 Okay, thank you everybody for being here. The paper I will present today is Responses to Temperature Shocks, Labor Markets and Migration Decisions in El Salvador, and this is joint work with Ana Maria, who is here today, Juliana Quigua and Jimena Romero. And measuring the effects of extreme temperatures is, completely, is very relevant nowadays. Given that, as you see in this graph, both the frequency and length of heat waves have intensified in the last decade. In addition, measuring the effects of these temperature shocks can be particularly damaging in regions that depend on rain-fed agriculture and for small agricultural workers that, in developing countries, who seldom have access to re-scoping mechanisms. And in regions that might not have a non-agricultural sector that is big enough to absorb the displaced workers. So in this paper, we're going to make a contribution to this discussion by exploring what are the responses to temperature shocks in El Salvador. We are going to focus on two effects, the migration effect and the mechanism that through which weather shocks are going to affect migration that is going to be through labor markets. But in order to tell you this story, first we want to show you that actually, the people that are going to be the most affected are indeed agricultural workers. So the first piece of evidence that I want to show you is how these extreme temperatures are affecting agricultural production. And when I talk about agricultural production, I'm going to focus on the effect on corn. And we focus on corn because it's the main staple in El Salvador, okay? So just to give you a little bit of a preview of what we are going to do, what we see is that there is indeed a negative effect on the total production on corn. And then the question is, if there is like this negative effect that is going to affect agricultural income, how are farmers going to react? And we see that we are going to see important adjustments through the local labor markets. What are they going to do? We're going to see that there is going to be a reduction in the labor demand of agricultural workers. They are going also to adjust hourly wages, but very importantly for El Salvador, and this is an important difference with other settings, we see that the non-agricultural sector cannot absorb these place workers. And then what we see is that this effect that we see through the labor markets is the mechanism through which these weather shocks are going to affect the decision to migrate. Importantly here, when I talk about migration, I'm always going to talk about international migration. Unfortunately, we don't have data that allows us to explore domestic migration. So always when I talk about migration is international migration, okay? And then we're going to explore some heterogeneity, and this is where we would like all of your feedback. We're going to see that there are important heterogeneity by land ownership and access to re-scoping mechanisms like access to migrant networks and access to financial markets, okay? By answering these questions, we make three main contributions to the literature. First, we make a contribution to this growing literature that see the effect of weather shocks and natural disasters. Now, the effect of temperature shocks on migration has been established before, but most of the literature has used aggregate data. Now we're going to see individual data that is going to allow us to understand who are the farmers who are migrating and that is going to allow us to understand better the mechanism and then to identify more fine-grained policy recommendations. In addition, although there is a growing literature and understanding this effect in developing countries, most of the literature has focused in developed countries. And of course we know that the conditions are going to be very different. As I was telling you, in developing countries, most of these farmers are small farmers that don't have access to re-scoping mechanisms. And in addition, some of them are constrained by migration costs. And then we have this important policy discussion and here is again important to clarify that when we talk about the effects of temperature on migration, we are talking about short-term temperature shocks on migration. We cannot identify long-term adaptation to climate change. However, by seeing that the farmers are making adjustments and maybe costly adjustments, there is an important discussion on how there has to be a global responsibility to address the economic effects of these decisions and of climate change overall. As I was telling you before, we are going to focus in El Salvador. This is a country where, with excellent international migration to the US, by 2017, 30% of working-age individuals born in El Salvador were living in the US. And when we were starting to think about this question, what motivated us is that you see this rising trend. It hasn't stopped and you see that it hasn't stopped even though in the last 15 years immigration enforcement policies have been getting tougher and tougher in the US. So then you wonder what is happening in El Salvador that you see this persistent rising trend on international migration. Now, the dynamics of migration from El Salvador are very complex. There are different motives. But what we see is that weather shocks and the reason of why they are migrating because of climate change is having more importance. El Salvador, in fact, is very vulnerable to weather shocks. During the last decade, they had three extreme droughts. And this is important for this contest because most of these farmers are subsistence farmers with very small plots of land and without irrigation systems. So they really depend a lot on the temperature and the rain. Okay, in order to answer the question, we're going to use three different sources of data. We're going to start with an agricultural survey. This is the ENAMP and we have information of the survey from 2013 to 2018. These are cross-sectional surveys and we are going to use this. This is representative at the national level and we are going to use the survey to look at agricultural production and labor outcomes of agricultural farmers. We are complimenting this with a household survey. This is the EHPM. This is also a cross-sectional