 All right. Hello, everybody. We'll go ahead and get started for our afternoon presentations. I'm going to keep talking, but I will filibuster a little bit while people take their seats. So after what I think was an incredibly interesting morning, we are now going to turn our attention to a broad overview of the state of macroeconomics and how it more with an eye toward how it incorporates climate and how macroeconomic models, the challenges that macroeconomic models have in incorporating climate. We have a big chunk of our panel is joining us virtually. So we'll be going back and forth a little bit. But I'm going to hand things over to Amy Nakamura. Amy, I'm here there. I can't see you, but I bet you are. Thank you very much, Wendy. It's a pleasure to be here. My name is Amy Nakamura. I am a macroeconomist who works on topics like monetary and fiscal policy, but I'm also been involved a lot in topics related to macroeconomic measurement. So how do we measure things like GDP and inflation topics that came up this morning? The purpose of this session is to sort of provide an overview of the different kinds of macroeconomic models that are out there and how they can contribute to the debate and discussion of issues related to climate change. Just by way of introduction, I would imagine, since there are many scientists in the room, that one thing that could be a little jarring for scientists say, you know, familiar with the standard model in physics is that all these macroeconomic models are approximations. And for that reason, macroeconomists tend to use different models for different questions, like, for example, long run growth questions versus short run questions related to business cycles, monetary policy or fiscal policy. I think this sort of heterogeneity and scope of the models is something that will hopefully give you a sense of in this session. So the logistics of the session, each of the speakers will have eight minutes, and at the end we'll have questions just like in the session this morning. So let me start with our first speaker, Jim Stock, who is joining us virtually from Harvard. Yes, I think so my driving or are you driving. I think you're driving. Okay. Oh, I've got a I can't share the screen. We're trying to figure it out Jim it might be that we're driving. That would be fine. It's the same deck that I sent you. Jim, we're going to control the slides. Sounds good. We ready to go. Yeah, just say next slide when you want to go to the next one. Terrific. Okay, so, so thanks very much. I'm looking forward to this. Let's next slide. So, the, the first few decks the first few slides I have actually are essentially a synopsis of what occurred this morning so I'm not going to spend much time on that at all it was a terrific session. And it really laid out, I think, a very practical in a very practical way the task at hand. So let's just move to the next slide. I wanted to just do some level setting me mentioned that there's a lot of heterogeneity in the different types of models and different considerations come in at different horizons so what we have here is just a just a fixed ideas we have a picture of real GDP and employment. On the left and native units and on the right and logs. I think what you can see the main point of this is that over longer term horizons over decadal and multi decadal horizons if we think about, you know, long term challenges. The whole name of the game is what's the growth path going to look like and what's the uncertainty in the growth path. It's certainly not a constant or even growth path but it's but but that's what really drives it at the other side. If you're thinking about business cycle analysis and using things for monetary policy, or potentially thinking about systemic stability questions like Adele Morris was talking about that really gets into these more these gray bars and these business cycle fluctuations so it can be very large, but temporary and ephemeral compared to the long term growth. Let's go to the next slide please. So, on this actually is probably not even worth spending much time on this. This just it turns out it's just a summary which you can have of what the discussions were this morning about how different horizons address different questions. Because there's different horizons and different questions. That means that the models that are going to be used are going to be different so that underscores this point about heterogeneity of models. One thing for those of you who are not sort of steeped in the acronyms, I have to have a little chart here of who the various players are in this game or the leading players are in this game. Next slide please. Okay. There's one point that I think I'm going to flag here and this was implicit or this actually is something I asked in a discussion. I think a question for us. I have an opinion but we can, I don't, you know, my opinion isn't important I think it's more that it's more that this is something to discuss. The discussions this morning focused and the modeling efforts that the government has done so far which are terrific has focused on going from climate risks, both physical risks and transition risks to their impact on macro models. For example, one might build into a baseline, the effect that we're going to have some hits to GDP, and that's coming through a variety of different channels productivity and capital accumulation and so forth we're going to have some hits to GDP as climate impacts get worse over the coming multiple decades. That's something that one would incorporate into the baseline and that's the red arrow going from climate risks to macro models and then given that baseline, there's going to be a lot of things that one computes having to do with labor markets and financial markets and so forth. So you can almost think about this as just being an exogenous shifter due to climate risks. A question is a question on the table for us to discuss is whether to close that loop to go from macro model and say macro policy considerations back to climate risks. So if we pass the US passes the IRA, what does that do to the climate risk path. That's would be a province of an IM integrated assessment model. That is I think technically feasible from my understanding of how the CBO and OMB Troika models work. It may or may not be something one would want to do. I think I'm just going to flag that. There's a good discussion of GDP this morning and I think I'm just going to leave it at that because there's a couple of other points I want to make so next slide please. Okay, here's where it gets complicated, I