Counting missing women: a reconciliation of flow and stock measures

Ebert, Cara; Klasen, Stephan; Vollmer, Sebastian Article — Published Version Counting missing women: a reconciliation of flow and stock measures Journal of Population Economics Suggested Citation: Ebert, Cara; Klasen, Stephan; Vollmer, Sebastian (2025) : Counting missing women: a reconciliation of flow and stock measures, Journal of Population Economics, ISSN 1432-1475, Springer Berlin Heidelberg, Berlin/Heidelberg, Vol. 38, Iss. 4, https://doi.org/10.1007/s00148-025-01132-0 This Version is available at: https://hdl.handle.net/10419/333223 Standard-Nutzungsbedingungen: Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden. Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. Terms of use: Documents in EconStor may be saved and copied for your personal and scholarly purposes. You are not to copy documents for public or commercial purposes, to exhibit the documents publicly, to make them publicly available on the internet, or to distribute or otherwise use the documents in public. If the documents have been made available under an Open Content Licence (especially Creative Commons Licences), you may exercise further usage rights as specified in the indicated licence. http://creativecommons.org/licenses/by/4.0/ Vol.:(0123456789) Journal of Population Economics (2025) 38:82 https://doi.org/10.1007/s00148-025-01132-0 ORIGINAL PAPER Counting missing women: areconciliation offlow andstock measures CaraEbert1 · StephanKlasen2· SebastianVollmer2 Received: 2 August 2023 / Accepted: 22 September 2025 © The Author(s) 2025 Abstract Stock estimates of missing women suggest that excess female deaths are concentrated in South and East Asia and among young children. In contrast, flow estimates suggest that gender bias in mortality is much larger than previously estimated using stock measures, is as severe among adults as it is among children in India and China, and is larger in Sub-Saharan Africa than in India and China. We show that the different stock and flow measure results rely on the choice of the reference standard for mortality and an incomplete correction for different disease environments in the flow measure. Alternative reference standards reconcile the results of the two measures. Keywords Missing women· Gender bias· Mortality· Disease· Age· Sub-Saharan Africa· China· India JEL Classification J16· D63· I14· I15· O15 Responsible editor:Kompal Sinha Stephan Klasen passed away on October 27, 2020, after battling the incurable disease Amyotrophic Lateral Sclerosis (ALS) for 5 years. We are grateful for what we have learned from Stephan through our joint work and previously as students. * Cara Ebert cara.eber[email protected] Sebastian Vollmer sv[email protected] 1 RWI – Leibniz Institute forEconomic Research, Berlin Office, Zinnowitzer Str. 1, 10115Berlin, Germany 2 University ofGoettingen, Waldweg 26, Göttingen37073, Germany C.Ebert et al. 82 Page 2 of 26 1 Introduction The extent of gender bias in mortality in developing countries has been a research topic that sparked substantial controversies in the economic and demographic literature since Amartya Sen first pointed in 1989 to the millions of “missing women” in South and East Asia (Sen 1989). This literature focused largely on calculating stock measures of missing women. Stock measures compare the actual sex ratio (males/ females), using census information in a country, with an expected sex ratio that would occur in the absence of gender bias in mortality. An alternative method to the stock measure poses the flow measure of missing women (Anderson and Ray 2010). The flow measure compares the ratio of male-to-female age-specific mortality rates with an expected ratio of male-to-female age-specific mortality rates from a reference population without gender bias. Because flow measure estimates can be better disaggregated by time, age, and disease—whereas the stock measure estimates the total deficit in alive women at one point in time—they are of great value to target policies and stimulate future research. However, these two measures have very different implications about the overall number of missing women and their distribution by age and region. This paper examines the flow measure’s sensitivity to alternative assumptions and methods and reconciles the findings from the two measures. Missing women estimates based on the stock measure, and many more studies on gender inequalities using micro data from individual countries or regions, have converged on two key findings (e.g.,Sen 1989; Coale 1991; Coale and Banister 1994; Das Gupta 1987, 2005; Murthi etal. 1995; Klasen and Wink 2002, 2003; Das Gupta etal. 2009).1 First, the problem of missing women is much more prevalent in China and India than in Sub-Saharan Africa. Second, in China and India, missing women are largely attributable to excess female mortality before birth and/or the first few years of life. The stock estimates amounted to around 100 million missing women in 2000 in the most affected countries, i.e., China, India, Pakistan, Bangladesh, the countries of Western Asia, and some countries of North Africa (Klasen and Wink 2003). Compared with a previous census round in 1990, the number of missing women had increased by about 6 million between 1990 and 2000 but fallen as a share of females alive (from 6.5 to 5.7%) (Klasen and Wink 2003).2 1 The micro-level literature has developed into a massive evidence base documenting for East and South Asia not only the existence and determinants of gender discrimination in mortality, but also in numerous other health and behavioral outcomes, particularly at birth and the first early years of life (e.g., Asfaw, Klasen and Lamanna 2010; Jayachandran and Kuziemko 2011; Barcellos etal. 2014; Jayachandran and Rohini P, 2017; Ebert and Vollmer 2022; Nath 2023). The impact policies or factors that reduce these gender imbalances have also been studied (e.g., Anukriti 2018; Tandel etal. 2023; Luo and Zou 2024; Javadekar and Saxena 2025). 2 A major controversy about these stock estimates of missing females arose when Oster (2005) claimed that the “natural” sex ratio at birth in some of these countries was much higher due to the fact that the widespread prevalence of the hepatitis-B-carrier status raised the sex ratio at birth. She claimed that this way, some 50–70% of missing females could be accounted for; i.e., their absence was unrelated to discrimination. After an intense debate and further research (Abrevaya 2009; Das Gupta 2005, 2006; Lin and Luoh 2008), Oster retracted the finding in 2009 (Oster etal. 2010) and the previous estimates of around 100 million were reinstated. See Klasen (2008) for a review of this debate. Counting missing women: areconciliation offlow andstock… Page 3 of 26 82 Contrary to findings from the stock measure, estimates of the annual flow of missing women for the year 2000 in China, India, and Sub-Saharan Africa suggest that excess female mortality is (a) as severe among adults as it is among children in India and China, (b) higher in Sub-Saharan Africa than in China and India, and (c) largely due to excess adult female mortality in Sub-Saharan Africa (Anderson and Ray 2010; World Bank 2012). For China, India, and Sub-Saharan Africa together, excess mortality based on the flow measure is also much larger than implied by the existing literature on stock measures. The flow measure suggests that in a single year, about four to five million women died in excess in India, China, and Sub-Saharan Africa