 Alright, thanks everybody. It's been a wonderful day. This is the last session, which I'm very excited for. So, you know, this session is going to be focused on what do we know empirically about the relationship between climate or meteorological variables that are related to climate and aspects of the economy that either describe the macro economy or might plausibly be inputs to things that are part of the macro economy. And I think this session, I mean just for context, there was a, you know, a raft of research that was done in the 90s and the early 2000s, helping us all think theoretically about climate change as an economic problem, you know, thinking about contributions of Bill Nordhaus and Nicholas Stern and Marty Weitzman. These were all theoretical models that were sort of describing the nature of the problem, sort of qualitatively but using mathematical models. And very little of those models were calibrated to sort of real world data. And so there's been, since then, something of a revolution in terms of measurement, measuring these types of relationships. And that's what we're going to hear about right now. I think the challenge to this audience is to ask the question how do we take those measurements and integrate them into the models that we're using for these types of decision making purposes. So I'll get out of the way. Marshall Burke is our first speaker. Marshall is an associate professor of global environmental policy at the door school at sustainability at Stanford. And he's the deputy director at the Center for Food Security and the environment. And he has research focuses on social and economic impacts of environmental change, and I'm measuring the understanding the economic livelihoods across the global landscape. So, Marshall you have 10 minutes and I'll interrupt you at eight. Thanks a lot, Saul. Thanks for the invitation to be here. Sorry to be remote. You guys are all wearing ties so being remote and as a California means I don't have to wear a tie, which is wonderful. So I'll start us off here thinking again as Saul mentioned about data driven estimates really focused on the physical climate risk so this will be on the physical side now on the transition side. So looking at both the US and globally we're going to use some global data and then try to think about what those data mean for the US specifically so I can't control the slides, but next slide. Thanks. Okay so two basic approaches that you're going to see between me and and Tama and Jeremy I believe. And then what I'm not going to do is as bottom up or what is sometimes called the enumerative approach so this is really nice it uses trusted micro data we can go sector by sector and get really high quality causal estimates of climate impacts in different sectors so in mortality and agriculture in labor supply. And these are great again, high quality micro data, really strong econometrics, great estimates. So those are the pluses a challenge in these especially when you're trying to have them inform the sort of modeling exercises that we saw before, or think about broader sort of aggregate economic impacts. You need to be able to measure things across all sectors so you need to be able to enumerate measurement in all the sectors that we care about that contribute to economic output. And then you need to be able to somehow integrate these estimates right so it's the integration of many sort of partial equilibrium estimates over sectors and across space and that can be challenging. Next slide. So opposite approach and one I'm going to walk you through today is what we call a top down approach so instead of a new rating impact sector by sector and adding them up we're going to let the economy add things up for us and study the aggregate output of that adding up so looking at which indeed is what we'll look at what's nice about this is the adding up is done for you so that's nice. An added feature is that at least some of the costs and benefits of adaptive measures so things like sector reallocation that you might be worried about are going to be embedded at least to some extent in these measures so that's nice so that adding up is done some of the adaptations are embedded. But there's challenges to number one, we don't have a ton of GDP data, right. Especially if you limit to a individual country. And so that's why we're going to use global data. And maybe more importantly you're going to miss that you care about that are not in GDP and this has already come up it will come up again I assume and Tamas remarks things like how we value lives lost due to a changing climate. That's really important we think about these impacts. That said we're going to see how far we can push the GDP data in what follows the next slide. Okay, so I'm not going to take you through the econometrics happy to talk about that more if folks are interested. Really what the econometrics boiled down to is the following thought experiment. So in a year that is hotter than average. In the US economy or any economy that we're going to look at does it grow faster or slower in that year, and in subsequent years. Okay, so again, take your the taller than average, look to see is an economy growing faster or slower in that year, and in some other year. Why is this a useful thought experiment. So the variation in temperature tends to be somewhat random right there's a long term trend but about that trend there's some randomness some years or hotter some years cooler, and we can use that randomness to think about its impacts on economic aggregates in a way that reflects the nice micro level causal inference that I mentioned on the previous slide so basically we're going to take the same causal and approaches we're going to apply it to the GDP data, and think about across a lot of different economies and a lot of different years. Next slide. So we're going to do this globally. We're looking at country level outcomes. Again, we don't have a lot of data if we just focus in the US so we're going to go globally. We're using data on per capita GDP growth so real per capita GDP growth, which we have for most countries around the world back to the 1960s. So think of roughly 200 countries roughly 60 years of data per country. And then we're going to merge that with temperature and precipitation data and run this thought experiment actually in the data so this is going to be a panel regression that again doesn't try to compare Nigeria to Norway it compares Norway to itself over time. It compares Nigeria to itself over time as the temperature fluctuates, and then we're going to try to fill in this plot on the left and say okay what what does this overall response function look like. And before showing you the result let's just think intuitively what we would expect it to see. So if you're in some of the coldest economies in the world so think Iceland think Mongolia. This is on the on the so on the X axis here is average temperature at really really low temperatures what might we expect what we might expect that these economies are actually a little bit better off in years that aren't terribly cold right they might benefit from a bit of warming. Take that out to the extreme we don't produce much in the Antarctic part of the reason is because it's very cold right so you warm the Antarctic up, maybe we can produce more there humans are more comfortable. What happens at the other end very very hot, you might expect that productivity is going to fall right output is going to fall as temperatures increase if you're in a very hot place to start and temperatures further increase so that would suggest a sort of hill shaped relationship between temperature and output. Next slide. And I primed you to see this and sorry for the slightly blurry figure here that is indeed what we see so this is the global non linear responsive GDP growth to temperature. We're still at at a cold average temperatures this thing is upward sloping we see in the data that very cold countries tend to benefit. Most countries are around the peak of this curve or are in the tropics to the right of it and so most of the world begins to be harmed as you slowly warm them up or push them to the right on this plot as I'll show you in a second. Next slide. Quickly this thing is very robust again sorry this is so blurry very robust to different ways of estimation I won't spend a lot of time on this happy to chat next slide. One immediate thing you might be interested in is is adaptation. And so can we study adaptation in this framework. Using 60 years of data that the world has gotten much richer three to four x richer over the last 60 years depending on how you count surely we've adapted surely we've gotten better at dealing with warmer temperatures or temperature fluctuations. An easy thing we can do in these data is actually split up the data