The impact of mass gatherings on the local transmission of COVID-19 and the implications for social distancing policies: Evidence from Hong Kong

The study investigates whether mass gatherings following the relaxation of Hong Kong’s Prohibition on Group Gathering Regulation (POGGR) in June 2020 contributed to a resurgence of COVID-19 infections (the third wave). The context includes global evidence that mass gatherings facilitate transmission of respiratory diseases and WHO guidance that even medium-sized events can elevate risk. In Hong Kong, case numbers had stabilized through late June 2020, after which restrictions were relaxed (group size limits raised from 4 to 8 and then to 50), coinciding with demonstrations related to the National Security Law (NSL). The research question is whether these large public gatherings increased local transmission and what implications this has for social distancing policy. The study is important for informing pandemic control strategies when considering easing restrictions on gatherings.

Prior work links mass gatherings with increased transmission of respiratory infections (influenza, measles, meningitis) and emphasizes event and venue characteristics in shaping risk (indoor vs outdoor, density, duration). During COVID-19, evidence on small indoor gatherings’ role is substantial, but rigorous assessments of large-scale gatherings are limited. Notable cases include the Sri Petaling Mosque event in Malaysia and festival-related gatherings in Borriana, Spain, which were associated with outbreaks, while some US studies suggested protest-related increases in non-participants’ stay-at-home behavior could offset transmission among participants. Broader literature highlights roles of mobility, socioeconomic, demographic, and climatic factors in epidemic spread. WHO guidance recommends suspending or strictly managing mass gatherings. This study addresses the empirical gap by evaluating large outdoor public demonstrations’ impact on local COVID-19 transmission in Hong Kong.

Design: Comparative case study using the Synthetic Control Method (SCM) to construct a counterfactual “synthetic Hong Kong” from a pool of over 200 Chinese cities (mainland and Macau) that did not experience the same mass gathering events. Intervention and periods: Pre-intervention (control) period from May 1 to June 19, 2020. The primary intervention date is June 20, 2020 (start of mass public gatherings/demonstrations after relaxation of POGGR). Robustness tests shift the intervention to each day from June 21 to June 30 to account for incubation, testing, and reporting lags. Outcome variables: Daily new confirmed cases and total cumulative confirmed cases of COVID-19. Predictors/covariates: Epidemiological variables (14-day pre-intervention averages and sums: average total cases, average total cases per 10,000 population, sum of daily new cases), demographic and socioeconomic indicators (population, households, population density, GDP, GDP per capita, number of hospitals, doctors and hospital beds per 10,000), and climatic parameters (daily average temperature, relative humidity, wind speed, air quality index). Data sources: City-level COVID-19 epidemiological data from China Data Lab (Harvard Dataverse); demographic/socioeconomic data from China City Statistical Yearbook (2019; reflecting 2018) and official yearbooks for Hong Kong and Macau; meteorological data from the China Meteorological Data Service Centre (hourly station records, aggregated to daily via Empirical Bayesian Kriging interpolation in ArcGIS), Hong Kong Observatory, and Macau Meteorological and Geophysical Bureau; AQI data from Harvard Dataverse for mainland cities and the World Air Quality Index project for Hong Kong and Macau. SCM specification: The method estimates weights on donor cities to match Hong Kong’s pre-intervention outcomes and predictors, minimizing mean squared prediction error. The model assumes the untreated outcome follows a factor model with observed predictors and unobserved common factors and factor loadings. The treatment effect at time t is the difference between actual Hong Kong outcomes and the weighted sum of donor outcomes post-intervention. Donor pool: 282 cities after data cleaning. Two main models are estimated: Model 1 (daily new cases) and Model 2 (total cases). Robustness checks vary intervention dates to accommodate incubation/testing delays. Confounding control: Small gatherings allowed after POGGR relaxation (≤50 people) could confound effects; donors generally maintained stricter mass gathering restrictions (often prohibiting >200 people), reducing this confounding. Extended pre-intervention matching and inclusion of climate/socioeconomic predictors address time-invariant differences and other confounders. Recognized limitations include unmeasured socio-political differences between Hong Kong and mainland cities. Hypothesis: Mass gatherings following relaxation of POGGR materially increased local COVID-19 transmission in Hong Kong, contributing to the third wave beginning in early July 2020.

