 Hello everyone, I'm Paola Vila. I'm a PhD student at the Paris School of Economics and today I'm gonna present you a joint work with Dominique van de Val and Sylvie Lambert, sorry, which are here in the room. And I'm gonna talk about marital trajectories and women's well-being in Senegal. So first I'm gonna give you a bit of motivation. So in Senegal divorces are frequent and widowhood is a common situation for most women, more particularly at a certain point in their life. And this is mainly due to the fact that they marry men that are at least 10 years older on average. And indeed if we look at the data, what the the data I'm going to present you within a few slides says, in 2006 and 2007 in Senegal, among ever married women we had almost 90 percent of women that were current widows or remarried after a widowhood and 30 percent that were current divorced women or married after a divorce. So if you sum these two figures you have about at least one-third of women that had experienced a marital dissolution in Senegal in 2006. Also the remarriage is pretty frequent and when it takes place it takes place rather rapidly, for example, for women that remarried after widowhood the medium duration between widowhood and remarriage is one year and for divorce is two years. So when it happened it happened pretty fast. And so given how common these broken trajectories are we think that it's of interest to see whether and how it affects women's well-being. So there's actually little evidence in the economic literature at least on the impacts of marital dissolution on women's well-being in Sub-Saharan Africa, in particular when it comes to divorce. So for widowhood what we know actually mainly comes from the study of female-headed households in which most of the studies looks at households that are headed by widows or divorced women and then analyze the welfare impacts of this situation, marital situation. But actually what happened is that these papers don't look at or neglect the impact of remarriage. So because most of those women that remarried are absorbed within a household that is headed by a man. So here I'm citing some of the papers of the literature that looks at female-headed households and in particular at widows. And what they find is pretty mixed. So actually more negative than positive. But for example Appleton, Oral and Christian and Van der Waal find more negative impact of widowhood in a certain set of welfare indicators whereas Appleton and Oral find very heterogeneous impact of widowhood on land access. For divorce the core of what we know is mainly given by other types of social sciences, mainly sociology and demography and anthropology. And what they find is really more linked to the fact that divorce can be seen as a means of emancipation, mainly because we're in a context where the first marriage is often an arranged marriage between the family of the groom and the bride. And so divorce is a way for the woman to escape this marriage that is not a love marriage. And through remarriage she can choose a better or better suited husband for her. But at the same time there's some evidence that divorce because it's almost always linked also with some kind of social stigma can be linked to some difficulties in terms of material support. In the case for example where the divorce can be a bit stigmatized by our social network. So for both divorce and widowhood we can see that there's we can see some potential association with negative consequences and these consequences can be probably smoothed by remarriage or not. So there's potentially some long lasting effects of this dissolution on welfare. So what this paper does we think that it's at least to our knowledge one of the first to directly study the relationship between marriage dissolution and women's well-being in Senegal. It uses a rather recent and nationally representative data from a household survey that I'm going to detail a bit more within a few slides. And we also use some DHS data to firstly document Senegalese women's marital trajectories then to try to provide some evidence on the correlation between current consumption and other dimensions of welfare. So the link between the marital situation and some levels so some indicators proxies of welfare. And lastly to analyze the effects of selection into widowhood divorce and remarriage that might be a play because obviously there's a lot of selection and the genetic going on. So really we're not claiming that we have a proper one directional causal link because of all these reasons of selection and the genetic but we do think that we have some pretty salient patterns and correlations that are worse documenting. So some preview of the results we find that divorce and widowhood are associated with very different consequences for those women. Divorce seems to be a way to gain some relative autonomy. Education playing really a positive role mainly on the selection part. And whereas widowhood sorry is associated with more negative consequences and that are not mitigated by remarriage. And in particular we find that leviatic marriages which are the fact that a widow can marry the brother of some relative of the deceased husband play a very negative role at least correlated with the level of consumption. So now I'm going to give you sorry this is going to be the outline of presentation I'm going to start giving some institutional backgrounds of the marriage market or the marriage legislation in Senegal. Then I'm going to detail you the data we use provide you some descriptive statistics and then analyze differences in