CONTEXT: Qualitative studies assessing the impact of the household delivery of family planning services on women's social status have yielded contradictory findings. Given the resumption of these services in Bangladesh in 2003, it is important to reevaluate the social impact of the doorstep program using quantitative techniques.
METHODS: Longitudinal and cross-sectional data from 3,783 women using doorstep services in two rural districts of Bangladesh are used in ordinary least-squares and logistic regression analyses to assess the effect of doorstep services on changes in women's status between 1988 and 1993.
RESULTS: In analyses controlling for background characteristics, women's status in 1988, previous service use and visit selection bias, household outreach is associated with increases in women's status between 1988 and 1993. However, this effect is largely attributable to the impact of doorstep services on women's ability to regulate their fertility rather than to the home visits themselves.
CONCLUSION: The decision of the Bangladesh Ministry of Health and Family Welfare to resume household family planning services should not be detrimental to women's status, and may be associated with gender benefits to female clients served by the program.
International Family Planning Perspectives, 2003, 29(3):128-145
From 1978 to 1997, the Bangladesh government hired and trained married women to provide family planning counseling and services to couples in rural households. At the peak of the program, a total of 28,000 of these "family welfare assistants" were working in hamlets throughout the country.*1 By the early 1990s, evidence showed that doorstep service delivery had helped to increase family planning awareness, as well as the rate of method uptake and the continuity of method use, among rural couples.2 Furthermore, by increasing access to reversible methods, the program helped to reduce the proportion of contraceptive users who adopted female sterilization3—a method that the government family planning program had heavily promoted. In areas where family welfare assistants had been trained to administer the injectable, the reduction in the reliance on sterilization was particularly pronounced and was also accompanied by a reduction in reliance on the pill.4 Even so, overall pill use increased steadily following the initiation of doorstep delivery, with family welfare assistants providing nearly 85% of all pills dispensed in rural Bangladesh between 1978 and 1997.5
Despite these successes, various studies identified shortcomings in the program. First, doorstep services did not include child or reproductive health care, in large part because a legacy of organizational problems had separated health care and family planning in the government ministry responsible for the program.6 Second, health and social service training, and supervisory and management problems affected the quality of services in the program. Client surveys suggest that services unduly focused on particular contraceptive methods and often omitted health service needs.7 And recent qualitative research suggests that operational problems have persisted, even after health sector reforms.8 Third, some family welfare assistants focused on better-educated women; targeting uneducated or poor women would have been more effective.9 Finally, the policy of reaching couples was implemented as outreach that focused on female clients rather than on couples. The exclusion of men constrained the range of services offered and reduced the efficiency of outreach efforts.10
Whether doorstep service delivery affected women's autonomy and status is also unclear. Finding out the answer is important because Bangladesh is a patriarchal society, and Bangladeshi women have a lower status than do women in other Asian countries.11 The low status of women in rural Bangladesh is detrimental to maternal and child health12—a finding that is consistent with research conducted elsewhere in South Asia, showing that women are more likely than men to contract infectious diseases.13
The incorporation of gender equity into sexual and reproductive programs and the consideration of gender effects during program evaluation were emphasized in the International Conference on Population and Development in 1994.14 Consequently, evaluation of the doorstep program shifted from the assessment of its effects on family planning to the measurement of its impact on women's lives. However, studies on the program's social effects have so far reached contradictory conclusions.
A qualitative study conducted by Simmons and colleagues15 in the Matlab project area concluded that doorstep delivery of services enhanced women's status. For example, family welfare assistants benefited directly from receiving cash wages, and indirectly from gaining mobility, prestige and authority from their work. In turn, the large-scale deployment of female family planning workers changed people's perceptions of women's roles: Every hamlet in Bangladesh had a family welfare assistant acting as household visitor, adviser and confidante, clearly showing that women were employable, mobile, socially gregarious and autonomous, without depending on male partners or the extended family. Young female clients were particularly influenced by encounters with family welfare assistants. The program was thus characterized as having a "beyond-supply" social effect:16 In addition to helping clients achieve control over their fertility—and therefore their lives17—doorstep delivery provided benefits extending beyond those arising from the receipt of family planning services.