alone. For the 1990 s, the stock measure would be consistent with a flow of about one million missing women in a given year.3 This paper discusses three key methodological aspects of the flow measure and shows that the flow measure results are consistent with the stock measure results when a set of different assumptions is made. The first two points regard the reference standard, which is used to simulate a gender-equal society, and the third point regards the availability of quality mortality data required for the calculation of the flow of missing women. The ideal counterfactual for assessing the amount of gender bias in mortality would be a society in which males and females are treated equally and in which there are no differences in sex-based behavioral patterns across countries with significant mortality implications (Coale 1991; Klasen and Wink 2002, 2003; Anderson and Ray 2010). There is no such society, past or present, that exhibits these features, so it is inherently difficult to generate a reference standard that is entirely beyond reproach. However, because it is so difficult to identify an adequate reference standard, it is important to understand the implications of the choices made. Throughout the paper, we refer to the flow measure based on the methods and data used in Anderson and Ray (2010)—the seminal paper on the measurement of the annual flow of missing women—and therefore the number of missing women in 2000. The citation count of Anderson and Ray (2010) is 431, with the highest citations per year very recently in 2021 (38 citations) and already 28 citations this year (2025).4 Anderson and Ray (2012, 2019) use the same methods and estimate excess female mortality in Indian states and among unmarried women in developing countries. First, we review the reference population of high-income countries in 2000 used in flow estimates.5 Today’s high-income countries are an environment with extremely low overall adult mortality and, therefore, already small differences in male and female mortality rates result in large changes in the male-to-female mortality rate ratio of high-income countries. This is empirically relevant as non-natural causes of death or 3 Note that flows of missing cannot be calculated as differences in stocks between time periods; i.e., the difference of 6 million in stocks does not imply 600,000 missing women per year. This is due to the fact that some women in a particular age group that have died in excess in a year might not count as missing in a stock measure at the next census if in the time interval some men in that age group have also died due to high overall mortality; i.e., to the extent that an excess female death in a year just advances mortality between two census periods, it would not add to the increase in the stock of missing women at the next census. 4 Based on Google Scholar and as of December 2024. 5 High-income countries refer to the former World Bank country classification “Established Market Economies” as used by Anderson and Ray (2010), which includes Western Europe, Canada, United States, Australia, New Zealand, and Japan. C.Ebert et al. 82 Page 4 of 26 consequences of behavioral patterns, such as accidents, violence, suicides, and smoking, affect males much more than females. Second, we discuss that disease environments do not differ across countries for males and females in the same way, but they differ across sexes and countries. Therefore, controlling for differences in the disease environment in the reference standard requires controlling for sex differences within disease groups across countries. For example, the male-to-female mortality ratio from AIDS is around 4 in highincome countries in 2000 because AIDS mortality was confined to male-dominated highrisk groups (gay men and IV drug users), whereas in Sub-Saharan Africa AIDS spread from the beginning to the entire population, with heterosexual transmission playing a key role, and AIDS mortality is more similar for males and females. Thus, using today’s highincome countries as a reference population for AIDS mortality suggests massive excess female mortality in adulthood in countries with high rates of AIDS mortality. The World Development Report 2012, which uses the flow measure approach, indeed shows particularly high rates and increases in excess female mortality in AIDS-affected countries of Sub-Saharan Africa (World Bank 2012). The two points on the low mortality environment and the disease correction are related, but each has a separate effect on the estimates of missing women and is therefore also discussed separately. In a third point, we highlight a general concern that underlies the use of mortality data in Sub-Saharan Africa, which is that reliable vital registration data is only available in one of 46 countries. For the remaining countries, mortality data must be estimated and, therefore, the flow of missing women is inherently imprecise. We use a variety of alternative reference standards for sex-specific mortality and find patterns in the flow of missing women for the year 2000 that are consistent with previous findings on the stock of missing women. We find that excess female mortality is much more serious in India and China than in Sub-Saharan Africa. In China and India, excess female mortality is largely driven by gender bias pre-birth and during the first few years of life. The magnitude of the problem is also much reduced, with our estimates suggesting that 1.5–1.8 million women die in excess in the three regions in 2000. This number is substantial and worrying but below the number of five million females per year estimated in Anderson and Ray (2010) or 3.3 million under the age of 60 by the World Bank (2012). Our estimate is also lower, but to a lesser extent, than the flow measure estimate by Bongaarts and Guilmoto (2015), who show trends in excess female mortality from 1970 to 2050 using data from 93 countries with different mortality levels to create a reference population.6 Bongaarts and Guilmoto (2015) 6 The estimations of Bongaarts and Guilmoto (2015) amount to excess female deaths after birth of 0.54 million in China and 0.82 million in India in 2000. They predict the reference sex ratio of mortality based on a linear regression of the log sex ratio on life expectancy for combinations of different age groups and years (304 separate regressions). The countries included in the computation of the reference population are Algeria, Angola, Argentina, Australia, Austria, Belgium, Benin, Bolivia, Brazil, Burkina Faso, Burundi, Cambodia, Cameroon, Canada, Chad, Chile, Colombia, Côte d’Ivoire, Cuba, Democratic Republic of the Congo, Denmark, Dominican Republic, Ecuador, Egypt, El Salvador, Eritrea, Ethiopia, Finland, France, Germany, Ghana, Greece, Guatemala, Guinea, Haiti, Honduras, Indonesia, Iran, Iraq, Israel, Italy, Japan, Jordan, Kazakhstan, Kenya, Kyrgyzstan, Laos, Libya, Madagascar, Malawi, Mali, Mexico, Morocco, Mozambique, Myanmar, Netherlands, Nicaragua, Niger, Nigeria, North Korea, Papua New Guinea, Paraguay, Peru, Philippines, Portugal, Rwanda, Senegal, Serbia, Sierra Leone, Somalia, South Africa, South Sudan, Spain, Sri Lanka, Sudan, Sweden, Switzerland, Syria, Tajikistan, Thailand, Togo, Tunisia, Turkey, Turkmenistan, Uganda, the UK, Tanzania, United States, Uzbekistan, Venezuela, Yemen, Zambia, and Zimbabwe. Counting missing women: areconciliation offlow andstock… Page 5 of 26 82 do not provide flow estimates by age, so comparisons of age patterns cannot be made. Further, our