welfare levels before really looking at the selection into widowhood and divorce and remarriage. So the marriage market in Senegal first to give you like a broad picture of what is Senegal for a woman in 2006. So Senegal is a Muslim country it's characterized by several ethnic groups from the world of being the majority group and the patrilocal and patrilineal norms prevailing in Senegal. Households are mainly extended so on average there's 18 individuals per household but this is really an average because you can find very small and nuclear households for example in Senegal where the husband lives with the wife and in rural areas very large households with more than 30 individuals living together. And they're extending the sense also that there are several adult married adults living together that do not share a nuclear link so they're not necessary brothers and sisters and mothers and sons or daughters. Some marriage features so on average women marry at 18 with an announcement that it's at least 10 years older. During the ceremony there's some payments that are exchanged so here it's a bride price so from the groom to the wife or the family of the wife. So I'm not going to talk during this conversation about the effects of a bride price but bride prices more at stake during the first marriage and not that much during the remarriage or the second or third marriage. And women had their first child on average one year after marriage. Other type of features of the marital arrangements that we're in Senegal so there's some polygamy so one third more or less of married women are in polygamous unions a bit more and there's also non-corresidency meaning that a woman can live in a separate dwelling than her husband's one and in case of widowhood there's a leverage marriage. So leverage marriage in our sample we have half of widows that remarried that remarried within a leverage marriage and the other half that don't and so leverage marriage can be seen somehow as some kind of insurance maybe for the women because it is a guarantee that after the disease of the husband she can she can have a male protector and have some kind of social insurance but at the same time because we're going to see that leverage marriage is also correlated with a lot of more traditional and a poor background for the women. Let me see some at some kind of for the tribe where women are forced to remarry within this this poor marriage condition. So now some institutional context. So what are the rights and resources of women I'm going to start with divorce. So the rights or resources that divorce is are many going to depend on the type of marriage they contracted before a divorce obviously. So whether the marriage was recording a civil register or whether they marry under a customary law. So if they're married they were married under a civil register the divorce can be initiated by them. The judge decides on the custody of the children and the husband can be required to provide for the subsistence of the wife. Whereas if they marry under a customary law there's no available legal recourse for either party and there's mainly a very asymmetric situation between the women and the men. So the men can repudiate his wife so it's prohibited by the law but the fact that there's still some repudiation going on in Senegal. And so the women can also ask for separation but the final decision is up to other up to the family of the groom of the up to her own family etc. So they don't have a lot of voice in this process. And also children custody and child support is at her ex-husband's discretion. So yeah for divorce this is going to be one of the main variables we're going to look at. For widowhood so here the consequences of widowhood are more a link to the position the deceased husband had and the number and the composition of children she had with him. So if her husband was a serious servant she gets a monthly pension equal to one third of the husband's wage and this amount is going to be shared among co-wives in case of polygamy. If the deceased husband was in the formal sector it's at the firm. The fact that she benefits from a pension or not is going to be at the firm's will. And for the more general case so the family code which is the set of all the laws in Senegal says that the individual has to choose between two options so that what we call the general case or the Islamic of customary law. So if the wife to the general case wife's most inherited share equal to that of the children but under the Islamic and customary law which is the case the option that is the most chosen in our sample not in our sample in Senegal. Wives inherited of one eighth the total bequest and it's going to be it has to be shared among co-wives in case of polygamy and here some inherit more than daughters but in practice what happened is that wives are more often excluded from the bequests. So now the data so we're using two two main data sets so the first one is called poverty and family structure PSF hereafter. It's been collected in 2006-2007 it's nationally representative so we have about 1,088 households and so about 15,000 individuals. For our women's sample we're going to have so you know it's not written here about three thousand individual records. So this data set is very suited for our purpose because of two main reasons first is get detailed information on marital trajectories which is not always the case in the main service we can have in Sub-Saharan Africa so we have a lot of information on the circumstances of the last dissolution for example the type of occupation that the husband had the number of children the women had with the CDC's husband or the ex-husband in general the number of dissolution that they had the time of dissolution etc etc and also it records consumption at a very almost individualized level so what we call here the cell level so it records in a consumption at the subgroup level within a household so really recording groups that are budgetary independent somehow so for example a married couple with the kids okay so it really records the consumption that are only for the subgroup expenses that are shared between some other groups within the household and consumption that are really shared for all the individuals all the members in the household for example everything that is related to food so this allows us to have like a very more detailed and precise consumption or that consumption that indicates really more precisely what an individual consumes within the household then we're gonna compliment this data set with using DHS so DHS for Senegal in 2005 so it's more or less the same period and in this data set we have more information also on variables that we think are very linked to women's agency decision-making and resource constraints so here yes one of the problems that for DHS we only have women from 15 to 49 and if you want to compare PSF with DHS and the national census the rates the rates the proportion of women that are currently divorced or widow are pretty comparable between PSF and these two other data sets okay now some descriptive statistics so in PSF we have at least 18.5 women that are that ever experienced widowhood so I'm here among ever married women and 30 13% that ever experienced divorce at least so I'm saying at least because we have the information only on the last the reason of the last breakup and most women so here about one part also had another dissolution before this one so but we have no information on what happened so this is really a lower bound but then if we assume that for all widows that had more than one dissolution they also had a divorce before that we can have a upper bound of the proportion of women that ever experienced widowhood and in that case we find 21% and if we do the same exercise but for divorcee we find 17% so remarriage is very common in particular for every divorcee women because about two-thirds of every divorcee women are remarried so if we look now at our cross-sectionals the sample among every divorcee we're going to have two-thirds that are now remarried whereas for a widow we have one quarter of our ever widow women and that are now remarried so again these are still and when I'm going to talk about every divorcee and every widow is using the information of the last breakup so here again it's going to be a lower bound of this proportion. The remarriage almost always happens in a polygamous union so for a remarried divorcee there are about half of them married in a polygamous union whereas for a remarried widow this year is higher so almost three-quarter of them married in a polygamous union if we compare this with the share of first marriage women that are in a polygamous union these shares are pretty high because it's just one quarter of the first marriage women that are in a polygamous union and as I was saying in the introduction half of remarried widows are in a leviatic marriage and the other half are not and for an international comparison the percentage of widows and divorcee are similar between Senegal and other West African countries so here I'm plotting the proportion of ever widow women by age group so we see that here age is a very positive gradient of the proportion of the likelihood to be an ever-women woman so starting with almost no one being a widow to almost a universal percentage of women that are widow when you're 70 and more for divorce the peak is more around 40 okay but actually this hides the fact that the likely the the marriages that are more at risk to end up by your divorce are in the first five years of marriage so when marriage happened it happened more likely in in the first five or ten years after the first marriage there's also a very sorry a generational component of divorce so here I plot the survival function by marriage duration here the exit event is divorced and you can see that so the right the red line is for women more than 40 years old in my sample and the blue one is for those that are less than 40 so the younger you are the more likely you are to have a to experience a divorce and there's also a a geographic component so here it's the same exercise but using women that were in a rural area time of dissolution and comparing them to women that were in the urban area time of dissolution you see that urban women are more likely to to divorce okay so now I'm going to look at the differences in welfare levels so I'm going to start with the DHS data looking at more non-monetary welfare indicators and then I'm going to look at consumption levels so here I'm showing you some statistics again you is using DHS so it's women between a 15 and 49 these are person percentages proportions except for obviously age and the DHS asset index and every significance test is compared to married ones women okay so firstly we see that obviously these share are very different from those we have in PSF because of this age restriction and so we have very few ever with the women and more married divorces and three quarters of the samples obviously married ones so one thing here is that ever with the women are more likely to be household heads followed by ever divorced women and lastly there's the