decade by decade or here I'm grouping them by two decades and study whether that response function changes over time and we find absolutely no no change in the response and it's totally rock solid over the last six decades so as one measure of adaptation you might expect this thing flattening out as we become better at dealing with really hot temperatures we see no evidence that that has happened in historical data. Next slide. Okay last thing to look at and then we're going to run the world forward and think about future economic impacts. We want to understand whether a hot year this year affects output or growth and output in this year but also potentially in subsequent years. This work suggests there could be lagged effects this has important econometric implications to for thinking about whether we're actually picking up growth effects or just what we call level effects. And so we want to run a lagged model where heat can affect output in this year and in subsequent years. I'm going to show you these in terms of marginal effects so think back to that plot I just showed you the hill shaped response function now I'm taking the derivative of that right so I'm just measuring the slope of that function. If that thing is positive that means it's upward sloping and country's benefit from a little bit of warming if it's downward sloping means countries are harmed. I put the USA average temperature in here, population weighted average temperatures about 14 degrees Celsius. And so us is very close, if not just a little bit past the peak of that plot. Okay, so this is, this is in the same year the effects of a hot year on output in that year. Next slide. Over two years. So in that year and the following year and then adding up the effects. If you look closely you can see that things coming down just slightly next slide. And we can run this up to three years or five years or even out to 10 years the picture looks pretty similar. As you add more years you accumulate what looks like more and more damage so more countries have this line now that is below zero, and even substantially below zero, the cumulative effects of temperature are much larger than the sort of immediate effects of temperature in a way that really matters for the calculations that we're about to do. So again the cumulative effects when you include these lagged effects, get larger and larger the more lags you include out to about five years and then to us it looks like it's fairly stable at five years next slide. Okay, so what are the implications for near term GDP this is the topic of today. So what we're going to do is we're going to take those response functions really seriously and anticipate that economies in the next few decades are going to respond to temperature increases just as they have in the last few decades so we're going to run the world forward. We're going to assume a baseline growth rate so imagine this is just that we picked 2% 2% baseline growth rate. Assume that's a no warming growth rate and then we're going to perturb that based on that response function. I just showed you we can do this for the US we can do it for anywhere in the world I'll just show you results for the US. So here's the change in global temperature out to 2100 under different emission scenarios I won't dwell on this let's just focus on the green one which is the path that we think we're on right now, relative to today so this is all done relative to today, 2020. We're going to warm additional degree and a half maybe two degrees by 2100 under this SSP three 7.0 sort of the again the scenario we think we're on right now. Okay, that's a temperature warming next slide. So to pump that through our response function we can measure then the impacts on the growth rate. So, folks here mainly interested in 2050 so if you go to the middle of this plot on the green line. This is our zero lag model so this would be our more conservative model. We find that you lose about a 10th of a percentage point of annual growth rate by 2050 every year. If you were going at a 10th of a percentage point slower than you would have been absent warming under an emission scenario that we think of as sort of business as usual right now. Okay, next slide. Okay, what does that mean for total output so here now I just run it out to 2050. So this is sort of budgetary window we're interested in. We find a number that was actually pretty close to the one Bob used earlier and was used in the CBO I think they used 1% by 2050. We find one and a half percent by 2050. Again under our most conservative estimate. If you run the lag model you include the cumulative effects this thing is actually four to five times larger. So this is much larger so this would be I think a very conservative interpretation of the results you get out of the GDP data if you take these lagged effects seriously but this would be the number 1.5% last slide and I'll finish up here. So how to integrate these estimates into a longer term budget outlook so if we assume they're right, which of course they are. We think these historical data offer a useful empirical constraint on the relationship between growth or output and temperature right and an empirical constraint or a set of moments that these models can try to match. Next slide. I think one thing to point out so so sort of two conclusions there one thing that was sort of pointed out or worried about in the very nice OMB CEA white paper on this that Fran mentioned earlier is if it's the case that temperature has already slowed productivity growth in the US. So our budget outlooks are based on historical slowing of TFP growth and maybe climate is already sort of baked in to our estimates. Our data would suggest that's not true the US has been right at the optimal temperature and so at least historically had not been affected but will be affected as future temperature increases push it off that optimum. And I would also suggest that the TFP slow down, at least in the in the GDP data is is not a result of warming in the US specifically but will be in the future so projections that reflect the slow down or not already making a point and the recommendation here is again and I think very close to what folks are doing right now which is great. So take your preferred model with TFP or with factor specific productivities and adjust those productivity sets that the model output actually matches the empirical constraints that we see in the data and if I understand Bob earlier it sounds like this is very close to what the WHO is doing. And I think that's great I think that would be a nice way to integrate these estimates, so I will stop there and hand it off thanks. Excellent. Thank you Marshall. Now I'm going to introduce Tama Carlton who's an assistant professor at the Brent School of environmental management, environmental science management at UC Santa Barbara, and Tama's research combines economics with remote sensing data science and climate science to quantify how environmental and economic development shape one another. Thanks all. Thanks so much for the invitation to be a part of this event. As Marshall said I'm going to focus on the bottom up approach to measuring climate damages and with a particular lens on how that approach and some of the new empirics that we've been working on shed light on the inequalities of climate change. Next slide please. I'll think about climate change as a global phenomenon of course, but we don't experience climate change in the form of global mean surface temperature global sea level rise we're experiencing these impacts at a local level and often in the form of extreme events, perhaps no better example than the Pakistan floods that hit last summer. Next slide. But these local level effects, of course, manifest very differently around the world while Pakistan was getting these floods the southwestern US was getting hit with a series of heat waves, and the actual welfare effects of those two events with very, very different despite both being driven to some part by this global phenomenon of warming. Next slide. So accurately capturing these local level damage estimates in a sector specific way can be really important for climate policy in two different ways so one. We're not going to be getting aggregate climate damage estimates correct if we ignore the fact that they're very differentiated effects across heterogeneous population. And the other is that as we're building plans for adaptation looking forward we need to have accurate estimates of what's going to happen in a sector specific manner on the ground to make it a locality. Next slide. So, as Saul mentioned, a lot of the work that we all here are contributing to sort of built out of early global climate damage assessments that were very influential but gave us estimates like three degrees of warming is associated