There Is a Glaring Gender Bias in Death Registrations in India

In the absence of a reliable Civil Registration System in India, the sample registration system, beginning in 1970, has been the only source of information that allows us to track the Sustainable Development Goals, calculate the human development index, and measure sex ratios. Since 2001, however, there has been no attempt to examine the quality of the sample registration system. In this context, the present article carries out such an exercise and finds that there was an undercounting of deaths in India by around 4.3% for males and 11.3% for females during 200110. 


The Civil Registration System (CRS) in India is grossly deficient and cannot be used directly to analyse vital rates, such as fertility or mortality rates. Operating under the Office of the Registrar General and Census Commissioner of India (ORG), the CRS registers vital statistics such as births, deaths, and stillbirths from all over the country. However, its implementation on the ground and the quality of its registrations is severely inadequate. Having serious effects on the way in which vital rates such as the underfive mortality rate, fertility rate, sex ratio or the human development index are calculated for India, a faulty CRS could lead to misrepresentation and thus incorrect policy interventions. 

For instance, in 2011, the CRS was only able to account for 81% of the total births in the country (Office of the Registrar General of India [ORG] 2011). That is, it registered only 21.8 million births, as opposed to the expected 26.3 million births (calculated using birth rates from the sample registration system [SRS] and the 2011 population census). Adjusting the population in 2011 for its net omission rate of 2.298%, the expected births would have been 27 million. Surprising as it may be, this is in fact a relative improvement compared to the data accumulation from 10 years back, when it was only able to account for 50% of the total births. Similarly, the CRS counted only 65% of total deaths in 2011. That is, it registered 5.74 million total deaths, as opposed to the expected 7.79 million. In the case of infant deaths, the system miserably failed, as it only accounted for 1,78,172 deaths against the total of 11,88,000 infant deaths in India (ORG 2011). 

Looking at the historical insufficiency of the CRS, the SRS was launched in the late 1960s to provide data for calculating key rates, such as fertility and mortality rates, etc. After the launch of the SRS, in the late 1960s, the ORG conducted two in-house comprehensive evaluations for the SRS in 19801981 and 19841985 (RGI 1988). The evaluation in 1980-81 had an undercount rate of 3% for both births and deaths, whereas the evaluation during 1984-85 had an undercount rate of 1.8% for births and 2.5% for deaths. There is, however, no information available by gender, which could have posed a problem in discussing national-level, gender-based improvement in the mortality rate or the sex ratio.

Though the SRS maintained a 95% completeness in death registration for males during 1971–1997, this fluctuated in the case of females, from 91% during 1971–1981, to 88% during 1981–1990, and 93% during 1990–1997 (Bhat 2002). The deterioration in the quality of registration was noticeable in Punjab, Haryana, and Uttar Pradesh in northern India, Andhra Pradesh and Tamil Nadu in southern India, and Maharashtra and Gujarat in western India. These may be suggestive of larger issues, such as the ingrained gender biases prevalent in these states (Bhat 2002).

This gender-based gap is exacerbated while evaluating the SRS data using indirect techniques for the period 1981–1990, which indicate an omission of 5% of deaths among men and 12% of deaths among women on average; at the same time, births were also under-reported by 7% (Bhat 2002). Such a differential in omission has serious implications for studying the gender gap in mortality rates, and consequently, the implementation of women-centric interventions by the government. 

Data for the 1990s and the first decade of the 21st century have also not been put to systematic scrutiny, nor analysed with respect to the quality of counts in deaths in the SRS. Thus, it becomes much more important to evaluate SRS data quality, not only in the context of changing sampling areas for each SRS, but also in the context of changing vital rates. Therefore, the focus of the present work is to evaluate  mortality data from the SRS for the periods 1991–2000 and 2001–2010.


The SRS data is an initiative of the ORG, with the goal of generating reliable and continuous data on demographic indicators. It was introduced as a pilot scheme in some major states in 1964–65 and was made into a full-scale system in 1969–70. The SRS collects data using a dual record system, which first enumerates events on a continuous basis, and later verifies the same by conducting retrospective half-yearly surveys. This is followed by the process of matching the two records and subsequent field verification of the unmatched and the partially matched events. We used the percentage of deaths by age and sex from the SRS, the percentage of deaths by age and sex used for estimating the number of deaths, and the size of population by age and sex for India and its states between 1991 and 2010 via the census. The states analysed are Punjab, Haryana and Rajasthan (from the northern region), Uttar Pradesh and Madhya Pradesh (from the central region), West Bengal, Odisha and Bihar (from the eastern region), Gujarat and Maharashtra (from the western region), and Andhra Pradesh, Karnataka, Kerala and Tamil Nadu (from the southern region). Himachal Pradesh and Assam could not be included in the analysis as the data is deficient for both. The states covered in the analysis account for nearly 87% of all of India’s population, as per the 2011 Census. 


