Factors Influencing the Use of Prenatal Diagnostic Techniques and the Sex Ratio at Birth in India
T
Economic and Political WeeklyJune 16, 20072293takeadvantage of recently released data from the 2001 Censuson fertility and age-sex distribution of the population insingleyearsto study the influence of some key factors on theSRB in India.Evidence from Census of 2001Information on the sex ratio of children of age 0-six years wasone of the first data to be released from the 2001 Census. It causedwidespread anguish as it showed significant fall in the proportionof females in this age group, indicating dramatic increase in theincidence of pre-birth elimination of females [Registrar General,India 2003]. But data on child sex ratios are also affected bysex differentials in child mortality, under-enumeration and agemisreporting [Bhat 2002]. Recently, census data on populationby single years of age have been released. This information canprovide further clues to the nature of changes in the child sexratio and its causes. In analysing this information, before com-puting sex ratios, we applied a three-point moving averageformula to smooth the single year age data. In Figure 1, for all-India and the state of Punjab, we have plotted the difference inthe sex ratio (males per 100 females) at the same single-year ofage between 1991 and 1981, and between 2001 and 1981.However, the graph shows instead of age, the year of birth ofchildren implied by their reported age in the census. For bothPunjab and India, the increase in the sex ratio is more pronouncedfor more recently born children (ie, at younger ages). As perthe 1991 Census data, the sex ratio steadily increased between1985 and 1990 by 3 percentage points for India as a whole, andby 9 percentage points for Punjab. As per the 2001 Census data,the sex ratio steadily increased between 1995 and 2000 by another3 percentage points at the all-India level, and by another 11percentage points in Punjab.The pattern of change observed in the census data discountsthe possibility of this change arising from a rise in excess femalechild mortality in recent times. If it were the cause, owing tothe cumulated impact of the mortality differential, sex ratios atages three-four years would have shown greater change than atages one-two years. The systematic age (or time) pattern in thesex ratio increase raises doubts whether it could be explainedby more accurate reporting of children’s age. However, the roleof age misreporting in distorting the trend in sex ratios cannotbe completely ruled out since for the overlapping period of1988-90, the rise in the sex ratios indicated by the 1991 and 2001Census are not identical–while the former census indicatessignificant increase, the latter indicates no change or a even adecline (in comparison to the sex ratios of 1981 Census atcorresponding ages).To a large extent, the systematic rise in the sex ratio in theyears preceding the census must have been due to the rising trendin the SRB. If it were the only cause, the implication is thatbetween 1985 and 2000 the sex ratio at birth increased by 6Table 2: Child Sex Ratio and Sex Ratio at Birth by SelectedBackground Characteristics, All-India, 2001 CensusSex Ratio at BirthSex RatioBirthsChildrenBackground Characteristicsof Childrenduring theEver-bornof AgePrecedingto Women 0-6 YearsYearAged 20-34All107.8110.4106.7ResidenceRural107.1110.4106.3Urban110.3110.6108.3ReligionHindu108.2110.9106.9Muslim105.3107.4105.3Christian103.7103.8103.0Sikh127.3129.8119.1Buddhist106.2108.4105.2Jain115.0118.0110.5Other religious communities102.5106.5102.3Caste/TribeScheduled tribe102.8106.4103.1Scheduled caste106.6108.6105.8Others108.8111.5107.4Mother’s age< 15na105.9na15-19na108.2na20-24na111.4na25-29na113.2na30-34na112.1na35-39na109.1na40-44na103.7na45-49na99.7na50+na75.7naMother’s educational levelIlliteratena108.7106.0Literate but below primaryna110.0106.3Primary but below middlena111.8107.1Middle but below matric or secondaryna113.0107.5Matric or secondary but belowgraduatena115.3109.4Graduate and abovena114.1109.7Total births/children (in thousands)1,63,82019,8872,37,622na - not available/applicable.Source:Census of India, 2001.Table 1: Sex Ratio for the Age Group 0-2 Years in 1981, 1991and 2001 CensusesStateSex Ratio, 0-2 Age GroupChange1981199120011981-01All-India102.7105.4108.35.6*North-westJammu and Kashmir104.2na109.04.8*Himachal Pradesh103.2107.1113.910.7*Punjab107.3116.2127.220.0*Haryana107.8114.7124.216.3*Delhi106.1109.1116.510.4*North-centralRajasthan103.0108.1111.48.4*Uttar Pradesh103.1105.9108.85.6*Bihar100.5103.5105.34.8*Madhya Pradesh101.5103.6106.14.6*EastAssamna103.4103.80.4North-east@101.6101.8103.31.7West Bengal101.9103.4104.32.5Orissa101.0103.3105.84.8WestGujarat104.6107.4114.39.7*Maharashtra105.1107.0111.36.2*Goa104.5103.7108.13.6SouthAndhra Pradesh100.5102.5103.63.2Karnataka102.3104.4106.03.7*Kerala102.7104.8104.11.4 Tamil Nadu102.9105.2106.03.1Notes:* Linear change with age statistically significant.@ Arunachal Pradesh, Manipur, Meghalaya, Mizoram, Nagaland,Sikkim and Tripura.Source:Census of India, 1981, 1991 and 2001.
