Care to trust? Gender and trust in leaders during the Coronavirus pandemic

ABSTRACT
 There has been praise of how female leaders have handled the Coronavirus pandemic relative to their male counterparts by presenting a more “caring” leadership. Of similar coverage has been the role of public trust for how successful governments have been in containing outbreaks. In this paper, we build on these two literatures to understand different determinants of trust during the pandemic between men and women. Following social role theory, we argue that female citizens’ trust judgements are more likely to be driven by the perception that leaders are more caring than are men, whilst men's judgements are more likely to be driven by competence judgements than women's. We test this argument using original survey data from three countries. We find that this relationship holds in the United States, but not the United Kingdom or Italy. This adds to variation in gender gaps in the USA and Europe; at the same time, it also suggests that the propensity for women to be less trusting than men is not down to (perceived) leadership traits.


Introduction
In the current Coronavirus pandemic and in previous epidemics, public trust is seen as fundamental in determining public and government responses. Typically, this is seen as a positive relationship: as trust increases, so does the likelihood of compliance with the required but strict measures (van Bavel 2020; Blair, Karim, and Morse 2017;Siegrist and Zingg 2014), though there is also evidence of the opposite effect, with high trust leading to complacency and low trust related to greater risk perception (Dryhurst et al. 2020;Wong and Jensen 2020). In either case, understanding what drives trust in government and in leaders is important for understanding government responses in the early days of the pandemic and what government action may look like as the pandemic unfolds.
The existing political science literature on determinants of political trust is typically interested in overall levels, or the average effect of some variable on trust. However, this likely masks significant differences between demographic groups, which are only crudely uncovered by using demographic groups as control variables. In many, if not most, instances there are theoretically driven reasons we would expect different determinants of trust depending on individual characteristics. If trust is central to public and government responses to COVID-19 our urgency to understand it is even more warranted.
In this article, we develop an argument that the determinants of trust in leaders during the pandemic are likely to be heterogeneous between genders. In the early stages of the pandemic, female leaders were praised for their better handling of the crisis owing to their more empathetic, relatable and personable communications alongside quick and unambiguous action (Hassan and O'Grady 2020), though arguably their objectively stronger performance is confounded by state capacity and wealth (Piscopo 2020). Drawing on social role theory, we argue that trust in leaders may follow the same gendered patterns, in which the perceived "caring" of leaders has a stronger effect for women than it does men. Using original data from three democracies, we show that whilst women prize caring more than men and that the perceived level of caring increases trust on average, the effect on trust is only greater for women in the United States.
These results contribute to three literatures. First, we contribute to the political trust literature by adding to our understanding of political trust and the COVID-19 pandemic (Devine et al. 2020) and by developing the argument that the determinants of trust (and, by extension, its consequences) may be different between demographic groups. Second, we contribute to the gender literature by advancing our understandings of the gendered dynamics of leadership during the pandemic, and by adding to the established literature on gender gaps in political attitudes.
Voters' judgements on a leader's traits mediate support for leaders (Graefe 2013;Funk 1997). Furthermore, gender stereotypes can mediate this relationship. Whilst men in leadership contexts may be judged primarily on, and presumed to possess more, competence, female politicians seeking or in leadership positions face a finer balancing act between demonstrating competence and warmth and may find their stereotyped association with compassion is incongruent with desired "masculine" leadership traits such as strength and decisiveness (Eagly and Karau, 2002;Huddy and Terkildsen 1993;Jamieson 1995). The prevalence and application of these gendered stereotypes vary by context and voter reliance on gender cues is contested by some (Bauer 2015). One of the contexts in which differences in desired leadership traits varies is times of crises such as the Coronavirus pandemic.
At the same time, the coronavirus is primarily a health crisis, though one with widespread consequences, and health is a policy area traditionally associated as "feminine" (Huddy and Terkildsen 1993); it is a crisis that is being played out in the domestic, private sphere, traditionally associated with women. Unlike previous crises in which "male" leadership traits are seen as preferable, the current context faced, particularly in Western countries that have little experience of pandemics, may mean citizens express preference for more "feminine" leadership styles or traits. These are traits such as compassion and communality (Huddy and Terkildsen 1993), and indeed, this would be supported by the positive early coverage of female leaders.
However, we may expect that these preferences, or at least the strength of them, are also dependent on voter characteristics. There is ample evidence for a series of "gender gaps"often measured as a binary definition of sex in political attitudes and voting, although these vary across time and countries (Giger 2009;Shorrocks 2018). On specific issue positions, women have been found to support what are categorised as more communal issue positions (Schneider and Bos 2019). For instance, they are less supportive of force and torture than men, and more supportive of welfare policies and government spending (Campbell 2012).
The causal mechanisms behind this are debatable. In part, it is driven by macro-level changes in parties, public policy and socioeconomic factors (Iversen and Rosenbluth 2006;Giger 2009). However, we follow Schneider and Bos (2019) and contend that the socialisation of gender may assist in explaining these differences. Social role theory posits that differences in the gendered division in social roles held by men and women have individuallevel effects on political attitudes. Women have traditionally, and continue to, occupy more communal roles in society, and men agentic roles. As a result, most women develop more communal traits (e.g. caring and empathic) and men agentic traits (e.g. aggressive, ambitious). Socialisation reinforces these traits as citizens are expected to behave in certain ways dependent on their sex. These diffuse gender roles explain, at an individual-level, differences in political attitudes: women are more likely to have a diffuse gender identity with communal personality traits leading to more communal political opinions (Schneider and Bos 2019). From this, we contend that women will value these traits more than men when evaluating leaders.

