 It's nice to be back to WIDER and make a presentation which links up to the kind of issues that WIDER is usually involved with. This is a work which is nearing completion and we would like to thank the project Dranties, which is the South Asia Network on Development and Environmental Economics, which gave us the funding to do this kind of work. This is of course part of a five country project that we are looking at and trying to link up with the variability and migration via the agricultural route. So the question that we are asking is how do we really try to analyze this and this is the method that I am going to use, that we are going to look at the context and the objectives of the study and how the migration patterns are in the Indian context and that in some sense sets the stage for how we are modeling and how we are looking at the analysis and then of course the methodology and the final results. So if we really look at the linkage between climate change and agriculture, there has been substantial amount of evidence to show that the two are linked in the sense there are adverse effects and if you see the first graph out there which is basically trying to link up the case of the South Asia where you can see that without climate change the blue part is basically showing increases in per capita calorie availability and the two scenarios are for 2050 and we start with 2000 and you see that the 2050 availability is lower in the presence of climate change scenarios, two different scenarios depending on the cases. If you look at the focus on just the two Indian examples, they are done by my co-author for this particular paper as well who has done substantial work for India and you do see that the proportion of people in the bottom two expenditure classes in the rural areas that's the bottom poor is actually increasing depending on the intensity of the climate change. These three are different scenarios for climate change and this gives the most intensive scenario of climate change over the period of time and you see the effect is more severe in the rural than in the case of the urban population. There is also some interesting study that was subsequently done in 2011 about last year where there was scope for adaptation that was basically trying to bring it through the modelling of spatial autocorrelation which shows that without spatial autocorrelation there is a higher decline in the net revenue from agriculture as a percentage of the 1990 base value when you allow for adaptation. So this kind of brings us to two interesting linkages that there is an impact of climate change on agriculture and that some of the impact does have a lesser effect when we give scope for adaptation and in this sort of literature is where the migration comes in and so there are several reasons why migration can happen. One of them is this most standard literature that we know of coming from the Lewis model or the Aristotle model which is basically that the process of migration, the process of urbanization brings with it a certain amount of migration but what is also important to notice that the pace of migration is somewhat lower for the kind of growth rates that these regions are showing. So people are not moving as much as they should for the kind of overall economic growth that they are showing but they also see that there are a lot more shocks that can happen in these regions through cyclones, droughts and floods which are also exacerbated by the climate change events and migration can happen. Some of it can be short term, some of it can be long term. There is also a short term migration that is increasingly being recognized particularly in the Indian context due to the distress in agriculture and rural livelihoods because of the kind of investments that are taking place or the lesser productivity that agriculture is bringing in. If we focus more on climate change and migration then there is substantial amount of literature that has been coming in since mid-2000 looking into the migration as an adaptation strategy to climate change in the sense that people want to move out of the kind of distress that they have and that becomes a kind of adaptation strategy and if you look at the particular work that came out in 2011 by Black et al. it largely synthesizes this kind of work into different sub-heads of migration due to environmental conditions, migration due to social, economic and demographic reasons and sort of puts the climate stress as one of those environmental reasons. And if you look at the kind of the way the studies have been looking at it, we can say that migration could either be a cross-country movement like from Mexico to the US or from the Asian countries like the Philippines to the other nations as well as within-country migration. So the migration is discussed in the sense from both perspectives and the migration can also be either planned which means there is a very clear framework which says that there is inundation going to take place and there can be a planned migration or there can be an autonomous migration where people just migrate in an immediate response to a distress or a shock. And coming to the more focused issue of this particular topic that we have here, I think the tone was set in by the paper in 2010 by Feng et al. which was looking at the link between the impact of climate change through its effect on maize and betelis and the emigration from Mexico to the US and they sort of used a certain methodology which to some extent we are also going to follow and we find that there was in their study that there is some amount of a weak connection between these three linkages. Then they subsequently came up with a more recent work early this year linking it up to the US case where they then talk about internal migration within the US on the impact of climate change productivity in certain parts of the US and its impact on internal migration. Then again you see another recent work that basically looks at the sub-Saharan region and in fact it very nicely splits up the migration into two levels. One is the intra-country migration and the other is the inter-country migration. And they very nicely summarized their work from the rural, the migration that happens from the rural to the urban areas caused by largely the economic reasons that is the more productive regions tend to draw more people while they are moving but there are also there is what is called as also the amenity channel in the sense that once people move to these better off regions within their own countries they