survey representative at the national level and we are going to use this survey to look at compliment labor outcomes and this is going to be our main source to look at migration outcomes. And of course, now importantly to measure the temperature shock, we are going to use data from the NASA. This is going to be at the one kilometer grid and we are going to have averages at the weekly level and at the municipality level. So what is going to be our shock? And this is going to be very important in this literature and just like to give you a little bit of a preview, we have different measures of the shock and the results always hold. But what is going to be our shock is going to be the number of hard weeks during the main agricultural season where hard is going to be if the temperature is two standard deviations above the historical mean. And this is just to tell you what has happened with the temperature level in El Salvador and our empirical strategy is going to exploit this temporal variation that we see in temperature and the geographic variation. So it's not only that we see that there is temporal variation but it hasn't affected all the regions in El Salvador in the same way. You might start to wonder and I don't want this to just like stay in your brain the rest of the presentation that 2015 was a pretty particular year. We were worried about that too. 2015 is a particular year, but it's not driving the effects. We can drop to 2015 and the results hold. Okay, so now let me tell you all the story of like our predictions and the results. And the first prediction is that when temperature increases, then we're going to see a negative effect on agricultural production. And we are going to see different measures of agricultural production, but I'm going to show you today the effect on total production and production per hectare. I'm going to go kind of fast through the model. So we're going to have the log of the crop yield. This is going to be for each producer living in municipality J in the year T during the harvest season. What we care about is this delta one that measures the effect of an additional week with an additional hot week on the production and we're going to have a bunch of controls. We're going to have controls at the producer level. We're going to have controls at the municipality level. And importantly, we have municipality and year fixed effects. Maybe there is a pre trend that could be correlated with this shock. So we are going to control for these pre trends. All the standard errors are going to be cluster at the municipality and year. But now we have also checked that the results are robust to using only standard errors that are going to account for spatial correlation. Okay, so what happens with total production as the first prediction tells us there's going to be a negative effect both on total production and on production per hectare. So what this tells us is that one additional week with an extreme temperature is going to decrease total production by 2.8% and land productivity by 5.4%. This is already telling you that the farmers are also adjusting the use of land, okay? And very nicely, this is consistent with results found for Peru. Let me just go here. The second prediction is that once you see that there is a negative effect on income, then farmers need to adjust, right? And they need to adjust like the cost of short term inputs. The first thing that they are going to do is that the adjustment through use of agricultural inputs, we see that there is a negative effect on chemical agents, mostly used for post harvest activities. And as we like after the results were suggesting before, they are increasing land allocated to current production, which could be very costly in the long term. And now we see that there's also going to be an adjustment through local labor markets. So what do we see? The prediction is that there's going to be a lower demand of agricultural workers to diminish those costs. And also there should be a substitution between hired and household workers, again to diminish those costs. And that is exactly what we find. So what we see here is a contemporaneous temperature shock on the number of workers who are in the farm. And we are going to, in order to know whether there is a substitution, we're going to see the effect on non-household workers that are going to be hired workers and household workers. And what we see is that all these negative effect that we see in all workers is being driven by hired or non-household workers. This is not significant, so we should interpret this with caution, but the signs of these effects is kind of like telling us the story of the substitution effect. So now what happens on the other side? What is happening with the farmers, with the workers, agricultural workers that are being displaced? So here what we suggest and what we predict is this is going to be different for those who are landowners and for those who are landless. So our prediction says, okay, if you are a landowner, the demand and supply of labor is going to be decided simultaneously. What is going to happen is that you are going to increase your working hours on your own land, but also, and this is based on a model of Sima Jayachandran, there's going to be a reduction in wages. And these may provide like an insurance mechanism for landowners, particularly if they don't have access to re-scoping mechanisms. If you are a non-landowner, then you are going to be displaced from the agricultural sector and you have two decisions. You are going to either reallocate within the agricultural sector to a sector that was not impacted by the temperature shock, you are going to reallocate to the non-agricultural sector and this has been found before in the literature, or you will need to migrate, okay, domestically or internationally. So what do we find? We do find that there is a response in the labor market. Individuals in non-landowner households are decreasing their labor force participation as we were predicting, yes. And we see, like importantly, we see no evidence of reallocation in the non-agricultural sector. We have investigated this in different ways and in no way we see that they are going to the non-agricultural sector, which again is very different to what they have found in other settings. For