calculations confirm severe gender bias in mortality among adults in SubSaharan Africa, likely linked to elevated mortality of women due to AIDS, although it is of smaller magnitude than the original flow estimates. These results are important because missing women estimates guide policy-making. As evident from the World Development Report on “Gender Equality and Development,” the flow of missing women poses a relevant metric to highlight grievances in gender equality in developing countries as they allow for comparisons by age, disease, and over time. The World Development Report concludes, “Excess female mortality is slowly shifting from early childhood in South Asia to adulthood in Sub-Saharan Africa, declining in all low-income countries except in Sub-Saharan Africa” (World Bank 2012, p. 77). The report further points out that some of the results were not detected byor are even in contrast with earlier stock measure estimates: “Less well known is that excess female mortality is a continuing phenomenon beyond childhood and a growing problem in Sub-Saharan Africa. […] While missing girls at birth are indeed concentrated in India and China, consistent with the earlier discussion, excess female mortality after birth is highest in Sub-Saharan Africa, the only region where the numbers are going up over time” (World Bank 2012, pp. 118–120). While our flow estimates differ from the previous ones in the size and age patterns of missing women, they do reinforce that the number of missing women aged 15 to 25 in Sub-Saharan Africa is large and should be a subject of future research. This paper is organized as follows. In Sect.2, we present the conceptual framework of the flow measure of missing women. In Sect.3, we discuss the problem of low-mortality reference populations and the sex ratio as a functional form of the reference standard, and present estimates of missing women using alternative reference populations and functional forms. In Sect.4, we discuss the relevance of sex-specific differences within diseases across countries and present estimates of missing women using an alternative reference standard for AIDS. In Sect.5, we draw attention to the issue of lacking mortality data in Sub-Saharan Africa. Section6 concludes. 2 The flow measure ofmissing women Using the flow measure, excess female mortality ( EFM ) for age group a can be calculated the following way: MRf(a) and MRm(a) refer to the actual mortality rates of females and males in age group (a) , respectively.  MR f (a ) and  MRm(a) describe the mortality rates of females and males in age group (a) in the reference population, respectively.  MRm (a)∕  MR f (a ) is the male-to-female sex ratio of the reference population and is EFM (a)= ( MRf(a)− ( MRm(a)  MR m (a)∕  MR f (a) )) ∗Popf(a ) C.Ebert et al. 82 Page 6 of 26 used as a correction factor to scale the actual male mortality rate MRm(a) to account for, for example, differences in mortality from risky behavior or specific diseases across sexes. The scaled MRm(a) provides the reference mortality rate of females in age group a which is subtracted from the actual mortality rate of females to receive the population share of missing women in age group a . Multiplying the population share of missing women in age group a with the number of women in age group a , Popf(a), results in the total flow of missing women in age group a , EFM(a) . Based on Eq. (1), the essential ingredients to calculate excess female mortality are the actual age-specific mortality rates and the correction factor consisting of the agespecific male-to-female mortality rate ratio of the reference population. As a reference population, the flow measure of missing women uses high-income countries in 2000. The reference population and the functional form of the correction factor, i.e., dividing by the sex ratio in mortality, together build the reference standard. Table1 presents the mortality rates for adults aged 15 to 59 by sex for Sub-Saharan Africa, India, and China in 2000 as actual mortality rates ( MRf(15 −59) and MRm(15 −59) ) and for high-income countries in 2000 as reference mortality rates (  MRm (15 −59 ) and  MRf (15 −59 ) ). The actual sex ratios of mortality rates range between 1.13 and 1.46. Adult males thus die at 13–46% higher rates than adult females. The sex ratio in high-income countries in 2000 is about 1.92. Thus, adult males die at 92% higher rates than adult females. This sex ratio in high-income countries constitutes the key component of the flow measure’s reference standard and the estimates of missing women will be highly sensitive to this reference standard. Following Eq.(1) and using high-income countries in 2000 as a reference population, Table2presents the flow of missing women in Sub-Saharan Africa, India, and China in 2000 (Anderson and Ray 2010).7 With this reference standard, the problem of missing women as a share of the female population is worst in Sub-Saharan Africa. 0.47% of females died in excess in 2000 in Sub-Saharan Africa, compared to 0.35% in India and 0.28% in China. Table2further illustrates that in India and China, there is a massive problem of missing women among adults that is larger than among children. At last, adding the total number of missing women across Sub-Saharan Africa, India, and China results in 5 million females that went missing in 1 year. The extent to which these excess female deaths are entirely attributable to direct gender discrimination, rather than other causes, has gotten careful attention: “it is possible that certain aspects of poverty can create ‘unintended’ gender biases relative to the developed-country benchmark. The accounting methodology that we follow is entirely silent on matters of interpretation. The case for (or against) lack of similar care has to be made separately” (Anderson and Ray 2010, p. 1292). Also, this paper is silent about the causes underlying a certain number of missing women; instead, it discusses the estimation of the number itself. 7 For example, in Sub-Saharan Africa, the mortality rate for females aged 1–5 is 1.39%, while the male one is 1.27% (Anderson and Ray 2010, Table1). Since in high-income countries the male–female sex ratio is 1.25 in that age group (Anderson and Ray 2010, Table1), the number of missing females aged 1–5 in Sub-Saharan Africa is (0.0139 − (0.0127/1.25))×42m. = 0.16 million as shown in Table1. Counting missing women: areconciliation offlow andstock… Page 7 of 26 82 3 Low mortality reference populations andthesex ratio asafunctional form ofthereference standard 3.1 The problem The flow measure of missing women employs male-to-female age-specific mortality rate ratios that prevail in high-income countries in 2000 as a reference standard. The main shortcoming of this reference standard is the very low overall mortality in high-income countries. In very low mortality environments, even the smallest differences in the numbers of deaths across sexes result in very high mortality sex ratios and therefore low female reference mortality rates and high flows of missing women. In the adult age group of 15to 29-year-olds, which accounts for most of the estimated flow of missing women in Sub-Saharan Africa (see Table2), the female mortality rate in high-income countries in 2000 is 0.42 per thousand, whereas that for men is slightly higher at 1.08 per thousand, resulting in a sex ratio of 2.6; i.e., the probability of men dying in this young age group is 160% higher than the corresponding likelihood for women. By contrast, the female mortality