Ultrasound, Birth Order, and Sex Ratio

V. Results

The reported estimates in this section are from linear probability re- gressions in which the dependent variable is an indicator variable that equals 1 if the birth is male. Linear probability models are useful to this study because the fi tted prob- abilities are close to 50 percent. Probit estimation produces almost identical results. Standard errors are adjusted for serial correlation by clustering at the county level.

A. Sex Ratio at Birth by Parity

We fi rst explore how the probability of male births differs across birth parity. 21 Only the birth- order indicators with fi rst births as the omitted category are included as the explanatory variables in the fi rst column of Table 2. The next column includes a set of mother- and pregnancy- specifi c covariates, which could potentially affect the likeli- hood of a male birth. Mother- level controls include the mother’s ethnicity Han versus ethnic minorities, education, and a quadratic of maternal age at conception. Pregnancy characteristics include gestation length and indicators for the timing of initial prenatal care visits. Additional controls, including the year of conception effects and county fi xed effects, are added sequentially in Columns 3–4. Column 5 presents the results from a regression that controls for provincial characteristics, including per- capita GDP, per- capita fi scal expenditure, the number of hospital beds and doctors, and the number of teachers all in logs. The fi nal column shows the specifi cation with county- specifi c linear time trends. The coeffi cients of the birth- order indicators are interpreted as the difference between the probability of male higher- order births and male fi rst births. The regression results reported in Table 2 reveal that sex ratios increase with birth order. The estimates of birth- order effects are positive and statistically signifi cant at the 1 percent level, and neither the point estimates nor the standard errors are consid- erably affected by the inclusion of additional controls. In the preferred specifi cation, which contains the richest set of controls fi nal column, Table 2, the estimates imply that second births are 2.0 percentage points more likely to be male than fi rst births, whereas third- and higher- order births are 4.3 percentage points more likely to be male compared with fi rst- order births. Thus far, our results are consistent with the empirical regularity found in censuses and other fertility surveys, which indicate that the sex ratio in China tends to increase with birth order. These empirical fi ndings are suggestive of prenatal sex selection. This issue is explored more directly in the fol- lowing subsection.

