SLIDE 1 Men’s Labor Migration and Women’s Health and Mortality in Rural Mozambique
Victor Agadjanian, University of Kansas Sarah R. Hayford, Ohio State University Natalie A. Jansen, University of Kansas Abstract As in many other parts of the world, labor migration is widespread and growing in southern
- Africa. Yet the economic outcomes of migration diversify, and so do its consequences for the
wellbeing and health of non-migrating household members. We use data from a longitudinal project conducted over eleven years in rural Mozambique to examine the association between men’s labor out-migration and their non-migrating wives’ self-rated health and mortality. The analyses detect no significant differences when comparing non-migrants’ wives to migrants’ wives in the aggregate but point to instructive variation among migrants’ wives according to the economic success of migration. Specifically, women married to unsuccessful migrants are less likely to report being in good health and have higher mortality risk over the project span than women married to successful migrants. These patterns generally persist whether migration success is defined based on reported remittances or on the wife’s subjective assessment and are impervious to the addition of individual and household-level controls.
SLIDE 2 1 Introduction Labor migration has been shown to improve non-migrating household members’ health and wellbeing by generating remittances that increase household food supply and access to health care (Adams & Page 2005; Adams & Cuecuecha 2013; Böhme et al. 2015; Frank & Hummer 2002; Gupta et al. 2009; Hamilton et al. 2009; Ponce et al. 2011). However, the migrant’s absence from the household may also increase psychological strain, depressive symptoms and
- ther health impairments among non-migrating adult household members (e.g., Adhikari et al.
2011; Chen et al. 2015; Huang et al. 2016; Lu 2012; Lu et al. 2012; Menjívar & Agadjanian 2007; Nobles et al. 2015). In addition, considerable research has documented an association of men’s migration with greater perceived and actual risks of HIV and other STIs among their left- behind wives (Agadjanian et al. 2011; Agadjanian & Markosyan 2016; Agadjanian et al. 2013; Hirsch et al. 2007; Sevoyan & Agadjanian 2010; Weine & Kashuba 2012). In this study we use data from a longitudinal project in rural southern Mozambique to examine the relationship of men’s labor migration with their marital partners’ health and mortality while focusing in particular on the variation in the economic impact of migration. As in many parts of the world, labor migration, both internal and international, is widespread and growing in southern Africa. Most international labor migration flows in the region are directed toward the Republic of South Africa (RSA) and originate from its poorer neighbors, including the Republic of Mozambique. Given its proximity to RSA and volatility of its rural economy, the southern part of Mozambique has traditionally been a major source of primarily male labor migration, and this migration continues today. However, in recent times this labor migration has increasingly shifted from highly regulated employment mainly in the South African mining industry to less formal and temporary work. Accordingly, the economic
SLIDE 3 2
- utcomes of migration and the impact of migration on sending households, which is largely
shaped by economic remittances, have also become more diverse. Our analyses seek to account for this growing diversity of migration outcomes. Conceptualization and hypotheses To the extent that migration-generated remittances are used for food or health care needs and enhance the household overall economic security, we would expect a positive association between migration and the health and well-being of members of the sending household that varies depending on the level of remittances. By the same token, migration that supplies the household with stable resources should be associated with lower mortality risks of non-migrating
- members. In contrast, migration that does not produce a steady flow of remittances may be
