Forecasting Long-term Wuhan Coronavirus Cases
A simple methodology with a longer forecast horizon
After seeing the impact to markets on Monday, I posted my first thoughts on the Coronavirus the next morning. Then, yesterday, I offered a very basic forecast for confirmed cases from the World Health Organization (“WHO”).
Thus far, the forecast has done well but seems to be undershooting reality, predicting 5,676 cases for WHO report 9 versus 6,065 actual. It is predicting 7,400 for report 10 but already one news outlet tracker is reporting 7,815 cases.
I decided to not make any adjustments to yesterday’s forecast but also recognize that the Wuhan Coronavirus cannot grow exponentially forever. So, I’d like to offer another simple approach that can provide a forecast for the long-term development of the virus:
Like the exponential forecast, this one is not very technical. I stand by my rationale in the previous post for why I want to keep things simple - it’s very early and there are so many unknowns. However, a forecast is still a useful tool for benchmarking. For example, now that we have seen actual cases exceed our forecast for two days, the virus should warrant some extra attention for exceeding expectations.
This forecast has a really long time frame, so the assumptions are incredibly sensitive. Change parameters a little bit and it’s the difference of tens of thousands or, theoretically, even millions.
What is valuable about this prediction is the shape. It’s generally something we should expect to see as the situation develops. There is a period of exponential growth and then an inflection and a long tapering. I achieved that by assuming a flat, slightly negative assumption for the acceleration of the growth rate.
Under this forecast, the next few days develop a little more slowly than the exponential forecast from yesterday:
January 30 - 9,842
January 31 - 12,139
February 1 - 14,690
February 2 - 17,467
February 3 - 20,438
Now, we can keep a close watch on the pace of change between the number of cases reported each day. That makes two scenarios that have been forecasted, which provides even more benchmarking capability. Let’s continue to hope the actual results are less than anything presented here.
Enjoy the article? If you haven’t already, subscribe for free below and never miss a post. No spam. Cancel whenever!