Trust and gender
Despite these expected differences, the COVID-related political trust literature does not differentiate between demographics, nor gender specifically. This is also despite the majority of existing work which argues that there is a gender gap in political trust in Western democracies (though this is by no means uncontested, e.g. Schwindt-Bayer and Mishler 2005). Whilst there is some work on what determines a gender gap in trust where one exists, most focuses on the importance of descriptive representationfor instance, seeing more women in parliament or as leaders (Cowley 2014;Hinojosa, Fridkin, and Kittilson 2017).
Instead, we build on the previous section and argue that women's levels of trust may be more driven by the perception of a caring leadership than men's, and less driven by perceptions of competence than men's. We draw this from the above argument that men develop more agentic traits whilst women develop more communal traits. This does not mean, of course, that women do not also value competence, only that men's trust judgements are driven to a greater degree by competence perceptions than are women's, and women's trust judgements are driven to a greater degree by perceptions of leader "caringness" than are men's. Overall, our argument is only that the socialisation of men and women into (differing) gendered roles may lead women to value communal and caring behaviour. 1 Given our cross-national research design, described next, this also begs the question of whether we would expect similar relationships across countries. Indeed, given different ideological distributions of the population, healthcare systems, and responses to the pandemic, to name just a few, it would not be unreasonable to posit alternative relationships. Yet our theoretical argument suggests that we would expect a similar relationship between trust and sex: our three countries are similar in cultural background, socialisation of genders, and a dominance of male leaders. Whilst we might expect variation indeed, we show there isin trust on average, we expect that the gendered differences in trust and leadership are broadly consistent across similar countries.
To summarise, we posit the following testable hypotheses: H1: Perceiving the leader as caring is associated with higher trust in leaders H2: The effects of caring on trust are greater for women than men H3: The effects of competence on trust are greater for men than women.

Data and cases
To test these hypotheses, we use original nationally-representative survey data from three countries: Italy, the United Kingdom and the United States.
Fieldwork took place over four days from 19-23 May 2020 by Ipsos Mori. In Italy, a sample of 1314 adults were surveyed (51.3% of those were women), in the UK this was 1167 (52.5% women) and the US sample contains 1150 adults (48.3% women). Sampling quotas were placed on gender, age, region and working status. Italy was the first to experience the pandemic, and in response Prime Minister Conte enforced a swift and strict lockdown; his handling of this has meant being described as "authoritative" by national media, and also being praised for his "mediation skills". The country experienced high COVID mortality rates, but many believe this was eased effectively and Conte's approval rating was up to 71% in March, and still at 59% around the time of fieldwork in May. The UK, however, has experienced a slower and less coherent response to the pandemic from Prime Minister Boris Johnson, accompanied by less public praise: during the period of fieldwork in May, 59% of respondents to a poll said they agreed "that the government's response had been confused and inconsistent" (Ipsos Mori 2020). There have been briefer and more inconsistent lockdowns in the USA, alongside the highest mortality rates for COVID-19, which increased significantly in the weeks before survey fieldwork. The mainstream critique of Trump is that he showed "clear willingness to trade lives for the Dow Jones" (Wilson, quoted in The Guardian, 5 May 2020). Yet, before the time of fieldwork, President Trump's approval rating was 49%, up around 5% from January 2020 (Gallup 2020). Our three country cases, therefore, represent male leaders with high, moderate and low approval ratings, which represent perceived successes and failures in overall leadership and during the pandemic.