rate ratio of 25to 29-year-olds in Sub-Saharan Africa is 14.8, and that of males is 10.8. In Fig.1, we simulate how the mortality sex ratio of 25to 29-year-olds in the reference population would fall as the mortality level increases from very low mortality levels prevailing in high-income countries to high levels prevailing in Sub-Saharan Africa, while the difference in mortality rates between sexes remains constant. Figure1 also shows the number of missing women associated with these mortality levels and sex ratios. Increasing the female mortality rate of 25to 29-year-olds only slightly from 0.42 to 2.42 reduces the sex ratio by half from 2.6 to 1.3, and the number of missing women falls from 258,118 to 152,434, a difference of over 100,000 just in the age group of 25to 29-year-olds alone. Further increases in the mortality level have much smaller impacts on the number of missing women because the mortality sex ratio is increasingly less sensitive as mortality levels rise. Increasing the female mortality rate of 25to 29-year-olds all the way to the level prevailing in Sub-Saharan Africa, which is 14.8, reduces the number of missing women to below 110,000.8 This sensitivity of the number of missing women is particular to the flow measure because the stock measure does not rely on the number of deaths but on the number of males and females surviving, which make up, of course, much larger numbers. The above calculations illustrate the sensitivity of the number of missing women with respect to the mortality level for 25to 29-year-olds only, which is the age group with the highest mortality sex ratio. Figure2 depicts the maleto-female mortality ratios for all age groups in high-income countries in 2000. The ratios are particularly high – above 2 – in the age groups between 15 and 35. Between the ages of 35 and 55, the mortality ratios are somewhat smaller, 8 At high mortality levels, also the difference between the male and female mortality rate is often larger. In Sub-Saharan Africa, the difference is 3.9. If we were to use that difference instead, or increasing the difference from 0.66 to 3.9 as we increase the mortality level from 0.42 to 14.8, the implications of this simulation exercise would not change. C.Ebert et al. 82 Page 8 of 26 averaging around 1.86. A final peak arises between the ages of 55 and 70, with mortality rate ratios over 1.94. For older ages, the ratio falls sharply, but always remains above 1. The first peak between 15 and 35 is largely a result of the role of injuries which particularly drives up male mortality rates in these young age groups; in the demographic literature, this has sometimes been referred to as the “testosterone spike” linked to injuries related to the combination of alcohol abuse, dangerous traffic behavior, and violence among young men (Kalben 2000). Because of the overall low mortality among 15to 29-year-olds in high-income countries, deaths from injuries are particularly prominent. In this young age group, injuries make up 71% of all deaths of males (and 51% of females), and the sex ratio of mortality from injuries is 3.7, whereas it is 1.6 for all other deaths. Therefore, the reference standard is heavily affected by this disproportionate importance of injuries in adult mortality for males in high-income countries. The second peak is to a large extent caused by the mortality consequences of high rates of smoking among adults. As shown in detail by Crimmins etal. (2011), what matters for mortality are smoking rates two to four decades prior, i.e., smoking Table 1 Adult mortality rates (15–59) by sex in 2000 Notes: Adult mortality rates refer to the likelihood of having died by age 60 if one was alive at age 15. Male refers to the adult (15–59) mortality rate among males. Female refers to the adult (15–59) mortality rate among females. Sex ratio refers to the ratio of columns (1) and (2), i.e., the male-to-female mortality rate ratios. To derive mortality rates in Sub-Saharan Africa and high-income countries, population (15–59 years old) weighted averages of country specific adult mortality rates were calculated (see supplementary materials). Sub-Saharan African countries include: Algeria, Angola, Benin, Botswana, Burkina Faso, Burundi, Cameroon, Cape Verde, CAR, Chad, Comoros, Congo, Cote d’Ivoire, DRC, Equatorial Guinea, Eritrea, Ethiopia, Gabon, Gambia, Ghana, Guinea, Guinea-Bissau, Kenya, Lesotho, Liberia, Madagascar, Malawi, Mali, Mauritania, Mauritius, Mozambique, Namibia, Niger, Nigeria, Rwanda, Sao Tome and Principe, Senegal, Seychelles, Sierra Leone, South Africa, Swaziland, Tanzania, Togo, Uganda, Zambia, Zimbabwe. Highincome countries include: Australia, Austria, Belgium, Canada, Denmark, Finland, France, Germany, Iceland, Ireland, Italy, Japan, Luxembourg, Netherlands, New Zealand, Norway, Portugal, Spain, Sweden,Switzerland, United Kingdom, United States Sources: Country specific adult mortality rates in 2000 were taken from World Health Organization (2001). Population numbers were taken from United Nations, Department of Economic and Social Affairs, Population Division (2000) Male Female Sex ratio (1) (2) (3) Sub-Saharan Africa 0.511 0.454 1.125 India 0.287 0.213 1.347 China 0.161 0.110 1.464 High-income countries 0.125 0.065 1.918 Counting missing women: areconciliation offlow andstock… Page 15 of 26 82 mortality from AIDS than men. The stark difference in sex ratios of AIDS mortality across regions emerges because AIDS incidence and fatalities in high-income countries are mostly driven by gay men and IV drug users (also predominantly men), whereas heterosexual transmission plays a significant role in Sub-Saharan Africa (Ainsworth and Over 1997). Sex-specific differences in mortality across countries also exist, as previously discussed, for deaths from risky behaviors. However, whereas sex-specific differences in risky behavior are likely similar in Sub-Saharan African countries and high-income countries, the patterns for other diseases, such as AIDS, evidently diverge to a large extent. To account for differences in disease environments, the original flow measure neglects sex-specific differences and builds up the number of missing women disease group by disease group. Specifically, the sex ratio of age-specific mortality rates for each disease group in high-income countries is used as the reference Table 3 Number of missing women (in 1000 s) by age group using alternative reference standard Notes: Ratio refers to the functional form of the reference standard that divides the male mortality rate in the region of interest by the sex ratio of mortality rates in the reference population. Diff. refers to the functional form of the reference standard that subtracts the difference in sex-specific reference population mortality rates from the male mortality rate in the region of interest. RC refers to high-income countries in 2000; MLT are Princeton Model Life Tables West. Numbers do not sum to total because of rounding Sources: Population and death numbers by age for Sub-Saharan Africa, India, and China, and the reference population of high-income countries in 2000 were taken from the Supplementary Materials 2 of Anderson and Ray (2010). Mortality rates for Princeton Model Life Tables West were taken from Coale etal. (1983). To determine the correct mortality level in Coale etal. (1983), life expectancy at birth for Sub-Saharan Africa, India, and China in 2000 was taken from United Nations, Department of Economic and Social Affairs, Population Division (2019) Sub-Saharan Africa