.4280*
	(−2.24)	(−2.85)	(2.50)	(−2.15)	(−2.52)	(2.07)
Holbrook and Van Dunk index	−0.0001	−0.0324	−0.0011	0.0000	−0.0338	−0.0011
	(−0.05)	(−0.93)	(0.72)	(0.02)	(−0.94)	(0.84)
Union membership density	0.6683	31.3218	1.1241	0.6520	32.3524	1.1057
	(0.65)	(1.65)	(1.45)	(0.64)	(1.70)	(1.45)
Population (1000s)	−0.0001**	−0.0027**	−0.0001*	−0.0001**	−0.0024**	−0.0001*
	(−3.94)	(−3.06)	(2.29)	(−4.28)	(−2.75)	(2.10)
Personal income, per capita	−0.0161	0.0539	0.0018	−0.0144	0.0944	0.0032
	(−1.87)	(0.44)	(0.40)	(−1.71)	(0.76)	(0.72)
Population density	0.0022	0.0989*	0.0038	0.0023	0.0850	0.0032
	(1.63)	(2.03)	(1.83)	(1.85)	(1.76)	(1.65)
Urban population (%)	−0.1513	0.0770	0.0115	−0.1642	−0.9741	−0.0304
	(−0.54)	(0.01)	(0.04)	(−0.60)	(−0.11)	(0.10)
Population over 65 (%)	−4.7575	46.5743	1.7118	−4.2117	50.1498	1.7692
	(−1.26)	(1.33)	(1.25)	(−1.10)	(1.39)	(1.29)
Population under 19 (%)	−0.3768	−25.1038	−1.0082	−0.9463	−28.0206	−1.1298
	(−0.14)	(−0.48)	(0.52)	(−0.35)	(−0.55)	(0.62)
College degree (%)	0.7700	9.1584	0.4123	0.7900	5.7909	0.2770
	(1.03)	(0.65)	(0.80)	(1.03)	(0.41)	(0.55)
Ethnic fractionalization index	−0.6397	61.6170*	2.2291	−0.4033	49.7329	1.6608
	(−0.76)	(2.19)	(1.80)	(−0.46)	(1.80)	(1.48)
Percent of neighboring states preempting	0.0290	1.3745	0.0415	0.0214	1.0424	0.0258
	(0.31)	(0.88)	(0.70)	(0.23)	(0.66)	(0.45)
Number of previous preemptions	−0.1151**	−2.0748**	−0.0766**	−0.1163**	−2.1912**	−0.0780**
	(−5.76)	(−5.29)	(2.84)	(−5.81)	(−5.49)	(2.80)
Observations	707	707	707	707	707	707
Number of states	31	31	31	31	31	31
Within R2	0.105	0.439	0.409	0.108	0.439	0.425

Notes: Robust t-statistics in parentheses for OLS; CLOGIT grouped by state; FELOGIT includes state fixed effects and average partial effects are displayed.

**

p < 0.01,

*

p < 0.05.

Table 3

Count models

Variable	(1)	(2)
Average state legislative majority ideology	3.0139**	—
	(3.32)
Average state legislative chamber ideology	—	5.3743**
		(3.06)
State legislative professionalism	−1.8216	−2.7394
	(−0.33)	(−0.47)
Folded Ranney index	−13.4670**	−11.2735
	(−2.68)	(−1.90)
Holbrook and Van Dunk index	−0.0094	−0.0104
	(−0.20)	(−0.21)
Union membership density	15.5634	13.1893
	(0.74)	(0.57)
Population (1000s)	−0.0026	−0.0022
	(−1.82)	(−1.35)
Personal income, per capita	0.1932	0.2219
	(1.28)	(1.38)
Population density	0.1079	0.0930
	(1.56)	(1.20)
Urban population (%)	−4.8044	−5.7575
	(−0.49)	(−0.57)
Population over 65 (%)	57.5955	62.4133
	(1.39)	(1.43)
Population under 19 (%)	−15.2943	−15.8238
	(−0.23)	(−0.22)
College degree (%)	−5.7608	−6.5912
	(−0.34)	(−0.34)
Ethnic fractionalization index	47.6238	32.1677
	(1.32)	(0.82)
Percent of neighboring states preempting	1.5346	1.0207
	(0.83)	(0.51)
Number of previous preemptions	−1.9594**	−1.9882**
	(−5.91)	(−4.54)
State fixed effects?	Yes	Yes
N	1,059	1,059

Notes: Outer product of gradient (OPG) z-statistics in parentheses.

**

p < 0.01,

*

p < 0.05.

