SPECIAL ARTICLEjanuary 17, 2009 EPW Economic & Political Weekly56Tamil Nadu and the Diagonal Divide in Sex Ratios Sharada Srinivasan, Arjun S BediBetween 1961 and 2001, India’s 0-6 sex ratio has steadily declined. Despite evidence to the contrary, this ratio is often characterised in terms of a diagonal divide with low 0-6 sex ratios in northern and western India and normal 0-6 sex ratios in eastern and southern India. While unexpectedly high rates of female infant mortality have been reported in Tamil Nadu, it is still regarded as lying outside the ambit of states with unusually low 0-6 sex ratios. Based on an analysis of patterns in sex ratio at birth, infant mortality rates and under-5 mortality rates for Tamil Nadu, this paper traces the development of daughter deficit in the state and examines the validity of the diagonal divide in sex ratios across India. We find evidence of daughter deficit in more than half the state’s districts with a majority of the shortfall arising before birth. The evidence presented here, combined with earlier work on declining 0-6 sex ratios outside north-western India, suggests that the diagonal divide is no longer an appropriate distinction.We thank the Directorate of Public Health, government of Tamil Nadu for VES data, Anandhkumar for assistance, K Nagaraj, Sayeed Unisa and Sunita Kishor for information on data sources and a referee for comments.Sharada Srinivasan (sharada@yorku.ca) is with York University, Canada and Arjun S Bedi(bedi@iss.nl) is with Institute of Social Studies, The Netherlands. 1 IntroductionThe phenomenon of missing women is well-established in India. Based on Census 2001, Klasen and Wink (2003) estimate that about 39 million women are missing. Under-lying this estimate, the population sex ratio in India has declined from 972 in 1901 to 933 in 2001 and the 0-6 sex ratio, that is, the proportion of girls to boys has declined from 976 in 1961 to 927 in 2001 (Table 1, p 57). Various factors such as nutrition (Goodkind 1996), incidence of hepatitis-B virus (Oster 2005), and parental hormonal levels (James 1996), among others, may influence sex ratios at birth (SRB) and therefore the 0-6 sex ratio. In the Indian context, Oster (2005) argues that, in part, the level of the unexpectedly masculine SRB may be driven by the incidence of hepatitis-B virus while Jayaraj and Subramanian (2004) argue that, in part, the declining trend in the female-maleSRB may be attributed to improvement in women’s health status. While these explana-tions may account for some of the inequality in sex ratios, the differential survival of girls and boys due to practices like infanticide, neglect, and sex selective abortion cannot be denied. For instance, based on an analysis of SRB, Jha et al (2006) find that about 10 million female foetuses may have been aborted during 1985-2001.Conventionally, much of the daughter shortfall has been ac-counted for by states such as Punjab, Haryana and Delhi in the north and Gujarat and Maharashtra in the west. An early essay on India’s demography (Visaria 1967) demonstrated that sex ratios were persistently lower (i e, more masculine) in the northern states, and higher in the south. More than a decade later, Sopher (1980: 296) commented that the “completeness of the regional dichotomy is impressive”. Dyson and Moore (1983) confirmed this pattern and grouped the major Indian states into two broad demographic regimes divided by a diagonal line that approximated the contours of the Satpura hill range, extending eastward tojoin the Chhotanagpur hills of southern Bihar. Miller’s (1989: 1232) analysis of changes in juvenile sex ratios between 1961 and 1971 confirmed the diagonal divide. More recently, analysis of changes in the 0-6 ratio between 1991 and 2001 shows that, except for Lakshadweep, Sikkim, Mizoram, Tripura and Kerala, all states have registered a decline in the ratio, prompting the conclusion that “the traditional north-south divide stands significantly modified” (Agnihotri 2003: 4351). For example, Agnihotri (2003) shows that the 0-6 sex ratio for Orissa has declined from 971 in 1991 to 953 in 2001 and that in 2001, 12 districts had a 0-6 sex ratio lower than 952 as against two in 1991.1 A similar pattern is observed in other eastern states. 
