If you don't remember your password, you can reset it by entering your email address and clicking the Reset Password button. You will then receive an email that contains a secure link for resetting your password
If the address matches a valid account an email will be sent to __email__ with instructions for resetting your password
Address correspondence to: C. Ross Hatton, MPA, Department of Health Policy and Management, Johns Hopkins Bloomberg School of Public Health, 624 North Broadway, Baltimore MD 21205.
Department of Health Policy and Management, Johns Hopkins Bloomberg School of Public Health, Baltimore, MarylandSNF Agora Institute of Johns Hopkins University, Baltimore, Maryland
U.S. residents had varying experiences of the COVID-19 pandemic and social safety net policy in 2020. Past research has suggested that partisanship, ideology, racial attitudes, and personal experience may each influence policy attitudes. In this study, we explore whether variation in support for social safety net policy in 2020 is predicted by negative experiences of the pandemic when controlling for racial attitudes, partisanship, and ideology.
Methods
Support for 12 social safety net policies in 2020 was estimated using data from a nationally representative panel survey of U.S. adults conducted in 2020 (n=1,222). Logistic regression was used to examine differences in the predicted probability of supporting a majority of social safety net policies related to health, housing, and employment by partisanship, ideology, racial attitudes, and negative experiences of the pandemic. Analyses were conducted in 2021.
Results
Higher levels of symbolic racism was a consistently strong predictor of lower social safety net policy support across health, housing, and employment policies; as was identifying as either Conservative or Republican. Negative experiences of the pandemic were generally unpredictive of support for the social safety net policy.
Conclusions
Despite the pandemic's consequences as well as the potential for social safety net policy to address these consequences, negative experiences of the pandemic failed to predict policy support, even as racial attitudes, partisanship, and ideology strongly predicted these preferences in 2020. Building public support for social safety net policy requires communication strategies that identify the shared benefits of these policies.
INTRODUCTION
Owing to the coronavirus disease 2019 (COVID-19) pandemic, millions of Americans have lost their jobs or income, experienced the health effects of COVID-19 directly, or fallen behind on rent or housing payments.
Despite the far-reaching consequences of the pandemic, not everyone has felt these shockwaves. For example, racial disparities have been observed in essential worker status, which has contributed in part to observed disparities in COVID-19 exposure, hospitalization, and mortality.
Unemployment is similarly concentrated among specific industries, and low-income residents are more likely to report losing a job, taking a pay cut, or having trouble paying bills.
The federal government has, in response, passed unprecedented spending packages and adopted major regulatory change aimed at preventing and alleviating the social, economic, and health damage wrought by the pandemic. This assistance has involved massive increases to the breadth and benefits of the social safety net, including broadening the eligibility for unemployment benefits, issuing stimulus checks, and prohibiting tenant evictions.
Given the importance of these policies as well as the prevalence and variation of these shocks, in this study we examine whether negative experiences of the pandemic explained the variations in pandemic-related policy support. Past research suggests that personal experience can shape policy attitudes, though its impact is often smaller than one might expect.
The COVID-19 pandemic provides an opportunity to expand upon this research. Specifically, the reach and severity of the pandemic may have created conditions in which negative experiences of the pandemic were highly salient at the same time that federal policymakers were actively considering policies to promote recovery (e.g., investing in affordable housing and requiring employers to provide sick leave).
Exploring the extent to which personal experience predicted policy support during the pandemic can help inform the types of framing likely to be effective in shaping public support.
However, the public's experience of the pandemic and the government's policy responses did not occur in a vacuum. In 2020, the murder of George Floyd sparked widespread protests, and the presidential election was highly politicized around the pandemic response and Black Lives Matter protests.
For social safety net policy in particular, White Americans’ beliefs about racial inequality often strongly predict support —more than economic self-interest and partisanship.
It is therefore necessary to consider these potential explanatory factors alongside negative experiences of the pandemic.
This study uses the nationally representative survey data collected in 2020 to examine whether personal experience (i.e., negative health, housing, and employment experiences), after controlling for partisanship and ideology, explained pandemic-related social safety net policy support. This question was then examined solely among non-Hispanic White respondents, for whom racial resentment was included as an additional potential explanatory variable. Specifically, this study asks whether negative experiences of the pandemic were associated with higher or lower policy support for health, employment, and housing social safety net policies, after controlling for ideology, partisanship, and—for non-Hispanic White Americans—racial resentment.
