Son Preference in India: Shedding Light on the North-South Gradient Son Preference in India: Shedding Light on the North-South Gradient Daniela Klaus, Arun Tipandjan Abstract: Son preference is widespread in India and deep-rooted in its history. It is a matter of concern because it produces an imbalanced juvenile sex ratio. There are far fewer girls than boys. The fi gures vary greatly among the Indian states suggest- ing a major north-south gradient in son preference accompanied by a minor west- east gradient. The aim of this paper is to explain the regional pattern. We provide an application of the value of children-approach according to which the decision to have children is made on the calculus of benefi ts and costs related to children. In the light of the socioeconomic and sociocultural background in India, we propose that (potential) parents’ expectations of benefi ts and costs are biased in favour of sons. This is suggested, therefore, as the key motivation for the preference for male offspring. However, region-specifi cs in the level of affl uence, the educational level, the mode of production, the meaning and importance of religion, and the kinship regime are assumed to produce stronger son preference in north India compared to south India. This mediation-model is tested using the Indian sub-sample of the international Value of Children-study. Data were collected in Uttar Pradesh (north- central India) and Puducherry (south-east India). Mothers aged 16 to 65 were inter- viewed in 2002 and 2010. Based on 1,173 respondents, a structural equation model was carried out to test the hypothesised composition effects related to the region and the mediating position of sex-specifi c benefi ts and costs. Initial fi ndings con- fi rm that the national son preference pattern is more likely to be found among north Indian mothers than south Indian mothers. As assumed, the sex-specifi c balance of benefi ts and costs contributes to the explanation of son preference. However, there is little evidence that the benefi ts and costs mediate between the region-specifi c socioeconomic and sociocultural profi les and son preference. Son preference is most pronounced among mothers of the north-urban sample after controlling for region-specifi c distributions of socioeconomic and sociocultural background vari- ables. Variations in son preference across the regional sub-samples are partly ex- plained by the respective background variables and the benefi t and cost-structure. But independent regional effects continue to be signifi cant and thus, a considerable part of the north-south gradient remains unclear. Keywords: Son preference · India · Value of children · Kinship regime · Dowry Comparative Population Studies Vol. 40, 1 (2015): 77-102 (Date of release: 18.12.2014) © Federal Institute for Population Research 2014 URL: www.comparativepopulationstudies.de DOI: 10.12765/CPoS-2014-18en URN: urn:nbn:de:bib-cpos-2014-18en5 • Daniela Klaus, Arun Tipandjan78 1 Introduction Son preference is widespread in India and deep-rooted in its history. Although it is common in the Middle East, North Africa, and in South and East Asia, it is most pronounced in India and China. There is great public concern about this issue for at least two reasons: First, it is widely argued that son preference is slowing the tran- sition to low levels of fertility as couples might continue to bear children until the desired number of sons is reached. But empirical evidence is mixed and suggests a complex interplay (Arnold et al. 1998; Basu 1999; Bhat/Zavier 2003; Clark 2000; Das Gupta/Bhat 1997; Mutharayappa et al. 1997; Pande 2003; Srinivasan 2005). The spread of the preference for smaller families might add to the pressure to prevent the birth of daughters because the probability of having a son decreases with the lower overall number of preferred children (UNFPA 2012: 42-43) Second, son preference implies gender discrimination which is manifested in a broad range of sex-selective behaviour (Basu 1999; Behrman 1988; Borooah 2004; Das Gupta/Bhat 1997; Oster 2009; Pande 2003; Pande/Astone 2007; Rosenzweig/ Schultz 1982; Sengupta/Agree 2002; Srinivasan 2005): Sex-selective investments in education, autonomy, and self-determination reproduce the subordinate social and economic position of women. Sex-selective rates of abortion and infanticide and sex-specifi c health care and nutrition produce sex-selective mortality and thus distorted sex ratios. India, like some other Asian countries, faces the problem of “missing women” (Sen 2005: 225) which implies, among other things, a shortage of women on the marriage market – termed a “marriage squeeze”. Within the last decades, the juvenile sex ratio, defi ned as the number of boys aged 0 to 6 per 100 girls of the same age, became increasingly disproportional: It has increased from 102 in the middle of the 20th century to 110 in 2011 with an accelerated increase in the 1980s (Bhaumik 2012). This rise does not indicate an intensifi cation of the son preference but its stronger manifestation, supported by the growing availability of methods of sex diagnosis technology. The development of prenatal diagnostics is the driving factor behind the shift from the traditional postnatal discrimination to the more effective practice of female abortion (Goodkind 1996). A harmful side effect of such a rigid avoidance of giving birth to a female child may be the impairment of women’s mental and physical health due to (multiple) abortions. The alarming development of the sex ratio in India can largely be attributed to the north-western states. Latest fi gures for the juvenile sex ratio suggest a major north-south gradient which is accompanied by a minor west-east gradient. Apart from few exceptions, states in the northern plains and in the central upland show ratios that are well above 110. Ratios in south-east India reach around 105. In some selected states like Kerala and Puducherry they are almost balanced (Bhaumik 2012; Government of India 2011; UNFPA 2012). This uneven distribution of the sex ratio across the regions is accompanied by a similar distribution of son preference (Bhat/ Zavier 2003: 639, 644; International Institute for Population Sciences (IIPS)/Macro International 2007: 106). In fact, the alleged north-south divide is an imprecise re- fl ection of India’s diversity which needs to be refi ned: Son preference is prevalent in Son Preference in India: Shedding Light on the North-South Gradient • 79 north-west India and is less common in south-east India. For reasons of simplifi ca- tion, we talk about north-south differences in the following. The aim of this paper is to contribute to the explanation of this regional pattern, which in turn implies understanding the rationale behind the son preference. We apply the revised value of children (VOC)-approach (Nauck 2001) which is grounded in the rational choice tradition. Accordingly, we argue that gender bias is caused by the sex-specifi c expectations of benefi ts and costs on the part of (potential) parents. These child-related expectations arise from the respective mix of parents’ socioeco- nomic characteristics and the cultural setting in which they live. Prominent regional variations in these background variables imply a favourable cost-benefi t balance of sons compared to daughters in north India. We test the developed theoretical model using data from the international Value of Children-replication study (Trommsdorff/ Nauck 2005). We use a sample of 1,173 mothers collected in the north-central Indian state of Uttar Pradesh in 2002 and in Puducherry, which is located on the east coast in south India in 2010. Son preference is captured by an attitudinal measure. Since the study neglects the male perspective, the prevalence of son preference may be underrated but the interrelations between the explanatory variables and the out- come should not be biased. The remainder of our article is organised as follows: In the next section, the theoretical model is outlined. Its major implications are discussed in the light of the research on son preference in India in section three. This is followed by a sketch of the theoretical model and the presentation of the hypotheses. Section fi ve provides the data and measures employed for the analyses. Findings are shown in section six and are discussed in the fi nal conclusion. 