with the losses globally of about 1.3%. This gave us one sort of aggregate number across sectors and across locations around the world. Next slide. But where the frontier sort of moving now is much higher spatial resolution that allows us to really dig into inequality in a way that wasn't possible before. So these are six examples from work by collaborators at the climate impact lab over measuring climate damages for each of these sectors at the scale of about 25,000 different regions around the world. It's not just us so I'm of course going to talk about our work at the climate impact labs throughout this talk but I want to point out that a variety of different scholars working with very different tool kits are also working to increase the spatial resolution of these estimates to start building a much more comprehensive understanding of inequality under climate change whether you're using spatial equilibrium models like you're seeing on this slide or the type of bottom up empirical approach that I'll highlight through the rest of the talk. Next slide. Another important sort of feature of the new era of climate damage estimation assault pointed out is its empirical foundation so on the left is a histogram of the dates of publication of empirical papers behind the original I am going into the original US so on the right you're seeing evidence of an explosion of empirical work in the space and this is the type of empirical work that we think can and should be brought into both bottom up and top down damage estimations. Next slide. Another feature is our ability to integrate probabilistic projection so this means thinking both about statistical uncertainty but critically about climate uncertainty both in terms of the total magnitude of warming were likely to see by looking at an ensemble of climate models and that's on the X axis on this graph, but also the spatial distribution of that warming so each climate model is going to have slightly different projections about who's going to face what climate and when we're thinking about inequality, getting a probability of six cents of local level warming is critical. Next slide. The last feature I want to highlight in terms of where this literature is that is using that empirical data and those empirical approaches to capture differential vulnerabilities so we're thinking about differential abilities of populations around the world to adapt to the exact same climate event. So you're seeing mortality risk on the Y axis against daily temperature and recover those response functions for two very different places Oslo Norway and a frog on it. And just the differences in those curves and in those temperature distributions shows us that if we kind of ignored that heterogeneity we with something like a global average response function would be pretty dramatically mischaracterizing projected effects of climate changes to localities. So I think these sort of innovations leave us is in a position to really transform the way that climate policy is particularly treating equity. So we can build climate damage estimates that are empirically grounded and globally representative accounting for that differential vulnerability in different populations and characterizing and valuing uncertainty this will be important for like aggregate metrics like the SEC, but also when we're trying to specifically think about equity as part of the kind of policy problem. Next slide. I'm going to sort of give a window into how we're doing this type of work at the climate impact lab and then I want to wrap up with just some thoughts on some of the challenges of the bottom up and sector specific approach Marshall already pointed to a few of them that that sort of will end. Next slide. This bottom of approach basically means that we're going to be going sector by sector constructing empirical estimates of climate damages using a lot of the same tools that Marshall just highlighted generating productions for each sector and then pulling those together into an integrated analysis for something like a social cost of carbon or even just an aggregate estimate of damages at a given time period. Next slide. I'm going to start by looking at mortality and then I'll sort of point to some results from across different sectors that we've studied. Next slide. So this is a empirical approach so we're going to start with data collection we in the case of mortality to have mortality records for about 55% of the global population this type of data collection will look the same in other sectors like energy or labor etc. We then build empirical those response functions just like you saw this sort of looks like an upside down version of Marshall's graph where we see mortality risk elevated under cold and heat conditions particularly for the elderly population. Next slide. We then capture this idea of differential vulnerability by allowing that for that you just saw to vary across space based on conditions in each location. For example, we find that in the oldest regions of the world that are highlighted on the left, the effect of a hot day is very extreme, because people are not prepared for that excess heat that they're going to get on that day. In contrast, in the next slide when we move to the hottest regions of the world, we see that the effect of a hot day is much more muted because these are populations that have adapted to that climate. Similarly on the next slide we want to allow for economic resources to influence the shape of that curve so it's not just that people are adapted to their climate but it's also that levels of income in a given population and facilitate adaptation through lowering sensitivity of mortality for example to heat and that's what we see in the data. Next slide. So that type of analysis that empirically looks at how mortality risk and temperature look to that relationship varies across space and time can be used to construct these type of locations specific to those response functions that capture that heterogeneity and therefore inequality in a way that hasn't been done in the past. Next slide. We can pair that with climate model projections to build estimates for example here you're looking at end of century of climate change damages to a given sector accounting for the fact that different populations are differentially able to respond. The map is showing you the average but then you know the value of this probabilistic approach is that for each of those locations for recovering a full distribution of projected impacts. What we're seeing sort of systematically across these sectors is that these impacts are highly unequally distributed around the globe although the spatial patterns look somewhat distinct so on the next slide we can see that mortality damages. Next slide. Are born primarily by today's global for that's what you're seeing on the left, but on the right you can see that the adaptation spending we estimate to be to be extended in order to protect against mortality risk to see that that is born more by today's populations. A very different pattern emerges on the next slide for agriculture where we see that the losses are actually greater in the world's bread baskets which are more temperate and generally wealthier, which creates a very different pattern of inequality relative to mortality risk. On the next slide we can see that in the case of labor, the inequality really falls along sectoral lines so it's the workers who are spending their time in agriculture mining and construction industries where they're exposed to the climate, where they're really suffering effects on their labor supply and therefore a disability of working under extreme conditions. And then on the next slide in energy. Again we see that income is really critically important so we find that for the vast majority of the global population people don't have the resources to increase energy expenditures and the response to extreme temperatures and so we're really not seeing effects on the energy system in these places it's really isolated to wealthy places on the right of the slide. Next slide. So this is sort of evidence of the examples of that bottom up approach but of course we often want to pull that together to build aggregate estimates of damages one way of doing that is the social cost of carbon the monetary value of damages that are caused when we release one additional to. So, next slide I want to talk about how we can integrate that extreme amount of spatial heterogeneity into the access and calculation so I won't get into the weeds here but the basic idea is that we can compute a spatial certainty equivalent damage function that that basically when we aggregate damages across space we're going to place higher weight on the damages that accrue to poor regions, where every dollar is worth more in terms of its marginal