also want to move to even better off regions once they are better equipped in terms of their skill sets and so on and then they tend to emigrate. And here the weather plays a slightly different role whether directly has an impact on emigration whereas in the case of the rural urban migration they find that the weather agriculture and then migration is the kind of linkage that they are able to show and this is again a very large macro study that they have conducted in the sub-Saharan region. Apart from that we also have a few micro studies which are basically linking up looking at the household level data and particularly look at a panel data looks at agricultural risk and it's linkage between weather variability and migration only focusing on one country that's Nigeria and these are of course largely the econometric models that one is more commonly familiar with but we also have simulation techniques which are basically looking at forecasting approaches where in the case of the Brazilian study they look at the influence of agriculture cost by climate change impacts and in turn how it influences migration. A very very recent study about a month back it was just published which doesn't use an econometric model which doesn't bring in agriculture directly but sort of looks at it through the shocks and finds that there can be about 3 to 10 million internal migrants that is within the country people can move in the next 40 years within Bangladesh. So that's the kind of tone that we set to this particular work that I'm bringing into and understanding very well that there can be several reasons why migration can happen and being a developing country there can be several triggers to migration we are basically trying to focus on the weather variability induced migration operating through the channel of agricultural productivity changes so that's the large question and focusing on the nature of information that we have we have this very specific objective which is to look at interstate migration that's at one level states are provinces in the Indian context they are administrative boundaries which actually are distinguished on the basis of the languages that are spoken in different parts of the world and at a second level we also look at districts within each of those states that's where we sort of look at it at an interstate level because there's a larger variability in the information that we have and we also bring in a certain difference in the kind of results that we have on the kind of crop choices that are made in this particular case and I'll just tell you why it may be interesting to have a crop choice very shortly. So how does a migrant get defined? A migrant as far as the data said that we are using is based on the national census the country level census that is conducted every 10 years and in that case the question that is asked is when they go to a particular household the question is is the place of enumeration different from the place of last residence and if the answer is yes then you are a migrant but then there is also another interesting angle to that in terms of how long and what is the duration you've stayed in and that question is also added along with that so even if you have been staying in a place for 30 years sometimes you can be called as a migrant for the simple reason that you were born somewhere else and so on so there is a little bit of issues in terms of how the migrant is exactly defined and some of it is discussed in a little bit more detail in the paper and the migrants are then further classified into these 405 categories some of which we will be using for our analysis durations of stay this is what I was just mentioning how long a person has been in a different place of enumeration and the gender issues are important in the Indian context then you have the source and the destination whether you move from rural to urban or rural to rural within the state between the states and so on and then there is also a very interesting category in the Indian context that when because of the practice of what is called as exugamy and endogamy that is people marry within families and people marry within their own villages as well as there are several parts of the country where people marry outside their village and so on so the question of have you migrated also very strongly affects the reason for migration so you will find that a large number of women call themselves as migrants though they are actually for migrating for social reasons so there are several reasons and some of which is not directly relevant to our study and we will be using that for our analysis this is where I just said that there has been some amount of interesting patterns that are emerging if you see the dark line out there that is the all India migration rates that you see there across the 4 census years that we have used the data for our analysis and you see that the migration rates for men is just about a little around 10% but it is kind of declined whereas for the women the migration rates are higher at about 16% and the difference is largely because of the reason for migration that women tend to give which is not only for employment and other reasons but also for marriage but what you also see is that there has been a decline more in the case of male migrants than in the case of female migrants you expect that less to happen here but the fact that it has still happened is also a puzzle and there is not much work going on there but what has really bothered the people who are working on the migration literature is that why is it that when the base of urbanization has been somewhat good when the economic growth has been fairly high the migration rates have actually declined over the years and that people are not moving as much as they should for the level of development that we are having here so that's the kind of debate that's still going on and figuring out and we are still our latest 2011 census data has been collected but it's not yet collated and put out for information and here when we start splitting it up into the source and the destination the first letter R or U stands for the source and the second letter U or R would stand for the destination so the possible ways for us to classify the data set is to look at rural to rural urban to rural and so on combinations we would largely be focusing on rural to rural and urban to rural analysis in this particular case for our work and what you see is that there is also another pattern overlaid on top of that whether they move within the district