Disease-economy trade-offs under alternative epidemic control strategies

The COVID-19 pandemic caused unprecedented health and economic impacts, including over 448 million infections and more than 6 million deaths worldwide by early 2022, and a historic contraction of U.S. GDP at an annual rate of 32.9% in Q2 2020. These events highlight the need for coupled frameworks linking epidemiological dynamics with economic behavior to assess disease-economy trade-offs. The authors develop such a framework that maps individual choices in consumption and labor-leisure, which generate infection-risking contacts, to epidemic dynamics. They identify a key coordination failure: infectious individuals impose externalities on susceptible individuals, prompting the latter to withdraw from economic activity (voluntary isolation), which deepens recessions. The study compares three control strategies—voluntary isolation (decentralized behavior), targeted isolation (coordinated isolation of infectious individuals), and blanket lockdowns (uniform isolation)—to determine whether and how targeted isolation can overcome disease-economy trade-offs.

Four strands inform the assessment of control strategies: (1) epidemiological studies of disease dynamics and heterogeneous impacts of interventions; (2) epi-economic studies modeling individual behavior as a driver of transmission and evaluating costs/benefits of controls; (3) macroeconomic analyses embedding disease-behavior interactions into broader economies or projecting macro outcomes without explicit epidemiology; and (4) statistical analyses relating disease-related behaviors to economic activity. Gaps include tractable structural mapping from economic activities to infection-relevant contacts and calibration using social contact surveys; limited integration of detailed microbehavioral and transmission mechanisms into full-economy models; and a narrow, simplified set of control strategies in existing coupled models, notably lacking individual-focused targeted isolation. The paper addresses these gaps by building a calibrated coupled epi-economic model with an activity-to-contact mapping and by explicitly analyzing targeted isolation under information and compliance frictions.

The study constructs a dynamic coupled epi-economic model integrating a SIRD transmission framework with forward-looking individual economic choices. Individuals of health types S (susceptible), I (infectious), and R (recovered) choose consumption and labor, generating contacts that drive infection risk. Contacts are linked to activities via a contact function calibrated from pre-pandemic social contact matrices and economic data. Baseline contact function is linear: ψ(A) = Σ ρ_i A_i A_i, where coefficients convert dollars spent (consumption) and hours worked (labor) into contacts; robustness checks consider ψ(A) = Σ ρ_i (A_i)^α with concave/convex forms to capture heterogeneous contact rates. The SIRD model updates states as S_{t+1} = S_t − r ψ(A) S_t I_t; I_{t+1} = I_t + r ψ(A) S_t I_t − (p_R + p_D) I_t; R_{t+1} = R_t + p_R I_t; D_t = D_t + p_D I_t, with β decomposed into biological (r) and behavioral (ψ(A)) components. Individual optimization: forward-looking agents choose consumption and labor to maximize lifetime utility with risk of infection and mortality (utility of death Ω calibrated to a value of statistical life), subject to per-period budget p_c c_t^m = w φ_t^m l_t^m (normalized consumption price, wage, and type-specific productivity; φ_I < 1 reflects productivity losses for infectious individuals). Two problems are solved: (1) decentralized (voluntary isolation) where agents maximize personal utility; and (2) social planner (targeted isolation) coordinating type-specific choices to maximize aggregate utility and internalize transmission externalities. Blanket lockdown is modeled as a constraint forcing all types to isolate regardless of disease status. Calibration: U.S.-specific age-location contact matrices (Prem et al., 2017) grouped into consumption, labor, and unavoidable activities; next-generation matrix methods set R0 ≈ 2.6 at the disease-free state; pre-epidemic consumption and labor map contacts to dollars/hours; unavoidable contacts normalized. Baseline assumes perfect information (agents know health status) and full compliance; applications introduce frictions: (a) test quality (single metric capturing sensitivity/specificity) and reporting lags (8→5→3 days over time) affecting perceived states and actions via weighted choice rules; (b) compliance modeled as a fraction deviating to voluntary isolation behavior under targeted isolation or lockdown mandates. Numerical solution uses value function iteration to compute policy functions for consumption and labor by health state and policy regime. Sensitivity analyses vary asymptomatic prevalence via productivity losses, activity contact ratios (consumption vs labor; avoidable vs unavoidable), and contact function curvature (α).

