Partisanship, Policy, and Americans' Evaluations of State-Level COVID-19 Policies Prior to the 2020 Election.

The COVID-19 pandemic was a key policy issue during the 2020 election in the United States. As such, it is important to analyze how voters evaluated government responses to the pandemic. To this end, in this article, we examine factors that influenced Americans' evaluations of state-level COVID-19 policy responses. We find that during the pandemic onset period, Americans typically rated their state governments' responses more favorably if their governor was a co-partisan. In contrast, during the re-opening period, we find that Democrats relied on both partisanship and policy to evaluate their state-level responses, while Republicans continued to rely solely on partisanship. We contend that given the complex policy environment surrounding COVID-19, Americans may have not been fully aware of the policies their state governments adopted, so they relied on partisan cues to help them evaluate their state-level policy responses. But by the re-opening period, Americans likely had enough time to better understand state-level policy responses; this allowed Democrats to also evaluate their state-level responses based on policy. These findings shed light on how Americans evaluated COVID-19 responses just prior to the 2020 election.

Introduction

In the early months of 2020, policymakers throughout the US suddenly had to address a complex public health policy problem at the start of a contentious presidential election year.[1] A novel coronavirus had developed in China and as it spread throughout the world, it became clear that it was highly contagious and potentially more deadly than similar respiratory viruses. As lawmakers scrambled to develop policy responses to mitigate the spread of COVID-19 throughout the US, public officials and Americans became divided along partisan lines about the government’s proper role in addressing the crisis. Democrats favored a more stringent response, but such a response could come with significant economic costs. Republicans, on the other hand, favored a less-stringent response from the government, which meant COVID-19 could spread more quickly and overwhelm health care systems throughout the US. It appeared that whichever approach Democrats and Republicans adopted in responding to the virus, there would negative side effects and potential electoral consequences in November.

President Trump, a Republican, failed to adopt a strong national-level response to the pandemic, so state governments, led largely by governors, began adopting policies to address the crisis. On average, Republican governors adopted less strict responses while Democrats adopted stricter responses (Fowler, Kettler, and Witt 2021). Evaluations of these policy responses from public officials and media pundits dominated the news cycle, with Republicans typically criticizing Democrats for overreaching and Democrats criticizing Republicans for not doing enough to mitigate the spread of the virus. Yet to be explored, however, is how state-level COVID-19 policy responses influenced individual Americans’ evaluations of state-level COVID-19 responses. Given that COVID-19 was the defining policy issue of the 2020 elections, it is vital to understand how voters evaluated how well actors at various levels of government handled the crisis.

In this article, we explore how the adoption of COVID-19 mitigation policies and their effects influenced Americans’ evaluations of state-level officials’ handling of the crisis. We have two theoretical expectations. We expect that both political considerations (partisanship of one’s governor) and policy considerations (stringency of COVID-19 response) will influence attitudes towards state-level COVID-19 response. First, if Americans care about policy outcomes, we should expect Democrats to have rated their state-level policy responses more highly the more stringent the policies became. In addition, we should expect Republicans to have rated their state-level policy responses more highly the less stringent the policies became. We should also expect that if per capita case rates and unemployment rates are lower, then Americans will rate state governments’ responses better. Nevertheless, given that COVID-19 was a complex policy issue, it is possible that Americans were not fully aware of the exact policies their state governments adopted and what the side effects of those policies were. In such an environment, individuals likely relied on partisan cues to evaluate their state’s policy responses. In other words, it is possible that Americans evaluated state-level policy responses higher if their governor was a co-partisan.

To explore how COVID-19 policies, their outcomes, and individual partisanship influenced Americans’ evaluations of state-level responses, we used data from two public opinion polls, two survey experiments, and data on the stringency of US state-level policy responses to COVID-19 (Shvetsova et al. 2020b). The findings of our analyses suggest that during the pandemic onset period, partisanship, not policy, influenced Democrats’ evaluations of their state-level COVID-19 policy responses. Democrats were simply more satisfied with their state-level COVID-19 responses if their state was led by a governor who was a Democrat. Yet, we also find that during the re-opening period, both partisanship and policy outcomes influenced these evaluations. Democrats remained satisfied if a Democrat led their state and that satisfaction increased the stricter their state-level COVID-19 responses became. For Republicans, we found that only partisanship influenced their evaluations during the pandemic onset and re-opening periods, not policy outcomes. Republicans were simply more satisfied with their state-level policy responses if their state was led by a Republican governor.

The findings of this analysis shed light on how voters evaluated COVID-19 responses at the state level during the election. While state-level responses did not seem to influence Republicans’ evaluations, they did seem to influence Democrats’ evaluations in the second half of 2020. It is clear that Democrats favored a stronger and more effective response at the state level. Heading into the November election, these evaluations likely weighed on their minds as they decided whether to vote and who to vote for. Hence, it is unsurprising that just prior to the 2020 election, slightly more than 80% of Democrats said the state of health care and the COVID-19 outbreak were very important issues they were considering when making their vote choices (Pew Research Center 2020).

In the next section, we briefly review the US government policy response to COVID-19. Then, we discuss Americans’ attitudes about and their evaluations of these responses. We then provide support for our expectations using data from public opinion polls and two experiments. The last section discusses the implications of the results in understanding how COVID-19 evaluations influenced the 2020 election.