think, which is, it's easy enough to think, I'm sure that the climate scientists will disagree, because I'm not a climate scientist, it seems easy enough for me to think about physical consequences like sea level rise, especially if I'm focusing on a 10 year or even the 25 year long term CDL projection. And you can think about some of these consequences. And I think one of these issues some of these direct physical effects which is what I put in blue as being things that would affect a baseline of longer term growth, and you can see how you can enter those into certain models. I think that things start to get complicated when we start thinking about the sorts of questions that Lars was asking when he asked this morning about uncertainty and variability, and, and events we don't have much, much understanding of I'm going to offer up a way to partition this and this is going to be imperfect. And I suggest partitioning these effects as on the one hand long term growth effects. And then the other hand, effects that would affect the distribution of shocks in the future. Now we don't really know either of those there's a large literature saw Sean and others have been leaders in in that, looking at the effects of temperature and precipitation and other climate variables. And GDP and other important economic measures of economic activity and measures of welfare. That can be thought of as long term growth effects. And secondly, there's questions of energy security policy uncertainty energy price volatility over the course of the transition effect, especially if the transition the economic transition is a quick one. And that I think can be thought of as changes in the distribution of shocks. I'll have to say that, from my own perspective I'm especially interested in some of those issues, those issues are directly pertinent to shorter term projections and to some of the micro macro macro macro issues that the Fed is facing next slide please. So here's a fan set of different types of models. So, on the left is an example of a purely empirical six variable model. It's used by the this been used it's out there in the literature to look at the effect of oil price changes on industrial production in the United States. So that's an example of a type of model, if you had scenarios or distributions of shocks, you could apply that sort of model to look at the effects of that on GDP. The intermediate one is a CBO model, which is looking at long term budget implications. And that's just taken from a recent publication of the CBO. That's a big model has hundreds of equations they report. They have hundreds of equations because they have many, many variables that you need. When you're doing the CBO budget detail analysis, and, and that's a that's a really completely different model for a different purpose that baseline could also be adjusted. Jim, we're on a tight time schedule, so almost at a time. Oh, I'm at eight minutes. I'm sorry, I'm not paying attention to that the upper right hand corner is is a long term growth identity and ways that the climate can affect that is going to make a lot of difference in terms of long term efforts. Okay, so why don't I just call it quits there I think the question is where we can put in all these different effects at these different horizons and these different types of models. That's good. Thank you very much and I think that was a very good transition to Lars Hansen from the University of Chicago. Lars. I need my slides somewhere. Okay, let me get started here. This is largely about so I've been very interested in in model building and in uncertainty. And so climate change really brings both those things together. And this is all about I guess what Jim called closing a loop which I think if you really want to engage in formal policy analysis it's important to consider closing that loop. So let me just start off that saying that modeling large scale macro economic systems let me just indicate what I see is has been the conventional practice first of all we have economic agents individuals enterprises and other entities. And these differ from physical particles and so the sense which you know the economics models are different than physical models, because they're forward looking, I mean we can debate exactly the nature that forward looking behavior but there's forward looking behavior when you want to think about making an investment you're looking forward into and trying to guess what's going to happen in the future. That makes the analysis of the models more challenging. And because they have to incorporate this forward looking behavior. Many of the existing macro economic models are right now are approximately linear this opens the door to numerical methods that are tractable to implement at a very large scale. Many of the models that are analyzed are basically approximations around balance growth paths. And so, so, yes, so yes I did it was kind of this nice growth path I can think about doing approximations around that. And some of the challenge modeling challenges get sidestepped by a considerable use of loosely structured models aimed at capturing empirical patterns and potentially dynamic responses to macro shocks reflected by historical data, but but but but they also limit the type of policy questions that can be addressed. As been mentioned by others I'll go fast on this one is that no one size fits all and in terms of macro economic models, some have sectoral richness some really focus on micro economic head region adi and the role of micro economic uncertainties. A small number or fear are highly nonlinear. And most of these have been tailored to date to the study of financial crises, but but they're otherwise very very highly stylized. Um, there's an interesting contrast between macro, the conventions and macro economics and the in the conventions in our related literature called macro finance modeling in macro models and aggregate uncertainty as often has second order implications, maybe micro economic uncertainty is important but the macro economic uncertainties are treated as more or less a second order type type calculations and in macro finance models are going to certainly is necessarily a big deal because that whole literature is about well, how do financial markets respond to air good shocks and so on. So I would uncertainty is featured there. Long term to an economist not to a geoscientist when economist long term is like you know, three or four decades not like, you know, many centuries. It is featured in a substantial body that literature along with uncertain extreme