Daughter Elimination in Tamil Nadu, India: A Tale of Two Ratios

Abstract

A disturbing feature of demographic trends in India is the sharp decline in the proportion of girls to boys. Most existing analyses of the Indian child sex ratio present a country wide picture and focus on trends across states. Such state level analyses may hide intra state variation. This paper uses district and village data on sex ratio at birth and infant mortality to examine the extent, geographical spread and nature (before or after birth) of daughter deficit within the South Indian state of Tamil Nadu. Our analysis shows that (i) daughter deficit in Tamil Nadu occurs in nearly half the state's districts; (ii) a large proportion of daughter deficit occurs before birth; (iii) daughter deficit rises with birth order and (iv) daughter elimination is not confined to particular socio-economic groups.

Acknowledgements

The authors would like to thank Christophe Z. Guilmoto and Moyna Bammi Bedi for preparing the maps presented in the paper and Roy Huijsmans, D. Jayaraj, Admasu Shiferaw, S. Subramanian, Dubravka Zarkov and two anonymous referees for their comments.

Notes

1. Kerala has witnessed a decline in child (0–6) sex ratio between 1961 and 1991 and although a slight improvement has been recorded in the 2001 Census, there are indications that this may be temporary (Sudha and Irudaya Rajan, 1999; Patel, 2002).

2. The 0–6 sex ratio reflects the combined effect of the SRB, infant mortality (0–365 days) and child mortality (1–6 years). Male mortality is typically higher than female mortality in the age group 1–6. Thus, the 0–6 sex ratio is likely to hide the extent of female deficit at birth and in the age group 0–365 days.

3. The terms female deficit and daughter deficit are used synonymously. However, we draw a distinction between daughter/female deficit and daughter elimination. Deficit implies that there is a gap between the number of expected daughters and the number of daughters actually born. Elimination, whether it is sex selective abortion or infanticide, is treated as a potential cause of the observed deficit.

4. For example the Tamil Nadu government's Girl Child Protection Scheme is described as a scheme intended for ‘the education of girl children in poor families; to promote family planning, to eradicate female infanticide and to discourage preference for male child’ (GoTN Policy note 2003–2004, Demand No. 43, G.O.Ms.No.131dt 17.7.2003, accessed at: www.tn.gov.in). The scheme explicitly targets families below the official poverty line. Programmes run by non-governmental organisations (NGOs), such as Poonthalir in Salem and the Indian Council for Child Welfare in Madurai, offer economic support and training to poor families in order to avert infanticide (see Srinivasan, 2006: 224).

5. The definition of urban and rural areas used in the Vital Event Surveys (VES) is the same as the definition used in the census. Urban is defined as all places declared by the state government under a statute as a municipality, corporation, cantonment board or notified town area committee. Any area, which is not covered by the definition of urban, is rural. More details are available at: http://www.censusindia.gov.in/Census_Data_2001/Census_Newsletters/Newsletter_Links/eci14mail.htm (accessed on 15 June 2008).

6. We do not use the survey which covers the reference period 1995 as there are some doubts about the quality of these data. According to Athreya (1999), the first round of the VES was conducted at very short notice, the planning and training of enumerators was inadequate and the sample was not selected on the basis of the specified design. Subsequent rounds of the survey were designed, planned and implemented keeping in mind the shortcomings and the lessons learnt from the first round.

7. Birth and death records in the nutrition and health centres were used to infer female infanticide (and, to a lesser extent, sex selection). Health records in Tamil Nadu use the category ‘social’ causes to record infant deaths due to infanticide or neglect. However, not all such deaths may be recorded accurately. In cases where the cause of female infant death was not reported, discussions with health and nutrition workers revealed whether death occurred due to unnatural causes or occurred in suspicious circumstances (for example, a newborn baby rolling across a bed and falling on the floor).

8. Countries included are, Egypt (1999) 946; Canada (2002) 947; Guatemala (1999) 965; Chile (2003) 955; Japan (2003) 948; Saudi Arabia (2000) 951; Sri Lanka (1996) 951; France (2002) 953; Netherlands (2002) 948; Poland (2003) 944; Romania (2003) 940; Spain (2002) 939; United Kingdom (2003) 951; and Australia (2003) 944.

9. Based on their analysis of almost 2 million births which occurred in hospitals and health centres during the period 1949–1958 (presumably a period when there was no pre-birth interference), Ramachandran and Deshpande (1964) conclude that during the period under analysis, the SRB in India was 943. The ratio varies between 926 in central India and 961 in north India. Despite the fact that sex ratios based on hospital data may reflect only a small proportion of births, the number of births analysed in their paper is quite large and the SRB lies in the range reported for other countries. Based on an estimate of the link between SRB and life expectancy at birth, Klasen and Wink (2003) compute an expected SRB of 962 for India.

10. The medical and social science literature identifies several medical and environmental factors that may have an effect on the sex of children and result in unusually skewed sex ratios at birth. These include: maternal nutrition (Goodkind, 1996; Jayaraj and Subramaniam, 2004); Hepatitis-B (Oster, 2005); father's occupation (Dickinson and Parker, 1997); father's presence at home (Norberg, 2003); maternal dominance (Grant and Yang, 2003); smoking (Fukuda et al., 2002); and time taken to conceive (Smits et al., 2005).