Brass (1975) used the concept of balancing equation to figure out the relationship between age distribution of deaths and population. In the absence of migration and under the assumption of population stability, it was logically argued that the proportion of population at an exact age x to total population above age x would equal growth rate (r) plus ratio of death above age x to population above age x. However, it has been argued that this method makes strong assumptions, particularly of population stability and, therefore, it may not work when the population under study is unstable. Preston and Hill (1980) proposed another method that does not require the assumption of stability, but which does not take migration treatment into consideration either. Interpretation of completeness of death registration above age five or so was linked to relative quality of enumeration of two censuses being used. However, they proposed another method based on the assumption of stable population and the logical concept of stationary population. In stationary population, all deaths above age x are the same as survival at age x, as all those who are at the survival age would die in the remaining years of their life.

After the emergence of the generalised population model (Coale and Trussell 1981), treatment of relations in various population parameters underwent a drastic change. In mathematical demography, this model has been demonstrated to be useful not only in the estimation of vital rates using census data (Coale and Trussell 1981), but also for understanding population momentum and distance to stability. Using this model, Bennett and Horiuchi (1981) extended previous analyses and proposed a method to estimate completeness of death registration during intercensal period. These methods were used to estimate the completeness of the CRS in India, and in some of the Asian countries (Pathak and Ram 1981). Bhat (2002a) reformulated the general growth balance method incorporating migration to estimate the completeness of the count of deaths. 

In the present research work, we have used the Bennett and Horiuchi (1981) method. Results were obtained using the latest version of the MortPak software package1 to obtain the data set for India and its selected states, for the periods 1991–2000 and 2001–2010. 


The results using the Bennett and Horiuchi (1981) method for estimating completeness of death count (age five+) in the SRS in India, for the periods 1991–2000 and 2001–2010, are discussed in this section. At the outset, it is pointed out that there may be an undercount in relation to the quality of the respective census in 1991, 2001, and 2011. It may be observed that among males, during 1991–2000, nearly 7.5% of (ages five+) deaths were missed in the SRS at the all-India level, whereas only 2.1% of female (ages five+) deaths were missed. During 20012010, 4% of male (ages five+) deaths and 11% of female (ages five+) deaths were missed.

At the state level, there was an unexpectedly huge variation not only across states but also within the states during the two decades. Based on these estimates, we may conclude that Maharashtra and Punjab seem to have had full coverage of (ages five+) deaths. The states of Andhra Pradesh, Gujarat, Karnataka, Madhya Pradesh, Odisha, and Tamil Nadu indicate an invariably high undercount during 1991–2000, but almost 100% coverage during 2001–10. States like Assam, Bihar, Haryana, Himachal Pradesh, Kerala, Rajasthan, Uttar Pradesh, and West Bengal show a moderately high undercount of (ages five+) deaths in both decades.


The SRS has undoubtedly helped us understand the demographic transition in the country in the last five decades. However, clinically examining the quality of data raises questions. With a moderately high undercount rate in select states for a decade, that is, 20012010, macro-level indicators measured using the SRS turned out to be faulty. This would mean that subsequent policy interventions may have been a result of serious misdiagnosis as well. It is, thus, pertinent to have an extensive internal auditing of the data collected through the system.

Moreover, indirect methods for estimating population parameters in the absence of reliable data have contributed to our understanding of demography in developing countries. Nevertheless, these methods rely on certain assumptions and require data to be in a certain format. The current analysis provides an undercount for ages above five or 10 only. There is no way to get the completeness of count for deaths below age five. It has been observed (ORG 2010) that under civil registration, deaths below age five, especially those below age one, are grossly undercounted. This raises doubt over the current demographic literature for India. 

Particularly, with a few exceptions, the analysis clearly indicates a comparatively larger undercount of female deaths, as compared to males. In other words, female mortality is underestimated in the SRS; more so in recent times. Such a situation leads us to believe that whatever gains for females have been observed in terms either of under five mortality rates (U5MR) or female mortality may not be true. 

In order to understand the dynamics of the quality of data from the SRS, an in-depth analysis of unit level data is essential. The Registrar General of India must regularly get its data analysed; not just for internal consistency, but also for its validity vis-à-vis census age distribution. 

In Conclusion

In the absence of a reliable civil registration system in India, the SRS provides estimates of vital events. However, the data collected under the SRS has limitations, as shown by preliminary calculations. The analysis in this article indicates a huge undercount of deaths, especially females as compared to males. Moreover, the SRS counting/coverage appears to have deteriorated over time. This puts a huge question mark on the credibility of the SRS data, especially with regards to the gender gap observed in mortality. It may be concluded that there is a need to review the SRS procedure/design of data collection and update the methodology of deaths registration (for ages five+). Doing so will give better and more reliable estimates of deaths (ages five+) at the national and state levels by gender, and update mortality indicators like infant mortality rate, U5MR, and female mortality rate.

The authors would like to thank Patrick Gerland for technical comments, and Shailja Thakur for providing editorial assistance.

Must Read

Do water policies recognise the differential requirements and usages of water by women and the importance of adequate availability and accessibility?
Personal Laws in India present a situation where abolishing them in the interest of gender justice also inadvertently benefits the reactionary side.   
Back to Top