Data.

SafeGraph is a data company that aggregates anonymized location data from numerous smartphone applications to provide insights about physical places (www.safegraph.com). Using geo-location data obtained from the activities of more than 35 million smartphones, it provides an aggregated dataset (Social Distancing Metrics) to researchers. For each census block on a given day, the Social Distancing Metrics data report the total number of devices, the number of devices exhibiting full-time work behavior (spending more than 6 hours between 8 a.m. and 6 p.m. in one location outside the home), devices exhibiting part-time work behavior (spending more than 3 but less than 6 hours between 8 a.m. and 6 p.m. in one location outside the home), and devices that stay home all day. SafeGraph determines the home of a device by its common nighttime location over a 6-week period. We aggregate this data to the county-day level and merge it with county data on demographics, election results from the 2016 presidential election, daily local COVID-19 cases and deaths, and daily minimum and maximum temperature, precipitation, and wind speed. We exclude Alaska and territories of the U.S. Moreover, following SafeGraph warnings, we drop two dates (2/25/2020 and 1/27/2021) with anomalous data. To enhance privacy, SafeGraph excludes a census block group if it has fewer than five devices observed in a month. We obtain county-level demographic information from [21], 2016 presidential election results from [22], COVID-19 case and death numbers from [23], and weather data on temperature, precipitation, and wind speed from gridMET as maintained by [24].

Methods.

We examine how the shares of devices that left home differ across counties of varying political slant between February 2, 2020, and February 7, 2021. Our main specification is: (1) where s_ct is the share of devices in county c on day t exhibiting a particular behavior. Specifically, we calculate the shares of devices that left home for any reason, for full-time work, for part-time work, or for non-work reasons using the total number of devices in the sample as the denominator and the number of devices exhibiting each of these behaviors as the numerators. SafeGraph reports a sampling bias in favor of detecting devices that are moving (see [25]). S1 Appendix shows the robustness of our conclusions to several different adjustment approaches.

The independent variable of interest, P_c, is the percent voting for Clinton in the 2016 election. We use this vote share as a proxy to capture a county’s political slant. The coefficients of interest, β_w(t), are allowed to vary by week w(t). We also let control variables D_ct to vary weekly in their influence. They include demographic and socioeconomic controls, daily weather controls, and variables to capture differences in COVID-19 risks across counties. Demographic control variables are percent living under the poverty line, the population share of women, the population share of individuals living in poverty, population shares of whites, blacks, Asians and other non-whites, percent of adults with a bachelor’s degree, percent of adults without a high-school degree, the unemployment rate, and the employed share of the county population. Weather controls include daily minimum and maximum temperature, precipitation, and wind speed. Pandemic risk proxies include the logarithm of one plus the number of COVID-19 cases, the logarithm of one plus the number of COVID-19 deaths, the logarithm of population density, population share over the age of 65, and a binary variable indicating whether the county is in a metropolitan area. We include county fixed effects (ζ_c) to control for cross-sectional differences and state-date fixed effects (η_s(c)t) for state-wide changes in social-distancing policies. Because each observation is a share in a county, we weigh observations by the total number of devices in each county. We cluster errors at the county level.

Data.

The survey data comprises 13,334 responses to a nationally representative (in terms of age, ethnicity, gender, income, and geographic regions) survey of the U.S. adult population. We used https://luc.id/theorem/ to target a nationally representative sample. We obtained respondent demographics (age, race/ethnicity, gender, educational level, annual household income, presence of children at home, and political affiliation) from Lucid. Table S2.1 in S2 Appendix confirms that our sample tracks well with the U.S. national benchmarks in age, race/ethnicity, gender, and regions of the U.S., albeit slightly under-sampling whites. The survey was fielded across 5 time periods between April 1, 2020, and February 15, 2021: early April 2020, late April 2020, June 2020, August 2020, and February 2021. We map zip codes to counties and merge them with the external datasets to obtain county-level population density, COVID-19 infection, and death counts. The University of Michigan Institutional Review Board (IRB) reviewed the surveys and determined that they are exempt from ongoing review (HUM00148129, HUM00180582). The IRB has also approved the merge between survey responses and county-level data (HUM00180640).

