 All right, wonderful. Well, thank you very much for having this paper on the program. This is joint work with Nora Pankratz, who's at UCLA right now. We are excited to present this paper here. The title is Climate Change and Adaptation in Global Supply Chain Networks. So over the last couple of years, a lot of research in finance and economics has shown that exposure to physical climate risks has all kinds of effects on farms. It affects their capital and labor productivity, it affects corporate earnings, and it affects stock returns, among a number of other outcomes. So we know that these effects exist. As a result, regulators and investors have been looking for ways to mitigate these risks and to manage their exposure to these physical climate risks. But while we know that these effects on firms exist, a lot less is known about how firms adapt and manage these types of risks. And one of the areas of the economy that's typically the most exposed to this kind of risk are supply chains because they run through different parts of the world that are very strongly affected and exposed to these global climate change effects. And this is exactly the place where our paper tries to shed light on. Specifically, what we're trying to do is we're trying to examine if firms that are engaged in these global supply chain networks are adapting to climate change risks by adjusting their supplier input network. Specifically, what we're looking at is the question if customer firms adjust their supply chain networks when the perceived change in the distribution of physical climate risks of weather shocks changes. We're looking at the question if it affects the termination of existing supplier relationships as well as the startup of new relationships that are replacing those those previous relationships. Before we get to this main research question of our paper, we also study first sort of the necessary condition the question if these customer firms have a financial incentive to make these adaptations by looking at the question if these shocks. First of all, these weather shocks propagate within these within the supply chain networks, estimating the propagation of heat and floods at supplier locations for the financial and operating performance of customer firms. So to give you a quick preview of the results in the first stage of our paper which we sort of think of as sort of a necessary condition for the second stage. We first show that both heat and flood shocks to different degrees but both of them affect and decrease the performance of the supplier firms in the first place and then in the second place also of their customer firms as these shocks sort of propagate and spill over through the through the network. And then in our main part of the paper in our main contributions, we show that when these shocks, when the occurrence of these shocks starts to exceed what a customer could have reasonably expected coming into this relationship from a benchmark period before. The customers start to terminate these existing supplier relationships and they then move on and replace suppliers with other firms that appear to be less exposed at least when looking at the benchmark period before the before the termination of the of the relationship. So our paper sort of sits in the middle of several different strands of the literature, we're looking at adaptation in production networks so we're speaking to this emerging literature on the formation of production production networks and economics. We're looking at specifically the role of climate change when it comes to the formation of these production networks we're also speaking to the literature on climate change and finance and the propagation of shocks and networks that has has grown quite a bit recently by shown that not only do dramatic massive natural disasters, propagate through these networks but only also smaller climate related weather shocks that are not big enough to be classified as sort of big natural disasters and then we also contribute to this literature on climate change and learning. By by focusing on the question how these customer firms learn about changes in the underlying distribution that sort of giving this these these weather shocks. Alright, so for the rest of the talk I'm first going to speak about the data that we use in this paper for a few minutes. I'm going to introduce the conceptual framework that we have in mind when thinking about these results and then I'm going to talk about our main empirical strategy and results focusing sort of on the adaptation part in the most part. Alright, so our data comes from sort of two or three main sources. First of all, we need to identify these existing supply chain relationships we use facts that revere for that. It has the advantage to not only identify existing relationships but also comes with start and end dates of these relationships which allows us to even study the determination and start of these relationships in the first place. A second advantage is that it covers many different countries around the world, which is important because climate change is sort of heterogeneous in the way that affects different parts of the world, which is exactly our underlying assumption when it comes to these these adjustments we supplement this with information on the locations of individual establishments of the firms in our sample to be better able to measure which firms are exposed to weather shocks at any given point in time. Of course our sample is sort of biased a little bit towards larger firms since facts that revere only has listed firms in it since we need financial information. We're a little bit biased towards developed markets as well in the same way because of the reporting requirements that give facts at the data. We think that these data biases would sort