2 Theoretical Background: The Value of Children-Approach Numerous studies are dedicated to son preference in India but most of them fo- cus on its drastic consequences for society in general, and for girls and women in particular. Much less research explores son preference as an outcome. The few ex- ceptional studies conclude that son preference is deeply entrenched in India’s eco- nomic, social, and religious structure (Chamarbagwala 2011; Das Gupta et al. 2003; Diamond-Smith et al. 2008; Dyson/Moore 1983; Gaudin 2011; Nasir/Kalla 2006; Pan- de/Astone 2007; Probst 2009; UNFPA 2012; Vlasoff 1990). More precisely, the sex- specifi c value of children in India stems from widespread agricultural (wheat) pro- duction, the great importance of religion, the persistence of the caste system, the patrilineal and patriarchal organisation of society and the related custom of dowry. Yet a theoretically driven and stringent conceptualisation which relates signifi cant background characteristics, child related benefi ts and costs, and the correspond- ing sex bias, is rarely provided. Moreover, previous empirical work did not aim at disentangling the explanatory power of the socioeconomic and cultural variables stressed (except the study by Pande/Astone 2007) and also omitted quantifying the sex-specifi c benefi ts and costs. • Daniela Klaus, Arun Tipandjan80 To fi ll the theoretical gap, we apply the revised (VOC)-approach (Nauck 2001). Some theoretical traditions argue that the decision on the number of children is made on the calculus of child-related benefi ts and costs (Becker 1996; Friedman et al. 2007; Leibenstein 1975; Liefbroer 2005). So does the VOC-approach. It is built on the idea that individuals strive to optimise their wellbeing. For this purpose they invest in their personal resources like educational attainment, money, professional career, leisure time activities, or in social relationships such as the relationship to offspring. Relying on the theory of social production functions (Ormel et al. 1999) the value of children comprises children’s potential to provide comfort, social es- teem, and affection as main aspects of wellbeing. This analytical distinction basical- ly matches what the initial value of children research calls the economic, the social- normative, and the emotional value of children (e.g., Arnold et al. 1998; Kagitcibasi 1982). In addition to the benefi t-side, direct and (indirect) opportunity costs arising from child-rearing are taken into account. The degree to which children are linked to benefi ts and costs depends on the characteristics and resources of the (potential) parents as well as the contextual setting in which they are embedded. It is argued that child-related preferences and fertility behaviour are determined by the child- related benefi ts and costs. Even though relying on this rationale we are aware of the individual’s bounded rationality (Simon 1982) as we assume a subjective perception of costs and benefi ts based on incomplete information. The VOC-approach has been successfully applied to explain cross-national vari- ations in fertility rates (e.g., Nauck 2007) and the historical decline in fertility (e.g., Klaus 2010). Because it is reasonable to assume that boys and girls differ in their benefi ts and costs in many countries worldwide, it seems to be promising to adopt this approach for the investigation of sex preference. Thus, in this paper, we follow the theoretical proposition of the VOC-approach and the corresponding view for- mulated elsewhere: “The calculus of preferences for children of a particular sex can be understood by extending the concept of the value of children to the two sexes separately. If the net utility of having a son outweighs that of having a daughter, parents are likely to prefer sons to daughters. A complex interplay of economic and sociocultural factors determines the benefi ts and costs of a child” (Arnold et al. 1998: 301). 3 The Value of Sons and Daughters in India Against India’s socioeconomic and sociocultural background, existing studies em- phasise the children’s economic-instrumental value, the marriage transaction costs due to the custom of dowry payment, and children’s social-normative value rooted in the dominant patrilineal kinship system and traditional beliefs (e.g., Arnold et al. 1998; Vlasoff 1990). We observe a widespread consensus that daughters are less valuable to parents than sons. Moreover, a strong socially and economically motivated pressure to bear at least one son and to limit the number of girls is em- phasised. In the following, the relevant research literature is reviewed in the light of Son Preference in India: Shedding Light on the North-South Gradient • 81 the VOC-propositions. The focus will be on potential inter-individual variations and variations across regions. 3.1 The Economic-Instrumental Benefi t of Children Agriculture still plays an important role in India’s economy. Today the agricultural sector contributes to only around 17 percent of the gross domestic product of India (Reserve Bank of India 2012: Table 3) but engages around half of the country’s man- power. Strong dependence of agrarian production on the climate and the weather implies an economically hazardous environment for a large part of the population. This is aggravated by the overall lack of public old age security and other insurance schemes covering the risks of life. As a consequence and in line with the key propo- sitions of the VOC-approach, economic incentives to have children are prevalent in India. However, the economic-instrumental utility fl ow is usually in favour of sons (e.g., Cain 1986; Chen et al. 2011; Probst 2009; Vlasoff 1990) for at least two reasons: First, India is a patrilineal society in which patrilocal household formation is inherent (e.g., Dyson/Moore 1983; Singh 2005). Daughters are absorbed into their husband’s line once they are married. Although very valuable for their husbands’ family, wom- en are of limited long-term value for their own parents. Usually the son and his fam- ily of procreation are obliged to take care of his elderly and frail parents. Second, the son’s responsibility is accompanied by his higher productive potential. Having a son means a reliable source of labour in farming where manual labour is still im- portant despite the growing spread of modern technology. Men also have greater opportunities for gainful employment since they are higher educated and qualifi ed than women, who have gained in education and paid labour participation within the recent decades but “are still not encouraged to work outside the home” due to persisting patriarchal norms which restrict their presence in the public sphere (Singh 2005: 150). Men’s superior economic status is reinforced by the dominance of patrilineal inheritance rules which ensure their ownership of productive assets like the land, farm, or the family business. Nevertheless, regional variations can be assumed, fi rst of all, due to the varying strength of the patrilineal kinship regime across the country. In their seminal work Dyson and Moore (1993, see also Das Gupta et al. 2003) identifi ed two systems that meet in India and which are part of larger sociocultural areas. The north Indian system as part of the West Asian model is very rigid as patrilineality is the rule and thus, prevailing gender discrimination is reinforced. Daughters are largely seen as an economic drain on their families. This differs from the south Indian system, which belongs to the South and East Asian kinship model. Here, the logic of patrilin- eality is still dominating but is more fl exible. Endogamy is practiced which refers to the custom to marry within a certain (social) group similar to that of the own family. The newly married couple usually reside separately from the husband’s family (neo- locality) which enables daughters to remain (physically) close to their parents after their marriage. Women may inherit and transfer property rights (Agarwal 1988). As a consequence both sons and daughters might render help for their families • Daniela Klaus, Arun Tipandjan82 and parents in south India. This coincides with women’s higher contribution to the family subsistence in the rice-growing agriculture which dominates in south India. Women play a major role in the rice-production from weeding to harvesting. Thus, they are of relatively high economic value in contrast to central and north India where wheat production is widespread (Reserve Bank of India 2012: Table 23) and basically requires male ”muscle power“ (Pande/Astone 2007: 5). Economic prosperity and educational attainment are also argued to contribute to a smaller gender gap because of a lower overall economic-instrumental benefi t of children. Parents gain fi nancial independence from their offspring if their socioeco- nomic resources increase. On the other hand, impoverished households and low educated segments of the population depend on (minor) children’s contribution to the family’s living. In this respect, again, noteworthy regional differences can be ob- served. In the southern states (as well as a few states in northernmost India) a great part of the population can read and write (around 80 percent) whereas literacy is much lower in the states of the northern central upland (Government of India 2011). A similar pattern is found for affl uence: The north-central strip of India exhibits the lowest gross domestic product, embracing Uttar Pradesh with 1,586 USD per capita in 2009. This contrasts with the much higher GDP in south India, which amounted to up to 6,602 USD in Puducherry in 2009, for instance (The Economist 2012). To summarise, the widespread gender gap in terms of the economic-instrumen- tal benefi t is assumed to be smaller in south India. Here we observe a unique setting where attenuated patrilineality, rice production, and trends of economic prosperity and educational expansion converge. 