 early pregnancies in low and middle-income countries. And my name is Vincent Somville. I'm writing this chapter together with L'Archivar Berger, Chetil Burvat, and Amina Mohammed Mali, and Bertil Tungodan, who are all based in Norway. So just some words about the issue. First of all, early pregnancies are much more prevalent in low-income countries than in high-income countries. This is an important reminder. So according to the latest number from the World Bank, we have 13 births in 1,000 women in the age of 15, 19 in high-income countries. But it's seven times higher in low-income countries. So much, much more prevalent in low-income countries than in high-income countries. And why is it problematic? There are different consequences. First of all, early pregnancies are associated with negative effects on women and child health. So this was also mentioned by Siwan Anderson this morning. They also have negative consequences on economic aspects. So young women who have a child are less likely to complete school. They will have lower education. They will have lower wages and lower job opportunities. So you probably know all of this. So what I'm going to do is this chapter is dealt in two parts. In the first part, we want to review existing research. What do we know? What have we learned about what kind of interventions are successful in reducing early pregnancies in different contexts? In the second part, so one of the conclusions from that review will be that we actually know little about the mechanisms through which different interventions affect or don't affect early pregnancies. And then we want in the second part to use a current project of ours to dig more into the mechanisms and provide evidence of what mechanisms are actually at play when you have different types of interventions. And then we will conclude. And the conclusion from this exercise will be that women need jobs probably more than condoms. This may be obvious to most of you. We think it's not obvious necessarily to the rest of the world, and it's worth writing it down. So let's start with the literature review. First of all, the scope of the review. So we will only be looking at research about interventions, policies, programs, whatever you call them, that affect pregnancies or sexual behavior of young women. They're typically in the age range of 15 to 20, sometimes a little bit younger, sometimes a little bit older. But in most papers, it will be that range. And we will only be looking at low and a few middle income countries. And we're restricting ourselves to papers published in this century. So this means that the interventions of the studies actually started much earlier. So in fact, we will cover the past 20 years of policies. We think that we have some contributions in the way we look and the way we read the papers. Because the classical way these issues have been reviewed. And the last review is a co-claimed systematic review. So it's a very medical approach, where the reviewers usually look at the divided papers between demand-side interventions on the one hand, supply-side interventions on the other hand. That means interventions that are targeting the patients, customers, and interventions that are targeting the health providers. So for example, training of nurses would be supply-side interventions. So sexual and reproductive health information is a demand-side intervention. And they generally conclude that you need both. At the same time, and doing just one type of intervention is not as efficient, yes. We have a very different approach. We want to think in economic terms. So the way we structure the review is to say that people have defined preferences and beliefs. And they have different opportunities. So in our minds, the girls that are studied, they face a set of different opportunities that can be very wide or very restricted. And then they have different preferences and beliefs about each of those opportunities. And then those preferences and beliefs together with the constraints that they face on this set of possibilities will determine what they actually turn up doing. This is a very different way of looking at the literature. And it leads us to different conclusions compared to what the medical doctors have been doing in the past. So it will be clearer in a few minutes. So we group together preferences and beliefs. It's not always clear how to differentiate the two in the literature with what we call mindsets. So we will have first interventions that affect mindsets. And in a separate part of the review, what are the interventions that affect opportunities or constraints? So one example of intervention that affects the mindset is information about HIV prevalence by partner H. That's a paper of Pascal Indipa that you probably know. She says, no, if you have sex with someone who is 19 years old, this is a probability that you will be HIV infected. If you're not protected, if you have sex with a man who is 45 years old, this is a probability that you will be HIV infected. So that does not change the set of possibilities that the girl is facing. But it might change the belief that they have about the consequences of their actions. An opportunity intervention would be, for example, what is described in the paper by Jensen, providing links between to linking young women in rural villages in India to jobs in the outsourcing industry, links that did not exist before, without giving any information about sexual reproductive health. But this is increasing the opportunity set. It's giving them an other alternative to motherhood, which is moving and starting a wage-paid job in