Abstract
Although partisan politics tend be set aside during crisis, the timing of gubernatorial actions in response to COVID-19 is telling about how partisanship is shaping the way elected officials are reacting to this pandemic. Using an event history analysis, the authors find that Democratic governors responded to the White House’s attempts to downplay the severity of the pandemic by declaring emergencies in order to draw citizen attention to and to prepare for a public health crisis. On the other hand, Republican governors resisted doing so until Trump declared a national emergency on March 13; however, Republican reactions were conditional on the president’s job approval in their states. While the COVID-19 pandemic has pushed governments into uncharted territory, state governors appear to be following patterns of vertical partisan competition that mirror those of more conventional policy areas in recent years.
The COVID-19 1 pandemic has swiftly become a global health crisis that is challenging governmental capacities around the world to contain the spread of the virus and provide health services to the infected. While critics, domestic and abroad, have largely panned the U.S. federal response efforts in initial weeks, state governors, in contrast, were quick to take action as the pandemic arrived on their doorsteps (Nossel, 2020; Rubin, 2020; Slaughter, 2020). For example, on February 29, Washington Governor Jay Inslee, at the early epicenter of the outbreak in the U.S., was the first U.S. elected official to formally declare a state of emergency in response to COVID-19. California Governor Gavin Newsome followed suit within days, and by March 16, governors of all 50 states had officially declared emergencies (Povich, 2020). Notably, this is the first time in American history that every state chief executive has done so in response to a public health crisis (Federal Emergency Management Agency [FEMA], 2020). While partisan politics tend be set aside during such times, the timing of gubernatorial action is telling about how partisanship shapes the way elected officials are reacting to this pandemic. For instance, 20 of 23 Democratic governors declared emergencies in their states before President Donald Trump declared a national emergency on March 13, while less than half of Republican governors did (12 of 27).
Historically, emergency declarations tend to be in response to natural or man-made disasters that tend to be defined by unplanned, dangerous situations that threaten human life for a temporary period of time, and as a result, create social or economic disruption (e.g., hurricanes in the Gulf Coast states, wildfires in the West) (FEMA, 2020; Rutkow, 2014; Rutkow & Vernick, 2017). Less common are instances of emergencies related to public health crises. While also unplanned, these events tend to be viewed as less socially or economically disruptive and/or less acutely threatening to human life in the short-term. Additionally, mitigating public health crises typically unfolds over the long-term, which diminishes the temporary element of an emergency. Of course, this has not stopped experts from using “public health emergency” in political discourses to describe events such as the Zika virus (Gulland, 2016), the opioid addiction crisis (Gostin et al., 2017), or even gun violence (Koop & Lundberg, 1992). While a select few states issued public health emergency declarations to address H1N1 and the opioid addition crisis, this has been rare (Rutkow, 2014; Rutkow & Vernick, 2017). Nevertheless, the COVID-19 pandemic has been distinctly different from previous public health crises in that the timing, potential severity, and disruption to social and economic norms in the U.S. has been more akin to a disaster, which has been a major catalyst for elected officials to employ the “emergency” moniker and invoke powers reserved for dire situations.
For governors, declaring an emergency has both substantive and symbolic benefits. While authorizing legislation differs from state to state, once a governor declares an emergency via an executive order, she is granted an array of powers in addition to the broad crisis management authorities possessed during non-emergency situations. 2 These may include: activating emergency response plans and/or operations centers; access to funds and/or personnel, equipment, or supplies reserved for critical situations; statutory immunities and liability protections for responders; and, suspension, waiver, or streamlining of rules or regulations that may hinder response efforts (Association of State and Territorial Health Officials [ASTHO], 2020). From an administrative perspective, declaring an emergency “change[s] the legal and operational landscape in which governments, private organizations, and the public operate” (ASTHO, 2020, pp. 1–2). Additionally, there may be symbolic value in as far as governors use their moral authority to focus public attention. Louisiana Governor John Bel Edwards summed this sentiment up on March 11, when he declared an emergency: “We want to make sure people take this serious” (Hilburn, 2020). Specific to COVID-19, a March 13 to 14 poll found that 72% of respondents trusted information from state and local governments a great deal or good amount, while only 50% trusted the news media the same amount, which would suggest that gubernatorial declarations and state public awareness efforts are likely to have a bigger impact on the public than news coverage in terms of communicating the severity of the pandemic (Montanaro, 2020).
Despite the litany of scholarship on crisis and emergency management across several disciplines (e.g., public administration, public health), the literature on why emergencies are declared is rather thin (Noe & Furay, 2013). Among the few studies, Rutkow (2014) finds that most emergency declarations are used for coordinating state agencies and resource planning, particularly if out-of-state resources may be needed to meet demand. Furthermore, in comparison to disaster declarations that focus on managing aftermath (e.g., loss of electricity, unsanitary conditions), public health declarations tend to focus on preventive and/or containment measures to reduce the spread of illness (e.g., prevention and treatment activities). To this end, in analyzing the seven states that declared emergencies in response to the H1N1 outbreak in 2009, Rutkow (2014) determined that declarations were at least partially driven by the first confirmed cases within states or by federal actions (i.e., U.S. Department of Health and Human Services determining H1N1 constituted a public health emergency). This would suggest that governors may take some cues from the federal government, but are also responsive to the particular circumstances occurring within their state. Notably, five of those states were led by Democrats and two by Republicans, with the Republican governors leading the pack on emergency declarations. However, previous scholarship provides little insights into how partisan polarization and/or contested federalism that define the contemporary political landscape impact gubernatorial reactions during emergency situations.