individuals in the agricultural sector, we do find that they are increasing the hours and it is completely consistent with the idea of the substitution effect that we were seeing before. And as predicted by the model of Zima Jaya Chandran, we see that they are adjusting also the cost by decreasing the hourly wage. Okay, so now we see that there is this important effect through local labor markets and we see that this is going to play an important role, this is going to be the mechanism through which we see this effect on migration. So are they migrating or not? And we see that there is an important effect on migration and importantly, this is all driven by agricultural workers. We know that other things could be happening but there's no effect at all for non-agricultural workers. One additional week with extreme temperature increases the likelihood of migration by 20.1% but this is all driven by agricultural workers. And for non-agricultural households, we see no effect at all. Okay, the last point and this is again where we want to pick your brains the most and kind of tells you also the suggestive story that this is being driven by what is happening in the labor market is that the effect on agricultural labor markets and migration is going to depend on access to re-scoping mechanisms. If there's no access to re-scoping mechanisms then the effect is going to be transmitted to a local labor market. Agricultural wages are going to go down and there's going to be a higher reliance on international migration. However, if you have access to re-scoping mechanisms the effect is not going to transmit as strongly to a local labor market and therefore you don't need to rely on international migration as strongly. So let's see if that is what is happening. So let me just like tell you what is happening here in this figure. In order to measure access to re-scoping mechanisms we are going to have this measure of informal networks. So this is going to be the share of migrants in your municipality. You could think this is going to be indigenous so we are going to measure this at baseline before migration takes place. And what we're going to have is some municipalities that are going to be below the median of share migrants so this is going to be the proxy of you have less access to this informal insurance and this is going to be the mechanisms that are above the median of share migrants so you have more access to this informal mechanisms. What I'm showing you here is the effect on local labor markets. So the first column is the effect on higher workers like non household workers. This is weekly hours and this is going to be wages. And you can see that when we split the municipalities in this way all the effect that I was showing you in the labor market is all driven by municipalities that have fewer access to this informal insurance. We do this with remittances, we do this with formal credit and the story is always consistent. So if you have no access to re-scoping mechanisms more effect on local labor markets and now let's see if that is going to translate on migration. And once again, that is what we find. We find that the impact on the likelihood of migration is lower in regions with a higher share of migrants and remittances which suggests once again that the effect that the weather that the temperature shocks are having on local labor markets is playing an important mechanism that explains the ultimate decision to migrate internationally. So this suggests that receiving remittances might help to alleviate the negative temperature shock and could increase the likelihood of staying in the place of origin. In addition, and I didn't show you these results we see that credit constrained households and non landowners are more likely to migrate. Okay, we have a bunch of roblesness checks. How we measure the shock, whether the effects are being driven by the shock during the agricultural season. Something else that you could be wondering particularly for El Salvador is what is the role of crime? We control for crime in different ways and there's always a consistent story that the effects that we are finding are driven by agricultural workers and by a temperature shock and crime is not driving the effects in this case. So now to conclude what we find. We find that there is a negative impact of extreme temperatures on agricultural production. One additional week with a temperature shock is going to decrease total production by 2.8% and productivity by 5.4%. We see that agricultural producers need to respond to this negative shock and they are going to respond by adjusting through the local labor market. They are contracting the labor demand for non-household workers and using other inputs as well. The labor market therefore is going to act as a transmission mechanism of these negative weather shocks and we see that there is important level of heterogeneity depending on whether they have access to informal and formal insurance mechanisms. Those that cannot have access to these risk coping mechanisms are going to have a higher likelihood of migrating and what we find is that agricultural one additional week with a temperature shock is going to increase international migration by 25%. This might be like you see this number and you might think like, yeah, this is very big, but again, this could be just like because this could be temporal migration or permanent migration and we cannot disentangle this. So just to conclude here, what we see here in this relationship are two types of migration. Migration as a strategy to survive and compensate for income losses or migration as a way out of poverty in regions with untenable conditions including because of climate change. And when we think about policy recommendations we need to think about policy recommendations that are going to tackle these two types of migration. In order to deal with this migration we need to think if it is still possible to be productive let's think about mechanisms that are going to give technical assistance and re-scoping mechanisms in order for them to be productive. But when it's not an option we also need to think about removing obstacles that are going to allow that migration to happen. And I think that is the challenge that we are facing nowadays. Thank you very much.