Abstract
The COVID-19 pandemic struck during a period of extreme polarization in American politics. Unsurprisingly, responses to it quickly became politicized despite increasingly clear findings from scientific and public health communities about the most effective approaches for limiting its spread. We ask how the politicization affected pandemic response at the state level. We document and explain several kinds of state-level actions, beginning with 2020 variations in collecting and publishing COVID-related data and early mitigation strategies. We find that state capacity explains the former and partisanship the latter. We show that divisions within the Republican Party also meaningfully affected state responses. Inter- and intraparty divisions—rather than geography or severity of COVID—in fact continue to influence state policy following the inauguration of President Joe Biden, the availability of vaccines, and the rise of the Delta variant. These findings document that U.S. federalism often created obstacles to effective governmental responses.
The COVID-19 pandemic hit the United States in waves. As the life-threatening virus spread around the world starting in early 2020, American communities, states, and regions experienced the dire consequences in uneven phases and from different standpoints. Citizens and leaders dealt with health and economic effects at different moments and leaders often disagreed about what mattered most. Nevertheless, as most experts and observers realized soon enough, only a nimble coordinated response could have saved the United States from the outsized COVID-19 death trends that unfolded over the course of 2020 and into 2021. National coordination informed by honest data was needed to ensure rapid testing, deploy and shift health resources, and implement mitigation efforts while limiting damage to the economy and schools. Later, when new COVID-19 vaccines were approved at the end of 2020, further rounds of coordinated effort were required to get them into American arms as quickly as possible.
To be sure, the United States has a federal system of government, in which responsibilities for policymaking and implementation are divided among at least three levels of government: national, state, and local. Responses to big crises such as the COVID-19 pandemic can involve national authorities claiming emergency powers to force uniform actions at lower levels. More often, U.S. federal government authorities orchestrate cooperation that includes allowances for varied policies or processes of implementation by state and local authorities serving diverse sets of citizens and institutional stakeholders. However, even a cursory overview of what actually happened in the United States during 2020 and 2021 underlines that neither type of theoretically possible national response occurred.
The global pandemic spread into the United States during 2020—at a politically fraught juncture. Controversial incumbent President Donald J. Trump, ever distrustful of expert advice on any matter, including public health, soon became obsessed with the impact of the unfolding crisis on his reelection prospects. After a brief effort to act nationally, he and his closest advisors took over messaging from the Centers for Disease Control and Prevention (CDC) and other federal health authorities. By April, the administration decided to downshift responsibility for economically disruptive public health measures to the states. This was an abdication of federal responsibility to force uniform national responses or orchestrate smooth federal cooperation––especially because the Trump White House did not just step back. Instead of leaving state authorities, especially governors, entirely on their own, the Trump administration both dumped policymaking responsibilities on them and simultaneously began attacking the decisions made by Democratic governors and mayors (Shear and Mervosh 2020). Nor were many Republican governors allowed space to devise their own consistent strategies, because Trump himself often intervened to play one branch of state government against the other or encourage localities or businesses to strike out on their own (Olorunnipa, Witte, and Bernstein 2020). From 2020 forward through 2021, disagreements grew and intensified across the multiple tiers of the U.S. governmental system inflected by clashing understandings and priorities of the pandemic response. No sustained federal takeover imposed a disciplined response; no coherent federal policy orchestrated cooperative state and local actions. Instead, state authorities, governors, and legislatures and sometimes state courts were forced to jump in where federal authorities faltered or displaced responsibilities.
How to make sense of America’s subnational COVID-19 responses undertaken by the fifty states is the focus of this article. In the following sections, we document and seek to explain several kinds of state-level actions during successive phases of responses to COVID-19.
THEORETICAL INSIGHTS FROM POLITICAL SCIENCE LITERATURES ON FEDERALISM
Policymaking for the COVID-19 pandemic is not, of course, the only time U.S. federal divisions of public authority have shaped decisions and results. Under the Constitution, many aspects of public policy have always been up to the fifty states and sometimes to localities within them. Scholars have long analyzed federal divisions of labor and dynamics of cooperation or disjuncture. Further, political scientists have long studied the dynamics by which new laws or regulations diffuse––that is spread (or do not spread)––from state to state (for an overview, see Karch 2007). Especially in the early twentieth century, many nationwide policy innovations happened through just such a process––as one state copied another until all or most had adopted a given measure, such as early state laws to increase high school graduation rates (Goldin 1998), early workmen’s compensation laws (Guyton 1999), and early welfare benefits (Skocpol et al. 1993). Even after the New Deal and the later Great Society, when the federal government became more actively involved in funding, coordinating, and uniformly legislating major interventions into the industrial economy and social welfare provision, states have continued to play major roles in implementing and co-funding critical public programs, as can be seen in the diffusion of auto safety laws (Wagenaar, Maybee, and Sullivan 1988) or preschool (Karch 2013).