married ones for the widowed women so we have about one third that is household head and 20 percent that is remarried so these percentages really mirror those mirror those of the percentage of women that receive most of husband property so this actually can be a reason why they they don't remarry and when we look at the DHS asset index which is a a index that is computed on the assets of the household so the higher the the index the more rich asset rich the household in which those women are living is um you see that in urban areas uh remarried women are among the the poorest so married ones remarried widows and remarried divorces and the non-married ones are among the richest here remarried widows are the poorest group or live are living in houses that are among the poorest and this pattern is uh the same in rural areas or at least again the poorest group is or the women that live in the poorest households are the remarried widows so now using the same database I show again some of these pretty statistics using some variables so they're more in the paper but some variables that we think are linked to some to the voice that these women have within their household or to the resource constraint that they have etc so um so again it's pure correlation but still so when we look at the proportion of women that declare they have no say on or their own health care um we see that um so again married women are pretty bad and non-married women are pretty okay when we look at constraints on seeking health care we see that permission is not really an issue in synagogues costs are more a problem and um widows are the most constrained women now more more interestingly when we look at the proportion of their uh um the proportional women that declare that their own earnings are spent on household which is something very interesting so here again in Senegal income are not necessarily pulled within the household so the women's earnings so what they earn from their um the labor is usually supposed to be uh spent on for herself or or dependent on non-food items so the fact that they spend some of their earnings on the household can be seen uh like the fact that they have to contribute to the household expense so they're living in the in a poorest household and here we see that uh on average one third of women contributes uh to the household expenses with their own earnings uh but um sorry none of these women one third don't contribute at all to um their household's earnings but remarried women are more more likely to contribute and indeed they contribute uh one third contribute to more than one uh spend more than one half of their um earnings in household expenditures so again they seem to be uh in a weak position um then looking at the proportion of women that justify the beating here again uh remarried women really uh stands out so this is really um not controlled by anything it's pure correlation but it seems to say something about remarried women that's even if they um remarried they're in a weak position so this is maybe due to the fact that they were very uh um they they come from very poor backgrounds to begin with or because they because of remarriage they they are now in a weak position um so now um so yeah so this but just to say that this part are gonna be seen also in psf so we really think that marital status is uh very interesting and clear um dimension through which we can look at welfare for women so differences in consumption level now i'm gonna use um psf data so using this disaggregated uh consumption that i've uh uh tell you about so here i plug the unconditional means for the log of total cell consumption per capita so um by current status of the woman and you can we can see firstly that um married women are among the poorest and remarried widow again stands out as the more of the poorest group then i break up this consumption by women that are in the verity marriage and uh those who aren't and we see that this lower average is driven by the women that are in the verity marriage and those who aren't in the innovator marriage marriage actually have a level of consumption that is uh quite similar to married ones women and remarried divorces now to understand a bit more the correlation between these levels of consumption i regress the log of this consumption on a set of um cover is that uh we do think that um play a role on the level of this consumption for example age is square um the household composition the fact that the women went to school or not the fact that she had a son the ex husband of the current husband occupation etc then i'm gonna use uh so for so i'm gonna regress this for each group and i'm gonna use um the parameters of this regression and i'm gonna try to predict the consumption of a certain group of women if they would have been in the another situation okay so trying to find the counterfactual the factual consumption of each group in another in another marital status um okay so here the results so first these are the results of the ols regression for each group of the consumption um so here i just want to point out like the high returns of education for every group um then the fact that having a husband uh working in the public sector is also highly positively associated with a high level of consumption as well as the fact that these women live in urban residents so all these coefficients are not significantly different from another so this one is from the others but those aren't and also the fact that okay for a divorces you find a very positive effect of having a son 18 or older then i'm gonna use these