Yves here. Even though this post, being VoxEU, is pretty dry, it addresses a hot topic: the idea that the economic devastation of skyrocketing unemployment and failing businesses is due to government action to tamp down the spread of coronavirus. In fact, it concludes that reactions to coronavirus are the main driver.
My small sample from a not-terribly-sensitive area (a liberal pocket in the heart of Trumpland) doesn’t lend strong support. On the one hand, when the national news started braying about infections in New York (and then in neighboring states), pretty much everyone in the geriatric age group at my gym stopped attending. But even though I started doing take-out at our favorite restaurant (which would normally never do that but the chef knows us), it was packed when I’d go to the bar (wearing a mask even then) for pick-up, and the average age was over 40. And I doubt this spot will make it; I don’t see how the better independent restaurants get back on their feet. But more generally, I saw almost no changes in behavior here, even after the governor closed all the schools. Only when non-essential businesses were shuttered as well did some people seem to internalize the message that coronavirus might be serious.
The flip side is that by being self-employed, I have little contact with worker-driven demands to shutter operations if necessary safety measures aren’t taken. Remember the closure of auto plants due to union pressure. And see in our companion post today on how meatpacking plant workers don’t want to come in and risk getting sick if their employer won’t ante up to provide adequate protection.
So that may weigh heavily in the data, as well as knock-on effects from supply chain disruptions. For instance, an auto-industry supplier I know said Mercedes closed its US plants a week sooner than they had wanted because they were unable to get a critical part from Europe due to coronavirus disruptions there.
And Alabama is one of the five worst states in observing social distancing,1 so being a a less bad section may not make much difference.
By ChaeWon Baek, Peter B. McCrory, Todd Messer, and Preston Mui, all PhDs candidate in Economics, University of California, Berkeley. Originally published at VoxEU
Stay-at-home orders have been imposed in many countries to flatten the COVID-19 pandemic curve, but it’s not clear how much economic disruption is caused directly by the orders and how much by the coronavirus. This column disentangles the two by comparing the implementation of stay-at-home policies across the US and high-frequency unemployment insurance claims. The direct effect of stay-at-home orders accounted for a significant but minority share of the overall rise in unemployment claims; unemployment would have risen even without such orders. Unless the underlying public health crisis persists, undoing stay-at-home orders will only bring limited economic relief.
Closing schools, restricting economic activity, and promoting social distance are necessary to flatten the pandemic curve (Ferguson et al. 2020), but the economic costs of implementing such lockdown policies are difficult to pin down. In the US, the most prominent of these policies is the stay-at-home (SAH) order, which typically requires people to remain at home except for tasks and work deemed essential.
Researchers have already begun quantifying the general economic disruption in the US associated with the coronavirus pandemic in terms of economic uncertainty, reduced economic activity, increased firm risk, job losses, and reduced labour-force participation (see, respectively, Baker et al. 2020, Lewis et al. 2020, Hassan et al. 2020, and Coibion et al. 2020). However, little work exists quantifying the economic effects of stay-at-home policies themselves.
Starting in mid-March, state and local officials in the US implemented stay-at-home policies in order to limit the spread and severity of the coronavirus disease. Between 16 March 2020 (when the first was issued in the Bay Area, California) and 4 April, the share of the US population under stay-at-home orders rose from 0% to 95% (see Figure 1). In the same period, 16 million US workers filed for unemployment-insurance benefits.
Figure 1 Cumulative share of population under stay-at-home order in the US
Sources: Census Bureau, The New York Times; author’s calculations.
These orders have recently come under public criticism for exacerbating the economic disruption. However, it is unclear how much of the disruption was due to the stay-at-home orders themselves and how much is due to factors that would have occurred in their absence.
In Baek et al. (2020), we take a first step in disentangling the direct effect of stay-at-home orders from the general economic disruption brought on by the coronavirus pandemic. We do so by studying the impact of such policies on initial claims for unemployment insurance, a high-frequency, regionally disaggregated indicator of real economic activity in the US.
Implementation of Stay-at-Home Orders in the US
We use data from the New York Times to track implementation of stay-at-home orders in cities, counties, and states, measuring a state’s exposure to stay-at-home orders as the average number of weeks a state’s workers were subject to the orders. By 4 April 2020, California had the highest exposure to stay-at-home orders, with workers exposed on average to two-and-a-half weeks of such orders. Conversely, five states (Arkansas, Iowa, Nebraska, North Dakota, and South Dakota) had no counties under stay-at-home policies by 4 April. Figure 2 reports exposure to stay-at-home orders for workers in each state in the US, showing that there was considerable variation between states.
Figure 2 Employment-weighted state exposure to stay-at-home policies through week ending 4 April 2020
Sources: Census Bureau, The New York Times; author’s calculations.
