There is a really nice post over at Dean Dad about financial literacy. The part that I really liked was the issue of stock market returns. In the post he discusses how planning tools often assumed an 8% return while his own IRA has performed nominally negative. While it is possible that there could be a growth surge that makes an 8% return happen over a 30 year period, it has got to be one heck of a surge to cover up 12 years of negative nominal returns.
This is actually a more general principle and one that we should keep in mind as researchers -- for many phenomena that we study, there is limited data as to what could happen in the future. Extrapolating disease rates, for example, is hard based on current rates (especially for infectious diseases where exponential growth can matter). You also have subtle issues like vaccines and herd immunity (or, in a similar sense, behavioral firebreaks such as careful hand washing). Very small changes in circumstances can make it impossible to predict the future.
So my own lesson from this anecdote is that we need to be very skeptical about the assumptions that go into forecasting. We want to do it (because some information is better than no information) but we need to be ready to revise assumptions as data comes in.
Semiparametric efficiency and nearly-true models
2 hours ago