Scholars who recognize these realities of U.S. public policymaking and implementation have studied many explanatory factors to make sense of variations in the timing and content of state choices in adopting new programs––either programs adopted in parallel by states learning from others, or programs that the federal government helps fund while leaving key decisions about adoption or administration to the states. Researchers studying state to state diffusions of innovations or federally encouraged state policy adoptions or refusals have identified key causal variables. We might expect governments to act sooner and citizens to respond more collaboratively if a given problem––such as COVID-19 cases or deaths––is more acute in their jurisdiction (Elcheroth and Drury 2020). Scholars have also shown that state actions can be influenced by economic factors or fiscal resources or by existing institutional governmental capacities (Capano et al. 2020). That could mean that wealthy states, or states with solid public finances, or states with strong data-collection or public health capabilities would have responded differently than others to the pandemic. Additional lines of argument in classic literatures probe for regional effects: perhaps nearby states influence one another in adopting a policy response (Berry and Berry 1990; Glick and Rose 1999). More recent literature suggests that it is not proximity but instead shared ideology and culture among states that encourages policy diffusion (Butler et al. 2017; Karch et al. 2016).
All of these ideas consider factors inherent to each state’s particular internal situation or neighborhood ties. Other theories in the literature consider whether individual states’ relationships to the federal government, including ties to agencies or copartisan ties to the presidential or congressional authorities, might be more important than severity of difficulties or comparative adequacy of state-level resources. The COVID-19 pandemic presented a host of unprecedented conditions for American politics: a global pandemic during a period of intense political polarization, vitriol, and stalemating under the direction of a president with authoritarian tendencies. Given this context, we began this research with few favored explanations and instead sought to test the validity of these existing theories from the literatures on federalism and state politics. Although all of the previous approaches for examining policy responses across states and within U.S. federalism suggest factors we can use to characterize and explain America’s state-varied responses to the COVID-19 pandemic, we show that traditional perspectives are insufficient because they pay too little attention to extreme party polarization in the current era, and because they say next to nothing about factional struggles and clashes of ambition among politicians using state-level platforms to define national profiles.
We have defined and tested a range of possible explanatory variables that might explain differences in early state-level data collection and dissemination of plans, initial pandemic mitigation measures, and eventual stances after vaccination becomes possible amid a new surge of Delta variant infections. In each successive analysis, we introduce our dependent variable definitions and measures along with the sets of possibly relevant explanatory factors we examine for each outcome. When we held clear expectations for a particular outcome, we state as much within each section. In all of our explanatory models and accounts, we probe relevant political or institutional factors across the states in the context of appropriate economic and demographic controls (for a full list of variables, see table A.1).1 We describe more specifics as we introduce each model, but our guiding principle was to use a cumulative COVID-19 case rate per ten thousand residents measured the month before the collection of the outcome variable.2 In other words, if our outcome was measured as of July 2020, then we use case rate data as of June 1, 2020. Although our results are robust to multiple specifications of the COVID-19 case rate, we use this approach because it makes the most intuitive sense that public officials consider recent trends in the virus’s spread when making policy decisions. Depending on the nature of the dependent variable, we run either ordinary least squares regression or logistic regression models. Full results are included in the appendix. For the binary outcome variables, we report the regression results in the appendix and discuss the average marginal effects in the text.3
We now move on to flesh out our successive sets of dependent variables and explore relevant explanatory factors for successive phases of U.S. state-level responses to the COVID-19 pandemic. We examine state response in three phases. First, we study state approaches to collecting and publicizing data early in the pandemic (March–July 2020), when public health experts and elected officials alike were scrambling to understand the virus and its impact. We turn to the second phase (March–December 2020), when the consensus on effective mitigation strategies was clear, to understand which states were more likely to implement research-backed public health policies. 4 In this phase, we also examine legislative responses to gubernatorial powers. In the third phase (January–December 2021), we examine initial mitigation strategies and conclude with a discussion of state actions during 2021, after vaccines became available and partisan control of the federal government changed. In the discussions to come, we report many null findings, patterns that do not fit inherited social science expectations about U.S. federal responses to crisis. Basically, we find that the severity of the COVID-19 crisis is not associated with the speed or intensity of policy response in either 2020 or 2021. Existing institutional capacities appear to have had a modest influence on early state-level tracking of the epidemic.
Partisan divides and shifting partisan dynamics through presidential regimes have been especially decisive. The role of party confirms existing findings from the state politics and federalism literatures on the filtering power of ideology in policy adoption. However, we refine the existing emphasis on party—it is not just party label but alignment with Donald Trump that best predicts state-level response during both the Trump (March–December 2020) and early Biden presidencies (January–December 2021). Those dynamics, moreover, turn out to be more than just Democrats versus Republicans, because on the Republican side intraparty factional differences have also influenced state-level responses, especially in 2021. These conclusions broadly echo findings elsewhere in this issue that the composition of state government at the outset of the pandemic influenced subsequent political and policy outcomes (Evans et al. 2022, this issue).
HOW STATES GATHERED AND PUBLICIZED DATA (MARCH–JULY 2020)
Public health officials in America’s fifty states initially had limited knowledge of COVID-19 and thus scrambled to understand and communicate its contagiousness and modes of spread and devise reasonable mitigation strategies and effective therapies. Federal experts in the CDC did offer guidance, but their messages were often disrupted or muted by the Trump administration (Olorunnipa, Witte, and Bernstein 2020), leaving state governments to meld general federal guidance with their local context and preferences to determine appropriate responses to the pandemic.