In contrast, qualitative work by Schuler and colleagues18 found that recurring household visits might have inadvertently reinforced patriarchal customs of female dependency, seclusion and purdah. Although doorstep services were not shown to harm women's status, results suggested that they prevented improvements in women's status that might have occurred if female clients had been required to travel to a clinic. Obtaining services away from home could have exposed female clients to influences fostering independence, autonomy and mobility.
Other studies concluded that door-to-door outreach was no longer justifiable on demographic grounds. For example, some observers argued in 2000 that the culture of practicing family planning had become so ubiquitous in Bangladesh by the 1990s that doorstep services had become unnecessary.19 Moreover, some family welfare assistants targeted higher-status women, hence departing considerably from the intended client coverage.20 These findings suggested that providing family planning services at a static service delivery site would not affect contraceptive prevalence.
Prompted by these findings and by concerns about the cost of sustaining a large workforce of family welfare assistants, some donors cut support for the program and others called for the program to be discontinued. In 1997, the national provision of doorstep services ended, and each subdistrict health team received instructions to develop static community "satellite clinics" that provided basic health care and family planning services.21 The Ministry of Health and Family Welfare continued to employ 23,500 family welfare assistants in these satellite clinics.
Despite the changes in central policy, some program managers continued to encourage staff to provide doorstep delivery, and some local officials and senior administrators continued to advocate such services on the grounds that they successfully promoted family planning. In March 2003, the Ministry of Health and Family Welfare issued a policy reinstating doorstep delivery, countering established loan and assistance agreements between the government of Bangladesh and the World Bank and other donors.22
Given the decision to reinstitute doorstep service delivery and the lack of consistent evidence from qualitative studies on the program's social impact on women, we designed a quantitative study to assess the social effects of the scheme when it was fully operational. In this study, we used multiple linear and logistic regression techniques to examine the association between women's status and the number of visits that clients received from family welfare assistants.
The data for this study were drawn from the records of the Maternal and Child Health-Family Planning (MCH-FP) Extension Project.† The International Centre for Diarrhoeal Disease Research, Bangladesh, launched the project in 1982 to test whether successful strategies of the Matlab Family Planning and Health Services Project could be implemented by the Ministry of Health and Family Welfare in Sirajgonj and Abhoynagar—rural districts in central and western Bangladesh, respectively.23
The MCH-FP Extension Project monitored contraceptive and reproductive behavior and contact with government health and family planning workers among a random stratified cluster sample of women in regions of Sirajgonj and Abhoynagar served by family welfare assistants. The original sample consisted of 5,600 rural women of reproductive age who were surveyed by trained interviewers every 90 days between 1982 and 1993. The interviewers collected information on women's background characteristics in 1982 (including husbands' educational level) and compiled longitudinal panel data on contraceptive use, the number of visits received from female welfare assistants in the past 90 days, and demographic events (such as births, deaths and migrations) since the last interview.24 The project included two cross-sectional surveys that investigated women's autonomy and status within their households: the Women's Status Survey (WSS), conducted in 1988, and the In-Depth Survey (IDS), conducted in 1993.
Of the original sample, 602 women were displaced from their homes by severe flooding in 1988 and a further 1,215 left the study areas or were lost to the study during the 11 years of the longitudinal survey (e.g., because of migration or death) or did not participate in the WSS and IDS for these reasons or because of temporary absence. Hence, we analyzed data from the remaining 3,783 women.
To analyze the relationship between exposure to family welfare assistants and women's status, we followed the methodology proposed by Balk,25 using longitudinal data on doorstep visits from the recurring interviews and data on women's status from the two cross-sectional surveys. We hypothesized that female clients who received more visits from the outreach workers were more likely than those receiving fewer visits to experience an increase in their status.