detrimental to health and mortality outcomes of the household members, as such migration imperils the household’s economic security, thus further aggravating the social and psychological consequences of the migrant’s physical absence from the household. Our approach builds upon previous studies that have illustrated the importance of accounting for the diversity of migration economic outcomes. For example, Yabiku, Agadjanian, and Cau (2012) found that migration was associated with lower mortality among migrants’ left-behind children in rural Mozambique, but that this association was only present for economically successful migrants. Other studies, including those carried out in the same context as the present analysis, also show that the effects of migration on health and general wellbeing of non- migrating household members vary according to the economic returns of migration to the households (e.g., Agadjanian and Hayford 2017; Agadjanian, Arnaldo, and Cau 2011;
SLIDE 4 3 Agadjanian, Yabiku, and Cau 2011, Lu 2010; Lu et al. 2012; Yabiku and Agadjanian 2017; Yabiku, Agadjanian, and Cau 2012). While the economic returns from migration to the household are critical for its members’ wellbeing and health, migration may affect these outcomes through other pathways. Thus, migrants’ exit from the community may expose them to greater risks of sexually transmitted diseases, including HIV, especially in settings of high HIV prevalence, which may, in turn, affect their marital partners back home (Agadjanian et al. 2011; Agadjanian & Markosyan 2016; Agadjanian et al. 2013; Hirsch et al. 2007; Sevoyan & Agadjanian 2010; Weine & Kashuba 2012). Migrants’ prolonged absence from the household may also lead to heightened anxiety, stress, and depression among their non-migrating partners and other household members (e.g., Adhikari et al. 2011; Chen et al. 2015; Huang et al. 2016; Lu 2012; Lu et al. 2012; Menjívar & Agadjanian 2007; Nobles et al. 2015). Yet, as Lu et al. (2012) argued in their study of rural China, migrant remittances may partly offset the negative effects of migration on non-migrant household members’ mental health. Finally, migration may lead to the dissolution of migrants' marital unions, exposing their partners, children, and other family members to economic and social hardships and, by extension, heightening their health vulnerabilities. However, again, as Agadjanian and Hayford (2017) showed in Mozambique, marital unions of economically successful migrants are less likely to dissolve than those of unsuccessful ones. Importantly, because labor migration is typically a long-lasting state, and because improved nutrition and improved medical care may have long-term impacts, the impact of migration on health of non-migrants may accumulate over time. However, the cumulative effects of migration
- n health of non-migrating household members are not well understood largely because adequate
longitudinal data to capture and measures these effects are often lacking. Arguably, this
SLIDE 5 4 knowledge gap is especially critical for the understanding of health implications of migration for migrants’ non-migrating marital partners, who typically experience the most direct impact both
- f the economic returns to migration and of the social and emotional disruptions and strains
caused by the departure and prolonged absence of the migrant. In this study, we seek to fill this gap in the literature by investigating the association between men’s migration and their marital partners’ self-rated health and mortality in rural southern Mozambique. For both outcomes, we compare wives of migrants and non-migrants, but we also seek to capture the effect of diverging economic fortunes of migrant households. Following the previous research that highlighted the importance of this divergence, rather than a simple migrant vs. non- migrant dichotomy, we contrast self-rated health and mortality of women whose husbands have been economically successful migrants with those of women married to economically unsuccessful migrants. Importantly, migration success is defined here from the standpoint of the migrant’s household in the community of origin, i.e., the actual or perceived impact of migration
- n the household wellbeing, rather than the absolute amount of migrant earnings. Guided by
previous work (Agadjanian, Arnaldo, and Cau 2011; Yabiku, Agadjanian, and Cau 2012), we conceptualize migration success both objectively, in terms of remittances received by the migrant’s household, and subjectively, i.e., in terms of the assessment of the impact of migration
- n the household wellbeing by non-migrating wives. We formulate and test two general
- hypotheses. First, we hypothesize that women married to unsuccessful migrants will have worse
self-rated health that those married to successful migrants, net of other factors (Hypothesis 1). Second, we hypothesize that mortality risks will be higher among wives of unsuccessful migrants compared to successful migrants’ wives (Hypothesis 2).