The Effect of Stay-at-Home Orders on Unemployment Insurance Claims
To provide evidence of a causal link between the implementation of stay-at-home policies and the observed increase in unemployment insurance claims, we couple the spatial and regional variation in stay-at-home implementation with high-frequency unemployment claims data by state. This allows us to isolate the economic disruption resulting from the stay-at-home policies. To make states comparable, we scale initial claims by state employment as reported in the 2018 Quarterly Census of Employment and Wages.
We compare the evolution of scaled unemployment insurance claims of ‘early adopters’, defined as those states being in the top quartile of exposure to stay-at-home orders through 4 April, to those of ‘late adopters’, defined as those states being in the bottom quartile. As shown in Figure 3, in the first few weeks, early adopters initially had a higher rise in unemployment claims relative to late adopters. By the week ending 4 April, the relative effect of adopting stay-at-home orders early largely disappears, reflecting the fact that by this point approximately 95% of the US population was under a stay-at-home order.
Figure 3 Box plots by week of initial unemployment-insurance claims relative to employment for early and late adopters of stay-at-home policies
We find a positive correlation in cumulative unemployment insurance claims and our measure of stay-at-home exposure, both measured through 4 April (Figure 4). Each bubble in the figure represents a state, with the size of the bubble indicating population and the colour indicating the severity of the local, reported coronavirus outbreak.
Figure 4 Scatterplot of stay-at-home exposure to cumulative initial weekly claims for weeks ending 21 March thru 4 April
We formalise this methodology and find that an additional week of exposure to stay-at-home policies increases unemployment insurance claims by approximately 1.9% of a state’s employment level (using 2018 employment levels by state, as reported in the Quarterly Census of Employment and Wages). This result is robust controlling for factors potentially related to both stay-at-home implementation and the magnitude of new unemployment insurance claims. For example, we control for sectoral composition related to job losses occurring in the week before the first stay-at-home policy went into effect, the share of jobs likely able to be done at home, and the severity of the COVID-19 outbreak in the state.
We calculate that the direct effect of stay-at-home orders is accountable for 4 million unemployment insurance claims between 14 March and 4 April, which accounts for approximately a quarter of the overall rise in unemployment claims in that period. The direct effect of stay-at-home orders on unemployment is therefore small relative to the aggregate increase in unemployment insurance claims, suggesting that a large majority of the increase in unemployment may have occurred in the absence of such orders.
Corroborating Evidence from the Google Community Mobility Index
Along with unemployment claims, we also investigate the effect stay-at-home orders have on mobility using the Google Community Mobility Index, which tracks visits to different categories of locations (e.g. grocery stores, workplaces, or parks). We focus on mobility to retail establishments, as the main goal of stay-at-home orders was to limit non-essential business activity. The Google mobility data provide additional evidence of the effect of stay-at-home orders as it relates to a ‘demand channel’.
Figure 5 shows how a county’s retail mobility changes when stay-at-home orders are implemented (x-axis equal to 0). The extremely high-frequency nature of this data allows us to address the concern that our results are driven by anticipation effects, whereby people expect policymakers to implement stay-at-home policies imminently and reduce retail activity because the public health threat is much more salient. We find no evidence of this type of anticipation effect.
Before stay-at-home orders were implemented, retail mobility evolved similarly across counties, as evidenced by the flat line. The day they were announced, the orders reduced retail mobility by 5%. For the following two weeks, retail mobility stayed close to 10% lower relative to other counties as a result of the stay-at-home orders. Considering that during this period mobility fell by 40%, this data suggests a similar share of economic decline attributable to stay-at-home orders as the unemployment insurance data.
Figure 5 County retail mobility index
Sources: Google, The New York Times; authors’ calculations.
Our results support the idea that flattening the pandemic curve implies a steepening of the recession curve in the absence of any government support (Gourinchas 2020). Stay-at-home orders account for a significant but minority share of the overall rise in unemployment claims during the pandemic, implying that much of the rise in unemployment during this period would have occurred in the absence of these orders.
This suggests that any economic recovery that arises from undoing stay-at-home orders will be limited if the underlying pandemic is not resolved. Weak consumer confidence, supply-chain disruptions, and self-imposed social distancing are just a few examples of economic headwinds that could persist even in the absence of stay-at-home orders. At the same time, the orders are likely to have public-health benefits from slowing the spread of the coronavirus.1 Taken together, our results caution policymakers not to expect the reopening of the economy to be an economic panacea.
See original post for references
1 This metric has some problems. Home health care aide, nurse, and physical therapist visits to my mother would all incorrectly be deemed “non-essential,” as would visits by relatives to pick up or care for children of “essential workers” when they had to start their shifts. I would assume the people keeping track would argue it nets out across states but does it?