2001 Census data 1991 Census data 
Punjab 2001 India 1991
India 2001
Legend Sex ratio at birth NA < 105 105-109 110 and above Legend Sex ratio at birth NA < 105 105-109 110 and above SU = MU FU
M
S = N N FN
MU ⎛ 11 ⎞
AU =− FU = MU ⎜⎜− ⎟⎟
S SS
⎝ U ⎠
Legend Per cent NA 0.00-9.99 10.00-19.99 20.00 and above Legend Per cent NA 0.00-9.99 10.00-19.99 20.00 and above AU SS
= 1− F = 1− .
FU + AU UMU SU
MN ⎛ 11 ⎞
AN =− FN = MN ⎜⎜− ⎟⎟
S SS
⎝ N ⎠
AU + AN F + A + A
UU N
+ FU + A
MU + AU N
M + F + M + F + A + A
UU NNUN
Economic and Political WeeklyJune 16, 20072298India as whole, the SRB of 112 among those who used PNDTindicates that 6 per cent of female foetuses were aborted afterPNDT. But if it is assumed that some of those who did not usePNDT may have actually used it, the estimated percentage offemale foetuses aborted rises to 17 per cent, and the adjusteduse of PNDT is 14 instead of 13 per cent. Those who did notavail antenatal care (ANC) services clearly could not have usedPNDT and resorted to female feticide. Their SRB of 102 alsotestifies to this. But the SRB among those who availed ANCfrom a doctor is 112. Twenty-five per cent of them had reportedPNDT, which rises to 27 per cent if adjustment is made for thesuspected non-disclosure of its use. If the normal sex ratio is105, the SRB of 111 among PNDT users indicates that 6 percent of the users may have used the technology to abort femalefoetuses. But this estimate rises to 23 per cent if an adjustmentis made for the possibility that some of the reported non-usersof PNDT users did not disclose its use (since the SRB amongthis group is 113). Among women who availed PNDT fromproviders other than the doctor, 5 per cent had reported the useof PNDT, and it is estimated that 24 per cent of them had misusedit. In this group, among those who did not report the use of PNDT,the SRB is 104, indicating no misreporting of the PNDT status.In urban areas, 31 per cent of live births were reportedlysubjected to PNDT compared with only 8 per cent in rural areas.On the assumption that reported PNDT use is correct, 8 per centof those who used PNDT misused it in urban areas whileTable 5: Reported and Adjusted Use of Prenatal Diagnostic Technologies and Estimated Percentage of Females FoetusesAborted After the Use of These Technologies by Selected Background Characteristics, NFHS-2, 1998-99Variables TotalPer Cent of BirthsSex Ratio at BirthPer Cent of Female FoetusesBirths inSubjected to PNDTAllPNDT NotPNDTAborted after PNDT the SampleReportedAdjustedBirthsReportedReportedEst 1Est 2Total 32,22812.813.9107.7107.1111.86.116.9ANC status at pregnancyNo ANC 10,9870.00.0102.4102.40.00.00.0ANC by doctor 15,74524.527.0112.6113.1111.15.522.8ANC by others 5,3955.05.3104.7104.0119.512.112.1ResidenceRural 25,0647.78.8107.6107.4109.84.424.4Urban 7,16430.831.8108.1105.7113.67.58.9RegionEast 4,4215.08.7114.2114.6106.51.462.8South 6,13227.927.9102.0102.7100.20.00.0West 4,42126.027.3108.8107.1113.97.812.6North-west 2,04717.222.0119.2114.6144.427.346.7North-central 15,2114.65.2106.4105.5126.416.924.3Birth order1 9,29920.821.8106.9108.3101.80.010.32 8,33315.516.9108.9108.8109.33.919.43 5,7089.310.5108.0106.3126.116.725.24+ 8,8924.35.3107.3105.3166.937.140.0Sex composition of siblingAll sisters, no brother 7,87413.918.9119.1116.2138.724.351.2All brothers, no sister 7,17611.711.7100.5100.3102.20.00.0Other combinations 17,18212.813.2105.9106.2104.00.07.2ReligionHindus 25,53412.213.5108.3107.4114.78.521.1Muslims 5,04212.012.0101.4101.898.00.00.0Others 1,57026.128.3111.9112.5110.24.819.9CasteScheduled caste/tribe 9,4916.37.8108.7108.8107.72.535.8Other backward caste 10,35814.014.1105.4105.3106.61.52.9Others 12,00017.419.2110.0108.7116.59.922.0Parents’ educational levelBoth parents illiterate 8,9222.72.8105.1105.396.70.010.2Only father literate 9,8905.35.7104.3103.4121.513.613.6Mother less than middle 5,82114.214.2103.2103.3102.50.00.0Mother middle or more 7,51033.837.6119.0121.4114.28.127.8Standard of livingLow SLI 11,6383.84.5105.9106.593.90.024.6Medium SLI 15,06811.812.2106.1106.0107.42.28.4High SLI 5,12535.838.8116.2114.2120.012.523.3Mother’s media exposureNo exposure 15,3142.72.9105.2104.9115.99.49.4Regular exposure 16,88522.023.7110.0109.7111.35.718.0Mother’s work statusNot working 22,28214.916.7109.8108.6116.39.723.9Working, not for wage 5,1359.19.1104.1104.797.90.00.0Working for wage 4,8047.47.4102.4103.291.90.00.0Mother’s ideal family size1-2 15,48021.524.4113.8114.3112.16.428.03 8,6755.87.7109.5108.8120.713.043.8 4+ or non-numeric 8,0733.93.995.195.196.20.00.0Source:NFHS-2 micro data.
