12 April 2020

Acting on Forecasts Rather than Proof

The forecasts for COVID-19 deaths are falling. That's wonderful news. It turns out that measuring the cost for exponential growth of a virus has something in common with measuring the value of startups. It is subject to error but can still inform you how to act.

One key lesson is to move first and move fast. By the time you have data proving the value of a startup, you pay far more for it. Similarly, by the time you have data proving the severity of a virus, you pay far more for it.

Number of deaths thru 11-Apr:
San Francisco: 14
New York: 6,898

"[San Francisco mayor London] Breed ordered businesses closed and issued a citywide shelter-in-place policy effective on March 17, at a point when San Francisco had fewer than 50 confirmed coronavirus cases. (California Governor Gavin Newsom followed with a similar statewide order 19 March.) On that date, New York City already had more than 2,000 positive cases. But New York Governor Andrew Cuomo and New York City Mayor Bill de Blasio, reluctant either to shutter schools or issue a stay-at-home directive for the nation’s largest city, didn’t take similar action for several days. By the time New York City fully shut down on March 22, more than 10,000 cases were reported across its five boroughs." [From the Atlantic, "The City That Has Flattened the Coronavirus Curve"]

5 days can make a big difference when facing a virus that spreads or contracts exponentially.

I see people wondering how stock prices can go up when we're still in a pandemic. Stock prices represent an attempt to price an endless stream of future profits. It is true that the Dow is up 30% in the last couple of weeks. It is also true that it is still down 25% from its peak a couple of months ago. Stock prices fluctuate because people are trying to estimate something in the future that is continually changing as events and best estimate methodologies change. Investors still agree that the coronavirus and measures to protect against it have lowered the value of future profits; their margin of error in estimating that means that stock prices are going to fluctuate. A lot.

I see people dismissing the models forecasting coronavirus deaths as being wrong. Models are always wrong but that doesn't mean that they aren't helpful. One catch-22 with models and policy is that the group taken least seriously could be proven most accurate. What do I mean? Let's say that we had ignored the coronavirus warnings from experts and continued as normal - never socially distancing and not changing anything. In that scenario, New York would be a best case for cities and fatalities would be multiples of what they are now. We could easily have 1 million deaths rather than 100,000. The best-case forecasts would now seem tragically naive. On the flip side, if we listened to those who warned of the worst and took serious measures to protect against that, moving fast and dramatically, the worst-case forecasts would now seem morbidly pessimistic. We would have far less than 100,000 dead and would never come close to a million. Even without behavior change, forecasts of anything that grows or contracts exponentially are likely to be off. Hugely. The outcome could easily be 10X or 1,000X better or worse given just small changes in the rate of contagion or mortality.

Early investors in Apple likely never once stopped to think that it could be worth a trillion early the next century. But they didn't have to know it would be worth that much to know it was a good investment. Given the Bay Area is the epicenter for trying to value the future, it is unsurprising that it would become a model for how to minimize the harm of a virus. Among the many things the folks in the Bay Area have learned is that it is better to move first to pursue a possibility - whether that possibility is avoiding fatalities from a pandemic or owning shares of a startup that later make you rich - and then learn from and adapt to reality than it is to wait for the data to come in and by then to have missed your opportunity.

As the future comes at us with increasing speed, the ability to quickly assess what models suggest rather than what data confirms could make all the difference.

No comments: