![Sample our Social Sciences journals, sign in here to start your FREE access for 14 days]({"type":"image" , "src" : "/sda/110822/TOC-Subject-Sample-2021-Social-Sciences.jpg"})
Abstract
This paper examines the determinants of women's knowledge regarding HIV/AIDS using data from a nationally representative survey in India. Although around 45 per cent of sample women had heard about the disease, their knowledge regarding its modes of transmission and prevention is found to be limited. To explore the possibility that there may be a different process that determines awareness as opposed to quality of knowledge regarding HIV/AIDS, a negative binomial hurdle model and a two-stage ordered probit model are estimated. The results show that the effect of several covariates, such as education and mass media, on awareness is different from their effect on quality of knowledge.
Notes
1. Knowledge levels have been found to be somewhat higher in countries with a longer history of infection and higher infection rates, but even here significant knowledge gaps persist. For instance, recent surveys in seven sub-Saharan African countries that have been hit hard by AIDS (Burkina Faso, the Central African Republic, Côte d'Ivoire, Senegal, Tanzania, Uganda, and Zimbabwe) found that only 40 to 70 per cent of people with a recent non-regular sex partner could name condoms as a means of preventing HIV (World Bank, 1999).
2. Report at the First International Workshop on AIDS for Developing Country Media, Ottawa, Canada, 2–3 June, 1989 cited in Netter (1992).
3. HIV is transmitted in human body fluids by three routes: (1) sexual intercourse; (2) direct injection with HIV contaminated drugs, needles, syringes and blood; (3) from HIV infected mother to fetus in utero, during child-birth, or during breast feeding (Stine, 2003).
4. Balk and Lahiri (1997) used data from the NFHS-I survey of 13 out of 25 states in India in 1992–93 to examine determinants of women's knowledge regarding AIDS. Their study was one of the first to use data from the general Indian population. Gwatkin and Devshwar-Bahl (2001) used data from the Demographic and Household surveys (DHS) in 23 countries to examine the socioeconomic and gender differentials in knowledge about HIV/AIDS. Pillai et al. (2003) used data from the Zambia DHS to examine the role of family planning services in AIDS prevention, but they take knowledge levels as exogenously given. There are several other studies that have examined knowledge levels among sub-populations such as Gupta and Mitra (1999) for migrants in Delhi slums and Bryan et al. (2001) for Indian truck drivers.
5. Balk and Lahiri (1997) distinguish between awareness and knowledge, but they estimate these equations independently. Thus, their estimates are subject to sample selection bias (a detailed discussion follows in Section IV).
6. Details on the survey are available at http://www.nfhsindia.org/research.html
7. For simplicity, the survey used the term ‘AIDS’ to refer to both HIV and AIDS.
8. The means reported in the paper are unweighted means.
9. In an earlier study, using data from the first round of NFHS conducted in 1992–93 in 13 out of 26 states in India, Balk and Lahiri (1997) found that only 17 per cent of women had heard about HIV/AIDS. According to the most recent behavioural surveillance survey carried out in 2001, around 70 per cent of women in the 15–49 age group had heard about AIDS (NACO, 2001).
10. In general, it is true that on such open-ended questions, there is likely to be some interviewer bias as the number of responses obtained depend to a certain extent on how insistent the interviewer was on probing. In the specific case of the NFHS, we believe this bias to be small as the field investigators underwent extensive training and supervision. Moreover, even if the bias is present, it is likely to be randomly distributed.
11. Some other ways of prevention, such as prevention of mother-to-child transmission (by not getting pregnant if you are HIV positive or taking a prophylactic drug) and reducing breastfeeding if you are an HIV mother, were not included in the precoded list. Very few women in 1998–99 were likely to know about theses two methods. Only 1 per cent of women reported a mode other than the pre-coded methods. The data files do not contain information on what exactly they reported, so we do not know whether it was correct or incorrect.
12. Kenkel (1991) also discusses the psychological phenomenon of cognitive dissonance, which occurs when the individual attempts to reconcile her self-image as a smart person with possibly contradictory actions. Thus, for instance a smoker will not admit that he believes that there are serious adverse effects of smoking.
