Downtown Downturn: The Covid Shock to Brick-and-Mortar Retail

Introduction

Nearly three years after its onset, work-from-home (WFH) technologies and their effects are one of the most important and enduring economic legacies of the COVID-19 pandemic (Barrero, et al., 2021). According to the U.S. Census Bureau, 27.6 million Americans reported primarily working from home in 2021, triple the size of that population in 2019.¹  In response to this revolution in work and commuting patterns, high-income residents in primarily business services industries have moved from downtown to suburban neighborhoods and out of large, expensive cities (Ferreira & Wong, 2022; Ding & Hwang, 2022; Athloff, et al., 2022; Li & Su, 2022; Brueckner, et al., forthcoming; Ramani & Bloom, 2021). To date, most of the literature on the responding change in firm locations has focused on those whose workers have made the WFH transition. They find that these firms are also leaving downtown locations and substantially reducing the size of their commercial office space (Dalton & Groen, 2022; Rosenthal, et al., 2022; Gupta, et al., 2022).

Less well studied is the impact of WFH on the location decisions of brick-and-mortar establishments owned by retail firms. For retail establishments, exposure to WFH adoption happens more directly through changes to the location and behavior of their customer base, rather than through their workforce (which remains largely in-person). For instance, a substantial share of consumer shopping trips begin or end at home locations, leading many firms to locate near residential areas (Relihan, 2022; Farrell, et al., 2017). This may drive retail firms to similarly shift to where consumers have moved residency. In addition, consumers tend to chain visits to retail stores with their commutes (Miyauchi, et al., 2022). With fewer consumers commuting each day and commuting from different locations, traditional employment centers may be unable to support the same level of retail establishments. Furthermore, this retail activity may not simply move to residential locations if consumers who work from home make fewer and different trips—for instance, people who work at home can prepare their own lunches instead of visiting a local restaurant. ²

This report provides the first detailed evidence on the changing locations of retail establishments due to the widespread adoption of WFH technologies. To do so, we build a novel dataset to track the entry and exit of retail establishments through Q4 2021 based on the credit and debit card activity at card terminals located in each establishment. With this dataset, we show that there is wide variation in the extent of the retail establishment recovery across space. We find that retail establishments have paralleled the exodus of populations from large, expensive cities in preference for smaller, Sun Belt cities and from city centers to suburbs. These effects counterbalance, such that the aggregate number of establishments is almost equal to its pre-pandemic baseline. In addition, we show that neighborhood exposure to the WFH shock is a key driver of the changing composition of retail establishments in downtown versus suburban areas, even controlling for neighborhood changes in population. This is likely due to the different shopping behaviors of those who work from home, such as reduced demand for clothing and personal care services.

Understanding the scale and nature of the effects of population growth and WFH on retail locations is important because these establishments account for a large portion of commercial space and economic activity. Prior to the pandemic, retail goods and services accounted for 29 percent of principal commercial building activities, compared to 16 percent for office.³  If establishments are changing in composition and location, then this will have important spillover effects to other sectors of the economy, including labor and housing markets. These effects also have implications for the production and consumption value of cities and the policy interventions, such as changes in land use regulations and tax codes, that may be necessary to support continued urban economic health.

Data asset

We use transaction data from over 70 million Chase customers to identify brick-and-mortar establishments from Q1 2017 to Q2 2022. Our definition of an establishment is the unique combination of a merchant identifier (e.g. “Bob’s Restaurant #1234”), ZIP Code Tabulation Area (ZCTA), and product corresponding to a point of sale where in-person transactions take place. To ensure sufficient card activity to pinpoint the entry and exit of establishments, we limit our study of establishments to 16 major cities in which JPMorgan Chase has a significant customer base and make observations at a quarterly frequency.⁴ Our definition of a city is a Core-Based Statistical Area, which includes both the primary city and outlying suburbs.

We focus on a subset of everyday retail products that are well-captured by card transaction data. These include groceries, general goods, clothing, restaurants, home goods, leisure, personal care services, professional consumer services, and pharmacies. The following are non-exhaustive lists of examples for select product types: grocery stores include traditional grocers and specialty food stores. General goods include department stores, discount stores, and other retailers that sell everyday goods (e.g., florists or booksellers). Restaurants includes bars that sell food to be consumed on premises in addition to full-service and fast-food restaurants. Home goods includes furniture and home improvement stores. Personal care services includes salons, barbershops, and dry cleaners. Leisure includes venues like movie theaters, bowling alleys, and gyms. Professional consumer services includes veterinarians, tax preparation, and childcare.

Our establishment identification process defines activity for over 1.7 million establishments during our study period. We consider an establishment to be active in each quarter that we observe any transaction at that establishment, while we consider establishments to exist for all quarters inclusive of their first and last active quarters. For example, consider an establishment that is first active in Q3 2019, is active for the next two quarters before becoming inactive during a local shutdown in Q2 2020 and Q3 2020, and then is active again in Q4 2020 until Q2 2021. We identify the entry date for this establishment as Q3 2019 and its exit date as Q2 2021. For an external benchmark, we compare our results for establishment levels in 2019 and changes from 2018 to 2019 to the establishment survey conducted by the Census County Business Patterns and find them sufficiently consistent.

Interpretation of our establishment measure depends on several important features. The first is that we do not cover all establishments. Absent from this dataset are establishments with cash-only payments and those outside retail goods and services, such as in manufacturing and business services. The second is that establishments are distinct from firms. This can lead to a different pattern in entry and exits when the focus is establishments versus firms. For instance, a firm may close an establishment but still exist. A firm can also open and close establishments at the same time across a city leading us to measure entry and exits across space without any change in the number of firms.

We combine these data with information on consumer locations from the residential addresses reported by customers with at least 10 transactions in each reference month, changes in neighborhood spending offline and online derived from Chase card data, pre-pandemic neighborhood demographic data from the 2015-2019 American Community Survey (ACS), and residential and employment workforce data from the 2019 Longitudinal Employer-Household Dynamics Origin-Destination Employment Statistics (LODES). Data is merged by the ZCTA geography. Throughout the analysis we refer to ZCTAs as neighborhoods for ease of exposition.