 I'm from now working at the International Fund for Agriculture and Development. This work is the result from my work at the University of Sussex with Julie Litchfied, who is also here and presented earlier today. And I've worked together with Julie for a long time in the Migrating Out of Poverty project. We also have partners here based in Ghana, and this is the result of data collection conducted by our partners here. And an analysis we attempted using this data. So the title changed a little bit because in exploring the data further and progressing in the analysis of what we had in mind, this is kind of what the focus of the paper is now. It's called the Nature and Impact of Repeated Migration within Households in Rural Ghana. And the motivation comes from, basically very broadly speaking, from the fact that there's still a gap in the literature of answering the question, what impact migration has on the left behind households. There is a lot of research done on that, but it has a lot of different results due to, on the one hand, differences in the context that we are in, because of course if you look at rural area in Ghana, you will find a different result than in a rural area in Vietnam and at different points in time. Additionally, a lot of the differences come also from the fact that different data is used, often only cross-sectional data. So there's issues of indigeneity that the researchers address with different instrumental variables or other strategies that can of course then also lead to differences in the result, but furthermore in the theory of the new economics of labor migration, it is also not possible to predict whether there is a positive or negative impact of sending a migrant from the household on the household left behind due to the counteracting factors that you experience, because on the one hand you have a member who might have generated income in this household who is now leaving this household. So you're losing an income source, let's say, but at the same time this migrant is expected to send remittances while also the migration is incurring costs. So of course you cannot, this migrant member is not moving without having to pay for this move or to, yeah, it would take time to actually find a job. So all these factors make it a very, very complex question and we are trying to contribute to this literature by paying a bit more attention to the fact that also within the same household you actually have a very diverse pattern of migration experience. So there's often more than one migrant within a household, at least for our data, we can see that and I'm pretty sure also in other data you would find this. And these migrants appear to move for various reasons and not all members within the same household move for exactly the same reason. And so we are kind of in the context of the new economics of labor migration framework when we think about households that have a migrant, but we want to allow a little bit for this diversity of the nature of migration. And then our other contribution is that we are using the first migration focus panel study for the Ghanaian context where a lot of the data in Ghana is coming from the Ghana, the GLSS, which of course has the advantage of national coverage, like national representative, our study is very focused, but we gain a lot of insights from this study. So the specific questions we are answering in the papers, first we are trying to give a very rich overview descriptively over the characteristics of these new migrants, of the kind of second round migrants compared to the first round migrants and the households that have such repeated migrants to draw some conclusions or hypotheses that we can make what could be going on in these households. And the second part of the paper focuses on trying to make an impact assessment whether having such an additional migrant has any impact on the welfare of the household left behind, conditional on already being engaging in migration. So the data that I talked about that we are using is a household panel study that was collected in 2013 and again in 2015, always around March, April, so there shouldn't be any issue of different seasons in five regions of Ghana, which I will show in a second, and it was collected by the Center for Migration Studies here, based here at this university, funded by DFID, and really the big advantage of this data set is that it has a strong focus on migrant households, so it specifically oversampled households to increase the number of observations of households that have migrants, because if you look at national representative surveys, they will always be a relatively small end of households with migrants, making it a bit more difficult to continue the analysis. And additionally, this household, the survey is very rich in covering questions, any questions around the migration decision, how migration was financed, what the migrant was doing before, they left what they are doing now, the destination, about remittances, remittance receipt, the use of remittances, and also about return migrants, so that's really kind of the selling point of this data. So those were the regions that we covered, and the regions were chosen based on the most recent census data, which showed that these regions are those that are migrant sending regions in Ghana. So the blue ones are the ones very related to this. So just to give you a better insight in what we mean by saying a new migrant and migration experience within households, so imagine this example household where you have five members, ABCDE, and at the baseline period in 2013, member A and member D were away as migrant members. So the interview was conducted with the household head, or the person appointed to be eligible to answer the questions, and said that person A and person B are currently away as migrants. And the other three people with the zeroes, they were present in this household at that time of the interview and the survey, and they are household members present here in this household. In the follow-up survey, when we went again, we observed then that household member A is still away as a current migrant, household member B is still present, household member C is back, household member D has returned, is not anymore away, and household member E is away as a migrant. So we now consider in our context, and this is a household that has migration experience at baseline, because at baseline, there's at least one migrant currently away. And in the follow-up, this household has a new migrant. So even though the total number of migrants is still the same, another member actually moved. And in the meantime, someone else returned. And these are all factors that could have different effects. So if we would just, there are some studies that look just at the change in the number of migrants, but they might miss what happens that you have a return migrant, and the fact that there is someone new moved, so that could have incurred another additional costs to send this migrant, but also it could reduce the income, because this person used to work in the household. The return migrant, on the other hand, could bring back money and knowledge, or maybe it was a failed migration, and this person now puts another burden. So all these intervening factors, we're trying to somehow account for that in this context. So our sample is restricted to only households who have a migrant at the baseline. So they all have migration experience. And then we're going to look at those that have a new migrant compared to those that don't have a new migrant in the second wave. So from the descriptive analysis, what we find first when we look at the households that have such a new migrant, compared to those households that have no new migrant, we find that they are, first of all, they are much larger on average by two household members, which makes a lot of sense that you think, well, to send another migrant, you need to have the amount of people at home to be able to send another person. So it has something to do with