The survey asked respondents about their adoption of protective behaviors, comfort with engaging in certain activities after stay-at-home mandates were lifted, worries regarding health and economic well-being, and beliefs regarding the impact of the virus. The survey also elicited pre-existing health risks, the role of constraints as a driver of personal choices during the pandemic, news consumption habits, and the amount of attention respondents paid to different sources of information regarding the pandemic. S3 Appendix presents the survey instrument.

Methods.

We first study partisan gaps across time in protective behaviors, comfort with different activities, health worries, and beliefs. We then study how partisan gaps in these measures vary with the individuals’ health risks, consumption of bipartisan news, and level of attention to information about the pandemic.

To document partisan gaps over time, we use the following regression equation: (2) where Y_iτ is an outcome variable of interest for person i at time τ. There are four groups of outcome variables, related to, respectively, protective behaviors, comfort with different activities, worries, and beliefs. First, to elicit protective behaviors, the survey asked “Which of the following changes have you personally made to protect yourself from the coronavirus infection?” for a set of behaviors including: (1) do not meet friends or extended family; (2) avoid large gatherings and public transportation; (3) wash hands more often; and (4) wear a mask when out and about. The first two are social distancing actions that may contribute to the mobility differences we also examine with observational data, and the latter two are health precautions individuals were encouraged to take.

Second, to elicit comfort with activities, from June 2020 on, as most of the stay-at-home orders were lifted across the nation, the survey asked respondents whether or not they “feel comfortable in engaging in” the following activities: (1) eat in a restaurant with indoor seating; (2) eat in a restaurant with outdoor seating; (3) be part of a gathering with more than 10 people; (4) go to a coffee shop; (5) go to a bar; (6) go to a gym; (7) go grocery shopping; and (8) go shopping for non-food items. For both protective behaviors and comfort with activities, the dependent variables are coded as indicators (0/1) on each of the items.

Third, the survey asked how worried respondents felt about their own health, and the health of their partner, kids, extended family, members of their community, and the whole U.S. (on a scale of 1 to 5, 1 being not worried at all, 5 being extremely worried). It also asked how worried they feel about the economic well-being of the same groups of people (using the same scale).

Fourth, the survey elicited two types of beliefs: (1) Individuals’ predictions on their own health risks, i.e., the chance of becoming infected in the next three months and the chance of having no/mild or serious symptoms should they get infected. (2) Their predictions of the number of U.S. deaths by a certain target date assuming the state policies remain the same. The target date was July 1, 2020, in the April waves, September 1, 2020, in the June waves, December 1, 2020, in the August wave, and May 1, 2021, in the February wave. For worries and beliefs, we use standardized (z-score) measures as dependent variables.

As for the independent variables, P_i is a vector of indicator variables representing person i’s political affiliation (Democrat, Republican, or Independent). We allow the impact of partisanship, captured by α_τ, to vary over time. We determine respondents’ political affiliation as Democrats, Republicans, and Independents based on their responses to the political affiliation question asked by Lucid. We indicate an individual as a Democrat (Republican) if they choose one of the following responses to the political affiliation question: “Strong Democrat (Republican),” “Not very strong Democrat (Republican),” “Independent Democrat (Republican)” or “Other—leaning Democrat (Republican).” We indicate an individual as Independent if they choose one of the following responses to the political affiliation question: “Independent—neither” or “Other—neither.” Alternative specifications of this variable do not substantively change our results.

The control variables X_iτ include individual-specific demographic controls (categorical variables for the respondent’s gender, race, age, educational level, annual household income, and presence of children) and an indicator for whether the respondent has any of the chronic conditions listed by the CDC as a high-risk factor. Control variables X_iτ also include geography-specific covariates that proxy for local pandemic severity over time (natural logarithm of one plus the number of cumulative COVID-19 cases in the respondent’s county at the time of the survey, natural logarithm of one plus the number of cumulative deaths in the respondent’s county at the time of the survey, natural logarithm of the population density of the respondent’s zip code). Summary statistics of control variables are in Table S2.2 of the S2 Appendix. We include state-time fixed effects, μ_s(i)τ, to capture systematic differences across states at the time.