As every good crisis tells us something about our institutions, we examine gubernatorial emergency declarations in order to understand the deeper party cleavages that manifest during this unprecedented challenge and have thrust us into uncharted territory. Given the propensity for Republicans to trust in President Trump and/or unwillingness to challenge his policy leadership, Republican governors were far less likely to consider COVID-19 as a serious threat early on as the White House strategy was to downplay the severity of the pandemic. As the White House strategy shifted towards a more aggressive campaign, marked by the announcement of a national emergency on March 13, Republican governors began following suit by declaring their own emergencies, but presidential job approval serves as an important conditional factor in this trend. On the other hand, the inherent dissatisfaction that many Democrats have in President Trump, their high levels of concern for social issues (including as public health), and their higher levels of trust in scientists and health experts, suggests that Democratic governors were much more likely to read the early White House strategy of downplaying as a lack of leadership or as another sign of an ineffective administration. As such, most Democratic governors likely believed that the federal government was ill-prepared for the pandemic, so they began to use their emergency powers as platforms to draw citizen attention to and to prepare their governments for a public health crisis.
Governors as Partisan Actors
Given the global scale at which this pandemic has spread and previous experiences with outbreaks of Severe Acute Respiratory Syndrome (SARS) in 2002 and H1N1 influenza in 2009, it is unprecedented to see states take a more aggressive approach to containment than the federal government. However, states have become increasingly active in contesting federal leadership and shaping policy in recent years, and this activism has taken on a distinctly partisan character. Jensen (2017) contends that while there is no specific pivot point, governors are “mirroring an increasingly polarized landscape around them” (pg. 314) which emerged sometime in the early to mid-2000s. Specifically, governors are using their positions to engage in partisan conflict through federalist institutions by “challenging the federal government when it is controlled by the opposing party” (Goelzhauser & Rose, 2017, p. 286). This manifests not only in the way governors advocate for policies within their own states, but also how they lobby federal officials and wade into conflict over disputed policy areas. Legislative gridlock and the appearance of federal inaction on key policy issues, such as climate change, only serves as a catalyst for more aggressive state action, as state elected officials find political advantage in occupying a contested policy space, advocacy groups venue shop for policymaking bodies friendly to their cause, and voters of the opposing party become dissatisfied with presidential or Congressional actions (Conlan & Posner, 2016; Jensen, 2017; Rose & Bowling, 2015; Rose & Goelzhauser, 2018).
In a broader contest, this is a manifestation of contested federalism with roots in the mid-2000s as states and the federal government ran into more conflict over who would lead on emerging policy issues. In large part, this resulted from a shift towards bottom-up activism from subnational governments and the federal government attempting to reclaim policy spaces that states occupied in response to national inaction (Gamkhar & Pickerill, 2012; Riverstone-Newell, 2012). For instance, throughout the George W. Bush administration, states led the way on climate change as the federal government largely followed a policy of benign neglect, but under President Barack Obama, the national government tried to wrestle away control, leading to conflict over who controlled the climate policy agenda (Rabe, 2011). In many cases, the federal government serves as a mitigating force between states pushing for more progressive policies at the national-level and states pushing to uphold the status quo. Consequently, the federal government tends to find itself being challenged for doing too much by one group of states and not doing enough by another group of states, with these challenges being at least partially driven by partisanship as policy preferences are more likely to align if state and national leaders share a partisan affiliation.
While asymmetries in federal-state relations are nothing new, scholars argue partisan conflicts in federalism have only grown more brutal in the aftermath of the 2016 elections and a switch of Republican and Democratic positions on federal versus state authorities (Goelzhauser & Konisky, 2019). Since taking office, Trump has pushed new policies onto states relating to immigration, climate change, and healthcare through administrative action that have gone mostly unchecked by Congressional Republicans and Republican Governors, even though the Republican Party has traditionally advocated for states’ rights. On the other hand, Democrats, who have largely advocated for a big federal government since the New Deal and supported similar administrative actions by Obama, are now focusing their efforts at sub-national levels in order to challenge federal authority over these issues (Goelzhauser & Rose, 2017). To this end, Democratic state leaders have relished playing spoiler to the goals of the Trump administration by refusing to implement policies or filing lawsuits to block federal laws from being enacted (Somin, 2019). The result is a divergence in not only the types of policies that Republican and Democratic governors enact, but also how they react to federal leadership on any policy issue.
