 Hello, this is a joint work with an Egyptian researcher, Shirin Alazawi, who is at Santa Clara University in California. And it's preliminary work. I'm embarrassed to say that we don't have very strong results. And the project moved from the initial objectives and plan to, in various ways, that I will describe because of data limitations and in the various surveys that I'll be talking about. So conceptually, we started with thinking that there's an inequality puzzle in the Middle East, that some measures of inequality make it look like there's very little inequality. Inequality of incomes has been found low in a number of many countries. On the other hand, there's some literature describing large inequality of opportunities, inequality in early childhood development, inequality in access to education or access to health or nutrition. And so one goal of this project is to solve the puzzle, to see what's the role of regional migration that could reconcile these different findings. The thinking is that migration can affect opportunities and outcomes directly. And it also affects the measurement of inequality. If individuals move, they fall out of the household survey. So their backgrounds and outcomes cannot be measured. And we wanted to see whether this can explain or the puzzle or it would even exacerbate the puzzle. If we measure correctly all the people who started in different regions, whether the corrected measure of inequality would become more reasonable or even give us a greater puzzle. And we also started with the realization that in the Menor region, there is substantial migration, but now that we're in sub-Saharan Africa, maybe these migration flows would not look as significant as when I just worked with the Menor region data and comparing them to other regions in the world. So that was the motivation to understand migration and return migration, how it affects different dimensions of inequality directly and how it affects the measurement of the different dimensions of inequality. And the idea was to use a large sample of large harmonized surveys to study the migration experiences of individuals across the Menor region, the backgrounds of these people and their economic outcomes at multiple points in time. So not only the initial situation and the final economic outcomes today, but with the data from these surveys to measure how the individuals moved through different income levels or occupations to their situation today. And so with this, the larger objective was to say something about lifetime social mobility of individuals in various dimensions, incomes or wealth or education level or employment status, and also to say something about across generations. Because some of these surveys have modules on family backgrounds and parents, and we could compare the economic outcomes of parents to the economic outcomes of children. And some of those... So these ideas, when we started working with actual data, they were confirmed that we can evaluate some of these things. We can study the social mobility over time and to some degree across generations, but we had to change some definitions and change maybe the focus of the paper as we realized some limitations. And generally we want to use something like the difference in difference approach to see how the economic outcomes of individuals changed over time. As a function of circumstances such as the experience with migration and so on. And so we want to study it at the micro level for individual people, individual migrants, how their life situation changes over time, and also look at the aggregate level to say something about the measure of inequality, how does our estimate of the genie coefficient change if we correct for migration and the dropping out of migrants from the sample. And whether we can say something about intergroup inequalities and the different dimensions of inequality. So as I said, the big piece of work here was working with the surveys. And so today I will present some statistics of the backgrounds of migrants versus non-migrants and some information about their outcomes. But no regression analysis because we were fighting more with the data distributions in different countries, different samples, and thinking, let's say, at what level should we pool these surveys together? Is it okay to pool multiple waves for a country? Can we use the panel component or do we have to use each wave separately? So the surveys are three waves of the Egyptian labor market panel survey, 98, 2006, 2012. There's also a 1988 wave, but that's a little bit more different. But even among these three waves later on you will see that there are some, what appears to be systematic differences, and it's causing to question whether it's okay to pool these panels together. Jordan 2010, 2016, that's a very, very recent survey, which is currently still being processed by the survey administrators. But we can already start using the data. Maybe the results will slightly change as the process of formatting the data set will continue. And Tunisian 2014 sample, and I've been told by an knowledgeable person that survey has some problems. And so when we see statistics for the Tunisian survey or when we compare statistics across the surveys, we might find some unexpected results. So this is a simple table showing the sample size in each survey. I will be focusing on 25 to 55 year old men as a relatively homogeneous group. And even among this group we'll see a lot of differences across the survey waves. So among these 25 to 55 year old men, we can classify people who, as a return migrants, if they respond it in some way to several survey questions. One question is, have you worked abroad for more than six months? Another question is, in your last move of residence, did you move from abroad? In the move of your residence before the previous time, was it from abroad? So there are several questions that are asked across most of these surveys and we use them to classify individuals as return migrants or non-migrants. One footnote here is that originally we wanted to study migration rather than return migration, but the data on migration is either missing or because it's answered by the household head or the household responder. So we were skeptical about using data on migrants and we changed the focus to return migrants. But that introduces other problems such as what we consider as non-migrants are not really non-migrants, it's the people who haven't migrated yet. When we look at the ages between return migrants and non-migrants, we'll see that there are systematic age differences. There's still a chance that the non-migrants are planning to migrate in the following years and maybe they should not be considered non-migrants at all. The sample sizes differ and they're all supposed to be representative of the national population, but we might worry about standard errors and the comparability of how precise things are because each observation represents a different number of individuals in the real population, which can be a serious problem. So we try to identify who are the return migrants versus non-migrants versus maybe people who cannot be classified. Impute economic outcomes in real terms at various points in time and between generations. So any outcome in monetary terms is only available for the responding individual for the current time period. So one idea we had was we can use employment status and occupation of the person in previous time periods and use the distribution of incomes today across occupation groups to impute the incomes in previous time periods. And that's possible for several points in time where we may not have the exact date when the previous occupation was held by the individual but we just generally know that we can compare the occupation and income in the current job and occupation and income in the previous job or in the job before the previous job. So there's information about current migrants. I think there have been presentations at this conference about that but we decided to ignore the issue of current migration. I think I will show one table for current migrants simply because the amount of information we have about current migrants is much more limited and maybe less trustworthy. So just quickly about as I said about the earnings we know what incomes are typical in each occupation or occupation group in the current year and using the information