3.2 The High-Cost Custom of Dowry The majority of child-related costs are of a fi nancial nature. This comprises expendi- tures for children’s upbringing (nutrition, investments in health care), their education and their marriage (direct costs) but also covers the loss of a mother’s contribution to the household’s subsidence or income when she takes care of her children (indi- rect costs). Whereas the latter is assumed to be of the same value regardless of the child’s sex, higher investments into (the income potential of) sons may cause higher son-related direct costs. This sex-related ratio, however, might be reversed if mar- riage transaction costs are taken into account. Whereas the practice of a bride price raised by the husband’s family dominates worldwide, India is one of the few socie- ties to feature the payment of a dowry. Dowries entail a transfer from the bride’s family to the groom and his family at the time of marriage. Studies widely agree that the custom of dowry is a key issue in the understanding of the persistent son prefer- ence in India (e.g., Diamond-Smith et al. 2008; Jaggi 2001). In the literature, several interrelated reasons are discussed that may account for the existence of dowry (e.g., Dalmia/Lawrence 2005). It is common where women’s productive input to the household is low and where patrilineal and patriarchal struc- tures prevail (Dyson/Moore 1983; Jejeebhoy/Sathar 2001; Singh 2005). The dowry payment aims at maintaining the status of the bride’s family since it is considered as a symbol of prestige. Its absolute amount “increase[s] with both the wealth and Son Preference in India: Shedding Light on the North-South Gradient • 83 social status of both sides of the marriage bargain” (Anderson 2007: 163). Moreo- ver, high dowry payments are a strategy for social upward mobility. This is related to the caste-based structure of India which implies that social status is independent of fi nancial capital and is inherited through caste. This motivates hypergamy, i.e. the marriage to a husband of a higher status than the bride’s family. High dowry payments used to be an upper class phenomenon but have been increasingly adopted by lower status and low caste families within the last decades (Miller 1981; Singh 2005). This trend might be triggered by the growing affl uence and welfare dispersion among the Indian population which has enabled a growing share to make use of dowry as a means of increasing status and a rise in family’s prestige (Anderson 2003). Even in traditional bride-price regions in south India like Tamil Nadu, a spread of dowry has been observed (Anderson 2003; Caldwell et al. 1982; Rao 1993; Singh 2005). Since in south India the preference for one child of each sex is common due to the high benefi ts attached to boys and girls, the growing practice of dowry started to raise concern among families about the immense fi nan- cial pressures, at least after having two daughters (Diamond-Smith et al. 2008: 697). To summarise, dowry is mandatory in north India and rural areas due to stronger patrilineality and stronger patriarchial norms and payments are argued to be high- est among the upper class. 3.3 The Social-Normative Benefi t of Children Depending on the social and cultural context in which the family is embedded, children enhance parent’s status attainment (Nauck 2007: 617). In social structures where the continuance of the lineage is a highly salient goal, the birth of a child increases the parental status and provides behavioural confi rmation. In patrilineal societies like India, only male offspring can carry on the family name, continue the family line and therefore, reproduce the family’s social status. Moreover, religious customs and traditions in India bear a high religious bias towards sons. A son pro- vides prestige not only during parents’ lifetime but above all, after their death (Cald- well et al. 1984; Das Gupta et al. 2003; Probst 2009; Vlasoff 1990). For Hindus, who make up the majority of Indians (81 percent), having a son is imperative (Singh 2005: 131). He is the only one who can light the funeral pyre of his deceased par- ents, offer ancestral worship and thus ensure the salvation of their souls. Therefore, in terms of the social-normative value of children, the sex gap is most evident. But since patrilineality and the aim to achieve immortality are most prominent in north- west India (Das Gupta et al. 2003), the social-normative value of sons should be more pronounced here than in south India. To sum up, there is strong evidence that the cost-benefi t balance of children is generally in favour of sons and it is argued that this is the main reason for a wide- spread son preference. Nevertheless, regional differentials can be expected which come into effect due to region-specifi cs in terms of the kinship regime, the signifi - cance of religiosity, and the socioeconomic profi le of the inhabitants. The access to (higher) education and monetary resources as well as sociocultural identity depend on the individual’s place of residence. • Daniela Klaus, Arun Tipandjan84 4 Theoretical Model and Hypotheses Figure 1 provides a sketch of our theoretical model. Whether sons or daughters are preferred or no preference at all exists depends on the respective combination of sex-specifi c benefi ts and costs attached to children by their (potential) parents. The cost-benefi t structure is, however, not arbitrary but a result of the individual’s so- cioeconomic status and its sociocultural embedding – both systematically varying across regions. The review provided has exposed the educational and the welfare levels as meaningful facets of individual’s socioeconomic status. The strength of patrilineality and the magnitude of religiosity are found to be the most relevant fac- tors characterising the individual sociocultural background. The model suggests a complex path dependency of our major outcome variable. The explanation of son preference in India includes the aim of shedding light on re- gional variations by means of the explanatory components in between. This implies a multiple mediator model as illustrated in Figure 1. A set of subsequent hypotheses can be derived. As a starting point, we assume regional variations in the prevalence of son pref- erence. (H1) Son preference is more prevalent in north India than south India. (H2) The regional variations in son preference are mediated by individual charac- teristics rather than being independent (mediation effect). The mediation is assumed to be based on multiple mediators. First of all, regions differ in terms of socioeconomic and sociocultural characteristics (composition ef- fects). (H3) The socioeconomic status is higher in south India than north India. (H4) Patrilineality and religiosity are stronger in north India than south India. The benefi ts and costs related to sons and daughters are the core of the media- tion. They are assumed to depend on the individual’s background variables and, at the same time, to predict the son preference directly (benefi t-cost effects). Fig. 1: Theoretical model Region Sex Preference Socio- Economic Factors Socio- Cultural Factors Sex- Specific Costs Sex- Specific Benefits Source: Own design Son Preference in India: Shedding Light on the North-South Gradient • 85 (H5) Son preference is more likely the higher the relative benefi ts related to sons. (H6) Son preference is more likely the higher the relative costs related to daugh- ters. (H7) The higher the socioeconomic status, the lower the relative benefi ts related to sons. (H8) The higher the socioeconomic status, the higher the relative costs related to daughters. (H9) The higher the level of patrilineality and religiosity, the greater the cost- benefi t structure of children in favour of sons. 5 Method 5.1 Data Analyses were made using data from the international Value of Children-replication study (Trommsdorff/Nauck 2010; Trommsdorff/Nauck 2005). Data were collected in 17 countries worldwide. Within this research program, an Indian sub-sample was realised in two geographical areas that are prototypical for north and south-east India in respect of our outcome and the potential predictor variables. In 2002, the north Indian sample was carried out in the state of Uttar Pradesh which is located in the central upland. Interviews were conducted in Varanasi City and in several villag- es of the Naugarh Block of the Chandauli district. In 2010 a south Indian sample was realised at the south-east coast inside the state of Tamil Nadu. An urban sample was conducted in the city of Puducherry and a rural sample in some villages nearby in the districts Villupuram and Cuddalore. The 8-year gap in fi eldwork might produce biased fi ndings with respect to the prevalence of son preference. An underestima- tion of the gap between the north and south Indian sample can be expected since the juvenile sex ratio has slightly increased between 2001 and 2011 in Uttar Pradesh but not in Tamil Nadu (Government of India 2011). Nevertheless, the bias should be small and regression coeffi cients are not affected. Respondents were interviewed face-to-face with a standardised questionnaire. The questionnaires were available in Hindi (north India) and Tamil (south India). In fact, the study has realised a three-generation sample design of mothers, grandmothers and adolescents. For the present analysis we have selected the sub- sample of mothers. Fathers were not included in this study. Convenience sampling method was applied: Mothers were recruited in institutions like schools, kindergar- tens, and self-help groups. Rural families were approached directly in their homes. Altogether 1,173 valid interviews are available that are equally distributed over four regions we differentiate in order to capture the hypothesised north-south gradient but also to control for possible urban-rural differences: north-rural, north-urban, south-rural, and south-urban. The surveyed mothers are in different stages of their life: They are aged between 16 and 65 years with an average age of 32 years. Moth- ers in north India are somewhat older (rural = 33; urban = 35) than their southern counterparts (rural = 32; urban = 33). This corresponds to the higher number of • Daniela Klaus, Arun Tipandjan86 children in the north Indian sample (rural = 3.0; urban = 2.3) compared to the south Indian sample (rural = 2.0; urban = 2.0). 