Furthermore, it is difficult to separate partisanship from ideological orientation, which may contribute to divergent response strategies to public health crises, like COVID-19, as there are fundamental inter-party differences in policy preferences. In general, Democrats tend to want more public services supported by a larger government, while Republicans prefer a smaller government that provides fewer services (Grossmann & Hopkins, 2015; Pew Research Center, 2015). To this end, a majority of Democrats now strongly favor or somewhat favor a public health insurance option or a national health plan, which would indicate that they view public health as a public, rather than private, good (Lopes et al., 2020). Thus, simply from the perspective of partisan orientations toward the role and size of government, it seems likely that Democratic governors will favor using the government to address public health issues, and therefore, be quicker or more aggressive in their response to COVID-19, including declaring a state of emergency.
While this pandemic is pushing government into uncharted territory, how political parties orient themselves towards this crisis is likely informed by citizen preferences, and previous scholarship argues that citizen preferences and related health behaviors are highly correlated with partisanship (Baum, 2011; Gadarian et al., 2020; Kriner & Reeves, 2014). For instance, in analyzing 120,000 individual survey responses spanning the 1990s and 2000s, Kriner and Reeves finds that party identification is the single most important predictor of support for healthcare reform. Specific to public health crises, Baum (2011) found Republicans were far less likely to be concerned about swine flu or to become vaccinated during the 2009 H1N1 outbreak, in comparison to Democrats. This is largely driven by information availability and messaging in a fragmented media marketplace, where exposure to information that challenges preexisting beliefs is limited and individuals tend to obtain information from sources that confirm their preexisting beliefs (Baum, 2011; Gollust & Cappella, 2014). Thus, at the individual-level, Republicans and Democrats are exposed to different facts and opinions on the threat of public health crises, which is exacerbated by inconsistent messaging from elected officials (Baum, 2011; Deslatte, 2020). This ultimately leads to differential attitudes and behaviors as individuals rely on their partisan lens to interpret ambiguous circumstances.
Preliminary evidence indicates these trends extend to COVID-19. Gadarian et al. (2020) indicates “consistent partisan differences among Americans not only in terms of their desired public health and public policy responses, but also on health behavior, like hand-washing and social distancing practices” (p. 1). Additionally, survey evidence from early March indicates that Republicans reported less concern about COVID-19 and less willingness to take individual action in response to the threat, compared to Democrats (Aleem, 2020; Badger & Quealy, 2020; Beauchamp, 2020; Reuters/Ipsos, 2020). Another poll from mid-March found that 76% of Democrats, compared to only 40% of Republicans, thought the coronavirus was a real threat (Montanaro, 2020). Furthermore, preliminary research (by political scientist Brian Schaffner) finds that Google searches for hand sanitizer were lower in U.S. media markets with higher proportions of the population voting for Trump, with this trend continuing until a national emergency was declared on March 13 (Beauchamp, 2020). This would indicate that a partisan divide is emerging among the public, which may further influence government responses. Anxiety and trust in experts may at least partially explain some of this trend. Looking at previous experiences with pandemics, Albertson and Gadarian (2020) find that “[w]hen people were more anxious about H1N1 and smallpox, they were more likely to trust the experts.”
While Democrats and Republicans report fairly high levels of trust in information about the coronavirus from public health experts and subnational governments, there is substantial partisan difference in citizens’ trust in the media with a majority of Republicans distrusting information on this subject (Montanaro, 2020). Granted that the media is the key avenue by which public health experts communicate with the public and the conflicting messages about potential severity of COVID-19, how citizens view the legitimacy of information and its sources may also be important in determining their views. For instance, a mid-March poll found that while a majority of Republicans (74%) trust information they hear from President Trump on the coronavirus a great deal or good amount, only 8% of Democrats do (Montanaro, 2020). With President Trump’s early dismissive approach to the coronavirus and critical comments about the media’s coverage of the pandemic, Republicans may be more willing to accept that COVID-19 poses little threat, while Democrats are much more skeptical of that position. Thus, partisan differences on this subject may be a function of an information divide, with Democrats highly anxious about the potential threat posed and Republicans far less concerned. Given these inherent partisan differences, we expect Democratic governors to be more likely to declare emergencies sooner than Republican governors, highlighting partisan divisions in state responses to this pandemic.
Pandemic Timeline
COVID-19 first appeared in December 2019 in Wuhan, China, but the first case in the U.S. was not confirmed until January 21, 2020 (Taylor, 2020). Throughout January and February, the virus spread worldwide. Due to the exponential growth in the number of cases and increased testing, numbers exploded throughout March. By March 26, there were 462,684 cases in 172 countries including 63,570 in the U.S., making the U.S. the new world leader in confirmed cases (McNeil, 2020; World Health Organization [WHO], 2020). Notably the total number of cases during these early weeks has been debated as the number of COVID-19 tests available to state and local public health officials were limited and many potentially infected individuals had yet to display symptoms (Buchanan et al., 2020). Despite uncertainty, state leaders began to prepare for worsening conditions. On February 29, Washington Governor Jay Inslee declared a state of emergency in anticipation of the same growth curve experienced by other countries, and by March 16, all 50 governors had declared emergencies in their own states (COVID Tracking Project, 2020; FEMA, 2020).