Results

Table 2 shows the results of three regression or logit models where the dependent variable is binary, indicating one if a state had a legislative workers’ rights preemption of any kind in a given year. The first specification (columns 1 and 4) is estimated using OLS. The second specification is estimated using conditional logit (columns 2 and 5) and the third specification (columns 3 and 6) is estimated using fixed effects logit. Across all specifications, both ideology variables are positive and statistically significant, suggesting that increasing within-state conservatism in the political majority or the legislative chamber is associated with a higher risk of preemption in any given year. The marginal effects between the OLS and FELOGIT models are quite similar, with an increase of 1 unit in average ideology (neutral (0) to conservative (1)) associated with a 10–13 percent increase in the risk of preemption. When we examine the chamber ideology, the effect size roughly doubles to an 18–23 percent risk of preemption in any given year.

Legislative professionalism exerts no statistical influence on the risk of preemption. Given the within nature of our analysis, this finding suggests that within-state increases or decreases in legislative professionalism are unrelated to preemption activity, as we expected based on Kim et al. (2021). Similarly, as measured by the Holbrook and Van Dunk index, electoral competition is unrelated to preemption activity, contrary to our expectations. Lastly, an increase in the folded Ranney index, indicating increasing competition for control of the state government, is associated with a lower probability of preemption. These latter two results suggest an interesting dynamic. Electoral competition does not appear to alter the probability of preemption; however, competition among political parties for control of the government does, with lower levels of political control leading to fewer preemptions. This result is similar to Barber and Dynes (2023), who assert that politically unified states are more likely to preempt. To test further whether single-party control drives the results, we unfolded the Ranney index, effectively measuring the strength of Democratic Party control in state government, and re-ran the analysis. Across all models and specifications, the results were statistically indistinguishable from zero while preserving the main ideology finding.¹⁵ This result boosts our confidence in the main finding that competition, regardless of party, for government control contributes to a lower likelihood of preemption of local workers’ rights laws.

Contrary to our expectations (and prior findings), state unionization density is statistically unrelated to the within-state probability of preemption. While Kim et al. (2021) ascertain that high unionization states are less likely to preempt, we find no evidence that within-state changes in unionization rates are associated with preemption. Taken together, these results suggest that while high unionization rates may protect workers against preemption, increasing unionization rates likely does not confer similar protections.

Demographic characteristics, as measured by population density, percent of the urban population, and ethnic fractionalization, are not statistically associated with the likelihood of preemption. Likewise, we observe no evidence of policy diffusion through the influence of neighboring states preempting workers’ rights. In general, larger states are less likely to preempt as are states who have previously preempted in this policy area. The remaining control variables are largely unrelated to the probability of preemption.

Count Model

We extend the analysis to relax the aggregation of all preemptions into a single dichotomous variable. Rather than indicating any preemption in a given year, the number of preemptions by state-year is the dependent variable.

Table 3 shows the results of two negative binomial regression models where the dependent variable is the count of legislative preemptions of any kind in a given year. The first specification includes the average state legislative majority ideology, and the second model includes the average state legislative chamber ideology. In both models, ideology is a positive and statistically significant predictor of the count of legislative preemptions. Negative binomial coefficients may be viewed as semielasticities; therefore, a one-unit increase in averaged state majority ideology is associated with 301 percent more preemptions, all else equal.¹⁶ Given that the median number of preemptions is zero, this large coefficient makes some sense. The effect size of the second specification is not quite double that of the first. Taken together, increasingly ideologically conservative state legislatures are associated with an increase in the count of legislative preemptions. Like the prior results, increasing competition for control of the state government is associated with a decrease in the predicted preemption activity. A 0.1 unit increase in the folded Ranney index is associated with a 134 percent decrease in preemption activity. This finding is fragile, with only the results from model one being statistically different from zero. Like the primary results in table 2, the remaining political institutions variables, unionization, demographic factors, and policy diffusion do not influence preemption activity.

Discussion and Conclusions

The intent of this analysis is to examine why states preempt their local governments in the area of workers’ rights. Our cross-sectional, time-series approach aligns with nascent literature that introduces the element of time (Weissert et al. 2021) to the study of state preemptions (Bucci and Jansa 2021; Goodman and Hatch 2023), allowing us to explain factors associated with changes in preemptions over time. Using data from the Economic Policy Institute on five workers’ rights policies, we tested several hypotheses associated with preemption adoption: ideology, political institutions, interest groups, demographic factors, and policy diffusion. We find support for our hypotheses that more ideologically conservative state legislatures will preempt more often, and less politically unified state governments will preempt less often. This is consistent with previous narrative and cross-sectional studies that emphasize the correlation between Republican and unified control and preemption of workers’ rights laws (Bunch 2021; Flavin and Shufeldt 2020; Riverstone-Newell 2017). However, our time-series analysis confirms changes in ideology are associated with workers’ rights preemptions, rather than the previous understanding that current party control correlates with all existing workers’ rights preemptions at a given time. We also find support for our hypothesis, consistent with Kim et al. (2021) and Flavin and Shufeldt (2020), that legislative professionalism is not statistically associated with workers’ rights preemptions. In contrast, Goodman and Hatch (2023) and Fowler and Witt (2019) observe correlations (although in opposite directions) between affordable housing preemptions and preemption in general, respectively, and legislative professionalism, providing further evidence that factors associated with preemption adoption likely vary by policy area (Grossman 2013; Goodman and Hatch 2023).