utility, then it is in a wealthy place and by comparing the damage functions on the screen you can see the one on the right that accounts for that inequality leads to much more damages relative to a level of consumption then when we ignore that inequality. And on the next slide we can see how that manifests for for the SEC in that comparing the first column to the third column with versus without this accounting for equity can really dramatically change these aggregate estimates of damages. Okay, I'll wrap up quickly. So one key feature I want to point out that sort of relates to Marshall's presentation is that I think this approach that the bottom up approach is showing us that non market damages are really important component of aggregate damages of climate change. You look at just the blue bars on the right of the screen these are our sector specific estimates of social cost of carbon. And you can see that labor and mortality, which are the two non market components of our SEC so far are really a significant overall number. That's true in our work at the climate impact lab but also on the next slide, you can see that it's true and other work as well for example rendered at all. New paper on the SEC last year that shows again, mortality is a really important component of the aggregate number just meaning that we need to pay a lot of attention on market damages. I think Marshall is standing up so I'll wrap up really quickly on the next that I have a couple challenges I want to highlight in how we are going to build these sectors specific damages into aggregate damage metrics I think the first is how we do monetization converting from physical units to monetary units. You know that that's going to depend really critically on strong assumptions. Next slide Marshall already mentioned this that dealing with feedbacks and interactions it's tricky in a bottom up approach in general but particularly when a lot of these damages are non market. And then finally I think that migration is likely to be really important in characterizing location, the next slide sorry, and location specific sector specific damages and I think this is a really tricky problem that we're still actively working on. Thanks so much for your time and sorry for using a little extra. Excellent. Thank you, Tama. Our third and final speaker is here in person with me. Jeremy Martinage is the chief of the climate science and impacts branch within the EPA. And the branches work includes developing indicators tracking the observed effects of climate change, developing the climate science and economic analysis supporting EPA regulation, according to zero project, a collaborative modeling effort to quantify and monetize the impacts of climate change in us. Thank you, Saul. Thank you for the invitation. Marshall and Tama are very hard acts to follow so I will try my best. So we heard top down global econometric analysis. We next heard from Tama bottom up global analysis across a number of different sectors and so I'm going to come down to the US and talk a little bit about some of our modeling and EPA specifically focused on bottom up modeling. And so I'm going to talk a little bit more about what's going on in our country. So, a lot of this work really generated about 16 years ago when Max Maxman Marquis was being considered in Congress and we were, we were asked to present to a number of staff members and the time of what can you tell us about the tools that we can just sell to our constituents about, you know, describing the benefits of this rule and we worked a lot with integrated assessment modeling community and CGE models at that time and when we would talk about, you know, changes in productivity or consumption or GDP, or this new concept of the social cost of GHGs that was just hitting the street, they're like no no no we don't want any that we want to be able to tell our farmers or our public health practitioners that our lives are going to be benefited as a result of this rule, and that got us down this path. You know, many years later that we now have with the CIRA project, which is, if you can go to the next slide. So then we have an EPA. Yep, to quantify climate impacts it's a lot of bottom up process based modeling to look at a large swath of the ways in which climate impacts human health our infrastructure ecosystems. And really the intention is to be as comprehensive as possible it's certainly not comprehensive. The problem with this approach is that every time you have a new set of questions, you're dealing with lots of modeling teams. And it takes a long time to rerun everything and so we try to develop a next slide please are reduced form reduced complexity version of the CIRA project and we call that Freddie. We're currently actively in working to further build out and further apply. It has a number of sectors that I will show to you in a second again, this uses all the detailed bottom up. Very high resolution high temporal resolution spatial and temporal resolution modeling that you do from a from an impact sectoral modeling perspective, and using that create to create damage functions that you can then build and use in Freddie. Next slide please. So here's some of the sectoral coverage within Freddie it's certainly not complete. We've worked with a climate impact lab. So those folks are here Tama saw Bob and others, James rising to get this information into Freddie. We want to continue to expand this certainly climate impacts modeling is happening well beyond just our two groups and so we are actively, you know, wanting to build as much in here. There's a lot of coverage here but there's a lot of these are incomplete not only because there's some missing but also because the ones that we have may only capture a part of the total effect. Again, this is entirely within us borders so some of these impacts are happening. You know a lot of impacts are happening globally that would have effects on us interest and those aren't captured here. Next slide. So, these specific numbers are not are not the point of what what I want to show you here it's really just the capabilities of Freddie to take answer questions of, you know, what do total damages, you possibly look like for the sectors that are represented. How do those map out across sectors so we can try to understand at an aggregate level or nationally or, or regionally, you know what types of impacts are occurring in different parts of the country and where they may be most significant. Next slide please. We can also look at the regional sectors that we map to the seven national climate assessment regions shown here. We're working to build that down to even finer resolutions this year. We can also look at both at national and regional scales, how these impacts might be distributed across different populations of the US shown here are. Populations based on race and ethnicity and the likelihood that some populations may face greater risks rather than others. Next slide please. So, going back to this Freddie coverage. Fran, John and Andrew Wilson in the back and others we've been working with to think about how we can use Freddie, given that it has a decent sectoral coverage to inform a lot of the work that they described earlier. And one of the things that we're running up against is that, you know, the levers that we have to connect this physical impacts modeling to, you know, the mouse model to other, you know, broad macroeconomic models, those levers are not well established. And you can see that, you know, a lot of the impacts in Freddie here are not, you know, directly relevant to capital or just don't have a good lever to pull on. If you were observing it and you saw before the largest damage sectors for the five of those are not shown in purple or red so that's meaningful. And there's a lot of important climate damages that are not, you know, captured would not be captured if we were to include anything right now in any economic or macroeconomic framework. Next slide please. So, I'm going to just talk through some, some takeaways here that that first one like I just mentioned is that there's, you know, a lot of additional work that we need to do to, you know, think through these connections between physical climate impacts to the sector level and then what are the levers within these macroeconomic models that we can connect to. That's something that we're doing that we're actively working on right now. And hope to hope to be in a better position, you know, in future iterations of the processes that Fran and john and others are leading in the White House. Freddie is far from comprehensive. There's a lot of impacts that are that are not included. As I mentioned, we also tend to, you know, like most physical climate impacts models underrepresented streams. That's something that starts with a lot of the