5 Comments
User's avatar
Lyman Stone's avatar

You have not read https://www.pnas.org/doi/10.1073/pnas.1722364115 ???

Expand full comment
Inquisitive Bird's avatar

No, I haven't. Thanks for the link

Expand full comment
Max More's avatar

Just reading the abstract, this appears to reinforce the claim by forecasting expert J. Scott Armstrong about the power of simple methods in many cases. (I highly recommend The Principles of Forecasting.)

Expand full comment
Emil O. W. Kirkegaard's avatar

Why not backtest the method on older cohort data? Presumably you can find the age specific fertility rates for some countries for the past and compare to the cohort fertility rate.

Expand full comment
Inquisitive Bird's avatar

Though I didn't discuss it at all, I actually did backtest it -- just to be sure. The predictions were in general very accurate when tested on real data at this time horizon (up to 15 years forward). For example, when testing on the Danish time series for the years with complete data, all predictions were less than 0.1 off (e.g., if the predicted fertility was 1.8, the true fertility would never be smaller than 1.7 or greater than 1.9).

Expand full comment