India China RC MLT RC MLT RC MLT RC MLT RC MLT RC MLT Ratio Ratio Diff Diff Ratio Ratio Diff Diff Ratio Ratio Diff Diff (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) At birth 0 0 0 0 184 184 184 184 644 644 644 644 0–1 32 22 − 235 13 146 181 8 156 109 148 65 119 1–4 160 55 55 56 164 140 121 141 23 29 13 20 5–14 70 5 18 2 93 71 53 69 2 12 − 9 4 15–29 578 241 279 237 258 111 127 113 24 − 9 − 9 − 37 30–44 345 − 3 − 56 − 51 93 − 70 − 63 − 80 73 − 2 53 − 16 45–59 84 − 55 − 144 − 93 120 − 39 − 114 − 75 89 − 9 15 − 26 60–79 213 − 40 − 18 − 40 541 76 103 41 490 27 170 12 80 + 44 − 10 10 − 8 113 21 84 39 272 64 165 64 Total 1526 213 − 91 116 1712 674 503 590 1727 903 1107 783 % of females 0.47 0.07 − 0.03 0.04 0.35 0.14 0.10 0.12 0.28 0.14 0.18 0.13 C.Ebert et al. 82 Page 16 of 26 standard, and the ageand disease-specific flows are then summed up over all ages and diseases to achieve the total number of missing women in a year.19 Building up the flow of missing women disease by disease only has minor implications for the estimated total number of missing women. In Sub-Saharan Africa, the annual flow falls from 1.526 million to 1.385 million, in India from 1.712 to 1.637 million, and in China from 1.727 to 1.592 million (Anderson and Ray 2010). The key findings of the overall high magnitude (4.6 million per year), the preponderance among adults, and the larger share of missing women in Africa are not affected significantly by this disease correction. The reference standard used in this disease correction presumes a counterfactual that without gender bias, AIDS in Sub-Saharan Africa would have also been primarily a disease affecting gay men and IV drug users and suggests that 58% of all adult missing women in Sub-Saharan Africa die from AIDS (see Table4, column (3)). This counterfactual is likely false. AIDS in Africa has been a disease that has spread mostly through heterosexual intercourse (and mother to child transmission) due to the early emergence of the disease in Africa and its spread among heterosexual couples long before it was identified as a disease, together with a very high incidence of STDs (Iliffe 2006; Ainsworth and Over 1997; Oster 2005). It is likely to be true that the particularly high incidence and fatality of AIDS among young women is to some extent due to gender discrimination (e.g., unwanted sexual intercourse with older men), but this is not the main reason why AIDS is a disease affecting mostly heterosexual people in Sub-Saharan Africa across adult age groups. Assuming the ratio from high-income countries as a reference standard likely overstates the relevance of AIDS in causing missing women considerably. To test the sensitivity in the number of missing women to alternative assumptions for an AIDS-specific reference standard in environments with a predominantly heterosexual transmission path, we assume a sex ratio of mortality in high-income countries that weighs mortality rates by the share of heteroand homosexual populations. For example, the sex ratio of HIV among heterosexual men (84.1%) and women (86.9%) as well as homosexual men and women (defined as those who are not heterosexual) in the US, as one of the established market economies, is around 1 (Lansky etal. 2015; Centers for Disease Control and Prevention 2019).20 This number is much smaller than the actual male-to-female mortality ratio from AIDS of 4 prevailing in most adult age groups of high-income countries in 2000. The sex ratio of AIDS mortality of 1 would still allow for high female mortality from AIDS in the 15–29 age group—in which gender discrimination in the transmission of AIDS 19 When female deaths from a disease were less than 100 in the considered age group, Anderson and Ray (2010) use average disease group-specific reference mortality ratios, rather than the mortality ratio for that one disease. When deaths were essentially zero in the reference population, such as STDs among children in developed countries, a reference sex ratio of 1:1 was used (Anderson and Ray 2010, p.1279). 20 Specifically, the sex ratio is calculated as: sexratio = (sharehetero.men∗HIVcasesper100,000hetero.men)+(sharehomo.men∗HIVcasesper100,000homo.men) (sharehetero.women∗HIVcasesper100,000hetero.women)+(sharehomo.women∗HIVcasesper100,000hetero.women) . Counting missing women: areconciliation offlow andstock… Page 17 of 26 82 through forced sexual intercourse likely matters—as the male-to-female ratio of deaths from AIDS among 15to 29-year-olds is only 0.53 in Sub-Saharan Africa.21 Table4presents the flow of missing women from AIDS using a sex ratio of mortality from AIDS of 1 in columns (4) and (5) and a sex ratio of 1.2 as a more conservative reference standard for AIDS in columns (6) and (7). The flow of missing women from AIDS reduces by about 400,000 excess female deaths, from 611,000 to 19,000–171,000, essentially all accruing in the 15–29 age group. Further, the share of adult women that go missing due to AIDS reduces from 58% to a more realistic percentage share of 4 to 28%. 4.2 Maternal mortality, injuries, andsuicide There are numerous other diseases in addition to AIDS that require attention when the reference population is not drawn from a comparable disease environment. Some diseases are particularly rare in high-income countries, which renders the sex ratio of mortality prevailing in high-income countries an overall unsuitable reference standard for that particular disease. As there is no such thing as paternal mortality that can be used to derive expected maternal mortality, the flow measure of missing women assumes that any maternal death above the rate prevailing in high-income countries is an excess death. Because maternal mortality is rare in high-income countries, maternal mortality among 15to 45-year-olds accounts for the second most excess female deaths in Sub-Saharan Africa (after AIDS, which would change if the alternative reference standards for AIDS were used) and for most excess female deaths in India. Hence, the assumption of all maternal deaths being excess female deaths is a determining factor for the total number of missing women and the distribution of missing women across age groups in Sub-Saharan Africa and India. Maternal mortality is typically related to the poor overall health conditions and services prevailing in many developing countries. Figure3 presents the relationship between child mortality rates of both sexes, which is an indicator of overall health conditions in a country, and the maternal mortality ratio. Child and maternal mortality show a close fit, indicating that maternal mortality goes hand-in-hand with poor overall health. If Sub-Saharan African countries were outliers in this relationship, then this might be due to gender inequality in health access that goes beyond poor overall health conditions. The dashed lines in Fig.3 present the predicted maternal mortality ratios based on a simple linear regression model of maternal mortality on child mortality and a dummy for Sub-Saharan Africa. The results suggest that on average, about 11–16% of the maternal mortality in Sub-Saharan Africa is not 21 While partly this high mortality could indeed be due to discrimination linked to unwanted sexual intercourse of young women with older men, part of the high female mortality could also be due to earlier voluntary sexual activities of females (and thus earlier AIDS risks) which is observable in most countries of the world, including high-income countries; see World Bank (2012) for a discussion. C.Ebert et al. 82 