The Economics of Fertility in Developed Countries: A Survey

Abstract

This paper will be included as a chapter in the forthcoming Handbook of Population and Family Economics by Rosenzweig and Starks. The paper describes trends in US fertility and presents an assessment of static and dynamic models of fertility and econometric methods of analysis. It is argued that theories of the allocation of time of household production and human capital investment help to improve knowledge about fertility decision making in developed countries. How these theories can be included in empirical assessments of the determinants of fertility results in the identification of exemplary strategies for obtaining credible estimates of causal relationships. Improvements in earlier models accounted for the timing of first births the spacing of children and contraceptive behavior. The emphasis now is on econometric models of fertility behavior within a dynamic or life cycle context. The review includes the strengths and weaknesses of different strategies for identifying price and income effects on household-level fertility and analyses of aggregate time-series data choices hazard models and estimated structural models. Ordinary least squares models are viewed as potentially inconsistent because endogenous variables such as womens earnings or government policies are potentially correlated with unmodeled choice issues or other unobservables. Fixed effects models attempt to include controls for omitted variables that produce an endogeneity bias. However these aforementioned reduced form effects of prices and income on fertility models are static. Rosenzweig and Wolpin offer alternatives for identifying the effect of exogenous variation and quantity-quality trade-offs in fertility. Another option is to apply static models to age specific fertility rates or to use hazard approaches for modeling incomplete histories. Structural dynamic approaches rely on estimable stochastic models of life cycle fertility.

Cite this work

No public copy is stored for this record, so only the metadata is cited.

Stored only when both papers are in the corpus — 32 citations exist upstream.