The susceptibility of fertility measures to random fluctuations

Yen-hsin Alice Cheng, Academia Sinica
Kathryn Grace, University of California, Santa Barbara
Joshua R. Goldstein, Max Planck Institute for Demographic Research

The total fertility rate (TFR) is the standard measure used to compare the fertility levels of populations over time and across space. This measure has been used to document the recent rapid fertility decline in developing countries and very low TFRs in some European countries have led researchers to explore the phenomenon of “lowest-low” fertility. Arguably, however, the TFR has significant limitations, primarily a sensitivity to an increase (or decrease) in age at childbearing, that may inaccurately present a deflated (or inflated) value. The Bongaarts-Feeney tempo adjustment technique (where the TFR is adjusted based on the change on the mean age at childbearing) has therefore become an important tool in the demographer’s toolbox. In an effort to increase the generalizability of the TFR adjustment technique by taking into consideration age specific variation in fertility postponement, Kohler and Philipov in their 2001 article, developed an alternative strategy for adjustment. The Kohler Philipov technique incorporates variation in postponement – age-period interaction – but reduces to the Bongaarts Feeney approach if the timing changes are assumed to be invariant. Each measure, traditional TFR, BF- adjusted TFR and the KP-adjusted TFR are nearly always presented as fixed values (as opposed to random variables where variation is present). However in some cases there are only small differences in the adjusted measures which may be the result of random variation. The purpose of this research is to evaluate the random variation found in the TFR and the mean age at birth, and the resulting variation in the two adjusted measures, BF-TFR and KP-TFR. Analysis will be conducted using data from several different populations that have experienced notable changes in period TFR. The results will highlight the behavior of the TFR and the adjusted measures with consideration to random variability.