Economic and Political WeeklyJune 16, 200722994percent did so in rural areas. But the sex ratio of those whodid not report the use of PNDT shows that 32 per cent in urbanareas and 9 per cent in rural areas may actually have used it,and among them 9 per cent in urban areas and 24 per cent inrural areas may have misused it. Thus the use of PNDT is certainlymuch higher in urban areas than in rural areas, but it is notaltogether clear as to where the misuse is higher.The use of PNDT is relatively high in south and western regionsof India where the reported use is 27-28 per cent. The reporteduse is less than 5 per cent in east and north-central regions ofIndia. In north-western parts of the country, where the child sexratios are high, the reported use is moderate (17 per cent). Thestrong regional pattern in the reported use is clearly seen inFigure4 where we have mapped the use-rate for 77 for naturalregions of the country. One reason for the relatively high useof PNDT in south and western parts of the country could be thatthere the use of ANC services, especially from a doctor, is high.However, as the map in Figure 5 shows, even when only caseswith ANC by a doctor are considered, the regional pattern inthe use of PNDT does not disappear. Reported use of ultrasoundor amniocentesis is rare in east and north-central India, even whenANC is availed from a doctor.Although south India shows the highest use rate of PNDT,misuse of the technology is rare; as a consequence, the SRB isless than 105 in this region among both users and non-users(Table5). In the north-central region, though the use of PNDTis rare, misuse among those who use it is high (17-24 per cent).But it is in the north-western region, where the misuse of PNDTis most frequent (27-47 per cent). For north-western India, whenthe reported use is adjusted for its possible under-reporting, theimplied use rate increases from 17 to 22 per cent. In other regions,the implied corrections for underreporting are small. Althougheast India shows the highest percentage of misuse if the highSRB for non-users is taken into account, it is most probably dueto sampling errors in the data than the actual misreporting ofPNDT status.The reported use of PNDT declines as the order of birthincreases. But the misuse of the technology increases with thebirth order. The reported use is 21 per cent for the first birthorder but the estimated misuse, at the maximum, is only 10 percent. On the other hand, when the birth order is four or more,only 4 per cent report the use of PNDT but nearly 40 per centresort to abortion if the fetus is female. Thus in the case of firstbirth, the predominant reason for the use of PNDT is to detectabnormalities of the fetus, while at higher orders the main reasonfor the use is to detect the sex of the child. The reported useof PNDT doesn’t show significant variations with sex compo-sition of previous children born to the woman. But the misuseof the technology is highest when the woman had no son, buthad one or more daughters. Among such women, the misuse isestimated to be 24-51 per cent. When women had no daughterbut had one or more sons, the SRB is 101, and 