Arbeitsbereich

Fertilität und Wohlbefinden

Auf einen Blick Projekte Publikationen Team

Projekt

Improvement and Evaluation of Fertility Forecasts

Mathias Lerch, Peng Li, Mikko Myrskylä, Christina Bohk-Ewald (MPIDR / University of Helsinki, Finnland)

Ausführliche Beschreibung

Forecasts of completed fertility predict how many children will be born on average to each woman over her entire reproductive lifetime. The forecasts serve as a basis for further inquiries, e.g., to produce population forecasts and to help decision-makers devise social security plans. A literature survey on the forecasting methods identified more than 20 major methods and approximately 180 variants. In light of this enormous methodological variety, today's forecasters face a key question: Which method should they use to produce accurate and reliable fertility forecasts?

To answer this question, we systematically evaluated all major methods, using full-data cross-validation for the first time ever. That is, we assessed the overall accuracy of the methods and uncertainty estimates (if provided) across as many forecast periods as possible, using the fertility data of all birth cohorts and countries available in the Human Fertility Database and in the United Nations World Population Prospect Estimates database. Accounting for so many fertility levels, shapes, and trends by cross-validation ensures valid and robust evaluation results.

Looking at the accuracy of fertility rate forecasts, we have found that the baseline method of Freeze rates is outperformed by only four methods, which are more complex. Furthermore, many probabilistic methods tend to underestimate forecast uncertainty. Evaluating the forecast performance of the current methodology not only reveals the best methods, it also helps us to detect unresolved issues. We aim to overcome these issues by revising the existing fertility forecasting methods and developing new ones.

Following good scientific practice and to allow for the reproducibility of results, the main techniques adopted for forecasting fertility and assessing forecast performance have been published in R-packages.

Schlagworte:

Geburtenentwicklung, Statistik und Mathematik

Publikationen

Bohk-Ewald, C.; Li, P.; Myrskylä, M.:
Proceedings of the National Academy of Sciences of the United States of America 115:37, 9187–9192. (2018)    
Das Max-Planck-Institut für demografische Forschung (MPIDR) in Rostock ist eines der international führenden Zentren für Bevölkerungswissenschaft. Es gehört zur Max-Planck-Gesellschaft, einer der weltweit renommiertesten Forschungsgemeinschaften.