boundaries which is a sub level of the administrative boundaries within the states where they move between districts so these two are within the provinces or within the states and this is the third category that you see is between states so the reason for looking at between states is because there's a lot of difference in the levels of development across the states and you do see that there is quite a bit of variation in the economic growth rates as well as the rates of industrialization and so on so that kind of pans out in the result that we are seeing for the men is that the rural to rural migration bar that you see which is the first bar is actually declining in numbers so though the migration rates have been declining the numbers of rural to rural migration is actually declining and that to some extent is responsible for the decline in migration rates but what I would like you to observe is that the rural to urban migration is actually picking up and by 2001 it's been quite substantial what is also substantial to note is that this colored bricklayer mark that you see there which discusses about the interstate movement is also actually picked up in the case of men so two things have changed for men one is that more and more of men are moving from rural to urban areas and that it's mostly interstate movement while in the case of women you can see that it's predominantly rural to rural that is what is increasing whereas all the other categories of migrants are very small in number so what we see is that if you really again go back by the reason of migration you will see that marriage still dominates us about something like 90% of the women would report marriage as the only reason for migration so that's how it's also important for us to choose what form of migrant we are trying to look at and while we also try to look into the case of another data set which is not coming from the census but called as the National Sample Survey Organization or NSSO which recently was able to collect data on what is called as long term migrants and short term migrants and here you can again see a very clear difference of what we call as the monthly per capita consumer expenditure deciles so this is the poorest and this is the richest decile here and you can see that anyway women dominate when you try to look at them in a long term sense but what we also see is that this one is the urban male so the people who have migrated among men to the urban areas are largely from the return sections that means the most skilled have been able to make use of the transitions that have taken place during this period and when you look at short term migrants this is predominantly male migrant and they are mostly from the poorer sections so there is also an issue of how the migrants are or what kind of backgrounds they come from and what kind of migration that they undertake a short term migrant is one who stays away from his place of residence for less than 6 months whereas the long term migrant is one who has moved away from his place of residence for any time beyond one year that's how they try to classify these two groups so having said the tone for this data that we are trying to look at and we know that there is quite a bit of issues in terms of understanding what could be the reason finding out whether induced migration could be somewhat difficult for the information base that we have and we are looking at it through both these channels that was discussed in the literature that's the agriculture channel and the amenity channel and so what is the data set that we have as I said we have two levels of analysis carried out the first one is the state level and that looks at interstate out migration rate that is somebody moves from one state to the other state we have three years of data but within those three years of data we have what is called as durations of stay so between any two census year we have somebody who said I have been here for the last one to four years and I have been here for the last five to nine years and using that we sort of classify our information on the past so somebody who says in 1981 that he came five to nine years before to the place where he gets enumerated then the person must have come between 1972 to 1976 then we have similarly somebody who says one to four years and the person would have come between 1977 to 1981 so we are able to splice the information that is given in the census to further get more data sets over time so we are able to create about six time points which cover these five years I do not know individual years so I have to work with these periods that we are looking at but we are able to work with 15 large states in the Indian context and we exclude people who say marriage as a reason for migration as well as people who say place of birth as a reason for migration because that is another social issue when the mother while delivering her child tends to go back to her natal home and the child gets registered in the natal home rather than where the child is staying so we have to exclude these two cases of reason for migration when we come to the state level the challenge is slightly different in the sense that we do not know the source of migration we do not know where the people came from but we know where the people are getting into and we also are able to work with only one census the advantage of this is that we have a panel data with about 90 points that is six years, 15 states whereas here we have one census with two years of duration but a whole lot of districts about 500 or districts that we have in the country so you can see that the variability in the data set is likely to be different between these two contexts and because of the district level we do not have reason for migration given we have done one estimation for total migration rates and one for male migration rates so that sort of very quickly tells you the way we have to organize our data set the temperature and rainfall was also not an easy data set to work with but luckily the Indian Meteorological Department gave us very detailed data sets made it available to us and one of our co-authors I mean one of the co-authors in another work was actually working on this data set and we are happy to have the data set that he provided to us and we sort of updated that to correspond to our years the agriculture data set has been taken also at two levels in the sense that one we look at the net state domestic product for agriculture which is per capita and which is the value added from agriculture and