Although states vary in their capacities to collect data (Brambor et al. 2020; James 2022), state-level officials had no choice but to collect and disseminate information on COVID-19 from early on in the crisis. Proving real-time information useful to citizens, institutions, and local authorities became an essential function of state government. We expected that some existing capacity for data collection would facilitate rapid response to the COVID crisis. Even so, given the antiscience stances many Republicans trumpeted in the Trump era, it is reasonable to consider whether the nature and timing of data collection and dissemination depended on a state’s ideological and partisan leanings. Fortunately, we can track information on such state activities using an internet archive, The Wayback Machine, and archives of gubernatorial press releases. Using these sources, we developed three measures, the first about the publication of a statewide county-level dashboards tracking of cases and deaths, the second about the publication of long-term reopening plans, and the third about the commitment to contingent economic reopening plans with clear, health-related thresholds for moving from one phase of reopening to the next.
County-Level Dashboards
By the six-month anniversary of virus’s arrival in the United States, all fifty states had a regularly updated and publicly available dashboard documenting trends in COVID-19 cases and deaths, but the speed with which these dashboards became available varied. We documented the publication date for each state’s COVID dashboard using The Wayback Machine (Internet Archive, n.d.). This gave us the date and time that each dashboard’s URL became public (see figure 1).
Month of URL Publication for Public COVID Dashboard
Source: Authors’ calculations based on data from The Wayback Machine (Internet Archive, n.d.).
Four states—Alaska, Mississippi, South Dakota, and Virginia—published the first dashboards on March 6, 2020, forty-four days after the first documented case in the United States. Most states published their dashboards in April and the last ones, from California and Maine, in late June 2020. The average state published about 2.5 months (eighty days) after the country’s first documented COVID-19 case.
Long-Term Reopening Plans
Comparing state data collection and publication early in the pandemic offers some insight into the type and quality of information available to officials making decisions. We also wondered about the extent to which public officials systematically incorporated this information into a coherent pandemic response plan. The publication of long-term reopening plans offers some evidence of whether public officials anticipated using data to respond to the pandemic.
Many governors outlined long-term reopening plans within the first six months of the pandemic. These plans outlined phases of reopening state economies, guidance for businesses, and suggestions for guiding the public, industry, and government through the uncertainty of the pandemic. We counted a state as having a long-term plan if the state website linked to a standalone document outlining reopening (and in some cases, reclosing) phases, along with guidance on how to partake safely in public life during each phase. Twenty-six states had published a long-term plan by July 2020 (see figure 2). Interestingly, many of the plans reflected state motifs, themes, and traditions in their title, text, and graphic design (see table B.1).
States with Advertised Long-Term Reopening Plans, July 2020
Source: Authors’ calculations based on state websites.
Contingent Reopening Thresholds
Some governors committed to progressing only through reopening phases when key measures—such as testing positivity rate, hospital capacity, ICU bed availability, and number of new cases per day—reflected a lower risk of contracting the virus and a high probability that the health system could handle any new cases. We call these contingent reopening thresholds or contingent thresholds. For example, in June 2020, Rhode Island committed to moving to the next phase only in the event of a “14-day downward trend in the number of cases OR a 14-day stable trend in declining hospitalizations” (Rhode Island 2020, 11). The Rhode Island plan, like other states with thresholds, also insisted on having at least 30 percent of ICU hospital bed space available before progressing to the next phase. Theoretically, committing to specific thresholds for making substantial public policy decisions (such as opening certain businesses or allowing interstate travel without quarantines) without the knowledge of when the state will reach them offers an unusual example of public officials constraining their future selves to data-driven decision-making. Advertising the objective thresholds to induce compliance with pandemic measures creates an agreement (albeit informally) between the public and public officials regarding the prioritization of economic and health outcomes.
Twenty-six states advertised contingent thresholds for at least some portion of their reopening phases (see figure 3). Although this is the exact same number of states that published long-term plans, the two lists are not perfectly aligned (see table B.1). Some states, such as Texas, published detailed reports outlining the guidance for reopening businesses and government but did not commit ex ante to metrics that would trigger progression or regression through the opening phases. Others, such as California, did not publish a long-term plan but did put reopening thresholds on its website indicating when the governor would consider moving the state to the next phase of reopening.
States with Advertised Contingent Reopening Thresholds, July 2020
Source: Authors’ calculations based on state websites.
Explanations of Dashboards, Plan Publication, and Thresholds
What explains these state-level variations in data tracking and publication of reopening plans? Our mapping and multivariate models controlling for region do not suggest any consistent or coherent regional clustering with publication of data. Partisan leanings and divides could, of course, have played a role. Beyond politics, however, the states’ existing capacities for data collection and analysis may have influenced responses (James 2022). Specifically, the collection and publication of data could be the result of state-level infrastructures and public health system resources before COVID-19. We test both sets of possible explanations and find that partisanship does not account for variations in data collection and publication of data and plans (for the full results and models, see tables D.1–D.10).