of push us away from finding bigger effects that if we had of the full world sample, we might find larger effects because smaller less global firms might have an even stronger exposure to these types of shocks. We combine this data with information from World Scope on firms financial performance operating performance and other firm characteristics. And then we add our two types of weather shocks to this to this data shock particularly what we're focusing on our floods and heat. And we focus on these types of shocks because both of them are well known to be expected to change going forward. As a result of climate change this is different from other papers looking at natural disasters like earthquakes, which are much harder to tie to climate change. The two shocks that we're looking at floods and heat waves are very directly connected and linked to changes in the in the climate. And both of these types of shocks have very different effects on how firms operate so it allows us a little bit to to compare the magnitudes and the types of the channels through which these shocks affect firms and their and their adaptation. Alright, so before I get to our main result. I want to introduce sort of a, a simple theoretical framework that kind of guides our analysis and illustrates how we think about this about this issue so our first stage results are first part of the paper focuses first of all on the propagation of this of these shocks. And so the way we think about this is as an economy in which a number of customers produce certain outputs under perfect competition and for producing these outputs they need capital, they need labor, and they need a set of inputs from a group of suppliers. In our setting as is common in the literature adjustments in the short run are not possible so capital is fixed. And as a result, both customers and suppliers make sort of strategic decisions on who to source inputs from to to maximize their profits. In this, in this in this world, we have weather shocks that are related to the to the climate distribution that are drawn from the underlying climate distribution, and the underlying climate distribution as indicated by this omega t can change over time. And as these weather shocks occur as as has been shown in other papers, for example, Baro and Sylvania and affect the supplier firms in this network. It can constrain sort of the availability of inputs that are important for the customer's production away from the optimal and therefore affect the performance of the customer firms and that's exactly what we do in the first stage of the paper document. Evidence consistent with this idea that not only large natural disasters but also climate related weather shocks can distort these these production networks and affect negatively affect the performance of the affected customer firms in the interest of time. I'm going to mostly focus on our adaptation results which are the main the main focus and the main contribution of the paper. All right, so what we're focusing particularly on is customers decisions to either continue or terminate existing relationships in any given period we call this PT as sort of the decision of the customer to either continue with a supply chain relationship that they have or terminate this relationship. So as we show in the first part of the paper, the occurrence of these weather shocks is costly. And so customers who understand this, before entering a relationship, assess the risks of these weather shocks before entering a supplier relationships. So, exactly before entering the relationship they try to get an idea of their exposure to these shocks by estimating omega zero so the exposure to the shock at the beginning of the relationship. And as a result, all in sort of perfect competition, all existing supplier customer relationships are such that if the risk of weather shocks increases, it would, it would, customers would not be willing to continue with this relationship because the risk of an increase in these weather shocks would increase the the cost that the supplier represents to the customer and make this existing supply chain relationship potentially suboptimal. Right. And so over the course of the supply chain relationships customers continuously observe the actual sort of shocks that occur at the at the supplier firm and reevaluate what they think is the underlying distribution of of the climate. And so, as long as the realization of these types of shocks sort of is below what the customer expected ex ante, there's no reason for the customer to make any adjustments. As soon as the occurrence of these weather shocks indicates that the climate distribution has shifted, and shocks are now more, more likely than they were anticipated to be the customer has an incentive to to terminate and move away from this supplier and and replace them with a different supplier. All right, so how do we, how do we think about how customers estimate both the, the, the omega zero so the expectation of climate shocks, as well as well as well as well as how do they update the their their beliefs well the way we think about this is that customers typically in our setting rely not on forward looking information because this is really difficult for firms to process with with the abundance of different climate models and the complexity in those climate models. It's extremely difficult for firms and practice to use this type of information and so we we operate under the assumption that customers mainly rely on experience shocks as they move through the supply chain relationship. And so the way we construct our main measure of interest our main available of interest is as follows for each customer supplier relationship. We look at a benchmark period before the supply chain relationship was experienced was initiated to estimate