Understanding Distributional Impacts is Key to Community Resilience | LBJ School of Public Affairs | The University of Texas at Austin

By D. Cale Reeves, Vivek Shastry and Varun Rai

In the wake of large-scale shocks like the COVID-19 pandemic, community well-being requires policy responses that put equity and opportunity across all sections of the community at their center. A common policy response in the first few months of the pandemic was some form of local lockdown. For example, in March, 2020, the mayor of Austin declared a local disaster and issued the "Stay Home—Work Safe" (SHWS) order, "requiring all individuals in the City to stay home or in their place of residence except to perform certain essential activities, or to perform work in or obtain service from an Essential Business, Essential Government Service, or in Critical Infrastructure." While orders like these clearly reduced the local impact of COVID-19, shutting down nonessential business and government functions had a wide range of secondary impacts, which were not all positive. It is now clear that some already-vulnerable populations bore a disproportionate share of the negative health, economic, and social outcomes.

Resiliency Toolkit
Full Report (University of Texas Libraries)

Not all communities—nor the individuals that comprise them—have equal access to the resources that enable resiliency. Many people quickly adapted to the new behaviors that the lockdown required, for example, picking up their groceries through "curbside" delivery programs. Those that could shift to working from home did so in compliance with the SHWS order, continuing to earn their incomes in isolation, while reducing their risk of infection. But many employed in the service industry were unable to work at all, as bars and restaurants saw substantially reduced business or in many cases shuttered entirely. Many employees in roles declared essential were able to retain their income streams, but at the expense of increased exposure to COVID-19. The inequitable impacts illustrated here are problematic in and of themselves, but they are exacerbated when they aggravate underlying inequities—such as when service industry or essential employees are disproportionately composed of lower income earners or marginalized groups. Unless holistically designed, even necessary and effective policy actions can cause irreparable long-term damage to certain populations.

Effectiveness and equity together must form the core criteria by which policy outcomes are evaluated—the long-term resilience of the community as a whole demands it. But with little precedent to identify what might constitute a holistically good response in the face of high levels of uncertainty, a wide variety of policy designs are inevitable. Shelter-in-place—similar to the SHWS order in Austin—started as early as March 17 in some California counties, and two weeks later in Texas and Florida. The duration of the shelter-in-place order was only a month in Texas and Florida and two months or longer in Washington and Michigan. Additionally, varying definitions of essential services target shelter-in-place orders to different groups of workers. Each of these policy design elements—timing, duration, and targeting—evokes a different pattern of behavior among individuals in the jurisdiction. At a systems level, the effectiveness of the response depends on the degree to which individuals do or do not comply with it—a decision stemming from individual risk perceptions, risk tolerances, and social influences.

Unless holistically designed, even necessary and effective policy actions can cause irreparable long-term damage to certain populations. Effectiveness and equity together must form the core criteria by which policy outcomes are evaluated—the long-term resilience of the community as a whole demands it."

For policymakers to balance effectiveness and equity, they need access to tools that allow them to assess the system-level dynamics of both the pandemic and of their own (policy) responses, so they can resolve potential problems of unequal impacts. Top-down "mass-action" type models—i.e., those that focus directly on population-scale dynamics (e.g., progression of the pandemic at the county or city level) while only coarsely resolving socio-demographic aspects—are well-understood tools for describing entire systems and have played a central role in guiding policymakers through the uncertainties of this pandemic. But top-down modeling often lacks the individual- and community-level insights that can drive nuanced, equity-based decision-making.

Agent-based modeling (ABM) is a bottom-up approach that focuses on the investigation of individual-level outcomes and then links those micro-level responses and impacts into a system-wide macro-level estimate of efficacy. By focusing on individuals or households, ABM also incorporates aspects of individual level decision-making—such as the decision to comply with a shelter-in-place order—that most mass-action models only resolve at aggregate levels. The ABM approach permits a nuanced understanding of the distributional impacts of policy choices, an understanding that is critical for enacting just and equitable policies that help build and preserve the resilience of communities.

To understand the full spectrum of interactions among policy design elements and behavioral responses, we introduce the COVID-19 Policy Evaluation (CoPE) tool, developed recently at the LBJ School of Public Affairs.[1] CoPE is a flexible, modular, empirically and epidemiologically grounded agent-based model of the spread of COVID-19 that compares, ex ante, the impacts of different policy design elements on the effectiveness and distributional equity of policy outcomes. Using tools such as CoPE, which use equity as a central criterion for policy evaluation, can help policymakers choose the responses that allow communities to better absorb and recoil from large-scale shocks in the short-term, while safeguarding their resilience in the long-term.

