key: cord-0979189-tq21e1fu authors: Weissert, Carol S; Uttermark, Matthew J; Mackie, Kenneth R; Artiles, Alexandra title: Governors in Control: Executive Orders, State-Local Preemption, and the COVID-19 Pandemic date: 2021-07-24 journal: Publius DOI: 10.1093/publius/pjab013 sha: f912f04660343481cf6e3d7855226106bb782b5f doc_id: 979189 cord_uid: tq21e1fu The nation’s governors took strong and decisive action in responding to the 2020 COVID-19 pandemic, often directly affecting their local governments. These actions allow us to examine this question: Will governors’ actions in an unprecedented emergency situation centralize the authority of the state or rely on local governments to deal with localized problems? Additionally, what factors affect those decisions? We examine all governors’ executive orders affecting local governments in the first five months of the 2020 pandemic. We find that preemption did occur, especially in the early months of the pandemic. States that gave their localities more autonomy were associated with preemption throughout the pandemic; the governor’s party affiliation and her ideological match with local officials were associated with greater preemption in some phases of the pandemic but not others. While extreme, Governor Kemp's actions were not unique in the spring and summer of 2020, as governors negotiated an unsteady and politically charged policy trajectory that directly affected their local governments-the governmental unit with the broadest intergovernmental health responsibilities. Governors sometimes relied on cities and counties to meet their citizens' needs by implementing more stringent measures than those issued by the state, but other governors expressly forbade any action that conflicted with the governor's executive order. At issue is more than policy implementation or political piques; rather, the experiences of 2020 reflect state actions that broadly affect local governmental responsibilities and powers. Of particular concern is whether these actions-taken by governors in times of emergency-reflect or accentuate longterm trends of centralization of authority in relation to their local governments or whether they are a one-off in an emergency situation not likely to be repeated. The COVID-19 response in 2020 provides a new venue for this research on state centralization and preemption in several ways. First, it is an emergency situation where residents of the state turned to their governors for reassurance and solutions (Bowman and McKenzie 2020) . Many legislatures were out of session when the pandemic hit, and those in session typically relied on governors' executive orders. Second, these actions were taken over a short period of time, limiting court action that is common in previous state preemptions (Swanson and Barrilleaux 2020) . Third, many of these preemption directives were bipartisan, with both Republican and Democratic governors acting to close their states or inhibit local actions. Finally, these executive orders are not closely tied to corporate and other interests, as is commonly the case with preemption measures in other contexts (Kogan 2018) . We examine state preemption of localities utilizing an original data set of actions that governors took affecting their local governments. We conduct a fiftystate cross-sectional analysis of governors' executive orders to deal with COVID-19 and then reopen the states' economy. We code the content and timing of policies related to local governments, focusing on whether these actions empower or preempt local governments and how these actions changed over the course of the pandemic. We find that COVID-19-related executive orders typically preempted local authority and centralized the authority of the state rather than empowering local governments to deal with localized problems. When considering the factors associated with a greater likelihood of governors preempting local authority, we find that in states that generally allowed local governments to exercise more autonomy governors were more likely to issue centralizing orders throughout the pandemic. Democratic governors were more likely to issue preemptive orders than Republican governors; but in a multivariate analysis, the effects persist in only the final months of the study period. Ideological asymmetry-where the party of the governor and the ideology of local communities does not match-is associated with levels of preemption in the early and late months of the study, though in different directions depending on the time period. In the early months, the governor in an asymmetric state was more likely to issue a preempting order; in the final phase of our study, asymmetry led to more empowerment. Federalism research has long focused on centralization by the federal government and to a lesser extent by the states. Kincaid (1998) was an early skeptic of federal governmental rhetoric about devolution where the federal government was giving up power to the states. Subsequent studies in the U.S. and internationally have supported Kincaid's skepticism-the trend toward centralization benefiting the federal government has been a fairly consistent one (Kincaid 2019; Weissert and Uttermark 2017) . In turn, states also tend to centralize their control over localities. Stephens (1974) concluded that although there was wide variation across states in terms of how centralizing they were toward local government, there was a general trend toward increased centralization. Bowman and Kearney (2011) applied the Stephens centralization measure to more recent data and added a survey of city officials and their perception of centralization. They confirmed the trend of centralization, even during a time when some attempt was made at devolution through the Temporary Assistance for Needy Families (TANFs) program. Of particular interest in recent years is states' use of preemption, including instances where states preempt local action without setting a statewide standard in the policy area (NLC 2019) . Examples of popular preemptions include single-use plastic bags and straws, gun control, fracking, minimum wage, paid leave, homesharing, establishment of public broadband services, discrimination in public facilities, ride-sharing, rent control, and immigration enforcement (Bulman-Pozen 2018; Flavin and Shufeldt 2020; Fowler and Witt 2019; Hicks and Weissert 2018; Kogan 2018; Riverstone-Newell 2017) . Although state preemption is not new, it has become more prevalent and more aggressive over the past decade, with states adding penalties such as fines, state aid cutoffs, or personal liability for damages for local governments and officials that do not respond as directed (Briffault 2018) . One example is from Florida where in 1987 the state preempted local gun laws. In 2011, the state upped the ante by putting in place harsh penalties if the local government enacted gun restrictions. Local officials can now be fined up to $5,000 and must personally pay court costs for any violation of the law. Some