Page 18 of 26 explained by poor health conditions and might be due to gender bias.22 Thus, a small number of deaths from maternal mortality—around 22,000 to 32,000 rather than all 200,000 maternal deaths—might indeed be due to gender inequality. Another noteworthy cause of death is injuries, specifically suicides and fires, for which differences in the distribution of deaths by sex potentially account for a substantial number of missing women. The sex ratio of deaths from suicides in India is relatively balanced and likely driven by the economic hardship of families (especially in rural areas), whereas deaths from suicides in high-income countries Table 4 Number of missing women from AIDS (in 1000 s) by age group using alternative reference standards Notes: Column (1) presents the total number of missing women when calculated disease group by disease group as presented in Table 6 of Anderson and Ray (2010). Columns (2) to (7) refer to different reference standards that were applied in the estimation of missing women from AIDS. To calculate the % of total in columns (5) and (7), the numbers in columns (4) and (5) were divided by the new total number of missing women that were corrected for the number of missing women from AIDS, i.e., 180/(406-(277–180)) = 0.58. Figures in columns (1), (2), (4), and (6) are rounded to the nearest thousand. Percentage shares in columns (3), (5), and (7) are based on numbers rounded to nearest thousand in columns (1), (2), (4), and (6). Numbers do not sum to total because of rounding Sources: Anderson and Ray (2010) Table6 for column (1). Population numbers and death numbers from AIDS by age for Sub-Saharan Africa and the reference population of high-income countries in 2000 were taken from the Supplementary Materials 2 of Anderson and Ray (2010) Sub-Saharan Africa: Flow of missing women Total From AIDS Reference: sex ratio, high-income countries Method: disease-bydisease group Reference: sex ratio, high-income countries Reference: sex ratio = 1 Reference: sex ratio = 1.2 (1) (2) (3) (4) (5) (6) (7) Age No No % of total No % of total No % of total 15–29 406 277 68% 180 58% 213 62% 30–44 278 240 86% − 91 – − 8 – 45–59 134 78 58% − 58 – − 29 – 60–79 217 15 7% − 11 – − 5 – 80 + 25 0 – 0 – 0 – Total 1,060 611 58% 19 4% 171 28% 22 11–16% is the confidence interval of the mean of the Sub-Saharan Africa fixed effect (81 deaths) divided by the predicted maternal mortality ratio (mean = 13.5%, CI = 10.8–16.2%). Counting missing women: areconciliation offlow andstock… Page 19 of 26 82 are male-biased and driven by psychiatric conditions (e.g.,Aditjanyee 1986; Hiroeh etal. 2001; Mortensen etal. 2000; Nordentoft etal. 1993; Qin etal. 2000). Similarly, deaths from fires are heavily male-biased in high-income countries (sex ratio of adult mortality (15–59) of 2.1), whereas in India, fires afflict relatively more women (sex ratio of adult mortality (15–59) of 0.44). An estimated 106,000 out of 163,000 fire-related deaths in India in 2001 accrued to women (Sanghavi etal. 2009; Bhalla and Sanghavi 2020). It was further estimated that more women died from fires than from giving birth and at a rate 20 times more than in the rest of the world. These fire-related deaths are said to be due to the use of open fires, candles, and kerosene during domestic chores as well as self-immolation and domestic violence (Sanghavi etal. 2009). Thus, a considerable share of fire-related missing women is likely due to gender bias.23 Yet, the flow of adult (> 15 years) missing women from fires using the sex ratio from high-income countries as a reference standard is 72,000 in India, which makes up over 50% of missing women from unintentional injuries and 80% of all fire-related deaths of adult women would have to be due to gender discrimination, which seems unlikely. The discussion illustrates that when using the sex ratio of mortality prevailing in today’s high-income countries as a reference standard in the estimation of missing women, disease corrections must address the differences in the sex-specific disease environments between high-income countries and China, India, and Sub-Saharan Africa. Differences in disease environments may be based on different structures within disease groups (as in AIDS or injuries) or can be largely a result of the overall worse disease environment in developing countries (as is the case with maternal mortality). Overall, the necessity for a detailed disease correction depends on the chosen reference population. The more appropriate the general reference population is, the less such a disease correction is required. For example, estimates using the Princeton Model Life Tables are much more similar to the within-disease corrected flow of missing women using high-income countries as a reference standard already because the Princeton Model Life Tables offer reference mortality rates for mortality environments that are similar to the country or region of interest. 5 Availability andquality ofmortality data fromSub‑Saharan Africa Putting the methodological question of appropriate reference standards aside, an important reason why flow estimates of missing women have not been produced before is the lack of reliable data on age-specific mortality rates by sex in many developing countries. Disease-specific data are even harder to come by as they require a complete vital registration system. 23 Due to death from interpersonal violence and women would also be less likely to conduct household chores if the division of domestic and wage work was equal across sexes. C.Ebert et al. 82 Page 20 of 26 Our analysis follows previous estimations of the flow of missing women and uses mortality data for the year 2000 from the World Health Organization’s 2002 Global Burden of Disease estimates (Anderson and Ray 2010; World Bank 2012).24 A detailed description of data sources and methods used in the derivation of the WHO mortality data is provided in Mathers etal. (2004; see text, Table4, Annex Table 6, and Annex Table7). While the majority of countries have fairly recent data on infant and child mortality from DHS and related surveys (the modal time period being 1995–1999; see Mathers etal. Table3), data to estimate adult mortality are often lacking. In Sub-Saharan Africa, adult mortality estimates for 2000 are based on observations from only 4 out of 46 countries, two each for the early and the late 1990 s (see Mathers etal. 2004, Table2). In fact, reliable vital registration data are available for only one country in Sub-Saharan Africa, which is the island of Mauritius (see Mathers etal. 2004, Annex Table6). Although China and India also lack full vital registration systems, in China, age-specific deaths could be estimated from a sample vital registration and the 2000 census, and in India, data from Fig. 3 Cross-country relationship between child mortality (x-axis) and maternal mortality (y-axis). Notes: Child mortality is measured as the probability of dying between birth and exact age 5, expressed per 1000 live births, while the maternal mortality ratio (MMR) is the number of maternal deaths per 100,000 live births. Predicted maternal mortality ratio is based on a linear prediction of maternal mortality on child mortality and a Sub-Saharan Africa dummy (both predictions result from the same model). Sources: United Nations Inter-agency Group for Child Mortality Estimation (2021) for child mortality in 2000. Hogan etal. (2010) for maternal mortality ratio in 2000. United Nations, Department of Economic and Social Affairs, Population