the household structure. They are mostly family farmers, so the majority of them against most of the income from farming of their own farm, and relatively more of the migrants of these households actually stay to have a job at the destination to compare to those without a new migrant, and relatively more of them also have a return migrant. So it seems that migration in this household is a very common pattern. And then we look at the individuals themselves to move the new migrants compared to the migrants at baseline. So this is a bit more about, okay, is there maybe like a hierarchy or what happens when you move after someone else in your family already has moved in the past? They seem to come relatively more from a younger generation, meaning the child of the household head, but also more the grandchildren or the nieces, compared to the baseline migrants who are relatively more often the household head or the spouse of the household head, or the sibling of the household head, or maybe also the child. So there seems to be some sort of generational hierarchy as well. Relatively more of them have been in educational unpaid work before moving compared to the baseline migrants, often a lot have been working, and they move for work, they move for education and for marriage and for family reunion or joining other family members, and relatively more of them move for education, for marriage and for other family reasons, compared to the baseline migrants whose majority was really moving forward. And what is also very striking is that they move for the costs that they have to pay for moving are relatively lower. So here in this table what you can see is they're comparing those new migrants to the baseline migrants, what they state, what their migration costed, and then we also look at where we can already see that those new migrants who moved for the very first time pay only 160,000, almost half, pretty much half of what the baseline migrants paid, and those migrants who have actually even moved in the past before and who returned and who moved now again, they pay even less. So we can clearly see there seems to be some pattern of experience, migration becomes cheaper with this experience on the one hand on the individual level, but also within the household. So households with primary experience sending another migrant, they seem to also save, yeah, have some sort of information advantage that reduces the migration cost. And overall for all migrants that we have in the sample, and most the majority of them are permanent migrants, we also see some seasonal migrants, and the majority of them finances migrations with savings. So we are in a setting where the households are credit constraints, they need to rely on their savings, they can, there's barely any household who claims, who states that they used any sort of formal loan. In some cases it's a loan from a friend or family, but it's mostly savings. So then we do some econometric analysis where we want to look at this impact of having a new migrant compared to not having a new migrant and how that affects the welfare. So we run a first difference model of a wealth index that we construct from indicators for housing quality on a dummy, whether you have that new migrant or not, and control for household characteristics that we think can change over time and can affect the outcome, as well as some local characteristics, trying to capture the local labor market. And of course there's the issue of indigeneity. So on the one hand the first difference model accounts, takes care of any time invariant unobservable factors of the households that can affect both the welfare outcome as well as the decision to have a new migrant. And additionally we use a matching method, we cannot really rely on an instrumental variable approach or we could not find one that we think works in this context, because we are not trying to define are you more likely to have a migrant or not have a migrant at all, but actually all these households engage in migration, all these households have experience with migration, so we need something at the household level that makes these households very comparable in order to try to control for the selection issue that we face here. So what this matching approach does is just to construct weights that create a balanced comparison group to our treated new migrant households. I can show you some more statistics for this, but I will now move on. So for the dependent variable, so we use this asset index and just so you have a kind of the first view of what we are looking at is that at baseline the treated, so those households with a new migrant in the blue line and the dashed line is the comparison group, they have a fairly similar distribution of this asset index and this of course also results from creating these weights that make this household's look in observable characteristics almost the same, pretty much the same. So that results in also the asset index, their wealth status is pretty much the same and then two years later we can see that they've shifted apart to some extent. So the control households have a slightly higher asset index than the new migrant households and in the regression however what we find is that this effect, this kind of slight difference which would be interpreted as a negative effect is insignificant and rather small. So in the first column it's just the normal or less without any controls and without weighting and then we use the weights to make these households much more comparable and also add some controls that we think can change over time and affect the outcome. There seems to be some small negative effect of having a return migrant which we haven't unpacked yet but it's also yeah only weakly significant and we also try to investigate does it maybe depend a bit more on the gender of the migrant, whether it's a seasonal permanent migrant or where the migrant moved to, whether they're still close to the origin household or further away and also there we cannot find any significant impact however we see there might be something there if we look at the signs but we don't want to give too much emphasis on that. So far what we take away from this non-result is that we have different aspects that could explain this so on the one hand is that asset indices tend to change only slowly over time so maybe two years is a bit short to actually see a significant change in the asset wealth of these households however we saw in the distribution that there is something there is something there furthermore one thing is that if there was a positive impact of having an additional migrant by receiving more remittances also there it could be that the period is too short to actually capture these positive returns maybe that new migrant first has to settle at destination establish their yeah earn enough income to actually send this income back but we also see that the costs for these new migrants are much lower therefore these households do not experience any negative effect of sending a migrant by incurring migration costs so there seems to be some compensation there and the other important aspect is that in order to actually increase your asset wealth a household has to invest more into their housing right to actually change this variable and if these households are actually using their savings to pay for migration they might not be able to use it to pay for more investments so that's why we might also not see either like a larger positive shift or we see that small kind of negative difference because these households couldn't invest as the other households said so the conclusion is basically just a summary of all these results yeah with the main kind of pointing at okay we need in this area of research to answer the question of what the impact is of having migrants on the households left behind we need more longitudinal data that is able to unpack this and be emphasize that it is important to also account for these dynamic and repeated patterns of migration within households thank you