thirty municipalities and three counties brought suit against the state, arguing that the provision violates the state constitution. The trial court agreed, saying that while the legislature has the authority to prevent local governments from passing gun regulations that are stricter than state laws, local officials cannot be punished for enacting such measures (Haughney 2019) . Previous research on state preemption has focused on statutes, not executive orders (see, for example, Flavin and Shufeldt 2020; Fowler and Witt 2019) . This is understandable since state legislatures are the traditional source of local preemption in substantive areas such as the environment, gun control, and wages. Executive orders tend to be used relatively infrequently in traditional policy areas and typically deal instead with appointments, agency actions, and creation of task forces and boards. Executive orders related to disaster response accounted for some 11 percent of the executive orders analyzed by Ferguson and Bowling (2008) . There is little reason to think that preemption would be less appealing to governors than legislators. In fact, it might be more attractive, given the high stakes in the COVID-19 responses. Nebraska Governor Pete Ricketts provided an example of aggressive action when he threatened counties that failed to open their offices to the public with loss of federal COVID-19 response money allocated to the state. He then followed up by threatening loss of funding if local officials required masks in courthouses and other county offices (Hammel 2020) . The pandemic required quick and decisive actions by governors across all states. Kincaid and Leckrone (2020) called these actions "unprecedented in U.S. history," demonstrating that the police power of states remains potent. Governors exercised their power through executive orders that were often sweeping and sometimes controversial. This use of preemption by governors has raised some hackles with state legislators. At least seven state legislatures sued their governors for exceeding their authority (Quinton 2020) . At issue were extended emergency declarations, mask orders, and business-closing orders. Less noted was how governors used their executive orders to preempt local government. State and local governments were pivotal in responding to COVID-19. Governors, in particular, were in the hot seat, when the national response, led by the White House, essentially delegated authority (and blame) to them. As the Washington Post put it, "As Washington has stumbled, governors of both parties have acted to fill the void . . . This spring, the balance of power has been flipped, with states forced to compensate for failings at the national level" (Balz 2020) . The intergovernmental response to the pandemic in 2020 can be divided into three time periods: the shutting down; the opening up; and the partial second shut down. The three time periods are important since they reflect differing political and public health contexts that can affect the policy responses of governors. The first state of emergency was issued by Washington State Governor Jay Inslee on February 20 (a national state of emergency was signed by President Donald Trump on March 13). Shortly after, the first stay-at-home order was issued by California Governor Gavin Newsom on March 19. Governors across the country issued stay-at-home orders, closed bars, restaurants, movie theatres, and gyms, and limited the size of gatherings, including religious ones-often preempting local governments' early actions to close businesses and limit the size of gatherings. Governors of Illinois and Wisconsin saw courts strike down their stay-at-home order (Wisconsin) and limits to the capacity of bars and restaurants (Illinois). In both states, some local officials put in place their own orders mirroring the state policy (Petrella 2020; Vetterkind and Schmidt 2020) . Congress acted quickly in March 2020 by providing funding for vaccines, paid sick and child-care leave and free COVID-19 testing, and passing a massive $2.2 trillion Coronavirus Aid, Relief and Economic Security (CARES) Act with cash payments to individuals, increased unemployment benefits, forgivable loans for small businesses and $340 billion for states and local government. The White House recommended limits on large gatherings and encouraged work and school from home in mid-March. Governors of both parties sought federal assistance for testing, protective equipment for health workers, ventilators, and other supplies, but the White House instead set up a system where states competed with each other (and sometimes the federal government) to purchase supplies and equipment. Governors' leadership seemed to have paid off. By mid-April, governors, on average, had a 69 percent approval for their handling of the coronavirus; President Trump's approval for this responsibility was 44 percent (Montanaro 2020) . The second period began on April 20 when South Carolina became the first state to start reopening its economy and loosening restrictions on behavior. Over the next weeks and months, many states followed. Most governors called for a phased reopening of bars, restaurants, movies, gyms, beaches, and parks tied to reduced cases and deaths from the pandemic. President Trump encouraged states to reopen, arguing that the cure should not be worse than the problem itself, and urging protestors to "Liberate" Michigan, Minnesota, and Virginia, states with Democratic governors who had refused to lift stay-at-home orders (Mauger and LeBlanc 2020) . The third distinctive time period began on June 22 when states first paused their phased reopening, closed bars, and gyms that had been reopened, and began to impose mask requirements in response to a step-up in new COVID-19 cases and deaths. By mid-June over twenty states and more localities had put in place mandatory face masks (Chappell 2020) . In this period, intergovernmental differences within states became more visible. Face masks, in particular, became a political flashpoint. A number of localities in states without statewide bans put them in place, but some localities in states with bans objected to them. For example, at least twelve Texas county officials announced they would not abide by Republican Governor Greg Abbott's executive order mandating face masks in public for any county with over twenty confirmed positive cases of COVID-19 and banning public gatherings of more than ten people (Martin 2020). The three periods of gubernatorial executive orders are important since they involved different types of action in response to different contexts. Closing up businesses and services initially engendered some public and local pushback but reflected largely non-partisan public health responses to an unexpected and terrifying scourge. By the second phase, reopening businesses, economic concerns added to or replaced public health concerns, and partisanship in the form of advice from the president began to appear. Finally, the third phase