what is the average number of shocks that happens during that period of time. And then as the supply chain relationship continues and starts each year a customer assesses how many shocks have occurred on average since the beginning of the relationship to try to estimate if this the climate distribution has shifted and if the number of shocks now exceeds what you could have expected going into this into this relationship. All right, so this is our sort of main specification outcome variable on the left hand side is either a one or zero depending on whether the customer continues with the existing relationship or terminates it. A main variable of interest on the right hand side is this indicator variable that indicates if the observed number of shocks since the beginning of the relationship has exceeded on average what could have been expected coming in from this benchmark period for for the initialization. We saturate this model with fixed effects at the supplier customer industry year level, as well as at the supplier customer and country customer country and your level to account for any other trends or industry trends that might explain what's going on. All right, so our main result is in this in this table what we find is that the exceedance of expected shocks positively affects the likelihood that existing relationships are terminated in the following year. So we find that this is true both for heat shocks as well as flood shocks to give you a sense of the magnitude we also estimate this linear probability model as a logic model instead. We find that the likelihood of termination increases by about 6 to 8% for heat by about 9 to 11% for floods. We do a number of robustness checks estimating this with alternative specifications. We also find that there are a lot of potential hazards models dropping certain observations and find that the results are very consistent across all these different tests. Now, one issue of course that we face with the data is that we cannot clearly observe if the relationship is terminated by the customer or by the supplier. So to give some more confidence in the idea that this is a demand side decision from the from the customer, we are going to look at a number of different characteristics that are sort of indicative of the fact that this is related to a demand side decision. So first of all, we're looking at the question is the the decision that we're seeing is this is this consistent with a learning process on the side of the customer. And what we find is that in line with this idea of a stronger signal and the repetition of the signal, which are both sort of characteristics of the learning process, we find that this effect increases when the magnitude of exceedence of realized shocks relative to expected shocks is particularly high. And when these shocks are repeated several times so those are both characteristics in line with the learning process which which suggests that this has this is this is coming from the demand set decision from the from the customers. And at the same time we also look at transitory shocks. This is shocks by themselves and not compared to the to the benchmark period. And if this was indeed a disruption of the suppliers that is mainly driving this we would expect to find at least as a similar size effect there, which is not what we find, we find that there is a small effect there on these transitory shocks but it is much smaller, and does not have the same magnitude of effect. So our main measure survives the inclusion of this to have these transitory shocks. We also exclude firms that went out of business. We exclude firms that were in financial distress to further address concerns about the supply side and find that the results further to other assumptions that are sort of underlying our our main analysis is that customers and suppliers are both operating in perfect, perfect competition and perfect product markets. Of course that's not necessarily the case in reality. So if we are in the in an industry where customers or suppliers have certain market power, we could see that this decision is distorted and in line with this idea we find that the more competitive the supplier industry is so the closer to a perfect product market, the more, more sensitive customers are to exceedance of these shocks so the effect increases with competitiveness in the supplier industry. Of course it's also difficult for customers to precisely estimate kind of expected exposure shocks they have going into this relation relationship, which would again sort of make them less sensitive to to these observed shocks. What we see is that if we change the band of this of this benchmark period if we estimate it over five years over 10 years over 15 years, we find similar results across all of those across all of those different specifications. One thing that I have not talked about yet is this is the presence or the existence of climate forecasts and climate projections for the future so our main analysis as I said is based on this assumption that in reality it is very difficult for firms to operationalize and use climate projections, given the complexity of the models, given the abundance of different models and so forth. And so we mostly rely on this experience change for learning in our in our paper, but to also shed some light on the role of climate projections. So we get some we have some data on climate projections for the for the mid century between 2040 and 2059 from from from several climate models and then we specifically focus on those relationships where there is actually very little climate change projected over that period of time in the location of the supplier firm right so if if the location of the supplier firm is projected not to have a very strong increase in climate change in until mid century, we