102 for womenreporting PNDT use. While this may be suggesting abortion ofmale foetuses by such women, the difference from the normalsex ratio can also be due to sampling errors. On the other hand,the SRB for women who had no sons but had two or moredaughters is substantially high to be attributed to sampling errors(119 for all births and 139 for births with reported use of PNDT).The reported use of PNDT is 12 per cent among Hindus andMuslims, and 26 per cent among other religions (mainlyChristians and Sikhs). While those who misuse the technologyare negligible among Muslims, it is 8-21 per cent among Hindusand 5-20 per cent among other religions. As would be expected,the reported use of PNDT increases from 6 per cent amongscheduled castes/tribes to 14 per cent among other backwardclasses (OBCs) and 17 per cent among others. But the misuseof the technology is least among OBCs (2-3 per cent) while itis relatively high among other castes (10-22 per cent). The use of the technology increases with the level of educationof parents. When both the mother and father are illiterate, only2 per cent use PNDT. When only the father is literate, thisincreases to 5 per cent. When the mother has completed highschool or gone to college, the reported use is 42 per cent. Themisuse does not show a consistent relationship with education.It is low when both the mother and father are illiterate (0-10 percent), but relatively high when the father is literate but the motheris not (14 per cent) and when the mother has competed high schoolor gone to college (9-28 per cent). Interestingly, misuse is almostnegligible among mothers who are literate but have not completedmiddle school. To study the relationship of use and misuse ofPNDT with the prosperity of households we use an index ofTable 6: Results of Logit Analysis of Determinants of the Useof Prenatal Diagnostic Technologies, NFHS-2, 1998-99Explanatory VariablesCoefficientMaternal age0.105**Maternal age squared-0.0008Paternal age0.134***Paternal age squared-0.0017***Order of birth 3+ (no = 0)-0.260***No surviving male sibling and order 2 or higher (yes = 0)0.082No surviving female sibling and order 2 or higher (yes = 0)-0.206***Urban residence (rural = 0)0.743***Educational level (mother and father illiterate =0)Mother illiterate, father literate0.360***Mother less than primary0.683***Mother middle school0.819***Mother high school +1.189***Regular exposure to mass media (no = 0)0.661***Standard of living (low = 0)Moderate0.415***High1.107***Religion (Hindu = 0)Muslim0.101Christian0.485***Sikh0.094Others0.374**Caste/tribe (others = 0)Scheduled tribe-0.589***Scheduled caste-0.308***Other backward castes-0.086Female work status (not working=0)Working for wages-0.015Others-0.199**Ideal number of children reported by mother-0.449***Ideal number of children squared0.0278***Non-numeric ideal children (numeric = 0)-1.000***Health worker visit during pregnancy (no = 0)-0.275***Region (south=0)East-1.755***West0.060North-west-1.225***North-central-1.529***Constant-5.697***Number of births 31,481 -2 Log-likelihood 16,671Pseudo R20.231 Note:* p<0.05, ** p<0.01, *** p<0.001.Source:NFHS-2 micro data.
