Dissertation/Thesis Abstract

Evaluation of UN population forecasts and effect of urbanization on forecast accuracy
by Zhao, Fengmin, Ph.D., The Johns Hopkins University, 2010, 477; 3428575
Abstract (Summary)

Background. Even though the United Nations (UN) is the leading agency that systematically and consistently provides population forecasts, the accuracy of the UN forecasts at the country level is unclear. It is a well known fact that there is large gap between urban areas and rural areas in both fertility and mortality levels in developing countries, which makes the population heterogeneous by rural/urban residence. Rapid urbanization has been taking place in almost all developing countries in past few decades, which changes the composition of the population by residence, and consequently affects the national fertility and mortality levels. The cohort-component method the UN used for its forecasts does not explicitly take urbanization into account. This dissertation aims: (1) To evaluate the accuracy of the UN past population forecasts at the country level; (2) To explore how urbanization affects accuracy in national population size forecasts.

Method. The UN forecasts between 1968 and 1998 were included in the study, and 176 countries with all the forecasts were analyzed. The estimates in the 2008 UN forecast were taken as the "true" values for the evaluation. Six measures for forecast accuracy were defined and calculated for individual countries. The mean errors averaged across forecast rounds were calculated for each individual country to evaluate forecast accuracy at the country level. The mean errors averaged across countries were calculated for each forecast round to evaluate forecast accuracy at forecast level. Age-Period-Cohort cross-classified models were used to evaluate whether forecast accuracy has improved over time and to test whether urbanization speed was significantly associated with forecast errors. Simulation studies were conducted to compare forecast accuracy between the standard cohort-component method, stratified forecast method and multistate forecast method.

Results. The forecast errors were pretty large in some countries, and both over-estimation and under-estimation existed. Absolute forecast errors were almost linearly increasing with forecast duration in all 12 forecast rounds. The mean percent errors and mean adjusted percent errors were positive for all 12 forecast rounds, and the population and population growth rate were on average over-estimated in the UN forecasts. Between the 1968 and 1998 forecasts, there was an overall decreasing trend in the errors in national population size, TFR and life expectancy at birth forecasts, but no significant improvement in forecasting population growth rate and paces of changes of TFR and life expectancy at birth forecasts. Urbanization speed was positively associated with errors in national population size forecasts and population growth rate forecasts, adjusting for other factors. The standard cohort-component forecast method over-estimated the population size compared to the stratified and the multistate forecast methods when there was internal migration.

Conclusion. The UN forecasts have on average over-estimated the population growth rate, hence the national population size. Lack of consideration of urbanization in its forecasts contributes to the over-estimation. Stratified forecast method may be adopted to take urbanization into account explicitly in national population forecasts to provide better forecasts. Multistate forecast method can be considered for countries with sufficient data.

Indexing (document details)
Advisor: Becker, Stan
School: The Johns Hopkins University
School Location: United States -- Maryland
Source: DAI-A 71/11, Dissertation Abstracts International
Subjects: Demography
Keywords: Population forecasts, UN, United Nations, Urbanization
Publication Number: 3428575
ISBN: 9781124263946