CURRENT CONTRACEPTIVE USE IN INDIA: HAS THE ROLE OF WOMEN’S EDUCATION BEEN OVEREMPHASIZED?
Sarmistha Pal* and Gerald Makepeace Cardiff Business School, UK.
JEL Classification: J13, I12, O15 Key words: Family planning, Contraceptive use, Reversible and irreversible methods, women’s literacy, Selectivity corrected estimates Forthcoming in European Journal of Development Research 2003 Abstract Using the recent Indian National Family Health Survey (NFHS) data, this paper analyses the factors determining the current contraceptive use in rural and urban West Bengal in eastern India. A bivariate probit model with selection is used to determine the likelihood of not being sterilised and that of currently using some traditional or modern reversible method of contraception among non-sterilised women. Our results suggest that male and female sterilisation is a popular method among the poorer couples with little assets, poor education and more living children. More literate women are, however, more likely to use various reversible methods of contraception though the effect of husband’s education remains insignificant. Relative to women’s education or various household assets, the effect of belonging to an upper caste household is more pronounced on the current use of contraception, especially among rural women. Simulations of the effect of eliminating illiteracy suggest that the quantititive significance of education is small despite its robust statistical significance. Thus there is limited effect of household assets and women’s education on current use of contraception in our sample.
*
Corresponding author. Address: Cardiff Business School, Colum Drive, Cardiff, CF10 3EU, Wales, UK. E-mail:
[email protected]. Fax. 44-029-20-874419. We are grateful to the journal referees for very helpful comments on an earlier draft. We also wish to thank Cardiff Business School and Cardiff University for providing the funds to finance this project and John Cleland, Jocelyn Kynch and Pushkar Maitra for comments on an earlier draft of the paper. The usual disclaimer applies.
CURRENT CONTRACEPTIVE USE IN INDIA: HAS THE ROLE OF WOMEN’S EDUCATION BEEN OVEREMPHASIZED?
1. INTRODUCTION While India was one of the first developing countries to adopt family planning as an important objective of the central government policy as early as 1951, India’s population has more than doubled since 19611. The birth rate per thousand population was 29 during 1990-95 as compared to 21 in Sri Lanka and 14 in more developed regions (including 11 in Japan). In 1992, India strengthened its family planning efforts and aimed at increasing the use of contraceptives from 41% to 60% of married women in order to achieve its goal of a replacement level of fertility (Saxena, 1996). The question that we raise here is why irrespective of long-drawn government efforts, contraceptive use in general and use of modern methods like pills, condoms in particular has remained low in India. Using the recent Indian National Family Health Survey (NFHS) 1992-93 data, this paper analyses this question using selectivity-corrected estimates of current use of various traditional and modern methods of contraception among non-sterilised rural and urban women aged between 13-49 years in the Indian state of West Bengal. Though Kerala has often been quoted as a success story among the Indian states for its high literacy, high life expectancy at birth, low birth rate and low infant mortality rates (Drèze and Sen, 1995), West Bengal is an interesting case because of the large size of its population and its pattern of contraceptive use. In the post-independence period West Bengal started its economic development in a relatively favourable position among the Indian states in terms of its high literacy, low poverty, high industrialisation and urbanisation; but the state’s success has been rather moderate since then, especially over the 80s and the 90s (see Table 1). By the early 1990s, West Bengal lagged behind many states in terms of female literacy and female labour force participation rate. Though infant mortality rate (IMR) has declined in the state between 1981 and 1991, the rate of decline was surpassed or equalled by Bihar, UP, Gujarat, Kerala and Tamilnadu; the couple protection rate of 34.2
1
Irrespective of the fact that India’s National Family Welfare Programme has averted about 168 million births since its inception in 1951
1
in the state was also lower than the national average of 45.4. It is thus imperative to understand why contraceptive use in West Bengal has remained low if India is to achieve its family planning objectives. Existing studies have identified different factors affecting household fertility behaviour and contraceptive use in low-income countries. Van de Kaa (1996) summarises the findings of half a century of research in to the determinants of fertility and identifies different determinants of fertility. In particular, mention has been made of technological/biological (both natural, e.g., see Cleland and Hobcraft, 1985 and mortality decline oriented, e.g., see Chowdhury et al., 1976), economic (demand oriented, e.g., see Becker, 1965, 1981; Schultz, 1969 and also demand and supply oriented, e.g., see Easterlin, 1978), social (structural, e.g., see Caldwell, 1976 and psychological, e.g., Fawcett, 1970) and cultural (, e.g., Montgomery and Casterline, 1993) factors. While Van De Kaa’s review is rather broad-based, we shall concentrate in this paper on the literature on socio-economic and cultural determinants of contraception use. While most studies in this tradition unanimously suggest that women’s education plays a significant role in lowering fertility or in increasing use of contraceptives (Rosenzweig, M. and D.A. Seiver. 19822), the role of male education (Schultz, 1973) or household income/assets (including non-earned income or physical assets) on fertility and contraceptive may be ambiguous (Rosenzweig and Evenson, 1977, Anderson, 1983). It has also been argued that employment may give women greater command over resources and autonomy in household decision making, but less time for child care and breast feeding (e.g., Basu and Basu, 1991), which, in turn, may be responsible for higher morbidity among their children and, therefore, less use of contraceptives. There is also a complex two-way relationship between child survival and fertility: while some have argued that a high probability of child survival is necessary for couples to use family planning, others have found that unwanted fertility may exacerbate child mortality. Strong ‘son preference’ may further complicate decisions regarding fertility and contraceptive use, e.g., if couples continue to have children until they have a son (Jeffery and Basu, 1996). Cultural and/or locational factors (e.g., urbanization or caste/religion) may also explain some variation in fertility and contraceptive use across regions within the same country (Murthi et al., 1995). There has also been doubt as to the efficacy of family planning policy in many developing countries. Some research argues that improving the quality and quantity of contact between rural women and public sector health and family planning workers enhances the formers’ intention to use contraceptives in the future (see, for instance, Phillips et al, 1986). Others suggest that improving female autonomy through more equitable
2
More recently, using district-level Census data from India, Drèze and Murthi (2000) emphasize the role of women’s education in explaining fertility differences across the country and over time.