then we look at two crops separately that is rice and wheat which are the predominant crops that we have in the Indian context that people tend to grow the econometric methodology as I said follows a very simple simultaneous equations model approach so the first, the main equation of focus is migration influenced by yield and there is no other explanatory variable or any other variable directly in the model and all of the remaining variables are assumed to be either in the fix effects across cross section or the fix effects across time so we basically do a test and find out that that's the kind of model that fits in and then we test for whether the yield is endogenous meaning it's correlated with the error term in which case the yield equation is to be explained by the temperature variables that are sitting here and we have this kind of relationship in the framework that we have here so that's the large model that we have there if you look at the state level the interstate out migration rate and use the value added from agriculture we actually do not find any evidence for endogeneity which essentially means that we cannot estimate it as a two equation framework the weather is largely insignificant in the migration equation but the yield has an impact on the migration in a very small amount, very small rate that we are looking at in the particular case if we shift to particular crops that is the wheat and the rice you will see that there is a fairly strong relationship of endogeneity compared to what we had earlier more stronger in the case of rice than in the case of wheat which means we can use a two equation model that we are referring to but while we are doing that the weather in the rice equation is actually insignificant but overall we find that the yield has an impact on the migration two very quick observations is that the magnitude of this case in the case of wheat is lower than in the case of rice in the sense that because rice is more cultivated in larger regions rice is also more labor intensive so we expect the coefficient to be larger but you can see that the significance level is also not that high as one would have expected it to now we move on to the district level and in the case of the district level I would very quickly expect you to understand one thing which is different from the earlier case so what are we trying to say that if the yield declines then the migration rate would go up so therefore that's what it basically means or in other words if the yield increases then the migration rate would decline so there is an inverse relationship between yield and migration you can also see that the kind of variables that affect wheat and rice the temperature variables are also different and that's important from the perspective of policy which I'll come to towards the end when you move to in-migration data now we have to slightly look at the results differently that when somebody comes into a particular region they would only come in when the agriculture is doing well so in the case of the in-migration we would actually expect a reverse sign that is a sign that when yield goes up more and more people would come into a particular district and we would expect a positive correlation between the two so therefore there is a linkage that would change when I'm working with in-migration data but what is also important to note is that there is a mix of what is called as intra-district movements and inter-district movements which kind of affect our results so if you look at the wheat case there is endogeneity in both whether we look at total migrants or whether we look at male migrants there is a positive sign which sort of goes with what we look at that is if a district does well in the yield it draws in more migrants and people tend to move more within that particular region and you can also see that the weather variables are kind of mixed what you see here is that this is the impact of the suing temperature this is the impact of the harvesting temperature so you can see that if the suing temperature goes up then you have a different impact the harvesting temperature goes up then the yield declines so that's the kind of results that you see here and you can also see that the percentage of male migrants is always higher than the percentage of total migrants the emphasis is that wheat rice results are completely in contrast to what we see there the first contrast is basically in terms of seeing that the endogeneity is much stronger in the case of rice we see that the coefficients are also far more significant but they are negative in the sense that if a district does well in its yield in its yield of agricultural rice yield or it has a much higher productivity in rice yield then the mobility tends to decline that's the kind of results that we are finding which is in contrast to the case of the wheat case but what we also find is that the district level results also show a much better linkage to the temperature impacts so just by summarizing this we would basically say that there is some evidence of the kind of changes that you see on the migration and the patterns are also dependent on how we are trying to look at the linkage if we just do a simple hind casting exercise rather than a forecasting exercise we find that there is a moderate change in the migration rates that would have declined had there been a 1 degree centigrade temperature lesser increase in the last 30 years of that we are looking at so in another sense if the annual temperature were 1 degree more than what it has been in the last 30 years we would have seen about 0.44% increase in the migration rates than what we are seeing here and as you can see we had to choose the particular temperature and so on because the models were a little more complicated so just to summarize we do have a weak evidence if one were to look at it but it is largely depending on how the agriculture is showing up as androgynous we also see a very small migration rates but yet we find that there is some amount of linkage between the migration rates affected by the changes in crop yields due to the climate change so we see that there is we have sort of set the tone and this is for the first time in the Indian context we are able to organize this information base to this extent to be able to pull out the kind of information that we are trying to get and there is quite a bit of work that we need to do to take it forward particularly in the context of what it means for a very high proportion of rural to rural migration thank you so much