COVID-19 and the US Public Policy Response

On 31 December 2019, the WHO learned that there was an outbreak of a new respiratory illness in Wuhan, China. On 9 January 2020, it identified the illness as a novel coronavirus. As the virus quickly spread throughout Asia and then Europe, it became clear that it was highly contagious, hard to treat, and may have a high death rate compared to other similar respiratory illnesses. Early conclusions from public health officials and policy makers suggested that unmitigated spread could overwhelm unprepared healthcare systems. On 11 March 2020, the WHO declared the COVID-19 outbreak as a pandemic (World Health Organization 2020).

The US federal government began responding to COVID-19 in early January 2020 when it placed travel restrictions for travelers from Wuhan. As the virus spread throughout the world, the federal government began to adopt additional travel restrictions from countries with cases of known community transmission. The CDC began providing guidance to the public about how to mitigate the spread. However, throughout most of the outbreak, President Trump downplayed the threat and sent signals to the public that conflicted with advice from public health experts. Compared to other national governments, the US federal government adopted relatively few non-pharmaceutical measures to address the crisis and left it to the states to mitigate the spread (Adeel et al. 2020; Keith 2020; Shvetsova et al. 2020a; Taylor 2020).

It is not surprising that the President was reluctant to respond to the crisis. It was an election year and non-pharmaceutical mitigation measures such as shutting down businesses and stay at home orders would come at a significant economic cost. Economic well-being factors into many voters’ decision to either reward or punish the incumbent President’s party. An economic crisis months before the election could have significantly reduced the chances of Republican candidates winning election (Lewis-Beck and Stegmaier 2000). Hence, with the election looming in November, the President likely chose not to adopt strict mitigation measures in order to prevent a significant economic crisis that voters would blame Republicans for at the polls.

However, as perceptions of the economy have become increasingly polarized along partisan lines (Bartels 2002), there were likely other factors that could explain the President’s behavior. With respect to partisan considerations, Republicans and Democrats are highly sorted on economic policy and health care policy. One only need look into the recent past, under Obama’s administration, to observe the partisan divide over healthcare policy. Democrats favored the adoption of the Affordable Care Act and a strong response to H1N1 while Republicans opposed both types of measures (Baum, 2011; Gollust, Nagler, and Fowler 2020). It is possible, perhaps, that Trump’s identity as a Republican in an election year meant that public health considerations would be secondary to economic considerations. Additionally, as politics have become increasingly polarized (Mason 2018), Trump’s (and other Republicans’) distaste for COVID-19 mitigation measures could be explained by their desire to simply do what the Democrats did not support. While this may explain Trump’s downplaying of the virus as a “bad flu,” his racially charged comments referring to the “China Virus” or “Kung Flu,” do not fit nicely into this paradigm, suggesting that negative racial attitudes may also play a role in this (lack of) response.

The federal nature of the US political system can also explain, in part, why President Trump did not adopt strong mitigation measures. In a federal system, in the face of a complex policy problem with limited information, policymakers may delegate to other levels of government to solve the problem in order to avoid having to take responsibility (Greer et al. 2020). Republicans also tend to support state-led solutions to policy problems rather national-led solutions. Additionally, states can be laboratories for policymaking. Lawmakers at both the state and national level can learn from other states about which policies are effective and which are not (Boeckelman 1992; Shipan and Volden 2012; Shvetsova et al. 2020b, 2021). In sum, for a variety of reasons, the federal nature of the US political system may explain why the national government adopted a limited approach to mitigating the spread of COVID-19.

Work on policy response, generally, can also help to explain why we see such a politicized COVID-19 response in the U.S. When uncertainty exists with regard to the ideal policy response, as it did with the early stages of a novel disease outbreak, political considerations tend to become intertwined with the policy problem itself (Lindblom 1959). This can lead to disproportionate policy response, which is defined as either a policy overreaction, or a policy underreaction (Maor 2021). Importantly, different individuals can view the same policy response as disproportionate in either way, depending on their perceptions, especially when such a response is politicized (Conlan, Posner, and Beam 2014; Maor 2021). Major crises often force politicians into policy overreactions to give the appearance of doing something to combat them (Lodge and Hood 2002). Conversely, when major events occur, political leaders may underreact in attempt to avoid blame for policy proposals they are uncertain will work (Weaver 1986). Taken together, this literature suggests that the COVID-19 pandemic was uniquely situated to be politicized and to provide incentives for political leaders to both over and under react.

The diminutive response from the national government forced state governments, particularly governors, to take the lead in addressing the pandemic. With the partisan divide over economic policy and healthcare policy, there was a clear difference in the COVID-19 response by Democrat and Republican governors (Fowler, Kettler, and Witt 2021). To illustrate the differences in these policy responses during the early months of the worldwide spread of COVID-19, we use the state-level Protective Policy Indices (PPI) variable from the Institutional Origins of Protective COVID-19 Policies Dataset (Shvetsova et al. 2020b). The PPI is an index that measures how restrictive COVID-19 policy mitigation policies are on individuals in society. The index includes information about a variety of non-pharmaceutical mitigation measures such as school shutdowns, closing of businesses, and mandatory use of PPE that governments adopted. More stringent measures such as lockdowns are weighed more heavily in the index. Hence, higher values of the index indicate a state adopted more stringent policies. For a more detailed explanation on how the variable is constructed, see Adeel et al. (2020).

We calculated the daily average PPI for states with both Democrat and Republican governors. The data range from January 24 to April 24 which is the period where state governments were still adopting mitigation policies and not re-opening policies. We plotted these values against time in Figure 1. Democratic governors clearly adopted more stringent responses on average than Republican governors did. However, note that these differences tend to be small.

