LINKING GENDER AND KNOWLEDGE IN SUPPORT FOR GOVERNMENT INTERVENTION

Previous research consistently identifies a gender gap between men and women's scientific knowledge and consequent support for and trust in science (Gauchat Reference Gauchat2012; Miyake et al. Reference Miyake, Kost-Smith, Finkelstein, Pollock, Cohen and Ito2010). Explanations for this gap tend to focus on differences in the educational and cultural expectations of men and women, especially the stereotype that women are worse at science than men (Miyake et al. Reference Miyake, Kost-Smith, Finkelstein, Pollock, Cohen and Ito2010). Research in political science identifies a gender gap in men and women's policy preferences, with many differences persisting over decades (Lizotte Reference Lizotte2020; Shapiro and Mahajan Reference Shapiro and Mahajan1986). These gaps are concentrated, domestically, around crime and punishment, the environment, social welfare, and, most importantly, health care (Cochran and Sanders Reference Cochran and Sanders2009; Schlesinger and Heldman Reference Schlesinger and Heldman2001; Shapiro and Mahajan Reference Shapiro and Mahajan1986). Various explanations offered for this gap between men and women are summarized and tested by Schlesinger and Heldman (Reference Schlesinger and Heldman2001).

The differences identified by Schlesinger and Heldman that are relevant to assessing support for COVID-19 containment policies are likely captured in the different perceptions that women have toward government interventions and their higher average level of compassion for others relative to men. Specifically, these authors find that women support prosocial policies because of their higher level of compassion and perception of higher government effectiveness than men. Other relevant research identifies the linkage between policy preferences and differences in risk perceptions and aversion (e.g., Huddy, Feldman, and Cassese Reference Huddy, Feldman, Cassese, Stritke, Lewandowsky, Denemark, Clare and Morgan2009) and the influence of differing social roles and values between men and women (e.g., Diekman and Schneider Reference Diekman and Schneider2010).Footnote ¹

Finally, recent work directly examines the motivations for women's support for health care provision more generally and the Affordable Care Act (ACA) specifically, finding that women support the ACA more than men in large part because of their higher level of prosocial values (Lizotte Reference Lizotte2020). Furthermore, recent research on COVID-19 is consistent with this compassion and prosocial explanation for gender differences at the elite level. For example, Shay (Reference Shay2020) finds that female health commissioners and administrators were much more likely to adopt stay-at-home orders before their male counterparts.

All of this research supports the idea that the underlying mechanisms that inform policy preferences for women differ significantly than those for men. When women consider their support for COVID-19 containment policies, they are likely drawing on their prosocial considerations (possibly informed by their social roles), a general increased trust in government programs, and a reduction in the risk of harm for themselves and others. However, these considerations should only be considered by women if they know that COVID-19 is dangerous and has the potential to inflict great amounts of harm. Intuitively, the likelihood that any given individual should have this COVID-19 knowledge, we would expect, should be determined by their own level of general scientific knowledge, with greater levels increasing the probability that one understands the true gravity of COVID-19. Thus, we should expect that, all else being equal, women who possess high scientific knowledge should be more likely than men to support COVID-19 containment policies.

PARTISAN VARIATION IN INTERACTIVE EFFECTS

Against the backdrop of a deeply fractured political system, characterized by intense partisan and ideological polarization over a range of issues and demographics (Gibson and Hare Reference Gibson and Hare2016), COVID-19 presents an uncommon instance of a salient issue on which voters have few, if any, existing considerations to guide their policy preferences. Policies addressing COVID-19 are not only politically unfamiliar but also represent highly technical, “hard” issues (Carmines and Stimson Reference Carmines and Stimson1980) involving matters of medical and epidemiological expertise. Generally speaking, it is more difficult for voters to constrain their attitudes on hard issues along existing ideological lines (Pollock, Lilie, and Vittes Reference Pollock, Lilie and Elliot Vittes1993).

Under these conditions, public opinion should be more fluid and receptive to nonpolitical information flows (Zaller Reference Zaller1992). This is especially the case in situations involving anxiety and fear related to disease outbreaks and pandemics, which pushes individuals to be more likely to listen and heed the advice of medical or scientific experts (Albertson and Gadarian Reference Albertson and Gadarian2015). Hence, we expect that scientific literacy will play a larger role in shaping attitudes on COVID-19 policies than on sother scientific-based debates such as climate change (e.g., Kahan et al. Reference Kahan, Peters, Wittlin, Slovic, Ouellette, Braman and Mandel2012). However, these effects should also be conditioned by partisanship, as Republicans have received mixed, confusing, and conflicting cues from Republican elites and other right-leaning figures (Motta, Stecula, and Farhart Reference Motta, Stecula and Farhart2020). In contrast, Democratic messaging has been consistent in its support for scientific experts and the need for significant interventions (Green et al. Reference Green, Edgerton, Naftel, Shoub and Cranmer2020). Indeed, Shay (Reference Shay2020) finds evidence that implementation of stay-at-home policies were more prevalent in states with Democratic governors. Consequently, we expect that scientific knowledge should play a more muted role in shaping Republicans’ policy preferences.

RESEARCH DESIGN

To evaluate our hypotheses, we rely on nationally representative survey data provided by the Pew Research Center's American National Trends Panel Survey. Given potential concerns regarding endogeneity between COVID-19 attitudes and scientific knowledge, we rely on panel wave 32 fielded from January 7 to 21, 2020, to measure our key independent variable of citizen scientific literacy. To measure mass attitudes about COVID-19 policies, we rely on panel wave 64 fielded from March 19 to 24, 2020. In this survey, panelists were asked whether it was necessary for the government to restrict the following collective activities to contain COVID-19:

1. 1. International travel

2. 2. Most businesses (except grocery stores and pharmacies)

3. 3. Large gatherings of more than 10 people

4. 4. Major sporting and entertainment events

5. 5. K–12 schooling

6. 6. Restaurant dining

7. 7. Upcoming state primary elections (i.e., to postpone due to the virus)

These outcome variables were coded 1 if citizens felt it was necessary for the government to restrict such activity and 0 if they felt it was unnecessary. We construct separate composite measures of COVID-19 policy preferences and latent scientific knowledge using item response theory (IRT) models.Footnote ² Lastly, to measure our conditioning variable of gender, we rely on self- reported gender identification.Footnote ³

To test the conditional effects of knowledge and gender on support for intervention policies, we specify a logistic regression model for each of the seven outcome variables. We include an interaction term of knowledge and gender allowing us to estimate how the effect of scientific knowledge varies by gender. We also estimate a regression model evaluating the interactive effect of gender and scientific knowledge on our composite measure of policy preferences.

To provide a baseline comparison of scientific knowledge effects, we specify an additive model omitting the interaction terms. All of our models include the standard control variables of partisanship, ideology, age, education, race, income, and region.Footnote ⁴ To explore variation in our effects of interest, we estimate models by both partisanship and on the full sample, providing a total of 24 models evaluated.