should observe a much smaller reaction to these experienced shocks by the customers because the projections say that in this location of the supplier, there will not really be a big change going going forward. However, this is not at all what we find in the data so if we look at the sub samples of firms where the suppliers are in locations with very little expected change until mid century, the effect size that we document for climate for for exceedance of the shock expectation is pretty much exactly the same as in the full sample which also kind of supports this idea that the customers are mostly relying on this experienced experienced shocks and the experience changes to adapt and not so much on on model projections. Last in the paper we move on to the question on of replacement so what do customers do after terminating these existing suppliers and to study this question. We tried to match the terminated suppliers with other firms that are in the same for digit sec class so that do very similar things that initiated a supplier relationship with that same customer within a year to sort of try to get as close to actual replacements as we possibly can with the data that we have and then we estimate the exposure of the replaced suppliers and the replacement suppliers over both the period. The original existing relationship, as well as the period after the existing relationship to try to discern if the suppliers that were replaced had a lower expected sort of exposure, as well as a lower realized exposure after the after the termination of the initial after the termination of the initial relationship. Right so our main specification to examine this question puts a dummy variable on the left hand side that indicates whether or not the new suppliers had a lower exposure to these types of shocks than the old suppliers and puts again a the same indicator on the right hand side, which, which, which sort of indicates supply chain relationships that were likely terminated because of the exceedance of the expected number of shocks again we we saturate this model with a number of fixed to do to control for our trends in you know trade relationships industry patterns things like that. All right so before showing you the, the model results and estimations. Let me show you graphically kind of what we find. This is sort of the result for heat exposure, comparing the difference between replacement and replaced suppliers delineating between those relationships where the realized exposure was higher than the expected exposure, and those other relationships where that was not the case. Right and so if we focus here in this first chart on the exposure during the actual relationship, which led to the termination, we see that the replacement suppliers. In those cases where replacement happened because of because of a climate exceedance seem to have a lower annual heat days seem to have a lower exposure to these type of climate shocks than the, the replaced ones which is not the case as much in the ones that were that were terminated without such an exceedance. If we focus on the period after the termination this difference becomes quite a bit smaller. As I'll show you in the table there's still a significant estimate there but when it comes to the realized difference between the replacement and the replaced supplier after the termination the difference becomes becomes much smaller and if we then turn to climate projections for the mid century we do not see a difference anymore so we do not see that here if we consider the climate projections for the very long run, there is a large or any kind of difference between between terminations that had something to do with with climate expectations and and those that were terminated for for other reasons. Let me show you the corresponding quantitative exercise so this is basically just the table version of what I showed you in the in the chart just now we see that the difference between the likelihood that the replaced replacement supplier has a lower exposure than the replaced supplier increases or is much higher if we had this exceedance of climate exposure during the existing relationship this effect is especially strong during if we've measured the difference during the existing relationship it is still significant but much smaller in the period after and and it is much much less pronounced or not present when we look at projected days in the future, looking at the at the mid century if we looked at the results for flood at the corresponding flood results, they look roughly similar however we find a much smaller and not as significant effect when it comes to the difference after the initial after the initial relationship. Five minutes. Oh okay then I'm actually a little bit ahead in this case so I'll take my sweet time here on this on this conclusion slide and and give our discussing a little bit more time to talk about the paper that. All right so so what do we do in this paper let me summarize the the main takeaways. We first of all document that both heat and flood shocks decrease, not only the performance and profitability of the supplier firms, but also spill over to their customers through these supplier and customer networks. This has interesting policy relevance in the sense that it kind of has implications for what kind of disclosure we should require for climate related risks, and how we should think about climate stress tests in the sense that we should include these kind of input output linkages and production networks in our in our model of climate stress. Then in our main analysis in the paper in the in the main contribution. We show that customers are more likely to replace or to terminate existing suppliers. When the occurrence of weather shocks exceeds what the customers could have possibly expected going