5.2 Measures Son preference is covered by an attitudinal indicator. Respondents were asked for their sex preference: “If you had exactly one child, would you prefer it to be a boy or a girl or is it all the same to you?” A more implicit measuring was realised in the Indian National Family Health Survey (NFHS) where sex preference was assessed by asking for the ideal number of children and its distribution on both sexes (Inter- national Institute for Population Sciences (IIPS)/Macro International 2007: 103). The forced choice questioning applied here may provoke higher rates of hidden sex preference but obviously does not as demonstrated later (Fig. 2). Overall, the major- ity state no preference (75 percent). Apart from that, son preference is more preva- lent (20 percent) than daughter preference (6 percent). To mitigate adverse conse- quences of a potential answer bias due to the fact of varying sex-compositions of children already born to the respondent, the number of sons is included as control variable (Gaudin 2011). For the multivariate analysis the indicator is dichotomised: son preference (score 1) versus daughter or no preference (score 0). The sex-specifi c costs of children are measured by a single question asking: “Who costs the most fi nancially: sons or daughters or do they cost about the same?” Cor- responding to the distribution of the sex preference, the share of those who declare that child’s sex does not matter in terms of expenditures is highest (71 percent). Only 6 percent of the respondents assess sons to be more expensive and 23 per- cent vice versa. Again, the categorical variable is recoded into a dummy indicator with score 1 indicating relative higher costs attached to daughters. Two scales are available asking for the benefi ts expected from grown-up daugh- ters and grown-up sons. Among others, they cover instrumental and fi nancial as- pects by two items which refer to care in old age and fi nancial assistance provided by the adult child. Respondents rated them using a response-scale ranging from “not very important” (score 1) to “very important” (score 5). Both items are signifi - cantly interrelated (p < 0.01) which applies with respect to sons (r = 0.56) and with respect to daughters (r = 0.45). They were summed separately for daughters and sons. To obtain a measure of the relative benefi t expectations, both sex-specifi c scores were subtracted and then respondents were classifi ed: Mothers with a value within the range of one standard deviation were assigned to the group with no sex-specifi c benefi ts. They have provided nearly the same level of benefi t-expec- tations towards daughters and sons. Again, this applies to the majority of our sam- ple (55 percent). Mothers beyond one standard deviation were classifi ed either as higher daughter-benefi ts (10 percent) or as higher son-benefi ts (34 percent). In our multivariate model higher son-benefi ts (score 1) are contrasted to any other expec- tations (score 0). The benefi t indicator just presented is restricted in two ways: Expectations to- wards minor children are not covered nor are social-normative benefi ts. The fi rst issue is less harmful as the relevant sex gap in instrumental and fi nancial terms Son Preference in India: Shedding Light on the North-South Gradient • 87 only starts to widen when children marry. Thus, the focus on long-term benefi ts should cover a large part of the expected sex differences. The second objection is more precarious because the gendered social-normative benefi ts are assumed to account for the son preference to a great deal. To alleviate this weakness, we rely on another scale asking for a number of reasons for having children. The item “a duty according to belief” is employed as a proxy to capture a son’s relevance to carry on the family name and maintain parents’ afterlife. The same fi ve-point response scale is used as for the sex-specifi c benefi ts. The overall mean score amounts up to 2.7 (SD = 1.50). The respondent’s socioeconomic position is captured by two indicators: the household’s welfare level and the educational level of the surveyed mother. The wel- fare is measured by a cumulative count of the actual property held by the household in which the respondent lives. It covers a summer house, gold, jewellery, stocks, and car. Home ownership and the holding of a plot of land were intentionally excluded. They are the rule for the rural population and we suspect them to be an indication of production means and the rural mode of living rather than a measure of affl uence. On average, households possess one of the fi ve items (M = 0.9; SD = 1.12). Education comprises four groups. The group of respondents with no degree (score 0) is the largest (32 percent). Respondents who have completed primary (score 1) and secondary school (score 2) are found in equal shares: 26 percent. The remaining 17 percent of respondents have a tertiary degree. Assuming that re- spondents can convert their education into a corresponding career and salary, this indicator provides approximate information about their wage potential and thus, their dependency on children’s fi nancial support. The sociocultural setting is described by patrilineality and religiosity. An attempt to measure the fl exibility of the patrilineal rules was made by using the respondent’s contact pattern to the affi nal (husband’s) kin in relation to the consanguineous (own lineage’s) kin. The data provide the contact frequency for the oldest brother, the old- est sister, the mother, the father and the four in-law pendants. Contact comprises face-to-face contact and contact mediated by letter, email, and phone. A fi ve-point answer scale was applied ranging from “daily” (score 1) to “less than once a month” (score 5). Separately for every four positions, a relative contact rate was computed. A positive score indicates more frequent contact with the affi nal family member and a negative score indicates more contact with the consanguineous family member. The four single indicators were then aggregated by computing an average score. This fi nal indicator stretches from “low fl exibility of patrilineal rules” (score +4) to “high fl exibility of patrilineal rules” (score -4). The score zero indicates a balanced contact rate to the members of both lineages. Overall, respondents tend to pro- nounced patrilineality (M = 1.0; SD = 1.81). Nearly all of our respondents are affi liated to Hinduism (96 percent). Regardless of their belief, they were asked to rate how important their religion is to them on a fi ve-point scale ranging from “not important at all” (score 0) to “very important” (score 4). The mean value proves a very high level of religiosity for our sample (M = 3.4; SD = 0.79). • Daniela Klaus, Arun Tipandjan88 To control for the different stages of life in which respondents were surveyed and for the diverse experiences they have with sons, we include in the analysis the respondent’s age (range = 16-65; M = 33; SD = 8.06) and the number of sons born (range = 0-8; M = 1.2; SD = 0.99). 5.3 Analytical Approach Structural equation modelling (SEM), specifi cally path analysis, is applied to test for the hypothesised composition effects related to the respondent’s regional location as well as to test for the suggested mediating position of the cost- and benefi t- structure. Direct and indirect effects are computed simultaneously and the sobel- test (z-score) is provided to test for the statistical signifi cance of the indirect effects. Weighted least square estimations (WLSMV) are carried out to account for the com- bination of binary and continuous indicators. Linear regressions are obtained in case of continuous dependent variables and probit regressions are obtained in case of binary dependent variables. Basically, standardised probit coeffi cients can be in- terpreted much the same way as OLS-regression coeffi cients. Positive coeffi cients indicate positive effects of the predictor variable on the outcome and vice versa. 