SLIDE 6 5 Data and Method Data Our data come from a longitudinal survey of rural women in four districts of the southern Gaza
- province. The initial wave of the survey was carried out in 2006. First, 56 villages, 14 per
district, were selected with probability proportional to population size. In each village all households containing at least one married couple (defined broadly to include both formalized and non-formalized marriages) were enumerated and separated into households with a married male migrant and households without. Then, fifteen households were randomly selected from each list and one married woman aged 18-40 from each household was interviewed. The procedure yielded a total sample of 1680 women, of whom just under one-half were married to labor migrants. The sample was revisited and re-interviewed in 2009 (Wave 2) and 2011 (Wave 3). In both Waves 2 and 3, the sample was randomly refreshed for respondents who could not be
- found. However, in follow-up attempts later in the survey year and early in the following year
some of the original respondents who had not been interviewed were located and interviewed. Their substitutes were retained in the sample. As a result, the total sample size increased to 1868 in 2009 and 2059 in 2011. In both Waves 2 and 3, detailed information on women’s demographic and socioeconomic characteristics, their health status, and their households was
- collected. A short survey (Wave 4) focusing on respondents’ key sociodemographic outcomes
and their adolescent children was carried out in 2014. No sample refreshment was done in Wave 4; in that year 1975 women were interviewed. Finally, Wave 5 was carried out mainly between May-July 2017; a follow-up targeting women who were not available during the main data collection period is still underway at this writing. At all waves after the first, for respondents who could not be located, a proxy interview was conducted with someone who knew the respondent
SLIDE 7
6 (typically another member of the household or a neighbor). These interviews recorded basic information about residential mobility and demographic events since the last survey interview, including the fact, year, and possible cause of respondent’s death. Overall, the Wave 5 (2017) survival status could be ascertained for 1570 (93.6%) of Wave 1 respondents. The Wave 5 survival data are used in the analysis of mortality. The current analysis of self-rated health uses data from Wave 3. Method Our analysis focuses on two outcomes – self-rated health measured at Wave 3 (2011) and mortality by Wave 5 (2017). Self-rated health has been widely used in health research, including in Mozambique and sub-Saharan countries, and has been shown to be correlated with objective physical health outcomes and to predict future health and mortality (e.g., Cau et al. 2016; Debpuur et al. 2010; Frankenberg & Jones 2004; Idler & Benyamini 1997; Jylhä 2009). It is typically measured in surveys by asking respondents to rate the state of their overall health on a scale from very bad to excellent, and there appears to be little variation in the external validity of self-rated health or its predictive power related to question wording or the number of response categories (Fayers & Sprangers 2002; Idler & Benyamini 1997). In this survey, self-rated health was measured in Wave 3 using a four-level scale – excellent, good, so-so, or bad. Based on the distribution of responses, we operationalize this variable as a dichotomy: excellent/good health (coded as 1) vs. so-so/bad (coded as 0). Accordingly, we fit binomial logistic regression models. To explore both cumulative and immediate associations of migration, and its economic outcome, with health, we use two modeling strategies. First, we model the effect of husband’s migration characteristics and other covariates are measured at Wave 1. These analyses are therefore restricted to Wave 1 respondents who were interviewed in Wave 3. These analyses assess the
SLIDE 8
7 longer-term impact of migration on self-rated health that may operate cumulatively via health care access and utilization, food security, physical and mental strain, marital stability, and other factors over a five-year time frame. A drawback of this approach is that it reduces the sample size by excluding Wave 1 women who were lost to follow-up or died by Wave 3. There is no way to soundly integrate the former of these two segments of the original sample in the analysis. However, we can include respondents who died by Wave 3. Because poor self-rated health and risk of death are highly correlated (Frankenberg & Jones 2004; Idler & Benyamini 1997; Jylhä 2009), we also fit a model in which add all these additional respondents, assigning them “0” on the outcome. Second, we fit models in which all predictors and controls are measured at Wave 3 to capture short-term impacts of migration on health. These cross-sectional analyses include all the respondents who were in marital union at Wave 3. To examine the relationship of migration with mortality, we fit logistic regression models predicting whether respondents interviewed in Wave 1 were still alive at Wave 5, i.e., approximately twelve years later. These analyses are limited to Wave 1 respondents for whom the survival status in 2017 could be ascertained. The main predictor is husband’s migration status. We define it as a set of dummy variables: not migrant, successful (“good”) migrant, and unsuccessful (“bad”) migrant. Following our conceptualization and previous research, we use two alternative approaches to define migration success – “objective” and “subjective.” The objective definition is constructed from the data on remittances sent by migrants. It includes three levels, separating migrants whose wives reported that they send remittances regularly or frequently, from those who send remittances occasionally, and from those who do not send any remittances. The subjective definition is derived from wife’s perception of improvement of the household’s economic conditions as a result of migration. This definition is dichotomous, contrasting women who thought that their household conditions had
SLIDE 9 8 improved to those who saw no improvement or reckoned that household conditions had
- worsened. Accordingly, for all outcomes, we fit three models: a model in which women with
migrant husbands are compared to women with non-migrant husbands; a model in which women married to migrants who send no remittances are compared to those whose migrant husbands send frequent remittances, those whose migrant husbands send occasional remittances, and women married to non-migrants; and a model that compares subjectively good migrants, subjectively bad migrants, and non-migrants. The models are restricted to women who reported being in a marital union (broadly defined to include both formalized marriages and non-formalized co-residential partnerships). They control for individual and household characteristics: respondent’s age; number of her living biological children; respondent’s educational level (no education, 1-5 years, 6 or more years of schooling), bridewealth status of union (at least some bridewealth paid vs. none); polygyny (in polygynous vs. monogamous union); respondent’s work outside subsistence agriculture (works
- vs. does not work), affiliation with a formal religion (affiliated vs. not), and household assets
- scale. The models also control for respondent’s self-assessed likelihood of being HIV positive
(confirmed or likely HIV+ vs. unlikely or impossible). Because self-reported HIV status was not included in the Wave 1 instrument, the models instead control for respondent’s worries about contracting HIV (very worried vs. not very worried or not worried). The multivariate analyses are carried out in STATA. Multilevel logistic regressions are fitted to account for clustering of observations and unobserved heterogeneity at the village level.