METHODS
Study Population
The Johns Hopkins COVID-19 Civic Life and Public Health survey was administered 3 times in 2020 to a longitudinal cohort of U.S. adults on April 7–23, July 7–22, and November 11–30. NORC's probability-based AmeriSpeak panel was used to field the survey online to a nationally representative sample of U.S. households. The sampling frame covered >97% of households and was generated using a combination of the U.S. Postal Delivery Service Sequence File and in-person listings.
AmeriSpeak participants received cash or other incentives for survey completion. Recruitment yielded a Wave 1 sample size of 1,468 (70.4% recruitment rate). A total of 1,222 respondents completed all 3 waves (83.2% completion rate). The analysis reflects data gathered from participants who responded in all waves. Appendix Table 1 (available online) compares sample demographics to the general population. Appendix Table 2 (available online) compares Wave 3 demographics to respondents who exited the survey. This study was approved by the Johns Hopkins Bloomberg School of Public Health IRB.
Table 1Predicted Probability of Supporting a Majority of Health, Employment, and Housing Policies
These statistically significant values were compared with the reference categories (first row of each covariate). Survey respondents were asked whether they supported the 12 social safety net policies. Responses of somewhat support and strongly support were coded as supporting. Policies were then categorized as belonging to the health, employment, or housing domain. This exhibit presents the average predicted the probability of supporting a majority of policies in a given domain (e.g., housing) for the primary independent variables after controlling for age, sex, race/ethnicity, the presence of children in the household, income, education, marital status, self-assessed overall health, the presence of comorbidities (e.g., high blood pressure), frequency of news viewership, economic sentiment, COVID-19 county incidence rate, and COVID-19 county case fatality ratio.
These statistically significant values were compared with the reference categories (first row of each covariate). Survey respondents were asked whether they supported 12 social safety net policies. Responses of somewhat support and strongly support were coded as supporting. Policies were then categorized as belonging to the health, employment, or housing domain. This exhibit presents the average predicted probability of White, non-Hispanic respondents supporting a majority of policies in a given domain for each of the primary independent variables, including racial resentment, after controlling for age, sex, race/ethnicity, the presence of children in the household, income, education, marital status, self-assessed overall health, the presence of comorbidities (e.g., high blood pressure), frequency of news viewership, economic sentiment, COVID-19 county incidence rate, and COVID-19 county case fatality ratio.
In Wave 3, survey respondents were asked to rate their support for 12 health, employment, and housing policies on a 5-point Likert scale (Appendix Text 1, available online, provides a list of policy questions and Appendix Table 3, available online, shows how these policies were categorized). Among others, the policies included raising the minimum wage to $15, guaranteeing 2 weeks of paid sick leave, and an extension of the federal eviction moratorium. Responses of somewhat favor and strongly favor were coded as supporting. Responses of neither favor nor oppose, somewhat oppose, and strongly oppose were coded as not supporting. Support for a majority of policies in each domain was the primary outcome of this analysis.
In Waves 1, 2, and 3, respondents were asked about their health, housing, and employment experiences. Health shocks were defined as losing health insurance, having a new physical or mental health issue and being unable to receive treatment, and receiving a COVID-19 diagnosis or knowing a friend or family member who was diagnosed as COVID-19 positive. A housing shock was defined as not being caught up on rent or mortgage payments. An employment shock was defined as being furloughed, laid off, having reduced hours, or having one's job eliminated during the pandemic. Individuals looking for a job or temporarily laid off at the start of the pandemic were also coded as experiencing an employment shock.
Respondents were also asked about their party affiliation and political ideology. For party affiliation, respondents selected Democrat, Republican, Independent, or none as well as the strength of their affiliation (e.g., strong, lean, or no lean). For ideology, respondents were asked whether they identified as liberal, moderate, or conservative. Appendix Text 1 (available online) provides the exact wording of the questions.
Wave 3 measured racial resentment using 4 questions identical to those used in the American National Election Study. Respondents rated their agreement with the following 4 statements on a 5-point Likert-type scale:
1.
Irish, Italians, Jewish, and many other minorities overcame prejudice and worked their way up. Blacks should do the same without any special favors.
2.
Generations of slavery and discrimination have created conditions that make it difficult for Blacks to work their way out of the lower class.
3.
Over the past few years, Blacks have gotten less than they deserve.
4.
It's really a matter of some people not trying hard enough, if Blacks would only try harder they could be just as well off as Whites.