into these supply chain relationships, and they're more likely to choose alternative suppliers who have a lower exposure to these types of climate shocks, at least when we consider the period before the termination when we consider the period after the termination, whether they actually achieve sort of an improvement in climate exposure is is a little bit murkier and a little bit less clear. So the three areas that we mostly contribute to with this with this paper are in climate adaptation related to other papers who are trying to understand how firms how agents in the economy learn about shifts in the in the underlying climate distribution are papers related to the to the literature on climate change and finance by term by by documenting the the financial implications of the propagation of these types of shocks. And then last to the to the literature on endogenous production networks, which has studied a number of different issues such as legal issues or property rights. So I'm going to do this by by focusing specifically on exposure to to shifts in in the in the climate change distribution. All right, thank you very much for the attention. I am very much looking forward to Michael's comments, and I'm happy to take any questions after the after the discussion. Thank you. Thank you very much, Christoph. So now let me guide Michelle Ketter from HALA Institute for Economic Research. The floor is yours. Thank you so much. Let me start by sharing my screen. All right. So, first of all, thanks to the organizers for for giving me the opportunity to discuss this wonderful paper and thanks to Christoph for an extremely lucid and clear presentation and also for giving me enough time to discuss this wonderful piece of research. Let me hit the ground running since this is our last intermezzo standing in between us and dinner, at least here in Europe. What this paper is doing it provides evidence on firm responses to climate shocks. And I think it is at least one, perhaps even two rich papers, since it is so deep and comprehensive. The paper basically isolates the effects of climate risk in two ways. First, it shows the climate risk realizations to affect directly supplier performance and in a way indirectly customer performance. An aspect on which Christoph didn't spend any time but which basically takes up around half of the results and content of the paper. Second, the paper isolates the effect of climate risk expectations and specifically on how customers and suppliers adapt to changes in these expectations in the form of choosing to terminate and form new production network ties. The main findings that I took away from the paper are summarized in the following bullets higher than expected climate risk in the location of the supplier directly reduces operating performance of these suppliers that indirectly leads to a deterioration of customer performance. And customers are more likely to terminate supplier links that are basically exposing them to higher climate risks. And this is a result which is irrespective of a couple of, I would call them modifiers like frequency or intensity of the expectation shocks, but also on the realization side, depending on the competition nature in the various sectors. And ultimately this leads customers to replace these stressed climate stress suppliers with less stress or less exposed suppliers indeed so I think there's little to add to your work what I had to do because it is such a such an animal with paper I had to landscape it a little bit so what I did some time is that basically the whole driver in an economic modeling sense is unobservable climate change and what I really liked about this paper was this angle that climate risk realizations are nothing but noisy signals of this ultimately unobserved climate distribution. So specifically, these are heat and flood events and deliberately not disaster risk materializations used elsewhere hurricanes that we have seen falling on land or earthquakes. The sections you haven't been seen in the presentation spent quite some time on showing how reduce our supply performance is reduced specifically basically how revenue scaled by total assets and operate results are depressed and subsequently how customer performance and then the second big part where apparently the focus in innovation lies is how expectations of customers about climate risk of their suppliers changes and these expectations then in that of themselves lead to ending relationships in section four. Whenever these realizations exceed expectations so it's a lot about the precision with which customers are able to to forecast if you will climate risk and driving the behavior of replacement which takes up a mere two pages of the paper that I got at least. So, let me use this landscaping to to pose a few questions here so this is and I want to start with this a very impressive empirical exercise I like very much the laudable efforts to combine climate and economic micro data. I have to reveal my ignorance I did not know this fact that revere around 8000 customers in 5600 suppliers over a long range of time, which is linked to financials I wasn't sure if you see financials for each firm on each quarter that are then linked via Orbus to various locations. In 92 countries for suppliers and 74 for customers now. Temperature flood data is very granular, for instance, 0.25 by 2.5 latitude longitudinal grid, but at the end the spatial unit remains a little bit unclear and I will come back to this when I post my questions. So let me start with three main questions on supplier and customer performance. Using my little map. So first of all, the paper is very transparent about taking an agnostic stance about the mechanism. It really doesn't take a stance on how exactly performance is depressed. Now I think that's not quite