Maternal Healthcare in Tajikistan: A Bargaining Frameworkстатья из журнала
Аннотация: Abstract This study examines women's declining use of maternal healthcare services in post-socialist Tajikistan. Using data from the 2003 and 2007 Tajikistan Living Standards Surveys (TLSS), the findings support previous evidence that a woman's use of prenatal and delivery care depends on her education, household income, and proximity to services. However, previous models have not specified who makes the decision to use maternal healthcare services. This study finds that in Tajikistan a woman shares decision making with her spouse and the eldest woman in the household. There is limited evidence that traditional proxies for bargaining power, such as relative earnings level, affect outcomes. The authors conclude that where women's exit options are limited, surveys evaluating the value of women's assets and their services in the home, as well as questions about decision making, will allow more refined measures of women's bargaining power. Keywords: Bargaining powereconomics of genderhealthcarehousehold behaviormotherhoodJEL Codes: C78I00J16 ACKNOWLEDGMENTS We would like to thank the editors and anonymous reviewers at Feminist Economics for their excellent comments and suggestions on an earlier draft of this paper. We also greatly appreciate comments, suggestions, ideas, and statistical support from Dr. Tom Hertz, Dr. Sibel Selcuk, Taggert J. Brooks, and John M. Nunley. Notes However, this positive relationship with prenatal care has not been found universally. Shiffman (Citation2000) finds that while assistance in childbirth does have a positive impact on reducing infant and maternal mortality rates, the same cannot be shown for use of prenatal care using cross-country aggregate data. Russian Orthodox and other Christians make up less than 5 percent of the population (Glenn E. Curtis Citation1996; World Bank Citation1999, Citation2003). The use of sample weights has been the subject of much debate. If the parameter one seeks to estimate is heterogeneous across people, then neither the weighted nor the unweighted results will yield a consistent estimate of the population mean of that parameter. On the other hand, if the parameter is in fact the same for all people, then the unweighted results are to be preferred, as they are more efficient. Thus, the weights distort the collected data but do not result in an efficient estimate (Angus Deaton Citation1997). We choose to use the weights to keep our work as comparable as possible with other research. We have also performed the analysis without weights, however, and the results are comparable in both their implications and statistical significance. Other work on this issue appears to have included all births reported in the sample (Falkingham Citation2003; Habibov and Fan 2008). Interviewees were free to identify any household member as their spouse. Only men were identified as “spouse” by the women respondents in this survey. As our sample includes only women with identifiable spouses, the results are valid only for this group. However, as we note, the results are very similar to those found by Habibov and Fan (2008) and Falkingham (Citation2003), who included all women who answered the questions regarding their last pregnancy. All researchers using these data must exclude the women who did not respond to questions about their last pregnancy. These might be women who are very conservative or who had very negative experiences in pregnancy. That might result in an undersampling of such women, but it is not clear why such exclusion would bias the results in any predictable way. We also examined a third outcome: birthing in a medical facility. The results were not statistically significantly different from the results regarding the more general question of delivery care, so we present the results for only two outcomes. Habibov and Fan (2008) use a two-stage ZINB regression method to model the number of prenatal visits given a decision to seek prenatal care at all using just the 2003 data. Due to the apparently inconsistent way that the data on number of prenatal visits were collected in 2003 and 2007, we chose instead to follow Hotchkiss (Citation2001) in presenting probit results for a dummy for multiple visits. In 2003, the number of visits varied between one and nine. In 2007, the number of visits varied between one and thirty-six, with 5 percent of the sample reporting over thirteen visits. It appears that the question was not asked, or coded, the same way in both years, and as a result we were disinclined to use all the data on number of visits. However, we performed two robustness tests for our modeling choice. First, to check for sensitivity to our choice of four visits for our dummy variable, we reran the regressions using a dummy for more than three visits and also for more than five visits. The results are identical, except that in the case of more than five visits the number of previous births has a statistically significant positive impact. Second, we followed Habibov and Fan (2008) in running a ZINB. The results for the second stage (the number of visits, given that at least one visit would occur) were identical to those resulting from the probit regressions presented here for the dummy variable Greater than Four Visits in sign and significance, with one exception. Number of Births is significant in the ZINB, but significant in the probit version only in the robustness test, where we used a dummy for Greater than Five Visits. The ZINB results are available from the authors. Both the 2003 and 2007 surveys include information on the existence of and distance to both the closest polyclinic and women's clinic. We use the shorter of these two distances. The correlation between a woman's education and that of her spouse is 0.35; correlation between the woman and the eldest woman is 0.16. The regressions control for the education of both a woman and her spouse, and continuous measures of relative education are colinear with those variables. Women's labor force participation is low, and data on employment were very incomplete. Among women who provided full answers to the questions about paid work, the vast majority of those who worked for pay were employed full time. There was little variation across different possible measures of access to income (hours worked for pay, weeks worked for pay), and the choice of measure did not affect the outcome. There is very little correlation between these bargaining power variables. Correlation coefficients range from -0.0027 (between the dummy for employed and woman head of household) to 0.0945 (between the education of the eldest woman and that of the spouse). Only about one-third of doctors in Tajikistan are women, which may contribute to reported feelings of being “ashamed.” Gorno-Badakhshan, which has statistically significantly higher levels of use of maternal health services, also has the highest share of women doctors, 67 percent (Andrew Dabalen and Waly Wane 2008). This was a closed question, and most options reflected an underlying model as in Figure 1. No options were offered regarding the woman's role in the choice.
Год издания: 2012
Авторы: Mieke Meurs, Lisa Giddings
Издательство: Taylor & Francis
Источник: Feminist Economics
Ключевые слова: Gender, Labor, and Family Dynamics, Global Maternal and Child Health, Global Health Care Issues
Открытый доступ: closed
Том: 18
Выпуск: 3
Страницы: 109–140