was the most contentious with the wearing of masks taking on a partisan flavor that sometimes led to assaults and well-organized protests. Given the different political and economic tenor of the three time periods, we will examine executive orders issued in each for any temporal impact on centralization or preemption of local governments. In the early days of the pandemic, governors across the country issued shut down orders, closed bars, restaurants, and gyms, and prohibited large gatherings. But the pandemic did not hit equally across states, leading to differential needs for closing and later reopening. These differences played out in governors' actions and in the courts. Although local governments sometimes provided the earliest response to the pandemic, when the states stepped in they did so big time. For example, Charleston and Columbia, South Carolina, issued stay-at-home orders to slow the spread of the virus in late March 2020. The South Carolina attorney general issued an opinion that only the governor, not cities or counties, could order citizens to stay at home during the emergency. The cities had argued that their actions did not contradict the governor's actions because he had not acted (Trainor 2020) . Similarly, Phoenix and Tucson, Arizona, ordered closure of bars and restaurants before a statewide mandate was implemented. But when the state did act, it constrained local action by forbidding counties and municipalities from making rules or regulations that conflicted with the governor's order (Pollett and Oxford 2020). Arizona Governor Doug Ducey explained his actions this way, "I will continue to believe that government closest to the people is best-except in a global pandemic," Sylvester Turner, the Mayor of Houston, described the situation from Executive orders and preemption the cities' view: "The state is driving the car. We are the passengers." (Abrahams 2020) . In June and July when a number of governors began to open up their states, conflicts arose over requirements on business and on mask mandates. Republican governors in two states-Arizona and Texas-refused to impose a statewide mask mandate and initially would not allow local governments to have one but then changed their minds. In mid-June, Arizona Governor Ducey allowed local officials to issue mask mandates but not restrict businesses in their communities. Texas Governor Greg Abbott issued an executive order expressly superseding local orders with regard to any type of fine or penalty for anyone not wearing a mask. He then changed his mind, issuing a statewide order with fines. The governor's order also authorized mayors and county officials to impose restrictions on some outdoor gatherings of more than ten people and require social distancing of six feet at large gatherings (Montgomery and David Goodman 2020) . In a September executive order, Florida Governor DeSantis lifted all statewide restrictions on businesses but allowed local mandates to remain in place-but without fines. The executive order also barred local emergency ordinances that could prevent an individual from working or from operating a business and prevented local governments from requiring restaurants to operate below 50 percent capacity. It required local governments to quantify the economic impact and the public health need for limits on indoor capacity below 100 percent. In January 2021, the governor continued his opposition to local governmental attempts to impose their own rules regarding COVID-19 saying, "We will categorically not allow any local government to lock people down. We will not let any local government kick anybody out of their job. We will not let any local government fine individual Floridians." (CBS Miami 2021). As was the case with gubernatorial actions, court cases featured local officials sometimes not implementing state orders and sometimes acting beyond what the state thought their local powers allowed. For example, New Jersey sued Asbury Park for violating the governor's executive order banning indoor dining (Dolmetsch and McRoberts 2020) . However, the attorneys general of South Carolina and Texas issued opinions overturning local ordinances where local governments took actions in areas where the state had not acted (stay-at-home orders in South Carolina and restrictions on church services, essential businesses, and masks in Texas; Sabawi 2020; Trainor 2020). The Florida Appeals Court upheld the county mask mandate in Palm Beach County. Local officials were often left largely to appeal to the public to take steps to combat the virus. For example, Atlanta's Mayor Bottoms said she still had her "voice" and pledged to continue to ask Atlantans to "please stay home" (Amy 2020). The pandemic highlights both the importance of federalism and the intergovernmental political tensions that ensued. While the federal-state story is interesting in its own right, our concern is whether governors in issuing executive orders in response to the pandemic took away power of local governments. The question we address is whether governors acting in an unprecedented emergency situation will centralize the authority of the state or rely on local governments to deal with localized problems and what factors affect these decisions. Previous research would lead us to think that trends toward centralization will continue or even be extended when governors can act on their own and in times of emergency (Bowman and McKenzie 2020) . But there are contextual factors that might lead governors to give local officials more power. For example, in most states, COVID-19 cases were initially concentrated in certain areas of the state, leaving much of the rest of state in very different circumstances. Los Angeles had its first COVID-related casualty on March 11, 2020, months before more rural parts of the state saw any cases. Thus, governors could treat the affected areas differently from other parts of the state, giving officials there the ability to deal with their own needs. Other contextual factors that are important in predicting preemptive decisions include politics, phase of the pandemic, and local autonomy. The political party of the governor might be expected to play an important role in pandemic actions supported or opposed by a highly visible president who called out Democratic governors for closing down their states and lauded Republican governors who did not. These executive orders often preempted local governmental action. Also relevant is the political or ideological similarity between the governor and local officials. Governors might especially trust local officials who are close to their own views on issues. We are also interested in whether time plays a role in governors' power accumulation. Actions taken during the early period in March and April came in response to the sudden public health crisis and may have been especially centralizing as governors acted quickly and decisively. During the second period when states reopened partially or fully, there might be more