How Well Can Period Fertility Measures Help to Forecast Cohort Fertility Measures? Cross-Country Evidence
P. C. Roger Cheng, National Central University, Taiwan
Eric S. Lin, National Tsing Hua University, Taiwan
In our previous work, we propose a smoothed version of the tempo-variance-adjusted measure developed by Kohler and Philipov (2001) as a substitute for the cohort total fertility rate (CTFR), and find that our proposed method out-performs other popular ones in forecasting CTFRs. In this paper, we apply our proposed method to fertility data from 17 European countries, the U.S., and Japan, mark every 25 consecutive years as a sample, and carry out testing of predictive power for a few competing models to all cohorts covered in each sample period. Models such as the Naive method, the Evans method, the linearized Gompertz model, the primitive version and the smoothed versions (with various smoothing parameters) of period fertility measures will be included and compared, based on their approximation to the actual CTFR values. Distributions of forecast errors will be investigated and some useful conclusions will be derived from the comprehensive picture.