Businesses in high-income zip codes often saw sharper visit reductions during the COVID-19 pandemic

The COVID-19 pandemic drastically altered global mobility patterns. Traditional factors influencing mobility (travel time, cost, convenience) were augmented by infection risk and government policies like lockdowns and stay-at-home orders. This study investigates the impact of these changes on business visits in Minnesota, focusing on the disparity between high- and low-income zip codes. The research questions whether businesses in affluent areas experienced sharper visit reductions compared to their lower-income counterparts, both during and outside of lockdown periods. The findings are crucial for informing targeted post-pandemic economic recovery strategies. Figure 1 illustrates this disparity, comparing restaurant visits in Prior Lake (high income) and Hibbing (low income), revealing a more sluggish recovery in Prior Lake after the initial lockdown. The study aims to systematically compare pre- and post-pandemic visit changes for similar businesses across high and low-income zip codes to quantify this disparity and understand its implications for recovery efforts.

Existing literature demonstrates an overall decline in business visits during the pandemic, attributed to factors like job losses, infection risk aversion, and government mandates. However, this research distinguishes itself by focusing on the differential impact across income levels. While studies have shown a correlation between income and behavioral responses to social distancing (Weill et al., 2020), unpacking the mechanisms (income shock vs. risk aversion) remains a challenge. This paper addresses this gap by comparing visits to similar businesses in high- and low-income areas, controlling for business type and accounting for seasonality using 2019 data as a baseline. It also acknowledges limitations of difference-in-differences methods due to potential violations of the parallel trends assumption caused by varying risk aversion and responses to government policies across income groups. Previous research often used short observation periods and did not adequately account for seasonality, unlike this study.

The study utilizes SafeGraph's COVID-19 Data Consortium, containing data on weekly visits to millions of points of interest (POIs) across the U.S. The data includes information like NAICS code, address, zip code, and number of visits. Income and population data come from the American Community Survey. Zip codes were categorized as high or low income based on pre-pandemic median income. The analysis focuses on five periods: pre-pandemic (Feb 2 - Mar 16, 2020), first lockdown (Mar 17 - June 1, 2020), interim reopening (June 2 - Nov 23, 2020), second lockdown (Nov 24, 2020 - Jan 11, 2021), and second reopening (Jan 11 - May 31, 2021). A pairwise comparison method is employed, comparing the change in visits (relative to 2019) for all possible pairs of high and low-income zip codes for each business category and period. This approach mitigates the limitations of difference-in-differences methods by accounting for unobserved heterogeneity in risk aversion and policy responses. The key metric is the percentage of pairwise comparisons where high-income zip codes showed a sharper reduction in visits than low-income zip codes. Kernel density estimation plots visualize the distributions of these differences across all pairwise comparisons.

Table 2 presents the percentage of pairwise comparisons showing greater visit reduction in high-income zip codes for various business categories across different periods. The results reveal that in many business categories (e.g., full-service restaurants, religious organizations, movie theaters), high-income zip codes experienced a more significant decrease in visits than low-income zip codes, particularly during periods outside of lockdowns. This suggests that factors beyond mandated lockdowns, such as higher risk aversion, contributed to the reduced visits. The difference is less pronounced during lockdown periods, indicating that mandated closures had a more uniform impact across income levels. Figures 2 and 3 show kernel density plots for full-service restaurants and grocery stores, illustrating the shift in the distributions from being centered around zero (similar changes in high and low income areas) pre-pandemic to being skewed to the left (sharper reductions in high-income areas) post-pandemic for restaurants, but remaining centered around zero for grocery stores. This indicates that essential services like grocery stores were less affected by income-related differential behavior. Table 2 also reveals that during lockdown periods, low-income zip codes experienced more pronounced visit reductions across all business categories compared to high-income zip codes. This suggests the effects of income inequality on business visit reductions differed significantly between lockdown and non-lockdown periods.

The findings highlight the disproportionate impact of the pandemic on businesses in high-income areas, even outside of mandated lockdowns. This suggests that factors like heightened risk aversion or greater capacity for alternative consumption patterns (e.g., online ordering) in affluent communities played a significant role. The disparity in visit reductions has implications for targeted economic relief efforts. The concentration of sharper visit reductions in high-income zip codes for businesses promoting prolonged indoor visits (e.g., restaurants, theaters) suggests the need for targeted aid to these sectors. This is contrary to the argument that economic relief during lockdown periods should concentrate on low-income areas because they suffered a more significant economic impact during lockdowns. The persistence of these differences beyond lockdown periods indicates that these factors persist and that relief aid strategies must be carefully considered. The similar visit changes to essential services across income levels further supports the need for targeted strategies, not blanket policies.

This study reveals a nuanced impact of the COVID-19 pandemic on business visits, with higher-income zip codes experiencing sharper declines in non-essential, indoor-focused businesses even outside lockdown periods. This calls for a targeted approach to post-pandemic economic relief, prioritizing sectors and locations differentially affected. Future research should investigate the impact on spending, modes of transportation, and the role of factors beyond income in shaping mobility patterns.

The study uses visit counts as a proxy for economic activity, lacking granular data on spending or revenue. The lack of data on transportation modes and the potential for delivery services to affect visit counts are also limitations. The analysis focuses on Minnesota and may not generalize to other regions with different pandemic response strategies. The categorization of zip codes into high and low-income groups based on pre-pandemic data is a simplification.