leeway provided to local government, although this period was clearly more partisan than the earlier one. Finally, in the third, and most confusing, period, when there was enormous variation across the states in whether they reopened, closed up, or did both, the possibility of local control may well have been most important, in part because of blame casting that became more evident during this time. Finally, we examine the degree of local autonomy states have granted their local governments as a determinant of preemptive action. Dating back to de Tocqueville, the normative argument has prevailed that strong autonomous local governments are key to a strong federation. Yet local governmental authority relies solely on state law which gives local government the authority to govern matters of local concern, and states can freely intervene in local fiscal, regulatory, or substantive issues. State centralization and erosion of local autonomy have been documented for over forty years (Stephens 1974) . Clearly, autonomy is an important institutional factor-but the direction of the importance remains unclear. On the one hand, one might argue that in states where local governments have considerable autonomy, governors might use less preemption than in states where that autonomy is low. In these cases, there might be a history of local governmental action, institutional experience, and a tendency to leave local governments alone-or possibly have them take the responsibility or blame (Volden 2005) . Fowler and Witt (2019) found that home rule status increased the likelihood of state preemption across seventeen issue areas-but the effect was dwarfed by the overwhelming political impact of a Republican legislature. Flavin and Shufeldt (2020) found no significant relationship between home rule and the twenty-one policies they examined. However, it is possible that governors feel the need to curtail activities of localities possessing such autonomy. In this case, governors might not want variation in local response-particularly if that response flies in the face of political goals of the governors. For example, governors who wish to open up the state might not relish local governments responding by continuing restrictions on restaurants and other facilities or setting curfews. There are theoretical reasons as well since some scholars posit that autonomy is limited, and even unhelpful, in local governance. Frug and Barron (2008) argue that local autonomy simply amounts to some limited powers provided by the states, accompanied by commensurate restrictions. As a result, local governments must compete with each other in ways that can be settled only by the state, thus undermining the limited power local governments possess. Under this view, the level of local autonomy is so meaningless, it would not be a deterrent to expansion of state authority, including that of governors. Given the conflicting theoretical and practical possibilities, we are not hypothesizing the direction of local autonomy. We compiled a novel dataset of every executive order issued by governors affecting local governments relating to COVID-19 from March 19, 2020 to August 1, 2020. Our source of data was the Council of State Governments (CSG) COVID-19 Resources for State Leaders, which listed every governor's executive order issued to combat the pandemic. We identified those orders which included any mention of local government, city, municipality, county, local authorities, other governmental units, local orders, towns, townships, villages, and preempt(ion). We did not include the few executive orders related to school districts. Since some executive orders were comprehensive and included several items directed at local government, provisions within the executive order were the units of analysis. There were 897 provisions identified from the executive orders. Some states were more active users than others. New Jersey issued 106 provisions directed at its local government; Missouri issued one. The mean state issued twenty executive order provisions affecting local government. We coded proclamations and declarations included in the CSG listing as executive orders. We did not include press releases or other informational communications that were at times listed by the CSG. Occasionally, the executive orders repeated local language in several orders. To account for this, we adopted the following rules: If the same language is repeated in different orders with a change (even small changes), we counted each separately. If the same language is repeated in orders, identical except for changing dates, we did not include it in the dataset. We coded the data for date of issuance and whether the executive order was statewide or directed to certain local governments. We used the three periods discussed earlier to capture any time-varying effects of our key independent variables on the degree of centralization in executive orders. Our key variable of interest was whether the provision was restricting (which we call preempting), empowering, or neutral, a classification used by Bowman and Kearney (2018) to examine state legislation directed at local government. Specifically: • Preempting provisions are those that place a burden on localities, remove a power currently held by local governments, prevent local government action, impose an administrative burden, or financial cost to local government, cease enforcement of local ordinance, take away previous authorization, or mandate local action. • Empowering provisions grant localities additional discretionary authority, allow local governments to implement more stringent measures, allow access to information or funding from another program, or in some way ease the burden for localities. • Neutral provisions are neither empowering nor restricting; for example, they do not lead to local governmental action, involve coordination, or if they contained provisions that both empowered and restricted. If there is no action by the local government flowing from the provision, it is neutral. year budget cycle may adopt a resolution instituting a grace period concluding on a date no later than June 1, 2020 for the payment of second-quarter property taxes. Neutral provisions AL: Governmental bodies can meet via phone or video conference. CA: The state health officer sets criteria to determine whether particular local jurisdictions may implement public health measures that depart from statewide directives. Local health officers may issue directives less restrictive than statewide if they provide certification to the state health officer. UT: Local health departments should coordinate with local businesses and appropriate industry associations to establish orders and recommendations that will govern food and beverage service providers and other businesses upon the termination of this Order. NJ: Nothing in this Order shall be construed to limit, prohibit, or restrict in any way the operations of law enforcement agencies. Two coders read every provision