parameters and i'm gonna choose so i'm gonna break up the sample by uh age group i'm gonna find uh so i'm gonna start with a reference group here is a widow i'm gonna find the average uh the the average uh women in this category group so i'm gonna give uh for the covariates the mean value the mean value of this group and i'm gonna use the parameters of the regression to have a sense of how though this average woman would have um been what would have been her consumption in another group okay so here so i start with the reference group is widow so here are the different um age groups and we find that widows if they weren't if they were remarried uh they would have consumed less so actually they are better off in their own uh current situation and this is true for urban and for rural areas okay so everything is not significant but we see that mostly um either the coefficients are pretty low or uh they're not um or they're negative whereas whereas if we look to remarried women we see that given their characteristics uh and using the parameters of the other groups we find that they would have consumed more in the other groups okay so there aren't there aren't in the better situation um so there's again a lot of um plausible um interpretations for this or maybe because okay remarriage is a bad thing for them or maybe because they were already um in a bad and they come from poorer backgrounds and they have no say on whether to remarry or not and they still marry in very um poor poor households so again this is true for urban areas and rural areas um then doing again the same exercise but with married ones women we see that these women are really better off in their situation and married to this solution is almost always associated with a negative effect on conception we did the same thing with uh using as a reference uh divorce currently divorce and uh remarried divorce and we find that there's no effect so for um for divorce at least it seems that women really select into um the status that is better for them uh now we replicate the exercise but instead of looking at age group at current age group we look at um uh groups of age at at the time of this solution but we find again the same pattern so we know that this is really irrespective of uh any uh unobservables and actually I have to say here that this is probably due what I was saying like with where it is like due to these unobservables that this woman come probably with the of more traditional backgrounds and again if I go back if you look at the constant here it's pretty low for married women so this is can be um we can signal the fact that there's probably a lot of um unobservables that we're not controlling um here and that are negatively linked with the level of consumption so yes I said that uh same exercise so for age of dissolution sorry and yeah and we did the same exercise with duration since dissolution and the same pattern supply so here widows that are not remarried are better off than the situation where they would have uh remarried um whereas remarried widows would have been better uh not remarrying again conditional on the average um levels of the covariates and irrespective of unobservables so okay so now let's see what are the selection that is at play and why those women like are now in this situation why they remarry not or at least what are the covariates that are um uh linked with this position so I'm going to start first with the selection into widows and divorce then look at remarriage and then give you some uh idea of what is correlated with quality uh the quality of remarriage so here it's uh I'm showing the marginal effects of a logit model on the probability of uh being a widow or in the southern columbine a divorcee so we see that um women that are from more um lower backgrounds so that had the per disease husband was um not uh in the private formal sector are more likely to be widowed and also though it's not very significant women that are were in french attend at least once a french school are um less likely to be a widow whereas for divorcee what is more striking is that there's a positive selection of um educated women into divorce so women with uh say good attributes so women that were more uh educated and are more in urban areas are more likely to divorce and also women that uh had a husband with bad bad attributes here a husband that were not in the private formal or public sector are more likely um to um uh not to divorce so women with bad husband attributes are more likely to divorce so say the selection here is driven with good attributes whereas for widow it is like more bad attributes like in a broad sense uh for remarriage so i'm going to start with the the divorces um we saw that the more educated you are the more likely you are to uh remarry whereas here sorry to be a divorcee whereas here education plays a negative role so when you're educated you divorce and you're more likely not to remarry um and for widows and also for we divorcee we see that the priority of remarry remarry is clearly linked to correlates that are um proxies for a traditional for women living in traditional settings for example living in rural areas having a polygamist father have been fostered before age uh 14 and what is the most important um um determinant of the marriage is obviously the the age the younger you are the more likely you are to remarry oh sorry i forgot to say um then for divorce says uh the fact to have a son a termitive solution to have a son when um uh when the woman divorced is negatively correlated with the priority of divorce whether it is