Together, these 4 questions provide a measure of symbolic racism—also known as racial resentment—which has been defined as “a new form of racism… blending racial animus with perceptions that Blacks violate traditional American values.”
In the race and politics literature, these responses are typically summed and scaled 0–1 (1 corresponding the highest racial resentment) to examine the association between racial attitudes and policy support.
The same scale is used in this analysis. Because this measure was initially developed to examine racial resentment among non-Hispanic White Americans, the use of this explanatory variable in this study is similarly limited to a secondary analysis of non-Hispanic White respondents.
Logistic regression was used to examine the association between shocks, partisanship, ideology, racial resentment, and support for a majority of policies in each domain. Controls included age, sex, race and ethnicity, children in the household, income, education, marital status, self-assessed health, comorbidities, frequency of news viewership (generally and Fox News specifically), economic sentiment (i.e., has the economy improved or gotten worse in the past year), COVID-19 county incidence rate (at the midpoint of Wave 3), and COVID-19 county case fatality ratio (at the midpoint of Wave 3). The predicted probability of supporting a majority of policies in each domain was then calculated for each of the primary independent variables. An adjusted Wald test was used to determine whether estimated differences in predicted probability were statistically significant.
Multiple sensitivity analyses were also conducted. First, the prior models were run without economic sentiment, case fatality ratio, and incidence rate ratio as these could have limited the association between shocks and policy support. Each individual policy was also analyzed to identify conflicting results. Finally, a different cut point for the outcome—supporting all policies—was analyzed to test whether the cut point influenced the findings. All analyses were conducted using Stata, version 16, with survey weights applied to produce nationally representative estimates.
In total, 45.3% of the weighted sample reported experiencing ≥1 pandemic-related health, housing, or employment shock. Specifically, 26.3% reported an employment shock, 26.1% reported a health shock, and 7.2% reported a housing shock. A total of 10% reported experiencing both a health and employment shock, and 1.3% reported experiencing all 3 shocks. Receiving a COVID-19 diagnosis or knowing a diagnosed friend or family member was the most common health shock (21.5%), followed by losing health insurance (5.7%) and having an untreated physical (1.1%) or mental health condition (0.3%). Among employment shocks, having one's hours reduced at work was most common (23.3%), followed by being laid off (8.7%), being furloughed (8.4%), and having one's job permanently eliminated (3.0%). A total 4.6% of respondents were unemployed and looking for work at the start of the pandemic whereas 0.9% were on temporary layoff.
Figure 1 presents the share of respondents who supported most policies in each domain, as well as support for individual policies. The share of respondents supporting most policies was >50% in all 3 domains. Employment policy had the highest support, with 64% of respondents supporting a majority. Individual policy support ranged from 72% (increasing spending on construction projects) to 42% (longer and higher unemployment benefits). A total of 59% of the respondents supported a majority of housing policies. Support was highest for extending the Department of Housing and Urban Development mortgage rule (67%) and lowest for increasing spending on housing vouchers (54%). A total of 55% of the sample supported most health policies, ranging from 67% (guaranteeing 2 weeks of paid sick leave) to 46% (having a national health plan like Medicare-for-all).
Figure 1Support for health, employment, and housing social safety net policies.
Note: This exhibit shows the share of respondents who support each of the 12 analyzed policies, as well as support for a majority of policies within the health, employment, and housing domains. Policy support was assessed using a 5-point Likert scale. Responses of somewhat favor and strongly favor were coded as supporting. Respondents of neither favor nor oppose, somewhat oppose, and strongly oppose were coded as not supporting.
CDC, Centers for Disease Control and Prevention; HUD, U.S. Department of Housing and Urban Development.
Table 1 shows the predicted probability of supporting a majority of health, employment, or housing policies as a function of shocks, partisanship, and ideology; full results appear in Appendix Table 4 (available online) for the full sample and Appendix Table 5 (available online) for the non-Hispanic White sample. The one significant association between policy support and shocks was among individuals who experienced health shocks. A total of 62% (95% CI=58, 65) of the individuals who did not experience a health shock were predicted to support most housing policies, compared with 54% (95% CI=48, 60, p<0.05) among those who experienced a health shock.
The association between ideology and policy support was statistically significant across all 3 domains. Moderates and conservatives were less likely to support a majority of policies in each domain, compared with liberals. For example, 74% (95% CI=67, 81) of liberal respondents were predicted to support most health policies, compared with 40% (95% CI=33, 48, p<0.001) of conservatives. A total of 78% (95% CI=7, 86) of liberals were predicted to support a majority of employment policies, whereas 55% (95% CI=47, 62, p<0.01) of conservatives were predicted to support a majority. Finally, 76% (95% CI=69, 84) of liberals were predicted to support most housing policies, compared with 48% (95% CI=40, 55, p<0.01) of conservative respondents. A similar pattern was observed for moderate respondents (Table 1).