the outsourcing industry, right? I think in these cases, the differences are quite clear. It's not always the case. Sometimes we've been also internally arguing, should this paper be a mindset intervention? Should it be an opportunity intervention? It's not always obvious. But we still think that it's a useful way of thinking about the different interventions. So let's go about the mindset interventions. They will be typically, and this is the most classical approach to deal with this issue, typically sexual reproductive health information provided in schools. And sometimes, but not always, accompanied by training of nurses or supply of contraceptives. Now, what mechanisms do we expect in these cases? There is a first direct effect of the information because the girls will learn how to avoid STDs or how to avoid pregnancy if they didn't know how to. And in some cases, there will also be some learning about how to say no to sex if you don't want to have sex or how to negotiate sex. So this information could have a direct effect on the outcome of the relationship. Then there might be also some effects on the beliefs. The example that I gave earlier was the probability to get infected if you have sex with such or such person. And in addition, there is an effect of access because it's usually also accompanied by improved access to contraceptives. So you reduce the cost of safe sex. Now, all those different channels should, in theory, reduce early pregnancy rates. The implicit assumptions in this type of program is that early pregnancy is not the best option. And that girls have clear, well-defined, better alternatives. And if they get pregnant, it's because they're not able to do otherwise. So this kind of program does not open for the possibility that maybe they don't have a clear, better alternative. So I don't want to review each paper one by one, but I try to give you a summary. There are almost all sexual, reproductive health information are quite standard with improvement in health services. So that is either providing youth-friendly health services locally or subsidizing the distribution of contraceptives, condoms, in particular. There are two studies that are quite different. One is Cabezon, what is here? That was just an abstinence-only information campaign in Chile. And DUPAT 2011 does HIV risk information. I will come back to it. And then I want to see if those interventions typically have, first, do they change the knowledge and the attitudes of the girls? If they change the knowledge and the attitudes, then the next step is, do they change the behavior? Do the girls are more likely to protect themselves? Or do they have two or partner, for example? And then if they change the behavior, do they change the end outcome, which is the early pregnancy? So what we find is that, in general, this kind of campaign is very successful in changing knowledge and attitudes. Attitudes, I mean, do you think it's OK to use a condom when you have sex? Or are you likely to use condoms in the future, for example? So if you have a program where you teach the girls how to use condoms and then you ask them if they know how to use condoms, usually they will know. But then very few studies actually found an impact on behavior. So reported use of contraceptive is usually one measure, number of partners, age of partners, and so on. So green is, I don't know if you can read, green is a positive change, red is a negative change, orange is when the results are a little bit unclear. And then so very few changes in behaviors. What is really surprising is that we usually don't know if there is an effect on pregnancies. So I would assume that if there is no effect on sexual behavior, there is presumably no effect on pregnancies, but I don't know. And the reason it's not reported is that in most studies this was not the primary objective because they were designed towards HIV infection. So they're most interested in condom use, but in pregnancies themselves. Yes, so we have two studies, but we do have two studies that have a strong effect on reduction in pregnancies. So summary, knowledge generally improves. Behavior is hard to change. Pregnancies usually not reported. We find decreased pregnancies in two studies, which is the abstinence campaign in Chile from Cabezan and the do-flow paper that I will come back to. There are strong limitations to those studies. So first, as I said, we usually don't know if there is an effect on pregnancies. I have a very small number of observations. This is 20-something studies from a few countries in specific contexts. It's different interventions, different study design, different measurements. So it's hard for me to conclude, but thank you. And then I think the measurement is problematic, especially one of the most frequent measures is condom use. Of course, it's not observed. It's self-reported for obvious reasons. So if you spend time in a campaign repeating to young girls that they should use condom when they have sex, and then a few weeks later you go back and you ask, did you have a condom the last time you had sex? They will say yes, most likely. So I think this is problematic. And my own understanding of those studies is that whatever we see as an estimate should be an upper bound on the true effect. OK, not the second type of interventions. Not interventions that affect mindsets, but interventions that affect opportunities, that change the economic opportunities of the moment. So most of these, except a few that I will mention in particular, but most of these are one unconditional cash transfers. And then there will be some that have