SLIDE 10 9 Results Table 1 presents bivariate distributions of the outcome variables across different categories of migration status of respondents’ husbands. Following our interest in both longer-term and immediate effects of migration on health, Section A.1 and A.2 display percentages of Wave 3 respondents who reported being in excellent or good health among surviving Wave 1 respondents and among all currently married respondents in Wave 3. The first two rows suggest some advantage of all migrants’ wives, compared to non-migrants’ wives, in self-rated health, but the difference is modest, especially when husband’s migration status is measured at Wave 3. However, when the migrant-husband subsample is broken down by frequency of remittances, the contrast among these different subgroups of migrants’ wives is quite noticeable: women married to migrants who do not send remittances at all are noticeably less likely to report being excellent
- r good health, compared to women married to regular remitters. Women whose migrant
husbands send occasional remittances find themselves between the two extremes. The distinction between wives of “good” and “bad” migrants, defined subjectively, produces a similar pattern even though the contrast between the two categories is somewhat less pronounced than in the case of remittance-based classification. Regardless of the definition of migration success, the differences within the migrant subsample are more prominent when migration success is measured in Wave 3 (i.e., at the same time as self-rated health). Table 1 here Section B shows the fractions of Wave 1 respondents who died by Wave 5 across different migrant-husband categories. Again, the overall differences between migrants’ wives and non-
SLIDE 11
10 migrants’ wives in mortality rates is trivial. However, the breakdown of the migrant subsample by migrant quality produces a pattern that is very similar to that of self-rated health: women married to more successful migrants, defined either objectively or subjectively, have lower mortality than women married to unsuccessful migrants. In fact, mortality rates among wives of unsuccessful migrants are noticeably higher than among non-migrants’ wives. The results of the multivariate logistic regression models predicting self-rated health in Wave 3 from husband’s migration status in Wave 1 (Section A) and in Wave 3 (Section B) are shown in Table 2. The model in Section 2.A.1 compares women married to migrants and those married to non-migrants. It shows no net difference between the two categories of women. The model in Section 2.A.2 breaks down the migrant husband subsample by frequency of remittances. The reference category is women married to migrants sending no remittance. The results conform to the trend displayed in the bivariate distribution, but the differences do not reach the conventional level of statistical significance. As described in the methods section, to increase the sample size and the statistical power, we add to the analysis the Wave 1 respondents who died by Wave 3, coding all these additional cases as “0” on the outcome. The new model, presented in 2.A.3, shows the same pattern as in the previous one, but the difference between wives’ of regular remitters and wives of migrants who were sending no remittances at Wave 1 is now larger and is statistically significant. Section 2.A.4 shows the results of a model in which the migrant-husband subsample is disaggregated based of women’s assessment of the impact of migration on the household’s living conditions. This model shows no significant differences across migration status (the expansion of the sample to include women who died by Wave 3 did not change the effects of the predictor, and we do not present the results of this model here.)
