An Econometric Analysis of the Determinants of Fertility for China, 1952–2000

Abstract

This study investigates the determinants of the fertility rate in China over the 1952–2000 period. Consistent with theory, the key explanatory variables in our fertility model are real per capita income, infant mortality rate, female illiteracy and female labour force participation rates. The long-run results and the test for cointegration are based on the Johansen (1988) and Johansen & Juselius (1990) approach. Our long-run results conform to theory in that all variables appear with their expected signs, and the dummy variable used to capture the effects of the family planning policy indicates that in the years of the policy, fertility rates have been falling by around 10–12%. Our results suggest that socio-economic development – consistent with the traditional structural hypothesis – played a key role in China's fertility transition.

Keywords:

• China
• fertility transition
• cointegration
• family planning policy
• social economic development

Keywords:

• JEL Classifications: J13
• C22
• C52

Notes

Notes

¹Social and economic development is typically viewed as providing an aggregate setting that influences fertility directly. It may also influence fertility indirectly through family planning programs and behaviours through other variables more proximate to fertility. In this context, family planning programs and services are seen as facilitating an already pre-eminent relationship (Poston & Gu, 1987; Poston, 2000). Without the prior effects of social and economic development on reward structures, however, family planning programs would have little or no effect on fertility (Poston, 2000). See also Schultz & Zeng (1995), Feeney & Wang (1994).

²The total population in China increased to 820 million in 1970 from 550 million in 1950. The average natural increase rate per year was 27.8% during the two decades.

³The one-child policy met resistance from couples with only one child, especially if the first child was a girl, so an amendment was soon made. For example, it allows rural couples in some areas with only a girl to have a second birth after an interval of several years.

⁴See Hirschman (1994) and Masih & Masih (1999, 2000) for the comprehensive review of the conventional structural hypothesis. Interested readers may also find Macunovich (1999) useful.

⁵Masih & Masih (2000) postulate a theoretical model: fr = ψ (mr, st, ed, y, pr). Where fertility fr and infant mortality mr are the demographic variables; contraceptive use st and female secondary education ed are proxies for the ‘ideational’ or diffusion hypothesis; and real per capita income y and female labour force participation pr are proxies for the ‘structural’ hypothesis. They use cointegration, vector error-correction modelling and variance decomposition techniques to analyse the dynamics of fertility and its determinants with a particular focus on the role played by female education and the family planning program in the context of India. They found that the two policy variables – female secondary enrolment ratio and contraceptive use – played a significant and predominant role in reducing the initial fertility rate, despite the underdeveloped status of India.

⁶Like Masih & Masih, we also found urbanization and per capita income to be highly correlated. In order to avoid the multicollinearity problem we did not incorporate urbanization in modelling the fertility rate.

⁷In conducting the cointegration test we also imposed the structural breaks obtained from the ZA test in the form of dummy variables. However, our results were unaffected. These results are not reported here but are available from the authors upon request.

⁸The variance decomposition analysis is based on a vector error correction model with two lags of the first difference variables and one lag of the error correction term. The lag length of 2 was obtained using the Schwarz Bayesian Criterion.