Republican respondents also often had lower support for health, employment, and housing policies relative to Democrats; however, these findings were not as large as those seen for ideology, and support for health and employment policies among independents was not significantly different from that of Democrats. In the health domain, 66% (95% CI=61, 71) of Democrats were predicted to support most policies, relative to 42% (95% CI=36, 48, p<0.001) of Republicans. The predicted probability of support for a majority of employment policies was 72% (95% CI=66, 77) among Democrats, relative to 59% (95% CI=53, 66, p<0.05) among the Republicans. Housing policy support was lower for independents (56%, 95% CI=48, 63, p<0.05) and Republicans (53%, 95% CI=46, 59, p<0.01) relative to Democrats (67%, 95% CI=62, 73).
These trends were similar for non-Hispanic White respondents (Table 2). In addition, respondents with high racial resentment had lower support for social safety net policy than those with low racial resentment (Figure 2). Forty percent (95% CI=31, 49) of respondents with the highest racial resentment score were predicted to support most health policies, compared with 68% (95% CI=59, 77, p<0.01) of respondents with the lowest score. A total of 46% (95% CI=36, 56) of those with the highest score were predicted to support most employment policies, compared with 83% (95% CI=77, 90, p<0.001) of respondents with the lowest score. The gap in housing policy support was particularly large: 34% (95% CI=24, 43) of respondents with the highest racial resentment score were predicted to support most policies. Support among individuals with the lowest score was 79% (95% CI=71, 87, p<0.001).
Figure 2Racial resentment versus White, non-Hispanic support for a majority of health, employment, and housing policies.
Note: Survey respondents were asked whether they supported 12 social safety net policies. Responses of somewhat support and strongly support were coded as supporting, and policies were grouped together as health, employment, or housing policies. Respondents were then sorted into tertiles based upon their racial resentment scores (scores ran from 0 to 16, with higher scores indicating greater symbolic racism). This exhibit plots the predicted probability of supporting a majority of health, employment, and housing policies by racial resentment tertile, after controlling for age, sex, race/ethnicity, the presence of children in the household, income, education, marital status, self-assessed overall health, the presence of comorbidities (e.g., high blood pressure), frequency of news viewership, economic sentiment, COVID-19 county incidence rate, and COVID-19 county case fatality ratio.
When economic sentiment, COVID-19 case fatality ratio, and COVID-19 incidence rate ratio were removed as a sensitivity analysis, there were no changes from the previous results, indicating that these controls were not limiting the observed associations between shocks and policy support (Appendix Tables 6 and 7, available online). Changing the outcome variable to reflect support for all policies likewise did not alter the findings substantively (Appendix Tables 8 and 9, available online), suggesting that these findings were not due to the chosen cut point in outcome. Finally, each individual policy was analyzed (Appendix Tables 10‒15, available online). These results were consistent with prior models with 1 exception: respondents with employment shocks were more likely (48%, 95% CI=41, 54) to support enhanced unemployment benefits than respondents without an employment shock (40%, 95% CI=37, 44, p<0.05).
DISCUSSION
Because of the threat of the pandemic and the power of social safety net policy to protect from its effects, it would not have been surprising for negative experiences of the pandemic to explain greater policy support. Instead, these analyses found that racial attitudes, ideology, and partisanship were strongly associated with social safety net attitudes during 2020. With the exception of support for expanded unemployment benefits, these analyses found that negative experiences of COVID-19 were largely unpredictive of policy attitudes, both for specific individual policies and social safety net policy generally.
Why was policy support not predicted by the negative experiences that resulted from the pandemic? One answer could be the continued racialization of social safety net policy, which was observed for healthcare policy during the Obama administration.
The divergence in party messaging about the pandemic and racial protests, as well as the emergence of racial disparities as a defining feature of the pandemic, may have also dwarfed the role of shocks.
Finally, many Americans received the benefits of pandemic policy action (e.g., stimulus checks), which may have tempered support for further policymaking.
These findings have multiple implications for policy and research. First, lawmakers are presently debating policy targeted at promoting recovery from the pandemic and expanding the social safety net. The Build Back Better Act contains several policy options like those examined in this study, including paid family and medical leave and increased funding for affordable housing.