on the occupation group of the individual in the previous job in the job before the previous job or eight years ago and the occupation group of the father at the same age as the individual is now we can use those occupation groups to impute the approximate income that the previous generation was earning and the respondent was earning in prior years. Of course that creates noise because we are not taking the whole distribution of incomes at each point in time but we are essentially summarizing those incomes by the average in each occupation group. One nice thing about the maybe small advantage of this approach is that this way we get real incomes. We don't have to worry about inflation or regional cost differences because we are taking today's earnings and we are using them to impute prior earnings. And then another measure of outcome can be the wealth index. So here we take advantage of the different variables in the surveys on household assets and maybe productive agricultural assets the amount of land that households own and in this way try to improve a little bit on the standard approach taken. So we really try to include all of the available household assets particularly because we are thinking that some of these migrants come from rural areas and it is important to include the assets that are typically owned in rural areas rather than cities. Now because of the way we impute earnings and wealth we will be very careful whether we use these outcomes cardinally or just ordinarily. We will just use outcome quantiles so we will look at the relative distribution of earnings and wealth so we will just look at which fifth of the sample the individual falls in terms of their earnings and in terms of their wealth and we will tabulate joint densities of migrants and also of non-migrants on the distribution of outcome quantiles at two alternative points in time and to get a sense of the mobility how the individuals moved over time across earnings quantiles or wealth quantiles we will compute this Shorok's mobility measure or mobility index which really compares it's a simple indicator of the percent of sample who retained their position in the distribution of earnings compared to the individuals who moved up or down from the previous position I think some descriptive statistics about the destination countries of international migrants so this is among the return migrants and this is about their most recent migration spell one conclusion from this table is that across the different years and different countries and years there are different patterns of out migration and there is also different concentration of the spread of migrants to the rest of the world so for example in Egypt we find that 97 or 98% of return migrants went to one of 10 countries which is a relatively narrow region to which they migrated but let's say in Jordan only 85% of migrants went into one of the 10 countries and the other 15% of migrants went to other countries in the world so I think one finding here is that there are systematic differences between where people migrate to from Egypt versus Jordan and Tunisia somewhere in between we can also look at let's say because of historical differences and geographic differences Tunisian migrants tend to go to Europe whereas Jordanian and Egyptian migrants go to the rest of the MENA region and to the Gulf countries and that might affect the economic outcomes of these migrants a little bit of who are the migrants what can we say about their backgrounds well we can for the current migrants we can just compare there's limited information on them but we can compare the earnings in the occupation group from which they migrated so here the sample is limited to people who were in the labor market and they were working before they migrated and these numbers are the average earnings in the occupation group where they worked compared to the overall average and we find that typically migrants came from lower income occupation groups it would appear that in Egypt in 2006 these migrants earned 125 Egyptian pounds compared to 304 so less than half of the general mean but we see it so in Egypt 2012 again the migrants came from relatively less paid occupation groups in Tunisia the same thing but in Jordan we see a completely different picture these migrants came from much richer occupation groups than the general population regarding urban residents at birth typically we find that migrants are more likely to be from rural areas except for Jordan so in Jordan we see a completely different pattern of the background of migrants than in Tunisia and Egypt we see that also about their education level Jordanian migrants are much more likely to be college educated than non-migrants whereas in Egypt and Tunisia it's the less educated individuals who tend to migrate mean age the people who went through migration already are older than the people who have not gone through migration yet so that creates the concern that we need to control for age to really distill who is a non-migrant or who is not yet migrant in terms of economic outcomes we can look at different outcomes individual wage earnings, household wage earnings per capita or household wage earnings wealth index per capita, wealth index whether the individuals have a contract job or formal job generally I think we find that the return migrants are doing better than non-migrants in the current time period so now we would have to carefully look at how they compared before their migration spell to how they compare after the migration spell in this table we can look at average earnings between return migrants and non-return migrants currently in the previous occupation group in the occupation group before the previous occupation eight years prior and fathers occupation I think one interesting fact here is that when we look at fathers earnings we don't see too much difference there is no systematic pattern whether fathers of migrants or fathers of non-migrants earned more but if we look at the following outcomes in the current job, previous job, before previous job systematically migrants become more successful than non-migrants we can draw a picture like this showing that generally we see some upward trend in the occupation groups held by parents and held by the individuals eight years ago positioned before the previous job, previous job and today one conclusion from this graph is that well we see this premium that so the dashed lines show non-migrants and the full lines show migrants and we see this premium to migration that migrants always generally outperform non-migrants but they always did it eight years ago in the previous, before previous job oops, as well as today so it's hard to attribute this migration premium to the migration spell there should be some self-selection where these individuals so maybe one way to summarize is that they don't appear to be family effects but individual effects that these individuals were performing better even eight years ago regardless of when they actually migrated and I will just show to compare that's the last slide I will show we can compare the income quintiles that the non-migrants or return migrants are today compared to where they were eight years ago or we can use a different point in time such as the father's occupation group earnings and we can see what is the extent of mobility if this individual started in the poorest quintile eight years ago oops, eight years prior this one would they still be in the first quintile or all the way up to the fifth quintile in the current time period so to get a simple measure of the social mobility we would look at the individuals who are on the diagonal of this matrix and individuals who are off the diagonal and we might also compare individuals who are up here compared to individuals who are down here who improved their relative standing in the society or whose economic standing became worse off the one thing that I will just mention is that return migrants show much more mobility than non-migrants just looking at the Shorrocks Mobility Index we would see across all the survey waves that the index of mobility is higher among return migrants than non-migrants so whether that's because of selection problems or because of the actual return to migration we see that migrants tend to move more between their original economic position and the economic position today thank you so much