The COVID-19 pandemic presents legitimately difficult policymaking tradeoffs, as society seeks to limit the social and economic damage caused by the pandemic while mitigating the spread of disease. This is apparent with respect to school reopening. Keeping schools closed has the potential to exacerbate existing inequalities in educational attainment and outcomes, and makes it difficult for those who have school-aged children to return to work. At the same time, opening schools for in-person instruction has the potential to exacerbate community transmission of the virus.
Ideally, policymakers would be able to reflect some amount of input from the public in their responses to the pandemic, but this is especially difficult in cases such as these. Many people don’t have one set preference regarding whether schools in their community should be open, and instead have multiple, conflicting preferences that depend on relevant conditions. Specifically, many citizens would prefer that schools be open for in-person classes on the condition that this would not create undue risk for themselves or their communities, and would prefer that they stay closed otherwise. Survey items that only ask for one summary preference without making such conditions explicit do not give respondents an opportunity to articulate this complexity.
This paper outlines systematic approaches for measuring these conditional policy preferences using a combination of survey experiments and recently-developed machine learning approaches for analyzing them. On the question of school reopening in particular, we find that the vast majority of respondents are open to becoming more supportive of having schools open for in-person classes under certain conditions -- especially if members of the scientific community say it is safe and if the virus becomes less prevalent. However, we argue that this analytical framework should be applicable to public opinion in many other areas of policymaking where, in conversation, a question about whether someone supports an idea is likely to be met with “It depends.”