SPECIAL ARTICLEEconomic & Political Weekly EPW january 17, 200957Nine out of 18 districts in West Bengal have an urban 0-6 sex ra-tio below 952 while it is 927 in Kolkata. Fourteen of 23 districts in Assam display an urban 0-6 sex ratio less than 952. Despite evidence that the diagonal divide may no longer be valid, there is a tendency to persist with this distinction. For instance, based on Census 2001 data onSRB, Bhat and Zavier conclude that theSRB in the southern and eastern states of India “is well within the range observed in normal circumstances”. They also note that, “a line drawn diagonally separating south and eastern India from north and western India would neatly demark the two re-gions of low and highSRB. Except for some isolated pockets (suchas around Salem district in Tamil Nadu)... (ibid: 2296, emphasis added)”. However, even based on their definition of a normal range, between 943 and 971, the prevailing SRB of 935 in Tamil Nadu (919 in rural Tamil Nadu) and of 926 in Orissa, clearly falls outside the normal range.2 Similarly, Bhaskar and Gupta (2007) note that south and east India exhibit international patterns in child sex ratio as against the unbalanced sex ratios in the northern and western states. While noting that “pockets of southIndiaare infamous for infanticide” (Patel 2007: 41, emphasisadded),arecent book on sex selective abortion does not contain anyessayonthe prevalence of the practice in southern and easternstates. Set against this background, this paper has three aims. First, to use data from Tamil Nadu to examine the validity of the diago-nal divide, second, to use information on SRB, infant mortality rates (IMR) and under-5 mortality rates (U5MR) to analyse state-level temporal patterns in daughter deficit and third, to provide a district-level analysis of daughter deficit. A district-level assess-ment highlights intra-state variations while analysis of theSRB, IMR andU5MR provides an estimate of the extent of pre-birth, early (within a year after birth) and late (between age 1 and 5) post-birth daughter deficit in Tamil Nadu.3 The following section of the paper describes the data, Section 3 discusses benchmarks established for assessing the deficit, Section 4 presents the extent of daughter deficit while the final section concludes. 2 DataAn intra-state assessment of daughter deficits requires district-level data on SRB, IMR, and U5MR. Indian data on vital events comes from the Sample Registration System (SRS), a countrywide annual survey of vital events covering 1.1 million households in each round. TheSRS supports state-level but not district-level analyses. Fortunately, the 2001 Census provides, for the first time, district-specific data on live births. Accordingly, the district-specific SRB presented in this paper are based on 8,95,765 births recorded in Tamil Nadu in Census 2001. In addition, we use various rounds of Vital Events Surveys (VES) conducted by the Directorate of Public Health (DPH) of Tamil Nadu for the birth years 1996 to 1999 and for the birth year 2003. These surveys track vital events at the district level. The data available in eachVES, nine million individuals and about 1,74,000 births, may be contrasted with theSRS surveys which cover 3,55,000 individuals and 6,000-7,000 births in the state in each round. Each VES covers 3,00,000 individu-als per district and provides data on SRB and infant deaths disaggregated by district and gender (for details see Athreya 1999). Information from VES 1996 to 1999 ispooled to provide SRB and IMR estimates for the late 1990s while information from VES 2003 is used to compute district-specificIMR estimates for the early 2000s. Information on gender-specificU5MR is available only at the state-level andasdiscussed below, this information is combined with infant mortality data to compute estimates of mortality in the age group 1-5. 