and coded whether it was restricting, empowering, or neutral. The intercoder agreement across measures was 96.3 percent (Krippendorff's alpha equals 0.938). A full set of intercoder reliability statistics is reported in Appendix table A1. We coded preemptive provisions as 1; enabling as À1; neutral as 0. With these scores, we calculated two related dependent variables. Our first measure uses states as the unit of analysis; the dependent variable is the average level of preemption from all gubernatorial executive orders. Our second measure uses individual executive order provisions as the unit of analysis; the dependent variable is the score of each provision. In both instances, we examine the total executive orders issued and then analyze each of the three time periods discussed earlier. We found that governors were more likely to be preempting in executive orders directed to local government rather than enabling. The map in figure 1 shows the breakdown by state. The figure highlights the variation in preemption over the course of the pandemic. Montana, Nebraska, North Dakota, and Oregon were the most preempting states. The most empowering states were Hawaii, Iowa, and Rhode Island. The map reveals that there is no regional bias in terms of these tendencies. The South, Midwest, and Far West show no discernable patterns, highlighting the fact that governors' decisions were based on their own state's public health and political situations. 1 Overall, over half of the local executive order provisions were preemptive; only a little over one-fourth were enabling. When we look across the phases, there was more preemption in the Phase 1 closing orders (59 percent) than in Phases 2 and 3 where preemption was around 50 percent (see table 2 ). It is important to note that the number of executive order provisions differs across the three time periods. In the initial shut down (Phase 1) there were 423 provisions targeted at local governments. In Phase 2, there were 373 and in Phase 3 there were 101, for a total of 897. Table 3 compares the mean rate of preemption between Democratic and Republican governors across the phases of the pandemic. We find that across all phases of the pandemic both Democratic and Republican governors favored preemption over empowerment. While the mean rate of preemption is higher for Democratic governors in Phases 1 and 3, Republicans report a higher rate of preemption in Phase 2. However, none of these differences is statistically significant at the 0.1 level. 2 While these findings suggest that governors did not shy away from preempting their local governments in responding to the pandemic, we need to more fully analyze the results, controlling for factors that might influence the descriptive findings. We do so using the following independent variables. The key independent variables were party and institutional power of the governor, whether the governor was a lame duck, whether the governor is an ideological mismatch to the state's local governments, partisan composition of the state, local government autonomy, measures of public health needs and the state's economy, and whether the executive orders were targeted to selected areas of the state. Note: This map reflects how preemptive or empowering toward local governments governors' executive orders were on average. States with the darker colors are more empowering (À1) and those with lighter colors were more preempting (þ1). Given the salience of the issue and the toxic national political environment, the role of the political party of the governor in her decisions seems obvious. However, we are examining only executive orders that affect local governments, and here national and state executive convergence may not play the key role they do in other emergency responses. Ideologically, we might expect to see Democratic governors more likely to preempt localities since Democrats are generally less supportive of decentralization than Republicans (Dinan and Heckelman 2020), but we do need to keep in mind that most large cities are governed by Democratic mayors who likely oppose preemption. We include a dummy variable of 1 for Democratic governors; 0 for Republican governors. We also include a variable for governors' powers. States where governors' powers are strong might see more preemption of local governments in the coronavirus response. We use Ferguson's (2017) additive index reflecting these components: separately elected executive branch officials, tenure potential, appointment power, budget power, veto power, and party control. Each component is scored from 1 to 5 for a possible total of 30. Governors with stronger powers might be more likely to preempt local governments. Also included in the analysis is a dichotomous variable for whether the governor is unable to run again in the next election, where 1 is a lame duck and 0 where the governor can run again. The rationale here is that a lame-duck governor might be less concerned with pushback from local governments than a governor who might be seeking reelection. The partisan composition of states is potentially an important determinant of how governors responded to the coronavirus given the contentious nature of presidential-gubernatorial relations during the period. To measure partisanship, we used the percentage of the state's electorate voting for Trump in 2016. Robustness checks using state-level ideology estimates (Berry et al. 2013 ) result in substantively similar results. Also important in the likelihood of preemption is local governmental autonomy. If local governments have a great deal of autonomy, the governor might be less willing to preempt their power or alternatively might feel preemption is the only way to guarantee desired local action. We measured local government autonomy using the measure developed by Wolman et al. (2010) . The measure of local autonomy reflects three dimensions that flow from the literature on local autonomy: local government importance, local discretion, and local government capacity. The authors then identify twenty-one variables within these dimensions and use factor analysis to isolate seven root factors or factor scores (two for local government importance, three for local government discretion, and two for local government capacity). A final score for every state is the average of the seven-factor scores aggregated to an overall local autonomy ranking ranging from À1.23 (Hawaii) to 0.861 (Kansas). Higher scores indicate greater levels of local autonomy. The measure applies to all levels of local government in a state. We used local autonomy rather than home rule for both substantive and technical reasons. First local autonomy is a broader measure that encompasses the ability of local government to act apart from legal authority-what Bowman (2017 Bowman ( , 1121 calls "the imprecision of the concepts of Dillon's Rule and Home Rule." Second, there is considerable difference in "home rule" take-up among localities that can be problematic in using the variable