2
social and economic development will have a far greater impact on fertility than the provision of family planning services (Bhattarcharya, 1998). The present study examines the relative significance of these socio-economic and cultural factors. Using the detailed data now available from the NFHS for rural and urban West Bengal, we shall jointly estimate the probability of being non-sterilised and of using some reversible (modern or traditional) method of contraception among non-sterilised women, using a bivariate probit selection model, which allows us to correct for any selectivity bias. Our study is significant in a number of ways: (a) while many studies do not distinguish between various modes of contraception, we distinguish between various reversible (e.g., pills, condoms or abstinence/withdrawals) and irreversible (e.g. male/female sterilisation) methods. Given Indian government’s somewhat heavy handed policy in popularising the irreversible method of sterilisation (see discussion in section 2.1), our study focuses on the factors determining the use of various reversible (traditional and modern) methods of contraception among non-sterilised women. (b) We consider the variation in the demand for contraceptive use between rural and urban areas of the state. This dichotomy has been emphasised in the literature, but has seldom been examined, especially in the Indian context. (c) We consider literacy of each spouse and examine the relative roles of each spouse’s literacy on the current use of contraception. In addition to the traditional role of female literacy to delay the age of cohabitation and also to help choosing and using contraceptives more effectively, we argue that female literacy is also an indicator of differential bargaining power of women in decisions to use contraceptives through a reduction in social conservatism. In this respect, we also estimate how a reduction in illiteracy among the sample women would encourage the use of contraceptives in rural and urban areas of the state. (d) Finally, we examine the importance of other possible factors including household caste and assets in relation to woman’s education on the current use of contraception. The paper is developed as follows. Section Two explains the data and the methodology used for determining the current use of contraceptives in the survey year. Section Three estimates a bivariate probit model with selection to empirically determine the current use of contraception among non-sterilised women, with correction for selectivity bias and section four concludes.
2. DATA AND METHODOLOGY The analysis in this paper is based on the National Family Health Survey (NFHS) householdlevel data for 1992-93 from rural and urban West Bengal. The NFHS is a nationally representative survey of ever-married women age 13-49. The NFHS covered the population of 24 States and the 3
National Capital Territory of Delhi to provide demographic and health data for interstate comparisons. The main objectives of the survey was to collect reliable and up-to-date national-level and state-level data on fertility, nuptiality, fertility preferences, knowledge and practice of family planning, the potential demand for contraception, the level of unwanted fertility, utilization of ante natal services, breast feeding and food supplementation practices, child nutrition and health, vaccinations and infant and child mortalityThe survey was carried out as a major component of a project to strengthen the Survey Research Capabilities of the Population Research Centres in India, initiated by the Ministry of Health and Family Welfare, Government of India and funded by the United State Agency for International Development. The International Institute for Population Sciences, Mumbai (I.I.P.S), was designated as the nodal agency for providing co-ordination and technical guidance to the NFHS. The data collection for the NFHS was undertaken by seven Consulting Organizations in collaboration with the concerned Population Research Centres in each State. The East-West Center/Macro International, United States of America, provided technical assistance for all of the survey operations. The success of NFHS 1992-93 in creating an important demographic and health database in India has paved the way for repeating the survey. The second NFHS undertaken in 1998-99 is designed to strengthen the database further and facilitate implementation and monitoring of population and health programmes in the country. The Principal objective of NFHS-2 remained as before though it included some additional information (not collected in NFHS 92-93) relating to measurements of the nutritional status (e.g., height and weight) of all eligible women, blood test of women and children for haemoglobin. However, the information regarding contraception use that we use remained very similar. In the NFHS, a total of 88,562 households were covered, and the interviewers collected information from 89,777 ever-married women age 13-49 (23,455 in urban areas and 66,322 in rural areas). The fieldwork was conducted in three phases between April 1992 to September 1993. Here we use the data collected from the eastern Indian state of West Bengal. The household and women questionnaires of the NFHS provide information about the personal and family characteristics of women aged between 13-49 years. We exclude women who were pregnant at the time of the survey, which yields samples of 3131 and 848 rural and urban women respectively.
2.1. Current Use of Contraception We classify the available methods of contraception into reversible and irreversible methods. While male or female sterilisation is classified as an irreversible method, traditional methods such as abstinence or withdrawal from any intercourse, or modern methods such as use of pill, condoms, 4
injections, IUD/copper etc. are referred to here as reversible methods. We distinguish sterilisation from reversible methods because it has very different consequences. Sterilisation implies that no choices can be made about contraception in the future and the method of contraception is therefore predetermined for the sterilised individuals in our sample. The policies of the Indian government make it likely that outcomes involving sterilisation were determined in a different way to the other methods. Table 2 suggests that about 31% of rural couples and 24% of urban couples were sterilised while as many as 45% of rural women and 40% of urban women in our sample did not, at the time of the interview, use any contraception. This means that only about 18.5% of rural and 24.7% of urban women were currently using some modern contraception (e.g., pills, IUD, condoms etc.) while a 5.4% rural and 11.5% urban women were currently relying on traditional methods. Table 3A examines the supply sources of sterilisation and other modern methods of contraception in order to assess the extent of the public provision of these family planning services in our sample. Among different sources, we can broadly distinguish between government (e.g., government/municipal hospitals, primary health centres and their subcentres, government paramedic etc.) and private sources (e.g., pharmacy/drug store, private hospital/clinic, private doctor etc.). While government-funded hospitals tend to be the major providers of sterilisation services in both rural and urban areas, take-up of government provided pills and condoms seems to be rather insignificant among both rural and urban women in our samples. Users of IUD in rural areas, however small in number, seem to receive them from various government hospitals. In other words, pills and condoms seem to be largely self-provided by the sample women under consideration. Thus the availability of sterilisation in government hospital may reflect a policy priority towards particular long-term, permanent method for lowering fertility rates. In spite of the persistent government policy of encouraging sterilisation as the preferred mode throughout the country, fertility rates were between 4.4. and 5.1 in the northern states of UP, Bihar, MP and Rajasthan while it was only 1.8 in Kerala in 1991, which in turn can be attributed to significantly higher literacy, especially female literacy in Kerala. These regional contrasts within India strongly argue for collaboration (based inter alia on active participation of educated women) as opposed to coercion (Drèze and Sen, 1995). Among the women who are currently not using any contraception and not sterilised, 571 (among 1404) rural and 166 (among 335) urban women in the sample explained the reason for not using contraceptives. This is summarised in Table 3B. Mention was made of (a) desire for further children including a son; (b) menopause /hysterectomy; (c) difficult to be pregnant afterwards and other health problems; (d) family opposition including opposition by husband; (e) religious reason; and (f) inconvenience and dislike for existing methods. Very few women suggested that the current non5
use of contraception was attributable to the lack of knowledge, fear of sterilisation or difficulty to work afterwards. Different contraceptive methods have different levels of efficiency and also involve some fixed and variable costs. Fixed costs include costs of consultations, installation and monitoring of contraceptive performance while among variable costs one can consider the costs of learning to use the contraceptive effectively and also other psychological, monetary and time costs of searching for the best method. For example, the monetary/time costs in using condoms are low compared to pills or IUD but there are variable costs in terms of its effectiveness. By contrast, sterilisation is an irreversible method that cannot be used for birth spacing. There are also various health hazards involved in different techniques. Consequently different couples characterised by different characteristics will optimally choose different techniques. This is further analysed in the next section.