In addition to studying the partisan gaps over time using Eq (2), we also examine whether these gaps and their persistence are heterogeneous across individuals with different health risks, bipartisan news consumption, and attention to information about the pandemic using Eq (3) below: (3) where Y_iτ (the outcome variable of interest), P_i (the political affiliation dummy variables), the control variables X_iτ and μ_s(i)τ (the state-period fixed effects) are the same as those in Eq (2). Z_i is a binary variable capturing individual heterogeneity of interest. In this specification, the coefficient α_τ gives the partisan gaps among individuals for whom Z_i = 0, and the sum α_τ + γ_τ measures that for individuals for whom Z_i = 1.

We examine three dimensions of heterogeneity. First, we study differences between individuals with high or low health risks. We consider an individual to be at a higher health risk (Z_i = 1) if they are either 65 years or older or have at least one pre-existing health condition that qualifies them as at-risk by the CDC, and at lower risk otherwise (Z_i = 0). The survey asked respondents to indicate any conditions they have from a list of diseases obtained from the CDC website on March 30, 2020: Moderate to severe asthma; COPD or other chronic lung diseases; Serious heart conditions; Diabetes; Conditions that can cause a person to be immuno-compromised including cancer treatment, smoking, bone marrow or organ transplantation, immune deficiencies, poorly controlled HIV or AIDS, and prolonged use of corticosteroids and other immune weakening medications; Severe obesity (BMI of 40 or higher); Chronic kidney disease and currently undergoing dialysis; Liver disease. 57% of our respondents are considered high-risk. Second, we examine differences between individuals who consume news across vs. along partisan lines. The survey asked individuals to indicate the top news outlets they regularly consume. We define an individual as consuming bipartisan news (Z_i = 1) if the top news sources they regularly report consuming span the liberal-conservative divide, or as consuming news along partisan lines (Z_i = 0) otherwise. While there are partisan differences in choices of news outlets, a considerable fraction of people consume news across the partisan line: 46% of Democrats indicate Fox News as a source they regularly watch, and 42% of Republicans consume a more liberal news outlet in addition to Fox News. We consider CNN, NBC/MSNBC, NPR, Huff Post, The New York Times and Washington Post as liberal news outlets, but the results are robust to narrower classifications. Third, we examine differences between individuals who pay more vs. less attention to information about the pandemic. The survey asked participants to indicate the amount of attention they pay to various information sources (on a 5-point scale ranging from 1 “not at all” to 5 “very much so.”) spanning from scientists/researchers to friends and family members regarding the pandemic specifically. A respondent is said to pay more attention to information about the pandemic (Z_i = 1) if their average attention score across different information sources is above 3.5 on a 5-point scale, and as paying less attention (Z_i = 0) otherwise. Again, while there are partisan differences in information attention, there are within-Democrats and within-Republican variations in the amount of overall attention they pay to others in terms of the pandemic, calculated as the average attention score across all items. Summary statistics of news consumption and attention to different sources broken down by political affiliation can be found in Table S2.3 in S2 Appendix.

Persistence in partisan gaps.

Fig 2 plots the Democratic—Republican partisan gaps obtained from the estimates of α_τ’s in regression Eq (2). Differences between Independents and Republicans are more muted overall, and are reported in Fig S2.2 in S2 Appendix.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Partisan gaps in actions, attitudes, worries, beliefs.