3 Benchmarking and Measuring Daughter Deficits Daughter deficit, defined as the gap between the number of expected daughters and the number of daughters born or alive in a certain age group, may occur before birth, within a year after birth (early post-birth deficit) or in the age group 1-5 (late post-birth deficit).4 This section considers how these deficits may be identified and establishes an “expected” ratio against which pre-vailing ratios may be compared. Pre-Birth Daughter Deficit International evidence on expectedSRB in countries without pre-birth interference is available from several sources. Based on 240 years of Swedish data, Johansson and Nygren (1991), henceforth JN, conclude that theSRB is “biologically very stable” and close to 952 females per 1,000 males. JN also analyse data from 12 other western industrialised countries for the period 1970-84 and con-clude that the SRB in these countries conforms to patterns found in Swedish data. We updated these numbers using data from United Nations (2004) and computed SRB for all countries with a rela-tively complete civil registration system (at least 90%) with 1,70,000 or more live births per year (about the number recorded in Tamil Nadu’s VES) and with no record of pre-birth interference. The SRB for the set of 16 countries which satisfied these criteria lies between 932 and 965 with a weighted mean of 948.5 Clearly, there is variation in the SRB across countries but given the number of fac-tors that may have a bearing on SRB, it lies in a narrow range. In the Indian context, the expectedSRB may be obtained by examiningSRB at a time when accurate pre-birth sex selection must have been difficult. The earliest source that we have located is Ramachandran and Deshpande (1964), henceforthRD. Based on 1.93 million births in hospitals and health centres between 1949 and 1958,RD report an all-India SRB of 943 and based on Table 1: Population Sex Ratio, 0-6 Sex Ratio in India and Tamil NaduYear 190119111921193119411951196119711981199120011: Population Sex Ratio-India 972 964 955 950 945 946 941 930 934 927 9332: Population Sex Ratio-Tamil Nadu 1,044 1,042 1,029 1,027 1,012 1,007 992 978 977 974 9863:0-6 Sex Ratio-India . . . . . . 976 964 962 945 9274: 0-6 Sex Ratio-Urban India . . . . . . . . . 935 9035: 0-6 Sex Ratio-Rural India . . . . . . . . . 948 9346: 0-6 Sex Ratio-Tamil Nadu . . . . 1,010 a 999a 985 974 967 948 9427: 0-6 Sex Ratio-Urban Tamil Nadu . . . . . . . . . 955 9518: 0-6 Sex Ratio-Rural Tamil Nadu . . . . . . . . . 945 931Ratios are defined as number of females per 1,000 males.a 0-4 sex ratio, from Chunkath and Athreya (1997). Sources: Figures are based on census data (a) Rows 1 and 2, from 1901 to 1961, Visaria (1969), (b) Rows 1 and 2, from 1981 to 2001, Planning Commission (2002), (c) Rows 3 and 6, Premi (2001), (d) Rows 4, 5, 7 and 8, and (e) Census India, Issue 15 (2003), Office of the Registrar General, India. 
SPECIAL ARTICLEjanuary 17, 2009 EPW Economic & Political Weekly587,36,216 births aSRB of 943 for south India. While hospital data reflect a small proportion of births, RD conclude that given their large sample size it is likely that the hospital sex ratio is a “valid indicator of the order of magnitude of the SRB for the population”. WhileRD’s figures are based on actual birth data, estimates of the IndianSRB based on the link betweenSRB and life expectancy at birth have been computed by Klasen and Wink (2003) and Sudha and Rajan (2003). Both report an expectedSRB of 961-962. To benchmark pre-birth daughter deficit, we work with the av-erage of the (south) India-specific estimates of 943 and 962. That is, we use 952 female live births per 1,000 male live births as the sex ratio which may be expected in India and Tamil Nadu in the absence of interference. Thus, pre-birth daughter deficit (DDpre), defined as the gap between the number of expected girls and the number of girls born is calculated as, DDpre = NmSRBe – Nf. ...(1)Nm andNf denote the number of live male and female births and SRBe denotes expectedSRB. 