• Targeted isolation vs voluntary isolation and blanket lockdowns: Disease trajectories under targeted and voluntary isolation are nearly identical (overlapping infection curves), while blanket lockdowns can reduce cases more during enforcement but induce large rebounds upon relaxation (nearly 100% of cases averted recur later). Despite similar disease outcomes to voluntary isolation, targeted isolation dramatically reduces economic losses by shifting isolation burdens to infectious individuals.
• Economic impact: Peak-to-trough contraction is 66% under voluntary isolation and 84% under blanket lockdown, but only about 3% under targeted isolation. Targeted isolation averts approximately 91–95% of individual economic losses relative to voluntary isolation (Fig. 2B/C), translating to up to roughly $3.5 trillion in avoided recessionary losses; estimated required spending to support isolation (e.g., two weeks’ pay for infected individuals) is around $428 billion, yielding net savings of up to $3.5 trillion relative to voluntary isolation.
• Mechanism: Under voluntary isolation, susceptible individuals reduce consumption (~3 hours/day less) and labor (~6 hours/day less at peak) due to higher infection risk from active infectious individuals, raising economy-wide costs. Targeted isolation reduces infectious individuals’ activity by similar amounts, enabling susceptible individuals to maintain work and consumption without increasing overall contacts or prevalence at activity sites.
• Trade-off frontier: Further reductions in cases beyond the targeted/voluntary isolation levels require sharply higher costs (nearly tenfold increase in economic losses to achieve minimal cases), indicating an optimized disease-economy frontier.
• Information frictions (testing): With limited and delayed testing (10% quality, 8-day lag), targeted isolation recovers only ~13% of the baseline economic gains and ~30% of infection control gains. With improving test quality (to 95% by day 75; 8-day lag), targeted isolation recovers ~92% of economic and ~94% of infection control gains. Reducing lags (to 5 days at day 60 and 3 days at day 75) adds little beyond improved quality.
• Compliance: With 0% compliance, targeted isolation is ineffective. With 75% compliance and perfect information, targeted isolation recovers just over ~76% of the baseline economic benefits and nearly all infection control benefits; with 75% compliance and improving information, roughly ~95% of both economic and infection control benefits are recovered.
• Blanket lockdowns vs targeted isolation: Blanket lockdowns deliver stronger short-term infection control but cause rebounds and deep economic contractions; targeted isolation and voluntary isolation dominate economically across scenarios, while blanket lockdowns may deliver temporary hospital load relief early when testing is poor.
• Robustness: Main conclusions persist across variations in asymptomatic prevalence (via productivity losses), shifts in contact allocation between consumption, labor, and unavoidable activities, and alternate contact function curvatures; these alter the magnitude of economic gains but not the relative disease outcomes.
• Additional quantitative notes: Compliance cost thresholds to overturn targeted isolation savings would need to be very high (on the order of $8,000–$20,000 per person) under many plausible lockdown designs; rebounds after blanket lockdowns approach nearly all cases averted during lockdown.

The findings demonstrate that the principal driver of pandemic-induced recessions is a coordination failure: infectious individuals do not internalize the risks they impose on susceptibles, leading the majority (susceptibles) to withdraw from economic activity under voluntary isolation. Targeted isolation resolves this failure by incentivizing or mandating infectious individuals to isolate, enabling susceptibles to work and consume with minimal additional disease burden. Thus, targeted isolation minimizes the disease-economy trade-off that characterizes voluntary isolation and blanket lockdowns. Blanket lockdowns, although capable of temporary case reductions and potential early relief for hospital systems when testing is poor, induce severe recessions and significant rebounds when lifted. The results underscore the importance of early, high-quality, widely available testing and compliance incentives to unlock most benefits of targeted isolation; test timeliness matters less than quality. The calibrated mapping from economic activities to contacts is central to explaining why targeted isolation can allow more economic activity without increasing infections. Policy implications include prioritizing strategies that internalize externalities from infectious individuals (e.g., paid isolation, compliance checks) and potentially prioritizing vaccines to individuals least able or likely to isolate when infectious to reduce systemic withdrawal by susceptibles.

The paper develops a tractable, calibrated epi-economic model linking individual economic behavior and contacts to epidemic dynamics, and uses it to evaluate voluntary isolation, blanket lockdowns, and targeted isolation. Targeted isolation emerges as an optimal strategy that maintains disease control comparable to voluntary isolation while avoiding deep recessions, averting up to 91–95% of individual economic losses and up to about $3.5 trillion in recessionary costs. The benefits are robust across parameter variations and can largely be realized even with partial compliance and improving testing quality. Blanket lockdowns may serve as complementary, time-limited tools early in a pandemic when tests are poor, with targeted isolation taking over as test quality improves to manage rebounds and sustain economic activity. Future research should refine incentive design for isolation, integrate richer heterogeneity (e.g., superspreading, sectoral work-from-home capacity), and improve data-driven mappings from mobility and social networks to contacts and transmission risks.

Key limitations include simplifying assumptions for tractability: baseline perfect information on health status and full compliance; limited heterogeneity (e.g., age, occupation, social network structure) beyond calibration; reliance on pre-pandemic contact matrices and simplified mapping that assumes consumers interact only with consumers and workers only with workers; baseline linear contact function with reduced-form alternatives; abstraction from detailed costs and logistics of testing systems (costs common across compared strategies) and from nuanced testing strategies; using productivity losses to represent asymptomatic/presymptomatic work capacity; and not explicitly modeling all kinds of interventions or cross-regional policy variations. These choices may affect quantitative magnitudes and generalizability though robustness checks suggest main qualitative conclusions persist.