Global Empirical Forecasts of COVID-19 Trajectories Under Limited Information on the Efficacy of Intervention Strategies

11 Pages Posted: 2 Apr 2020

See all articles by Kai Lin

Kai Lin

Coolabah Capital

Christopher Joye

Coolabah Capital

Nathan Giang

Coolabah Capital

Adam Richardson

Coolabah Capital

Date Written: April 2, 2020

Abstract

As the novel coronavirus and its associated disease COVID-19 started to rapidly transmit around the world in early 2020, the financial, social and health impacts represented a 1-in-100 year shock, the likes of which had not been observed since the last global pandemic in 1918 and the Great Depression in 1929. A key question for policymakers, medical researchers, and financial market participants was how the disease would propagate in an environment in which it was left unconstrained as compared with preferable alternatives where nation states implemented assertive efforts to mitigate the disease’s adverse effects. Medical researchers seeking to advise governments produced theoretical forecasting models, drawing on the epidemiological literature, which have often been too inflexible and abstract for use by financial markets. For this niche user group, empirical, agile, and intervention-aware forecasting methods are paramount, especially those that can accommodate the subjective judgements of different users. This paper outlines two such empirical forecasting frameworks for the daily confirmed case counts, eventual case counts, and time to peak daily new case counts for major countries. The first framework uses a linear mixed effect model for the case growth rate, accounting for the presence of intervention measures and idiosyncrasies of individual countries. The second framework allows users to forecast the case trends of a target country by substituting in the observed effects of interventions from qualitatively similar countries with customisable calibrations to reflect lower efficacies. Combined, these two frameworks are especially useful in the early days of the outbreak, when the effects of different countries’ imminent interventions have not yet shown up in observed data, but which can be inferred from similar countries further along their intervention path. When first applied and published on March 23, these models projected the peak in daily new COVID-19 case counts for the US and Australia would arrive in early-to-mid April 2020. To the best of our knowledge, this was one of the first early-to-mid April peak projections published globally. Whilst not theoretically founded in the mechanisms of infectious disease, such empirical forecast frameworks offer versatile and parsimonious projections for financial market participants seeking to make decisions under conditions of uncertainty apropos the efficacies of different intervention measures around the world.

Keywords: Coronavirus, COVID-19, Forecasts, Financial Markets

Suggested Citation

Lin, Kai and Joye, Christopher and Giang, Nathan and Richardson, Adam, Global Empirical Forecasts of COVID-19 Trajectories Under Limited Information on the Efficacy of Intervention Strategies (April 2, 2020). Available at SSRN: https://ssrn.com/abstract=3566596 or http://dx.doi.org/10.2139/ssrn.3566596

Kai Lin (Contact Author)

Coolabah Capital ( email )

Suite 2502 of Level 25, Westfield Tower 2,
101 Grafton Street, Bondi Junction
Sydney, New South Wales 2022
Australia

HOME PAGE: http://coolabahcapital.com/

Christopher Joye

Coolabah Capital ( email )

Suite 2502 of Level 25, Westfield Tower 2,
101 Grafton Street, Bondi Junction
Sydney, New South Wales 2022
Australia

HOME PAGE: http://coolabahcapital.com/

Nathan Giang

Coolabah Capital ( email )

Suite 2502 of Level 25, Westfield Tower 2,
101 Grafton Street, Bondi Junction
Sydney, New South Wales 2022
Australia

Adam Richardson

Coolabah Capital ( email )

Suite 2502 of Level 25, Westfield Tower 2,
101 Grafton Street, Bondi Junction
Sydney, New South Wales 2022
Australia

Here is the Coronavirus
related research on SSRN

Paper statistics

Downloads
1,280
Abstract Views
3,997
rank
16,892
PlumX Metrics