Our remaining hypotheses are not supported. Based on previous work by Bucci and Jansa (2021) and Kim et al. (2021), we expected changes in unionization rates to explain workers’ rights preemptions. High unionization rates may be protective against preemptions, but changes in unionization rates do not influence the likelihood of a new preemption or unionization works in different directions in different states—as a protective factor against preemptions in some states and an incentive to preempt in others. Further research is needed to explore the relationship between unionization and the preemption of workers’ rights. Despite evidence of state-level labor laws diffusing across jurisdictions (Pugh 2012), we find no evidence of preemption diffusion among neighboring states. This may be because preemption is aimed at stopping local policy diffusion (Mallinson 2021), which we do not measure due to data limitations, or because the appropriate “neighbor” is ideological rather than geographic (Grossback, Nicholson-Crotty, and Peterson 2004). Future research could explore both of these hypotheses with the addition of local-level policies and measures of preemption adoptions by ideological neighbors.

Our results appear to conform to the popular narrative of more ideologically conservative state legislatures preempting policies of more ideologically liberal city governments. The findings here add to the recent work by Swanson and Barrilleaux (2020), who observe a similar result for court-driven preemption, and Goodman and Hatch (2023), who study the preemption of local affordable housing laws. Ideology appears to play an important role in the decision to involve the state in the affairs of local governments in recent decades. Using ideology rather than partisanship allows us to be more precise. If Republican legislatures drive preemption, as prior analyses contend, why does Tennessee preempt when New Hampshire does not? Both are Republican-dominated legislatures; however, they are ideologically distinct—Tennessee is more ideologically conservative than New Hampshire. Ideology, rather than partisanship, may explain why some studies observe that Republican state legislatures are more likely to have workers’ rights preemptions (Bucci and Jansa 2021; Bunch 2021; Flavin and Shufeldt 2020) while Kim et al. (2021) do not find the same relationship. Including ideology, rather than partisanship, in models explaining state preemptions may help future studies explain these apparent inconsistencies and anomalies.

Our analysis is not without limitations. Given our data, it is not possible to measure the ideological distance or widening of such distance between state legislators and local officials. Barber and Dynes (2023) demonstrate that this distance is an important predictor of perceived preemptions across several policy areas. Swanson and Barrilleaux (2020) reveal a similar result with court-driven preemptions. Combining these two approaches, observed preemptions and ideological distance, would be fruitful for understanding the relationship between state and local policymakers and state preemption. Another potential limitation is our preemption data’s relatively narrow, policy-specific nature, which somewhat restricts the generalizability of the findings to other policy areas. However, we take Goodman and Hatch’s (2023) approach that scholars should examine preemption by policy area because it is likely that the factors that influence preemption differ by policy area.

Local governments rely on their relative autonomy to provide local solutions to local problems. States are well within their Constitutional powers to adjust the level of autonomy provided to local governments; however, some stability in autonomy is required. If, as the results of this analysis suggest, more conservative state legislative ideology is an important predictor of preemption activity, it is possible for these preemptions to be undone with changes in ideology. If local autonomy begins to change with state election cycles, local elected officials and administrators will have difficulty planning for future events. Local government provides public services that, by their nature, require stability—services like police and fire departments, public parks, and schools. Related specifically to this analysis, local workers’ rights ordinances require businesses to make changes to accommodate new regulations. Uncertainty over whether local business regulations will change could significantly alter business decisions and ultimately harm local economies. Likewise from an advocate perspective, the threat of preemption combined with the varying policy approaches across jurisdictions makes it difficult to build and coordinate workers’ rights (Johnson 2021). Michener (2023) observes that housing preemptions cause local organizations to change their policy goals, which can be particularly harmful for race-class subjugated communities. This should be a concern for workers’ rights policies as well because they are most likely to help these same communities. While there is hesitancy to the idea (Johnson 2021), workers’ rights advocates may want to consider taking a federal approach instead in order to ensure greater policy stability.