to represent local "pushback" (Weissert and Ice 2014) . The technical concern is that researchers often use a dichotomous variable that does not reflect variations in authority. Additionally, we account for whether the governor and her local officials are politically or ideologically similar (or different). This variable reflects whether the median county in a state matches or differs from the governor's ideology-a measure of ideological asymmetry. The variable builds on the county ideology measure developed by Tausanovitch and Warshaw (2014) who used election returns data and multilevel regression with poststratification to produce estimates of the ideological leanings of counties. If there is similarity in the ideologies of the governor and median county, we might expect to see less preemption; if there is dissimilarity we should see more preemption. The variable is dichotomous, coded 1 when the median county within the state is liberal and has a Republican governor or if the median county is conservative and has a Democratic governor. Conversely, the variable is coded as 0 when the median county within a state is liberal and the governor is a Democrat or when the median county is conservative and the governor is a Republican. This measure is a way of capturing the political division between governors and their local governments given most local officials do not run on partisan tickets. Public health and economic variables may also push governors to take action in response to the pandemic in ways that affect local government. To capture public health demands, we include a count of positive COVID-19 cases per 100,000 population derived from daily counts available from the COVID Tracking Project (https://covidtracking.com). As higher unemployment is an indicator of economic distress among residents of the state, we include a measure of the average weekly unemployment insurance claims rate by state obtained by the U.S. Department of Labor. Not only does this measure capture the rate of economic pressure experienced by the workers within the state, but since these are claims rates, they also reflect the economic pressures on the states as well. Finally, we include the percentage of our local-based executive order provisions that were targeted to specific regions or municipalities within the state. This is a control for the variance in the scope of executive order issuance across states. If the provision is targeted to a region or jurisdiction, it is coded 1; otherwise, it is a 0. For our individual-level analysis, we also code whether the executive order was statewide or targeted to certain designated areas of the state. Appendix table A2 provides descriptive statistics for the dependent and independent variables. We estimate two sets of models, one with the state as the unit of analysis and the other with the executive order as the unit of analysis. In each model, we include the same independent variables, aside from a transformation of the targeted variable discussed earlier. Each set of models is estimated four times, one for each of our three phases and one for all periods combined. For the first set of models, we use OLS regression with the dependent variable as the average amount of preemption occurring in the state. The dependent variable ranges from À1 (all orders within a state where empowering) to 1 (all orders within a state where preempting). We include the OLS regression models in the analysis as they allow us to evaluate the impact of each state's actions with respect to the pandemic equally. This can help guard against a particularly executive-action heavy state-for example, New Jersey-from providing too much influence on our results. For the second set of models, we use ordered logistic regression, with the dependent variable coded 1 if an executive order preempted local governments, 0 if it was neutral, and À1 if it was empowering. To aid in interpretation of the model, we provide predicted probabilities for each variable that was statistically significant. This provides the predicted change in the probability of preemption given a oneunit increase in the independent variable. We begin by estimating a series of state-level regressions to evaluate our hypotheses. In table 4, our dependent variable captures the proportion of executive order provisions that are preempting versus empowering-ranging from À1 to 1. In the first model, we estimate centralizing trends of provisions across the entirety of our time frame from March 19 through August 1, 2020. In Models 2-4, we evaluate state centralization across each phase of the pandemic: the initial shut down (Model 2), reopening (Model 3), and partial second shutdown (Model 4). If a state did not issue an executive order during a particular phase, it is not included in the analysis. Local autonomy is the only variable in our model which associates with a state's centralizing tendencies across all executive orders affecting localities (see table 4 ). Notes: Two-tailed tests. *p < 0.10, **p < 0.05. In Model 1, a one-unit increase in local autonomy is associated with a substantively large 0.46-point increase in total centralization. Over the three phases of the pandemic, the local autonomy variable reports a positive coefficient, which fails to reach conventional levels of statistical significance. Our political independent variables capturing the governor's party and ideological match with local governments were only weakly associated with centralization. Neither variable was significant when all executive orders were examined (Column 1). However, when the total is broken down across phases, some effects emerge. In the third phase, Democratic governors were more likely to issue preemptive executive orders with a modest significance level of 0.10. During Phase 3, a Democratic governor issued executive orders which were 0.73 points more centralizing compared to a Republican governor. Similarly, the variable measuring a mismatch between the governor and local ideology was significant at 0.10 level in two phases-but in different directions. In Phase 1, the initial lockdown, a mismatch in ideology was positively associated with preemption; in Phase 3, the second lockdown, the mismatch led to less preemption. A governor in an ideologically asymmetric state issues a collection of orders which were, on average, 0.50 points more preempting and À0.79 points less preempting, respectively. The differing effects of ideology mismatch might be due to the uncertainties and public health concerns prevalent in the initial shutdown. Governors acted more preemptively and were more likely to do so given an ideological mismatch with localities, as we had expected. Three or four months later, governors were less likely to be preemptive when their ideology mismatched their local governments. Also, at play in the third phase may have been more confidence on the part of governors that localities of differing ideologies needed fewer preemptive orders. Other independent variables including