Economic and Political WeeklyJune 16, 20072300standard of living (SLI) computed from household amenities andassets [IIPS and ORC Macro 2000]. The reported use of PNDTincreases from 3 per cent in low SLI households to 38 per centin high SLI households. The estimated misuse among reportedPNDT users also increases from zero per cent in the low SLIgroup to 2 per cent in the medium SLI group and 13 per centin the high SLI group. But when adjusted for the possiblemisreporting of PNDT status, the estimated misuse of PNDT inthe low SLI group (24 per cent) turns out to be as high as inthe high SLI group (23 per cent) and substantially higher thanthe medium SLI group (8 per cent).The reported uses of PNDT increases with the mother’s ex-posure to media. But it does not show a clear relationship withmisuse. When only the reported use of PNDT is considered,misuse is marginally lower among women regularly exposed tomedia. But when adjusted for possible underreporting of PNDT,the estimated misuse is higher among women regularly exposedto media. Women’s work status however has a clear relationshipwith both use and misuse of PNDT. Both use and misuse arehigher among non-working women compared with womenworking for wage or those who are self-employed or workingin the family farm or business. Reported as well as the adjusteduse rates increase with mother’s ideal family size. When theidealfamily size is one or two, 22-24 per cent had PNDT.Whenthe ideal family size is four or more, only 4 per cent usePNDT. The misuse is highest when the ideal family size isthree (13-44 per cent), moderate when it is one or two (6-28 percent) and negligible when it is four or more.The foregoing analysis shows that the use of PNDT tends tobe systematically related to socio-economic factors, and theadjustments needed for possible under-reporting of its use arerelatively minor. Therefore, without significantly biasing theresults, one can apply the multivariate techniques to the reporteddata on PNDT use to study the independent effects of key socio-economic variables on the use of PNDT. Accordingly, Table 6shows the results of logistic regression of the determinants ofPNDT use. As the table shows, most of the variables used inthe regression have significant, independent effects on the useof PNDT. The use of PNDT increases with maternal as well aspaternal age. But the use of this technology is lower if the orderof birth is three or more. It is particularly low if the mother hasalready given birth to a son.Urban residence, educational level, exposure to media andstandard of living show strong, independent and positiveeffects on the use of PNDT. Its use is higher among Chris-tians, and if thereligionisother than Hindu, Muslim or Sikh.The use is lower among members of scheduled tribes and sched-uled castes, even after controlling for standard of living andeducational level. When compared with non-workers, the use islower among mothers who work but not for wage. The use ofPNDT is lower among mothers having larger ideal family sizesand among those who gave non-numeric answer to the questionon ideal family size. Women who were visited by health workersduring the pregnancy report lower use of PNDT. This findingis directly in contradiction with the claim of some activists thatpublic health workers act as conduits of this technology in ruralareas. Even after controlling these variables, eastern, north-central and north-western parts of India show lower use of PNDTthan southern states. In other words, the geographical patternseen inthe PNDT use cannot fully be explained by the observedsocio-economic variations. This indicates the influence ofneighbourhood on the use of technology.Determinates of Sex Ratio at BirthIn recent years the use of PNDT has emerged as the keyintervening variable through which other factors influence theSRB in India. There are, however, a number of factors thatindependently affect the SRB. Teitelbaum (1972) provides anearly review of the literature on this subject. James (1987) andWaldren (1998) provide a more recent review of the literature.Chahnazarian (1988) has applied multivariate techniques to testthe independent effects of some of these variables using vitalstatistics data for several countries. For India, Retherford andRoy (2003) have used the NFHS data to test the significanceof a limited set of factors. Here, using the same data set, an