3.1 Early Post-Birth Daughter DeficitIMR, defined as mortality in the age group 0-365 days and ex-pressed as infant deaths per 1,000 live births is usually higher for males. Based on a review, Waldron (1983) concludes that, “in most available data males have had higher mortality than females during the first year of birth”. JN (1991) analyse male and female infant deaths in several countries for the period 1976-84, and conclude that there is a “natural” ratio of 77 female deaths for every 100 male infant deaths. Based on data from United Nations (2004), and for the same set of countries that satisfy the criteria used to establish the SRB benchmark, we find that the ratio of female to male infant deaths ranges from 71 to 83 with a weighted mean of 76.5.6 Working with a naturally occurring SRB of 952 females per 1,000 males and a potential range of 71 to 83 female infant deaths for every 100 male infant deaths, yields an expected female IMR of between 75 (71/0.952) and 87 (83/0.952)% of the male IMR. Based on the weighted mean of the ratio of female to male infant deaths we de-fine the expected female IMR as 80% of the male IMR (76.5/.952). Thus, early post-birth daughter deficit (DDepost) is written as, DDepost = {FIMRs – (0.8 * MIMRs)}Nf. ...(2)FIMRS andMIMRS are estimated female and male infant mortal-ity rates, respectively. Alternatively, (2) may be written as, DDepost = {FIMRs – FIMRe}Nf, ...(3)where FIMRe denotes the expected femaleIMR. Post-birth deficit for Tamil Nadu is computed fromVES based estimates (whereas pre-birth daughter deficit is based on census data). The VES data yield estimates of male and female IMR which help identify districts with evidence of early post-birth daughter deficit. However, an assessment of the absolute post-birth deficit, as in (3), requires information on the total number of female live births (Nf). This information is only available from Census 2001, while the IMR data are from VES 2003. As long as there has beenno major change in birth rates between this period, census information on the total number of female live births may be used to compute (3). According toSRS Bulletin (Vol 36, No 2), in 2001, the birth rate in Tamil Nadu was 19 per 1,000 while it was 18.3 in 2003 (SRS Bulletin, Vol 39, No 1). Both of these are survey-based estimates and it is unlikely that they are statistically differ-ent from each other, which suggests that information on female live births from Census 2001 may be used. 3.2 Late Post-Birth Daughter DeficitTo compute post-birth deficit in the age group 1-5 we need gender-specific mortality rates for this age group. While district level mortality rates for this age group are unavailable, state-level information on U5MR, at least for some years, is available and mortality rates for the age group 1-5 may be obtained by subtract-ing the IMR from theU5MR.7 However, before computing this rate we need to establish a benchmark for U5MR.To establish this, we examined mortality patterns in the same set of countries used to establish benchmarks forSRB andIMR. Unfortunately, the various editions of the UN Demographic Year-book contain information on age-specific mortality rates for the age group 0-4 and not for 0-5. While we are forced to use this in-formation, it is unlikely that the male-female mortality gap in the two age groups differs substantially. Data on mortality patterns for these countries shows that the ratio of female to male deaths in the age group 0-4 ranges from 72 to 83 female deaths for every 100 male deaths with a median of 78 and a weighted mean of 77. These figures are similar to those obtained for defining the IMR benchmark and suggest that expected female mortality rate in the age group 0-5 may also be pegged at about 80% of the male mortality rate. Thus, post-daughter deficit (DDlpost) in the age group 1-5 is defined as the gap between the estimated and ex-pected female mortality rate, times the number of female births, and is written as, DDlpost = {(FU5MRs – FIMRs) – (0.8* (MU5MRs – MIMRs))}Nf. ...(4)FU5MRs andMU5MRs are estimated female and male under-5 mortality rates. Alternatively, (4) may be written as,DDlpost = {FCMRs – FCMRe} Nf, ...(5)where FCMRs andFCMRe are estimated and expected female child mortality rates in the age group 1-5. Combining (1), (3) and (5), we obtain total daughter deficit (TD) as,TD = (NmSRBe – Nf) + {FIMRs – FIMRe}Nf + {FCMRs – FCMRe)Nf. DDpre DDepost DDlpost ...(6)4 EstimatesThis section presents estimates of 0-6 and birth sex ratios as also the temporal and spatial patterns of infant mortality.4.1 Temporal and Spatial Patterns in 0-6 and Birth Sex Ratios We begin by discussing the figures in Table 1. As shown there, between 1961 and 2001, the all India 0-6 sex ratio fell from 976 to 927, a decline of 49 points. Over the same period the 0-6 sex ratio 