because it's more difficult to have the custody of uh of the son while being remarried uh or because as we saw from the previous regression um having a son at time of divorce was positively linked to a level of um for the level of consumption for widows having a um what is it having um children at time of dissolution so the complement of this coefficient is positively linked to remarry most certainly because um having children helps the fact to have a lyricic marriage whereas not having children with a deceased husband is uh more difficult for them to be in a lyricic marriage and actually if we plot the um priority to be in a lyricic marriage um in comparison with a non-library marriage we see that having a son at time of widow who is closely linked uh with it whereas here um education thanks protects them from uh being in a lyricic marriage and uh so helps them to be non-library in a non-library marriage and as we saw um women in this non-library marriage are in a richest uh consumer or have um higher levels of consumption now um here i'm going to try to give you some sense of the quality so of uh of the marriage and how it links with the uh status of the women so when we look at the correlation between dhs data and um marital characteristics and women's autonomy plus with all the qualitative interviews that we did and um all the literature we can expect that a good marriage is firstly a monogamous one one where the women don't coincide with the in-laws or with the husband and um in addition we can expect that the following um variables supposedly link with the quality of marriage so having a civil marriage because it gives more rights and results for women having a husband working in the formal sector for all the pension that this can provide to the women in case of widowhood plus the level of consumption that is just linked to it and um the possibility to live with the the children's of the previous union so as a benchmark here i use of women that are in the first marriage and i look how these characteristics uh correlates with other uh covariates so here i'm just going to stress out the positive role of of uh having attending friend's school in the quality so for first marriage women um and also sorry the role of being in a urban area so for women that attend school and living in uh and uh and living in urban areas they are more likely to uh don't coincide with the husband to be less polygamous uh and to have more a husband that works in the formal sector and this also apply to the probability to have a a civil marriage and not to live with in-laws so we're expecting that this we're going to see want to yeah analyze the fact uh that probably this uh correlate these variables also correlate in the same sense for um remarried widows and remarried divorces and actually what we see is that so here you have again the different types of uh remarriage and the sub samples for widows and divorces um so i'm going to start with divorces so we saw firstly that education it was positively linked with the priority of divorce and negatively for a remarriage but uh when we look at those divorces that remarried having uh been to friend's school at least prevent them to be uh in a polygamous union and it's also closely linked to the priority to have a formal husband working in the formal sector and finally uh to have a husband to have a civil marriage registered whereas for women widows sorry this education doesn't play a role on at least the likelihood to be in a polygamous union so does not prevent them to be in a polygamous union so i don't have much time so i'm just going to show you here the priority of uh what we call here upward mobility so okay it's a very small sample but we're taking all these remarried women we're selecting those who had a previous uh husband that was uh working in the informal sector or the agricultural sector and we look at the priority that the new husband is now working in the formal sector what we see is that here again the two variables that are really playing a role it's education and living in a urban a rural area urban area sorry which are like positively linked to the priority to have some kind of upward mobility and there's no difference between um married widows and married divorces so now i'm going to conclude so divorce and widower that associates very different consequences in terms of welfare for divorce it seems to be uh a way to for women to gain a relative autonomy so current divorces are the richest group in our sample and education as i just said today is a very um positive role because divorces are likely to be more educated women their education is negatively correlated with marriage and even among remarried divorces correlated with very equality unions and for widowhood we have a more negative consequences that are not negated mitigated by a remarried first a double negative selection so poor women are more likely to experience widowhood um and most likely those who have to remarry are the most vulnerable groups so at least we saw that they came from more traditional families um and uh yeah and there's also something related with the role of lairviet marriage either because it's can be seen as an insurance for those women because they have this at least this social insurance uh provided by the fact that they can remarry the the this the the brother of some relative this is husband but at the same time maybe this is what uh this is some kind of poverty should poverty trap because we saw that this lairviet marriage also correlated with lower levels of uh at least of uh consumption thank you