2.2. Methodology We consider a one-period framework, which represents household choice at one point in their life cycle and classify women into sterilised and non-sterilised groups to examine the current use of contraception among the non-sterilised women. This allows us to consider the current use of modern and traditional (reversible) methods of contraception among the non-sterilised couples at the time of the survey. Thus, the current use of reversible method (modern or traditional) for the i-th non-sterilised couple FPi is assumed to depend on individual characteristics (e.g., literacy) of each spouse (XF, XM ), joint demographic and socio-economic characteristics of the household (XH), e.g., household assets/income, caste/religion and also the public environment in which the couple live (XP). The latter determines the provision of utility servic es, namely, housing, sewerage, drinking water, education and health facilities, including the supply of contraceptives. In other words, FPi = g(XF, XM , XH, XP, ε ) where ε is the unobserved household and/or community specific characteristics (e.g., fecundity, efficacy of use) affecting contraceptive use, assumed to be uncorrelated with XF, XM , XH, XP. Here we focus on following hypotheses regarding current contraceptive choice and use in India.
Women’s autonomy: Given the unequal power structure between men and women in a patriarchal society, there prevails an asymmetry between husband and wife in household decision-making (Drèze and Sen, 1995). Thus, a couple’s preferences may not be conjugal and differences may exist in 6
parental preferences for expected family size, choice and use of contraception. In the absence of direct information on parental preferences as well as non-earned income of each spouse, we shall, examine if women’s literacy has a significantly different effect on contraceptive use3 relative to that of her spouse, in this case, via education. In particular, we emphasize the role of women’s education to lower social conservatism against using modern methods of contraception. In other words, educated women may be more receptive to modern social norms and family planning campaigns 4.
Fecundity: Woman’s age is an important argument here and is an observed indicator of fecundity. The use of contraceptives may be higher among younger women because their fecundity tends to be higher. There may also be unobserved heterogeneity in fecundity among women of a given age group, which will be taken care of by the random disturbance term.
Household Wealth: Assuming children to be normal goods, household income is expected to increase the demand for children and, therefore, reduce the demand for contraceptives. However, following the endogeneity argument as put forward by Benefo and Schultz (1996), we shall instrument income by various physical assets owned by the household. Caste: Caste is still an important factor in explaining the present day effects of earlier inequalities in the ownership and control of land and various non-land resources in the Indian society. This in turn is reflected in the economic and political power held by Indian households, especially if they belong to the lower caste (e.g., scheduled caste/tribe as identified by the Indian constitution). Thus generally, higher caste households are economically better off than the lower caste households are. There is also a close correlation between caste and women’s schooling in India: 1991 Census data suggests that while 71% female members (aged 7 and above) belonging to upper caste Hindu households in rural West Bengal were literate, as low as 22% and 5% female members from scheduled caste and scheduled tribe families were literate in the State. In other words, women from scheduled caste and scheduled tribe households in India are more likely to be averse to using modern contraceptive methods.
3
Traditionally it has been argued that women’s schooling may affect contraceptive use in a number of ways: (a) it typically delays the age at cohabitation. (b) More literate women can learn about and use contraception more effectively than uneducated women, thus reducing the number of unanticipated pregnancies. (c) More educated women are likely to be more effective in producing healthy children. 4
NFHS 1992-93 survey suggests that less than 60% of illiterate women in India consider family planning messages in the media to be acceptable as against over 90% of women who have completed high school education.
7
Birth-Spacing:If the family planning services were unavailable, especially in the period following a birth, the risk of pregnancy would be high, which in turn lower women’s capacity to produce healthy child. Frequent pregnancies also affect the time available for child care and subsistence activities. Accordingly it is argued here that the age of the youngest child born would be an important consideration in the use of contraceptives.
Son Preference: If infant mortality rates are high as in India, the choice of contraception may be affected by the number of surviving children. For example, an irreversible method like sterilisation is unlikely to be appealing unless couples have surpassed their reproductive goals and access to contraceptives is poor. More importantly, given the prevalence of son preference in many parts of India (Kishor, 1993), the number of living sons may be more important than that of girls in the choice of current mode of contraception.
Urbanisation: Urbanisation may reduce fertility because children are less likely to contribute to household production and more difficult to supervise in an urban setting. There may also be a concentration of women in urban areas with higher levels of schooling and higher levels of average income. There is also better provision of medical and health care in urban areas and people have greater exposure to mass media as well as wider opportunities to observe and discuss the lifestyles of other social groups. Thus, the demand for contraceptives might be determined differently in urban areas.
It follows from our discussion in section 2 that the provision of free sterilisation in government hospitals has undoubtedly an important role to play, inducing a large number of sample men/women to choose sterilisation as a preferred mode. It is however difficult to ascertain to what extent the choice of sterilisation has been influenced by the free provision of this service or by any other factors listed above affecting the demand for sterilisation. That is why we choose to focus on the choice of traditional and modern reversible methods among non-sterilised women in terms of the arguments listed above, using a bivariate probit selection model. This is done in the next section.
3. DETERMINANTS OF CURRENT USE OF CONTRACEPTION We use a bivariate probit selection model to examine the determinants of the current use of contraception among the rural and urban couples in the NFHS survey who were not pregnant at the time of the survey. Since significant proportions of both rural and urban samples were sterilised, we 8
select the non-sterilised couples to analyse the current use of various reversible methods of contraception, after correcting for the selectivity bias. This is explained in the next subsection.
3.1.
A Bivariate Probit Selection Model
Figure 1 outlines the sequential interpretation that we give to the model. Some people were sterilised so this outcome was pre-determined at the time of the survey. People who were not sterilised then made a decision whether or not to use a method of contraception. People who were sterilised did not make any decisions about contraception.
We therefore observe 3 groups of
individuals: Not sterilised and uses contraception, Not sterilised and does not use contraception, and Sterilised. In technical terms, we use a bivariate probit model for the joint outcomes of sterilisation and contraception with selection into contraception. This model is then estimated by maximum likelihood methods.
Figure 1: The structure of the model
Uses contraception Contraception Not sterilised Sterilisation
Does not use contraception
Sterilised
There is a selection of individuals into the group who make decisions about contraception. Particular types of individual are more likely to be sterilised than others so the individuals making decisions about contraception may not be typical of the general population. The model captures this in two ways. Some of the systematic effects are included as independent variables determining whether the individual is sterilised or not and whether an individual uses contraception or not. Thus we include inter alia education, wealth and caste as explanatory variables. Selection does not pose a problem if we can measure the variables that determine the outcomes. However, there will be other unobserved factors that affect sterilisation and contraceptive use. The model allows for the unobserved variables that affect the sterilisation outcome to be correlated with the unobserved variables that affect the 9
contraception decision. This addresses the implied selection problem.
Our results suggest that
unobservable factors that make an individual more likely to be non-sterilised tend to make the individual more likely to use contraceptives. To illustrate the possible effects of the unobservables, consider the possibilities that independence of mind makes a woman more likely to be non-sterilised and more likely to use conception or that depression makes a woman less likely to be non-sterilised and less likely to use contraception. More formally, the model has two dummy dependent variables, Not sterilised and Uses contraception. The variable ‘not sterilised’ takes the value 1 if the couple is not sterilised and 0 if sterilised. The variable ‘uses contraception’ takes the value 1 if the woman currently uses some reversible (traditional or modern) method and 0 otherwise. The value of each variable depends on a set of observed characteristics contained in the set of regressors and a set of unobserved characteristics contained in the error term. We assume that the error terms in the underlying latent variable model are normal and estimate a bivariate probit model to allow for correlation between the effects of unobserved characteristics in the two equations.