Thiruvallur Chennai Vellore Dharmapuri Salem Erode Thiruvanamalai The Nilgiris Viluppuram Cuddalore Peram balar Ariyalur Nagapattinam Nagapattinam Thanjavur Karur Thiruvarur Namakkal Coimbatore Puduk kottai Madurai Virudu nagar Thirunelveli Kanya-kumari Ramanathapuram Toothukkudi Theni Sivaganga Dindigul Thiruchira-ppalli SRB 838-908 909 928 929 951 ≥952 Kancheepuram 
SPECIAL ARTICLEjanuary 17, 2009 EPW Economic & Political Weekly60expected femaleIMR lies in a narrow range. For instance, in 1991 (Table 4, rows 4 and 5), according to census-based estimates the gap between estimated and expected femaleIMR is seven points while it is six points on the basis of SRS 1991. During the rest of the 1990s, the three available data sources for the period reveal the presence of early post-birth daugh-ter deficit with the gap between estimated and expected female IMR lying in a narrow range of 9.8 to 11.6 points (Table 4, rows 6-8). For the early 2000s, female IMR dropped from 39 in the period 1996-99 to 29.7 by 2003 (Table 4, row 9), but it was still higher than the expected femaleIMR of 24. The difference of 5.7 points between the estimated and expected female IMR is statistically significant sug-gesting the continued presence of early post-birth daughter deficit inthe 2000s. While statistically significant in both urban and rural areas, the gap is seven points in rural areas and relatively muted for urban areas where the gap is 1.8 points. While IMR figures based on SRS 2003 are higher than those based on VES 2003, a similar pattern of decline in femaleIMR from 53 in 1996 to 41 in 2003 (Table 4, rows 7 and 10) can be observed. The gap between estimated and expected female IMR in 2003 is about 5.8 points. Thus, regardless of data source the gap between estimated and expected female IMR are similar. Beyond state-level patterns, district-specific numbers pre-sented in Table 6 (p 61) show that the estimated female IMR is higher than the expected femaleIMR in nine districts.There is a large gap between urban and rural areas with evidence of early post-birth daughter deficit in the urban areas of one district as opposed to the rural areas of 10 districts. 4.3 Sex Ratios and Sources of Daughter Deficit To examine the development of sex ratios up to age 6 and to iden-tify the extent of daughter deficit occurring in each age group, consider the information on SRB, IMR, U5MR and the 0-6 sex ratio presented in Table 5 (p 61). Where possible we try to use data from a single source for each year. While the analysis is limited by data availability and is based on cross-section patterns (ide-ally one should follow a birth cohort over time) the figures are, nevertheless, revealing.Consider Table 5, row 1, according to these figures, in 1981, the SRB in Tamil Nadu was 969 suggesting that there was no pre-birth deficit at this time. Male IMR is pegged at 114 and female IMR at 93; given the higher male IMR this translates into a 0-1 sex ratio of 989. The male U5MR is 134 while it is 131 for girls, implying that in the age group 1-5, the male mortality rate (U5MR-IMR) was far less than that for girls (20 per 1,000 male live births versus 38 per 1,000 female live births). This 18 point gap trans-lates into a decline in the 0-1 sex ratio from 989 to the 0-5 sex ratio of 968. While, the decline from a SRB of 969 to a 0-6 sex ratio of 967 suggests post-birth daughter deficit, the main insight is that, higher 1-5 female mortality (1.9 times the male mortality) is the main source of daughter deficit. The gap between actual Table 3: SRB and Pre-Birth Deficit, Tamil Nadu, Census 2001 Overall Urban Rural SRB DeficitSRB Deficit SRBDeficit (1) (2) (3) (4) (5) (6)Tamil Nadu 935 11,471 960 1,637 919 9,834Ariyalur 923 181 955 0 920 181Chennai 979 0 979 0 . .Coimbatore 963 0964 09610Cuddalore 930 409 932 114 929 295Dharmapuri 869 2,212 936 58859 2,154Dindigul 934 256 