Figure 1.

Comparison of stringency of state-level policy responses to COVID-19 during the pandemic onset period.

Next, the histograms in Figures 2 and 3 compare the distribution of PPI at the end of the onset period (end of April 2020) and during the re-opening period (end of July 2020). For the April 2020 data, the PPI values range from 0.375 to 0.85 and for the July 2020 data, the values range from 0.1 to 0.825. In addition, while at the end of the onset period Democrat led states had a statistically higher stringent response than Republican led states did, the difference was not large. But in the re-opening period, Democrats clearly had a more stringent response overall.

Figure 2.

Distribution of PPI in US states, end of onset period.

Figure 3.

Distribution of PPI in US states, July 2020.

In the next sections, we discuss how individual Americans were also divided along partisan lines over the proper government response to COVID-19.

Americans’ Opinions of COVID-19 Mitigation Policies

Initially, Americans generally supported their governors’ approaches to mitigating the spread of COVID-19; however, public support for intervention waned over time as the economic side effects of the mitigation policies became apparent (Freking and Fingerhut 2020). Unemployment rose dramatically across the country, reaching its highest level at the national level in April at 14.8%, a huge shift from the low point of 4% 3 months earlier. GDP also declined 33.1% in the second quarter.

Within months, Americans became divided along partisan lines about what the proper response to the outbreak should be (Kushner Gadarian, Goodman, and Pepinsky 2020).[2] Republicans were far less supportive of non-pharmaceutical mitigation measures such as mask wearing (Utych 2020) and staying at home (Roberts and Utych 2020) than Democrats were. This could in part be explained by partisan and ideological considerations. Republicans are more likely than Democrats to favor protecting the economy than protecting public health. If there is a tradeoff, Republicans are more likely to support protecting the economy. Additionally, Republicans generally favor more local approaches to policy problems. Many Republicans felt that only their local governments should be allowed to decide which restrictions, if any, should be adopted to mitigate its spread. However, it is important to note that some Republican governors did not favor local responses if those responses included restrictions. In particular, Republican governors like Brian Kemp of Georgia, Tate Reeves of Mississippi, and Greg Abbott of Texas issued executive orders canceling local-level COVID-19 prevention measures.

Cues from partisan leaders could also explain variations in attitudes towards COVID-19 mitigation measures. The cues and signals from political elite factor heavily in how the public responds to policymaking and actions by government, especially in low information settings like the onset of the COVID-19 pandemic (Converse 1964, Kinder and Kalmoe 2017). President Trump, a Republican, downplayed the virus from the early stages of the pandemic. Republican voters may have trusted his perspective and subsequently began to believe COVID-19 was not a serious threat that governments needed to address. In contrast, prominent Democratic policymakers, such as Governor Gretchen Whitmer (MI), very publicly warned voters that COVID-19 was a serious threat. Democratic voters were more receptive to these cues and began to believe the government was justified in adopting strict mitigation measures to stop the spread of COVID-19 (Gollust, Nagler, and Fowler 2020; Hart, Chinn, and Soroka 2020).

Partisanship and media sorting can also in part explain Americans’ attitudes towards COVID-19 mitigation measures. Even prior to COVID-19, there was a clear difference in coverage of the proper government response to other similar healthcare issues such as H1N1 between conservative outlets such as Fox News and more liberal outlets such as CNN and MSNBC (Baum 2011; Gollust, Nagler, and Fowler 2020). After COVID-19 began to spread in the US, news coverage of the crisis once again began to become sorted between the media outlets. Ultimately, right-leaning outlets devoted far more time to stories that reference misinformation than more moderate and left-leaning outlets. Right-leaning outlets were more likely to promote stories that suggest the virus was a hoax (Motta, Stecula, and Farhart 2020). For example, Sean Hannity told his viewers that the “Deep State” was trying to spread panic and destroy the economy in order to hurt President Trump. Rush Limbaugh said that the Chinese government was trying to hurt the US through the spread of the pandemic. Furthermore, Fox News anchors told viewers that the response to COVID-19 was overblown and that Democrats were just using it as an excuse to impeach the president again (Hart, Chinn, and Soroka 2020).

In the current age of vast media choice, individuals tend to expose themselves to media they are most likely to agree with based on their partisan biases (Mutz 2006; Taber and Lodge 2006). Partisan cues about media are well-known, allowing individuals to easily select into co-partisan news sources (Iyengar and Hahn 2009). In the age of social media, this selective exposure can also be driven by social endorsements—when trusted individuals make claims, people are more likely to follow these claims and seek out that information (Messing and Westwood 2014). This selective exposure has been shown to polarize citizens in their individual level attitudes (Stroud 2010). Given that right-leaning and left-leaning media covered the COVID-19 crisis differently, based often on ideological biases, it is no surprise that COVID-19 attitudes polarized, related to how they view the role of government intervention in the crisis. Importantly, trusted sources within the right-leaning media environment, such as Hannity and Limbaugh, went as far as to perpetuate dangerous conspiracy theories about COVID-19. Given that the media fragmented along these multiple lines, it is entirely likely that selective exposure led to liberals and conservatives viewing the problem of COVID-19 very differently.