Explanatory Variables As discussed in section 2.2, the likelihood of not being sterilised and that of currently using some contraceptives depends on the individual characteristics of each spouse, joint household characteristics and the public environment. Given the distinctive features of rural and urban areas in India, we have estimated the models separately for rural and urban West Bengal. Let us first consider the determinants of ‘not sterilised’. Among the individual women’s characteristics, we include different categories, for age - ‘20 to 24’, ‘25 to 29’, ’30 to 34’, ‘35 to 39’, and ‘40 to 49’- and, woman’s education level – ‘literate but no schooling completed’, ‘completed primary school’, ‘completed middle school, ‘completed high school, ‘some university education’. The reference groups are age ‘13 to 19’ and ‘woman is illiterate’. We include a variable for whether the husband is literate 5. Sterilisation is unlikely to be appealing to couples who have not achieved their desired family size; accordingly, number of living sons and number of living daughters are included6. It is also expected that women who experience child death or problems of childbirth (e.g., still birth or
5
We also considered the husband’s employment status . However almost 99% of the husbands in our rural and urban samples are employed, so we omitted ‘husband is employed’ from the set of explanatory variables . 6
The analysis is based on a single cross-section data where we exclude the preganant women; hence without much loss of generality we assume that the family size is given at the point of the survey.
10
any delivery complications) are less likely to be sterilised. We consider a number of household asset variables to instrument household income: household ownership of land, radio, television, and bicycle, whether the house is built with bricks. Among household assets variables, radio and television are unique in the sense that they are also mass media of communication and, as such, they also capture the effect of family planning advertisements. Among household socio-cultural characteristics, we include if the couple belongs to a scheduled caste or a scheduled tribe. The variable for scheduled tribe is, however, dropped from the urban sample as there were very few women from scheduled tribe households living in towns and cities. Similar sets of explanatory variables are included for the second stage of estimation of contraceptive use in rural and urban areas. Among the available variables used in the two decisions, we argue that the current age of the youngest child is important in the decision to use contraceptives only. However, numbers of children, reproductive problems and child death are expected to be more important in the decision to be sterilised, but not in the decision relating to current use of contraception. In addition, there were very few rural women currently using some contraception who had literacy above primary level; hence, we have generated a variable to include all levels of education above primary level.
The Appendix contains precise definitions of the variables. Means and standard
deviations of the explanatory variables for rural and urban West Bengal are given in Table 4. One also needs to be cautious about the choice of explanatory variables primarily because of the endogeneity problems in modelling household decisions, which may bias the estimates. For example, income has been argued to be an important variable though household labour income is endogenous to household decisions regarding fertility (Benefo and Schultz, 1996). NFHS data set, however, does not provide information on the household income or expenditure. Accordingly, we use a number of instruments like various household assets, the education levels of the spouses, ownership of land and non-land assets and also caste/religion of the household. Level of schooling may be highly correlated with family income in less developed countries and favourably affect women’s labour market participation, but may have an opposite effect on fertility or child mortality. However, this will be less of a problem in our study, as we do not use information on household income sources. Secondly, in our study of single cross-section data we assume that past fertility as measured by number of surviving children (sons and daughters) and age of the youngest child are given and can therefore be treated as exogenous to the current use of contraception. Finally, it is often argued that decisions regarding women’s employment and fertility are jointly made. We also observe a close correlation between women’s education and employment. Hence, we exclude the woman’s
11
employment variable from our estimation and argue that women’s education can also be an instrument of their employment status.
3.2.
Maximum Likelihood Estimates
In this section, we consider the bivariate probit selection estimates of current contraceptive use, after selecting the women who are non-sterilised and not pregnant in 1992-93. These results are summarised in Table 5. Correlation coefficients between the two sets of unobserved characteristics in the joint determination of contraceptive use and sterilisation are significant and positive for the rural sample at the 1% level showing that modelling selection is crucial for this group. Selection is less important for the urban sample although the coefficient is bordering on significance at the commonly used 10% level. Determinants of non-sterilisation Rural women with low literacy, less land and from lower scheduled caste households are more likely to be sterilised in our sample, irrespective of their age group. However, women experiencing reproductive problems (e.g., still birth and delivery complications) and child death as well as with higher levels of literacy are less likely to be sterilised. More interestingly, rural women with more sons (but not daughters) are more likely to be sterilised. Women’s education, caste and household assets are also significant among urban women while ‘child has died’ and reproductive problems are not. In contrast to the rural sample, both the numbers of living sons and daughters are significant. In other words, sterilisation appears to be appealing to poorer couples identified by low education, low wealth and low caste who have surpassed their reproductive goals; there is also a pronounced son preference among rural women. Determinants of current use of contraception Selecting the non-sterilised women, we consider the determinants of current use of contraception. First observed fecundity as instrumented by age is an important consideration: compared to younger women, both rural and urban women aged 35 to 49 years are less likely to currently use any contraception. In addition, current age of the youngest child affects contraceptive use among urban women: the higher the current age of the youngest child, the higher is the likelihood of currently using any contraception. Women’s literacy has a highly significant impact on the current use of contraception for both rural and urban women. All schooling variables are highly significant for rural and urban women, thus suggesting that more educated women are more likely to use some form of contraception. In contrast, husband’s literacy does not turn out to be of much significance either in 12
rural or urban areas.
Among various household-variables, rural women from scheduled tribe
households are less likely to use any contraception though the caste variable is insignificant in both areas. Ownership of agricultural land among rural women and that of television among urban women significantly enhance the likelihood of currently using some reversible form of contraception. Taken together, women’s fecundity, education, tribal status and, to some extent, household wealth turn out to be significant determinants of the current use of contraception, among rural and urban women. It is only with respect to the ownership of assets that the current contraceptive use of rural and urban women seems to differ. In other words, joint determinants of the likelihood of being non-sterilised and that of currently using some contraception highlight the differences in factors affecting current methods of contraception, both reversible and irreversible. These are also indicative of the differences in contraceptive practices among rural and urban women, which to some extent reflect the differences in the socio-cultural environment as well as the public provision of health and family planning services.