The findings presented in this study support our contention that preemption should be analyzed by policy area. While ideology and political control of the government affect workers’ rights preemptions [or unionization and state minimum wages, see Kim et al. (2021), or Republican trifectas, see Flavin and Shufeldt (2020)], interest groups appear to dominate public health-related preemptions (Pomeranz and Pertschuk 2017), and legislative ideology, legislative professionalism, and renters (an interest group) are influential for affordable housing-related preemptions (Goodman and Hatch 2023). Some of this variation results from methodological differences: some of the referenced studies use a cross-sectional approach to analyze the cumulative number of preemptions as a function of various factors at the endpoint of the analysis. Our study argues strongly that a specific piece of preemption legislation arises due to numerous factors at a particular point in time. Future research should reflect this temporal reality of the legislative process. Such an approach will require the collection of comprehensive legislative histories of preemptions, data that currently exist only in an ad hoc manner. In addition, varying explanations for preemption are also likely to arise when including policy area-specific covariates. Aligning preemptions with time-specific covariates and a policy area-centric approach can lead to more nuanced understanding of factors that lead to legislative preemption than what currently exists in the literature.

Acknowledgements

We thank the three anonymous reviewers and John Dinan for their helpful comments and suggestions. We also thank Bruce D. McDonald for his unwavering encouragement.

Endnotes

1

While business interest groups are undoubtedly important, we have little way of including their influence over such a long period of analysis. However, we contend that the inclusion of prior preemptions in this policy area may be an indirect proxy for the influence of such organizations.

2

A sixth policy, gig economy, was added after we conducted this analysis. We chose not to add that type of preemption into our analysis because that area is more akin to restrictions on the sharing economy generally rather than workers’ rights specifically.

3

In an appendix, Bucci and Jansa (2021) most closely replicate our analysis by examining the relationship between Berry et al.’s (2010) government ideology and restrictive state labor laws. While direct comparison is impossible because of methodological differences in variable construction, these results are broadly supportive of our findings, albeit measuring different concepts (overall government ideology rather than legislator ideology and an index of various state labor policies rather than preemption specifically).

4

This necessarily limits variation in the preemption activity of a state. Theoretically, a state may have as many as five preemptions in a given year. This approach is nearly identical to the pooled components analysis in Boehmke (2009), allowing repeat adoption of similar components. Relaxing this restriction to examine the count of adoptions in any given state-year is explored in a later section.

5

There is some indication this spat of paid leave preemptions is tied to Wisconsin Bill 23, passed in 2011, to preempt the City of Milwaukee’s local paid leave ordinance passed by local referendum in 2008. The text of this legislation was distributed by ALEC (though it was never officially part of the group’s model legislation) and may have been adopted, in part, by numerous other states (Grabar 2016).

6

Some state ideology data are missing in the early parts of the time series. The missing data reduces the analytical dataset from a theoretical size of 725 observations to 707. The results presented are robust to choosing various new starting years that eliminate the missing data problem.

7

We use both the majority ideology and chamber ideology because the former excludes any possibility of bipartisan preemptions.

8

A folded Ranney index is calculated as $1-|0.5-\mathrm{Ranney}\mathrm{ }\mathrm{Index}|$⁠. Longer moving averages (6 year and 10 year) were tested and there is no appreciable difference in results.

9

Ethnic fragmentation is calculated as follows: $\mathit{ETHNIC}=1-\sum _i(RAC{E}_i{)}^2$ where $\mathit{Rac}{e}_i$ denotes the share of population for race $i$ where $i=\{\mathrm{white},\mathrm{ }\mathrm{Black},\mathrm{ }\mathrm{Native}\mathrm{ }\mathrm{American},\mathrm{ }\mathrm{and}\mathrm{ }\mathrm{Asian}\}$⁠. See Alesina, Baqir, and Easterly (1999) for further explanation of the construction of this variable and Vigdor (2002) for a theoretical perspective.

10

The results are robust to the exclusion of the baseline.

11

The latter is estimated in Stata using the logitfe command and the analytical corrected estimator. Average partial effects are displayed throughout.

12

The OLS results presented in table 2 are qualitatively robust to the inclusion of all states. As Beck (2018) predicts, the effect size on the two primary variables of interest are biased downward. The effect size of average state legislative majority ideology is lower by roughly 0.04 units and statistically significant at the 0.01 level. The effect size of average state legislative chamber ideology lower by roughly 0.05 units and statistically significant at the 0.01 level.

13

Fixed-effects Poisson regression is also an option; however, the LR test from table 3 suggests over-dispersion is a problem for all four models. Therefore, the estimation continues with negative binomial regression.

14

This is easily implemented in Stata using the nbreg command with the vce(ogp) option.

15

Full results are available upon request.

16

The incident rate ratio, $e^{\beta }$ is 20.37 suggesting relative to a state with a legislative ideology of 0, a state with a legislative ideology of 1 will have 20.4 times the incident rate of preemptions.