955 0 923 256Erode 934 362 962 0 910 362Kancheepuram 949 191 964 0 934 191Kanyakumari 9880100009680Karur 910 296 941 272 895 24Madurai 951 228 973 0 925 228Nagapattinam 982097509840Namakkal 859885916119829766Nilgiris 988096301,0230Permabalur 89623790329894208Pudukottai 935 219 879 145 945 74Ramanathapuram 990 0 1019 0 981 0Salem 838 2,567 880 694 806 1,873Sivaganga 955 0 952 0 956 0Thanjavur 962 45994 0 948 45Theni 895 444924 114 862 330Thirunelveli 9461379448394854Thiruvallur 9531709710936170Thiruvannamalai 9226801,016 0905680Thiruvarur 967099909600Thuthukudi 956 83 977 0 941 83Tiruchirapalli 9511119640941111Vellore 907 1250951 9 885 1,241Villupuram 9383289550936328Virudhunagar 945 180 966 0 930 180Total state deficit is the sum of the deficit in each district. Total deficit for each district is the sum of deficits in the urban and rural parts of the district. Source: Based on Census 2001, Table F-9. Table 4: IMR in Tamil Nadu Year/Statistic Source Overall Urban Rural Male Female Expected Male Female Expected Male Female Expected Female FemaleFemale1: 1981 NHDR 114 93 91.2 . . . . . .2: 1981 SRS 93 89 74.4 . . . . . .3: 1982-91 (mid-point 1986) NFHS-1 79 63 63.2 . . . . . .4: 1991 NHDR 55 51 44 . . . . . .5: 1991 SRS 60 54 48 43 40 34.4 69 61 55.26: 1989-98 (mid-point 1993) NFHS-2 50.2 51.8 40.2 . . . . . .7: 1996 SRS 54 53 43.2 42 34 33.6 59 61 47.28: 1996-99 VES 36 39 28.8 23 19 18.4 41 48 32.8Absolute value of test statistic 23.5 1.01 26.8 p-valuea (0.000) (0.313) (0.000)9: 2003 VES 30 29.7 24 24 21 19.2 32.6 33.1 26.1Absolute value of test statistic 7.20 1.43 7.11 p-valuea (0.000) (0.152) (0.000) 10: 2003 SRS 44 41 35.2 32 31 25.6 44 41 35.2IMR-number of infant deaths (age 0-365 days) per 1,000 live births.a H0: Estimated FIMR is equal to expected FIMR. Sources: (a) Row 1 and row 4, Planning Commission (2002), (b) Rows 2, 5, 7 and 10, SRS Bulletins; Row 2 (Vol 19, No 1), rows 5 and 7 (Vol 31, No 1), row 10 (Vol 39, No 1), (c) Row 3, Pandey et al (1998), row 6,www.nfhsindia.org/data/tn/tnchap6.pdf, and (d) Row 8 and 9 are based on VES.
SPECIAL ARTICLEEconomic & Political Weekly EPW january 17, 200961and expected femaleU5MR is 23.8 (131 – 0.8*134), most of which is accounted for by deficit in the age group 1-5 (38 – 0.8*20 = 22), while early post-birth deficit accounts for the remaining gap (93 – 0.8*114 = 1.8). Thus, in 1981 these numbers suggest no pre-birth daughter deficit; about 7% early post-birth daughter deficit and 93% late-post birth daughter deficit.9 In 1991 (Table 5, row 3) as in 1981, the SRB in Tamil Nadu was 952, suggesting no pre-birth daughter deficit. A male IMR of 55 and a female IMR of 51 translates into a 0-1 sex ratio of 955. The maleU5MR is 64 while it is 70 for girls, implying that in the age group 1-5 the male mortality rate was less than that for girls (9 per 1,000 male live births versus 19 per 1,000 female live births). This 10 point gap translates into a decline in the 0-1 sex ratio from 955 to 942 for the 0-5 group, which becomes favourable to girls in the age group 5-6 to reach 948 for the age group 0-6. The decline from a SRB of 952 to a 0-6 sex ratio of 948 suggests post-birthdaughterdeficit.The main point is that while the higher female1-5 mortality (2.1 times the male mortality) is still themainsourceof daughter deficit, its share has declined as compared to 1981 and early post-birth deficithasbecome higher than in 1981. The gap between actual and expected femaleU5MR is 18.8 (70 – 0.8*64), most of which is accounted for bydeficitintheage group 1-5 (19 – 0.8*9 = 11.8), while early post-birth deficit accounts for the remaining gap (51 – 0.8*55 = 7). Thus, in 1991 these numbers suggest no pre-birth deficit, about 37% early post-birth deficit and 63% late-post birth deficit. During 1996-99 (Table 5, row 5), unlike the previous years, the VES-based SRB of 935 is well below the benchmark of 952 suggesting the pres-ence of pre-birth daughter deficit. In contrast to 1981 and 1991, femaleIMR is higher than male IMR which translates into a less favourable 0-1 sex ratio of 929. Since we do not have informa-tion on U5MR or the 0-6 sex ratio for this period we cannot say