Measurements and methodology
As is common in gender gap literature we use binary sex (male and female) as a proxy for gender. 2 Our main variable of interest is perceptions of the leaders' caring traits in response to the pandemic. This is measured using a battery of four questions that ask about each country's leaders' approach to the crisis, which we call "agree caring". The questions list caring behaviours, such as listening to experts, their own party and opposition, and whether the leader "cares about people like me". Thus they ask the respondent to make an assessment of how communal the leader has been during the crisis. Responses are a fivepoint likert scale and we have taken the number of questions a respondent answered "agree" or "strongly agree" to these questions, to create an ordinal variable from 0-4. The more questions agreed to, the more caring and communal they think the leader has been. Italy has the most respondents that agree their leader is caring: Conte has the fewest who do not agree to any of the caring questions, indeed 73% of Italian respondents agree to at least one question, and there are few gender differences in the measure. Trump has the least agreement that his response has been caring: 48% of Americans did not answer agree to any questions, and women much less frequently answered agree to three or four questions. The UK's assessment of Johnson falls in between. A third of British respondents did not perceive him as caring, and slightly more of these people identify as female (Figure 1).
We also create a binary variable for whether the respondent prizes caring personally through a battery of five questions that ask the likelihood of performing care tasks as a result of the pandemic (such as shopping for a neighbour); we take anyone who gave the maximum likelihood for any of these five questions as someone who prizes caring. These people make up 30% of the full sample, and nearly three fifths of these are women. The gender distribution of this measure varies by country, as shown in Figure 2. Social role theory would predict that women disproportionately prize caring over men, due to being socialised to have more communal traits. This holds in Italy and the UK, where it is shown that female respondents are more likely to prize caring by 10 points; in the USA, however, men are slightly more likely.
Our final explanatory variable of interest is perceived competence of the leader. To measure this, we use a five-point likert responses to the question "[LEADER] is handling the coronavirus situation well" where 1 is "strongly disagree". This question serves as an assessment of one aspect of agentic traits in the leader and is depicted in Figure 3. American women are less likely to agree that Trump is competent, yet otherwise there are no substantive gender differences. Our dependent variable is trust in the country's leader, which is measured using responses to the question 'Using a 10-