direct effect on economic opportunities. And the mechanisms in the cash transfers, there are multiple. So the cash transfers are always defined by an eligibility criteria. So what also could have the transfer? It's usually poverty measure with geographical locations. But there is one case in Honduras where the criteria was that the household had to include a pregnant woman to receive the transfer or to have a child below 3 years old. Guess what happened? There is always a condition also under which you stop receiving the transfer. And you typically have two types of condition. One is linked to schooling. So the kids have to be enrolled and to have a minimum attendance level. There is also in most programs health services condition, which is that the woman in the household will have to attend a certain number of health services. Those two conditions could also have an effect on pregnancy rates. The first one because the transfer conditional on schooling gives you an incentive to stay in school and so not to get pregnant. The health checkups could also be an opportunity to provide health, sexual and reproductive health information. And so it could also have a direct effect on pregnancy. And then you have the transfer itself, which will have an income effect that could reduce the need for transactional sex, for example, or that could be used for investments or development of the family business that might also have an effect in improving economic opportunities and reducing pregnancy. And then you have the intervention that affect directly the opportunities. These are the papers that we have in the chapter at the moment. If you know more, I'm happy to include more. So there are all conditional cash transfers in Honduras, Nicaragua, Mexico, Mexico against the same opportune data in Malawi, in Pakistan, in Malawi again, but this is a comparison of conditional and unconditional cash transfer. I will come back to it. I think it's a very nice paper. This one, the WAC 2012, is a conditional cash transfer in Tanzania that is very specific because they had a group of people who would get a payment if they tested negative to STD tests. So it's not at all the schooling, health services, traditional cash transfer. It's really targeted at STDs. And they just test people regularly. And if people test positive, they don't get the money. If they test negative, they get the money. And then we have Janssen, that is a different type of approach that was really successful linking rural villages in India to also seeing industry and seeing what impact it has on pregnancy rates. Now, in all the cases where it was reported, those interventions had a positive effect, the effect of reducing risky sexual relationships. They also had the effects of reducing pregnancies, except in the case of Honduras, where we see an increase in fertility among the households that receive the transfer. And the other thing, that is because of this eligibility criteria that I mentioned. So again, it's really difficult to conclude from a small set of studies, which are many differences. They are different in many dimensions, those different programs. But still, from this table and already is that the interventions that are creating different economic opportunities have been much more successful in reducing early pregnancies compared to the classical sexual reproductive health information campaigns. Then I want to mention in more details a few selected studies. So personal preferences, maybe. I think these are the most interesting. The first one is Pascal Indupai in Kenya. What she did is in one group, she gives the standard abstinence curriculum. In a different group, she gives information about the risk of getting HIV if you have sex with men of different age. And she finds that in the risk treatment, you have a 20% decrease in teenage pregnancies. The abstinence course had no impact on teenage pregnancies. And what she observes is that the girls, in fact, in the risk treatment switch, did not have less sex. But they switched from risky sex with older male towards sex with people of the same age and protected sex. So I think what this shows, first of all, is because I don't want to say that information is always inefficient. Information can be extremely efficient, as is the case here. But it's a very specific type of information. Second, it also shows that young women have a choice. Some people say they have agency. They have this very specific information about the risk of getting infected when you have sex with different type of people. And they directly reacted to that and changed their behaviors. So the sexual behavior is not completely determined by external factors. Then a few words about Bayard Osler McIntosh 2011, comparison of conditional cash transfers and unconditional cash transfers. The condition was on attendance at school. And they find that the unconditional cash transfers reduced pregnancies by 44% more than compared to the control group. But the conditional cash transfer had no effect. In fact, they described that the effect comes from the girls that are out of schools. So they have a group of girls from the unconditional cash transfer treatment that dropped out of schools. But because it was unconditional, they keep receiving the payment. And that allowed them not to have sex and not to become pregnant because they had an economic alternative and a way to live. In fact, as they mentioned in the paper, 25% of the young women who are sexually active at baseline reported that they started their sexual relationships because they needed assistance or because they wanted gifts and money. So when