To measure variations in dashboard publication, we count the time to publication as the number of days from the first documented case in the country—January 22, 2020—to the first date the URL existed according to the Wayback Machine. For each additional health agency, it took an additional three days to publish the COVID dashboard, controlling for COVID case rates as of April 1, 2020, though this variable only reaches statistical significance at the 0.1 level. No relationship is apparent between party control of the executive or legislative branches and the time to dashboard publication. These results may suggest that coordinating the data collection and the technology to regularly update a county-level dashboard was facilitated by more centralized health agencies. Larger numbers of health agencies may have slowed responses. The null results for partisanship (either of the governor or the legislature) may reflect the insulated nature of those charged with collecting and maintaining public health dashboards; alternatively, they could reflect the fact that dashboards were created early in the pandemic, too early for the associated data collection to have been radically politicized.
Next, we turn to explaining reopening plans as published by July 2020 and include case rates as of June 2020, just before publication. Contrary to expectation, we do not find any partisan or bureaucratic explanation for which states published long-term reopening plans. Given the relatively low political cost and potential political benefit of publishing a reopening plan, both Republican and Democratic governors may have viewed published plans as a valuable signal that their administrations had roadmaps for emerging from the pandemic. Of course, the detailed content of the long-term plans from governors of different party affiliations diverge in ways our overall variable coding does not capture. In this study, we did not systematically evaluate plan contents, but doing so in the future may be worthwhile to better understand how governors of different partisan stripes described the pandemic and explained their approaches to addressing the many economic and social challenges it created.
Finally, we turn to findings about state advertisements of specific economic reopening thresholds during 2020. Beyond simply posting a general economic reopening plan, governors could choose to invoke and publicize data-based, contingent reopening thresholds. However, choosing to make such specific commitments posed political risk for state leaders. Governors’ responses to future pandemic developments could be constrained by specific data markers that supposedly had to be met before businesses could expand their operations, and specific commitments would also expose governors to political controversy. Although such actions can signal that the executive is observing public health guidance and using data as the primary indicator for evaluating conditions safe for reopening (potentially useful political signals), these commitments also narrow future options and gamble with shifting political pressures.
As it turns out, our statistical exploration shows that the number of health agencies in states, but not the governor’s party, was positively associated with the publication of contingent thresholds as of July 2020, controlling for case rates as of June 2020. Each additional health agency in a state is associated with a 6 percent increase in the probability that the governor published contingent thresholds.5 We cannot be sure what to make of this finding, but it could be that more agencies created greater pressures on governors of many political stripes to leverage public health data in announced choices for regulating economic activity during such an obvious public health crisis. Interviews with state-level officials could illuminate the mechanisms behind this finding.
STATE MITIGATION RESPONSES (MARCH–DECEMBER 2020)
Next, we turn to major efforts state authorities made to mitigate viral spread and illnesses during 2020, when vaccinations were not yet available and mitigation measures were paramount. State executive branches were responsible for implementing any emergency measures deemed necessary to contain the spread of COVID-19. All U.S. governors have some capacity to invoke additional powers during declared emergencies, and all fifty of them had in fact declared a state emergency by May 2020 (FEMA 2021). Many of these emergency measures directly impacted lives and livelihoods of state residents. We examine how, in the absence of federal action, governors varied the implementation of the two main strategies to limit the spread of COVID-19––specifically, mask mandates to require people to wear masks in public and stay-at-home orders to limit travel, congregation, and public activities for all except those performed by designated essential workers.
Mask Mandates
As the pandemic began raging, medical professionals and national public health officials were sending mixed signals about civilian mask wearing. Although scientists knew airborne particles were at least one method of spread, Dr. Anthony Fauci and others were concerned about the supply of personal protective equipment for medical professionals, and thus were not originally encouraging mask wearing among the public. By April 3, 2020, however, the CDC clarified its position and recommended cloth masks for civilian use (see Brewster 2020). Shortly thereafter, mask wearing became politicized, as President Trump refused to wear one (Victor, Serviss, and Paybarah 2020) and other conservative nonexperts asserted that requiring people to do so was an unnecessary infringement on personal liberty and an inappropriate response to an overblown threat of a virus no more dangerous than the flu (Chiu 2020).
By July 2020, mask wearing was an encouraged mitigation strategy, and the CDC and a plethora of other public health officials repeatedly argued that masking was one of the simplest and most effective strategies for preventing the spread of the virus. Yet despite this consensus, fewer than half of governors chose to implement statewide mask mandates (see figure 4). Coastal states with high-density cities, like New York City, Boston, Seattle, and several places in California, were among the first to experience crippling rises in cases that far surpassed the availability of medical equipment and hospital capacity. The initial clustering of mask mandates along the coasts largely aligns with these patterns of early case counts.
Statewide Mask Mandates
Source: Authors’ calculations based on NYT Reopening Tracker (New York Times 2020).
However, by the fall and winter of 2020, the crisis had seeped into the middle of the country, and Midwestern states began experiencing substantial increases in cases and deaths. One year into the pandemic, many more states had implemented mask mandates. However, more than ten states never implemented a statewide mandate in 2020, including Florida (Weber 2021) and Oklahoma (Jones 2021), which have had substantial spikes in COVID case counts and deaths.