In the context of these results, efforts to frame these policies in terms of personal experience may not resonate with voters.
These findings also reinforce the role of symbolic racism in non-Hispanic White attitudes toward social safety net policy, even during a public health crisis.
Recent work has found that racial attitudes may mediate the effect of racial framing on support for the government's pandemic response. For example, White Americans with negative attitudes toward Black Americans may be less likely to support pandemic restrictions after being informed about racial disparities in the COVID-19 death rate.
How information about race-based health disparities affects policy preferences: evidence from a survey experiment about the COVID-19 pandemic in the United States.
Although framing social safety net policy in racial terms can inform Americans about important disparities, it may lower policy support among non-Hispanic White Americans who possess negative racial attitudes.
Finally, the perceived receipt of benefits from pandemic-related social safety net policy may alter policy attitudes. This policy feedback, previously explored in other contexts, can influence individual policy attitudes depending on the perception of benefits received.
During the pandemic, public policy provided major protections and benefits to Americans (e.g., stimulus checks), which may have suppressed or altered support for additional policymaking. Researchers should work to explore the role pandemic policymaking may have played in shaping Americans’ support for greater policy action.
Limitations
This study has a few limitations. First, the study design is cross-sectional and cannot establish causal effects. Second, some characteristics (e.g., ideology) may also have affected respondents’ willingness to report shocks (Appendix Table 16, available online provides shock demographics), which could bias the findings. Third, the low number of shocks reported by respondents prevented thorough interaction analyses to examine variation in the role of shocks by ideology, partisanship, and racial resentment. Future research should seek to over-recruit individuals who experienced shocks to conduct these additional analyses. Finally, the shock definitions may fail to fully capture negative experiences of the pandemic (e.g., varying severity of COVID-19 symptoms, other types of housing instability). The presence of potentially uncaptured pandemic experiences in addition to variability in the severity of those experiences limit this paper's capacity to isolate the association between shocks and policy support.
CONCLUSIONS
These findings suggest that individual experiences of the pandemic were not relevant to social safety net policy attitudes. Instead, racial resentment, ideology, and partisanship were highly predictive of policy support, underscoring how race and party were dominant cleavages of American politics, even during a pandemic. Framing proposed social safety net policy in terms of self-interest may be ineffective, while framing the pandemic in terms of racial disparities may lead to schisms in public support. Policymakers require communication strategies that highlight how social safety net policy benefits everyone to build broad support for these policies.
ACKNOWLEDGMENTS
The Johns Hopkins Bloomberg School of Public Health and the Johns Hopkins University were not involved in the design and conduct of the study; management, analysis, or interpretation of the data; preparation, review, or approval of the manuscript; or the decision to submit the manuscript for publication.
The Johns Hopkins Bloomberg School of Public Health and the Johns Hopkins University Alliance for a Healthier World's 2020 COVID-19 Launchpad Grant funded data collection. CRH receives tuition and stipend support from the Bloomberg School of Public Health's Centennial Scholarship. The National Institute for Occupational Safety and Health (T42 OH0008428) provides tuition and stipend support to RJT.
None of these agencies were involved in data collection; design and conduct of the study; management, analysis, or interpretation of the data; preparation, review, or approval of the manuscript; or the decision to submit the manuscript for publication.
No financial disclosures were reported by the authors of this paper.
CRediT AUTHOR STATEMENT
C. Ross Hatton: Conceptualization, Formal analysis, Methodology, Visualization, Writing - original draft, Writing - review and editing. Rachel J. Topazian: Conceptualization, Methodology, Writing - review and editing. Colleen L. Barry: Conceptualization, Funding acquisition, Writing - review and editing. Emma E. McGinty: Conceptualization, Funding acquisition, Writing - review and editing. Adam S. Levine: Conceptualization, Funding acquisition, Methodology, Supervision, Writing - review and editing.
Federal eviction moratoriums in response to the COVID-19 pandemic.
Congressional Research Service,
March 30, 2021 (https://crsreports.congress.gov/product/pdf/IN/IN11516#:∼:text=The%20CDC%20eviction%20moratorium%20took,moratorium%20until%20June%2030%2C%202021. Accessed May 16, 2021)
How information about race-based health disparities affects policy preferences: evidence from a survey experiment about the COVID-19 pandemic in the United States.
00096-4&id=gr1.jpg){kind=link}
00096-4&id=gr2.jpg){kind=link}