Experiences in other countries indicate that in the absence of policy interventions to restrict the exposure of uninfected populations by carriers, cases grow exponentially and lead to rapid influxes of patients that overwhelm health services, leading to higher mortality rates. On the other hand, early testing and efforts to isolate have been instrumental in “flattening the curve” of patients seeking treatment, allowing for healthcare systems to more effectively absorb new cases (Cyranoski, 2020; Godoy, 2020a; National Institute of Health [NIH], 2020; Stevens, 2020). Consequently, the Trump administration came under significant criticism for its early strategy, which focused on downplaying the severity of the pandemic and posturing that domestic response efforts had largely contained the spread (Costa & Gregg, 2020). Despite appointing a Coronavirus Task Force on January 29 chaired by Vice President Mike Pence and restricting travel to and from China on January 31, Trump largely attempted to minimize the severity of virus and its potential to overwhelm public health systems (Blake, 2020; Chalfant & Samuels, 2020). For instance, on January 22, he remarked “We have it totally under control,” and on January 30, “We have very little problem in this country at the moment—five. And those people are all recuperating successfully” (Blake, 2020).
This sentiment was reiterated in numerous statements and tweets throughout February as cases both domestically and internationally increased, and representatives of the Trump administration supported this messaging. For instance, on March 8, the U.S. Surgeon General, Dr. Jerome Adams, said: “what we do know is that if we had massive numbers of cases, we would be seeing more deaths, and so we actually feel pretty good that some parts of the country have contained it” (Tapper, 2020). The White House’s early strategy of downplaying was met with criticism from Democrats, particularly governors, who consistently pushed for a more aggressive approach to testing and containment. However, White House strategy shifted on March 13 as Trump officially declared the COVID-19 pandemic a national emergency in order to “unleash the full power of the federal government” (White House, 2020). On March 16, he also backed away from claims that the virus was under control, stating: “If you’re talking about the virus, no, that’s not under control for any place in the world. . . I was talking about what we’re doing is under control” (Blake, 2020). In other words, the strategy moved from downplaying to acknowledging the severity of the pandemic.
Although gubernatorial statements justifying emergency declarations struck similar tones in terms of the need to prepare for a fluid crisis, protect vulnerable populations, and prevent healthcare services from being overburdened, they were relatively ambiguous as it relates to motivations and timing. However, some governors did indicate the sources of influence in their decisions. For instance, Georgia Governor Brian Kemp (Republican) explicitly indicated the White House’s influence in his decision when announcing his emergency declaration on March 14, stating: “Based on President Trump’s emergency declaration, I will declare a public health emergency” (McCullough, 2020). On the other hand, North Carolina Governor Roy Cooper (Democrat) explained on March 10: “We’ve analyzed the situation using available data and science. We’ve discussed options with health experts, business leaders and emergency management here and in other states” (Pitofsky, 2020). Notably, Georgia’s first confirmed case was on March 2 and on March 14, when Kemp declared an emergency, there were 99 cases; in comparison, North Carolina’s first case was on March 3, and on March 10, when Cooper declared an emergency, there were seven cases (COVID Tracking Project, 2020). While this is anecdotal, it would suggest that Democratic governors were more reliant on health expertise and the conditions within their own states when making decisions, while Republicans governors were taking cues from the White House.
Methods
Data and Dependent Variables
In order to find systematic evidence of the relationship between partisanship and the timing of gubernatorial emergency declarations, we use a quantitative analysis. While efforts to research COVID-19 and the corresponding response efforts are beginning in earnest, there are currently few published datasets on the political and/or policy dimensions. As such, we collected data from a number of publicly available sources, including news reports, the U.S. Census Bureau, and the COVID Tracking Project. Specific to news reports, we confirmed any data via multiple sources in order to ensure as much accuracy as possible. Given the circumstances of the COVID-19 pandemic, we attempted to measure variables with as much sensitivity to fluctuations between states and over time as possible in order to capture the dynamisms of the situation. However, there are some variables of interest that are observed at the national-level and do not fluctuate between states (e.g., presidential actions), and others that fluctuate between states, but are consistent over time (e.g., gubernatorial partisanship).
Our dependent variable is a declaration of emergency by governors, and is coded as such for the date in which it is officially announced. 3 As of March 16, all 50 states have declared an emergency in response to COVID-19. Unsurprisingly, given that he was at the early epicenter of the outbreak in the U.S., Washington Governor Jay Inslee was the first to declare an emergency on February 29 (Povich, 2020). Conversely, West Virginia Governor Jim Justice was the last to do so on March 16; notably, West Virginia was the last state to confirm a case within its borders (Mistich & Lofton, 2020). The median dates for declarations occurred on March 11 and 12, during which 12 governors announced emergencies; the mode case is March 13, when 13 governors did so. We obtained data on emergency declarations by reviewing news media coverage, and then confirmed dates via state websites.