Stay-at-Home Orders
Facing mixed compliance with mask wearing and, again, uncertainty about the mechanisms for the virus’s transmission, local and state officials saw keeping people in their homes as an effective strategy for limiting its spread. Many governors also implemented stay-at-home orders, imposing curfews, business closures, and limits on public gatherings. As of December 2020, all but seven governors had implemented a statewide stay-at-home order at some point as part of their emergency response (see figure 5).6 On average, the statewide stay-at-home orders lasted sixty days. Mississippi, at twenty-four days, had the shortest, and California holds the record for the longest, at 259 days, closely followed by New Mexico at 251 days.
Length of Stay-at-Home Orders, 2020
Source: Authors’ calculations based on NYT Reopening Tracker (New York Times 2020).
What Explains Cross-State Mitigation Efforts
Mask mandates and stay-at-home orders imposed noticeable changes and costs to state residents’ daily routines, economic opportunities, and even mental health. We might therefore expect such measures to be salient and garner more reaction from the public, making such moves more politically risky and controversial along partisan lines than simply posting dashboard data or advertising reopening plans or contingent thresholds. Thus we expected to find more clear partisan patterns to the implementation of mitigation strategies relative to those of data collection and publication.
We use various measures of partisanship and party loyalties in this article. Obviously, we code the partisan identity of the governor and the make-up of state legislatures. Beyond that, however, we account for a core reality of the present juncture––that the Republican Party, especially, is riven by growing internal divides about loyalties to the person and ever-changing public pronouncements of Donald Trump, the president during 2020 and early 2021 and then a highly politically visible former president active in shaping GOP politics for the future. The Trump ascendancy during the COVID crisis intensified and highlighted divides within the Republican Party, including those involving various Republican governors. To categorize governors based on their alignment with Trump’s approach to the pandemic, we conducted Google searches for each governor’s name and the terms “Trump” and “COVID.” We examined the top fifty news results for each governor to identify if and how each governor had publicly commented on the president’s handling of the pandemic. Some in the party, such as Doug Ducey of Arizona and Ron DeSantis of Florida, quickly became Trump supporters and have followed Trump’s lead on COVID policy, actively praising him for his handling of the crisis and invoking conspiracy theories and false statements to undermine public health guidelines for mitigating the virus. We call these governors the pro-Trump Republicans (represented by the R1 variable in our models). Other Republican governors, however, such as Charlie Baker of Massachusetts and Mike DeWine of Ohio, publicly denounced Trump’s handling of the pandemic. We call these the Trump skeptics (represented by R3 in our models). Last are those who have been careful to neither openly criticize nor laud Trump’s approach. Given that these governors avoided commenting on the public health response of their party’s leader during an unprecedented global pandemic, or sticking their head in the metaphorical sand, we refer to them as the Ostrich Republicans (for a full list of governors and categories, see table C.1).
Turning to our explanatory findings about partisanship and mitigation strategies, we find that the governor’s partisanship did in fact influence the implementation and length of mask mandates and stay-at-home orders. These findings echo our existing understanding of state policy response in a federal system. However, we provide evidence for additional nuance of the role of party and ideology, showing that internal divisions within the GOP also affect the length of the mandates. Notably, we do not find any significant relationship between the severity of COVID within a state and the type or length of mitigation strategies implemented. This finding undermines the traditional justification for federalism that, given the opportunity, states will respond to local conditions when developing and implementing policies.
For variations in the implementation of mask mandates by July 2020, we find a statistically significant relationship between the governor’s party and whether this step was taken, controlling for COVID rates in the state in the month prior. Compared with Democratic governors, Republican governors facing Republican-dominated legislatures were 38 percent less likely to implement a statewide mask mandate during the first seven months of the pandemic controlling for the severity of the COVID cases in their state as of June. Charlie Baker of Massachusetts and Larry Hogan of Maryland were the only Republican governors to implement a mask mandate by July 2020. However, by December 2020, 21 more states had implemented mask mandates, and at this point the relationship between the governor’s party and mask mandates disappears. It may be that outbreaks during the fall and winter months forced governors’ hands and led early resisters to adopt a new policy. An alternative explanation is that the new presidential administration represented a turning point toward depoliticizing masks, at least temporarily.
Republican-led states not only were less likely to have a mask mandate, when they did have one it was also significantly shorter than those in Democratically led states. On average, Republican governors imposed mandates nearly three months (eighty-eight days) shorter than those of their Democratic counterparts (see figure 6). The number of days a state remained under a statewide mask mandate in 2020 also highlights differences in approaches among Republican governors.7 As shown in figure 7, compared with Democratic governors, states led by pro-Trump and Ostrich governors had mandates on average four months shorter (121 and 119 days shorter respectively). Trump skeptics on average enforced mandates that were forty-two days shorter than their Democratic counterparts, but this difference is only significant at the 0.1 level.
Length of Mask Mandates in Days (OLS)
Source: Authors’ calculations.
Note: Democratic governors are the reference category for governor partisanship.
Length of Mask Mandates in Days with Detailed Republican Categorization (OLS)
Source: Authors’ calculations.
Note: Democratic governors are the reference category for governor partisanship.
A clear distinction between Trump skeptic governors and pro-Trump and Ostrich Republicans remains when we compare the length of mask mandates using our detailed Republican party categorization. The difference between Trump skeptics and Democratic governors is not statistically significant, but Pro-Trump and Ostrich Republican governors implemented mandates four months shorter (120 and 119 days, respectively) on average than their Democratic counterparts’ (see figure 7).