These partisan cues have long been shown to be a useful heuristic for citizens—provided politicians, and entire political parties, act in an expected manner, this can guide citizens on how to think about even novel issues (Arceneaux 2008). Citizens notoriously use party labels, rather than actual policy content, to guide how they view issues (Rahn 1993), and even change their policy positions to better align with their party (Lenz 2012). Citizens maintain these motivated beliefs even in the face of information that challenges these partisan attitudes (Taber and Lodge 2006). The COVID-19 pandemic is especially ripe for this type of motivated reasoning, as it is a novel issue where individual attitudes are unlikely to exist, much less by crystallized. Indeed, during the COVID-19 pandemic, we see evidence of the importance of partisan cues, as those in Republican counties are more likely to stay at home when Donald Trump’s Tweets showed he was taking the threat of the virus more seriously (Bisbee and Lee 2020).

This difference in media coverage sent divergent signals to viewers that influenced Americans’ opinions about the virus and the proper government response to it. Since COVID-19 was a novel issue, Americans likely sought media sources they trusted to learn more about it. Since conservatives and liberals tend to consume different media outlets, conservatives and liberals were exposed to different information and different perspectives about the threat the virus actually posed to the public and what the proper government response to the threat was. They also began to reject alternative information and viewpoints from other perspectives. Ultimately, Republicans were far less likely than Democrats to view COVID-19 as a serious threat, were less supportive of government intervention, and were more supportive of local responses to the pandemic (Gollust, Nagler, and Fowler 2020).

In sum, Americans were divided along partisan lines about the proper government response to COVID-19. In the next sections, we explore whether these attitudes influenced American’s evaluations of state governments’ responses to the pandemic.

Americans’ Evaluations of State-Level COVID-19 Mitigation Policies

Given that Democrats tended to favor strong government responses to COVID-19, we should expect that throughout 2020, Democrats should have been more satisfied with their state governments’ response to COVID-19 the stricter the measures became. And given that Republicans tended to favor weaker government responses to COVID-19, we should expect that Republicans should have been more satisfied with their state governments’ response to COVID-19 the less strict the measures became. This leads to Hypotheses 1 and 2:

Hypothesis 1: Democrats became more satisfied with their state governments’ COVID-19 response the stricter the response became.

Hypothesis 2: Republicans became less satisfied with their state governments’ COVID-19 response the stricter the response became.

Additionally, if Americans cared about public health and policy outcomes, we should expect COVID-19 case and deaths rates as well as economic indicators to have influenced Americans’ evaluations of state-level responses. As cases increased, positive evaluations should have decreased. And as unemployment rose, positive evaluations should have decreased. This leads to Hypotheses 3 and 4:

Hypothesis 3: Americans became less satisfied with state-level COVID-19 responses as cases increased.

Hypothesis 4: Americans became less satisfied with state-level COVID-19 responses as unemployment rates increased.

Alternatively, it is possible that COVID-19 policy responses and policy outcomes may have had no effect on Americans’ opinions of COVID-19. The response to COVID-19 was complex and voters may not have been fully aware of all the policies their state governments adopted and how those policies affected outcomes. Prior research on voters’ evaluation of government performance have found that often voters are unaware of the specific policies that leaders adopt and may not fully comprehend the consequences of those policies (Bartels 2002). Hence, given that the COVID-19 response was quite complicated, Americans may have not been fully aware of the specific measures officials in their state adopted and what the consequences were. Instead, they may have used partisan cues as a shortcut in determining what their state government did to address the crisis.

Given that Republican and Democrat officials very publicly disagreed on the appropriate government response to COVID-19, and these disagreements were well documented in different media outlets, it is reasonable to assume that Americans were generally aware that governors who were Democrats adopted stricter measures on average while governors who were Republicans adopted less strict measures on average. Relying on these generalizations, Americans in states with Republican governors may have simply assumed that their Republican governors adopted less strict policies while Americans in states with Democrat governors may have simply assumed their Democrat governors adopted stricter policies. If this is the case, Republicans should have been more satisfied with Republican governors’ responses because they assumed Republican governors adopted less strict policies and that is what Republicans preferred. Additionally, Democrats should have been more satisfied with Democrat governors’ responses because they assumed Democrat governors adopted stricter policies and that is what Democrats preferred.

Of course, it seems likely that, at least on some level, voters would be aware of their own states’ COVID-19 restrictions and how they compared to other states. This is important to consider because some governors’ responses were inconsistent with national trends. That is, some Republican governors, especially those in Democratic leaning states like Maryland and Massachusetts, did enact stringent measures to combat the spread of COVID-19. Yet, while voters may have been somewhat aware of their states’ policy response to COVID-19, partisanship still likely influenced their views of these policies. Previous research on selective evaluation suggests that partisanship can alter people’s perceptions of government performance (Cornelson and Miloucheva 2020; Iyengar et al. 2019; Tilley and Hobolt 2011; VanDusky-Allen and Utych 2021). For instance, Jorgensen et al. (2021) indicates that voters in Western Europe who voted for the governing party reported more positive evaluations of their government’s COVID-19 policy response compared to those who did not vote for the governing party. If Americans evaluated COVID-19 policy responses through a partisan lens, they may have perceived that state-level responses were adequate if their governor was a co-partisan and inadequate if their governor was not a co-partisan. Additionally, Americans may have been more likely to pay attention to positive stories about how their co-partisan governors were handling the crisis and less likely to pay attention to negative stories about it. If their governor was not a co-partisan, they were probably more likely to pay attention to negative stories about how they were doing a poor job at responding to COVID-19.