3.3. Differential Role of Women’s Literacy Until recently most policy analyses implicitly viewed the household as having only one set of preferences. However, a growing body of evidence suggests that more effective policy instruments will emerge if the processes by which households balance the diverse interests and preferences of their members (Alderman et al., 1995) are analysed. These processes give rise to differential bargaining power of the household members. Power is a multidimensional concept, which can include attitudinal attributes. We argue that literacy level of each spouse can be used as an indicator of a couple’s preferences and attitudes towards the choice and use of modern contraceptives. In this subsection we shall, therefore, formally test if women’s literacy has significant and differential effect on the use of contraception relative to her spouse or, alternatively, whether wife and husband’s literacy has the same effect on use of contraception. In doing so, we need to modify the specification of the bivariate probit selection model presented in Table 5. This is because in Table 5 wife’s literacy is decomposed into literate, primary, middle and high school and university levels for rural and urban women. However, similar information is not available for husband’s educational status; there is a single variable, showing whether or not the husband is literate at any level. In order to remove this disparity, we create a variable to reflect if the wife is literate at any level. These modified estimates are shown in Table A in Appendix II. Appendix Table A shows that wife’s literacy is highly significant while the husband’s literacy does not seem to have any impact on the current use of contraception among both rural and urban women. 13
This is confirmed by a formal test of the null hypothesis that the coefficients of the woman’s and husband’s literacy levels are the same. We perform a likelihood ratio (LR) test where the estimated statistic is distributed as a chi-square with one degree of freedom (equal to the number of restrictions imposed). The critical value of chi-square at 1% level is 6.6349 while it is 3.84 at the 5% level. The estimated likelihood ratio (chi-square) statistic is 18.22 for the rural sample and 6.993 for the urban sample. Thus the LR statistic is significant at 1% level for both the rural and urban sample. In other words, the LR test suggests that women’s literacy exerts a significantly different impact from that of her spouse on the current contraceptive use and encourages the use of contraception in both rural and urban areas.
3.4. Predicted Probability of Current Use of Contraception Bivariate probit selection estimates as shown in Table 5 show whether a variable has a statistically significant impact on the use of contraceptives but do not give any indication of their quantitative impact. We therefore computed the predicted probabilities of using contraceptives among non-sterilised women under various scenarios using the estimates from our model. We know, for example, that literacy has an important statistical impact on sterilisation and use of contraception. We would like to ask what would be the practical impact of raising education levels on use of contraception. Table 6 displays the predicted effects of changes in various significant factors explaining the current use of contraception among non-sterilised women. Line 1 shows the base for our comparisons. We have computed the predicted probability of using contraceptives for each non-sterilised woman. The figures show the average probability for all women in each of the rural and urban samples. The predictions for the ‘estimated values’ were computed by first predicting the probability of using contraceptives for each woman using the estimates from Table 5 and the values of the regressors for each woman. The statistics quoted on the first row (1) are the means of the probabilities across all non-sterilised women in each area. 21% of rural non-sterilised women use some contraception while the proportion is 26% among urban non-sterilised women. Thus the likelihood of currently using some reversible (traditional or modern) methods of contraception is generally higher among urban women. Current use of contraception also depends on the age of the woman. Compared to all women, the likelihood of using contraception is higher for both rural and urban women aged below 35 years (23% for rural and 29% for urban women). 14
On the basis of our estimates, we next consider what would happen if all illiterate women were to become literate. Row (2) of Table 6 shows the fraction of all women currently using some contraception in this regime. The changes in women’s literacy have their expected effects. Raising the educational standards of the women with the poorest achievements increases the likelihood of using modern method of contraception in both rural and urban areas. However, the effect is slightly more pronounced among rural women. For example, removing illiteracy alone would increase the fraction of all rural women of currently using any reversible method of contraception by about 0.04 points from 0.21 to 0.25 while it rises by about 0.02 points (from 0.25 to 0.27) for urban women. Similar effect is found for women aged below 34 years. Among various household assets, ownership of land among rural women and that of television among urban women are significant in the determination of current use of contraception in the bivariate probit selection model. Hence, we now consider the predicted probabilities of current use of contraception if (a) every rural woman without land owns some land and (b) every urban woman without television owns a television. These results are summarised in rows (3) and (4) for all women and rows (8) and (9) for women below age 35. These estimates indicate that the ownership of land among rural women has only a marginal impact, raising the probability of using some direct/indirect method of contraception by only 1 percentage point relative to their estimated value in row (1). Similarly if every urban woman had a television, the likelihood of using some contraception goes up by 1 percentage point. In other words, household assets seem to exert only a marginal impact for both rural and urban women in our sample estimation. Finally we assess the quantitative impact of belonging to a lower caste household on the current use of contraception. In doing so, we consider what would happen to the predicted probability of current use of contraception among rural women (since the caste variable is insignificant for urban women), if other things remaining unchanged, no sample woman belonged to the scheduled tribe category (since only scheduled tribe is significant in the determination of currently using some contraception). These estimates are summarised in row (5) and row (10) of Table 6 for all rural women and for rural women aged below 35 years respectively. The impact seems to be quite significant: relative to row (1), the probability of using contraception increases by 14 percentage point for all rural women and relative to row (6) by about 12 percentage point for rural women aged below 35 years. Let us now compare the effect of women’s literacy with that of household caste and assets on the current use of reversible (traditional and modern) methods of contraception among rural and urban women. The results significantly vary between rural and urban women in our sample. While raising 15
literacy seems to have a pronounced effect for both rural and urban women, impact of household assets seems to be negligible for both the samples. In contrast, the impact of caste is significant only among the rural women and the effect in this case is more pronounced than making all illiterate rural women literate. Thus in relation to caste, the effect of raising women’s literacy on contraception use is rather limited for the rural sample. Taken together, these results indicate the importance of reducing social conservatism and upholding modern social norms in encouraging the use of various reversible forms of contraception, especially among rural women in India.
5.
CONCLUDING COMMENTS
Using the recent NFHS 1992-93 household-level data this paper examines the causes of low contraceptive use in the eastern Indian state of West Bengal. In particular, our analysis focuses on the role of caste, household assets and women’s literacy on contraceptive choice and use among nonsterilised rural and urban women aged 13-49. Using bivariate probit selection model we jointly determine the probability of not being sterilised and that of currently using some reversible method of contraception among non-sterilised women. Rural and urban women from scheduled caste households with less land and more living children are more likely to be sterilised. In addition, experience of child death and reproductive health problems among rural women turn out to be significant in determining whether they are sterilised. There is also some evidence of son preference among sterilised rural women. Correcting for the selectivity bias, we next examine the probability of currently using some reversible mode of contraception among non-sterilised women in our sample. For both rural and urban samples, the most important variables seem to be women’s literacy and household assets (used as instruments of household income). Literate women are more likely to use some form of contraception. We also consider a likelihood ratio statistic to test if each parent's literacy has similar impact on these household decisions relating to contraceptive use in both rural and urban areas. The test suggests that literacy of each spouse have significantly different effects on contraceptive use. Some household assets are also found to be important; for example, ownership of land in rural areas and that of television in urban areas significantly enhance the likelihood of using some form of contraception. Women’s fecundity is also important: rural and urban women aged 35 or more are significantly less likely to use any contraception. However, caste seems to be affecting the rural and urban women differently: scheduled tribe households from rural areas have a significantly lower likelihood of using contraceptives while caste is insignificant for the urban sample. 16
Finally, we use our selectivity corrected likelihood estimates to compute the predicted effects of various changes for non-sterilised and non-pregnant rural and urban women. For example, removing illiteracy would increase the fraction of both the rural and urban samples using some contraception and the effect is slightly bigger for the rural sample. The effect of household assets on the likelihood of contraception use is marginal for both the rural and the urban samples. However, the effect of caste is more significant among rural women in relation to the effect of raising women’s literacy in our sample. Despite persistent governmental efforts over the past decades, current use of some reversible form of contraception continues to be low in West Bengal in the 1990s – a state with moderate level of literacy among the Indian state. The spread of contraceptives for limiting as well as spacing births among women of all background is necessary for the success of family planning objectives in India. While there is limited effect of household assets and women’s education in our sample, results emphasize the importance of removing inter-caste and rural-urban disparities in the country. The focus of the Indian family welfare program on sterilisation as indicated by the NFHS 1992-93 has been considered to be unsatisfactory and subsequently a new draft plan for the family welfare programme was undertaken in 1994. The new draft focuses on a number of key issues including moving away from numerical, method-specific contraceptive targets and incentives and expanding the use of male and reversible contraceptive methods (for spacing births) and broadening the choice of contraceptives. Effect of the new programme on the contraceptive choice and use would be reflected in the second round of the NFHS collected in 1998-99. We hope to analyse this new dataset to study the changing pattern of contraceptive use in India in the near future.