Predictor Variables
Our primary predictor variable is gubernatorial partisanship, which we use to test our central hypothesis that there is a partisan division in responding to this pandemic (i.e., Democratic governors are likely to declare emergencies sooner than Republican governors). Currently, there are 27 Republican and 23 Democratic governors. Given the polarizing responses to President Trump and his leadership, we also include net presidential job approval ratings by state, which are measured as the portion of respondents approving of his job performance minus those disapproving. We obtained data from Morning Consult (2020), with the most recently available job approval ratings based on survey data from February 20. Given that this was before any state emergency declaration were issued, this represents the most recent dated that is not unduly biased by the political timeline of the pandemic. Additionally, as President Trump’s leadership on this issue has been a particular point of scrutiny in recent weeks, we also use a dummy variable to account for the shift in White House strategy by comparing two time periods: (1) from January 21 to March 12 when the focus was on downplaying the crisis; and, (2) after March 13, when President Trump declared a national emergency.
If governors are responding to federal leadership (or lack thereof), then they should react differently before and after the shift in White House strategy. Additionally, there may be an interaction between partisanship and federal leadership, where Democratic governors are more likely to declare emergencies while the Trump administration is downplaying the severity of the pandemic, while Republican governors are more likely to do so after the Trump administration pivots their strategy towards acceptance of the threat. Furthermore, this may be a conditional relationship that depends on how popular President Trump is in specific states, with higher popularity likely leading governors to follow Trump’s lead (i.e., waiting on a national emergency declaration) and lower popularity likely leading governors to challenge the White House’s strategy of downplaying the problem severity (i.e., declaring emergencies sooner, rather than later).
We also use variables to control for the problem severity and population vulnerability within each state. To capture the severity and progression of the pandemic within specific states, we include four variables: the number of confirmed cases, the total confirmed U.S. cases, state cases as a percentage of U.S. cases, and days since first confirmed case within the state. To date, there has been extensive criticism of the availability of tests and the ability to confirm cases of the virus. To that end, there have been presumptive but unconfirmed cases in some states which has perpetuated speculation about the actual number of cases. We assume that in assessing the situation in their state, governors are relying on valid and reliable data on cases (i.e., confirmed cases), rather than speculative news reports. Nevertheless, this is a key limitation to this analysis because the confirmed cases are an indicator of the extent of the pandemic, but it may not be a wholly accurate measure. Since the number of infected cases tend to increase exponentially during pandemics and EHA assumes linearity (Box-Steffensmeier & Jones, 2004), we use the natural log of the count variables as a linear transformation. Additionally, we use a count of the days since the first confirmed case to control for the temporal dimensions of when cases actually appeared in specific states. We obtained data on the number and timing of confirmed cases from the COVID Tracking Project.
Finally, to capture the relative vulnerability of state populations, we include percentage of population over 60 years of age and population density. Initial epidemiology reports indicate that those over the age of 60 are both the most susceptible to COVID-19 and have the highest mortality rate, while those under the age of 60 are more likely to carry the virus asymptomatically (Godoy, 2020b; Guan et al., 2020). Consequently, governors in states with older populations may be more concerned for the potential havoc the virus could wreak if it were to spread, while those with younger populations may be more concerned about the undetected spread of the virus. Additionally, global health experts have advocated for social distancing as a mechanism to limit person-to-person transfer of COVID-19. Of course, in areas with high population densities, social distancing is more difficult and likely to be less effective, compared to rural areas where residents already enjoy a higher degree of interpersonal space. We obtained data on both age of population and population density from the U.S. Census Bureau; these variables vary between states, but not over time.
Event History Analysis and Risk Set
In keeping with previous scholarship that examines the temporal and spatial distribution of state-level policy interventions, we employ event history analysis (EHA). The utility of which is that it allows us to determine the probability of a qualitative policy change occurring during a specific time period (Berry & Berry, 2018; Box-Steffensmeier & Jones, 1997; Drolc et al., 2019). In this case, the qualitative policy change is state declarations of emergency, representing a special case of policy adoptions that occurred within an abbreviated time period in response to the spread of COVID-19. The advantage of EHA over more conventional methods of analyzing cross-sectional data is that it takes into consideration the fluctuation of variables over time, and indicates probability of not only policy adoptions occurring but policy adoptions occurring within a specific period of time (Box-Steffensmeier & Jones, 1997; Drolc et al., 2019). The latter is of specific interest here as the timeline of COVID-19 responses has been an important point of debate and criticism (Povich, 2020).
While EHA works similarly to other techniques that rely on a dichotomous dependent variable (i.e., analyzes how predictors affect the probability of a two-category outcome variable), the data is structured in a unique way, and is known as the risk set. Specifically, the data is treated differently for each of three time periods: pre-adoption, adoption, and post-adoption. During the pre-adoption period (which may include several observations per state), states that have yet to declare an emergency are observed, and the outcome variable is coded as 0. During the adoption period (which only includes the single time period in which adoption occurred), states are observed, and the outcome variable is coded as 1. Any observations following adoption (i.e., post-adoption) are then dropped from the dataset, as those states are no longer at risk of declaring an emergency (Box-Steffensmeier & Jones, 2004). Here, our observed time periods are days, so our unit of analysis is “state-days.” While previous studies tend to use state-years (i.e., observing time in year periods), the nature of these events and the availability of information suggest that days are the more appropriate unit. Our risk set, then, includes 2,530 state-days.
