(joint project with Peter Spencer)

Modeling the Covid-19 Epidemic Using Time Series Econometrics

  • The figure shows the forecast of daily deaths for the UK and the US using the ECDC data.

  • The forecast (dashed line) is based on our preferred gamma model that accounts for a weekend effect; the squares denote the weekend observations.

  • The gray area shows the 95% simulation-based confidence interval.

  • The history of forecasts (updated weekly) is available here.

  • There are many ways of analyzing the progress of an epidemic, but when it comes to short term forecasting, it is very hard to beat a simple time series regression model. These are good at allowing for the noise in day to day observations, extracting the trend and projecting it forward.

  • The classic `logistic' model has provided a realistic model of the behavior of Covid-19 in China and many East Asian countries. Once these countries passed the peak, the daily case count fell back, mirroring its initial climb in a symmetric way, just as the classic model predicts.

  • However, in Italy and Spain, and now the UK and many other Western countries, the experience has been very different. The daily count has fallen back gradually from the peak but remained stubbornly high. The reason for the divergence from the classical model remain unclear.

  • We take an empirical stance on this issue and develop a model that is based upon the statistical characteristics of the time series. With the possible exception of China, the workhorse logistic model is decisively rejected against more flexible alternatives.

  • Our regression models are designed to exploit this, using the daily statistics released by ECDC.

  • Our most current version of the paper is available here: Modeling the Covid-19 Epidemic Using Time Series Econometrics.

  • Our first report on the development of coronavirus in the UK is available here: Coronametrics: The UK Turns the Corner.