climate models that, you know, modeling projections that we use that are developed by others and then is further worsened just in how we tend to, you know, focus a lot on means and moving forward. So that's some place that we are continuing to work on. You know, one of my things I really want to impress upon is that And we heard this a couple times from from, you know, today is that, you know, focusing on, you know, national macroeconomic effects which is obviously the whole notion of this round table. It's really important for this community. I think from an eat from a kind of where I stand at ADPA I definitely see that, you know, or I'll just use an example I don't think there's a single person my extended family that could tell me what GDP means. And most of the stakeholders that EPA works with don't know either. So I think this, this whole process would be remiss. If we go through it and we're not thinking of ways that we can use the same modeling that we're doing to look at changes in productivity and consumption, but also making sure that we're driving endpoints from it that better connect with people because that's how the administration that's how EPA that's how others are needing to sell the value of what we're trying to do to reduce emissions and why it's important to people. So, you know, want to kind of nail that point home. And finally, I had touched a little bit we are and tamas presentation was great and thinking about and the climate impacts lab has been very progressive and thinking about distributional effects globally that's something that's really important here in the US and and something that's really important for the EPA to do when we release rules is to be able to talk about, you know, how would this rule benefit, you know, those who are, you know, overburdened, and whether it will or whether it won't. Being a climate scientist and working in a climate science group that does a lot of climate economics work. I'd be remiss in thinking that are without saying something about, you know, and as in the climate science world. What happens over the five next five to 10 years is not at all a timeframe that we think of because climate, you know, getting that signal out of the natural variability right now is really difficult and, you know, we are just at the tip of the iceberg. And so, you know, thinking beyond the next five to 10 years, where we know that the rate of damages will really be ramping up is really important. So, you know, would encourage this, this, this group and committee and round table to, to, to not, you know, of course we're going to, when we're thinking about this budget process we're not going to get out of the 10 to 15 year period, but also to be thinking about what's coming down further down the line and implications of that. So, thank you. Great. Thank you, Jeremy for that. While questions come in, I'll, you know, since I had two questions before that didn't get called on I'm going to use my privilege to ask two questions real quick. Well, for the three of you. Can you just speak for one second about the extent to which the approaches you described capture adaptation in ways that we might want to think about going forward. And also we saw three different presentations and we thought saw three different spatial resolutions. And so just given that you know this exercise that we're talking about in this room is a lot of macro to what extent is spatial resolution matter, you know, is that. I don't know if any of you guys want to step up to that and then I'll write down names for questions. Jeremy Tam will go ahead. I can start and go ahead. And on adaptation, the short answer is that adaptation is included to the extent that and in the manners that we've seen it happen in the past so we're using historical evidence on how say a heat wave looks very different for mortality and then it does in Seattle, or how electricity consumption responses to heat or differ in Delhi than they do in Florida to inform what we think the future will look like under those climates in the future and so to the extent that we've seen people adapt and respond. In the past, we're going to be capturing that but to the extent that there are big technological changes or really big shifts in people's behaviors that look very different from what we've seen in the past those, of course, would not be included because of sort of taking this empirical approach and relying really heavily on historical data to generate projections in the future. And then really briefly on space I think the short answer is that it matters a lot so we uncover really substantial heterogeneity within country boundaries and even within something like a state within the US and in many other regions of the world will be think about say the effect of a hot day or the effect of a cold day. In addition to the distribution of those hot and cold days and so I think, increasing the spatial resolution matters not just for adaptation planning but also forgetting the aggregate number correct, because there's so many non linearities and we ignore that. Great. I'll chime in and say, maybe first apologize that I'm seeing myself on a big screen here and I didn't button my, my collar. I think my style is still somewhat lost in the after effects of the pandemic. So in the Sierra project adaptation has really kind of been a start, because we do a lot of process based modeling when we when we think and you know wanted to develop a model to estimate, say, effects on coastal property and properties across the US, including that adaptive response was really imperative because you can't come up with a reasonable estimate without some sort of knowledge and assumption about how properties might respond to those risks. They're not just going to, you know, most properties will not just take, you know, losses and not try to do something if they can. And so, and that brings to a number of a number of sectors for example the high tide flooding and road infrastructure analysis as roadways that are on them during high tide. You know, travelers will not just, you know, sit and sit in line waiting for that puddle to recede for six hours they're going to, you know, punch their, you know, punch their address into Google maps of where they're going and find out some alternative way and so we capture all of that in our modeling and try to. So it's an important piece. Just that quickly on our side agree with everything Tim and Jeremy just said, a lot of the adaptation that we have seen will be baked into the estimate so the estimates we see in the historical data will be net of many of the adaptations. You know that that people will or won't have undertaken economics is a profession I would say is wildly optimistic on the ability of people to adapt. I, my own read of the data is that we should be less optimistic. If we get our act together on mitigation we are about halfway through the warming that we are likely to see by mid or end of century so we have experienced a lot of warming and had ample opportunity to adapt. And some data sets we see this but we see that adaptation is slow and then it's costly that's what Tamas data on mortality would suggest in the GDP data we see very little evidence of adaptation that response function has not flattened out over time. It is not differ strongly by income we can estimate it in wealthier and poorer countries and unlike in the results Tamas showed we do not see clear evidence of differential response by income suggesting that adaptation economy wide is not obviously solving this problem. So I am, I am less optimistic, given the historical data on on adaptation. Great Paulina. Hi, thank you. I guess. I'm not an economist. I do energy system modeling. So I have like potentially controversial statement. But none of the social cost of carbon estimates that are available and I've seen in the in the literature are consistent with a 1.5 degree target right there so we either believe that we need to reach 1.5 degree, which means net zero steel to emissions by the middle of the country to avoid catastrophic climate change, or we believe that the social cost of carbon is $100 but those two things are not consistent, because at $100, like at $100 per ton, we're not. That does not send the signal strong enough to do mitigation to reach 1.5 right so those two things are inconsistent one. And so the other question is, is that like, I would argue that the true social cost of carbon is unknowable. As a dollar value right. You have the bottom up models that are estimating the mortality impact and. And some of those could be monetized but there are things that we're just not going to be able to monetize like how do you monetize mass extinction events. And so, if the social cost of carbon is a novel as a in a monetary value. Like how do we first, I guess, how