FOUR STYLIZED FACTS ABOUT COVID-19

We develop a Bayesian method for estimating the dynamics of COVID-19 deaths and discover four key findings that expose the limitations of current structural epidemiological models . (i) Death growth rates declined rapidly from high levels during the initial 30 days of the epidemic worldwide. (ii) After this initial period, these rates fluctuated substantially around 0%. (iii) The cross-location standard deviation of death growth rates decreased rapidly in the first 10 days but remained high afterward. (iv) These insights apply to both effective reproduction numbers and their cross-location variability through epidemiological models. Our method is applicable to studying other epidemics.