B. Ultrasound, Birth Order, and Sex Ratio

In this subsection, the effect of ultrasound technology on the probability of a male birth is estimated, allowing this effect to vary by birth parity. The results are reported from the estimation of Equation 1. The fi rst column of Table 3 presents the results from the most parsimonious speci- 21. To make these results comparable with subsequent analyses, we restrict the sample to births for which the information on ultrasound is available. T he J ourna l of H um an Re sourc es 52 Table 2 Effect of birth order on male probability births 1975–92: linear probability model results Dependent variable: child is male 1 2 3 4 5 6 Second birth 0.020 0.023 0.022 0.019 0.020 0.020 0.002 0.002 0.002 0.002 0.002 0.002 Third or higher order birth 0.039 0.044 0.044 0.042 0.042 0.043 0.004 0.004 0.004 0.004 0.004 0.004 Individual controls No Yes Yes Yes Yes Yes County fi xed effects No No Yes Yes Yes Yes Year of conception fi xed effects No No No Yes Yes Yes Provincial controls No No No No Yes Yes County- specifi c linear time trends No No No No No Yes Observations 298,616 298,616 298,616 298,616 288,740 288,740 R - squared 0.0008 0.0012 0.0031 0.0034 0.0033 0.0051 Note: Individual controls include mother’s ethnicity, education, maternal age at conception and its squared term, gestation length and indicators for the timing of initial prenatal care visits. County fi xed effects are separate indicator variables for each county. Year fi xed effects are indicators that allow for unrestricted differences in year- to- year changes. Reported in parentheses are standard errors clustered by county. denotes statistical signifi cance at the 10 percent level, at the 5 percent level, at the 1 percent level. Che n, L i, a nd M eng 53 Table 3 Effect of birth order, ultrasound availability, and their interactions on male probability births 1975–92: linear probability model results Dependent Variable: Child Is Male 1 2 3 4 5 6 Second birth 0.013 0.015 0.014 0.011 0.012 0.012 0.003 0.003 0.003 0.003 0.003 0.003 Third or higher order birth 0.025 0.030 0.030 0.027 0.027 0.027 0.004 0.004 0.004 0.005 0.005 0.005 First birth × ultrasound 0.003 0.000 0.003 –0.007 –0.007 –0.008 0.003 0.003 0.003 0.004 0.004 0.004 Second birth × ultrasound 0.019 0.017 0.019 0.013 0.012 0.011 0.003 0.004 0.004 0.004 0.004 0.005 Third or higher order birth 0.030 0.028 0.031 0.024 0.024 0.026 × ultrasound 0.006 0.006 0.006 0.006 0.006 0.006 Individual controls No Yes Yes Yes Yes Yes County fi xed effects No No Yes Yes Yes Yes Year of conception fi xed effects No No No Yes Yes Yes Provincial controls No No No No Yes Yes County- specifi c linear time trends No No No No No Yes Observations 298,616 298,616 298,616 298,616 288,740 288,740 R - squared 0.0010 0.0014 0.0033 0.0035 0.0034 0.0053 Note: Individual controls include mother’s ethnicity, education, maternal age at conception and its squared term, gestation length and indicators for the timing of initial prenatal care visits. County fi xed effects are separate indicator variables for each county. Year fi xed effects are indicators that allow for unrestricted differences in year- to- year changes. Reported in parentheses are standard errors clustered by county. denotes statistical signifi cance at the 10 percent level, at the 5 percent level, at the 1 percent level. fi cation, which includes only the birth- order indicators and the interactions between ultrasound technology availability and a full set of birth- order indicators as the inde- pendent variables. The coeffi cient of the interaction between ultrasound technology and the fi rst- birth indicator is extremely small and not statistically different from zero, implying that access to ultrasound technology is not associated with any signifi cant change in the sex ratio of fi rst births. However, the coeffi cients of the interactions between ultrasound technology and higher- order birth indicators are both positive and statistically signifi cant. The estimates suggest that after the introduction of ultra- sound technology, the probability of male births increased by 1.9 percentage points for second births and 3.0 percentage points for third- and higher- order births. A quali- tatively similar pattern of birth- order main effects is observed even in the absence of ultrasound technology, albeit with a smaller magnitude. This implies that prenatal sex selection was possible, although certainly more costly, prior to the introduction of ultrasound technology in the mother’s county of residence. In Column 2, the results are reported for a regression specifi cation that controls for observed individual heterogeneity. A set of individual control variables similar to those in the previous table is included. The estimates of the birth- order indicators, and their interactions with ultrasound technology access, are largely insensitive to the inclusion of the individual covariates. The results likewise suggest an imprecise zero effect of ultrasound technology on the male probability of fi rst births. Column 3 adds county fi xed effects to the regression model, to eliminate potential bias in the previous estimates that are attributable to time- invariant omitted factors that vary across coun- ties. Again, the estimates are highly robust to this adjustment. In the fourth column of Table 3, unrestricted year fi xed effects are included to absorb aggregate shocks to the sex ratio at birth that may be correlated with ultrasound technology adoption. This exercise results in only minimal decreases in the coeffi cients for the interaction between ultrasound technology adoption and second- and higher- order birth indica- tors. The estimates suggest that local access to ultrasound technology increases the proportion of male births by 1.3 percentage points for second births and 2.4 percent- age points for third- and higher- order births, both of which are signifi cant at the 1 percent level. Moreover, the estimated effect of ultrasound technology on the gender of fi rst births is close to zero and not statistically signifi cant at the 5 percent level. Column 5 controls for provincial- level characteristics and Column 6 further controls for county- specifi c linear trends. The inclusion of these province- by- year covariates and county- specifi c trends has little effect on the results. Our estimates suggest that a large proportion of the increase in sex ratios can be attributed to the adoption of ultrasound technology in China. From 1980 to 1990, the proportion of male births at parity two increased from 0.535 to 0.559. The estimates from the preferred specifi cation Column 4, Table 3 suggest that local access to ul- trasound technology increased the proportion of male second births by 1.3 percentage points. This implies that local access to ultrasound technology accounts for 54.2 per- cent of the increase in the sex ratio at parity two from 1980 to 1990 0.013 0.559– 0.535 = 0.542. Similarly, regression analysis suggests that from 1980 to 1990, local access to ultrasound technology accounts for 38.7 percent of the increase in sex ratio at parity three and above 0.024 0.585–0.523 = 0.387. The sex bias originates al- most entirely from second- and higher- order births; thus, our fi ndings indicate that approximately 40 to 50 percent of the increase in sex ratio at birth can be explained by local access to ultrasound technology. This number tends to be interpreted as a con- servative estimate of the overall effect of the introduction of ultrasound technology on the sex imbalance in China. This is because individuals with a very strong preference for male offspring may cross county borders to obtain ultrasound scanning for sex selection.

C. Previous Children and Sex Selection