you receive the unconditional transfer instead, you have an alternative to sex. And then Janssen 2012, very different type of interventions. Three years of recruiting services to help young women in Indian villages to get jobs in the business outsourcing industry. The probability to have a child has reduced from 43% to 37%. So much bigger effect than what you see in sexual and reproductive health campaign from an intervention that is not at all related to sexuality in the first place. But that provided clear economic opportunities. And then you have actually two papers. I would love it if you know more, that explicitly combine mindset interventions and opportunity interventions. First one is Duflo-Dupac-Rameur a year, two years ago. They compare. They have three arms. One is an education subsidy and abstinence curriculum. And both combines compared to peer control. The education subsidy reduced the teen pregnancy rate from 16% to 13%. The abstinence education had no impact at all. When both were combined, the fertility fell less than when the education subsidy was combined, was provided alone. So that was the puzzle. What they say or their interpretation is that the abstinence education and the health information treatment really insisted on you shouldn't have sex before marriage. And the effect of that was that the girl got married earlier and got pregnant. And that contracted the effect of the education subsidy. So this is to say that I wanted to balance a little bit because I gave you the example of Dupac 2011, a very positive effect of the information on different risk of different age of the partners. But this one, you give additional information campaign that completely contracts the effect of the opportunity treatment. The other one that I want to mention is I don't think it's published yet, but it's a paper by Irina Bandiera, Imran Razzoul, Selim, and Marcus while in the room, which looks at the introduction of black clubs in Uganda. And in those clubs, the girls received vocational training combined with sexual reproductive health information. And they also have a meeting space where they're safe to meet and discuss. And that intervention had a really big impact. And first of all, in raising incomes and expenditures, but also in reducing teen pregnancies by 26% and reducing also entrance into marriage or cohabitations. And the share of girls supporting sex against their will also drops from 14% to around 7%. So this is an example of another intervention that is really successful, but that is again combining that is not just giving mindset opportunities, not just changing preferences and beliefs, but combining this together with vocational training that is supposed to increase the economic opportunities. OK. So the case that I'm trying to make, if it was not clear yet, I want to develop a bit further with, sorry, before last summary of all those papers together, I put on the horizontal axis the years of the intervention that is today's. On the vertical axis, whether they reduce or find a reduced or delay in childbearing, whether there is an impact in other sexual behavior, an impact on knowledge only, no impact, or increasing childbearing. And again, I hate to conclude on such a small sample, but it really seems like you have a group of studies here where you have no impact or impact on knowledge only, which are all sexual and reproductive health information campaigns. And then you have up here a group of studies that have impact on sexual behaviors and teenage pregnancies and that are mostly cash transfers or the combinations of providing economic opportunities with sexual and reproductive health information campaigns. Small sample. So I don't want to push it too far, but I think this is suggestive. No, we haven't learned a lot about the mechanisms. I've been saying this economic opportunity, they change the way girls think about the future. They give alternatives to motherhood, and that's what is driving the change. Is it the case? Is it, can we say that when you give sexual and reproductive health treatment, what kind of effect does it have on the beliefs and the preferences of the girls? When you give business training or vocational training, does it really change the set of opportunities that they see in the future? That's what we want to test. We want to use the current project to try to say more about the plausible mechanisms behind those findings. So we want to see if this mechanism is at play. Is it true that different interventions will have different effects on the extent to which the girls feel in control of their bodies and of their lives, and the extent to which they see alternatives to motherhood and better economic opportunities? Is this something we can test directly? To do that, we want to use the current project of ours in Tanzania that we short name Girl Power, which has four harms. The first is a more or less standard. It's implemented by a very feminist NGO called Femina Hip. The first harm is sexual and reproductive health information campaigns in schools. The other one is a business training, and then both combined. And then we have a pure control group. It's a cluster around the main striers, 20 schools per arm, around 3,500 girls in total in four regions, central regions of Tanzania. So that's more Gorododoma, Taborans, and Ginda. Almost all of the girls are between 16 and 18. Not the youngest, but yes, one thing that is particular to this case is that they're all girls attending secondary schools, and they're at the end of form 4. So they're at the end of their secondary school curriculum, and the interventions happens a few months before they finish school. And the idea was that most of those girls anyway will not have the grades to continue education. They have very limited job opportunities. So most likely what they're going to do after school is either to become a mother or to start small business self-employment activities. So we thought it would be really interesting to work with exactly that group if you target a specific age just before they get out of schools and into the labor market or marriage market, will you have a bigger impact? So the baseline, the treatment, and the first survey that we use in this chapter were done in 2013. We had another survey in 2014, and we are currently doing the last data collection. It's not completed. A brief description of the sample. We have on average six members per household. We surveyed the girls. 