17
REFERENCES Ainsworth, M., K. Beegle, and A. Nyamete. 1996. ‘The Impact of Women’s Schooling on Fertility and Contraceptive Use: A Study of Fourteen Sub-Saharan African Countries’, World Bank Economic Review, pp. 85-122. Alderman, H., P-A. Chiappori, L. Haddad, J. Hoddinott and R. Kanbur. 1995. ‘Unitary Versus Collective Models of Household: Is It Time to Shift the Burden of Proof’, The World Bank Research Observer. Aly, H.S. and M.P. Shields. 1991. ‘Son Preference and Contraception in Egypt’, Economic Development and Cultural Change, pp. 353-70. Anderson, K.H. 1983. ‘The Determinants of Fertility, Schooling and Child Survival in Guatemala’, International Economic Review 24, pp. 567-589. Basu, A.M. and K. Basu. 1991. ‘Women’s Economic Roles and Child Survival : The Case of India’ Health Transition Review, 1, pp. 83-103. Becker, G. 1965. ‘A Theory of Allocation of Time’, Economic Journal, 75, pp. 493-517. ………… 1981. ‘A Treatise on Family’, Chicago University Press. Benefo, K and T.P. Schultz. 1994. ‘Determinants of Fertility and Child Mortality in Cote d’Ivoire and Ghana’, LSMS working paper 103, Washington DC : World Bank. ………………….. 1996. Fertility and Child Mortality in Cote d’Ivoire and Ghana’, The World Bank Economic Review, 10(1), pp. 123-58. Ben-Porath and F. Welch. 1976. ‘Do Sex Preferences Really Matter? Quarterly Journal of Economics, 90, pp. 285-307. Bhattacharya, P.C. 1998. ‘Determinants of Fertility in Three North Indian States: Caste, Female Autonomy and Economic Development’, Canadian Journal of Development Studies, 19(1), pp. 37-53. Caldwell, J.C. 1976. ‘Toward A Restatement of Demographic Transition Theory’, Population and Development Review, 2, pp. 321-67. Chowdhury, A.K.M.A., A.R. Khan and L.C. Chen. 1976. ‘The Effect of Child Mortality Experience on Subsequent Fertility in Pakistan and Bangladesh’, Population Studies 30, pp. 249-61. Cleland, J et al. 1994. The Determinants of Reproductive Change in Bangladesh. World Bank Regional and Sectoral Studies, Washington D.C. Cleland, J. and J. Hobcraft. 1985. ‘Reproductive Change in Developing Countries’, Oxford University Press, Oxford. Dasgupta, M. 1987. 'Selective discrimination against female children in rural Punjab', Population and Development Review, 13. DeGraff, D.S. 1991. Increasing Contraceptive Use in Bangladesh: The Role of Demand and Supply Factors’, Demography pp. 65-81. Dolton, P. and G. Makepeace. 1996. “Female Labour Force Participation and Choice of Occupation:
The Supply
of Teachers”, 1993, European Economic Review, vol.37, pp. 1393-1411 Drèze, J and M. Murthi. 2000. ‘Fertility, Education and Development: Further Evidence from India’, Development Economics Discussion Paper no. 20, STICERD, London School of Economics.
18
Drèze, J and A Sen. 1995. India: Economic Development and Social Opportunity, Oxford: Clarendon Press. Dyson, T. and M. Moore. 1983. ‘On kinship structure, female autonomy and demographic behaviour in India’, Population and Development Review, 9, I, pp. 35-60. Duraisamy, M. 1996. ‘Women’s Choice of Work and Fertility in Urban Tamil Nadu, India’ in T.P. Schultz (ed.) Research in Population Economics, Vol. 8 Jai Press. Easterlin, R.A. 1978. ‘The Economics and Sociology of Fertility: A Synthesis’, in C. Tilly(ed.) Historical Studies of Changing Fertility, Princeton University Press, Princeton. Fawcett, J.T. 1970. ‘Psychology and Population’, Occasional Paper, Population Council, New York. Haddad, L, J Hoddinott and H Alderman (eds) 1997. Intrahousehold Resource Allocation in Developing Countries: Models, methods and policies, Baltimore: John Hopkins University Press. Jeffrey, R and A.M. Basu. 1996. ‘Girls’ Schooling, Women’s Autonomy and Fertility Change in South Asia’ (ed.) by R. jeffery and A.M. Basu, New Delhi, India, Sage Publications. Kabeer,N. 1994. Reversed Realities. London: Verso. Kishor, S. 1993. ‘May God Give Sons to All: Gender and Child Mortality in India’, American Sociological Review, 58, pp. 247-65. Kynch, J. and A.K. Sen 1983. 'Indian women: well-being and survival', Cambridge Journal of Economics, Vol 7, 363-80. Larsen, U., W. Chung and M. Dasgupta. 1998. ‘Fertility and Son Preference in Korea’, Population Studies, 52, pp. 317-25. Manser, M and M. Brown. 1980. 'Marriage and Household Decision Making : A Bargaining Analysis', International Economic Review. McElroy, M and M.B. Horney. 1981. 'Nash-Bargained Household Decisions: Toward A Generalization of the Theory of Demand', International Economic Review. Messer, E. 1997. 'Intra-household allocation of food and health care: current findings and understandings - an introduction', Social Science and Medicine, 44:11, 1675-84. Michael, R. and R. Willis. 1975. ‘Contraception and Fertility: Household Production under Uncertainty’, in N.E. Terleykyj (ed.) Household Consumption and Production, NBER, New York, pp. 27-93. Montgomery, M.R. and J.B. Casterline. 1993. ‘The Diffusion of Fertility Control in Taiwan: Evidence from Pooled Cross-Section and Time Series Models’, Population Studies 47, pp. 457-81. Murthi M., Guio,-Anne-Catherine and Dreze,-Jean. 1995. ‘Mortality, Fertility, and Gender Bias in India: A District-Level Analysis’ Population-and-Development-Review; 21(4), pp. 745-82. Pitt, M.M. 1993. ‘Women’s Schooling, Selective Fertility and Infant Mortality in Sub-Saharan Africa, Processed, World Bank, Washington D.C. Phillips, J.F. et al. 1986. ‘The Demographic Impact of the Family Planning-Health Services Project in Matlab, Bangladesh’, Studies in Family Planning, 13, pp. 131-40. Rosenzweig, M. and R.E. Evenson. 1977. ‘Fertility, Schooling and Economic Contribution of Children in Rural India’, Econometrica 45, pp. 1065-1079. Rosenzweig, M. and T.P. Schultz. 1982. ‘Market Opportunities, Genetic Endowments and intrafamily resource distribution, American Economic Review, 74, pp. 215-235.