Division (2019) for assignment of countries to SDG region Sub-Saharan Africa and to rest of world 24 We directly use the ageand disease-specific mortality data of Anderson and Ray (2010) as provided in their supplementary materials. Counting missing women: areconciliation offlow andstock… Page 21 of 26 82 a project vital registration system were available (see Mathers etal. 2004, pp. 10/11 and Annex Table6). Since the Global Burden of Disease estimates in 2002, there have been multiple revisions and updates of the WHO life tables, the latest update being from December 2019. However, the availability of quality vital registration data has not changed since 2002.25 Although a few more countries, for example, Cabo Verde and South Africa, now collect and report vital registration data to the World Health Organization, they were excluded due to insufficient quality (World Health Organization, 2020 a,b). In consequence, even the most recent WHO mortality data for Sub-Saharan Africa only include high-quality vital registration data for Mauritius. In the absence of vital registration data, ageand disease-specific mortality rates for Sub-Saharan Africa have to be estimated. This requires accurate calculations of overall age-specific mortality rates to provide the “envelope” for total deaths that can then be attributed to different disease categories (Mathers etal. 2004). When data on child mortality (the likelihood of reaching age five) and adult mortality (the likelihood of reaching age 60 given that one has reached age 15) are available, modified logit life tables can be used to estimate mortality rates for all age groups (Murray etal. 2003a, b a, b). Murray etal. (2003a, b) demonstrate that the prediction method performs well in simulations if reliable data on child mortality and adult mortality for each sex are available; as the relationship between child and adult mortality differs between countries and across regions (see also Coale etal. 1983), these two pieces of information are crucial to generate the full set of age‐ and sex-specific mortality rates. Because only child mortality data was available for the year 2000 in Sub-Saharan Africa, the corresponding levels of adult mortality were essentially assumed, using historical model life tables, as explained by Mathers etal. (2004, p. 11): “Based on the predicted level of child mortality in 2002, the most likely corresponding level of adult mortality (excluding HIV/AIDS deaths where necessary) was selected, along with uncertainty ranges, based on regression models of child versus adult mortality as observed in a set of almost 2000 life tables judged to be of good quality.”26 There is no independent way of confirming these adult mortality rates generated by the life tables, let alone the relationship between male and female adult mortality rates. In more recent years, there has been some progress in the estimation of adult mortality rates (15–59). Since 2012, sex-specific adult mortality estimates of the Global Burden of Disease studies are based on sibling survival data in countries where no vital registration data are available, which in the case of Sub-Saharan Africa mostly come from Demographic and Health Surveys (Obermeyer etal. 2010; Wang etal. 2020, see supplementary appendix 1, Sect. 2.3 and appendix Table1; World Health Organization, 2020a,b). Albeit not perfect, these adult mortality estimates are based on actual data from Africa and therefore allow for a more reliable estimation of 25 See https:// www. who. int/ data/ gho/ data/ themes/ morta lityandglobalhealthestim ates/ downl oadtherawdatafilesofthewhomorta litydatab ase for “Availability Contains an Excel file with the list of countries-years available for the mortality and population data. Last updated: 15 December 2019.” 26 Note that the 2000 life tables referred to here and further down in the paper by Murray etal. (2003a, b a, b) are from all over the world, not from Africa alone. C.Ebert et al. 82 Page 22 of 26 age-specific mortality rates using modified logit model life tables (Murray et al. 2003a,b; Wang etal. 2020, 2012, 2014; World Health Organization, 2020a,b). The enormous work of the Global Burden of Disease project in estimating ageand disease-specific mortality rates over the past decades is of great value and has enabled the measurement of the flow of missing women in addition to numerous other research. Still, the absence of vital registration systems makes it impossible to come up with precise age and disease-specific mortality rates. The availability of vital registration data would immensely improve the measurement of the age and disease-specific flow of missing women. When the flow of missing women is estimated using more recent or, potentially, in the future, using vital registration data, it is important to understand that this comparison over time will be confounded by the changes in the measurement of mortality. In contrast, the stock estimates of missing women rely on much more widely available (and reliable) census information and can be used as a reference point. 6 Conclusions The flow measure of missing women calculates excess female mortality by age and disease that occurs in a certain time period, whereas the previous literature focused largely on calculating the stock measure of missing women, i.e., the deficit among alive women at one point in time. Estimations of excess female mortality by Anderson and Ray (2010) and World Bank (2012) suggest that gender bias in mortality is much larger than previously found (about 4 to 5 million excess female deaths per year vs. around 100 million missing women in total in 2000), is as severe among adults as it is among children in India, and is larger in Sub-Saharan Africa than in China and India. However, these estimates depend on particular methodological choices with respect to the reference standard and the method of disease correction. If alternative, potentially better-suited, reference standards are used, the results align with those of the previous literature on the stock measure of missing women. The stock measure results show that more women are missing in China and India than in Sub-Saharan Africa and that, in China and India, missing women are largely attributable to excess female mortality before birth and/or the first few years of life. In addition to the question of appropriate reference standards, a lack of vital registration data in most developing countries makes it difficult to estimate precise numbers of excess female deaths. Gender bias in mortality is a serious issue that deserves urgent attention by policymakers. According to our calculations, 1.5 to 1.8 million females perish in excess every year. Although this number is much lower than the original flow estimates of 4 to 5 million yearly excess female deaths, our results suggest that the focus of the literature and associated policy proposals on South Asia and China, and on pre-birth and young children, is justified. Our estimates confirm high excess female mortality in Sub-Saharan Africa among young adults, mainly driven by excess mortality from AIDS and maternal mortality, which provides some intriguing further detail to the issue of gender bias in mortality in Sub-Saharan Africa studied so far (e.g.,Klasen 1996a, b; Svedberg Counting missing women: areconciliation offlow andstock… Page 23 of 26 82 1996; Klasen and Wink 2002, 2003; Ainsworth and Over 1997). Further investigation of gender bias in this age group in Sub-Saharan Africa should be a priority, as should be continuing efforts to improve vital data from developing countries to assess excess female mortality more accurately in the future. Acknowledgements We would like to thank Jörg Ankel-Peters, Jere Behrman, Sonia Bhalotra, Monica Das Gupta, Angus Deaton, Ken Harttgen, Alan Lopez, Ashok Rai, Debraj Ray, participants at seminars and workshops at Harvard, UPenn, Tufts, Dalhousie, and Göttingen, and editor Kompal Sinha and two anonymous reviewers for helpful comments and discussion. The authors declare that they have no relevant or material financial interests that relate to the research described in this paper. Funding Open Access funding enabled and organized by Projekt DEAL. Data Availability The data and reproduction package for this article are available in the supplementary materials. Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/ licenses/by/4.0/. References Abrevaya J (2009) Are there missing girls in the United States? Evidence from birth data. Am Econ J Appl Econ 1(2):1–34. https:// doi. org/ 10. 1257/ app.1. 2.1 Aditjanyee DR (1986) Suicide attempts and suicides in India: cross-cultural aspects. Int J Soc Psychiatry 32:64–73. https:// doi. org/ 10. 1177/ 00207 64086 03200 208 Ainsworth M, Over AM (1997) Confronting AIDS: public priorities in a global epidemic. World Bank Policy Research Report, No. 17285, Washington DC, The World Bank. http:// docum ents. world bank. org/ curat ed/ en/ 21121 14687 79168 446. Accessed 13 Oct 2025 Akbulut-Yuksel M, Rosenblum D (2023) Estimating the effects of expanding ultrasound use on sex selection in India. J Dev Stud. https:// doi. org/ 10. 1080/ 00220 388. 2022. 21541 52 Anderson S, Ray D (2010) Missing women: age and disease. Rev Econ Stud 77(4):1262–1300. https:// doi. org/ 10. 1111/j. 1467937X. 2010. 00609.x Anderson S, Ray D (2012) The age distribution of missing women in India. Econ Political Wkly47(47/48):87–95. https:// www. jstor. org/ stable/ 41720 413. Accessed 13 Oct 2025 Anderson S, Ray D (2019) Missing unmarried women. J Eur Econ Assoc 17(5):1585–1616 Anukriti S (2018) Financial incentives and the fertility-sex ratio trade-off. Am Econ J Appl Econ 10(2):27–57. https:// doi. org/ 10. 1257/ app. 20150 234 Asfaw A, Klasen S, Lamanna F (2010) Gender gaps in parents’ financing strategy for hospitalization of their children: empirical evidence from India. Health Econ 19(3):265–279. https:// doi. org/ 10. 1002/ hec. 1468 Barcellos SH, Carvalho LS, Lleras-Muney A (2014) Child gender and parental investments in India: are boys and girls treated differently? Am Econ J Appl Econ 6(1):157–189. https:// doi. org/ 10. 1257/ app.6. 1. 157 Bhalla K, Sanghavi P (2020) Fire-related deaths among women in India are underestimated. Lancet 395(10226):779–780. https:// doi. org/ 10. 1016/ S01406736(20) 30060-X C.Ebert et al. 82 Page 24 of 26 Bongaarts J, Guilmoto CZ (2015) How many more missing women? Excess female mortality and prenatal sex selection, 1970–2050. Popul Dev Rev 41(2):241–269. https:// doi. org/ 10. 1111/j. 17284457. 2015. 00046.x Centers for Disease Control and Prevention (2019) Estimated HIV incidence and prevalence in the United States, 2010–2016. HIV Surveillance Supplemental Report 24 (No. 1). http:// www. cdc. gov/ hiv/ libra ry/ repor ts/ hivsurve illan ce. html. Accessed 17 Mar 2021 Coale AJ (1991) Excess female mortality and the balance of the sexes in the population: an estimate of the number of “missing females.” Popul Dev Rev 17(3):517–523. https:// doi. org/ 10. 2307/ 19719 53 Coale AJ, Banister J (1994) Five decades of missing females in China. Demography 31(3):459–479. https:// doi. org/ 10. 2307/ 20617 52 Coale AJ, Demeny P, Vaughan B (1983) Regional model life tables and stable populations, 2nd edn. Academic Press, New York. https:// doi. org/ 10. 1016/ C2013-007295-7 Crimmins EM, Preston SH, Cohen B (2011) Explaining divergent levels of longevity in high-income countries. Washington, D.C., The National Academies Press. https:// doi. org/ 10. 17226/ 13089 Das Gupta M (1987) Selective discrimination against females in rural Punjab. India Population and Development Review 13(1):77–100. https:// doi. org/ 10. 2307/ 19721 21 Das Gupta M (2005) Explaining Asia’s “missing women”: a new look at the data. Popul Dev Rev 31(3):529–535. https:// doi. org/ 10. 1111/j. 17284457. 2005. 00082.x Das Gupta M (2006) Cultural versus biological factors in explaining asia’s “missing women”: response to oster. Popul Dev Rev 32(2):328–332. https:// doi. org/ 10. 1111/j. 17284457. 2006. 00121.x Das Gupta M, Chung W, Shuzhou L (2009) Is there an incipient turnaround in Asia’s “missing girls” phenomenon?. Policy Research Working Paper Series 4846. The World Bank. https:// hdl. handle. net/ 10986/ 4040. Accessed 13 Oct 2025 Deaton A (2004) Health in an age of globalization. In: Collins S, Graham C (eds) Brookings Trade Forum. Washington, D.C., The Brookings Institution. https:// doi. org/ 10. 3386/ w10669 Dyson T, Moore M (1983) On kinship structure, female autonomy, and demographic behavior in India. Popul Dev Rev 9(1):35–60. https:// doi. org/ 10. 2307/ 19728 94 Ebert C, Vollmer S (2022) Girls unwanted - the role of parents’ child-specific sex preference for children’s early mental development. J Health Econ 82:102590. https:// doi. org/ 10. 1016/j. jheal eco. 2022. 102590 Hiroeh U, Appleby L, Mortensen PB, Dunn G (2001) Death by homicide, suicide, and other unnatural causes in people with mental illness: a population-based study. Lancet 358(9299):2110–2112. https:// doi. org/ 10. 1016/ S01406736(01) 07216-6 Hogan MC, Foreman K J, Naghavi M, Ahn S Y, Wang M, Makela SM, Lopez AD, Lozano R, Murray CJL (2010) Maternal mortality for 181 countries, 1980–2008: a systematic analysis of progress towards Millennium Development Goal 5. Lancet 375(9726):1609–1623. https:// doi. org/ 10. 1016/ s01406736(10) 60518-1 Iliffe J (2006) The African AIDS epidemic: a history. Ohio University Press, Athens. https:// doi. org/ 10. 14296/ RiH/ 2014/ 2038 Javadekar S, Saxena K (2025) The seen and unseen: the unintended impact of a conditional cash transfer program on prenatal sex selection. J Popul Econ 38:32. https:// doi. org/ 10. 1007/ s0014802501091-6 Jayachandran S, Kuziemko I (2011) Why do mothers breastfeed girls less than boys? Evidence and implications for child health in India. Q J Econ 126(3):1485–1538. https:// doi. org/ 10. 1093/ qje/ qjr029 Jayachandran S, Rohini P (2017) Why are Indian children so short? The role of birth order and son preference. Am Econ Rev 107(9):2600–2629. https:// doi. org/ 10. 1257/ aer. 20151 282 Johansson S (1984) Deferred infanticide: excess female mortality in childhood. In: Hausfater G, Hrdy S (eds) Infanticide. Aladin, New York, pp 463–486 Johansson S (1991) Welfare, mortality and gender. Continuity and change in the explanation of male/ female mortality differentials over three centuries. Contin Chang 6(2):135–177. https:// doi. org/ 10. 1017/ S0268 41600 00013 14 Jones TR, Millington MJ, Price J (2023) Changes in parental gender preference in the USA: evidence from 1850 to 2019. J Popul Econ 36(4):3057–3070. https:// doi. org/ 10. 1007/ s0014802300957-x Kalben BB (2000) Why men die younger. N Am Actuar J 4(4):83–111. https:// doi. org/ 10. 1080/ 10920 277. 2000. 10595 939 Klasen S (1994) Missing women reconsidered. World Dev 22(7):1061–1071. https:// doi. org/ 10. 1016/ 0305750X(94) 90148-1 Klasen S (1996) Nutrition, health, and mortality in sub Saharan Africa: is there a gender bias? J Dev Stud 32(6):913–932. https:// doi. org/ 10. 1080/ 00220 38960 84224 46