We used WSS and IDS data to determine clients' social status in 1988 and 1993, respectively. We assigned women a score of one (for high status) or zero (for low status) on each of 42 indicators (23 from the WSS and 19 from the IDS) and calculated the proportion of women with high status at the time of each survey. However, because the survey instruments used in the WSS and IDS differed, only four indicators were common to both. Therefore, we constructed two multivariate regression models: In the first, the dependent variable was the overall status score in 1993; in the second, it was the overall change in status from 1988 to 1993 for the four indicators common to the two surveys. Statistical analyses were performed with STATA, version 6.
A multiple regression approach allowed us to control for other determinants of women's status such as socioeconomic and demographic characteristics and previous contraceptive use.26 We hypothesized that the background characteristics shown in Table 1 influenced women's status, contraceptive use and the interaction between the client and the service provider. We also hypothesized that bidirectional associations existed between contraceptive use and client-provider interaction, between women's status and client-provider interaction, and between women's status and contraceptive use.
Our first analysis used ordinary least-squares regression to examine the association between the number of visits a woman reported having received between 1988 and 1993 and her overall status score (range, 0-19) in 1993, which we calculated by summing her scores for the IDS indicators. Individual status indicators in the WSS were treated as independent variables to control for women's status in 1988. We also controlled for clients' background characteristics in 1982, the reported number of visits received between 1982 and 1988 (and thus the bias created by outreach workers' preference for visiting higher-status clients), and the number of times women were interviewed.‡ Regression coefficients were calculated before and after adjustment for clients' contraceptive use to examine if fertility regulation also influenced women's status.
Our second analysis used multinomial logistic regression to examine the association between the number of visits received and overall change in status from 1988 to 1993 for the four indicators that were common to the WSS and the IDS. For each indicator, women were assigned scores of +1, zero or -1 for an increase, no change or a decrease in status, respectively. We calculated the overall status change for each client by adding her four scores (range, +4 to -4). An overall positive score represented a higher status in 1993 and an overall negative score a lower status. Women with an overall score of zero were regarded as having the same status in 1988 and 1993. We examined the effects of the number of visits on the likelihood that a woman experienced an increase rather than no change in status and on the likelihood that she experienced a decrease rather than no change in status. We again controlled for background characteristics, women's status for the other indicators in 1988, visits received before 1988 and number of interviews, as well as contraceptive use before and during the study period.§ Model parameters were computed using the generalized estimating equation method of Liang and Zeger.27
As Table 1 shows, the study cohort and the original sample had similar background characteristics, such as mean age (29 and 30, respectively), mean years of schooling (1.5 and 1.4 years) and husband's mean years of schooling of the two groups (3.0 each). The household economic status, as reflected by the floor area of the home and the proportion of women owning a radio, was also comparable, as was the proportion desiring no more children. Selection bias due to sample loss was therefore minimal. Although sample loss affected the proportion of women working outside the home—37% in the final sample vs. 26% in the original sample—this difference is unlikely to have changed the direction of the associations revealed by the regression analyses. On average, women in the study cohort received 17 visits from female welfare assistants between 1988 and 1993, or approximately one visit every three months. During this period, women reported in about one-third of the interviews that they practiced family planning.
In 1988, almost all of the 3,783 clients in the study cohort reported having permission to visit parents or relatives, approving of women who earn money at home or who own property, or being able to take trips away from home (97-99%, Table 2). Only small proportions had permission to earn money away from home (13%), decided whether to keep a child in school (17%) or whom or when their daughter should marry (15%), or had a say in how their husband's earnings were spent (7%). In 1993, most respondents reported having permission to leave the village to attend a health center or to see a movie (91-92%), and said they usually decided family size and timing of births (90-91%). In the six months before the survey, however, only 5% had actually left the village to see a movie, and 28% had done so to attend a health center. On the other hand, 94% reported having ever taken a sick child to the hospital by themselves, although only 35% said they had permission to do so. The mean total status score in 1993 was 11.0 (standard deviation, 3.2).