do we reconcile the social cost of carbon and the case and the argument of 1.5 degree is our target. And then if we don't if we acknowledge that social cost of carbon is unknowable. Where do we go from there. Thanks. I'll just add, I appreciate your comments from an EPA perspective, which has been actively involved in the energy space and advancement of the development of the social cost of carbon. I don't think there's like a, you know, particular global temperature target in mind that's driving the advancement of those estimates I think it's, you know, looking at the science and economics and the literature and saying, you know, what are what's the robust science that we can bring into the process to help quantify these changes, acknowledging that there are still tons, even the latest estimates, still only include, you know, four to five sectors. You know, of potential damages at a global level so you know I think it's an iterative process and an imperfect process but one that, you know, takes time to develop. And I'll just clarify concepts a little bit so just to be really clear on the definition of the social cost of carbon at least as far as we're interpreting and using it. It's the benefits of reducing a ton of CO2. And if that's going to be used in policy making the idea that you then compare those benefits to the costs of achieving that emissions reduction. And so that doesn't, you know, it can be used to assess a proposed policy and determine whether it's cost effective or not, but it is not as a concept to design to pick an optimal target. It can be used in a broader model that models both sides of that equation both the cost and the benefits to derive an optimal level of warming and something like an optimal carbon tax but by definition and the way that it's used in the United States it is not necessarily tied to any type of optimal target number wise or in terms of warming or or carbon tax and then I think on the point about it being a noble I mean I think there's probably some philosophical discussions to be had about how you think about numbers that are difficult to quantify but I think Jeremy's totally right that what we've been continuing to see over the last few decades of work in this space is that as we improve our knowledge and existing sectors and add sectors in general we see that the number is rising that's not true in all cases, for example our energy numbers are negative. It's really critical to continue to improve and update it as opposed to not provide quantity of numbers of these estimates if we think that there are certain categories that are going to be more difficult. Can I follow up quickly on the definition? I'll give a chance to the other questions. Okay. Okay, thanks. Laurie. Hi you guys this was so interesting I'm learning so much today and I've been a sound like a broken record but as the population person I want to ask about population again so, Tammy you mentioned migration as a first order problem and that is that's so interesting to hear and you mentioned you're starting to think about it I would love to hear what that start looks like and then Jeremy I was wondering you know migration is a form of adaptation so I wonder if it shows up in the EPA's work at all and then I was super curious about suicide. Suicide showed up on your list of inputs at some point where how did that rise to that level of importance I'm just really curious. Thanks so much. Jeremy I can take migration really quickly and then pass off to you. Yeah, to be clear, I'm not an expert in this space I just see a really important unknown and sort of under studied and not sufficiently integrated and a lot of particularly the bottom up estimates of aggregate climate damages. So, the reason that I think this is difficult is that migration is there can be forces from climate change that are either pushing migration to happen. Or pulling people back and holding them from migrating and so it's a really complicated data generating process and one that gets really complicated when all locations are getting warm warmer and suffering changes and adaptation patterns at the same time and so I think some of the tools that we've been using to empirically characterize other sectors and other responses are going to be very difficult to pour over to this problem. So there's a lot of work in this space, showing us that there's many different heterogeneous migratory responses to climate change but how we build that literature into these types of macro global scale models of where we think people are going to go when I think that's, that's really challenging. I'll continue on that by saying that, you know, like I said before, the Syrah is very much just looking inside EPA in US borders and so what's not included in is that is, is the effects of, you know, global impacts where it could lead to migration to the US. And that's, you know, a huge, that would be having enormous impacts that we certainly our project has no ability to quantify. We have looked at migration the effects of migration just within the US where you know there might be places that are more inhabitable or, or how climate will interact with existing migration changes happening today, you know, pushes towards urbanization and so forth. And what we found is that a lot of the top tier including one of the PAs, the eye clues model, the gravity models within these migration models suggest very different things and very different patterns and so the uncertainty was so enlarge across these models that we, we sort of decided to, you know, stick with meeting cases and not kind of add that additional dimension of uncertainty even though we know it's probably really important, especially since a lot and a lot of what we do in the project is particularly when we talk about changes in risks to different populations a lot of that's who's living in that area in 2090 that we're projecting the impacts to and so it's, it's a, you know, I think a very important area that there's advancements to help us understand on the suicide piece. I actually know that Marshall and I think Saul had done some work on this so they're probably in much better spaces to, to answer this question but I'll just say, we recently completed an analysis and, you know, looking at, you know, what the effects of extreme heat on on on suicide incidents. And it's very meaningful I know I think Marshall and Saul maybe you guys have done work more at a global level is that right or I can't remember sorry. I think in the US and Mexico Tama has a paper and India. I think what's relevant to suicide and what distinguishes it maybe a little bit from the other mortality outcomes that Tama showed is you see a very different quote unquote dose response function so Tama's shows that and this is, you know, classic you see more mortality really cold temperatures and really hot temperatures for a few specific outcomes suicide homicide car wrecks. We don't see that relationship we see what looks like a linear upward sloping relationship and so that has very different implications for warming than the, perhaps cardiovascular during response to extreme temperatures that really dominate the sort of population health, temperature mortality response that Tama saw so Jeremy I was actually. Yeah, also interested in and sort of glad you had it up there as a separate thing because I think we, it might respond a little bit differently, especially if baseline rates are changing. It's differential from cardiovascular so we've seen suicide rates go up 25 30% in the last two decades in this country and so the baseline rate is changing in a way that that matters for future impacts. Thank you all it's really interesting. Eric. Thank you. I didn't mean to do that. Let me do this microphone. Hi. Thank you. This is for Marshall. When you were presenting your results about the multi year impact of a one year dry, a one hot year I mean some some climatic events happening clusters over multiple years and I would think that a sequence of hot years and what we've seen more often as a sequence of drought years would have a different impact than simply a sequence of one year, one year is separated by some gap. Are you able to look at that have you looked at that things. No question short answer no the results I showed you there, assume sort of additive separability right that the years don't interact they just sort of add up. And so to the extent they do and those amplify impacts then that's not in our results so that is an important empirical question I don't know that anyone has done this yet in the GDP data so it's an important. I'm going to write that down as something we should do. So thanks. Sarah, did you have your hand up. I wanted to go back to something that Jeremy said that I wanted to allow you to clarify since this is also put in public comment. You