Notes: This figure plots the estimated Democrat—Republican partisan gaps obtained from the estimates of α_τ in Eq (2) and the corresponding 95% confidence intervals. The x-axis indicates the period τ. In Panel A, a positive estimate means that, ceteris paribus, Democratic respondents are more likely than Republican respondents to have taken an action indicated in the legend. The actions studied are “Wash Hands”–wash hands more often; “Wear Mask”– wear a mask when out and about; “Not See Friends”– do not meet any friends or extended family; “Avoid Gatherings”– avoid public transportation and large gatherings. In Panel B, a positive estimate means that, ceteris paribus, Democratic respondents are more likely than Republican respondents to feel comfortable with an activity indicated in the legend. Activities studied are “Restaurant, In”– eat in a restaurant with indoor seating; “Restaurant, Out”– eat in a restaurant with outdoor seating; “10 + ppl”– be part of a gathering with more than 10 people; “Cafe”– go to a coffee shop; “Bar”– go to a bar; “Gym”– go to a gym; “Shopping”– go shopping for non-grocery items; “Grocery”– go grocery shopping. In Panel C, a positive estimate means that, ceteris paribus, Democratic respondents worry more about the health well-being of the group of people indicated in the legend. The groups of people are “Self”– respondent herself; “Partner”– respondent’s partner; “Children”– respondent’s kids; “Ext. Family”– respondent’s extended family; “Comm.”– members of the respondent’s community; “U.S.”– all people in the U.S. In Panel D, a positive estimate means that, ceteris paribus, Democratic respondents predict a larger number on the outcomes indicated in the legend. Outcomes over which expectations are elicited are “U.S. Deaths”– total number of deaths in the U.S. by a target date; “Chance of Infection”– chances that the respondent will get infected with the coronavirus in the next three months; “Chance of Serious Illness”– chances that the respondent will have serious symptoms should she get infected.

https://doi.org/10.1371/journal.pone.0287018.g002

Panel A depicts the estimated differences between Democrats and Republicans in the probability of taking protective actions. From this panel, we can see that although partisan gaps in social distancing and washing hands are negligible in the early days of the pandemic, they quickly grow to a substantial level by the end of April 2020 and remain high even by February 2021. The absence of partisan gaps in social distancing early in the pandemic corresponds with stay-at-home mandates many states had in place. As these restrictions are lifted, partisan gaps emerge. Regarding mask-wearing, a small partisan gap is present early in the pandemic, grows larger during the summer of 2020, and is diminished greatly by February 2021 when most businesses and local governments mandate masks indoors.

Panel B of Fig 2 presents partisan differences in comfort with various economic activities. Democrats are less likely to feel comfortable with engaging in economic activities overall. Except in the case of grocery shopping, the partisan gap is consistently between 10% and 20% across these activities, and stably so over time. These partisan gaps are substantial in magnitude. For example, over time, the overall share of respondents reporting feeling comfortable with dining indoors increased from 29% in June 2020 to above 42% in February 2021 (see Fig S2.1 in S2 Appendix). A partisan gap of 20% is, therefore, equivalent to about half of the overall change in comfort with indoor dining over time.

As Panel C of Fig 2 shows, from the beginning of the pandemic to the last period of our survey, Democrats report being more worried than Republicans about their own health that of their partner’s, local community’s, and the whole nation’s. The partisan gap in health worries varies between 20–40% of a standard deviation, and increases over time. Fig S2.3 in S2 Appendix depicts the estimated partisan gaps for economic worries, which are statistically significant and in the same direction, but more muted in magnitude.

Finally, Panel D of Fig 2 presents partisan gaps in predictions regarding own or systemic health risks. From the figure, we can see that while the magnitude of the gaps varies over time, the gaps are almost always present: Democrats are consistently more pessimistic than Republicans. For example, despite being more likely to take precautionary actions, Democrats’ prediction of their chance of becoming infected in the next three months from the time of the survey is, on average, 0.1 to 0.2 standard deviations higher than that of Republicans. Their predictions about their chance of getting serious symptoms conditional on getting infected are also 0.1 to 0.2 standard deviations higher. In terms of systemic risks, Democrats on average predict more deaths due to COVID-19 than Republicans by around 0.1 standard deviations.

Overall, the above results show that throughout the 10-month period, Democrats were consistently more worried about the health impact of the pandemic, more cautious about engaging in economic activities, and congruently, more likely to take social distancing and preventive actions. One might be concerned that partisan differences in health behaviors existed prior to the pandemic. Note that questions regarding social distancing, seeing friends and comfort with different activities are specific to the pandemic itself. Therefore, this concern would center around potential pre-existing partisan gaps in hand-washing or in health worries. According to our results, partisan gaps in hand-washing and health worries were small in early April but increased over time, implying that pre-existing partisan gaps, if there were any, cannot account for the results entirely.

Heterogeneity in partisan gaps.