More states implemented stay-at-home orders than implemented mask mandates in the first year of the pandemic, but partisan patterns remain. All twenty-four Democratic governors implemented home orders, and all seven of the states that did not were led by Republicans.8 Furthermore, orders in Republican-led states were nearly a month shorter than those in Democratic-led states, controlling for legislative control and COVID case rates as of November 2020 (see figure 8).9
Length of Stay-at-Home Orders in 2020 in Days (OLS)
Source: Authors’ calculations.
Note: Democratic governors are the reference category for governor partisanship.
We also find significant differences in the length of stay-at-home orders among the three types of Republicans. Compared with Democratic governors, pro-Trump governors implemented stay-at-home orders more than thirty-seven days shorter, on average (see figure 9). Ostrich Republicans are also associated with stay-at-home orders that were a month shorter than those of Democratic governors, though this is statistically significant only at the .1 level, and we find no meaningful distinction between Democrats and Trump skeptics.
Length of Stay-at-Home Orders in 2020 in Days with Detailed Republican Categorization (OLS)
Source: Authors’ calculations.
Note: Democratic governors are the reference category for governor partisanship.
INTERBRANCH BATTLES
In the ideal expression of federalist response to a pandemic, the branches of state government would cooperate to assess and respond to the particulars of the crises at the local level. Although this may have occurred in some states, in others the legislative branch prioritized limiting the emergency powers of the executive even while cases and deaths in the state rose rapidly. The National Council for State Legislatures has systematically tracked the introduction of state-level legislation aimed at curtailing executive power.10 We find clear partisan patterns to these occurrences. This suggests that partisan divides across branches of state government can hinder the idealized process of responsive state policymaking on which federalism relies. Legislative attempts at curbing gubernatorial power during 2020 also highlight the inter-Republican divisions we observed in the implementation of mitigation strategies.
State legislatures facing Republican governors were overall less likely to introduce legislation to curb executive power than legislatures paired with Democratic governors. This is perhaps unsurprising given that Republican governors were less likely to use their executive powers to impose onerous COVID mandates. Specifically, Republican-led legislatures paired with a Republican governor were 46 percent less likely to have introduced executive power curbing legislation than when paired with a Democratic governor. Majority Democratic legislatures were 52 percent less likely to introduce curbing legislation for Republican governors versus Democratic governors.
Patterns within the Republican Party further show intra-Republican conflict. Legislatures were about half as likely to have attempted curbing the powers of pro-Trump governors than of Democratic governors. The difference between curb attempts for Ostrich and Trump skeptic Republican governors and Democratic governors is not statistically significant, suggesting that legislatures in general may have had more tolerance for formal gubernatorial emergency powers as long as the governor was openly hostile to mitigation strategies (see table D.10 in the appendix for more details).
We can now assemble our findings about initial 2020 state responses to the COVID-19 pandemic. Table 1 summarizes the relationships between our key variables of interest—partisan control of state government and agency structure—and our outcomes of interest—data collection and publication, mitigation strategies, and legislative attempts to curb executive power. Our data show that by and large, partisan control of state government is associated with the implementation of different mitigation strategies and legislative power grabs, but not with the collection and publication of data. Existing state capacity instead plays a role in some state-level data collection and publication. The number of health agencies may be negatively associated with time to dashboard publication but have no impact on the publication of long-term plans. The number of agencies, however, is positively correlated with the use of contingent reopening thresholds. More public health agencies may have translated into more internal pressure in the executive branch to use data-driven decision rules. Our models also suggest that the governor’s partisan affiliation is associated with the use of mitigation strategies, but we find no effect of the legislature’s partisan make up. Confirming variation in COVID response within the Republican Party, we also find that among Republican governors, alignment with Trump is associated with shorter restrictive mandates.
Summary of Findings for 2020 Outcomes
LATE RESPONSE: THE BIDEN, VACCINE, AND DELTA ERA (JANUARY–SEPTEMBER 2021)
In the second year of the pandemic, the context of decision-making changed substantially. In late January 2021, President Joe Biden took office in the White House and brought science and urgency to bear on the challenges of COVID-19. Although state-level offices were also on the ballot in 2021, there were only two changes in partisan control: Republican Greg Gianforte took over the governorship from Democrat Steve Bullock in Montana and Republicans gained control of the New Hampshire statehouse. States were following CDC guidelines and rapidly expanding vaccine eligibility. The developments combined with the impending summer weather suggested that Americans might be able to return to some semblance of normal life by the fall. However, by July 2021, the emergence of the highly contagious Delta variant and plateauing vaccination rates called this progress into question.
In addition to a changing public health context, state elected officials faced a new approach to the pandemic from the federal government. Just as Democrats lauded the more proactive and fact-based approach Biden and his team brought to the pandemic, many Republicans previewed their aversion to any national attempts to impose COVID restrictions on their states. Yet, as Delta surged and the start of the school year loomed, some governors sought to tie the hands of local officials in backtracking to more restrictive COVID policies.