20% of the households are evicted by women. Half of the heads of the households are farmers, and the rest are either small business owners or public servants. At baseline, the vast majority of the girls said that they are sexually active. 80% agreed that it's common for girls of their age to receive money or gifts for having sex with older men. 60% declared that sexual harassment is also common. And they also know, like the first, when we ask them if they have friends or if they know girls that dropped out of schools, what was the main reason? It's typically pregnancy. It's the main reason why their friends are not in schools anymore. And they have a very limited sexual reproductive health knowledge. So for example, so we asked different questions about sexuality. For example, can you be pregnant during your first intercourse? We had almost 40% who said no, that's not possible. What we want to use is this material, which we think is really rich, but really hard to use. A few weeks after the treatment, we asked all of the girls to write a short essay about where they will be in five years. What do they want to do with their lives? And what are the big obstacles that they face? And then they had very few guidelines. So we're just blank page, write down where you see yourself in five years and what do you want to do? And what we want to see here is do the different treatments, the standard sexual reproductive health treatment, the business treatment and both combined, did they have different effects on how the girls think about their futures and think about different scenario? We love the material. We think it's really rich. We also struggle a lot with what do we do with this? So what we are doing at the moment is two complementary analysis. The first one is an in-depth, we took a random sample of 192 of those essays from the different treatment arms and we do an in-depth reading and categorization of the essays. So we had, that was actually the work of Amina, who's one of the co-authors, who's fluent in Swahili, she's from Kenya, who did this work of reading all the essays and really teasing out what the girls are saying in each of those 192 essays. And then of course we couldn't, well we did ask her to do that with the rest, but she declined. So we had to find a way to also analyze the rest of the essays and then we use text analysis algorithm. So we put all the essays into the computer and then we rewrite an algorithm that will pick up the words in the different essays and then we try to, we don't have, we know what are the main themes of the essays or what they're talking about. Of course the problem with the second one is that it's nice because it's large scale, it's automatic, so it's really fast, but you don't know if a girl says I really want to get pregnant, it will be coded as if she said I really don't want to get pregnant. It's really difficult to have a little bit of nuance in this, but that's exactly what we do in the in-depth reading. So the objective is to combine the two and see if we have consistent findings. And we also use specific words, I can go through the list if we have time later exactly to avoid this kind of a positive, negative discussion for example. For example, if we like to pick up words like profits, sales, customers, because we think that the girls, if they're talking about customers and sales and profits, they're very unlikely to be saying I really don't want sales, I don't like profits that I will avoid customers. So we believe we can plausibly interpret it as a positive discourse about business practices. Here is one example from the control group. My first priority in life is to educate myself and then I will use my education to prevent cruel practices that all the girls are subjected to. It was very difficult for my parents to pay my fees or buy school uniform. I have many problems that bother me. I'm completing from four and to date my parents have not paid the school fees. Sometimes when I go back home, I often cry. I would like to say that I have many aspirations in life however I feel that I will not be successful due to the difficulties I'm encountering. I do not have much more to say but I do have one worry. I would like you to give me advice. Once I finish school, what should I do in order to avoid the pressures of unwanted pregnancy and the expectation of early marriage? And this is the control group. So we had no discussion with her at all about pregnancy or sex or it just came out. That's what she identifies as the main worry. I would like to claim that the quotes are representative of the rest but I don't know. Here is a quote from the sexual reproductive health group. I have received training on fertility and health and I've been told how to say no when a man asks you to have sex