SLIDE 12 11 Table 2 here Section B of Table 3 displays the results of the cross-sectional analyses, in which both the
- utcome and all the covariates are measured at Wave 3. Again, there is no difference between
wives of migrants and non-migrants (Model 2.B.1). However, both the model where husband’s migration success is defined “objectively”, i.e., based on regularity of remittances (Model 2.B.2), and “subjectively,” i.e., based on wife’s assessment of the impact of migration on household’s material conditions (Model 2.B.3), show significant differences in the predicted direction. Thus, wives of regular remitters report significantly better health than do women receiving no remittances from their migrant husbands; in fact, women married to frequently remitting husbands also have better self-rated health than women married to non-migrants net of other factors (not shown). Women who think that their households are better off as a results of their husbands’ migration are in significantly better health than those who think that their husbands’ migration has worsened and have not changed the household’s conditions. In all, these results, especially those from the cross-sectional models, generally support Hypothesis 1. Table 3 displays the results of three logistic regression models predicting Wave 1 respondent’s death by Wave 5. As in the previous set of models, three approaches to husband’s migration are tested. Again, as in the analysis of self-rated health, Model 3.A shows no difference in mortality between wives of migrants and non-migrants. However, as in the previous analysis, the breakdown of the migrant subsample by frequency of remittances (Model 3.B) produces a statistically significant difference between the two extremes of the migrant quality spectrum: women receiving regular remittances in 2006 were more likely to be alive eleven years later than women whose husbands were not sending any remittances when they were first
SLIDE 13 12
- interviewed. Finally, Model 3.C produces a similar pattern although the differences in mortality
rates between “bad” and “good” migrants’ wives is only marginally significant (p<0.07) when
- ther factors are controlled for. Again, these results generally support Hypothesis 2.
Table 3 here Discussion The foregoing analyses offer intriguing insights into the association of labor migration with health and mortality of non-migrant household members. The findings illustrate the diversity of economic outcomes of contemporary labor migration in southern Mozambique and similar settings of rapid social change, as well as the health and mortality implications of that diversity. Not surprisingly the divergence between health and mortality risks of women married to migrants of differing quality was particularly pronounced when this quality was assessed “objectively,” i.e., on the basis of remittance frequency, but the findings also suggest that the “subjective” measure of migration success, which is arguably more inclusive and multilayered, could also be of value predicting health and mortality. It is important to reiterate, however, that both definitions of migration success reflect the non-migrant women’s (and their household’s) perspectives: migrant earnings only matters to these women’s wellbeing if they reach them. At this stage of the analysis we can only speculate about possible pathways through which the quality of men’s labor migration may affect their wives’ health and mortality. For example, as was mentioned earlier, migrant remittances may enhance women’s nutrition and improve their access to health care. Food and general economic security may require less time and effort from women for debilitating farming activities (which in this settings are performed primarily by
SLIDE 14 13 women). Regular remittances from their husbands may also enhance women’s optimism and psychological wellbeing, compensating for the strain resulting from spousal separation. Finally, in a setting of high HIV prevalence, HIV risks may vary across different types of migrants. Several limitations must be acknowledged. As with any observational data, causality cannot be fully ascertained. Specifically, the association of migration and its characteristics with health
- utcomes of non-migrating household members may reflect some endogeneity of migration with
respect to these outcomes. For example, it is conceivable that a man’s decision to migrate or his commitment to his marital partner, and therefore the amount of remittances he sends to her, may be influenced by the state of her health. The use of longitudinal data helps reduce the possibility
- f reverse causation. However, even with the longitudinal data at hand this possibility cannot be
fully eliminated. Another limitation is attrition from the sample due to loss to follow-up. Despite substantial and largely successful efforts to locate and interview the respondents who have moved from the villages in which they were first interviewed, some of those women could not be
- located. Proxy interviews with other community members aimed to determine the causes,
circumstances, and destination of such relocations point to a large share of divorced women among those who moved. Because in this patrilineal and virilocal setting women move to their husbands’ communities and households upon marriage, when they divorce they are typically ejected from the community and are often effectively effaced from the household’s and community’s memory and therefore are very hard to locate. The current analyses assume that these “missing” women do not have health profiles and mortality risks that are systematically distinct from the women that were interviewed or for whom survival status could be ascertained. Finally, we also acknowledge the statistical power of the analysis as a possible limitation.