said for at a five to 10 year level it's not possible to understand the signal out of the natural noise of climate change for events whereas we have probabilities showing the probabilities and magnitudes of events are increasing and are very much detectable right now so can you clarify what you meant there. Yeah, just that there's, you know, the effect of warming and perturbing the climate system is, you know, influenced by natural variability occurring today, and that will continue. But that warming signal becomes more clear over time. And so, not that plus the fact that, you know, the difference in emissions between policy cases and reference cases take time to show up in the climate system because of delays in the climate systems like there's, you know, you know, the greenhouse gases are very long lived and, and, and have, you know, lasting effects on our climate system so when we look at damages and impacts in the climate impacts world we tend to focus on timeframes like 2050 2070 2090, you know, end of century many decades out. And if someone wanted to know, you know, typically climate scientists would be more unsure answering questions of, you know, what are seal arise damages in 2026 going to be different than what they are today. I think most climate scientists would would have some some concerns that at tweaking that out. That said, there's a lot of other people in this room who could probably answer that question, either better or add to that than me. Do you have a debate online. Yeah, I'll clarify your answer and that is, there's natural variability, and that is phase so sometimes natural variability is in phase with the secular trend, and sometimes it's not. And that great effect is, you don't know right you can predict that so you have to assume that if it, you know, like the linear linear cycle, which is now going to be if there's a linear which will now be worse for the US. So I think that's an important nuance but I think we do know that the secular signal is now bigger than the national variability. That's one of the consensus in the science community so I think what we should be clear. Let me ask so I'm a climate scientist. I'm a measurement person and even climate modeling is very hard right. And so the way we dissect the problem is I call it case studies right that and the case studies we use our natural variability. So whenever there's some event, or a phase change between nature and secular finding we try to probe it and dissect it by simplifying the case. So in your impact human impacts scenario. I can see examples right for example hurricane Katrina had a devastating effect that's a case study that can calibrate in the US the models. There's this whole issue of people still moving to live in Arizona right, despite all the water stresses, people still moving to the West Coast it's very different in other countries where extremes hit and you have no choice. In America you still have resources and choices. So these are the case studies. I know we have a framework is hard but I think the more and this is happening in the climate community. So I presume I'm not an expert in your field, but the same thing needs to happen more. So you can get more robust disaggregated segregated cost, you know, cost functions or loss functions for all these things. This is just my approach and this is new to me but I think this is where actually the climate scientists have the same dialogue so the language is very similar the fields but I think we are a little bit more mature right we don't wait for warming for another two degrees and wait for oscillation, the Elanina cycle for example, and then we probe the carbon climate feedbacks, and you do the same with people right. So it's a very analogous system. So I just had to say that, and get your thoughts I mean you're probably earlier in the game climate science is what 2530 4050 years old since keeling, but, but you're you guys are nearest in North house and you're and you're you're you won't have enough so case studies calibrations contrast will be very important. Thank you. Anyone want to. Yeah, I really love that that question I think you're, you're right both that we're younger as a discipline but also that there's a lot of analogous conceptual frameworks behind what we're doing I think. Essentially, you can think about a lot of our statistical modeling as investigating many of those types of case studies in a larger data set and pulling out statistical relationships from seeing many different heat events in Chicago summer and many different cold events in Northern Europe. But I think you're also right that as climate change continues to progress in the benefit of that on the researcher side is getting more of those case studies or those observations in our data set out at the tail ends of some of these distributions to begin to parameterize those curves that we have to take out out of sample into the future and so you know Marshall does really exciting work on wildfire and the more that we're seeing these events the more data we have to begin to build up a systematic understanding of those case studies particularly at the edges of our distributions where we really need to parameterize carefully and we think about long run features. Yeah, thanks. Tim Linton climate scientist in the UK point of information surrounding sea surface temperatures are about four signal above the mean for June at the moment, and we're sweltering here. So I would love to hear, saluting all your great work I'd love to hear your thoughts on how we can push the frontiers of assessing the extremes and their impacts. But also I hope I'm not repeating souls original question. I know I take the point that we're not seeing necessarily evidence of adaptation within your econometric analysis, but to what degree, are you using this, or is anyone using this rapid advancing impact assessment to guide adaptation proactive adaptation intervention or spending your money to alleviate some of those impacts. Thanks. I can start just quickly by saying Tim I think exact that's exactly the right question I don't know yet who's using these estimates to guide out a patient I hope they do I think this is, in addition to contributing to what the CBO, CBO or CA is going to do on the budgetary side. I think understanding impacts and indeed understanding impacts of extreme events which I think Tama's group is actually doing a really good job of doing we're looking at historical extremes which might not be extreme and future climates but they're absolutely looking at extreme events and measuring impacts on mortality and other outcomes so I think to some extent that you know they are doing that pretty extensively. Absolutely right. You know this should inform budget it should inform our ambition on the mitigation side but it should absolutely inform adaptation priorities right because if we're seeing death that means we're not adapted and so we need to get our act together. Who's actually doing that I don't know. Jeremy. Yeah, I can take that a little bit. You know, again, we, we approach the adaptation questions because and we knew we knew we needed to include adaptation because we needed those assumptions in order to have robust estimates for purposes of determining climate damages. When we had those estimates a lot of people would reproach us of like oh we want to understand like what climate change means for how much money we're going to need to repair highways and roads in the US or for coastal damages and our coastal, you know flood insurance programs and it was like oh whoa whoa whoa like that's that's not the analysis we did like we didn't do like a, you know what does climate change mean for the flood insurance program like I don't know if I feel comfortable like it reformatting that program based on our estimates at all. So, you know, and I was actually just speaking an hour ago to someone in the room about how you know a lot of these, there's just not a lot of estimates out there. For a lot of these different sectors and so I think that's a an additional call to action of how we need to advance that side of it a lot of the damage quantification has been sort of more geared towards. How do we describe climate damages rather than rather than what are the economics of adaptation and, and so there, you know, that should be a, you know, I think an outcome of this, this workshop is a recommendation to really advance that. Rachel. Yeah, thank you. This has been just really a wonderful session. I just wanted to circle back to the question Sarah asked because the there is a clear signal already and the signal is relative to the baseline, you can use say the know has their most recent 30 or climate normal baseline. It's not about 2026 versus today, it's about the fact that already today we've warmed a little over degree Celsius