As of July 9, 2021, the CDC changed its recommendation on masking in schools, stating that vaccinated teachers and students did not need to mask indoors (Stobee and Binkley 2021). However, just over two weeks later, on July 27, the agency reversed its recommendations, citing the rise in cases and the contagiousness of Delta, and recommended masking indoors for all vaccinated individuals in schools (Sparks 2021).
Reinstating mask mandates, particularly in schools, was especially politicized. School board meetings leading into the 2021 school year reflected vitriolic debates between public health advocates and those arguing that mask mandates infringed on student and parent liberties (West, Johnson, and Linnane 2021). Some states committed to allowing local school districts to make decisions about masking based on local conditions; governors of other states proactively banned local or school mandates, against CDC recommendations (Durkee 2021). As of September 15, 2021, nine states had active bans on localities or school districts implementing mask mandates (see table 2). Republican governors lead all nine of these states.11 Eight are pro-Trump and Ostrich governors; all but one Trump skeptic governor have refrained from tying the hands of local officials. The exception is Spencer Cox of Utah, who in late August 2021 deliberated signing an executive order to roll back the state law that banned local or school mask mandates, which he had signed in May before Delta became widespread.
Governors Who Banned Mask Mandates as of September 2021
On September 9, 2021, President Biden announced a sweeping vaccination and testing mandate that affected all federal workers, federal contractors, and private businesses with more than one hundred employees. Federal workers and contractors are required to be vaccinated, and employees of private businesses must be vaccinated or submit to weekly COVID testing. Twenty-one of the twenty-seven Republican governors denounced the mandates within days and in some cases within hours of Biden’s announcement (Jackson 2021). Reflecting their continual opposition to the use of science-backed mitigation strategies, 100 percent of pro-Trump and 88 percent of Ostrich Republicans spoke against Biden’s plan. Only 28 percent of Trump skeptics did so. All but one of six Republican governors who refrained from commenting were Trump skeptics; the exception was Bill Lee of Tennessee, whom we coded as an Ostrich Republican (see table 3 for full categorizations).
Governors Denouncing September 2021 Vaccine Mandate
In 2021, state legislatures continued to defy the logic of federalism and limit the powers of the executive branch, adding local public health officials as targets. As shown in table 4, twenty-six state legislatures passed restrictions on the powers of public health officials (Weber and Barry-Jester 2021). Twenty-two of the twenty-six legislatures have Republican majorities, and legislatures in fourteen of the twenty-six states have also introduced legislation to curb the governor’s emergency powers during the pandemic. Well into 2021, partisan patterns persisted.
States That Have Attempted to Curb Power of Public Health Officials
Republican governors were far more likely to tie the hands of local lawmakers and experts than their Democratic counterparts were, both when it came to restricting schools or localities from mandating masks and when it came to passing restrictions on public health officials. Despite decentralization being a long-standing tenet of the GOP platform, many Republican governors acted proactively to restrict the ability of local entities to respond to the pandemic in their jurisdiction in the way they best saw fit. Most Republican governors supported Trump’s “locally executed, state managed, and federally supported” approach to handling the pandemic––yet many also instituted bans on local governments or school districts mandating masks. This indicates that the principle of federalism, or decentralization in the case of the Republican Party, is often implemented selectively in accordance with partisan priorities.
CONCLUSION
In this article, we have taken account of the centrality of state governments in shaping the overall U.S. response to the COVID-19 pandemic in 2020 and 2021. We have sought to describe important variations in state-level responses at various stages, from early efforts to track the growing pandemic county by county to early deployments of state authority to limit the spread of a deadly virus, to ongoing state responses to new surges of COVID spread even after effective vaccines were available for statewide and nationwide use by all adult Americans. To make sense of the timing and variety of state responses, we have paid careful attention to partisan compositions of state governments as well as to the divisions between Republicans who either publicly identify with their party’s controversial head or signal some distance from him. Along with variables referring to partisanship and factional loyalties––especially for state governors—we have also heeded classic political science literatures about the diffusion of policy innovations and theories about cross-level cooperation within U.S. federalism.
We first examined how states collected and published data in the midst of an unprecedented public health emergency caused by a new and therefore, at least initially, poorly understood virus. We find that existing state capacities, not partisanship, offer modest explanations for how quickly states gathered and publicized data in this first phase. Our findings also suggest a more nuanced relationship between state capacity and data collection than exists in the current literature. Decentralized public health agencies may hinder initial data-collection efforts, but more people and agencies in public health may provide additional pressure for the executive to leverage specific data in her decision-making.
COVID-19 policymaking in the United States—most of it through the fifty states and their dealings with local governments in their jurisdictions—turns out not to fit traditional theoretical expectations about divisions of authority, cooperative dynamics, and imitation of best practices across states in U.S. federalism. Across the fifty states, governors and legislatures all devised policy responses on the fly, largely in response to party-centered partisan and factional dynamics. The Democratic versus Republican affiliations of presidents and governors turn out to be the main drivers of government responses at each phase. On the Republican side, even the partisan label as such is not enough to make sense of state-level responses that have shifted with little scientific or principled ideological rationale. Fifty state governors have been the key players in America’s unfolding and uneven responses to the COVID-19 pandemic, and what each has been willing to do, when, depends on the party of the governor, on which party controls Washington, D.C., and on factional struggles within the Republican Party.
