Sex ratio in India —
Prabhat Jha a, Rajesh Kumar b, Neeraj Dhingra a
Our estimate of 0·5 million selective abortions is for 1997, and the estimate of 10 million refers to 1985—2005, not to the two decades before 1997. Overall sex ratios (female births per 1000 male births) have dropped further since 1997 according to the Sample Registration System (SRS). Sabu George highlights a striking increase in ultrasound machines since 1997. Thus, it is reasonable to use the 1997 annual estimate as an annual average for 1985—2005.
Anil Grover and Rajesh Vijayvergiya have misread our paper. The SRS is nationally representative: national rates include all states and union territories (except Jammu and Kashmir and rural Nagaland), but we presented only the larger states in the figures and tables. Full details, including the simple survey instrument, are in the 2005 Government of India report.1 The proportions of missing second girls are similar among Hindus and Muslims. We do not understand the suggestion of a “quarter of a million [ultrasound] machines”. Even if we halve George’s estimate of active ultrasound machines to 17 000, less than one ultrasound examination per weekday is needed to reach the estimated 3·6 million self-reported ultrasound examinations of pregnant women in the second National Family Health Survey (NHFS-2).
Our reported gap of 30% between expected and actual second (or third) girls is too large to be due to chance or other background factors (which might lower overall sex ratios but not cause the conditional ratios that we describe). Female undercounts do not account for the differences across states and settings. These undercounts tend to be more common in less educated families. However, we found that the gap in second girls is about twice as great in those with grade 10 or higher education than in illiterate women .
The SRS is more than 10 times bigger than the NFHS-2, and its female birth rates are higher owing to continuous and dual monitoring of households .However, the overall sex ratio at birth and birth order are similar. After publication, we found a careful analysis of NFHS data by N P Das2 also documenting conditional sex ratios for second or higher order births.
Analyses of 6000 births in the USA3 refutes the idea that “boys run in the family” as George implies. Admittedly, we find puzzling the slight excess in girls after first boys (also reported by Das2). We do not believe that the excess represents male feticide. One explanation might be differential misclassification by ultrasound machines. It is normally assumed that ultrasound detects male sex more accurately, but the opposite has also been reported.4
Mari Bhat’s estimate of selective abortions relying on NFHS-2 self-reports might not be accurate given that households may not reliably report an illegal activity. Nonetheless, our abortion/ultrasound ratio (0·5/3·6 million, or 14%) is consistent with Bhat’s estimate of 6—17% “misuse”. His suggestion of using overall demographic profiles at older ages might confuse sex differences in childhood mortality with sex-selective abortion. Our method is novel, specific, and robust, but it is not the only estimate. Indeed, others quote higher numbers of sex-selective abortions: one claims over 0·2 million in two states alone.5 Long-term, low-cost, representative, and routine monitoring of mortality (including before birth) is planned in the SRS.1
We will produce updated estimates soon. If we arbitrarily halve or double our estimate, the implications remain the same: modest sex selection is leading to profound demographic distortions.
We declare that we have no conflict of interest.