13. We are grateful to an anonymous referee for pointing this out to us.
14. Use of condoms can be classified under the sexual mode, but we take it as a separate category because of the following reasons. First, use of condoms is different from other responses under the sexual mode which deal more with modes of transmission while condom use is a method of prevention. Second, condom use has been promoted widely under the various family planning programmes, besides being a major focus of AIDS information campaigns.
15. First, the probit model of awareness is estimated and the inverse mills ratio is predicted. Then the ordered probit model is estimated using the inverse mills ratio from the first stage as one of the regressors. The corrected standard errors for this two-stage model are obtained by bootstrapping.
16. To test for the appropriateness of the negative binomial hurdle model versus other count data models, we carried out a number of specification tests. First, we used a test proposed by Cameron and Trivedi (1990) to test for over-dispersion in the data. The hypothesis of no over-dispersion was rejected using this test. Since the Poisson model is obtained as a parametric restriction of the negative binomial model, we used a Lagrange Multiplier test to for the Poisson specification against the negative binomial specification. The hypothesis of Poisson model was rejected using this test.
17. For the negative binomial hurdle specification, we used a test proposed by Vuong (1986) to test this hypothesis. The Vuong test statistic for both rural and urban specifications was found to be significant at 1 per cent, thus rejecting the standard negative binomial (single stage) model in favour of the hurdle (two-stage) model. For the ordered probit specification, the Heckman selection test was used.
18. Throughout the results section, significance only refers to the coefficient of interest found to be statistically different from zero.
19. Attaining universal primary education has been set as one of the UN Millennium Development Goals.
20. Higher here means education beyond middle school level.
21. However, in the ordered probit model, all levels of husband's education have a statistically significant effect on urban woman's awareness.
22. In the ordered probit model also, the direct effect of radio listening in quality of knowledge of both rural and urban women is not significant. However, some interaction terms of radio exposure with education are significant, but there is no pattern.
23. Although few of the coefficients associated with the media/education interaction variables were statistically significant independently, we found that, as a whole, they were jointly significant at the 99 per cent level.
24. The left out dummy here represents the upper castes.
25. Unfortunately, we do not have sufficient details in our data on the occupations of women and their husbands to examine this conjecture more rigorously.
26. Clusters are geographical location variables used in the sampling process. There are 266 rural clusters and 333 urban clusters in the NFHS data. The average number of women in a rural cluster is around 218 while the average number of women in an urban cluster is 80.
27. This was true for all the specifications tried except for the urban ordered probit specification where the effect was negative.
28. We also tried using lagged HIV infection rates instead of AIDS cases. However, as a referee pointed out, HIV infection rates are likely to be an inferior measure of people's personal experience with HIV/AIDS because it is asymptomatic. It is also likely that HIV infection rate itself is endogenous in our knowledge specifications. Since it takes several years for a full-blown AIDS case to be detected after an initial infection, using number of AIDS cases as an explanatory variable is likely to be less prone to the endogeneity bias.
39. The cubic terms were found to be very small and not statistically significant, so these were dropped in the specifications reported here.
30. A community questionnaire, with these questions on community services was canvassed only in rural areas.
31. The categorisation of the different states and union territories into regional dummies was done in the following way. First, preliminary regressions were run with all the states and union territories included as separate dummies. Then states with geographical proximity and similar coefficients were grouped together. Mizoram, Nagaland and Manipur had very different coefficients than the other eastern states, so these were categorised separately. The categorisation was done as follows North (includes Jammu, Punjab, Haryana and Himachal Pradesh), East (includes West Bengal, Orissa, Sikkim, Assam, Arunachal Pradesh, Tripura and Meghalaya), West (includes Maharashtra, Gujarat and Goa), South (includes Andhra Pradesh, Karnataka, Tamilnadu and Kerala), Central (includes Bihar, Madhya Pradesh, Rajasthan and Uttar Pradesh), Northeastern border (includes Manipur, Mizoram and Nagaland) and New Delhi includes the metropolitan area of the capital city.
32. Although New Delhi is not classified under the high prevalence category, being the national capital it has been at the centre of a lot of political activism and media attention related to AIDS.