While uncertainty among firms and professional forecasters is known to be countercyclical (Bloom et al. 2018), a large share of the current literature focuses on the role of real economic factors, such as stock returns or employment growth. Uncertainty surged over the course of the pandemic and, while it has currently subsided, it remains above trend (Baker et al. 2020a). An additional literature on belief formation has also emerged, emphasising the role of personal experience (e.g. Malmendier and Nagel 2011) and social networks (e.g. Makridis and Wang 2020) as overall determinants.
Why do you think? Tell me your party
Using novel and high-frequency data between March and June on a nationally representative sample of roughly 500-1,000 respondents per day from the Gallup Panel, our recent research investigates the determinants of beliefs about the pandemic and its economic implications (Makridis and Rothwell 2020). Our survey contains a wide array of demographic characteristics and location-specific information, allowing us to quantify the role of county-specific factors, such as COVID-19 infections per capita and unemployment claims, against socio-economic factors, such as age, education, race, and political affiliation, among others.
We focus on six major outcomes and create the following variables: (i) an indicator for whether the individual expects the economic disruption to last at least until the end of the year, (ii) an indicator for being somewhat or very worried about contracting coronavirus, (iii) an indicator for practicing social distancing very often or always in the past 24 hours, (iv) an indicator for mostly isolating and having little contact in the past 24 hours, (v) an indicator for wearing masks in the past week, and (vi) an indicator for visiting the workplace in the past 24 hours. These measures of the pandemic provide a strong complement to elicitations of expectations about the economy by Coibion et al. (2020).
Figure 1 Behaviour and expectations of US workers, by political affiliation
Figure 1 summarises the share of workers separately by political affiliation for each of these variables over time. We observe significant cross-sectional and time series variation in these attitudes. For example, in early April, roughly 32% of Republicans were visiting the workplace, whereas only 20% of Democrats did the same. By June, those shares climbed to roughly 43% and 33%, respectively, continuing to demonstrate a large partisan gap. Similarly, whereas only 15% of Republicans anticipated a large disruption as of early April, as many as 38% of Democrats expected a disruption. Moreover, these differences have widened over time with roughly 38% of Republicans expecting a disruption and 80% of Democrats expecting one.
In the first half of the paper, we take these data to a more formal regression framework. We find that political affiliation is the most important predictor (as measured by the t-statistic) of these beliefs. For example, Republicans are 18% less likely to believe that the COVID-19 disruption will last until the end of the year, whereas Democrats are 11% more likely, relative to independents or those who prefer an ‘other’ party.
To put that in perspective with other correlates, we see that a 10% rise in the number of new infections per capita is associated with a 0.2% increase in the probability of expected disruption. Moreover, political affiliation is even more predictive of economic expectations than employment status (employed are 8% less likely to expect significant disruption), education (those with graduate degrees are 3% more likely), or even health (those with a medical condition are 5% more likely).
We see similar patterns when we look at other outcome variables. For example, Republicans are 5% less likely to report being somewhat or very worried about getting the illness, while Democrats are 6% more likely, relative to moderates. Again, we see that increases in actual local infections raises concerns about contracting the virus, but only a fifth to a sixth as much as political affiliation. Here, employment status matters relatively more: those employed in a job are 6% less likely to worry about getting sick, perhaps reflecting that many are working remotely. Not surprisingly, we see that those with serious medical problems are 8% more likely to worry about getting sick, which reflects not only a potential selection effect, but also the possible heightened exposure to COVID-19.
One of the advantages of our data is that we observe a wide array of demographic characteristics and political affiliation, allowing us to compare how beliefs vary with and without political affiliation as a control. In particular, when we control for it, the significance of age, education, and, even to a large extent race, diminish as predictors of these different measures of beliefs.
The real costs of political partisanship
In the second half of the paper, we show that political partisanship also affects the economy in part by leading to overly restrictive or overly lax state policies in response to the pandemic. For example, we find that a percentage point rise in the state share that voted for Trump in 2016 is associated with a 1.9 percentage point lower probability that the state adopted a nonessential business closure and 1.5 percentage point decline in the probability that it adopted a stay-at-home order, which are both highly significant even after controlling for a wide array of demographic and industry factors.
Figure 2 Policy decisions and economic outcomes, by political affiliation
We subsequently explore the effects of these different state policies on county economic outcomes. As a simple illustration, Figure 2 shows that that political affiliation is closely tied with policy decisions. For example, we see that states that Trump won in the 2016 election have roughly five percentage point lower unemployment rates and 10%-20% more retail visits, small business growth, and consumer spending, relative to the pre-crisis trend. We subsequently take these data to the county-level using data made available by Chetty et al. (2020) and control for local characteristics.
While declines in retail visits are almost mechanical, the result for small businesses is unique: the introduction of a stay-at-home order (SAHO) and nonessential business closure is associated with a 3.3-3.7 percentage point decline in revenue growth for small businesses. Although we find an economically meaningful 1.3 percentage point decline for credit card spending, it is not statistically significant, which could reflect the offsetting increase in spending on digital goods through online platforms (Baker et al. 2020b, Chetty et al. 2020).
Implications for modelling aggregate costs
Beliefs play an important role in driving consumption and broader economic activity (Gillitzer and Prasad 2018, Benhabib and Spiegel 2019, Makridis 2020). The fact that political affiliation matters so much provides microeconomic evidence for models of belief distortions and their aggregate effects, as in Bianchi et al. (2020). We also speak to the potential for scarring, as in Kozlowski et al. (2020) and Portes (2020), that can arise when individuals with different political affiliations observe the same shock but arrive at different conclusions. Our results on the importance of political affiliation also relate with evidence from Allcott et al. (2020) and Bursztyn et al. (2020) who explore how these beliefs informed by political affiliation inform disease-mitigating behaviours.
Our ongoing work exploits the panel structure of our data to examine whether updates to beliefs are correlated with political affiliation and local and/or aggregate changes in economic activity.
Allcott, H, L Boxell, J Conway, M Gentzkow, M Thaler and D Y Yang (2020), “Polarization and public health: Partisan differences in social distancing during the Coronavirus pandemic”, NBER Working Paper 26946.
Benhabib, J and M M Spiegel (2019), “Sentiments and economic activity: Evidence from U.S. states”, Economic Journal 129(618): 715-733.
Bianchi, F, S C Ludvigson and S Ma (2020), “Belief distortions and macroeconomic fluctuations”, NBER working paper.
Bloom, N, M Floetotto, N Jaimovich, I Saporta-Eksten and S J Terry (2018), “Really uncertain business cycles”, Econometrica 86(3): 1031-1065.
Bursztyn, L, A Rao, C Roth and D Yanagizawa-Drott (2020), “Misinformation during a pandemic”, BFI working paper.
Chetty, R, J N Friedman, N Hendren and M Stepner (2020), “How did COVID-19 and stabilizing policies affect spending and employment? A new real-time economic tracker based on private sector data”, NBER Working Paper.