Time series models for epidemics: leading indicators, control groups and policy assessment

Abstract 

This article shows how new time series models can be used to track the progress of an epidemic, forecast key variables and evaluate the effects of policies. A class of univariate time series models was developed by Harvey and Kattuman (2020). Here the framework is extended to modelling the relationship between two or more series. The role of common trends is discussed, and it is shown that when there is balanced growth in the logarithms of the growth rates of the cumulated series, simple regression models can be used to forecast using leading indicators. Data on daily deaths from Covid-19 in Italy and the UK provides an example. When growth is not balanced, the model can be extended by including a stochastic trend: the viability of this model is investigated by examining the relationship between new cases and deaths in the Florida second wave of summer 2020. The balanced growth framework is then used as the basis for policy evaluation by showing how some variables can serve as control groups for a target variable. This approach is used to investigate the consequences of Sweden's soft lockdown coronavirus policy.