We also acknowledge that it is possible that voters may have simply supported whichever policies a co-partisan governor adopted and failed to support whichever policies governors of opposing parties adopted. This would be indicative of a partisan cheerleading effect, where individuals are more supportive of the decisions of co-partisans regardless of what the decision is. In the data analysis we employ survey experiments to gain additional causal purchase on whether partisanship or policy influenced voters’ evaluations of state government responses to COVID-19.

This discussion leads to the final hypothesis:

Hypothesis 5: Americans were more satisfied with their state-level government response to COVID-19 if they lived in a state where the governor was a co-partisan than in a state where the governor was not a co-partisan.

In the next two sections, we analyze public opinion data and data from experiments to ascertain which factors influenced Americans’ evaluations of COVID-19 responses.

AP-NORC Center COVID-19 Polls

To begin our analysis, we use data from two AP-NORC Center (2020a; 2020b) polls to ascertain whether state-level COVID-19 policy responses and their consequences influenced Americans’ evaluations of state-level responses. The first poll was taken between 14–18 May 2020. This poll was conducted immediately following the end of the initial onset period of the pandemic where state governments continuously adopted COVID-19 policies without those policies expiring or being repealed. The second poll was taken between 17–19 August 2020. This poll was conducted a few months after state governments began adopting re-opening policies. Note that both polls were conducted through either a computer-assisted telephone interview or a web-based survey. Sample demographics for the observations in our analyses are provided in the Appendix (see Supplementary Table A1).

For the dependent variable in the analysis, Evaluations, we used a question from the surveys that asked respondents the extent to which they approved, disapproved, or neither approved nor disapproved how their state government was handling the coronavirus outbreak. We re-coded the answers to range from 1 to 5 with 1 being strongly disapprove and 5 being strongly approve.

As per our hypotheses, we expect several different variables to influence variations in Evaluations. However, due to the interconnectedness of those variables, we needed to take into account the relationship between those variables in our model design. Figure 4 outlines the effect that we expect those variables to have on Evaluations and each other. Note that in our analyses we utilize path analyses using structural equation modeling in STATA.

Figure 4.

Theoretical expectations and model design.

Our first independent variable of interest is Partisanship. To code this variable, we used data from the AP-NORC polls that asked respondents their party identification. Respondents were coded as either a Democrat or Republican if they identified as so in the survey or if they identified as an independent, they stated they leaned towards one party over the other. Using this variable, we divided the respondents into two groups: Democrats and Republicans.

Our second independent variable of interest is Governor’s Party ID, which is coded 0 for Republicans and 1 for Democrats. As per Hypothesis 5, we expect Democrats and Republicans to give higher evaluations of their state governments’ COVID-19 responses if their governor is a co-partisan. Hence, we expect Governor’s Party ID to have a positive effect on Evaluations for Democrats and a negative effect on Evaluations for Republicans.

Our next independent variable is the state-level PPI at the end of April 2020 (for the May poll models) and the end of July 2020 (for the August poll models) (Shvetsova et al. 2020b). This variable measures the strictness of state-level COVID-19 responses. Higher values indicate stricter government responses while lower levels indicate less strict responses. For Democrats, we expect Governor’s Party ID to have a positive effect on PPI and as per Hypothesis 1, PPI to have a positive effect on Evaluations. For Republicans, we expect Governor’s Party ID to have a positive effect on PPI and as per Hypothesis 2, PPI to have a negative effect on Evaluations.

Next, we included variables that measured total cumulative Cases per capita at the state-level in May 2020 and August 2020, state-level Unemployment in May 2020 and August 2020, and Unemployment Change at the state level from January 2020 to May 2020 and January 2020 to August 2020. We gathered data from the to measure case rates and data from the to measure Unemployment and Unemployment Change. We expect PPI to have a positive effect on the unemployment variables and a negative effect on cases. And as per Hypotheses 3 and 4, we expect that as case rates and unemployment increase, Evaluations decrease.

We present the results of the models from Figure 4 in Tables 1 and 2 and Figures 5 through 8. Note that in all the models, as expected, Governor’s Party ID was positively associated with PPI and PPI was positively associated with Unemployment. But PPI had a positive effect on Cases in the May 2020 models and a negative one in the August 2020 models. For the May 2020 models, we re-ran the models assuming Cases influenced PPI instead and the substantive results of the models were the same.

Table 1.

Evaluations of State Government Responses to COVID-19 May 2020 Poll.

	Model 1 Democrats	Model 2 Republicans	Model 3 Democrats	Model 4 Republicans
Governor’s party ID	0.957*** (0.128)	−0.624*** (0.160)	1.023*** (0.099)	−0.691*** (0.131)
PPI	0.260 (0.353)	−0.436 (0.428)	0.033 (0.339)	−0.320 (0.432)
Cases	35.878*** (11.122)	−8.794 (17.437)	35.253*** (10.279)	−9.492 (18.452)
Unemployment	0.030* (0.016)	−0.033 (0.022)
Unemployment change			0.001* (0.001)	−0.001 (0.001)
Constant	2.348*** (0.291)	4.110*** (0.272)	2.483*** (0.210)	3.892*** (0.223)
N	510	360	510	360

Bootstrap standard errors in parentheses *p < 0.05, **p < 0.01, ***p < 0.001.

Table 2.

Evaluations of State Government Responses to COVID-19 August 2020 Poll.