attemptis made to test the significance of far more variables on the SRB.Accordingly, Table 7 shows the results of the logit analysisof determinants of the SRB using the data from NFHS-1 andNFHS-2. The results presented are with respect to the probabilityof having a male birth. Two sets of regression results are presentedfor each of the surveys, one using data on births that occurredduring the 0-14 years before the survey, and another using dataon births that occurred 0-4 years (for NFHS-1) and 0-3 years(for NFHS-2) before the survey. While the first set of regressionsis based on larger sample of births, the latter set takes advantageof having information on more variables for the more recentbirths. Nonetheless, none of the regressions is able to accountfor more than 1 per cent of the variation in the probability ofmale birth (as indicated by values of pseudo R2), which under-scores the random nature of sex determination at birth. Sincein such a situation statistically significant effects could be detectedonly in large samples, we have checked whether some additionalvariables become significant when 10 per cent probability levelis employed for rejection of null hypothesis.The most notable results from these regressions are with respectto the effects of sex composition of previously born children andthe use of PNDT on the probability of male birth. Although orderof birth doesn’t show statistically significant relationship in anyof the regressions, regressions using the NFHS-2 data set showthat if the mother had at least one previous birth but had no sonat the time of current birth, the probability of the birth beinga male is higher and this effect is strongly significant. Thissuggests that such women were not resorting to sex-selectiveabortion in the 1980s but had begun to do so in the 1990s. Thefemale-selection situation, characterised by women who had atleast one previous child but had no daughter, reduces the prob-ability of male birth, but its effect in the regressions is only mildlysignificant at 10 per cent level.Information on the use of PNDT is available only for the recentbirths from NFHS-2. The regression using this data shows thatPNDT use does not have a direct effect on the SRB. To testwhether its effect depends on the sex-selection situation, we haveinteracted its use with dummy variables for male- and female-selection situations. As the results show, when women in male-selection situation (at least one previous birth and no sons) usePNDT, it has strong positive effect on the probability of malebirth. Its effect in the case of women with female-selectionsituation is not in the expected direction (ie, negative), and alsonot statistically significant. It may be noted that male-selection
Economic and Political WeeklyJune 16, 20072301Table 7: Results of Logit Analysis of Determinates of Probability of Having a Male Birth NFHS-1, 1992-93 and NFHS-2, 1998-99 Births during 0-14 YearsBirths duringExplanatory VariablesBefore Survey1989-921996-98 NFHS-1NFHS-2NFHS-1NFHS-2Maternal age0.015**-0.004 0.013 0.012 Maternal age squared-0.0003**0.0002-0.0003-0.0002Paternal age-0.009*-0.006-0.005-0.030**Paternal age squared0.0001*0.00010.00010.0005***Paternal age not reported (reported = 0)0.027-0.070nanaOrder of birth0.0020.0000.0110.011No surviving male sibling and order 2 or higher (yes = 0)-0.0020.043***0.0040.110***No surviving female sibling and order 2 or higher (yes = 0)-0.025*-0.022*-0.035*-0.041Urban residence (rural = 0)0.024**-0.0170.016-0.039Educational level (mother and father illiterate =0)Mother illiterate, father literate-0.014-0.007-0.010-0.025Mother less than primary-0.0160.011-0.073**-0.017Mother middle school0.0050.0200.0090.090*Mother high school +-0.064***0.038*-0.127***0.068Regular exposure to mass media (no = 0)0.0000.020-0.020-0.009Standard of living (low = 0)Moderate-0.018-0.026**-0.014-0.005High0.007-0.0200.0290.044Religion (Hindu = 0)Muslim-0.028*0.013-0.0350.020Christian0.0130.045*0.0120.031Sikh0.0430.033-0.014-0.092Buddhist0.0280.0550.197*-0.107Others0.0400.057-0.0010.064Caste/tribe (others = 0)Scheduled tribe0.0150.0160.0270.006Scheduled caste0.0200.0100.0250.060*Other backward class0.011-0.006Female work status (Not working=0)Working for wages-0.013-0.010-0.059**-0.051Others0.004-0.0020.014-0.014Ideal number of children reported by mother-0.043***-0.070***-0.043**-0.064*Ideal number of