Abstract
This paper established a Susceptible-Infected-Recovered-Dead (SIRD) model to fit and predict the COVID-19 epidemic in China. On the premise of passing the Markov Chain Monte Carlo (MCMC) robustness check, we counterfactually assessed the effectiveness of partial unlocking and vaccination policies. Our key findings are as follows: (1) The SIRD model predicted that China could fully contain COVID-19 in July, 2020, which is consistent with reality; (2) If the strict social distancing policies are relaxed before the epidemic ends, it will lead to a second rebound. Even worse, the rebound increases with the advance of relaxation time and relaxation degree; (3) Timely vaccination is the best strategy if it is available. However, delayed vaccination is much less effective. (4) When further taking natural population growth rates into account, we find that the principal conclusions remain robust. The inclusion of demographic dynamics helps mitigate potential fitting bias resulting from changes in population size over extended epidemic periods like the COVID-19 pandemic. Empirically, the model shows a certain degree of applicability in regions with similar transmission trends. This paper theoretically and quantitatively supports the view that the effectiveness of China's comprehensive prevention and control measures, and empirically verifies that the model has potential reference significance for some regions with similar transmission trends.