	Model 5 Democrats	Model 6 Republicans	Model 7 Democrats	Model 8 Republicans
Governor’s Party ID	0.708*** (0.148)	−0.776*** (0.187)	0.983*** (0.141)	−0.729*** (0.157)
PPI	1.046** (0.373)	0.046 (0.440)	0.723* (0.398)	−0.031 (0.401)
Cases	−45.255*** (10.802)	10.416 (13.247)	−41.769*** (11.405)	10.118 (13.156)
Unemployment	0.130*** (0.027)	0.033 (0.036)
Unemployment change			0.273* (0.138)	0.095 (0.141)
Constant	2.012*** (0.243)	2.787*** (0.385)	2.645*** (0.306)	2.944*** (0.313)
N	513	408	513	408

Bootstrap standard errors in parentheses *p < 0.05, **p < 0.01, ***p < 0.001.

Figure 5.

AP-NORC Center Poll, May 2020, Models 1 and 2.

Figure 8.

AP-NORC Center Poll, August 2020, Models 7 and 8.

AP-NORC Center Poll, May 2020, Models 3 and 4.

AP-NORC Center Poll, August 2020, Models 5 and 6.

In the models for Democrats in the May 2020 sample, Governor’s Party ID had the expected positive effect on Evaluations. Democrats in states with Democratic governors evaluated their governors about one unit higher on the Evaluations scale than Democrats in states with Republican governors did. Unexpectedly, in the models for Democrats, Cases and the Unemployment variables had positive effects on Evaluations, and PPI has no statistical impact. Next, for Republicans in the May 2020 sample, Governor’s Party ID had the expected negative effect on Evaluations. Republicans in states with Democratic governors evaluated their governors about 0.6–0.7 units lower on the Evaluations scale than Republicans in states with Republican governors did. None of the other independent variables have a statistical impact. Hence, for the pandemic onset period, these results do not provide support for Hypotheses 1–4 but do provide support for Hypothesis 5. For the most part, during the early months of the COVID-19 pandemic in the US, actual policy outcomes did not drive Americans’ evaluations of state-level COVID-responses, partisanship did.

The results in the models for the August 2020 are substantively different from the results for the May 2020 sample for Democrats. During the re-opening period, Democrats in states with

Democratic governors evaluated their state responses between 0.7 and 1 unit higher on the Evaluations scale than Democrats in states with Republican governors did. But this time they also rated their state responses higher as PPI increased and Cases decreased. These results provide support for Hypotheses 1, 3, and 5 during the re-opening period. Democrats were more satisfied with their state-level responses the more stringent the policy responses became. And Democrats were satisfied with their state-level response the lower the number of cases per capita were. But Democrats were still more satisfied with their state governments’ response if their state governors were Democrats than Republicans.

Why would more cases of COVID-19 cause Democrats to rate their governor more favorably in May 2020, but less favorably in August 2020, all else being equal? There are perhaps a few explanations here—early in the pandemic, stringent measures had perhaps not had enough time to “work,” signaling that more stringent measures were need to combat the disease. Here, governors may essentially be getting a pass on previous COVID-19 prevalence in the earlier days of the pandemic, as the spread may have been viewed by Democrats as something that could not be controlled without stringent measures, which were not in place in very early stages. By August, the pandemic had become a long-standing part of life. An increase in cases at this point may signal that the measures implemented by the governor are not working, meaning that more cases would lead to worse evaluations of the governor, since they suggest struggles to combat the pandemic over a longer time horizon.

In contrast to Democrats, the results from the August 2020 sample were the same for Republicans as they were in the May 2020 sample. During the re-opening period, Republicans in states with Democratic governors evaluated their governors about 0.7–0.8 units lower on the Evaluations scale than Republicans in states with Republican governors did. The other independent variables did not have the expected statistical impact on Evaluations. Hence, for Republicans during the re-opening period, these results do not provide support for Hypotheses 1–4 but do provide support for Hypothesis 5. During the re-opening period, actual policy outcomes did not drive Republicans’ evaluations of state-level COVID-responses, partisanship did.

Robustness Checks

While it appears that during the August 2020 period Democrats considered their own states’ policy responses to COVID-19 when evaluating their state governments’ responses to the pandemic, it is possible that voters also took into consideration how their own state’s policy responses compared to other states. To this end, we ran additional models where we replaced the PPI variable with a variable that captured the numerical difference between a state’s response to the pandemic and the average response of states overall. We also ran additional models where we included a variable that captured the numerical difference between a state’s response and the average response of states with governors of the same party. We provide the results of the models in Supplementary Appendix B.

The results of the models in Supplementary Appendix B are consistent with the original models. In the May 2020 models, relative state-level restrictions did not influence citizens’ evaluations of state-level COVID-19 responses. But in the August 2020 sample, Democrats rated their state government’s response more positively the stricter the restrictions were in comparison to the average state government response. Democrats also rated their state government’s response more positively the stricter the restrictions were in comparison to the average state government response of states with governors of the same party. In contrast, Republicans’ evaluations were still not influenced by state governments’ responses. And in both the May and August models, a state governor’s party continued to have the same expected effect on respondents’ evaluations of government responses to the pandemic.

The results of this analysis provide additional support for our findings. Republicans and Democrats appeared to rely on partisan cues to inform their evaluations of state-level COVID-19 responses in May 2020. But by August, Democrats relied on both policy and partisanship while Republicans continued to rely on partisanship.[3]

Taken together, these results suggest that during pandemic onset period, partisanship, not policy, drove Americans’ evaluations of their state-level responses to COVID-19. While we acknowledge that Americans generally adopted strong policy preferences over the government’s proper role in mitigating the spread of COVID-19 during the onset period, it appears that those preferences did not influence how they evaluated their state governments’ responses to the pandemic early on. It is possible that given the complex nature of the policy response, along with the high level of uncertainty over the virus itself, in the early days of the pandemic, Americans may not have been fully aware of what their state governments’ responses were. Instead, we suggest that they relied on partisan cues to understand what approaches their state governments were taking and evaluated their state responses largely based on that information.