innocent. Why? Because it might just as well be climate policy in direct response of a materialization of climate change risk. And what I have in mind is that there's a lot of policy that follows exactly at the level of analysis it. The firm in year T due to a climate risk realization think of the flooding of the river elbow a little bit buyers here. Now that triggered a lot of transfer payments to mitigate the fallout of this climate shock in this case a flood. Likewise, there might be taxes that might be directed actually locally at firms in response to specific adverse climate events. And last but not least, think of insurance into which both suppliers but also customers can self select depending on their expectation of climate risk realizations. But of course these contracts will change regarding that price. The premiums will change if you want to have a flood insurance in the old city of Cologne. For example, hardly anything you can get. So there are a couple of events going on at the very level of variation it that might very well correlate with observed outcomes in terms of performance at the both supplier and customer level. Generally I was wondering when reading the paper how important measurement errors and I think it is great to think of it as a production network. But it might be hard to generalize from the sample of selected large firms why I think what the papers doing as you look at country D one having some suppliers that provide intermediate inputs to a customer see in country D two. Now, one thing from a simple smell check on the internet was that Volkswagen alone has 40,000 suppliers. We are looking here at some five to 8000 front so I wasn't sure how comprehensive this network is and we know from network analysis in France's interbank market studies that it is very important how complete the network is your study for instance. Second, the strategy to use headquarters to allocate both suppliers and customers is not entirely clear to me all the time. So for example, if I look at 3M company, which is the supplier of Apple according to their physical statement. They do have for instance in Guangdong, a plant probably, but I'm not sure if you would find financial information on China man land mainland 3M plans, knowing that Minnesota Minneapolis is the headquarter of 3M. So it's not so clear to me and the paper I think could benefit from being crystal clear here. What exactly is the spatial unit so as to understand also the variation source. Along these lines. Let me raise a few questions on on this first part and they carry on I think also to sections four and five which you emphasize during the talk. So let's say we observed suppliers probably from the trade literature we know that only very few firms do what I would coin FDI because the data you have here are basically coming from Apple having a plant in Guangzhou. But the majority of international trade and supply chains I think is going through exporters. So I was wondering if you have something to say about the importance of non incorporated firms. Second, what if Apple sells iPhones to Guangzhou. In the help model of the modes of international trade we know that the size of the foreign market determines very importantly how and whether Apple, the customer see would enter Guangzhou by means of setting up FDI or an export. And it of course raises the question of how clean a shock we are having. I am aware that you have this 500 kilometer perimeter, but still I would be interested in hearing this bidirectionality of the network being incorporated in your analysis. Then there are basically country time fixed effects, but within a country of course these climate shocks are highly localized. So you might be to the right of the river where it is hot and or underwater or you might be to the left of the river where it's cool and you're sitting high and dry. If there are firms as two that are also supplying to see, I don't think that the fixed effects that are DIT do the trick to totally isolate within firm shocks. So I would be interested to hear how this should work. Notabina, this carries over to your analysis of pairs where you have supply country year fixed effects, as you showed us in your presentation, I think, for as long as you have within country variation and exposure, non exposure to these localized climate shocks. Let me skip this point in the interest of time. I'm already my 11th minute and I also want to go through this slide. This is more for you. Christoph, this is summarizing my points I made. I want to spend one minute if I may on the cool new section expectation formation. Now this is a Bayesian supplier type of idea where basically the customer learns about the climate risk of the supplier. I have a very simple question. Why is this learning not taking place symmetrically? Suppliers, of course, also do see the noisy public signal about climate risk realizations and might take action just as customers take actions. They might sever the link. They might just as well take out insurance. And this, I think, would also correlate in a very similar way with something that is observationally equivalent, the termination of a supplier relationship. Say a supplier sees that she would have to ensure, but it's so expensive to get insurance, so she just stops to supply to Apple because it's no longer feasible. So symmetric learning is the heading that I wanted to put up here and there's a couple of examples here. One additional point that might be important is the fact that you are drawing from a normal distribution that's pushing it a bit since I think we are still looking at extreme weather events. So how normal are these aren't these more like tail events that's more technical note. I skip this point. I think the last point I want to make is the expectation formation on past