For three of the four indicators common to the two surveys, a larger proportion of women in 1993 than of those in 1988 had a high status score (Table 2). One-fifth of respondents had a higher score in 1993 than in 1998 for the question on whether they were allowed to take a child to a hospital outside the village—that is, they reported being allowed to do so in 1993 but not in 1988 (Table 3). Similarly, one-half reported in 1993 but not in 1988 that they were allowed to greet nonfamilial male visitors or go to the movies outside the village, and two-fifths that they could decide to see a doctor. Small proportions (0.3-10%) of women had lower status scores in 1993 than in 1988 on these four indicators, and 48-72% had no change in scores.
As we had hypothesized, contraceptive use—both before and during the study period—was positively associated with women's status in 1993 (Table 4, page 142). In the analysis controlling for contraceptive use, the educational level of clients' husbands and the size of the home were negatively associated with women's status, whereas clients' education, working away from home and owning a tube well showed positive associations. Women with some education and a job outside the home are likely to be exposed to external, modern influences, and investment in a tube well for a safe water supply may indicate modern views of health and child welfare.
Eight of the 23 status indicators in the 1988 WSS were positively associated with social status, suggesting that autonomy, authority, attitude toward independence, and mobility are valid indicators of women's status. Although we expected women of higher status in 1988 to be more likely than those of lower status to receive visits from family welfare assistants, visit selection bias was corrected for in the model, so the association between prior status and final status was independent of this bias.
All significant associations between women's status and 1988 status indicators and background factors (except husband's education) remained significant in the analysis that did not control for contraceptive use (Table 4). In addition, when contraceptive use was not controlled for, being able to travel to obtain a family planning method was linked with having a higher status, as was increasing exposure to family welfare assistants. This positive incremental effect of home visits on women's status could not be explained by the selectivity of outreach, thus supporting the hypothesis that household outreach fosters improved status for women. However, because this association was nonsignificant when contraceptive use was controlled for, the influence of outreach on women's status is attributable mainly to the effect of fertility regulation rather than to the effect of the social interaction during the visit. Still, we cannot rule out the indirect role of the contact between a client and a family planning worker on women's status in helping clients to adopt effective fertility regulation.
The results from the multinomial regression analyses shown in Table 5 indicate that only two background factors were associated with experiencing an increased or decreased status by the end of the study period: Women with husbands who had received a primary education were more likely than those with uneducated husbands to have a lower status in 1993 than in 1988. And clients from Abhoynagar were less likely than those in Sirajgonj to have an improved status.
The more visits that women received from family welfare assistants between 1988 and 1993, the more likely they were to experience an increase in status (relative to no change) during that period. In addition, the likelihood of increased status rose with the number of interviews during which women reported contraceptive use between 1982 and 1988. Neither factor was associated with decreased status (relative to no change).
To further examine the effect of contraception on women's status, we repeated the analysis with and without controls for contraceptive use (not shown). Comparison of these results showed that exposure to family welfare assistants had no effect on the likelihood of experiencing an increase or a decrease in status, other than through the indirect effects of visits on contraceptive use. Therefore, the impact of the doorstep delivery program on women's status derives mainly from its impact on fertility regulation; reduced unwanted fertility then likely contributes to an improvement in women's status.
Among the 1988 status indicators, being allowed to visit parents outside the village, having the authority to decide whether to buy medicine for a sick child and having a positive attitude toward women who decide on their or their child's medical treatment were related to having a reduced likelihood of experiencing an increase in status. In addition, having the authority to decide whether to buy medicine for a sick child was associated with having an elevated likelihood of experiencing a decrease in status (Table 5).
These findings are inconsistent with those reported by Schuler and colleagues, which suggested that household services may prevent improvements in women's status.28Their study collected qualitative data from 151 women and 139 spouses in six villages in northern and western Bangladesh served by a total of seven family welfare assistants. The sample was purposively selected to include relatively young, low-parity rural couples of low economic status who were practicing contraception. Selected households had at least one married woman aged 35 or younger with at least one living child; those with couples older than 35 or no couples younger than 35 were excluded, as were those with at least one acre of land. When we applied similar selection criteria to our analyses, results from the first regression model were comparable to those obtained by Schuler and colleagues (not shown), indicating that their study is biased by sample selection criteria.