Beyond the onset period, the results suggest that during the re-opening period, Democrats began to evaluate their state-level responses based on policy, although partisanship still played a role in their evaluation. For Republicans, they continued to rely solely on partisanship to make their evaluations. Perhaps it is possible that since Democrats paid closer attention to state-level policy responses because they wanted their state governments to adopt policies to mitigate the spread of COVID-19 more so than Republicans did. If indeed Democrats paid closer attention, they may have been better able to evaluate state-level responses based on policy outcomes than Republicans were. However, it is also possible that Republicans were simply not concerned with policy outcomes during this time and therefore policy outcomes did not influence their evaluations.

It is also possible, of course, that Republicans and Democrats may have different considerations about which factors are most important in support their governor’s actions. The COVID-19 crisis was a multi-pronged crisis—the obvious public health crisis brought on by the pandemic, but also the economic crisis brought on by vast and immediate changes in the economy due to efforts to mitigate the pandemic. It is possible that these economic considerations were more important generally for Republicans than Democrats (though note that we do not find a direct impact of unemployment rate on evaluations for Republicans). Additionally, it is possible that other factors influence policy responses of governors. Governors may, and probably did, base their decisions on COVID-19 mitigation policies on other factors, such as political factors within their state, economic factors, and case prevalence.[4]

While the results of the AP-NORC models illustrate the statistical relationship between partisanship and policy on citizens’ evaluations of state-level COVID-19 policies, given that we use a structural equation model analysis, these results do not prove our causal assumptions about the relationships between these variables. Instead, the results simply suggest that our expectations about the relationships between these are plausible (Bollen and Pearl 2013). In order to further the assess the causal impact that state-level COVID-19 responses and partisanship could have had on Americans’ evaluations of state-level responses to the pandemic, we turn to two survey experiments. In our experiments, we divide respondents into different groups and have them imagine they lived in hypothetical states with Democrat or Republican governors and different levels of COVID-19 responses. This will allow us to determine whether actually knowing the specific policy response would influence Americans’ evaluations of those responses. It also allows us to control for other factors that might influence voters’ evaluations, since factors like economics and case rates are not discussed, essentially holding them “constant” between the experimental conditions.

Survey Experiments

We ran two survey experiments: one in October 2020 with a sample of undergraduate students at a large Western public university in October and one in December 2020 with a nationally diverse sample recruited from Lucid Theorem. Note that the October 2020 survey took place just prior to the election while the December 2020 survey occurred right after. The differences in survey timing will also us to discern whether the election played a role in voters’ evaluations of state-level responses. Sample demographics for each study are provided in the Appendix (see Supplementary Table A2). In each study, participants were randomly assigned to one of four experimental conditions, where we varied Governor’s partisanship (Democratic or Republican) and the level of response to COVID-19 (stringent vs less stringent). This created a standard 2 × 2 experimental design. The full text of the treatment is given below.

Stringent

In response to the spread of COVID-19 in the early months of the pandemic, a Governor of a US state who was a [Republican/Democrat] declared a state of emergency, closed all schools, closed non-essential businesses, limited social gatherings, and only allowed people to leave their homes if it was necessary.

Less-Stringent

In response to the spread of COVID-19 in the early months of the pandemic, a Governor of a US state who was a [Republican/Democrat] closed all schools and closed non-essential businesses, but did not declare a state of emergency, did not limit social gatherings, and did not place limits on whether people could leave their homes.

After reading this treatment text, participants were asked to rate how much they approved of the Governor’s policy decisions, on a 4-point scale [Strongly Disapprove, Disapprove, Approve, Strongly Approve].[5] This creates the dependent variable, Governor Approval, used in all analyses.[6] The first independent variable in our analysis is Stringency. It is coded 1 if a respondent was assigned to the stringent condition, and 0 if assigned to the less-stringent condition. The second independent variable is Governor’s Party ID. It is coded 1 if the governor was a Democrat and 0 if the governor was a Republican.

While these survey experiments are hypothetical, we have reason to believe they provide additional causal evidence to support our arguments. Even though COVID-19 response was correlated with governor (or state legislative) partisanship, we find variation in stringency of response between governors of the same party, so we do not think our scenario of varying stringency and partisanship is, on its face, not believable. While, at the time of these studies, individuals would indeed have observed significant response to COVID-19 in the real-world, random assignment to experimental groups allows us to assume individuals whose responses are not “movable” due to real-world concerns are equally likely to be in each group. By focusing our analyses on evaluations of the hypothetical governor they have read about, we seek to minimize the effects of real-world observed behaviors on our analyses.

We divide our sample into two, one for Democrats and one for Republicans. As per Hypotheses 1 and 2, we expect Stringency to have a positive effect on Governor Approval for Democrats and a negative effect for Republicans. As per Hypothesis 5, we expect Governor’s Party ID to have a positive effect for Democrats and a negative effect for Republicans. We present the results of the OLS models in Table 3 below.

Table 3.

Evaluations of Hypothetical State Government Responses to COVID-19.

