Economic losses from COVID-19 cases in the Philippines: a dynamic model of health and economic policy trade-offs

The paper addresses how to balance economic recovery with public health safeguards during COVID-19 in the Philippines. By May 22, 2021, the country recorded over 1.17 million cases, and GDP contracted by 9.6% in 2020. Despite gradual easing of restrictions, surges (e.g., March 2021 in NCR) strained hospital capacity, prompting tighter lockdowns. Policymakers consider reopening (businesses, transport, age restrictions) alongside ensuring hospital capacity. The study extends the FASSSTER platform’s subnational SEIR model to quantify economic losses from infection and lockdowns and evaluates policy options subject to maintaining HCUR below 70%. It examines trade-offs over a 3‑month horizon across regions to inform calibrated reopening while safeguarding health system capacity.

The review situates pandemic economics within disaster economics, highlighting direct labor supply shocks (mortality, infection), indirect productivity and welfare impacts, and macroeconomic effects. Static SAM/CGE and microsimulation studies (e.g., H1N1, COVID-19 for Ghana, UK, global poverty) show substantial welfare and GDP losses, with prevention-related labor losses often exceeding health-related GDP losses, but they do not explicitly analyze dynamic health–economy trade-offs. Dynamic epidemiological models (SIR/SEIR) enable forward-looking policy analysis; integrations with economic modules show trade-offs between productivity and disease spread. Optimal control studies suggest targeted lockdowns minimize losses while saving lives, though many focus on fatalities rather than hospitalizations. Gaps remain for low- and middle-income contexts, subnational analysis, and explicit trade-offs with health system capacity. The paper fills these gaps by extending a regional SEIR with economic loss equations and constructing a policy trade-off tied to HCUR thresholds in the Philippines.

Model framework: The FASSSTER subnational SEIR model partitions the population into compartments: Susceptible (S), Exposed (E), Infectious asymptomatic (Ia), Infectious symptomatic (Is), Confirmed (C), and Recovered (R). Transitions are governed by effective transmission β=β0(1−λ), incubation and progression rates, detection rates (δa, δs), recovery and disease-induced death rates (εa, ε), and natural birth/death. Health system capacity (HSC) parameters reflect the prevent-detect-isolate-treat-reintegrate (PDITR) capability via detection and isolation rates. Recovered are assumed immune over the horizon. Economic extensions: Two differential equations quantify losses: - Infection-related losses YE(t): present value of foregone future gross value added (GVA) from COVID-19 deaths across age groups and contemporaneous output lost from isolation/hospitalization of infected and confirmed cases, parameterized by age-specific remaining productive years, social discount rate, employment-to-population ratio κ, and daily GVA per worker ω. This simplifies to YE(t)=L1·Ia(t)+L2·C(t), aggregating deaths and isolation terms. - Lockdown-related losses YL(t): foregone income from employed persons displaced by restrictions among the non-active-case population, YL(t)=κ·ω·φ·[S(t)+E(t)+Ia(t)+R(t)], where φ is a region- and period-specific displacement rate derived from sectoral operating capacities. Economic parameters: κ and daily GVA per worker (ω from annual z) are computed from Philippine Statistics Authority labor force, population, and national accounts by region and sector. Age-specific death distributions use DOH-EB data; the social discount rate follows NEDA. Displacement rates φ are region-specific and vary over reopening phases based on DTI operational capacity circulars; weights from labor force survey microdata yield regional φ. Policy trade-off: Minimize total losses ∫(YE+YL)dt over a 3‑month horizon subject to average HCUR ≤ 70% (ICU/bed utilization threshold guiding policy). Lockdown scenarios are combinations of four policy levels (Level IV to I) implemented monthly. Data and estimation: COVID-19 line list data (dates, PSGC location, case counts) from DOH are preprocessed with multiple imputation (predictive mean matching via R mice) for missing onset, specimen, admission, result, and recovery dates. The SEIR ODEs are solved using deSolve; λ (mobility/restrictions adherence) is fitted via L-BFGS-B (optim) with MSE, using parallel optimization (optimParallel) and bootstrap runs per region until correlation ≥90%, at least 50 iterations. Model outputs provide compartment trajectories to evaluate YE and YL under each scenario. Scenarios and regions: Eight 3‑month lockdown sequences per region (Table 2), simulated for NCR and four regions (Ilocos, Western Visayas, Soccsksargen, Davao). For NCR, two HSC settings are used (≈17.99% and 21.93%); for other regions, analogous lower/higher HSC settings (e.g., 16%/22%). Outputs include total losses as % of 2019 GRDP and average HCUR over the quarter.