executive power, lame-duck status, targeted orders, partisanship, pandemic severity, and economic distress were not significantly associated with the proportion of executive order preempting provisions within a state in any phase or in total. To more fully understand the role of executive actions in response to the pandemic, we follow the recommendation of King (1993) and evaluate actions within executive orders. To do this, we re-estimate our models-actions over the entire pandemic and the three phases-using a series of random intercept ordered logistic regressions. For these models, our dependent variable can take on the values of À1, empowering, 0, neutral/mixed, or 1, preempting provisions. We include the same suite of independent variables in the table. Predicted probabilities are reported in a column next to each model. Robust standard errors are included in all models (see table 5). Notes: Two-tailed tests. State random intercepts ordered logistic regression with robust standard errors. Observations reflect the population of COVID executive orders. Predicted probabilities denote the change in probability of a Governor issuing a centralizing executive order given a oneunit increase in the independent variable, holding all independent variables at their observed values. *p < 0.10, **p < 0.05, ***p < 0.01. Again, we find evidence of local autonomy associating with higher probabilities of preemption. In the total model and Phase 2, higher levels of local autonomy are significantly associated with preemption provisions. In Phases 1 and 3, the association is significant at the 0.1 level. The strength of the association is illustrated in Appendix figure A1. Across observed levelsof local autonomy, we find that the probability of an executive order preempting local government increases from 23.8 percent to 70.9 percent. This size of this effect remains persistent across all three phases of the pandemic with a one-unit increase in local autonomy leading to a 22.1 percent increase in the probability of preemption during Phase 1, an 18.5 percent increase in probability of preemption during Phase 2, and a 21.8 percent increase in the probability of preemption during Phase 3. As with the previously reported regression results, local autonomy is the only independent variable in our model which associates with preemption in the model examining actions across all phases in the pandemic. The multivariate analyses clarified the descriptive statistics with the findings related to both party and phases. There is a statistically significant difference between Democratic and Republican governors' use of preemption in these models and a mismatch between ideology of the governor and localities, and these effects are strongest during Phase 3 of the pandemic. Democratic governors were more likely to issue preemptive provisions in Phase 3. On average, a Democratic governor is 33.9 percent more likely to issue a preempting order compared to a Republican governor. Ideological mismatch was modestly significant in predicting preemptive measures in Phase 1 but was associated with fewer preemptive provisions in Phase 3. Substantively, on average, a governor in an ideological asymmetric state is 38.2 percent less likely to issue a preempting executive order during this phase. These results are interesting, if a bit puzzling. However, in two analyses-one where each state is given an equal weight in the analysis and one in which we examine each order directly-we come to the same conclusion. While these results are exploratory and should be interpreted gingerly-only twenty-eight states issued a total of 101 executive orders during this phase-these findings provide suggestive evidence of when partisanship became salient during the pandemic. We also find evidence that targeted orders are a significant predictor of preemptive executive orders in several models. In Phase 1, targeted ordersexecutive orders which do not impact the entire state-were positively associated with a 17.8 percent increase in the probability of a preempting executive order, on average-likely because the COVID-19 epidemic did not hit all areas of the state equally in the earliest period and were often focused in urban areas where preemptive measures were directed. In Phase 2, we see a different pattern where targeted orders were, on average, 30.6 percent less likely to be preemptive. This association is illustrated in Appendix figure A3. During Phase 2, a statewide executive order is expected to preempt local governments 51 percent of the time, while a targeted executive order is predicted to preempt only 24 percent of the time, on average. Thus, governors are giving areas of the state most impacted by the pandemic more discretion to meet the needs of their citizens. Again, we fail to find evidence that executive power, lame-duck status, Republican vote share, pandemic, or economic severity associate with executive order centralization. The only consistently significant finding-that governors were more likely to centralize power away from municipalities that had high levels of local autonomy-is at first counterintuitive. It seems more logical that governors would give more power to more autonomous localities-not less. However, it does make sense. As discussed earlier, some local governments refused to follow the governor's orders or acted against the governor's direction. These localities, likely those that are autonomous, may have been the target of governor's orders although this is not the only reason for the relationship. Variables which lead to greater empowerment of local governments were inconsistent across models. We find that targeted orders in Phase 2 and ideologically asymmetric states in Phase 3 are associated with increased empowerment. Yet, both variables were associated with greater rates of preemption in previous phases. These findings suggest that the decision to empower local governments is context-specific, and made on a case-by-case basis. For example, local officials in jurisdictions heavily affected by the pandemic have the knowledge and context to effectively handle problems within their own communities. There may also be a blame-sharing rationale where governors want to shift any blame on responses in hard-hit areas to other officials. State preemption is alive and well in governors' executive orders issued in response to the 2020 COVID-19 pandemic. In this sense, the results of our fifty-state analysis are consistent with previous research on centralization in federalism. In responding to the pandemic, the nation's governors took strong and decisive action in the face of little support from the federal government, and in doing so, they often preempted their local governments. We found that the party of the governor was not consistently significant in the tendency toward centralization, nor was the ideological mismatch with local officials. However, we did find some variation across time periods. In Phase 3, Democratic governors were more likely to associate