The Effect of Facility Characteristics on Choice of Family Planning Facility in Rural Tanzania
wp-02-61.pdf — PDF document, 72 kB (74521 bytes)
Author(s): Chen S, Guilkey D K
A major goal of most family planning programs in developing countries is to increase quality and access to family planning facilities. While it has long been hypothesized that contraceptive users are responsive to their supply environment, there is scant evidence showing that access and quality factors, have an effect on family planning use (for recent work on the topic see, for example: Bertrand, Hardee, Magnani and Angle 1995; Cochrane and Guilkey 1995; Frankenberg, Sikoki, Suriastini, and Thomas 2001a and 2001b; Koenig, Hossain, and Whittaker 1997; Steele, Curtis and Choe 2000; and Tsui, Ukwuani, Guilkey and Angeles 2001). Much of the work cited above on the role of the family planning supply environment on influencing contraceptive use has focused on access to care. Using distance and proximity as a measure of access to care, several studies have found that access is an important determinant of contraceptive use. Some studies that have tried to examine the role that quality plays on the use of contraceptives have found that quality is important, though it is hard to uncover significant effects for specific quality indicators (see, for example, Mensch, Arends-Kuening and Jain 1996). One study, Mroz, Bollen, Speizer and Mancini (1999), found that a community's subjective measure of quality had a significant impact on contraceptive prevalence in that community and that the size of the impact was larger than the other community measures such as time, distance and accessibility. In the past, studies that have examined the effect of quality on individual level family planning behavior have often used community level information on family planning service quality. For example, a knowledgeable individual within the community is selected to provide information on the nearest facility or a facility actually within the community. The facility is surveyed or the knowledgeable individual is asked about the type of services available at the facility. If the unit of analysis is the woman, each woman within the community is then assigned information from this facility. There are two problems with this type of analysis. The first is that within a community, each woman has no variation in the quality of family planning facility. Since they all are assigned the same facility, the quality attributes of the facility do not vary within a community. The quality variables are in some sense restricted since by construction, they can now only account for differences in behavior across communities (not within). Since the only other factors in the analysis are demand side factors such as education, these variables may be overstated as they are forced to explain the difference in behavior across all individuals. The second problem is that a woman may not attend the facility assigned to her. She may decide to go outside of her community to obtain family planning or to go to another facility which was not surveyed. This will result in measurement error in the quality attributes of the facility that the woman is purported to attend. Mensch et al (1996) in their study of facility quality used a situation analysis to obtain information on all the family planning facilities within 5 kilometers of each community which, in rural areas, is a better approximation to the market for services that the woman faces. They found that better services were associated with greater contraceptive use. Their measure of quality was however an index made up of a number of quality attributes so that the effect of any one factor is obscured. Even though they had information on the market of facilities that a woman could choose to attend, they were not able to make use of this information in their analysis. Our study represents a departure from the typical question asked in much of the family planning supply literature. Instead of examining the effect of the components of the supply of family planning on contraceptive prevalence or use we ask a more basic question. Among current users of family planning, what quality and access attributes influence a woman's choice of family planning facility? We are able to undertake this analysis because of an unusually rich data set which links rural women in Tanzania with their entire market of family planning facilities surrounding their community. In other words, we have data on all of a woman's options when she is contemplating where to go to receive family planning services. In addition, for 40% of modern contraceptive users in our sample we have information on the actual facility she attended even if it lies outside of the surrounding area. This means that we do not rely on community informants to determine where a woman goes for family planning and we have a more reliable accounting of the attributes of the facility that she attends. To determine the effect of specific facility attributes on facility choice we use McFadden's conditional logit model. In this model, the effect of choice characteristics are used as determinants of individual facility selection. We use distance as our measure of access, and two quality measures: at least one provider trained since 1992 and the number of modern family planning methods seen in stock. Compared to previous studies we have better data on trained providers and distance is more precisely estimated. The distance variable is based on a Global Positioning System (GPS) reading taken at the facility and measured from the center of the community. Our information on trained providers was based on a survey administered to all family planning providers and contained very specific questions on the type of training and the year that training was received. This type of study will help policymakers to identify what quality factors attract a woman to a facility and assist them in targeting their population programs to provide better care to women. After an introductory section, authors describe the data set used and present descriptive statistics. Section III presents the estimation methods and the empirical results. Authors conclude in section IV.
This document is not available in print from MEASURE Evaluation.