19
Rosenzweig, M. and T.P. Schultz. 1985. The Demand for and Supply of Births: Fertility and Its Life Cycle Consequences’, American Economic Review, 75, pp. 992-1015. Rosenzweig, M. and T.P. Schultz. 1987. Fertility and Investment in Human Capital’, Journal of Econometrics, 36 pp. 163-84. Rosenzweig, M. and D.A. Seiver. 1982. Education and Contraceptive Choice: A Conditional Demand Framework’, International Economic Review, pp. 171- 98. T.P. Schultz. 1969. ‘An Economic model of family planning and fertility, Journal of Political Economy, 99, pp. 582-606. Schultz, T P. 1973. A Preliminary Survey of Economic Analysis of Fertility’, American Economic Review, 63, pp. 71-78. Schultz, T P. 1993. ‘Investments in Women’s Human Capital and Development’, Editor, Symposium, Journal of Human Resources, 28, pp. 690-974. Schultz, T.P. 1990. Testing the Neoclassical Model of Family Labour Supply and Fertility, Journal of Human Resources, 25(4), pp. 599-634. Schultz, T P. 1994. ‘Human Capital, Family Planning and Their Effects on Population Growth’, American Economic Review, 83, pp. 255-60. Schultz, T P. 1997. ‘Demand for Children in Low Income Countries’, in Rosenzweig, M. and O. Stark (ed.) ‘Handbook of Population and Family Economics’, Elsevier, vol. 1A. Sen, A.K. 1997. ‘Population Policy: Authoritarianism versus Cooperation’, Journal of Population Economics. Thomas, D. 1990. ‘Intra-Household Resource Allocation: An Inferential Approach’, Journal of Human Resources, pp. 635-64. Strauss, J and Thomas. 1995. ‘Human resources: empirical models of household and family decisions’ in Handbook of development economics Vol3A, eds. J Behrman and T N Srinivasan, Amsterdam: North Holland. Visaria, P. 1993. ‘Demographic Aspects of Development: The Indian Experience’, The Indian Journal of Social Science, 6., pp. 219-42. W Van de Ven and B van Praag. 1981. ‘The Demand for Deductibles in Private Health Insurance’, Journal of Econometrics, 17, pp.229-252. Willis, , R. 1973. ‘A New Approach to the Economic Theory of Fertility Behaviour’, Journal of Political Economy, 81, pp. S14-S64. Wolfe, ,B.L. and J.R. Behrman. 1992. The Synthesis Economic Fertility Model’, Journal of Population Economics, 5, pp. 1-16.
20
TABLES Table 1. Comparison of West Bengal with Some Important Indian States Use of contraception, 92-93 States
Populat
Female
Female
Total
IMR per
Couple
Rate of
Current
Ever
n (in
literacy
labour
fertility
1000
protectio
women
use of
used
mn)
Age 7+
participn
rate
1996
n rate
sterlisn.
contrcn.
1991
1991
1991
(%) (1997)
Kerala
29
86.2
12.8
1.8
13
46.7
42
6
27
Punjab
20
50.4
2.8
3.1
52
76.9
32
17
32
Haryana
16
40.5
6.0
4.0
68
53.9
30
10
23
Maharas
78
52.3
26.5
3.0
48
51.0
40
6
16
AP
67
32.7
30.1
3.0
66
46.9
38
2
6
Tamil
56
51.3
25.1
2.2
54
51.7
38
6
16
WBengal
68
46.6
8.0
3.2
56
34.2
26
7
23
Source: Drèze and Sen(1995); Government of India web site: www.nic.in/mohfw/popindi.html
Table 2. Proportions (percentage) of Women Using Contraception By Method Method
Rural
Urban
None
44.8
39.5
Indirect (abstinence etc)
5.4
11.5
Direct (pills etc)
18.5
24.7
Sterilisation
31.3
24.3
Table 3A. Current Sources of Contraceptives Pills
IUD
Condoms Rural Urban
Sterilisation Rural Urban
Rural
Urban
Rural
Urban
Govt. hospitals health centres
20 (23.53)
2 (4.65)
31 (91.18)
10 (7.14)
5 (16.13)
2 (5.56)
938 (95.71)
161 (78.16)
Pharmacy Drug stores, Private hospital Other sources
61 (71.76)
41 (95.35)
3 (8.82)
4 (2.86)
18 (58.06)
24 (66.67)
31 (3.16)
41 (19.9)
4 (4.71)
-
-
-
8 (25.81)
10 (27.78)
11 (1.12)
4 (1.94)
85
43
34
14
31
36
980
206
Column total
21
Note: These figures omit those who are using traditional (abstinence, etc.) methods of contraception. The proportions of the column totals are shown in the parentheses.
22
Table 3B. Reasons for Not Using Any Modern Contraception Reasons
Rural
Urban
Desire for further children Menopause/hysterectamy Health hazards Family opposition Religious reasons Inconvenience of methods
42% 21% 10% 9% 6.5% 4%
23% 27% 15% 8% 3.6% 8%
TABLE 4. Means and Standard Deviations of Regression Variables
Variable
Mean
Rural Std.Dev.
Mean
Urban Std.Dev.
Uses contraception Not sterilised Age 20 to 24 25 to 29 30 to 34 35 to 39 40 to 44 45 to 49 Schooling Literate, no schooling Primary School Middle School High School Some university Women is employed Husband is literate Wealth Pucca Radio Cycle Agricultural Land Television Scheduled Caste Scheduled Tribe A child has died Still birth or similar problem Number of sons Number of girls Age of youngest child
0.24 0.69
0.43 0.46
0.36 0.76
0.48 0.43
0.20 0.20 0.15 0.14 0.10 0.08
0.40 0.40 0.36 0.35 0.31 0.27
0.12 0.21 0.17 0.18 0.14 0.11
0.33 0.41 0.38 0.39 0.35 0.32
0.17 0.11 0.08 0.03 0.08 0.25 0.26
0.38 0.32 0.27 0.17 0.09 0.44 0.44
0.15 0.13 0.15 0.15 0.13 0.18 0.19
0.36 0.34 0.35 0.35 0.33 0.39 0.39
0.11 0.41 0.53 0.62 0.08 0.11 0.06 0.30 0.15 1.40 1.34 0.33
0.31 0.49 0.50 0.49 0.27 0.31 0.24 0.46 0.35 1.27 1.32 0.26
0.51 0.59 0.54 0.15 0.50 0.04
0.50 0.49 0.50 0.36 0.50 0.20
0.21 0.11 1.26 1.17 0.35
0.41 0.32 1.21 1.16 0.25
Number
3131
848
23
TABLE 5. Bivariate Probit Selection Estimates of Non-Sterilisation and Current Contraceptive Use RURAL Not sterilised
Constant Age 20 to 24 25 to 29 30 to 34 35 to 39 40 to 44 45 to 49 40 to 49 Schooling Literate Primary School Middle School High School Some university High/Univ Husband is literate Wealth Pucca Radio Cycle Television Agricultural Land Scheduled Caste Scheduled Tribe Age of youngest child A child has died Still birth or similar problem Number of sons Number of girls
Correlation coefficient Log likelihood Sample size
URBAN Not sterilised Uses contraception Coef. T-ratio Coef. T-ratio 2.59
4.953
-1.142 -5.038
-0.708 -1.25 -1.42 -1.4
-1.329 -2.408 -2.720 -2.697
-0.127 -0.497 0.15 0.544 -0.33 -1.114 -0.63 -1.904
-0.836 -4.348
-0.99
-1.882
-1.49
-3.902
0.352 0.537
4.024 5.72
-0.21 -0.44 -0.22 0.28 0.33
-1.332 -2.381 1.166 1.217 1.316
0.505 0.76
2.481 3.586
0.888 0.023
8.811 0.341
-0.18
-1.326
1.44 -0.03
6.890 -0.196
-0.01 0.04 -0.23 0.04 0.22 -0.78
-0.086 0.037 -1.935 0.335 1.486 -3.324
Coef.