We used the CoPE tool to simulate a range of policy and behavioral response scenarios for Travis County, Texas (CoPE enables similar studies for nearly all other counties across the U.S.). In our baseline scenario, the shelter-in-place policy goes into effect 28 days after the first exposures occur and is partially lifted 45 days later. Occupations not designated as essential under the guidelines from the CDC and the Texas Governor's office cease having "on-the-job" interactions, while essential employees continue to interact both with their co-workers and with those receiving their services. CoPE also models social interactions outside of the workplace (this is important since mobility reports suggest that only about 75 percent of Travis County residents reduced their social activities in compliance with the shelter-in-place order).

Figure 1: An early shelter-in-place order substantially reduces hospitalizations compared to the baseline. The effect is most noticeable in peak hospitalizations. Late SIP orders increase both cumulative (left vertical axis) and peak hospitalizations (right vertical axis) over the baseline. Note: to highlight the tradeoff in efficacy and equity, the model that generates these results has a higher rate of infection than is empirically observed typically.

From this baseline, we investigate several hypothetical scenarios shown in Figure 1, including what would have happened had shelter-in-place been enacted one week earlier or later than it was. The CoPE tool shows, among other outcomes, the important impact of shelter-in-place timing on cumulative hospitalizations and peak hospitalizations. For the illustrative scenarios shown in Figure 1, we estimate that a shelter-in-place policy enacted one-week earlier than the baseline results in about 40 percent lower total hospitalizations and 30 percent lower peak hospitalizations compared to the baseline. On the other hand, a shelter-in-place policy enacted one-week later than the baseline results in about 50 percent higher total hospitalizations and over 200 percent higher peak hospitalizations compared to the baseline. These results from the CoPE tool are generally consistent with the consensus from several other modeling approaches, including top-down mass-action type models. From both top-down and bottom-up modeling approaches it is clear that when faced with a shock event such as the COVID-19 pandemic, policy responsiveness is paramount. But that is only half the story.

Because it uses a bottom-up approach, the CoPE tool can also explore the distributional impacts of policy choices. For the same illustrative scenarios described above, Figure 2 shows how a difference in policy design can impact income groups differently. A rapid policy response—enacting shelter-in-place only 21 days after initial exposures (i.e., one week earlier than the baseline scenario)—yields fewer hospitalizations (Figure 1), but the hospitalizations that do occur disproportionately burden the lowest income group (left panel vs. middle panel, Figure 2). In this case, the efficacy gain of an early shelter-in-place order is borne in part on the shoulders of those with the least access to resources that enable resilience.

Figure 2: Distributional equity of shelter-in-place (the shaded time period) impacts on hospitalization by income groups show that the lowest income groups are hardest hit in both extremes. Dotted lines show the baseline proportion of the population comprising each group, e.g. Middle-income earners make up nearly 50 percent of the population. Solid lines show the proportion of daily new hospitalizations from each income group. When the solid line is above the dotted line—as it often is for the lowest income group—that group is experiencing disproportionately higher hospitalization rates. Note: to highlight the tradeoff in efficacy and equity, the model that generates these results has a higher rate of infection than is empirically observed typically.

Policy that is designed and implemented without consideration of distributional impacts is no longer excusable—our communities, particularly those that are already marginalized and vulnerable, demand and deserve better.

The discussion above suggests two key takeaways relevant to community resilience. First, a critical aspect of resilience is to manage the scale and scope of shock experienced by communities. In the case of COVID-19, rapidly enacted policies helped prevent the situation from spiraling out of control in many places, thereby limiting the overall burden of COVID-19. This shows the importance of maintaining vigilant awareness of the threat horizon and of having contingency plans that can be implemented rapidly. Second, the distributional impacts of policy decisions can substantially disadvantage already underserved communities—for example, the lowest income groups, as illustrated above. When they are disproportionately impacted by events like hospitalization—which can incur unexpected costs and decreased income, alongside significant health impacts—individual and community survival are put at risk in both the short- and long-term. Effective policy is still a central value, but when a policy is expected to be effective and to disproportionately burden the vulnerable, complementary programs must be implemented specifically to support those vulnerable communities and minimize adverse impacts.

A socially just policy response must consider both effectiveness and equity. Policymakers need more tools like CoPE to enable them to make decisions and design policy with distributional effects at the fore of the discussion. Policy that is designed and implemented without consideration of distributional impacts is no longer excusable—our communities, particularly those that are already marginalized and vulnerable, demand and deserve better.

 

Dr. D. Cale Reeves is a Postdoctoral Research Fellow, Vivek Shastry is a Ph.D. student, and Dr. Varun Rai is a Professor and Associate Dean for Research at the Lyndon B. Johnson School of Public Affairs at the University of Texas at Austin. Dr. Rai is also director of the UT Austin Energy Institute.


[1] For more information and to access the CoPE tool, visit http://sites.utexas.edu/raigroup/

 

Faculty