Cointegration and Causality between Fertility and Female Labor Participation in Taiwan: A Multivariate Approach
Cheng, Benjamin S., Atlantic Economic Journal
BENJAMIN S. CHENG [*]
Applying Hsiao's version of Granger causality, this paper uncovers no causality from fertility to female labor participation and fails to find the expected relationship that female labor participation negatively predicts fertility in Taiwan. This indicates that working women in Taiwan do not necessarily have fewer children. The finding of this study contradicts the results obtained when using the conventional regression method which finds bidirectional relationship between fertility and female labor participation. In addition, this study detects that education exerts a great influence on female labor participation but not on fertility.
One of the most heatedly debated and contentious issues in demography and labor economics has been whether the female labor participation rate causes a lower fertility rate. Earlier cross-sectional studies (for example, Becker , Stolzenberg and Waite , Lehrer and Nerlove , Easterlin , Mahdavi , Becker , and Lehrer  among many others) suggest a close correlation between fertility and female labor participation. Although these earlier regression studies have merit, their weakness lies in their attempt to equate correlation with causation. As Cochrane  argued, the fact that two variables tend to be closely associated and move together does not necessarily mean that one causes the behavior of the other. Besides, the static assumptions of econometric models employed in these studies are open to question. Fertility decisions are inherently sequential, and this stochastic nature of the human reproduction process should be considered in this type of study [Heckman and Willis, 1976; Willis, 1987; Cheng, 1996; Cheng et al., 1997]. In other words, the effects of socioeconomic variables are rarely instantaneous. Moreover, it is not unusual in macroeconomics for a variable to be affected by its own past behavior. Thus, the relationship between fertility and female labor participation should be viewed not only in a dynamic manner but also as an autoregressive process.
Most recently, Cheng  applied the newly developed technique to U.S. data (by incorporating dynamic and autoregressive elements into his study) and found that fertility Granger-causes female labor participation without feedback. Applying the same technique to Japan's time series data, Cheng et al.  found that outside employment does not hinder women from having more children, but having small children at home strongly discourages employment. Cheng  and Cheng et al.  have made significant contributions by applying the recently developed technique in time series analysis when studying the relationship between fertility and female labor participation. However, these tests were conducted in a restricted bivariate framework. The causality results were found to be highly sensitive to model specification. Bivariate casualty tests have fallen out of favor in macroeconomics because causality test results are obviously extremely sensitive to omitted variables. Granger , Lutkepohl , and Serletis  have all demonstrated that Granger causality is severely affected by the bias due to the omission of relevant variables.
Therefore, the purpose of this paper is to extend Cheng's  study by adopting a multivariate rather than a bivariate model to examine the Granger causality between fertility, female labor participation, and education using Taiwan data. Particular attention is given to the proposition that working women tend to have fewer children. The second section of this paper presents the model and the methodology, and subsequent sections report and discuss the empirical results.
Methodology and Models
The theoretical relationship between fertility and female labor participation is well-documented in literature so it is only briefly discussed here. In studying the causal relationship between female labor participation and fertility, it is particularly important to consider female labor participation as a crude proxy for the employment and income earning opportunities available to women [Bowen and Finegan, 1969]. …