DISCUSSION AND CONCLUSION
Our analyses are based on the assumption that measuring the social impact of doorstep services requires statistical modeling of women's status in a large, random sample of women. After adjustment for clients' background characteristics, baseline status, previous visits received and visitation bias, results from our regression analyses support the hypothesis that the household service delivery program in rural Bangladesh leads to gender benefits. In addition, there is no evidence of detrimental gender effects from participation in the program.
Our study shows that use of doorstep family planning services is positively associated with women's status at the end of the study period, and an increasing number of visits is related to an improvement in status. However, gender benefits come from the program's impact on fertility regulation rather than directly from the social interaction involved in a household visit. Although these findings do not demonstrate direct "beyond-supply" social benefits, they suggest that encounters with family welfare assistants indirectly enhance the status of women by fostering reproductive autonomy.
In 1997, the Bangladesh Ministry of Health and Family Welfare abandoned the doorstep approach in favor of a passive approach using community clinics. Recent evidence has established that the fertility transition in Bangladesh has stalled, primarily because of a deterioration in the effectiveness of family planning.29 The results of our analysis suggest that a deterioration in the effectiveness of fertility regulation may be accompanied by significant effects on women's autonomy, thereby offsetting gains in women's status that would arise if family planning needs were fully met.
In 2003, amid considerable controversy, the Ministry instituted a policy reinstating doorstep services. Given the results of this study, this policy shift can now be reviewed in light of the gender benefits that may arise from the resumption of doorstep service delivery.
*Nongovernmental organizations also deployed female outreach workers, but most nongovernmental support for household-based family planning services has now ended.
†This initiative has now been renamed the Operations Research Project.
‡In the first model, women's status at the end of the study period is repre- sented by the regression equation wit = ,
where wit = women's status score for individual i in 1993, sijo = status indicator of order j among J such indicators for individual i in 1988, xik = number of times individual i was interviewed during the total of K interviews possible between 1982 and 1988, cik = number of visits from a female welfare assistant reported by individual i during the total of K interviews possible between 1982 and 1988, xil = number of times individual i was interviewed during the total of L interviews possible between 1988 and 1993, cil = number of visits from a female welfare assistant reported by individual i during the total of L interviews possible between 1988 and 1993, and zimo = baseline characteristic of order m among M such characteristics for individual i in 1982.
§In the second model, the natural log of the ratio of the probability that a woman has an increased status (pYai) to the probability that she has an unchanged status (pYbi) in 1993 compared with 1998 is represented by the multinomial logistic regression equation ln(pYai/pYbi) =
, where the variables xik, cik, xil, cil and zimo are defined in model one. A similar equation is used for the analysis of the likelihood that women have a decreased rather than an unchanged status.
1. World Bank, Project completion report: Bangladesh, Second Population and Family Health Project (Credit 921-BD), Washington, DC: World Bank, 1989, No. 7792.
2. Koenig MA, Hossain MB and Whittaker M, The influence of quality of care upon contraceptive use in rural Bangladesh, Studies in Family Planning, 1997, 28(4):278-289.
3. Phillips JF et al., A case study of contraceptive introduction: domiciliary depot-medroxy progesterone acetate (DMPA) services in Bangladesh, in: Segal S, Tsui AO and Rogers S, eds., The Demographic and Programmatic Consequences of New Contraceptives, New York: Plenum Press, 1989, pp. 227-248; Phillips JF et al., Improving the climate of choice in rural Bangladesh: the effect of organizational change on contraceptive behavior in rural Bangladesh, in: Bulatao RA, Palmore JA and Ward SE, eds., Choosing a Contraceptive: Factors in Method Choice in Asia and the United States, Boulder, CO, USA: Westview Press, 1989, pp. 212-236; Arends-Kuenning M, How Do Family Planning Workers' Visits Affect Women's Contraceptive Behavior in Bangladesh? Research Division Working Papers, New York: Population Council, 1997, No. 99; and Khan MA and Rahman M, Determinants of contraceptive method choice in rural Bangladesh, Asia-Pacific Population Journal, 1997, 12(3):65-82.