events again because because this has important consequences for valuation. Decision under uncertainty approaches had had have actually been much more prominent in macro finance setting than a more standard macro setting. To my taste there's valuable tool that modeling tools from that from the finance macro finance tradition that can be imported into the modeling of macro climate change linkages. Now, let me now turn to the challenges that I see posed by incorporating climate change considerations in this in the macro modeling. First of all the empirical challenges. We're going to be pushing economy into places that does not experience historically so this was idea we look at it rich historical data sets and they just tell us the answer and make us you know easy to calibrate the models going forward is just it seems like it's off the table. Yes historical evidence can tell us about parts of it but it's going to miss on other parts those models we really have to make these four looking extrapolations we really have to lie quite heavily on models. There's additional computational challenges because climate change is really about long term transitions. This idea of proximity around growth growth paths is off the table because we imagine that, because it's really all about this transitions to how we may eventually get to economies without fossil fuel emissions or the like. And so, so the thinking and how we do the both model solutions approximations and the like really are different. I think it's important that we incorporate new sources of where your court right on the air good uncertainty and I think this needs to be a first order consideration, including human impacts on the environment including economic adaptations to changes in the environment and the like. Now for an economic model builder interested in uncertainty there's two perspectives which you have to address. The first is there's economic agents inside the model, they, you know these business firms investors, they're facing these air good shocks is air climate shocks and so inside the model, we have to think about how they are responding to these. And then the model builders face the usual more kind of statistician outside the model types uncertainties which were kind of familiar with from various statistics literatures. So I think it's important and they kind of interact and potentially interesting ways, at least that's in my view. So now related to discussions that we had this morning is, I think it's important to kind of modify or extend the language that we use in terms of uncertainty and relevant for these quantitative models. And this is especially true with climate change I think it's also true with pandemics and others it's not just not just a climate change issue. And so I want to draw distinctions between three different constructs. One is risk. So, you know, lots of discussions right now just just make risk be a grab bag of everything. So the risk is as taught in economic classes is, you know probabilities and you don't know outcomes are almost out of time. Yep, this is my last slide. So, so I will finish soon. So each model has random impulses but they're in and so they give us a source of risk ambiguity. So, I mean, basically that these are situations where where multiple models should be consideration that should be under consideration, and each can give rise to different, different implications. So ambiguities this uncertainty across models now statistics has wrestled with this, but, but we have to remember especially in this climate realm there's a lot of subjective inputs going into things, and then there's uncertainty in those subjective inputs. And finally model mis-specification. Every model we write down and we mentioned this is just some rough approximation, each model is an abstraction and not intended to be a complete description of reality. There's a recent, there's decision theory that's emerged in the last couple of decades that provides ways to try to formalize these types of concepts and distinguish them. And it's trying to make operational notions of deeper radical uncertainty. But this, this is going to require, this requires refinement modifications to uncertainty quantification. And as I argued earlier today, from standpoint of policymakers on the table, it becomes very important we think about uncertainty aversion how adverse should society be to these uncertainties. Thank you. Thank you very much. Our next speaker is Saul Shang is also virtual from UC Berkeley. Great. Thank you. You can see this right. Okay, so I'm going to very briefly speak about sort of the larger landscape of how people are integrating climate change into economic models, focusing on those that are thinking about the impacts and much less on the transition risks. So just to give everyone a sense of where we're going as a community and what has been done so at a very high level, I think it's very useful to decompose the models into sort of three classes. There's one models with endogenous equilibria I think a lot of what Lars was talking about are these models in which you have a closed model, where you sort of specify the entire world and the model searches for an equilibrium that is internally consistent. Some of these are I am model are known as I am models, but there's also models known as I am models that I would call applied policy models that are scenario driven. These are designed to answer specific policy questions. They take a scenario as exogenous. And what's useful is that they can be globally comprehensive and account for a lot of different things. On the downside, like their models are not closed and so the models might be leaking somewhere and we don't actually specify what the entire world looks like. The third class of models are empirical models and so there's been a lot of exploration of real world relationships between different X's and Y's. And that's I think what any people in the room are probably familiar with. These are difficult to connect to policy because it's mostly specifying empirical relationships that don't necessarily map on to directly on to what decision makers are trying to do. So just at a very high level I actually that sort of as the take home prize at the end is you get this map of the literature. So it sort of think about this family tree of models on the left there's empirical in the middle are these and scenario driven, and on the right are these equilibrium models. And so there's been important progress, you know, thinking about sort from, you know, by Mendelssohn's work and the empirical side, the