We define the risk set as states at risk of declaring emergency in response to COVID-19, and include all 50 states from January 21 (i.e., date of first confirmed COVID-19 case in the U.S.) until the date in which an emergency is declared. Note that all states declared an emergency by March 16 (i.e., final date in the dataset). The parameters of the risk set are important in both a methodological and theoretical sense. First, given the abbreviated timeline with conditions changing on a daily basis and with unique circumstances for each state, our risk set is designed to be sensitive to both of those parameters. Second, by beginning it on January 21, we are assuming that once COVID-19 was confirmed to be present within the U.S. that all states were at risk of declaring an emergency, and that risk for any specific state did not end until an emergency was actually declared by the governor. After defining the risk set, we tested alternative distributions of the relationship between time and adoption, and determined the Gompertz distribution to be the best fit for the data. Additionally, we clustered standard errors at the state-level, as we assume standard errors within each state are correlated over time. We report hazard ratios, which function similarly to odds ratios but represent the ratio of probability of an event occurring (i.e., emergency declaration) to unit of time (i.e., days).
Results
Figure 1 displays four graphs: (1) the Kaplan-Meier failure estimates for the risk set; (2) the Kaplan-Meier failure estimates by gubernatorial party; (3) the smoothed hazard estimate for the risk set; and, (4) the smoothed hazard estimates by gubernatorial party. The Kaplan-Meier failure estimates (left-hand column) represent the portion of states declaring emergencies over time, while the smoothed hazard estimates (right-hand column) use a kernel function to compute a weighted average in plotting the hazard estimate. The failure and hazard estimates for the risk set are the best illustration of the overall timeline of state emergency declarations, with March 6 to 8 appearing to be a pivot point where the rate of declarations increases significantly in the ensuring days. Notably, 15 states confirmed the first case during that three-day span. The graphs by gubernatorial party also indicate a gap in the timing for Democratic and Republican governors that begins with the first emergency declarations and continues throughout the timeline. The median date for declarations by Democratic governors is March 10, and March 13 for Republican governors. On its face, this would suggest that partisanship has played a role in responses to the COVID-19 pandemic.
Kaplan-Meier failure and smoothed hazard estimates.
We use EHA models of state emergency declarations to provide further statistical analysis of these trends. Since our primary interest here is how partisanship affects gubernatorial action, model 1 focuses on the gubernatorial partisanship dummy variable as a simple dichotomous comparison, while model 2 estimates the effects of presidential job approval for governors of both political parties and model 3 estimates the effects of the interaction between gubernatorial partisanship and presidential job approval (see Table 1). Hazard ratios from model 1 indicate that Democratic governors were about 2.67 times quicker to declare a state of emergency than Republican governors, which largely conforms to our expectations of a partisan gap in responses to the pandemic. Additionally, the hazard ratio for Trump’s job approval indicates that as Trump becomes more popular in states, governors are slower to declare emergencies. This is consistent with findings from model 2 for both Democratic and Republican governors; however, this relationship is only statistically significant for Republicans, indicating no generalizable effect for Democrats.
Findings from EHA Analysis.
Note. Hazard ratios reported. Alpha levels *<.05, **<.01, ***<.001. Standard errors in parenthesis.
Model 3 provides further nuance to this by estimating the interaction effects. While the directional relationships for partisanship and presidential job approval are consistent with model 1, the interaction effect indicates that Trump’s popularity had a pronounced affect in influencing Republican governors in declaring emergencies, but little to no effect on Democratic governors (see Figure 2). Specifically, there appears to be no change in the likelihood of Democratic governors (dotted line in Figure 2) declaring emergencies across levels of Trump’s job approval. In comparison, there is a clear relationship between Trump’s job approval and timing of emergency declarations for Republican governors (solid line in Figure 2), with higher approval correlated with longer timelines for declaring emergencies. In fact, Republican governors in states where Trump is particularly unpopular may even be quicker to declare emergencies than their Democratic counterparts (e.g., Maryland Governor Larry Hogan was the first Republican to declare an emergency on March 5 in a state where Trump’s net approval is -24).
Graphs of effects of presidential job approval, models 3 and model 5.
To add further perspective, model 4 estimates partisan effects when considered in the context of the shift in White House strategy (i.e., before vs. on or after March 13), and model 5 estimates how presidential job approval influences decisions before and after March 13 for governors of each party (see Table 2). For model 4, we provide estimates with different base categories in order to understand the various interparty and intraparty comparisons that can be made. Focusing on the interparty comparison, we find that Democrats were about 2.18 times quicker to declare emergencies before March 13 than Republicans. While the hazard ratios also indicate that Democrats were slightly quicker (i.e., 1.34 times) to do so after March 13, the findings are not statistically significant, indicating that interparty differences were less consistent during this period. This suggests that interparty differences were most likely to emerge while the White House attempted to downplay the severity of the pandemic.
Findings from EHA Analysis.