4. Phillips JF, Hossain MB and Arends-Kuenning M, The long-term demographic role of community-based family planning in rural Bangladesh, Studies in Family Planning, 1996, 27(4):204-219; Cleland J et al., The Determinants of Reproductive Change in Bangladesh: Success in a Challenging Environment, Washington, DC: World Bank, 1994; and Larson A and Mitra SN, Family planning in Bangladesh: an unlikely success story, International Family Planning Perspectives, 1992, 18(4):123-129 & 144.
5. Streatfield K et al., Increasing Financial Sustainability of Family Planning Service Delivery in Bangladesh, Dhaka, Bangladesh: Population Council, 1997.
6. Phillips JF et al., The World Bank and Bangladesh's Population Program. Washington, DC: World Bank, 1991.
7. Bates LM et al., From Home to the Clinic: The Next Chapter in Bangladesh's Family Planning Success Story, Rural Sites, Washington, DC: The POLICY Project, Futures Group International, 2000.
8. Schuler SR, Bates LM and Islam MK, The persistence of a service delivery 'culture': findings from a qualitative study in Bangladesh, International Family Planning Perspectives, 2001, 27(4):194-200; and Bates LM et al., From home to clinic and from family planning to family health: client and community responses to health sector reforms in Bangladesh, International Family Planning Perspectives, 2003, 29(2):88-94.
9. Arends-Kuenning M, 1997, op. cit. (see reference 3); and Arends-Kuenning M, Hossain MB and Barkat-e-Khuda, The effects of family planning workers' contact on fertility p
10. Kincaid DL et al., Community networks and family planning promotion: impact of the Jiggasha approach in Trishal, Bangladesh, paper presented at the annual meeting of the American Public Health Association, San Francisco, CA, USA, Oct. 24-28, 1993.
11. Curtin LB, Status of women: a comparative analysis of twenty developing countries, in: Population Reference Bureau, Reports on the World Fertility Survey, Washington, DC: Population Reference Bureau, 1982, No. 5.
12. Hossain MB et al., The effect of constrained women's status on infant and child mortality in four rural areas of Bangladesh, paper presented at the annual meeting of the Population Association of America, Los Angeles, CA, USA, Mar. 23-25, 2000.
13. Lindenbaum S, Women and the left hand: social status and symbolism in East Pakistan, Mankind, 1968, 6(11):537-544; and Murthi M, Guio A-C and Drèze J, Mortality, fertility, and gender bias in India: a district-level analysis, Population and Development Review, 1995, 21(4):745-782.
14. United Nations (UN), Report of the International Conference on Population and Development, Cairo, 5-13 September, 1994, New York: UN, 1994.
15. Simmons R et al., Beyond supply: the importance of family planning workers in rural Bangladesh, Studies in Family Planning, 1988, 19(1):29-38.
17. Hoque MN and Murdock SH, Socioeconomic development, status of women, family planning, and fertility in Bangladesh: a district level analysis, Social Biology, 1997, 44(3-4):179-197; Keyfitz N, The family that does not reproduce itself, Population and Development Review, 1986, 12 (Suppl.):139-154; Kak L and Narasimhan S, The Impact of Family Planning Employment on Field Workers' Lives: A Strategy for Measuring Empowerment, Washington, DC: Center for Development and Population Activities, 1992; Davis K, Wives and work: the sex role revolution and its consequences, Population and Development Review, 1984, 10(3):397-417; Mazharul Islam M, Adolescent child-bearing in Bangladesh, Asia-Pacific Population Journal, 1999, 14(3):73-87; Perez AE, An analysis of the effects of fertility on women's special mobility in the Philippines, Asia-Pacific Population Journal, 1992, 6(4):35-66; and Hong S and Seltzer J, Impact of family planning on women's lives: toward a conceptual framework and research agenda, Working Paper, Research Triangle Park, NC, USA: Family Health International, 2002, No. 94-02.