T-ratio
Uses contraception Coef. T-ratio
2.190
12.449
-0.825 -9.497
-1.187 -1.645 -1.981 -2.003 -1.904 -1.400
-6.514 -9.148 -10.858 -10.887 -10.064 -7.216
-0.060 -0.152 -0.245 -0.429
-0.617 -1.241 -1.529 -2.321
-0.214 -0.030 0.147 0.655 0.245
-3.073 -0.318 1.357 2.986 0.763
0.059
1.027
-0.014 -0.052 -0.047 0.012 0.162 -0.199 -0.134
0.149 -0.908 -0.852 0.108 2.991 -2.421 -1.247
0.170 0.342
3.112 4.510
0.19 0.01
0.148 0.081
-0.161 0.003
-7.583 0.153
-0.24 -0.15
-5.636 -3.277
-0.040 -0.383 0.027 0.454 0.159 1.377 0.141 2.362 -0.044 -0.462 -0.502 -3.523 0.422 1.567
0.571 2.915 -2926 3131
-0.073 -0.528 0.113 0.898 0.3 0.01 -0.41
2.101 0.104 -1.020
1.54
2.792
0.21 1.608 -735 848
24
Table 6. Predicted Probability of Current Use of Contraception among Non-sterilised Women (Selectivity Corrected)
All women
Rural
Urban
(1) Estimated value
0.21
0.26
(2) Everyone with no education is literate (3) Everyone with no tele owns a tele (4) Everyone with no land owns some land (5) No one is a ST
0.25 0.22 0.35
0.27 0.27 -
(6) Estimated value
0.23
0.29
(7) Everyone with no education is literate (8) Everyone with no tele owns a tele (9) Everyone with no land owns some land (10) No one is a ST
0.28 0.25 0.35
0.31 0.27 -
Women below age 35
Note: We have not calculated the predicted effects if the variable is not significant in a given sample.
25
Appendix I: Definition of Regression Variables
Dependent variables: Uses contraception: Non-sterilised:
1 if the couple is currently using any contraceptives (excluding sterilisation) 1 if the couple is non-sterilised and 0 if sterilised .
Explanatory Variables: Age 20 to 24: 1 if aged between 20 and 24 years 25 to 29: 1 if aged between 25 and 29 years 30 to 34: 1 if aged between 30 and 34 years 35 to 39: 1 if aged between 35 and 39years 40 to 44: 1 if aged between 40 and 44 years 45 to 49: 1 if aged between 45 and 49 years 40 to 49: 1 if aged between 40 and 49 years Schooling Literate: 1 if literate Primary school: 1 if completed the primary school Middle school: 1 if completed the middle school High School: 1 if completed the high school Some university: 1 if has some university education High/Univ MIDSCHW+HIGHSCHW+UNIVW Wife literate: LITW+PRIMSCHW+MIDSCHW+UNIVW Husband literate: 1 if the husband is literate Number of sons: number of living sons Number of girls: number of living daughter Wealth/Income Pucca: 1 if the house is pucca Radio: 1 if the household owns radio Cycle: 1 if the household owns bicycle Agricultural Land: 1 if the household owns any agricultural land Television: 1 if the household owns television. Scheduled caste 1 if belongs to scheduled caste households Scheduled tribe: 1 if belongs to scheduled tribe households Age of youngest chile Current age of the youngest child A child has died 1 if a child has died Still birth or similar problem 1 if a child has been still born
26
Appendix II : Additional Table TABLE A. Bivariate Probit Selection Estimates
Variable Constant Age 20 to 24 25 to 29 30 to 34 35 to 39 40 to 44 45 to 49 40 to 49 Wife is literate Husband is literate Wealth Pucca Radio Cycle Television Agricultural Land Scheduled caste Scheduled tribe Age of the youngest child Child death Still birth etc. Number of sons Number of girls Correlation coef. Log likelihood Sample size
RURAL Not sterilised Coef. T-ratio
Uses contraception Coef. T-ratio
URBAN Not sterilised Coef. T-ratio
Uses contraception Coef. T-ratio
2.187
12.455
-0.817
-9.409
2.530
4.732
-1.158
-5.221
-1.162 -1.605 -1.947 -1.960 -1.883 -1.382
-6.403 -8.960 -10.713 -10.698 -9.994 -7.162
-0.046 -0.143 -0.265 -0.421
-0.495 -1.271 -1.795 -2.452
-0.635 -1.143 -1.284 -1.278
-1.167 -2.179 -2.431 -2.425
-0.042 0.288 -0.127 -0.347
-0.177 1.102 -0.462 -1.168
-0.079
-1.356
-0.784 0.502
-4.270 7.078
-0.892 -0.214
-1.667 -1.615
-1.119 0.842
-3.188 4.966
0.016
0.282
-0.036
-0.576
-0.223
-1.741
-0.202
-1.268
0.047 -0.027 -0.042 0.078 0.167 -0.195 -0.142
0.523 -0.480 -0.776 0.729 3.114 -2.394 -1.330
0.012 0.055
0.116 0.941
0.617 1.538
2.253 2.758 -0.487 -3.551
0.845 0.320 -2.183 1.018 1.343 -3.153
0.082 0.184
0.250 0.159 -0.045 -0.493
0.108 0.036 -0.253 0.129 0.192 -0.752
0.490 -0.026 -0.502
3.680 -0.172 -1.289
0.241
0.960
0.855
1.725
0.154 0.327 -0.168 -0.005
2.882 4.374 -7.995 -0.277
-0.006 0.082 -0.258 -0.163
-0.049 0.464 -6.681 -3.705
0.711 -2947.09 3131
4.028
0.381 -757.057 848
2.182
27