Note. Hazard ratios reported. Alpha levels *<.05, **<.01, ***<.001. Standard errors in parenthesis.
Comparing governors of the same party before and on or after March 13, Democratic governors were roughly 5.00 times quicker to declare emergencies before March 13, which again would suggest that they were reacting to muted federal leadership. On the other hand, we find no statistically significant difference between Republicans before and after March 13, but findings from model 5 provide nuance to this relationship. Specifically, there is only a statistically significant relationship between presidential job approval and gubernatorial partisanship for Republicans declaring emergencies on or after March 13, with Trump’s popularity positively correlated with delayed state declarations. Conversely, presidential job approval has no generalizable effects on Republican governors declaring emergencies before March 13 or on Democratic governors in general. The differential effects of Trump’s job approval on Republicans before and on or after March 13 indicates Trump’s job approval is an important explanatory factor for why some governors waited until March 13 or later to declare emergencies (solid line in Figure 2). On the other hand, for Republican governors declaring emergencies after March 13, Trump’s job approval had little or no effect (dotted line in Figure 2).
Looking at descriptive statistics furthers this point: Trump’s net job approval averaged −0.9 in the 12 states with Republican governors declaring emergencies before March 13, but +8.2 in the 17 states declaring emergencies after March 13. In other words, Trump is a relatively neutral figure for Republican governors that declared emergencies early, but a popular figure for those who waited. In comparison, Trump’s net job approval averaged -8.4 in the 20 states with Democratic governors declaring emergencies before March 13, compared to -10.3 in the three states declaring emergencies after March 13 (Morning Consult, 2020). Again, this would indicate that Trump’s popularity is not an explanatory factor in the timing of emergency declarations made by Democratic governors, but it is one in explaining why some Republican governors waited on the White House before declaring emergencies. Consequently, governors fall into three camps: (1) Democratic governors (who were most likely to declare emergencies while the White House downplayed the severity of the pandemic); (2) Republican governors in states where Trump is generally popular (who were most likely to declare emergencies after the White House signaled that the pandemic was in fact a real and pressing threat); and (3) Republican governors in states where Trump is generally unpopular (who followed no consistent pattern in declaring emergencies, which suggests some mimicked Democrats while others towed the Republican party line).
These models also indicate that state population characteristics played a role in the timing of emergency declarations. Specifically, as portion of state populations over 60 increased, governors were less likely to declare emergencies. On one hand, this could be explained by the potential that younger populations are more likely to spread the virus asymptomatically and expose older populations who are more vulnerable to severe illness and death. On the other hand, younger populations may be more concerned about the pandemic than older populations, which could indicate that governors are responding to public preferences rather than population vulnerability in determining how to proceed; however, initial survey data is conflicting on this point (Butchireddygari, 2020). Additional findings indicate that governors of states with higher population densities were quicker to declare emergencies declarations than governors of lower density states. Interestingly, this would suggest that governors may have been at least partially motivated in their decisions by the difficulty of implementing social distancing and other public health recommendations. If so, they may have declared emergencies earlier in order to give their citizens more time to prepare.
Findings for confirmed cases are a little more difficult to interpret. While confirmed cases at the state-level, percentage of U.S. cases, and days since first confirmed case were not statistically significant predictors, confirmed cases at the national-level indicated a higher number of cases resulted in later declarations. One issue potentially skewing these results is that the pandemic has spread asymmetrically across states, so that states hit early (e.g., Washington) waited weeks to declare emergencies while states hit later (e.g., Idaho) did so in response to the first confirmed case. Additionally, as the pandemic spread across the U.S., areas where it had yet to reach were more likely to be aware of the worsening conditions at the national-level than those early states as media outlets began to more aggressively cover the story. This would suggest that there is some interaction between the number of confirmed cases and the time period in which those cases were confirmed, where the substantive impact of new cases is dependent on rapidly changing conditions and available information. Thus, smaller numbers of confirmed cases later maybe more predictive than larger numbers of confirmed cases early on, because governors had more data on the emerging situation as the pandemic progressed.
Furthermore, not captured here are the presumptive cases that were not confirmed due to the lack of early testing, which remained an issue weeks after the initial cases. In many states, there was speculation by media that the official number of confirmed cases did not reflect the actual extent of the pandemic, creating an invisible problem (Baird, 2020). Therefore, governors may have also been influenced by the number of presumed cases or been reluctant to infer too much from confirmed cases in the absence of systematic testing efforts. While it is likely that governors were influenced by the growth in the number of cases within both their states and at the national-level in deciding to declare an emergency, the specifics of how they considered that information is not clear, and this relationship may be more complicated than what is modeled here and requires additional research.