18. Schuler SR, Hashemi SM and Jenkins AH, Bangladesh's family planning success story: a gender perspective, International Family Planning Perspectives, 1995, 21(4):132-137 & 166; Schuler SR and Hashemi SM, Defining and Studying Empowerment of Women: A Research Note from Bangladesh, Boston, MA, USA: John Snow International Research and Training Institute, 1993; Schuler SR and Hashemi SM, Effects of Rural Credit Programs and Women's Empowerment on Use of Contraception in Bangladesh, Boston, MA, USA: John Snow International Research and Training Institute, 1993; Schuler SR, Empowerment and family planning in Bangladesh, Network, 1994, 15(1):16-17; Schuler SR and Hashemi SM, Credit programs, women's empowerment, and contraceptive use in rural Bangladesh, Studies in Family Planning, 1994, 25(2):65-76; and Meekers D, Schuler SR and Hashemi SM, Can credit programs for women affect reproductive behavior? the case of rural Bangladesh, Working Paper, University Park, PA, USA: Pennsylvania State University, 1992.
19. Bates L, Schuler SR and Islam MK, From home to the clinic: the next chapter in Bangladesh's family planning success story, paper presented at the International Union for the Scientific Study of Population Seminar on Fertility and Family Planning Programmes in the 21st Century, Dhaka, Bangladesh, Jan. 18-20, 2000.
20. Arends-Kuenning M, Hossain MB and Barkat-e-Khuda, 1999, op. cit. (see reference 9).
21. Barkat-e-Khuda, Kane TT and Phillips JF, Improving the Bangladesh Health and Family Planning Programme: Lessons Learned Through Operations Research, Dhaka, Bangladesh: International Centre for Diarrhoeal Disease Research, Bangladesh, 1997; and Ministry of Health and Family Welfare, Bangladesh, Strategic Directions for the Bangladesh National Family Planning Program 1995-2000, Dhaka, Bangladesh: Ministry of Health and Family Welfare, 1996.
22. Ministry of Health and Family Welfare, Bangladesh, Conceptual Framework for Health, Nutrition, and Population Sector Programme (HNPSP), July 2003-June 2006, Dhaka, Bangladesh: Ministry of Health and Family Welfare, 2003.
23. Phillips JF et al., Transferring health and family planning service innovations to the public sector: an experiment in organization development in Bangladesh, Studies in Family Planning, 1984, 15(2):62-73.
24. Mozumder KA et al., The Sample Registration System: an innovative system for monitoring demographic dynamics, Asia-Pacific Population Journal, 1990, 5(3):63-72.
25. Balk D, Individual and community aspects of women's status and fertility in rural Bangladesh, Population Studies, 1994, 48(1):21-45.
26. Balk D, Defying gender norms in rural Bangladesh: a social demographic analysis, Population Studies, 1997, 51(2):153-172.
27. Liang KY and Zeger SL, Inference based on estimating functions in the presence of nuisance parameters, Statistical Science, 1995, 10(2):158-173; and Zeger SL et al., Models for longitudinal data: a generalized estimating equation approach, Biometrics, 1988, 44(4):1049-1060.
28. Schuler SR, Hashemi SM and Jenkins AH, 1995, op. cit. (see reference 18).
29. Islam MA, Islam MM and Chakraborty N, Plateauing of fertility level in Bangladesh: exploring the reality, Dhaka, Bangladesh: Dhaka University, 2002.
James F. Phillips is senior associate, Policy Research Division, Population Council, New York, and Mian Bazle Hossain is assistant professor, Public Health Program, Morgan State University, Baltimore, MD, USA.
This research was funded by U.S. Agency for International Development grant DPE-3005-A-00-8059-00. The authors thank Dr. Barkat-e-Khuda for helpful comments on an earlier draft of this manuscript.