children squared0.0018**0.0046***0.00160.0003Non-numeric ideal children (numeric = 0)-0.134***-0.156***-0.076-0.232***Consanguinity (no = 0)-0.016na0.012naHealth worker visited during pregnancy (no = 0)nana0.032-0.017Birth attendance (untrained birth attendant=0)nanaTrained birth attendant at homenana0.085***0.076*Trained birth attendant at institutionnana0.0400.070**Iron and folic acid supplementation (no = 0)Received tablets/syrupnana0.0160.022Fully consumednanana-0.041Number of antenatal check-upsnana0.008*0.000Maternal anaemia at surveyMildna-0.015na0.002Moderate/severena-0.044***na0.002Not testedna-0.025na-0.113**Mother’s body-mass index at survey (normal = 0)Low (below 18.5 kg/m2)na0.005na0.032High (25 kg/m2 or more)na-0.048**na-0.109*Not measuredna-0.202***na-0.119***Mother’s height at surveyna0.007na-0.001Mother’s height not measured (measured = 0)na-0.029na0.029Use of PNDT (no=0)nanana-0.056Interaction with PNDT useNo surviving male sibling and order 2 or highernanana0.228***No surviving female sibling and order 2 or highernanana0.015Altitude more than 1,000 metres (lower = 0)na-0.035*na0.029Region (south = 0)East0.0230.043**0.079**0.078*West-0.0020.041*0.0400.060North-west0.043**0.090***0.129***0.147***North-central0.051***0.082***0.090***0.090**Constant0.1100.317***-0.0510.550Number of live births1,75,815 1,58,840 53,591 31,353 -2 Log-likelihood2,43,4312,19,726 74,163 43,234 Pseudo R20.001 0.0020.0020.005Notes:* p<0.10, ** p<0.05, *** p<0.01.na - not available/applicable.Source:NFHS-2 micro data.
Economic and Political WeeklyJune 16, 20072302situation shows positive effect on the probability of having a malebirth even in the absence of PNDT use, indicating that somewomen must have suppressed its use.Another important result is with respect to the effects ofantenatal and delivery care. In both NFHS-1 and NFHS-2, dataon antenatal and natal care are available only for the recentbirths(for the last four years in NFHS-1 and last three years inNFHS-2). Both data sets show that the SRB is higher when trainedpersonnel attend the birth. This could be due to the reductionin stillbirths when they attend the birth. But number of antenatalcare check-ups and consumption of iron and folic acid tabletsor syrup during pregnancy appear to have no effect on the SRB.Urban residence, educational level, standard of living, expo-sure to mass media, religion and caste generally do not showstatistically significant or consistent relationship in all the re-gressions. Mothers with high school education or higher havereported lower proportion of males in their births in NFHS-1.But in NFHS-2 they have reported more males in their births,and the effect is mildly significant in the regression using datafor births during 0-14 years before the survey. Mothers workingfor wage report lower probability of male births in all the re-gressions but the effect is statistically significant only in dataon births during 0-14 years before NFHS-2. The regression resultsshow that the SRB increases as the mothers’ ideal number ofchildren decreases, though in a non-linear fashion. This effectis strongly significant in all the regressions. Also, those who gavenon-numeric response to the question on ideal family size hadfewer sons. This would seem consistent with the contention ofsome female activists that incidence of sex-selective abortionsincreases with the decrease in family size. However, the fact thatthis effect is also found in NFHS-1 data for the period when theincidence of female feticides were rare (as indicated by theinsignificance of the variable representing the male-selectionsituation) suggests that the observed relationship is spurious:Those who had sons tend to use contraception more than thosewhohad daughters; in order to justify their non-use of contraception,the latter group has the tendency to report higher ideal family size.Maternal and paternal ages show statistically significant re-lationship with the SRB in some of the regressions. In births thatoccurred during 0-14 years before NFHS-1, maternal age showsstatistically significant non-linear relationship, indicating ini-tially a rise in the SRB with maternal age and then a fall at olderages. This may be attributed to a similar relationship of maternalage with stillbirth rate. But this relationship is not seen in thedata from NFHS-2. As the role of men in sex determination ofthe