Bias-free Forecast-driven Guidelines for Opening Pandemic-ravaged Economies
13 Pages Posted: 28 Jul 2020 Last revised: 30 Jul 2020
Date Written: July 24, 2020
Abstract
Official guidelines for opening America after COVID-19 lockdowns are backward-looking. For example, they require a two-week downward trajectory of documented cases, without mentioning their statistical significance. We propose a forward-looking strategy using out-of-sample forecasts of deaths, after correcting for the biased data arising from a lack of randomized testing over the general population. We correct the so-called `selection bias' using the inverse Mills ratio neglected by epidemiologists. This paper updates and supplements three earlier papers. Here we adjust the preliminary forecasts by using an autoregressive distributed lag ARDL(1,1) model. The adjustment method is claimed to be relatively new with potential applications in improving disaggregate forecasts from Poisson link generalized linear model (GLM) for death counts. Our illustration forecasts deaths for July 27 in the hope that our state-by-state forecasts with maximum entropy confidence intervals will help local governors and mayors in their opening up decisions. A section adds forecasts for the week ending Aug.3.
Note: Funding: None to declare.
Declaration of Interest: None to declare.
Keywords: selection bias, inverse Mills ratio, forecasting new deaths
JEL Classification: C10, C33, I18
Suggested Citation: Suggested Citation