6 Results 6.1 Bivariate Findings Figure 2 illustrates differences in the sex preference across the regions. As hypoth- esised (H1), the percentage of mothers who state a preference for a son is signifi - cantly higher in the north Indian sample (32 percent in the rural and the urban sam- ple) compared to the south Indian sample (7 percent in the rural and 6 percent in the urban sample). Apparently, sex preference is the exception in south India: For a great majority of respondents, having a son or a daughter is all the same (82 percent in the rural and 88 percent in the urban sample). In fact, having no preference domi- nates throughout our sample, which contradicts the widespread view of a common son preference in India. This may result from the sample selection (mothers in two selected areas of India) and the use of an attitudinal measurement of sex preference instead of using a behavioural outcome. Using a similar measurement, the Indian NFHS reports very similar distributions for Uttar Pradesh and Tamil Nadu which promotes our results (International Institute for Population Sciences (IIPS)/Macro International 2007: 106). At the same time, this fi nding may also indicate (the onset of) a change toward greater gender equality starting with changing attitudes. It is also noteworthy that only a few more southern respondents prefer a daughter to a son compared to their northern counterparts and the likelihood for son preference does not vary between urban and rural mothers, neither in north nor in south India. At this aggregate-level, the region-specifi c pattern of sex preference revealed corresponds to the regional cost- and benefi t-structure in the way we have pro- Son Preference in India: Shedding Light on the North-South Gradient • 89 posed. The four columns in the left part of Figure 3 prove that signifi cantly more respondents attribute a relatively higher benefi t level to sons in north India than in south India. The percentage is highest in the urban sample of north India (41 per- cent) and lowest in the urban south Indian sample (16 percent). It is remarkable that the lower prevalence of higher son benefi t in south India does not result in a high- er percentage of balanced benefi t expectations but in a higher daughter benefi t: 39 percent of the rural and even 50 percent of the urban respondents in south India attribute higher fi nancial assistance and old-age support to daughters. This even ex- ceeds the percentage of mothers who expect higher son benefi ts. The widespread expectation of higher benefi ts linked to a daughter is in line with the relatively high percentage of daughter preference found in our southern sample (Fig. 2). The higher benefi t of sons to care for their elderly parents (fi nancially) in north India is complemented by their social-normative benefi t (not shown in Fig. 3): It is rated highest among mothers in urban north India (M = 3.2) and lowest in urban south India (M = 2.0). Both rural sub-samples range in between (M = 2.7). The mean differences are signifi cant (p < 0.001). Altogether, the overall son-benefi t is found to be lowest in our urban south Indian sample and the highest in urban north India. Regional differences are less pronounced in terms of child related costs but still match the sex preference pattern (see right four columns in Fig. 3). The vast major- ity of respondents in south India make no distinction between sons and daughters with respect to monetary expenses: 75 percent (rural) and as many as 83 percent (urban) of the respondents in south India assess both cost the same, coinciding with Fig. 2: Region-specifi c sex preference (in percent) Source: VOC-data India, N = 1,173 0 20 40 60 80 100 Rural Urban Rural Urban Daughter Same Son North South • Daniela Klaus, Arun Tipandjan90 their widespread lack of sex preference. In north India, equal costs are reported most often as well – by around two thirds. Higher costs attached to daughters form the second largest group in every considered region but with the highest share among village mothers in north India (35 percent). This may point to high dowry payments in relation to fairly limited fi nancial resources. Based on the bivariate observation, we can draw a preliminary conclusion. Whereas females’ relatively high economic and instrumental value in south India together with predominately equal costs attributed to sons and daughters may produce a high rate of those who have no sex preference at all, in north India the stronger son preference may be triggered by high benefi ts related to sons accom- panied by high costs related to daughters. Comparing rural and urban locations, no systematic variations are found. These initial comparisons of group-means only provide a rough picture and supply no information on associations at the individual level. The subsequent multivariate test provides an in-depth analysis of the internal structure considering additional potential predictors. Fig. 3: Region-specifi c costs and benefi ts attached to children (in percent) 0 20 40 60 80 100 Ru ra l U rb an Ru ra l U rb an Ru ra l U rb an Ru ra l U rb an Daughter higher Same Son higher North SouthNorthSouth Relative Benefits Relative Costs Source: VOC-data India, N = 1,173 Son Preference in India: Shedding Light on the North-South Gradient • 91 6.2 Multivariate Findings The theoretical model was applied to the data by SEM. In the full baseline model (not shown) the respondent’s degree of religiosity is found to be of marginal explanatory power which contradicts our hypotheses (H4 and H9). Religiosity is related neither to sex-specifi c child-related costs nor to benefi ts. Less prevalent son preference among south Indian respondents does not stem from their lower religiosity. Instead, mean values of religiosity are found to be high across the regional sub-samples with minor differences (M = from 3.3 to 3.5). As a consequence, we exclude religiosity from the fi nal model which is illustrated in Figure 4. In Table 1 the standardised coeffi cients and the z-statistics of the fi nal model are provided. We report all direct effects and the indirect effects related to the re- spondent’s regional residence. Two common fi t indices are provided to assess how good the fi nal model matches to our data: CFI = 0.970, RMSEA = 0.052. Following the cutting points suggested by Hu and Bentler (1999) values higher than 0.95 for CFI and lower than 0.06 for RMSEA indicate a relatively good fi t. Accordingly, our hypothesised model is found to be of reasonable quality. Mediation effect Whereas the bivariate test has already identifi ed the hypothesised north-south dif- ference in the prevalence of son preference (Fig. 2), the multivariate model adds some enlightening information: The overall region effect splits up in signifi cant di- Fig. 4: Explanation of son preference (fi nal model) Property Higher costs of daughter Social-normative son-benefit Son preference Higher instrumental- economic son-benefit Education Patrilineality Region Source: Own design • Daniela Klaus, Arun Tipandjan92 rect and signifi cant indirect effects. The indirect effects of respondent’s residence on the son preference are provided in section IV of Table 1: Mediated by region- specifi c background variables and the cost-benefi t structure of children respond- ents in north urban (-0.165), south rural (-0.140), and south urban India (-0.199) have signifi cant lower rates of son preference than mothers in north rural India (reference group). Basically, this confi rms our hypothesis (H2). The region effect is, however, only partly mediated because signifi cant direct ef- fects remain in existence (section I, Table 1): Corresponding to the indirect effects, south rural (-0.232) and south urban respondents (-0.210) are revealed as having a lower rate of son preference than respondents in north rural India (reference group). But remarkably, when controlling for the indirect paths, respondents in north urban India show the most pronounced son preference. The combination of their positive direct (0.159) and their negative indirect effect (-0.165) on son preference obviously levels out and thus, was masked in the bivariate test. These fi ndings suggest region- specifi c characteristics beyond the socioeconomic status and the kinship structure to be responsible for north urban respondents’ higher probability of son preference compared to their southern and even rural counterparts. In general, as direct region effects diminish but do not disappear in the mediation-model, unobserved factors or mechanisms relevant for son preference exert infl uence. Composition effects Still, a substantial part of the variation in son preference over the regions stems from region-specifi c distributions of the respondents’ characteristics considered – yet, some of the effects are unexpected or were not anticipated. For a detailed look we focus on section II in Table 1. Respondents in the urban sample in north India turn out to be upper class members. They live in the most affl uent house- holds (0.520) and are the highest educated (0.895). Respondents in south India are middle educated with no substantial rural-urban differences (0.475 for rural and 0.564 for urban respondents) and northern village respondents are least educated (reference group).The latter live in the most pronounced patrilineal environment as signifi cantly lower levels of patrilineal strength are found for the other three sub- samples: -0.374 for north urban, -0.452 for south rural, and -0.486 for south urban respondents. These fi ndings call for a refi nement of our composition hypotheses: There is no simple north-south gap but a rather complex rural-urban distinction has to been added: Whereas in