Fig 3 plots the estimated Democrat—Republican partisan gap for low-risk individuals and high-risk individuals separately, i.e., the estimated α_τ and α_τ + γ_τ in Eq (3) where Z_i is 1 for individuals with high risks and 0 otherwise. The estimates reveal two patterns: first, across all variables of interests (precautionary actions, attitude towards economic activities, health worries, and beliefs), the partisan gap among high-risk individuals is often statistically the same as (and sometimes smaller than) the partisan gap among low-risk individuals. Second, partisan gaps among high-risk individuals are substantial and stable over time. For example, high-risk Democrats are more likely to avoid seeing friends than high-risk Republicans by 10 percentage points (Panel A) and less likely to feel comfortable with a gathering with more than ten people by more than 10 percentage points (Panel B) throughout the sample. They are also more worried (Panel C) and hold more pessimistic beliefs (Panel D) than their Republican counterparts. These results are alarming as they indicate that as information about the disease becomes more accurate and widespread, partisan gaps remained strong even for the most vulnerable individuals.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Partisan gap heterogeneity across high-risk and low-risk individuals.

Notes: This figure plots the estimated Democrat—Republican partisan gaps for low-risk and high-risk respondents. In each panel, a dashed line and the marker on it give the confidence interval and the estimate for α_τ in Eq (3), which is the partisan gap among low-risk respondents, and a solid line and the marker on it give the confidence interval and the estimate for α_τ + γ_τ, which is the partisan gap among high-risk respondents. The x-axis indicates the period τ. The legends are the same as those in Fig 2.

https://doi.org/10.1371/journal.pone.0287018.g003

Fig 4 plots the estimated Democrat—Republican partisan gap for individuals who consume news across the partisan line versus those who consume a politically narrow scope of news, i.e, the estimated α_τ and α_τ + γ_τ in Eq (3) where Z_i is 1 for individuals consuming news across the partisan line and 0 otherwise. We find that throughout the period we study, partisan gaps in precautions taken, comfort with economic activities, health worries, and beliefs among individuals who choose to consume news across the partisan line are small in magnitude and oftentimes insignificant. In contrast, among individuals who do not consume news across the partisan line, while partisan gaps in actions and health worries quickly increase and become the main driver of partisan gaps. Moreover, partisan gaps in comfort with economic activities and beliefs are also large and significant in this group.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. Partisan gap heterogeneity across those who consume news across the political line vs. those who do not.

Notes: This figure plots the estimated Democrat—Republican partisan gaps for respondents who consume news across the political line and for those who do not. In each panel, a dashed line and the marker on it give the confidence interval and the estimate for α_τ in Eq (3), which is the partisan gap among consumers of narrow news, and a solid line and the marker on it give the confidence interval and the estimate for α_τ + γ_τ, which is the partisan gap among consumers of news across the political line. The x-axis indicates the period τ. The legends are the same as those in Fig 2.

https://doi.org/10.1371/journal.pone.0287018.g004

Fig 5 depicts the estimated Democrat—Republican partisan gap for individuals who pay attention to information about the pandemic versus those who do not. The results show that among individuals who pay a high level of attention to information from others regarding the pandemic (average attention score greater than 3.5 on a scale of 1 to 5), the partisan gaps in precautions taken, comfort with economic activities, health worries and beliefs are economically small and oftentimes insignificant. In contrast, individuals who report not paying much attention overall to others as a source of information about the pandemic show substantial partisan gaps in all metrics.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 5. Partisan gap heterogeneity across those who pay attention to other information sources vs. those who do not.

Notes: This figure plots the estimated Democrat—Republican partisan gaps for respondents whose average attention score on a 1–5 scale across other information sources (Friends, Family, Scientists, Pastor, Facebook or Twitter, CDC, Governor, President) is greater than 3.5, and those whose average attention score is 3.5 or lower. In each panel, a dashed line and the marker on it give the confidence interval and the estimate for α_τ in Eq (3), which is the partisan gap among those who do not pay a lot of attention to other information sources, and a solid line and the marker on it give the confidence interval and the estimate for α_τ + γ_τ, which is the partisan gap among those who pay high attention overall. The x-axis indicates the period τ. The legends are the same as those in Fig 2.

https://doi.org/10.1371/journal.pone.0287018.g005