Discussion
Party polarization has played a central role in contested federalism in recent years, as state elected officials use federal institutions to either challenge or acquiescence to national policy directions, depending on which party controls the White House. While the Republican Party has traditionally favored limited national government, they have largely embraced top-down leadership since Trump’s election. In contrast, the Democratic Party has taken up the mantle of opposing federal overreach in favor of states’ rights. Of course, as recently as the Obama administration, partisan stances were flipped, which would generally suggest this is less about ideology and more about partisan conflict. In other words, governors are likely to support national leadership if they are of the same party as the president, but advocate for states’ rights when they find themselves in the opposing party. Scholars have documented these trends across several policy areas, including immigration, health care, and climate change (Goelzhauser & Rose, 2017; Rose & Goelzhauser, 2018). However, the COVID-19 pandemic is an unprecedented crisis; one so severe that many may expected bipartisan sentiments to replace partisan animosity, if even for a short time. Unfortunately, our findings suggest that is not the case.
To this end, President Trump’s messaging in the early stages of the pandemic in the U.S. and his popularity are key factors in how state governors managed the response on the front-line. On one hand, Democratic governors were consistent in that they were willing to quickly occupy the leadership void on this issue in order to focus public attention and prepare governmental responses, as the White House downplayed the potential threat that COVID-19 posed. On the other hand, two varieties of Republican governors emerged: those who lead states in which Trump is a popular figure, and those who lead states in which he is not. Governors of the former were much more likely to follow Trump’s lead, and resisted calls to sound the alarm on COVID-19 before the leader of their party did. Conversely, governors of the latter were less willing to wait on the White House to shift their strategy. Although party doctrine and citizen perceptions may partially account for initial reactions, the evidence, and the unprecedented nature of the situation, would suggest that partisan gaps in the timeline of emergency declarations are as much about vertical partisan conflict as ideologies.
These trends are interesting in that they help clarify some of the competing messaging, recommendations, urgency placed on efforts to contain the pandemic, and the emergence of governors as key political actors (e.g., New York Governor Andrew Cuomo’s starring role as a critic of the federal response (McKinley & Goldmacher, 2020)). Additionally, it is notable that partisan polarization has seeped into how our government is responding to one of the most significant public health crises of the modern age. But, this begs the questions of whether partisanship will ultimately impact the extent or severity of COVID-19 in the U.S., or whether the timeline is a moot point. While still being investigated, several early studies find that government stay-at-home orders and social distancing restrictions were effective in helping slow or reduce the spread of COVID-19 (Courtemanche et al., 2020; Fowler et al., 2020). Evidence from other countries also suggest early intervention and containment is key (Sepkowitz, 2020). These initial findings suggest the partisan differences in gubernatorial behavior may have significant effects on outcomes, including rates of infection and/or mortality as well as economic impacts. Consequently, the ability of states to respond to the pandemic clearly demonstrate why and how states matter, even when the subject is of national and international importance.
Still, the COVID-19 pandemic is an evolving and fluid situation at a global-level, so while this provides some initial evidence on how partisanship and federalism are shaping governmental response efforts, more questions and evidence are likely to emerge as it progresses. For one, given that our analysis only takes into consideration initial actions, it is unclear how the partisan gap or vertical competition will continue to shape policy choices moving forward. Unfortunately, evidence from mid-March is not a promising indicator of a bipartisan future. For instance, while Congress at first seemed to take a bipartisan approach to relief, partisan conflict has re-emerged (Shabad et al., 2020). Additionally, as Democratic governors like Andrew Cuomo (Cole & Duster, 2020), Gavin Newsome (Conradis & Coleman, 2020), and John Bel Edwards (Ballard & Karlin, 2020) call for more aggressive containment measures, President Trump (Forgey et al., 2020) and Republican state politicians, such as Texas Lieutenant Governor Dan Patrick (Hennessy-Fiske, 2020), are already pushing towards an end to containment efforts in order to restart the economy. Furthermore, Republican governors followed a similar trend in issuing stay-at-home orders to declaring emergencies, in that they were on average 4 days behind their Democratic counterparts (March 28 compared to March 24) (Mervosh et al., 2020).
However, things may have evolved. While it is difficult to make comparisons about how social distancing and stay-at-home restrictions were lifted due to the differing stipulations and phased reopening plans across states, it appears that inter-party differences dissipated by the time restrictions were lifted, with Republican governors beginning to reopen parts of the economy around May 11, on average, and Democrats, around May 12 (Elassar, 2020). In general, messaging from governors would suggest that reopening plans were more reliant on structured administrative decisions that incorporated input from public health experts and economists, as compared to initial decisions on stay-at-home orders that were made in an environment of more uncertainty. On its face, this may indicate that when rapidly responding to ambiguous policy challenges, governors fall back on partisanship as a guiding light, but when afforded time and opportunity to consider policy alternatives, the influence of partisanship dissipates. If so, this may also partially explain the perplexing findings around the number of cases in that governors may have interpreted the same numbers differently due to the ambiguity in what they represent (Zahariadis, 2003). The evolving nature of this pandemic and the preliminary nature of our evidence makes it difficult to draw many conclusions on this point, so additional inquiry is needed as further evidence becomes available on both the timing and impact of decisions. The questions of partisanship need to be reevaluated over the coming weeks and months in order to understand how it impacts policy decisions, especially when those policy decisions may very well be directly connected to life and death.
Footnotes
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) received no financial support for the research, authorship, and/or publication of this article.