Results
First, we run the models with gender as a control variable, pooling respondents in all three countries and then separated by country. The results are presented in Figure 4. Overall, they show that perceiving the leader to be  caring is associated with higher levels of trust in all countries, providing evidence for the first hypothesis (that caring leadership is associated with higher trust). Indeed, since the variable is scaled between 0 and 4, the effect size is large: moving from the lowest to the highest would lead to an increase of 2.5 on the trust scale, a quarter of the entire scale of 0-10. Prizing caring, interestingly, is associated with lower trust, but not statistically significantly so, and given that it is binary, the full effect is quite trivial. Being female is associated with lower trust in all countries, but only significantly in the United States (and the pooled sample). Competence has a large effect in all countries, but 50% larger in the United States compared to Italy and the UK; given competence is measured on a scale of 1-5, the effect size is extremely large. Education is not associated with trust in the United States, negatively in the UK, and positively in Italy. Age is associated positively in the USA and UK but negatively in Italy. The picture is one of similar coefficients, but non-trivial variation between countries, particularly on the size of competence perceptions.
Second, we split the sample by gender, running the models separately for male and female respondents to test H2 (caring perceptions have a stronger effect on trust for women than men) and H3 (competence perceptions have a stronger effect on trust for men than women). The results of this analysis are presented in Figure 5. Perceiving that the leaders are caring has a positive effect on trust regardless of gender or country, and with approximately similar effect sizes across countries. Regarding H2, there is some support that perceived caring has a greater effect on trust for women. In the United States, the effect size is approximately twice as large for women than men, and the coefficients are statistically different. In Italy, the effect size is larger, but not statistically different, whilst in the UK the effect size is larger for men than women (though, again, not statistically different). Whether individuals prize caring is only significant for women in the United States. 3 Similar results are found for H3. In the UK and Italy, there are no observable gender differences in the effect of competence, but in the USA we again find that competence has a stronger effect on trust for men than women.
Overall, our results provide support for a gender gap in determinants of trust, at least as driven by leadership traits, in the US context, but not in the UK or Italy. To test our confidence in the results, we provide the following robustness tests. First, we interact "prize caring" and "agree caring", essentially asking whether respondents who prize caring more are more greatly affected by judgements of caring leadership, which may explain the (lack of) differential effects between genders and the negative effect in the United States. The results do not alter our conclusions. Second, we also control for the level of threat the respondent believes the Coronavirus poses to the country, them personally, or their finances. Whilst threat perceptions do determine trust, they do not impact our results. Finally, we control for whether the respondent aligns with the incumbent leaders' party. In this model, we see a uniform reduction in the effect of competence across genders. There is also a smaller coefficient for agree caring, but only for women, and being Republican has a greater effect on trusting Trump for women than it does men. This signals that partisan alignment explains more variance in trust for those in the USA who identify as female.

Conclusion
Our analysis suggests that citizens do seek traits that are caring and communal when evaluating trust in leaders during the Coronavirus pandemic, supporting early media coverage praising largely female leaders for displaying such traits (Hassan and O'Grady 2020;Piscopo 2020). Indeed, agreeing that the leader has exhibited caring traits is associated with increased trust, as hypothesised; evaluating the leader as competent is also associated with greater trust. However, we only find evidence for gender differences in trust judgements in the United States; there are no significant differences in the UK or Italy. We cannot therefore attribute females' lower trust to valuing communal traits more, or agentic traits less than men, outside of the USA. Indeed, British and Italian citizens of both sexes value both types of traits similarly. If gender gaps in trust are not through socialisation, as we posit through social role theory, then further research is needed to understand the determinants of trust gender gaps. It could, for instance, be the enduring lack of descriptive representation (Cowley 2014;Hinojosa, Fridkin, and Kittilson 2017) which develop over-time and cannot be identified in our cross-sectional surveys limited to countries with male leaders. Moreover, intersectional differences with age need to be further considered as gender gaps may vary by age cohort (Shorrocks 2016) as well as macro-level factors such as public policy (Sanders and Shorrocks 2019).
Gender differences are found in the USA. Here, we find evidence for H2 where caring has a greater effect on trust for women, and also for H3 where men's trust levels are greater affected by competence. This aligns with how modern gender gaps have developed at differential rates in Europe (Giger 2009). We are cautious, however, of attributing the US gender gap to socialized gender roles. The United States is also the only country where women are not shown to prize caring more than men. Although only one measure of gendered traits, social role theory would predict women are more likely to have communal traits, such as prizing caring. Therefore, we cannot be sure socialisation is driving this gendered difference. Moreover, although we find the agentic trait to have the greatest effect overall, between countries it is most strongly associated with trust in the USA. Understanding the roots of these differences requires further exploration.
What does this tell us about trust judgements during the COVID-19 pandemic? These initial results suggest leaders do not need to perform differently to ensure the trust of male and female voters. However, there is clear evidence that, unsurprisingly, (perceived) competence is by far the strongest predictor of trust. These results suggest that as the pandemic continues to unfold, rational judgement about the performance of leaders is going to sustain trust.
Of course, our analysis has limitations. First and foremost, the pandemic is a rapidly unfolding event, and our design is purely cross-sectional: a snapshot in time that does not capture over-time changes. Secondly, we use only three countries, all of which have male leaders. Our results would be interestingly contrasted with a similar analysis in a female-led country. Finally, we acknowledge that there is going to be overlap between our measures: for instance, people who value communal traits may consider this as competence as well, and as such our measures are going to be correlated.

Disclosure statement
No potential conflict of interest was reported by the author(s).