and there's a very clear climate signal. And the attribution science is solid right and it's getting better all the time in the last 10 years there's now rapid attribution when you have these extreme events like the Pakistan floods or heatwaves. I think, especially for an audience like this it's really important to be crystal clear that the signal is there, and the science is making the connections. Very clearly between events that are happening already around us. The question I have for Marshall and Tama is, you know I know you have to rely on data sets in the past, but we know going forward that they're significant non linearities that the climate science is showing a likely to be the case so the last one degree Celsius increase in temperature is is going to look like child's play compared to the next degree Celsius rise in temperature, every fraction of a degree from your own out is more consequential in many ways including to the economy but certainly to people's health. So how can we make sure even though we can learn some important lessons from past data that we're not getting overly anchored. We're already seeing that the insurance industry, it's the canary in the coal mine right it's already showing that the past is not a good predictor of the future. And then finally we have this mismatch where, while you, while it might be true that in the next five years, we see significantly worsening impacts but they're still not as bad as say mid century. We know from the IPCC reports and a whole body of science that these next five to 10 years are the critical years for the actions we can take to stave off some of those worse impacts. So, how can we make sure that we don't have this policy mismatch that we're waiting for the bad stuff to happen before we act because we say it's not happening in the next five years. That is our window to do a lot of things, not just on mitigation but also on resilience. And how can we collapse that future into the present because that's what if we have good information and we can bring that information into market incentives and policy incentives. Now, we can create that future right where there isn't that mismatch in time horizon so curious what you all think about the potential to do that through this workshop and others. You can start. Thanks Rachel. Super important points. So, we should think about two different types of non linearities non linearities in the response of the climate system to forcing and then non linearities in social systems in response to climatic change right and I think you're absolutely right that we need to be able to quantify both of those things. We're going to leave the as economists were going to leave the first one of the climate science folks and so that's their job to figure that stuff out. Our job is to not screw up then our estimates of what's going to happen to the world, given those non linearities right. And I think you're absolutely right that we're seeing evidence right now that past is not going to be guide to future so wildfires is one thing in California you you can help but study if you're a scientist in this area. And we've seen just in the last month just a complete unraveling of the insurance market in in California in response to wildfires you basically the main insurance are leaving the market it's just it's too risky. And, and so that I think goes to your point. And so that's one example of where it's going to be hard to really study the past well and say smart things about the future, because it's changing really quickly. And I think in other settings I don't think we should be so pessimistic so I think in the mortality temperature response right we've seen a lot of pretty extreme events, historically. And we have a lot of mortality data around the world that Tama and colleagues have put together to allow us, you know pretty good insights into what those extremes are going to look like and climate science has studied this a lot so I think in some settings we shouldn't throw up here we don't have any evidence on what the future might look like. But in the wildfire case and another really rapidly changing new climate impacts. I do think you're right that historical data are not a great guide. I'll just jump in on a couple points I agree with everything Marshall said and I think what what can be really critical there it's finding sort of the intersections in today's distribution of outcomes that allow us to think most responsibly about the future so for example in a lot of our projections we're thinking about places getting a lot hotter but also likely also benefiting from substantial economic growth so if we don't have data and places that are both hot and wealthy it's a lot harder for us to reliably make those projections about the future maybe and so just a working really hard to collect representative data which has improved a lot in the last few years, but also being transparent about when we're having to rely on out of school versus, you know, having experienced some of those events in the past and having, having more data to support that. And then on your second point about, you know, urgency of the present moment and not always focusing on projections in 2075 and beyond I think that's a really important one. On the one hand, just the construction of these dose response functions that we're talking about already begin to shed light on the damages of climate variability today. And maybe we could do a better job in our field of communicating what that means but just the fact that we recover you know Marshall's inverted you is giving you insight already into what happens when we get an exceptionally hot year in a tropical country right now. And even before you start projecting things forward and then I do think that the related to the prior question the science of impact attribution needs and should be given a lot more attention and time I know people in this room are actively working on it, but we can take some of the approaches you're seeing today and also present results for how damaging climate change has already been for mortality for labor for energy. And as developed as of a science as the future projections that we've been doing and I think that's important for our community. All right, we have two minutes left. James. Really interesting panel. This question mainly for Jeremy but I would welcome other people's thoughts as well. There's already been some discussion in here about the relative benefits of bottom up versus top down econometrics. A lot of the results in Freddie are inside process based models, their strengths and weaknesses to all three of these approaches and in addition, the approach effects for the station. I think that Freddie has sort of such an incredibly comprehensive or approaching comprehensive representation of all these, how is the PA or group thinking about combining the strengths and decisions of these different techniques to get at a better sense of best estimates. That's a great question. I think our first objective is just to characterize and include as many different impacts of climate change in Freddie as possible right like we're trying to cover the, you know, as much as we can. And so, there are both econometric and bottom, you know, and process based models within Freddie. There also needs to be intentional overlap right so you know there are I think one or two instances where we actually have an econometric approach to a specific impact and for that same impact also have a process bottom up and so having those differences is really important and something that we need to further build out right like just because I showed those 20 sectors it doesn't mean that I wouldn't want to have you know like that that that therefore, you know for road infrastructure we're done. You know and we don't want any more like no that's the exact opposite like if there are other road infrastructure studies send them to us and we'll build them build them in and so. But for as for like deep I think part of your question is like more deeper dives into looking at the relative differences and structural differences between them. It's not, frankly, something that we have a lot of bandwidth to do within within Freddie but it's something that's important for us to do. And we just haven't gotten to as much, unfortunately. All right, well, we're at time. So thank you. Thank you to the speakers. And what's our, what are we doing next project. Yeah so we're opening the slino one more time so people can add their ideas if you do have to hop off I just want to say that. Well first of all it's been an amazing workshop I've heard a lot of really great conversations and presentations. And for tomorrow we're convening at 930. So rather than 10 at 930 tomorrow. But otherwise, yes, I'm happy to let you all discuss and add your ideas to slido.