SLIDE 15
14 References Adhikari, R., Jampaklay, A., & Chamratrithirong, A. 2011. Impact of children's migration on health and health care-seeking behavior of elderly left behind. BMC public health 11(1), 1. Agadjanian, V., Arnaldo, & C., Cau, B. 2011. Health costs of wealth gains: Labor migration and perceptions of HIV/AIDS risks in Mozambique. Social Forces 89(4), 1097-118. Agadjanian, V., and S. R. Hayford. 2017. “Labor migration and marital dissolution in rural Mozambique” Journal of Family Issues (online first) Agadjanian, V, & Markosyan, K. 2016. Male labor migration, patriarchy, and the awareness- behavior gap: HIV risks and prevention among migrants’ wives in Armenia. AIDS Care. Agadjanian, V., S. T. Yabiku, and B. Cau. 2011. “Men’s migration and women’s fertility in rural Mozambique” Demography 48(3): 1029-1048 Cau, B. M., Falcão, J., & Arnaldo, C. 2016. Determinants of poor self-rated health among adults in urban Mozambique. BMC Public Health 16, 856. Chen, F., Liu, H., Vikram, K., & Guo, Y. 2015. For better or worse: The health implications of marriage separation due to migration in rural China. Demography 52(4), 1321-43. Debpuur, C., Welaga, P., Wak, G., & Hodgson, A. 2010. Self-reported health and functional limitations among older people in the Kassena-Nankana District, Ghana. Glob Health Action 3(Suppl 2), 54-63. Frankenberg, E., & Jones, N. R. 2004. Self-rated health and mortality: does the relationship extend to a low income setting? Journal of Health and Social Behavior 45(4), 441-52 Huang, B., Lian, Y., & Li, W. 2016. How far is Chinese left-behind parents' health left behind? China Economic Review 37, 15-26.
SLIDE 16 15 Idler, E. L., & Benyamini, Y. 1997. Self-rated health and mortality: A review of twenty-seven community studies. Journal of Health and Social Behavior, 21-37 Jylhä, M. 2009. What is self-rated health and why does it predict mortality? Towards a unified conceptual model. Social science & medicine 69(3), 307-16 Lu, Y. 2010. Rural-urban migration and health: Evidence from longitudinal data in Indonesia. Social Science & Medicine, 70(3), 412-9. Lu, Y. 2012. Household migration, social support, and psychosocial health: The perspective from migrant-sending areas. Social Science & Medicine, 74(2), 135-42. Lu, Y., Hu, P., & Treiman, D. J. 2012. Migration and depressive symptoms in migrant-sending areas: findings from the survey of internal migration and health in China. International Journal
- f Public Health, 57(4), 691-8.
Sevoyan, A., and V. Agadjanian. 2010. Male migration, women left behind, and sexually transmitted diseases in Armenia. International Migration Review, 44(2), 354-75. Weine, S. M., & Kashuba, A. B. 2012. Labor migration and HIV risk: A systematic review of the
- literature. AIDS and Behavior, 16(6), 1605-21.
Yabiku, S. T., and V. Agadjanian. 2017. “Father's labor migration and children's school discontinuation in rural Mozambique” International Migration (online first) Yabiku, S. T., V. Agadjanian, and Boaventura Cau. 2012. “Labor migration and child mortality in Mozambique” Social Science & Medicine 75(12): 2530-2539
SLIDE 17 16 Table 1. Descriptive results (percentages) Husband’s migration status
- A. Excellent or good self-rated
health (Wave 3)
(by Wave 5) N=1570
status in Wave 1 N=1328
status in Wave 3 N=1820 Non-migrant 72.81 74.87 9.38 Any migrant 77.47 76.93 10.00 Migrant, sending regular remittances 80.11 83.52 5.42 Migrant, sending occasional remittances 77.06 77.39 10.34 Migrant, sending no remittances 74.83 70.44 12.37 Migrant, household conditions improved since migration (“good” migrant) 79.29 81.36 7.29 Migrant, household conditions have not improved or worsened since migration (“bad” migrant) 75.49 72.08 11.53 All 74.70 75.60 9.75