with preemption, but ideological asymmetry led to more empowerment. We did find that governors preempted local government when those governments had more autonomy-perhaps in recognition of that autonomy. Furthermore, this was the most consistent and, in fact, only variable associated with centralization over the entirety of the pandemic. Interestingly, measures associated with COVID-19 severity, economic hardship, and gubernatorial powers, had no association with rates of preemption within the states. The contribution of this work is two-fold. First, as in previous research, we find a trend of centralization in state-local government action. Our analysis is unique in that it expands this discussion to gubernatorial executive orders. The trend occurs-and especially occurs-in local governments with more autonomy-a finding that should be pursued in future research. Second, we find that politics and ideology play a secondary role in governors' proclivity to preempt local government. This finding runs counter to legislative preemption, which often reflects specialized interests and concerns. COVID-19 was clearly an unprecedented national emergency that required quick action but also demanded much longer and revolving policy intervention than typical natural disasters. Will the centralization evident from governor's actions generalize to other more typical public policies? Hopefully, scholars can answer this question using this research as a baseline. We find some evidence that preemption attenuates over time. Whether this continues will be worth further study to fully answer the question of generalizability. Another topic of future research will be whether state legislatures will continue to take a backseat to gubernatorial powers, including those affecting local governments. For example, a full year after COVID-19 hit the state, the Republican-controlled legislature in Florida began to look at ways to rein in Republican Governor Ron DeSantis's emergency powers. "I think a part of our role is absolutely to look at what the Legislature's role is in a sustained emergency," said one Republican state senator. "We have never been in a situation where it's been this prolonged and ongoing." (Fineout 2021) . Other state legislatures are similarly stepping up their involvement in recognition of the governors' near-total authority in COVID-19, although there is little indication that state legislatures would be less preemptive toward localities than the governor, possible shifts in intergovernmental oversight are important to both scholars and practitioners. Notes 1. The map illustrates that there are states which issued only enabling or preempting executive orders. These instances occur mostly-but not entirely-in states that issued few (i.e., fewer than eight) executive orders over the course of the pandemic and constitute 3.3 percent of all executive actions. 2. Technically our data of executive orders constitutes a population, not a sample, thus traditional reliance on significance is less meaningful. Executive orders and preemption Figure A3 Executive order centralization across targeted orders-Phase 2. Executive order centralization across governor's party-Phase 3. Executive orders and preemption Figure A5 Executive order centralization across ideological asymmetry-Phase 3. Note: Two-tailed tests, clustered standard errors by state. *p < 0.10, **p < 0.05, ***p < 0.01. Mayor Turner Says COVID-19 Curfew Is 'Nuclear Option' and Asks Governor to Do More Georgia to Reopen Some Businesses as Early as Friday Georgia Governor Allows Local Mask Mandates, With Limits As Washington Stumbled, Governors Stepped to the Forefront A new measure of state government ideology and evidence that both the new measure and an old measure are valid The state-local government(s) conundrum: Power and design Second-order devolution: Data and doubt The legislative transformation of state-local relations Managing a pandemic at a less than global scale: Governors take the lead The challenge of the new preemption State-local preemption: Parties, interest groups and overlapping government DeSantis 'Will Not Allow Any Local Government to Lock People Down More than 20 U.S. States New Require Face Masks in Public Council of State Governments 2020. COVID-19 Resource for State Leaders Stability and Contingency in Federalism Preferences New Jersey Wins Injunction to Block Asbury Park's's Indoor Dining Plan. Bloomberg News Governors and the executive branch Executive orders and administrative control Florida Lawmakers Gave DeSantis Total Power Over Pandemic Aid. Now They Want it Back Explaining state preemption of local laws: Political, institutional and demographic factors Brian Kemp Forbids Cities, Counties from Requiring Masks as Coronavirus Surges in the State State preemption of local authority: Explaining patterns of state adoption of preemption measures City Bound: How States Stifle Urban Innovation Ricketts Tells Local Governments They Won't Get Federal COVID-19 money if They Require Masks. Omaha World Herald State, Local Governments Gird for Court Battle Over Florida's Firearms Preemption Penalties. The Center Square Home rule be damned: Exploring policy conflicts between the statehouse and city hall The devolution tortoise and the centralization hare Dynamic de/centralization in the United States Partisan fractures in U.S. Federalism's COVID-19 policy responses Executive orders and preemption The methodology of presidential research Means, motives and opportunities in the new preemption wars At Least 12 Texas Counties Won't Enforce Governor's Face Mask Mandate Trump Tweets 'Liberate' Michigan, Two Other States with Democratic Governors. Detroit News Coronavirus Latest: Trump's Fight with Governors is One He Can't Win. National Public Radio Texas Republican Governor Reverses Course and Orders Fac Masks Restoring City Rights in an Era of Preemption: A Municipal Action Guide Pritzker Urges Restaurants, Bars to Keep Following Safety Guidelines. Chicago Tribune Arizona Governor issues 'stay-at-home' order: Will take effect close of business Tuesday GOP Lawsuits Restrain Governors' COVID-10 Actions The rise of state preemption laws in response to local policy innovation Texas Attorney General Accuses Bexar County, San Antonio of Overreaching in Response to COVID-19 Pandemic. KSAT News State centralization and the erosion of local autonomy State government preemption of local government decisions through the state courts Representation in municipal government SC Attorney General Says Only Governor-Not Cities-Can Issue 'Stay at Home' Orders. Charleston Post and Courier Wisconsin Supreme Court Strikes Down State-At-Home Order Intergovernmental political competition in American Federalism Relations between state and local governments: The home court (dis)advantage Glass half full: Decentralization in health policy Comparing Local Government Autonomy Across States Executive orders and preemption Figure A1 Executive order centralization across local autonomy-all phases.