- During June 20–July 31, 2020, the mean daily growth rate of total cases in Hong Kong was 2.63%, versus 0.07% in the synthetic control. - Model 1 (daily new cases) indicates that in the absence of mass gatherings, Hong Kong would have had no more than 2 new cases per day from June 20 to end of July. Predicted sum of new cases: 16 in the first 14 days; 23 from June 20–July 31. Actual new cases: 120 in first 14 days; 2,145 from June 20–July 31. - Model 2 (total cases) shows cumulative cases would have been about 1,128 by July 31 without mass gatherings, compared to actual 3,272 (increase of 2,144 during the period vs 32 in synthetic control). - Baseline intervention date (June 20): Mass gatherings increased new infections by 62 over 10 days (87.58% of total new cases) and by 737 over 30 days (97.23%). - Robustness checks with shifted intervention dates: - Shorter lags (1–5 days): average +57 new infections over 10 days (75% of total), and +961 over 30 days (97.66%). - Longer lags (6–10 days): average +101 over 10 days (91.82%), and +1,553 over 30 days (99.36%). - Visual comparisons show divergence between actual and synthetic trajectories after late June, with synthetic trends stabilizing at very low levels while actual infections accelerated sharply. - Overall, results consistently indicate that public gatherings/demonstrations following relaxation of POGGR substantially amplified local transmission.

Findings indicate that mass public gatherings after relaxation of Hong Kong’s POGGR significantly accelerated COVID-19 transmission, contributing to the city’s third wave beginning in early July 2020. Despite mask use among participants, high crowding and close contact likely increased exposure risk. While some have argued that quarantine exemptions for specific overseas arrivals (e.g., seafarers) could have seeded cases, over 83% of cases during the study period were local, and a high fraction had unknown sources, underscoring community spread dynamics. The synthetic control analysis, controlling for socioeconomic and climatic differences, demonstrates that without mass gatherings, daily new cases would have remained very low, and cumulative growth minimal. The findings align with local expert views implicating increased social interactions around the Dragon Boat Festival (June 25) and July 1 holiday. Policy-wise, results support the importance of maintaining or promptly reinstating restrictions on group gatherings when community transmission risk persists, minimizing the potential for superimposed event-driven surges. The study provides empirical support for social distancing policies targeting large gatherings as an essential component of outbreak control, with relevance for jurisdictions considering relaxation of such measures.

The study provides empirical evidence that mass public gatherings in Hong Kong following relaxation of group gathering limits materially increased local COVID-19 transmission. Using SCM, the analysis shows markedly higher growth in both daily new and cumulative cases compared to a well-matched counterfactual. Contributions include: (1) complementing literature focused on indoor small gatherings by quantifying impacts of large outdoor public events; (2) informing policy on the necessity of sustained restrictions on mass/group gatherings amid ongoing transmission; and (3) offering insights generalizable to managing mass gatherings during early outbreaks of future novel infectious diseases. Policymakers should prioritize consistent, clearly communicated social distancing measures, especially POGGR-type rules, and adjust them promptly in response to risk assessments. Future research could incorporate behavioral and political context variables, more granular mobility/contact data, and extend to later periods with variant circulation and higher population immunity to assess external validity across epidemic phases.

- Donor-control differences: Socio-political and institutional differences between Hong Kong and mainland donor cities (governance, public trust, compliance) are time-invariant factors not fully captured in SCM, potentially biasing estimates. - Aggregated data: City-level aggregation limits assessment of within-city heterogeneity and specific transmission chains; many cases had unknown sources. - Concurrent changes: Relaxation of other social activities (e.g., dining, shopping) occurred alongside mass gatherings; while donor policies help mitigate confounding, residual confounding may remain. - Importation and exemptions: Possible contributions from quarantine-exempt arrivals (e.g., seafarers) are difficult to fully disentangle from community-driven spread. - Temporal scope: Study period ends July 2020 and does not include later SARS-CoV-2 variants (e.g., Alpha) or contexts with substantial acquired immunity, limiting generalizability to later pandemic stages. - Measurement lags: Incubation, testing, and reporting delays are addressed via robustness checks but may still affect precise timing of estimated effects.