south India, urban and rural respondents are fairly alike, we observe a great difference in north India. In line with our hypotheses, villagers in north India have a lower socioeconomic status (H3) and stronger patrilineality than respondents in south India (H4). This, however, does not apply for urban respond- ents in north India: In contrast to their rural counterparts, they have a higher so- cioeconomic status and do not differ in terms of patrilineality from respondents in south India. Nevertheless, the regional profi les contribute to the explanation of the regional variations in son preference. The region-specifi c levels of affl uence and education Son Preference in India: Shedding Light on the North-South Gradient • 93 are linked to son preference although somewhat different from our hypotheses. Whereas affl uence is found to be positively related to son preference, education is negatively related. This is evident for both variables’ direct impact (section I, Ta- ble 1: 0.083 for property and -0.148 for education) but also applies for their indirect impact via the individually held benefi t-and cost-expectations (see the following section). Benefi t-cost effects If we focus on the explanatory contribution of the benefi ts and costs attributed to children, we learn fi rst that they signifi cantly predict son preference in the expected way (section I, Table 1): Respondents are more likely to prefer a son if they per- ceive relative higher daughter-costs (0.186), relative higher instrumental-economic son-benefi ts (0.128), and higher social-normative son-benefi t (0.144). This is in line with hypotheses (H5 and H6). At the same time, one of the key propositions of the VOC-approach is confi rmed: The preference for children of a certain sex is at least partly based on a consideration of the benefi ts and costs expected from both sexes. If (potential) parents perceive a higher net-benefi t of a son, they wish to have a son instead of a daughter. This is reasonable in the light of the limited pool of (fi nancial and time) resources people have and the assumption that they tend to spend the resources with the objective of optimising personal profi t. However, the socioeconomic background variables show a fairly inconsistent pre- diction of the child-related costs and benefi ts and the sociocultural variables do not even add to their explanation at all (section III, Table 1). Neither child-related costs nor benefi ts are related to the degree of patrilineality (and religiosity, which has been already omitted from the fi nal model). This fi nally disproves hypothesis (H9). If we inspect the socioeconomic variables in more detail, we discover that re- spondents from affl uent families are more likely to expect that daughters cost more than sons (0.143). This confi rms the hypothesis (H8) which was informed by the assumption of large dowry payments in upper class families. At the same time, an unexpected fi nding emerges: Upper class respondents expect higher son-benefi ts of both kinds: instrumental-economic (0.079) and social-normative (0.199). This dis- proves hypothesis (H7) which was basically built on the idea of lower reliance on child-related (old-age) support and a less adherence to religiously-motivated ben- efi ts provided by sons among affl uent families. The fi nding might indicate that we have overestimated the greater independence of upper class families on child re- lated benefi ts and the related levelling effect on the gender gap. The unexpected higher son benefi ts in the upper class may be caused by persistently strong patri- archal norms. They may, in turn, imply that women are more likely to be excluded from labour market and are subordinate in religious ceremonies like the funeral. In total, the level of affl uence has a positive effect on son preference which challenges the few studies that proved minor but negative wealth effects on son preference (Bhat/Zavier 2003; Gaudin 2011; Pande/Astone 2007). • Daniela Klaus, Arun Tipandjan94 Tab. 1: Explanation of son preference (regression coeffi cients based on SEM) Standard Coeffi cient z-score (sig.) I Direct effects on son preference (probit) Region (ref.: north-rural): north-urban 0.159 2.192* south-rural -0.232 -3.596** south-urban -0.210 -2.997* Property 0.083 1.821+ Education -0.148 -2.170* Patrilineality 0.115 2.146* Higher costs of daughter 0.186 3.287** Higher instrumental-economic son-benefi t 0.128 2.422* Social-normative son-benefi t 0.144 2.884** II Direct effects on socioeconomic and sociocultural variables Property (linear) Region (ref.: north-rural): north-urban 0.520 15.795*** south-rural 0.101 3.295** south-urban -0.131 -3.333** Education (linear) Region (ref.: north-rural): north-urban 0.895 22.981*** south-rural 0.475 13.228*** south-urban 0.564 15.456*** Patrilineality (linear) Region (ref.: north-rural): north-urban -0.374 -9.278*** south-rural -0.452 -11.851*** south-urban -0.486 -12.804*** III Direct effects on cost- and benefi t variables Higher costs of daughter (probit) Property 0.143 3.236** Education -0.291 -5.675** Patrilineality -0.019 -0.414 Higher instrumental-economic son-benefi t (probit) Property 0.079 1.767+ Education -0.045 -0.895 Patrilineality -0.023 -0.489 Son Preference in India: Shedding Light on the North-South Gradient • 95 In line with previous fi ndings, women’s education has an overall negative ef- fect on son preference (Bhat/Zavier 2003; Pande/Astone 2007). Our fi ndings add to the existing research because they demonstrate that the negative effect is partly mediated by sex-specifi c costs and benefi ts: The higher the education, the lower the probability of stating higher daughter-costs (-0.291) which is found to promote son preference. This fi nding challenges hypothesis (H8). It suggests that women’s education is less an indication of socioeconomic class and thus linked to the need and amount of dowry (as we have assumed). Instead, high education obviously refl ects less gendered investment strategies in children and the same educational aspirations for any offspring regardless of the sex. It might indicate a more egalitar- ian environment in which women enjoy higher status and are less constrained by rigid patrilineal and patriarchal norms. This might also be applicable to the fi nding that social-normative reasons for sons become less important with better education of mothers (-0.128). Sex-specifi c attitudes and attributions might be less dominant among well educated women and thus, implying wider role defi nitions and oppor- tunities for them. Standard Coeffi cient z-score (sig.) Social-normative son-benefi t (linear) Property 0.199 5.220*** Education -0.128 -3.174** Patrilineality -0.019 0.577 IV Indirect Effects on son preference (probit) Region (ref.: north-rural): north-urban -0.165 -2.764** south-rural -0.140 -3.527*** south-urban -0.199 -4.083*** R² Son preference 0.330 Chi² (df) 63.690 (13)*** RMSEA / CFI 0.052 / 0.970 Tab. 1: Continuation Note: WLSMV estimation. ***<.001; **<.01; *<.05; +<.1. Controlled for age, number of sons, and interrelations between the cost and benefi ts aspects and between the socioeco- nomic and sociocultural variables. Source: VOC- data India. N = 1,173 • Daniela Klaus, Arun Tipandjan96 7 Discussion In this article, we aimed to explain son preference in India and its variations across the country. State-wise juvenile sex-ratios and individual sex preferences based on the Indian NFHS strongly suggest a persistent north-south gradient in son prefer- ence. This national pattern was confi rmed by our two Indian sub-samples which are exemplary for north and south-east India (compare Fig. 2). To shed light on the underlying mechanisms that may be responsible for these variations, we proposed a theoretical model following the VOC-approach. As im- plicitly done by most of the research on this issue, we explicitly relied on a cost- benefi t rationale behind the sex preference. We argued that son preference refl ects a cost-benefi t balance in favour of sons which in turn is triggered by a specifi c com- bination of individual characteristics. Hence, extending the previous research, the major purpose of our analysis was to provide empirical evidence for the explanatory power of the sex-specifi c costs and benefi ts and the region-specifi c distribution of relevant socioeconomic and cultural factors. The complex theoretical model was translated by SEM that is most adequate to disentangle direct and indirect effects. A unique opportunity was provided by the VOC-data because they cover attitudes towards the costs and benefi ts separately for daughters and sons. Moreover, data were gathered in two distinct regions of India. Some hypotheses were confi rmed whereas others were disproved: Son pref- erence was indeed more probable among mothers with higher expectations of son-benefi ts and daughters-costs. This signifi cantly adds to the empirical state of research as, to our knowledge, this has not been demonstrated by any other study. However, respondents’ affl uence, educational level, patrilineal contact pat- terns, and their degree of religiosity hardly affected their cost-benefi t-structure in the expected way. In fact, the direct effects of these background variables on son preference dominated. The almost