Missing Women and the Price of Tea in China: The Effect of Sex-Specific Earnings on Sex Imbalance2008

This paper shows strong empirical evidence against the unitary household model. Using two post-Mao reforms as exogenous variation, the paper finds causal evidence that an increase in the price of tea (which she argues proxies for an increase in women income) leads to an increase in the survival rate of girls¹11 additional mu of tea reduced the fraction of male children by 1.2 percentage points and in the educational attainment of both boys and girls²21 additional mu of tea increased female and male educational attainment by 0.383 and 0.501 years. Conversely, an increase in the price of orchard (which is the proxy for male income) leads to a decrease in the survival³31 additional mu of orchards increased the fraction of male children by 0.5 percentage points and schooling⁴41 additional mu of orchards decreased female educational attainment by 0.119 years of girls, while there is no significant effect on the schooling of boys. The paper convincingly argues that this is not because of the fact that households view girls as a luxury good, or that they view children as a form of investment, or through sex selection. The evidence presented in this paper points that the channel through which the causal effect operates is an increase in mothers' bargaining power through the increase of female-specific income. With these results Qian shows that the income pooling premise behind the unitary household model is humbuggery, no disrespect.

With planning, there existed three categories of crops in China, staple foods (1), regulated (2), and unregulated (3). Tea bushes, which are woman intensive, and orchards, which are male-intensive are in Category 2, and are the focus of this study. She presents three sets of main results. Firstly: 

\[ \text{sex}_{ic} = \sum_{l=1963}^{1990} (\text{cat}1_i \times d_l)\beta_l + \sum_{l=1963}^{1990} (\text{cat}3_i \times d_l)\delta_l + \text{Han}_{ic}\zeta + \alpha + \psi_i + \gamma_c + \varepsilon_{ic}\]

where \(\text{sex}_{ic}\) is the fraction of males in county \(i\) for cohort \(c\)1, \(\text{cat}1_i\) and \(\text{cat}3_i\) represent Category 1 and Category 3 areas, \(d_l\) is a dummy for birth year \(l\)1, \(\text{Han}_{ic}\) controls for Han ethnicity, \(\psi_i\) and \(\gamma_c\) are county and cohort fixed effects. The result in Figure IIIa shows that baseline trends on the categories that are not in focus (1 and 3) had no effect on child sex ratios, which helps make her strategy more credible. Next is the powerful result we previously mentioned:

\[\begin{split} \text{sex}_{ic} = (\text{tea}_i \times \text{post}_c)\beta + (\text{orchard}_i \times \text{post}_c)\delta + (\text{cashcrop}_i \times \text{post}_c)\rho \\ + \text{Han}_{ic}\zeta + \alpha + \psi_i + \gamma_c + \varepsilon_{ic} \end{split}\]

where \(\text{tea}_i\), \(\text{orchard}_i\), and \(\text{cashcrop}_i\) are the amounts of tea, orchards, and all Category 2 crops respectively planted for each county \(i\), while \(\text{post}_c\) indicates post-reform cohorts. This is where the results I previously mentioned come from. This is confirmed with a dynamic DiD approach and an IV 2SLS approach. The proposed instrument is hilliness at the county-level, since hilliness does affect wether a county grows tea, but is unrelated with culture and gender preferences.

\[\begin{split} \text{tea}_i \times \text{post}_c = (\text{slope}_i \times \text{post}_c)\lambda + (\text{cashcrop}_i \times \text{post}_c)\phi \\ + \text{Han}_{ic}\zeta + \alpha + \psi_i + \text{post}_c \gamma + \varepsilon_{ic} \end{split}\] The hilliness in the first stage shows a strong effect on tea cultivation. Then the second stage:

\[ \begin{split} \text{sex}_{ic} = (\widehat{\text{tea}_i \times \text{post}_c})\beta + (\text{cashcrop}_i \times \text{post}_c)\phi + \text{Han}_{ic}\zeta \\ + \alpha + \psi_i + \text{post}_c \gamma + \varepsilon_{ic} \end{split}\]

In this case the estimate is almost the same as the normal DiD, which lends credibility to her strategy. The instrument also seems credible. 

I liked this paper, it is true that it involves lots of proxies, but I believe her, and the conclusion is sound. This is a very creative paper, and it is unbelievable to me that anyone thought to use that model for a second. See many feminist and critical theorists, and even critiques from inside the discipline. Nancy R. Folbre was already criticizing this and the ethical assumptions (as well as other things) in her 1984 AER 74(3) comment. I don't have any complaints.