	Model 9	Model 10	Model 11	Model 12
	Democrats-undergraduates	Democrats-lucid	Republicans-undergraduates	Republicans-lucid
Stringency	1.169*** (0.119)	0.884*** (0.097)	0.112 (0.155)	0.343** (0.138)
Governor’s party ID	0.144 (0.119)	0.121 (0.097)	−0.395** (0.156)	0.185 (0.136)
Constant	2.269*** (0.103)	2.590*** (0.082)	2.764*** (0.134)	2.517*** (0.123)
N	262	515	225	337
R2	0.2758	0.1417	0.0314	0.0234

Standard errors in parentheses *p < 0.05, **p < 0.01, ***p < 0.001.

In Table 3, we observe even larger effects for Democrats, across the two samples. In the undergraduate sample, Stringency increases Governor Approval by over 1 point on the 4-point scale, and by nearly 0.9 points in the Lucid sample. In these cases, however, Governor’s Party ID has no effect—while its effect is positive, it is small in magnitude, dwarfed by the effect of stringency and does not approach even generous levels of statistical significance.

For Republicans, the story is considerably more muddled, as shown in Table 3. In the student sample, Governor’s Party ID has the expected negative effect on Governor Approval, but Stringency does not (note, however, that the effect of co-partisanship for Republican respondents in this sample is dwarfed by the effect of stringency for Democratic respondents). Unexpectedly in the Lucid sample, we see a statistically distinguishable (though considerably smaller in magnitude) positive effect of Stringency for Republican respondents. Here, Republicans rate a stringent governor about 0.35 points higher than a less-stringent governor, though this effect is considerably smaller than the effect for Democrats.[7]

These results taken together with the public opinion poll results from the previous section suggest that while in pandemic onset period partisanship mostly influenced Democrats’ evaluations of their state-level COVID-19 responses, after the re-opening period partisanship and policy outcomes both influenced Democrats’ evaluations. The results from the survey experiment suggest that in an ideal environment where Democrats know exactly what policies their state governments adopted, Democrats likely rely solely on policy outcomes to make their evaluations of their states’ COVID-19 responses. However, the results from the AP-NORC poll analyses suggest that when faced to make an evaluation on actual policy responses, Democrats rely both on partisanship and policy to make those evaluations.

For Republicans, the survey results taken together with the results from the AP-NORC polls suggest that partisanship mostly drove Republicans’ evaluations of their state governments’ COVID-19 responses in both the pandemic onset period and in the re-opening period. The results from the survey experiment with undergraduate respondents suggest that even when Republicans respondents are presented with specific hypothetical COVID-19 policy outcomes in their states, they still rely on partisanship to evaluate their hypothetical state-level responses. It is important to note, however, that the results from the Lucid sample for Republicans do not align with the results of the other models. The results suggest that neither partisanship nor policy are consistently driving Republicans evaluations of state-level COVID-19 responses. The Lucid sample is the most recent data that we analyzed, and it was in the post-2020 election period. It is possible that Republican evaluations of state responses changed in the post-2020 election era.

Discussion and Conclusions

In this article, we examined factors that influenced Americans’ evaluations of state-level policy responses to COVID-19. The findings of this article provide the literature with a better understanding of how voters evaluate government policies during a crisis period in a complex policy environment. The findings also shed light on how voters evaluated state responses to COVID-19 as they were deciding who to vote for in the 2020 election.

The results of this article suggest that there likely was a key partisan divide in how COVID-19 influenced the 2020 election. Heading into the election, Democrats were much more concerned about politicians taking stringent action to prevent the spread of COVID-19 while Republicans were less concerned with these policy responses to the pandemic. In addition, the results from a Pew Research Center pre-election survey illustrate that while the most important issue to Democrats prior to the 2020 election was health care and the COVID-19 pandemic, Republicans cared primarily about the economy (Pew Research Center 2020). These asymmetric preferences could have impacted how elected officials addressed both the pandemic and their own re-election strategies. Democratic politicians likely had to focus more on mitigating the COVID-19 pandemic while Republican politicians were perhaps less concerned about the pandemic itself and more concerned about the economy and the economic impacts of the pandemic.

Of course, the findings of this article only speak to a brief period in the span of COVID-19 and its effect on policies and public responses to these policies. While we are able to speak well to this period, we cannot speak at all to the post-vaccine availability phase. Since we have seen significant variation in response during the early phase of the pandemic, future work should unpack the relationship between public opinion and gubernatorial response to the later phases of the pandemic.

In future crises with a similar limited information environment to the one experienced during COVID-19, it is possible that Americans may again become divided along pre-existing partisan lines on the proper government response to the crises, especially if it is in the middle of a contentious election year. This could make it harder for government officials to properly address the crisis, especially if it requires individual Americans to take action to address the crisis, such as wearing masks or practicing proper social distancing. In the absence of unified support for government action, some Americans may ignore guidance and regulations from government officials, limiting the government’s ability to address the crisis (Cornelson and Miloucheva 2020). After the COVID-19 pandemic crisis passes, it would be worthwhile for lawmakers to reflect back on their experiences, consider how partisanship, public opinion, and the election influenced their ability to address the crisis, and consider how they can achieve bipartisan support for government responses during future crises.

Click here for additional data file.

Supplemental Material, sj-pdf-1-prq-10.1177_10659129211056374 for Partisanship, Policy, and Americans’ Evaluations of State-Level COVID-19 Policies Prior to the 2020 Election by Julie A. VanDusky-Allen, Stephen M. Utych and Michael Catalano in Political Research Quarterly