complete insignifi cance of cultural aspects is one of the most puzzling fi ndings because almost all work in this fi eld underlines the patrilineal structures and the Hindu Code as vital determinants for the pronounced son preference in India. At this point it cannot be ruled out that data limitations account for the small or even missing effects, at least to some extent. This especially applies to the meas- urements of the benefi t-concept of children. The social-normative facet is under- represented fi rst, because it is restricted to motivations based on religious grounds. It ignores children’s alternative value such as for the intergenerational transfer of the family’s wealth, the continuance of social approval and status (even beyond parents’ death), and thus the reproduction of the social order by the offspring. Sec- ond, it is not measured separately for daughters and sons. The indicators are thus suboptimal and do not pay enough attention to Indian’s specifi cs. This shortcoming results from the fact that the Indian VOC-data were collected within a larger inter- national project with a somewhat different focus, namely the provision of country- equivalent data and the exploration of the overall fertility instead of gender biased preferences. This might also adversely affect the coverage of the respondents’ cul- tural background. The strength of the patrilineality – as a very complex concept – Son Preference in India: Shedding Light on the North-South Gradient • 97 was only indicated by a signifi cant but single aspect: the intergenerational contact pattern. Marriage and inheritance practices are also highly relevant facets of the kinship regime but were not available in the present data. One explanation for the surprising insignifi cance of religiosity is related to its at- titudinal indicator which poorly differentiates between the respondents. As argued elsewhere (Glock 1954), it is more effective to add religiously-motivated behaviour such as attendance at religious ceremonies. This might be less sensitive towards response biases such as social desirability. Given these shortcomings, our fi ndings should be interpreted with caution. The effects might be underestimated. Validating the cost-benefi t argument based on a broader and more comprehensive array of measurements is strongly recommended. At the same time, however, the theoreti- cal perspective on son preference needs to be revisited. We come to a similar conclusion when we review the composition effects. Prom- inent variations in son preference across the regional sub-samples emerged. As expected, a signifi cant part of the variance could be interpreted in terms of socioec- onomic and cultural specifi cs, but a substantial proportion remained unexplained. Highly remarkable is the most pronounced son preference in the north-rural sam- ple when controlling for the background variables considered. A more detailed in- spection reveals that the majority of these respondents are married to husbands who work in well-paid positions, mainly as white collar workers (51 percent) and in the public or fi nancial sector (59 percent). We observe a quite homogeneous urban sample which is very modern and affl uent in socioeconomic terms but adheres to traditional, sex-segregated attitudes and norms, especially when compared to their southern counterparts. On the one hand, facets of costs and benefi ts that were not adequately considered in our study might help to dissolve the strong direct effects that emerged. On the other hand, the substantial amount of unexplained variance in son preference (R² = 0.33) calls for a critical review of the widely held set of argu- ments which are cited to produce son preference. For a satisfactory explanation, future research should pay more attention to mechanisms beyond the cost-benefi t argument. In this respect it might be fruitful to consider son preference separately for men and women and to discuss the marital decision making process in case of spous- es’ disagreement. One major limitation of our sample is its restriction to mothers. The high percentage of respondents without any sex preference shown might be overestimated since no men or childless women were available. It is reasonable to assume that male family members are more concerned about lineage related benefi ts like the continuation of the family and thus have a stronger interest in male offspring. Moreover, patriarchal structures imply that men are more powerful and therefore in a better position to assert their interests in terms of family planning. Thus, the exploration of sex-specifi c mechanisms and dyadic dynamics might add to existing knowledge. This also applies to the analysis of the development of pref- erences along the life course. A childless woman’s preference may change after her transition to motherhood and following every subsequent birth as an adaptation either to personal experiences with children of a certain sex (as a kind of correction) or to the present sex-composition (as an attempt to establish consistency). • Daniela Klaus, Arun Tipandjan98 A fi nal issue is related to the measurement of son preference: In general, sex preference is a strong motivator for sex-specifi c fertile behaviour which fi nally man- ifests in a more or less corresponding sex-composition of the children. But because of, inter alia, unequal access to sex diagnosis technology and abortion or potential spousal disagreement on preferences, individual sex preferences and the resulting behaviour do not necessarily correspond. Thus, attitudinal and behavioural meas- ures are interrelated but distinct aspects of sex preference. Whereas the attitudi- nal measure may be subjected to response biases due to social desirability during the interview the behavioural measure may be biased due to unequal abilities to achieve individual sex preferences. To improve the understanding of sex preference and sex-ratios, we thus suggest an extension of the present research by exploring the behavioural side and the link between both. Acknowledgment This study is part of the research project “Value of Children, Fertility and Intergener- ational Relationships in Cross-National Comparison”. The principal investigators are Bernhard Nauck (Chemnitz University of Technology, Germany) and Gisela Tromms- dorff (University of Konstanz, Germany). The Indian collaborators in the study are Ramesh Mishra (Banaras Hindu University, Varanasi, India) and Arun Tipandjan (In- ternational Centre for Psychological Counseling and Social Research, Puducherry, India) who organised the fi eldwork. The project was supported by the German Re- search Foundation (DFG). We would like to thank two anonymous reviewers for their helpful suggestions. References Agarwal, Bina 1988: Who sows? Who reaps? Women and land rights in India. In: The Journal of Peasant Studies 15,4: 531-581 [doi: 10.1080/03066158808438377]. Anderson, Siwan 2003: Why dowry payments declined with modernization in Europe but are rising in India. In: Journal of Political Economy 111,2: 269-310 [doi: 10.1086/367679]. Anderson, Siwan 2007: The economics of dowry and brideprice. In: The Journal of Eco- nomic Perspectives 21,4: 151-174 [doi: 10.1257/jep.21.4.151]. Arnold, Fred; Choe, Minja Kim; Roy, T. K. 1998: Son Preference, the Family-Build- ing Process and Child Mortality in India. In: Population Studies 52,3: 301-315 [doi: 10.1080/0032472031000150486]. Basu, Alaka Malwade 1999: Fertility decline and increasing gender imbalance in India, including a possible South Indian turnaround. In: Development and Change 30,2: 237- 263 [doi: 10.1111/1467-7660.00116]. Becker, Gary Stanley 1996: Eine ökonomische Analyse der Familie. In: Becker, Gary Stanley (Ed.): Familie, Gesellschaft und Politik: Die ökonomische Perspektive. Tübin- gen: Mohr Siebeck GmbH & Co. KG: 101-116. Behrman, Jere Richard 1988: Intrahousehold allocation of nutrients in rural India: Are boys favored? Do parents exhibit inequality aversion? In: Oxford Economic Papers 40,1: 32-54. Son Preference in India: Shedding Light on the North-South Gradient • 99 Bhat, P. N. Mari; Zavier, A. J. Francis 2003: Fertility Decline and Gender Bias in Northern India. In: Demography 40,4: 637-657 [doi: 10.2307/1515201]. Bhaumik, Saurav 2012: Declining Sex Ratio in India and Its Impact with Special Ref- erence to Punjab, Rajasthan and Haryana [URL: http://papers.ssrn.com/sol3/papers. cfm?abstract_id=2037847, 27.11.2012]. Borooah, Vani K. 2004: Gender bias among children in India in their diet and immuni- sation against disease. In: Social Science & Medicine 58,9: 1719-1731 [doi: 10.1016/ S0277-9536(03)00342-3]. Cain, Mead 1986: The Consequences of reproductive failure: Dependence, mobility, and mortality among the elderly of rural south Asia. In: Population Studies 40,3: 375-388 [doi: 10.1080/0032472031000142306]. Caldwell, John C.; Reddy, P. Hemachandra; Caldwell, Pat 1982: The causes of demo- graphic change in rural South India: A micro approach. In: Population and Develop- ment Review 4,4: 689-727 [doi: 10.2307/1972469]. Caldwell, John C.; Reddy, P. Hemachandra; Caldwell, Pat 1984: The determinants of family structure in rural South India. In: Journal of Marriage and the Family 46,1: 215- 229 [doi: 10.2307/351879]. Chamarbagwala, Rubiana 2011: Sibling