with him. You should have a firm stand in saying no and not fear him and you should start to bite, dig your feet into the ground. You have to stare at him in the face and be serious. All the girls should receive education on how to say no. So we believe this, you know, she picked up something, she has some clear advice on how do you deal with a man who wants to have sex when you don't want to and we think that might change, that will change her beliefs about what are the consequences of interaction with the child. Here is a quote from the business group. I'm a 17-year-old girl who studies at such secondary school. I received training on entrepreneurship that enabled me to understand the opportunities available to me after the training. I became more informed on how to start my own business that would make me work hard for my future. The next five years, I will have a huge business. I will bring, that will bring me income. After completion of my education, my life will be based in Morocco. I do not expect to have children during this time until when I'm employed or running my own business that will bring income and independence from anyone else. So that's what she picked from the business training. I thought that was interesting because in this business training, there is no mention of pregnancy, children, sexuality at all. It's completely focused on how do you keep the books? How do you deal with suppliers and customers? How do you calculate your profits and what kind of savings you should have? How do you get a loan from microfinance? But she immediately makes the links with the pregnancy. She's saying, no, I want to do my business. That means I'm not getting pregnant in the years to come because I first have to set this up. So then we try to, the example I gave in those course, we try to measure them quantitatively. So both in the full sample of essays using the text, the word count and algorithm and in the in-depth sample, we categorize the essays whether or in this case, the dependent variable is whether they have a locus of control. So in the essay, do they show, are they fatalistic or do they have discourse where they say that they are really in control of their bodies and in control of their lives? They will use words like, I'm confident, I'm capable, I can do this completely opposite to the first quote that I gave from the control group. And so what we see is that the three treatments, so the combined treatment, health, sexual reproductive health information and entrepreneurship, the three of them increase in a similar manner, the percentage of essays that show a strong control, a strong locus of control. Then we do the same with a different dependent variable which is whether they talk about sexuality. And there we see that another effect of those treatments, but only the health treatment, not at all the business treatment, is that they talk a lot more about sex and maternity and love. Then we move to the economic opportunity side. Do we see in the essays that the girls describe economic opportunities and an alternative to motherhood? So we code a dependent variable that's equal to one if it's clear in the essay that they describe another way to then not motherhood, but that they have other economic opportunities. And what we see there in both, again in both analysis, is that the effect is much stronger in the business treatment, in the combined treatment, but there is no effect from the health treatment. So it's, maybe slightly you would expect, but we're checking. So this one was quite general. It's whether they mention, so they mentioned what they want to do. Do I want to become a mother or do I want to start a business and get a job? Then we said, we thought, and this was actually a suggestion from Marcus, you have to be a much more specific. So here we code as one only if the essay mentions very specific activities, like not just I want to have a business, but I want to have a business selling vegetables and for this I will contact that bank that will make me a loan and I will go and sell in that region. So if they make very specific plans, we think they're more credible. And then we get starker differences between the treatments. And again, we show that the business treatment and the combined treatment have a strong positive effect on this variable compared to the control and compared to the health information. So this is the summary of what we do. What we conclude is that the sexual and reproductive health treatment has been, is very successful in inducing a higher locus of control. The girls after this treatment indeed show that they say that they are in control of their lives and that they are capable. They're also more likely to discuss sexual issues, but they don't show any clear alternative to motherhood. So they don't, when you ask them, where will you be in five years? Those health treated girls are not at all describing something else than being a mother. On the other hand, when you do the business treatment and the combined treatment, health and business, you have also the same increase in the locus of control, but you also have, then they also start to make clear plans, alternative plans to motherhood and clear plans about how to develop an economic activity. So we think that this evidence is useful in understanding the previous papers that we have discussed, because it really goes into the mechanisms in documenting how the two type of interventions that we have discussed before may affect exactly what the girls plans for the future. And we think that the picture is globally consistent. And I will stop there if you have questions.