Further, complicating the expectations of traditional federalism are interbranch battles for control over state policy. Effectively shut out of decision-making about COVID response, legislatures have sought to claw back authority from the governors and their public health officials who used emergency powers most liberally. These efforts further reflect the inter- and intraparty dynamics we documented with the implementation of mitigation policies.
When a public health emergency occurred in America, the Republican Party was not only in charge of most executive offices in Washington, D.C., and state capitols. It was—and continues to be—a party very much in the throes of rapid shifts in outlooks and leadership at all levels, and the GOP’s internal struggles during these shifts have had an outsized influence on state and local governments’ willingness and capacity to counter the COVID-19 pandemic.
APPENDIX
Data Description and Sources
Long-Term Plans and Reopening Thresholds by State
Source: Authors’ calculations.
State Reopening Plan Titles
Subcategorization of Republican Governors 2021
OLS Regression Results for Days to Dashboard Publication from January 22, 2020
Logistic Regression Results for the Publication of a Long–Term Reopening Plan in 2020
Logistic Regression Results for Advertisement of Contingent Reopening Thresholds in 2020
Logistic Regression Results for Statewide Mask Mandate, July 2020
OLS Regression Results for Length of Mask Mandates in Days in 2020
OLS regression Results for Length of Mask Mandates During 2020 in Days with Detailed Republican Party Labels
OLS Regression Results for Length of Stay–at–Home Orders During 2020 in Days
OLS Regression Results for Length of Stay-at-Home Orders During 2020 in Days with Detailed Republican Party Labels
Logistic Regression Results for Legislative Attempts at Curbing Executive Power in 2020
Logistic Regression Results for Legislative Attempts at Curbing Executive Power in 2020 with Detailed Republican Party Categories
FOOTNOTES
↵1. Our main controls include median income, 2016 Republican vote share, and population density. Our median income and population density measures come from the ACS 2018 5-year estimates. Population density is calculated as population per square mile.
↵2. We use the CDC’s cumulative case count on the first day of the month and population estimates from the ACS 2018 5-year estimate divided by ten thousand to get the cumulative case rate per ten thousand residents.
↵3. We calculated the average marginal effects using the margins package in R. We set all continuous variables at their mean. When we present results, we specify how we defined the binary variables for partisan control of the legislature and the governorship when calculating each respective average marginal effect.
↵4. Note that this phase does overlap with Phase I in order to capture policy actions taken by states early in the pandemic. We believe having some overlap in the phases is important given the variation in when the pandemic hit each state most severely.
↵5. This result holds regardless of the partisan control of the governorship or the legislature.
↵6. Based on data from the National Governor’s Association, which has been tracking state-level response.
↵7. Mississippi was the only state that started a mandate in 2020 to end it before the end of the year. Thus, to calculate the length of time a state was under a mandate in 2020, we counted the number of days between the start of each mandate and January 2, 2021.
↵8. Collinearity between the independent and dependent variables of interest precludes the use of multivariate logistic regression in modeling this relationship.
↵9. Interestingly, the length of stay-at-home orders is the only model in which cumulative COVID case rate is correlated with our outcome of interest. This said, the substantive impact of COVID case rate is extremely small, a coefficient of 0.005 per ten thousand residents, meaning that a state would see a single day increase in its mandate length only after an additional two million cases.
↵10. See NCSL, “Legislative Oversight of Emergency Executive Powers,” July 14, 2022, https://www.ncsl.org/research/about-state-legislatures/legislative-oversight-of-executive-orders.aspx (accessed July 22, 2022). The NCSL also tracks the fate of each piece of legislation. We focus on the introduction of legislation in this article, given that some of the attempts are still pending at the time of writing.
↵11. Perfect collinearity between the independent and dependent variables of interest precludes the use of multivariate logistic regression in modeling this relationship. This limitation applies to all models for the 2021 outcome variables, thus we present tables showing partisanship and policy choices instead of regression results.
- © 2022 Russell Sage Foundation. James, Sarah, Caroline Tervo, and Theda Skocpol. 2022. “Institutional Capacities, Partisan Divisions, and Federal Tensions in U.S. Responses to the COVID-19 Pandemic.” RSF: The Russell Sage Foundation Journal of the Social Sciences 8(8): 154–80. DOI: 10.7758/RSF.2022.8.8.08. The authors thank Lainey Newman and Harvard undergraduates Eliza Oehmler and Swathi Srinivasan for their valuable research work contributing to this article. For their feedback on early versions, we thank Meredith Dost, Shannon Parker, and Amy Lakeman. Direct correspondence to: Sarah James, at jamess{at}gonzaga.edu, 502 E. Boone Ave., Spokane, WA 99258, United States; Caroline Tervo, at caroline.tervo{at}duke.edu, 1737 Cambridge St., Cambridge, MA 02138, United States; Theda Skocpol, at skocpol{at}fas.harvard.edu, 1737 Cambridge St., Cambridge, MA 02138, United States.
Open Access Policy: RSF: The Russell Sage Foundation Journal of the Social Sciences is an open access journal. This article is published under a Creative Commons Attribution-NonCommercial-NoDerivs 3.0 Unported License.