11. Based on data covering the period 1976–1984, the ratio of male to female infant deaths was 133 in Canada, 133 in Japan, 131 in the United States, 129 in Hong Kong, 133 in France, 132 in Belgium, 131 in Austria, and 130 in Sweden. This pattern also prevails in less developed countries. For the same period the ratio was 128 in Malaysia and The Philippines (for more details see Johansson and Nygren, 1991).

12. Based on an expected SRB of 952 and the typically higher rates of male infant and child mortality, the lower limit for the 0–6 sex ratio in populations without any pre- or post-birth interference may be expected to be 952.

13. While the all-India infant mortality rate also exhibits a declining trend, the infant mortality rate in Tamil Nadu is considerably lower than the national IMR of 70 per 1000 live births in 1999. In the national figures, gender differences are not substantial although, contrary to international norms, the male IMR is lower than the female IMR (Planning Commission, 2001).

14. Given the concentration of medical facilities and professionals in urban areas, it is surmised that the lower SRB in urban India is driven by the widespread availability of the technology for sex selection (Agnihotri, 2003). In contrast, in Tamil Nadu, daughter deficit is mainly a rural phenomenon and may largely be driven by sex selective abortion. While a detailed analysis of the reasons for these differences is not pursued here, the extent of urbanisation and rural-urban connectivity in Tamil Nadu could have facilitated the spread, availability and access to sex selection technology. Tamil Nadu is the third most urbanised state in India with an urban population of 44 per cent. It has the highest composite index of urbanisation (as measured by urban population ratio, rural population served by urban centres and distance to the nearest town) resulting in better road facilities, development of transport network and migration (Rukmani, 1994). In addition, the focus of the state government on preventing female infanticide in rural areas may have created the demand for sex selection technology and led to its spread and availability in rural areas.

15. The total number of districts tends to vary over time. While our analysis is based on 29 districts, at the moment Tamil Nadu is divided into 30 districts (accessed at: www.tn.gov.in, 13 December 2006).

16. As mentioned earlier, we use gender differences in infant mortality rates as an indication of post-birth deficit. If we were to take into account that male IMR rates is expected to be higher than female IMR, it would lead to a larger estimate of the post-birth deficit. However, even with this adjustment the proportion of pre-birth deficit would be larger than post-birth deficit.

17. The male birth rate in 1999 was 20.9. Based on the state's male population of 31 million in 2001, this translates into 647,900 male births. An overall deficit of 20 females per 1000 males implies a female deficit of 12,958, that is, a pre-birth deficit of about 11,014 and a post birth deficit of 1943. Our estimate of about 2000 female infant deaths is somewhat lower than that reported in Chunkath and Athreya (1997). Based on data from primary health care centres, Chunkath and Athreya (1997) report that in 1995 there were 3226 female infant deaths due to social causes. Our figures refer to a slightly different period (1996–1999) during which the state government focused on preventing female infanticide in rural areas. This policy may have led to a reduction in female infanticide and, hence, the smaller number of cases but, at the same time, may have led to increased demand for sex selection technology in rural areas.

18. The incidence of abortion in Tamil Nadu is much higher than the corresponding national figures. According to International Institute for Population Sciences (1995) based on the National Family Health Survey 1992–1993, 11 per cent of pregnancies are aborted (spontaneous and induced) in Tamil Nadu, while the corresponding figure for the country as a whole is 5.8 per cent. Induced abortion rate for Tamil Nadu and India are 4.3 per cent and 1.3 per cent respectively.

19. More details on the patterns and the history of daughter elimination in this village are available in Srinivasan (2006).

20. Before carrying out their empirical work, Nillesen and Harris-White (2004) present a more detailed discussion of the potential link between household wealth, male and female education and the survival probability of females. They also draw a distinction between landed and landless households. Given the small sample at our disposal we do not attempt such distinctions and confine ourselves to examining the notion, implicit in intervention programmes, that daughter deficit is a problem of poverty.

21. Due to the small number of cases, we combine the two forms of daughter elimination rather than estimating separate models for female infanticide and sex selective abortion.

22. Comparisons across states shows that economically well-developed states tend to have lower female to male sex ratios and the most rapid decline in the ratio has also occurred in such states (see Premi, 2001). Based on data reported in Agnihotri (2003), in rural India, the correlation between the 0–14 female to male sex ratio and expenditure per capita is −0.82. Similarly, Siddhanta et al. (2003) show that in many large states the 0–14 female to male sex ratio declines with expenditure.

23. Based on the prevalence of Hepatitis B virus carriers and the effect of carrying this virus on SRB, Oster (2005) computes a ‘normal’ sex ratio of 935 for India. This may be compared with the national SRB of 899 in 1997.

24. A heartening feature is that for the first time, in April 2006, a doctor and a pharmacist in Haryana (a state with one of the lowest sex ratios in the country) were convicted and sentenced to two years in prison for violating the Pre-Natal Diagnostic Techniques (Regulation and Prevention of Misuse) Act, 1994. The state's health officials conducted 3200 inspections in 2005–2006 and revoked or suspended the licences of 114 ultrasound clinics. They also use decoy customers to test whether doctors are violating the law.