SLIDE 18 17 Table 2. Excellent/good self-rated health in Wave 3, multilevel logistic regression parameter estimates and standard errors Covariates
- A. Migration in Wave 1
- B. Migration in Wave 3
1 2 3 4 1 2 3
Migrant husband [non-migrant]
.153 0.076 (.141) (0.124)
Non-migrant husband
.000 .029 .050
.000 0.219 0.113 (.185) (.159) (.176) (0.181) (0.155)
[Migrant husband, no remittances]
.000 .000 .000 .000 .000
Migrant husband, occasional remittances
.163 .176 0.320 (.217) (.177) (0.229)
Migrant husband, regular remittances
.375 .469* 0.691** (.246) (.206) (0.273)
[“Bad” migrant, subjective] “Good” migrant, subjective
.252 0.397* (.229) (0.202)
Age
- .056**
- .057**
- .054**
- .056**
- 0.052**
- 0.052** -0.051**
(.016) (.016) (.015) (.016) (0.011) (0.011) (0.011)
Number of biological children
.069 .071 .110 .066 0.026 0.025 0.025 (.057) (.057) (.053) (.058) (0.038) (0.038) (0.038)
1-5 years education [No education]
.113 .104 .051 .107 0.185 0.166 0.183 (.154) (.156) (.146) (.155) (0.136) (0.137) (0.137)
6 or more years education
0.324 0.288 0.304 (.238) (.238) (.208) (.241) (0.205) (0.207) (0.206)
Bridewealth paid fully or partially [None paid]
0.123 0.097 0.103 (.150) (.150) (.138) (.149) (0.124) (0.125) (0.125)
In polygynous union [In monogamous union]
0.095 0.126 0.109 (.139) (.139) (.138) (.140) (0.136) (0.137) (0.136)
Has a religious affiliation [Has no affiliation]
.103 .098 .023 .097 0.194 0.180 0.183 (.139) (.141) (.123) (.138) (0.223) (0.223) (0.223)
Works outside subsistence farming [Does not work]
.313+
0.066 0.070 0.076 (.039) (.173 .151 (.173) (0.125) (0.125) (0.125)
Household assets scale
.173 .038 .017 .037 0.058** 0.054** 0.055** (.024) (.024) (.021) (.024) (0.020) (0.020) (0.020)
Very worried about getting HIV [Not very worried]
.004 .006
.011 (.158) (.159) (.151) (.158)
Very likely to be HIV+ [Not very likely]
(0.191) (0.192) (0.191)
Constant
2.39** 2.40** 2.077** 2.441** 2.304** 2.127** 2.199** (.400) (.446) (.396 (.434) (0.402) (0.418) (0.410)
Constant, level 2 (village)
.151 .158 .119 .159 2.304** 2.127** 2.199** (.068) (.069) (.056) (.071) (0.402) (0.418) (0.410)
Number of cases
1327 1327 1410 1327 1803 1803 1803 Notes: All predictors in Section A are measured at Wave 1; all predictors in Section B are measured at Wave 3; Model A.3 includes respondents who died by Wave 3; references categories in brackets; standard errors in parentheses; + p < 0.1, * p < 0.05, ** p < 0.01.
SLIDE 19 18 Table 3. Died by 2017, multilevel logistic regression parameter estimates and standard errors Covariates A B C Migrant husband [non-migrant husband]
(0.190) Non-migrant husband
.000 (0.259) (0.228) [Migrant husband, no remittances] .000 .000 Migrant husband, occasional remittances
(0.323) Migrant husband, regular remittances
(0.398) [“Bad” migrant, subjective] “Good” migrant, subjective
(0.272) Age 0.042* 0.045* 0.042* (0.018) (0.018) (0.018) Number of biological children
(0.074) (0.073) (0.074) 1-5 years education [No education]
(0.216) (0.214) (0.216) 6 or more years education 0.051 0.061 0.076 (0.276) (0.277) (0.283) Bridewealth paid fully or partially [None paid]
(0.169) (0.168) (0.174) In polygynous union [In monogamous union]
(0.187) (0.188) (0.185) Has a religious affiliation [Has no affiliation]
(0.221) (0.222) (0.222) Works outside subsistence agriculture [Does not work] 0.079 0.062 0.062 (0.198) (0.200) (0.198) Household assets scale 0.025 0.027 0.028 (0.029) (0.029) (0.030) Very worried about getting HIV [Not very worried] 0.176 0.170 0.171 (0.218) (0.220) (0.216) Constant
(0.583) (0.636) (0.558) Constant, level 2 (village) 0.046 0.049 0.059 (0.074) (0.074) (0.077) Number of cases 1557 1557 1557
Notes: All predictors measured at Wave 1 (2006); reference categories in brackets; Standard errors in parentheses;
+ p < 0.1, * p < 0.05, ** p < 0.01.