fetus is well established, their age could have a bearing onthe SRB. Some researchers have found that the SRB decreaseswith paternal age and speculated that it is due to the decreasein coital frequency with age [eg, Chahnazarian 1988]. All ofour regressions indicate that the SRB initially decreases withpaternal age and then rises at advanced ages. This effect isstrongly significant in births that occurred 0-3 years beforeNFHS-2 and mildly significant in the regression that used birthsduring 0-14 years before NFHS-1. While the initial decrease isas expected, the suggested rise after age 30 or so needs furtherinvestigation. The significance of the quadratic term in theregressions may simply be indicating that the rate of decrementdiminishes with age.The possibility of maternal malnutrition influencing the SRBhas been discussed in the literature on this subject. Anderssonand Bergstrom (1998) had found that short maternal stature andobesity were independently related to low SRB in Africa. Dataon mother’s nutritional status and anaemia are available only fromNFHS-2. When data on births during 0-14 years are used (ie,larger sample of births), moderate and sever anaemia show strongnegative effects on the SRB. Mother’s height does not show anyeffect on the SRB, but the results strongly suggest that obesewomen tend to have fewer sons. But it should be noted that NFHSdata on maternal anaemia and body-mass index refer to the timeof the survey rather than to the time of birth. Also to be notedis that women for whom data on body-mass index were notavailable had significantly lower SRB. This index was notcalculated for pregnant women and those who had given birthduring the two months before the survey. Such women were likelyto have had more female births in the past because in societieswith strong son preference, those who had sons are more likelyto use contraception and stop childbearing.Data on altitude are available only from NFHS-2. When birthsduring the 0-14 years before the survey are considered, there isa weak indication that higher altitudes reduce the chance of malebirth. This may be because of lower temperature and higher malefetal mortality at higher altitudes. It is also possible that peopleliving at higher altitudes are genetically different from thoseliving in low-lying areas. The information on consanguinity wascollected only in NFHS-1. It shows no statistically significantrelationship with the SRB. But owing to the limitations of theinformation on consanguinity collected in NFHS-1, it served asa poor proxy for the level of inbreeding. In all the regressions,even after controlling for the foregoing factors, the SRB issignificantly higher in north-western and north-central regions.This could be because of greater misuse of PNDT in these parts.However, as these regional differences are also significant (butsmaller) in the regressions that use NFHS-1 data, they may beindicating the presence of some genetic factors such as levelsof circulating gonadotrophin, or a tendency to under-report femalebirths in northern India.Summary and ConclusionsIn recent years the use of PNDT, followed by sex-selectiveabortion has emerged as a powerful determinant of SRB in India.In this paper we have analysed the factors influencing the useof PNDT as well as SRB. We present evidence from the recentlyreleased data on fertility from the 2001 Census as well as fromthe NFHS. The census fertility data suggest that SRB in Indiamay have increased to 110, and in some areas, to as high as 130.But this data may have been affected by under-reporting of femalebirths. However, even this data show that the SRB in south andeastern India is well within the range observed under normalcircumstances, and thus discount the possibility of widespreaduseof sex-selective abortions in these areas. The census data alsoshow that though there may be little rural-urban difference inthe SRB, among the educated class the SRB is abnormally high.The data collected in NFHS-2 (1998-99) show that 13 per centof live births were subjected to PNDT, and 6 per cent of femalefoetuses may have been aborted after PNDT. But, if possibleunder-reporting is taken into account, PNDT may have been usedin 14 per cent of the cases, and in 17 per cent of such cases femalefoetuses may have been aborted. The use of PNDT is higher insouthern and western regions of India while its misuse forlli