more. Combining information from Census 2001 andSRS 2001 (Table 5, row 6) allows us to construct the situation in 2001. The SRB remains at 935, implying pre-birth daughter deficit; female IMR continues to behigherthan maleIMR translating into a 0-1 sex ratio of 923. In sharp contrast to 1981 and 1991, by age 6 the sex ratio rises to 942. Thus, starting with a SRB of 935 and falling to a 0-1 sex ratio of 923, the 0-6 sex ratio has becomes favourable to girls, which in turn suggests that in 2001, unlike 1981 and 1991, male mortality between 1 and 6 is much higher than female mortality and that daughter deficit is unlikely to emanate from the age group 1-6. While we do not have figures for under-5 mortality, the increase in the sex ratio from (0-1) 923 to (0-6) 942 suggests that in 2001, male and female U5MR matches the internationally expected pattern of higher male mortality. Overall, these numbers show clear evidence of pre-birth and early post-birth daughter deficit and do not support a substantial role for late post-birth daughter deficit. Beyond 2001,IMR data from SRS 2003 andVES 2003 (Table 5, rows 7 and 8) is used to compute early post-birth deficit. Based on SRS 2003, early post-birth daughter deficit is 5.8 (41 – 0.8*44) and based onVES 2003 early post-birth daughter deficit is 5.7 Table 5: Development of Sex Ratios and Sources of Female Deficit in Tamil NaduYear SRB MIMR FIMR 0-1 Under 5 Under 5 0-5 0-6 Sex Ratio MMR FMR Sex Ratio Sex Ratio1: 1981 969 114 93 989 134 131 968 9671: 1981 952 93 89 951 . . . .1: 1991 952 55 51 955 64 70 942 9481: 1991 952 60 54 955 . . . .5:1996-99 935 3639 929 . . . .1: 2001 935 45 54 923 . . . 9427: 2001(SRB) and 2003 (IMR) 935 44 41 935 . . . .8: 2001(SRB) and 2003 (IMR) 935 30 29.7 934 34 . .Sources: (a) Rows 1 and 3, Planning Commission (2002)-SRB figures are weighted averages of the rural and urban SRB, (b) SRB figures-row 2 and row 4, are SRS-based estimates for the period 1981-90 to 1996-98, Retherford and Roy (2003). IMR figures for 1981 and 1991 are fromSRS Bulletin, Vol 19, No 1 and Vol 31, No 1, respectively, (c) Row 5, based on VES, (d) SRB figures, rows 6, 7 and 8, Census 2001, Table F-9; IMR figures, rows 6 and 7,SRS Bulletin, Vol 37, No 1 and Vol 39, No 1, respectively, and (e) IMR figures, row 8 are based on VES; U5MR, Health Indicators 2003, Government of Tamil Nadu. Table 6: Male and Female IMR and Post-Birth Daughter Deficit, Tamil Nadu(2003) MIMRFIMRDeficitMIMRFIMRDeficitMIMRFIMRDeficit Overall Overall Overall Urban Urban Urban Rural Rural RuralTamil Nadu 30 29.7 1,829 24 21 26 32.6 33.1 1,799Chennai 1714217142 . ..Coimbatore 23191115131425212Cuddalore 323198322203235110Dharmapuri 39 40* 219* 21 16 0 45 49* 258*Dindigul 31 48* 289* 33 35 37 30 54* 241*Erode 32250 26 180 3428 3Kancheepuram 232158 23 2121 23 2231Kanyakumari . . . 26 110 . . .Karur 343441263121373522Madurai 28 35* 230* 22 20 19 30 42* 138*Nagapattinam 322940252614343024Namakkal 26214139031267Nilgiris 28241220263332240Permabalur 31 38* 49* . . . 31 38* 42*Pudukottai 28 30* 92* 20 19 6 31 35* 97*Ramanathapuram 343136273122373111Salem 33 41* 283* 29 33 83 35 45* 178*Sivaganga 28 200 27 100 28 248Thanjavur 33230 18130 3826 0Theni 33 39* 88* 28 31 31 35 42* 46*Thirunelveli 373259 38 3111 363242Thiruvallur 252235161525282413Thiruvannamalai 302982251803234107Thiruvarur 28 21 0 13 26* 26* 34 19 0Thuthukudi 30273724207333027Tiruchirapalli 29 37* 230* 30 24 6 29 42* 182*Vellore 33 40* 349* 24 20 1 36 48* 312*Villupuram 3232143332203236*195*Virudhunagar 34303724285238313IMR-number of infant deaths (age 0-365 days) per 1,000 live births. The state total is based on districts where the difference between estimated and expected SRB is statistically different.* Indicates estimated FIMR is statistically greater than expected FIMR at at least 5% significance level. Source: Based on VES 2003. 
Annual Survey of Industries 1973-74 to 2003-04 (Vol. II) A Data Base on the Industrial Sector in India Second Edition (2006): Low Male-To-Female Sex Ratio of Children Born in India: National Survey of 1.1 Million Households in The Lancet, Published online 9 January 2006, DOI:10.1016/S0140-6736(06) 67930-0,