realizations. Clearly, I can see your argument that it is complex to decipher climate forecasts. Okay. But wouldn't financial market participants do exactly this aggregation of information on behalf of firms. So how, how defendable is this assumption that you rather rely on, on past experience only and not look at financial markets, for example, to learn about expected climate risk changes. Let me skip about the nitty gritty. I want to conclude on the main takeaways. This is a great paper. I recommend everyone reading it. Very carefully executed super rich and important topic that potential for high policy relevance and I personally learned a lot. I wish you all the best of luck with this very cool research. Thank you very much. Thank you very much Michelle for very detailed discussion and yet sticking to the time. We don't have questions in the chat so let me abuse my privilege as the moderator and then make a comment. I think one way of responding to the discussants question or is it really climate risk or something else is to focus on financing constraints, because firms don't have many suppliers because it's costly to find a new supplier. So, if you would find that credit constraint or financially constrained firms are less likely to switch suppliers in response to climate risks. That would be that would make you results more robust to the criticism that was raised. And actually in my own work I have looked at propagation of a fiscal shock or an increase in a border tax in Turkey. One thing that was very visible there was that firms that were hit by a shock were more likely to find new suppliers if they were not financially constrained. Anyway, so over to you Christoph you have, you know, couple minutes to respond. First of all, thank you both so much for the for the comments Michael super super in depth I really appreciate it. I think your comment about this being a monster of a paper is probably on point so I highly appreciate you sitting down and spending so much time and reading it in such great detail so much. I'm going to ask you of course to also send us all your your slides and everything so we can incorporate everything that you said. Maybe I'll take the second to just clarify a few things that maybe we need to make a little bit clearer in the paper regarding some of your comments so in terms of the first comment. So the the the propagation this is all done at the quarterly frequency, and it's also done within sort of existing firms with seasonal adjustments and within country by year fixed effects so any kind of country by time period fixed any kind of regulatory shocks that you were talking about in terms of subsidies or, you know, flood prevention measures or things like that should as long as they happen on the on the country level should be should be soaked up by that. In terms of the spatial variation that you talked about we kind of do quite a bit of work to specifically address this directly in the paper. This is why we collect data on the on the on the locations of the individual production sites and facilities, and we do show that this propagation is sort of stronger the more concentrated. These assets and the more concentrated these facilities are around the headquarter. So this is a conscious decision on our part to focus on the location as as a headquarter location. Because we're we're dealing with two sort of bad choices on the one hand you're right that headquarter locations are sort of an imprecise measure of of production activity. On the other hand, facility level data is super super noisy it does not tell you much about how much production activity is actually taking place at these individual locations and so it is just a less promising approach and introduces even more noise than than than it is what we're doing. So this is a conscious choice on our part and we try to deliberately address this by actually testing how sensitive our effects are to that we also have a constraint. And the paper that follows baron souvenir to limit our exposed firms to those that have some some minimal concentration of assets close to the to the headquarter to do guarantee that that's the case. And then last, if these firms were indeed all sort of spread around the world and not affected by this at all, you would have no reason to finding an effect. So I think this would to some degree bias us away from from finding anything. Another thing I wanted to say, you're, you're right about the sort of the granularity or the comprehensiveness of the supply chain data, of course, folks wagon has 40,000 suppliers but not all 40,000 are going to be in our in our in our data set. The reporting requirements that make it possible to sort of identify these these relationships are skewing to some degree towards larger and public companies right we need them to be public so that we can identify their financial performance. We need them to be public so that we can actually identify the existence of these relationships so to some degree this to skews us a little bit towards larger firms towards those kind of firms but we do think that these larger firms are less concentrated to some degree they're more likely to have done some mitigation they're more likely to be able to sort of deal better with these types of shocks. So our, our, our ex ante expectation is that if we were able to identify the entire set of firms. We would likely find a larger effect, at least on the on the supplier firms in the first and the direct effect because smaller firms being hit by these shocks should be affected more than larger firms that can do more to to deal with this kind of with this kind of shock. I'm happy to continue this conversation also at some other point are offline but for now my list of things that I wrote down while you were discussing is at its end. Yeah, thank you.