composition and selective gender-based sur- vival bias. In: Journal of Population Economics 24,3: 935-955 [doi: 10.1007/s00148- 010-0314-z]. Chen, Susan E.; Bhagowalia, Priya; Shively, Gerald 2011: Input Choices in Agriculture: Is There A Gender Bias? In: World Development 39,4: 561-568 [doi: 10.1016/j.world- dev.2010.09.012]. Clark, Shelley 2000: Son preference and sex composition of children: Evidence from India. In: Demography 37,1: 95-108 [doi: 10.2307/2648099]. Dalmia, Sonia; Lawrence, Pareena G. 2005: The institution of dowry in India: Why it continues to prevail. In: The Journal of Developing Areas 38,2: 71-93 [doi: 10.1353/ jda.2005.0018]. Das Gupta, Monica; Bhat, P.N. Mari 1997: Fertility decline and increased mani- festation of sex bias in India. In: Population Studies 51,3: 307-315 [doi: 10.1080/ 0032472031000150076]. Das Gupta, Monica et al. 2003: Why is son preference so persistent in East and South Asia? A cross-country study of China, India and the Republic of Korea. In: The Journal of Development Studies 40,2: 153-187 [doi: 10.1080/00220380412331293807]. Diamond-Smith, Nadia; Luke, Nancy; McGarvey, Stephen 2008: ‘Too many girls, too much dowry’: Son preference and daughter aversion in rural Tamil Nadu, India. In: Culture, Health & Sexuality 10,7: 697-708 [doi: 10.1080/13691050802061665]. Dyson, Tim; Moore, Mick 1983: On kinship structure, female autonomy, and demo- graphic behavior in India. In: Population and Development Review 9,1: 35-60 [doi: 10.2307/1972894]. Friedman, Debra; Hechter, Michael; Kanazawa, Satoshi 2007: A Theory of the Value of Children. In: Demography 31,1: 375-401 [doi: 10.2307/2061749]. Gaudin, Sylvestre 2011: Son Preference in Indian Families: Absolute Versus Relative Wealth Effects. In: Demography 48,1-28 [doi: 10.1007/s13524-010-0006-z]. Glock, Charles Y. 1954: Toward a typology of religious orientation. New York: Bureau of Applied Social Research: Columbia University. • Daniela Klaus, Arun Tipandjan100 Goodkind, Daniel 1996: On substituting sex preference strategies in East Asia: Does pre- natal sex selection reduce postnatal discrimination? In: Population and Development Review 22,1: 111-125 [doi: 10.2307/2137689]. Government of India 2011: Census of India 2011. Hu, Li-tze; Bentler, Peter M. 1999: Cutoff criteria for fi t indexes in covariance structure analysis: Conventional criteria versus new alternatives. In: Structural Equation Mod- eling: A Multidisciplinary Journal 6,1: 1-55 [doi: 10.1080/10705519909540118]. International Institute for Population Sciences (IIPS); Macro International 2007: National Family Health Survey (NFHS-3), 2005–06: India. Mumbai: IIPS. Jaggi, Tonushree 2001: The economics of dowry: Causes and effects of an Indian tradi- tion. In: Journal of Economics 4: 1-19. Jejeebhoy, Shireen J.; Sathar, Zeba A. 2001: Women’s autonomy in India and Pakistan: the infl uence of religion and region. In: Population and Development Review 27,4: 687- 712 [doi: 10.1111/j.1728-4457.2001.00687.x]. Kagitcibasi, Cigdem 1982: Changing value of children in Turkey. Honolulu, Hawaii: East- West Center. Klaus, Daniela 2010: Changing Value of Children and Fertility Transition in Turkey. In: Journal of Comparative Family Studies 41,5: 799-816. Leibenstein, Harvey 1975: The Economic theory of fertility decline. In: Quarterly Journal of Economics 89,1: 1-31 [doi: 10.2307/1881706]. Liefbroer, Aart C. 2005: The Impact of perceived costs and rewards of childbearing on entry into parenthood: Evidence from a panel study. In: European Journal of Popula- tion 21,4: 367-391 [doi: 10.1007/s10680-005-2610-y]. Miller, Barbara D. 1981: The Endangered Sex: Neglect of Female Children in Rural North India. Ithaca, NY: Cornell University Press. Mutharayappa, Rangamut et al. 1997: Is son preference slowing down India’s transition to low fertility? Mumbai: International Institute for Population Sciences. Nasir, Rosina; Kalla, Aloke K. 2006: Kinship system, fertility and son preference among the muslims: A review. In: Anthropologist 8,4: 275-281. Nauck, Bernhard 2001: Der Wert von Kindern für ihre Eltern. In: Kölner Zeitschrift für Soziologie und Sozialpsychologie 53,3: 407-435 [doi: 10.1007/s11577-001-0073-7]. Nauck, Bernhard 2007: Value of children and the framing of fertility: Results from a cross-cultural comparative survey in 10 societies. In: European Sociological Review 23,5: 615-629 [doi: 10.1093/esr/jcm028]. Ormel, Johan et al. 1999: Subjective well-being and social production functions. In: So- cial Indicators Research 46,1: 61-90 [doi: 10.1023/A:1006907811502]. Oster, Emily 2009: Proximate sources of population sex imbalance in India. In: Demog- raphy 46,2: 325-339 [doi: 10.1353/dem.0.0055]. Pande, Rohini P. 2003: Selective gender differences in childhood nutrition and immuni- zation in rural India: The role of siblings. In: Demography 40,3: 395-418 [doi: 10.1353/ dem.2003.0029]. Pande, Rohini P.; Astone, Nan Marie 2007: Explaining son preference in rural India: the independent role of structural versus individual factors. In: Population Research and Policy Review 26,1: 1-29 [doi: 10.1007/s11113-006-9017-2]. Probst, Wiebke 2009: The missing women phenomenon in India: Causes and effects. In: Journal of Young Investigators 19,17: 1-10. Son Preference in India: Shedding Light on the North-South Gradient • 101 Rao, Vijayendra 1993: The rising price of husbands: A hedonic analysis of dowry in- creases in rural India. In: Journal of Political Economy 101,4: 666-677. Reserve Bank of India 2012: Handbook of Statistics on the Indian Economics. Mumbai: India Printing Works. Rosenzweig, Mark R.; Schultz, T. Paul 1982: Market opportunities, genetic endowments, and intrafamily resource distribution: Child survival in rural India. In: The American Economic Review 72,4: 803-815. Sen, Amartya 2005: The argumentative Indian: Writings on Indian culture, history and identity. London: Penguin. Sengupta, Manisha; Agree, Emily M. 2002: Gender and disability among older adults in North and South India: differences associated with coresidence and marriage. In: Journal of Cross-Cultural Gerontology 17,4: 313-336 [doi: 10.1023/A:1023079219538]. Simon, Herbert Alexander 1982: Models of bounded rationality: Empirically grounded economic reason. MIT press. Singh, J. P. 2005: The contemporary Indian family. In: Adams, Bernd N.; Trost, Jan (Eds.): Handbook of world families. Thousand Oaks, California: Sage Publications: 129-166 [doi: 10.4135/9781412975957]. Srinivasan, Sharada 2005: Daughters or dowries? The changing nature of dowry prac- tices in South India. In: World Development 33,4: 593-615 [doi: 10.1016/j.world- dev.2004.12.003]. The Economist 2012 [URL: hhtp://www.economist.com/content/indian-summary, 30. April 2014]. Trommsdorff, Gisela; Nauck, Bernhard 2005: The value of children in cross-cultural per- spective. Lengerich: Pabst Science Publishers. Trommsdorff, Gisela; Nauck, Bernhard 2010: Value of children: A concept for better understanding crosscultural variations in fertility behavior and intergenerational relationships. In: Journal of Cross-Cultural Psychology 41: 637-651 [doi: 10.1177/ 0022022110373335]. UNFPA 2012: Sex imbalances at birth: Current trends, consequences and policy implica- tions. Bangkok: UNFPA. Vlasoff, Carol 1990: The value of sons in Indian village: How widows see it. In: Popula- tion Studies 44,1: 5-20 [doi: 10.1080/0032472031000144356]. Date of submission: 12.02.2013 Date of acceptance: 02.06.2014 Dr. Daniela Klaus (). German Centre of Gerontology. 12101 Berlin, Germany. E-Mail: daniela.klaus@dza.de URL: http://www.dza.de/en/about-the-dza/staff/klaus.html Dr. Arun Tipandjan. International Center for Psychological Counseling & Social Research. 605011 Puducherry, India. E-Mail: aruneps73@gmail.com URL: www.icpcsr.org Published by / Herausgegeben von Prof. Dr. Norbert F. Schneider Federal Institute for Population Research D-65180 Wiesbaden / Germany Managing Editor / Verantwortlicher Redakteur Frank Swiaczny Assistant Managing Editor / Stellvertretende Redakteurin Katrin Schiefer Copy Editor (German) / Lektorat (deutsch) Dr. Evelyn Grünheid Layout / Satz Beatriz Feiler-Fuchs E-mail: cpos@bib.bund.de Scientifi c Advisory Board / Wissenschaftlicher Beirat Paul Gans (Mannheim) Johannes Huinink (Bremen) Michaela Kreyenfeld (Rostock) Marc Luy (Wien) Clara H. Mulder (Groningen) Notburga Ott (Bochum) Peter Preisendörfer (Mainz) Zsolt Spéder (Budapest) Comparative Population Studies www.comparativepopulationstudies.de ISSN: 1869-8980 (Print) – 1869-8999 (Internet) Board of Reviewers / Gutachterbeirat Martin Abraham (Erlangen) Laura Bernardi (Lausanne) Hansjörg Bucher (Bonn) Claudia Diehl (Konstanz) Andreas Diekmann (Zürich) Gabriele Doblhammer-Reiter (Rostock) Jürgen Dorbritz (Wiesbaden) Anette Eva Fasang (Berlin) E.-Jürgen Flöthmann (Bielefeld) Alexia Fürnkranz-Prskawetz (Wien) Beat Fux (Salzburg) Joshua Goldstein (Berkeley) Karsten Hank (Köln) Sonja Haug (Regensburg) Hill Kulu (Liverpool) Aart C. Liefbroer (Den Haag) Kurt Lüscher (Konstanz) Emma Lundholm (Umeå) Nadja Milewski (Rostock) Dimiter Philipov (Wien) Roland Rau (Rostock) Tomáš Sobotka (Wien) Jeroen Spijker (Barcelona) Olivier Thévenon (Paris) Helga de Valk (Brussel) Heike Trappe (Rostock) Michael Wagner (Köln) © Federal Institute for Population Research 2015 – All rights reserved