- National Capital Region (NCR): A clear trade-off curve (parabolic) emerges between average HCUR and total 3‑month economic losses. Under lower HSC (17.99%), losses range from 16.58% (strictest) to 12.19% of GRDP as restrictions loosen, before rising again with further easing. Under higher HSC (21.93%), both average HCUR and losses are generally lower; scenarios 1–4 stay below the 70% HCUR threshold, with the lowest loss at 9.11% of GRDP. The abstract reports that the minimum cumulative loss under the HCUR constraint is 10.66% of GRDP, evidencing a policy trade-off in NCR. - Composition in NCR: As measures loosen, infection-related losses (YE) increase while lockdown losses (YL) decrease; at lower HSC, YE can comprise about half of total losses in looser scenarios (e.g., Scenario 7A). With higher HSC, YE rises more slowly and does not overtake YL, tempering total loss increases during reopening. - Regions outside NCR (Ilocos, Western Visayas, Soccsksargen, Davao): No parabolic trade-off is observed. Total losses generally decrease monotonically as restrictions ease, while average HCUR remains below critical levels, especially under higher HSC. In Davao, YE remains low even as restrictions ease, and YL steadily declines, yielding decreasing total losses. - Effect of health system capacity: Across all regions, higher HSC reduces both economic losses and HCUR. Improved PDITR (testing, detection, isolation, treatment) broadens the feasible policy space for easing restrictions without breaching HCUR thresholds.

The findings address the policy question of balancing economic reopening with health system constraints. In NCR, countervailing forces between infection-related losses (rising with mobility) and lockdown-related losses (falling with reopening) produce a trade-off curve; thus, minimizing losses subject to HCUR ≤70% yields a constrained optimum, consistent with hospital capacity being the binding policy constraint. Enhancing HSC flattens the infection surge, lowers HCUR, and dampens the marginal increase in YE during reopening, enabling lower total losses while meeting the capacity constraint. In contrast, in regions outside NCR, easing restrictions reduces total losses without threatening capacity, implying that stringent lockdowns are not warranted given epidemiological conditions; targeted small-area lockdowns and strengthened PDITR suffice. Overall, investments that improve HSC expand the policy frontier, allowing safer and more economically favorable reopening trajectories.

The study extends a subnational SEIR model with economic loss equations for infections and lockdowns to quantify health–economy trade-offs under HCUR constraints across Philippine regions. Simulations over eight lockdown sequences show a tight trade-off in NCR—minimizing losses while keeping HCUR ≤70%—but not in regions far from NCR, which enjoy wider reopening space. Strengthening health system capacity (PDITR) consistently lowers losses and utilization, broadening policy space. Policy design should be region-specific, considering local case dynamics, economic structure, mobility patterns, and health capacity. Given parameter sensitivity and evolving epidemic conditions (e.g., variants, behavioral changes), projections should be updated frequently and limited to short horizons (1–2 months). The methodology provides a tool for policymakers to evaluate economic impacts of lockdown policies and navigate health–economy trade-offs during COVID-19 and future epidemics.

Model outputs depend on timely and accurate parameter updates (e.g., transmission, detection, isolation rates) and data quality; behavioral changes and emerging variants can shift dynamics, causing divergence from projections. HSC and displacement rate estimates are approximations based on available administrative data and sectoral operating capacities. The assumed immunity of recovered individuals and other structural simplifications may not capture reinfections or heterogeneous contact patterns. Results are scenario-based over a limited 3‑month horizon and should be interpreted as indicative rather than predictive; authors recommend shorter (1–2 month) projection windows and regular recalibration.