scenario side in the middle here, and models like the dice model on the right populate this for you all so that when you look back you understand how things are connected. So if we think on the right hand side focusing on the endogenous models, you know the most famous one is the dice model, which is completely written on the screen for you it's a elegant beautiful model that helped us think through sort of the optimal planning problem of climate change. On the one side on the downside of having such a complete model is that it's got to be very simplified in order to be solvable. And I've also used these models to help think about uncertainty is that he was saying much more eloquently than I can capture, and that has become an important factor in terms of thinking about what is the value of of emissions today. And then a new branch of the literature building on this has been looking at models that are internally consistent and closed but contain information about physical space. So these are called spatial equilibrium models, and they're essentially like expansions of the dice model with many many pixels in which the individuals don't want to move between pixels. So, within this kind of equilibrium pillar. There was this very important work by Nordhaus in the beginning that gave rise to a large number of these models but also it was kind of hybridized with some important points made by by Marty Weitzman about the deep uncertainty that we should all be thinking about and so that's been combined in many different ways, and that has really developed into a clear branch of the literature. Also within this space are models that think about public finance problems and so whether or not we should be investing for example in innovation or thinking about distortions that come from different regulatory practices. And then finally thinking about space, you know the rice model was the direct descendant of the dice model trying to break the world into regions and then since then there's been the spatial equilibrium models. On the other side is the idea of having spatial having equilibrium in global markets and allowing for trade to occur and so this is kind of a different literature, but is also thinking about space. And so on to sort of have this whole literature thinking about these internally consistent models but is a little bit disjoint from some of the other models that are out there so the scenario driven applied policy models, sort of break the world into a few number of sectors and think about different impacts that can be specified from prior literature. And one important recent advance in this has been actually work that Jim has done thinking about developing sort of more internally consistent scenarios that can be used to drive these models and so that's something that I think people are really thinking a lot about now. So the page model gave rise to the stern report that had a lot of policy impact. These other models have gone and been sort of interwoven in a mini platform developed by David anthoff at Berkeley, and James James work sort of feeds into that. But thinking about the empirical side pillar of the research agenda. There's folks like myself who spend a lot of time looking at data and thinking about realizations of weather that are being drawn from the climate and trying to measure some sort of dose response some effect of some climate on some social outcome. And what that lets you do is it lets you develop a projection of some distribution of future social outcomes and a lot of people have spent a time has been time recently trying to understand how these dose response functions might evolve as the future changes and as populations adapt. So in this literature, our findings you may have seen so for example that like crop yields in the United States decline rapidly at around 30 degrees, or that labor productivity declines at high temperature which has been referenced here earlier, or that in some cases like violence among individuals can increase at high temperatures. Important to this group audience there have been several empirical findings about the growth effects that have been empirically derived but don't have a clear sort of theoretical foundation and reconciling the theory with the empirics has been a major task. So in this literature, there's been some empirical sort of branching of it thinking about the micro micro impacts thinking about macro impacts. And in that literature thinking about how to model adaptation has big a big innovation so there've been many papers developing that concept. And then when you think, sorry that I'm bouncing around in this tree here, it'll all come together at the end. You know, there's been some work on how do you that I've done with Bob and how do you account for all these micro level impacts to bring them into some policy decisions but it's not applicable to all policy decisions. And so in some new work that some of us at the climate impact lab have been doing. We've been trying to develop approaches that can bring together what we know about adaptation into these large scale empirical accounting frameworks and so this is the decent model that many people have probably seen. There's also been really important crossover studies. So for example, friend more in Delavane Diaz did an important study where they tried to take some of these empirical relationships on macroeconomic growth and embed them in these equilibrium dice models. Okay, and so there's been some crossover studies trying to sort of mix and match these models. There's also been some more recent work trying to take some of the global trade models and apply empirical approaches to them. Okay, so there's been some mixing. One minute. And then. And then, as I was saying so we at the climate impact lab have been bringing together have been doing some of this mixing as well. So modeling hurricanes and taking sort of atomistic behavior to model sea level rise impacts for example. And then one effort that that we've put a lot of time into has been how do we take these empirical findings from those models and integrate what we've learned about valuing uncertainty. And so that's sort of the climate impact lab has been focusing on trying to weave together a lot of the findings from the literature, but it's clearly not getting all of them and there's still a lot of other things that either can be brought into this model or have to be pursued with other models. So that this is just from what we're doing in the climate impact lab. So there's kind of like the take home map that you get for yourself and you can sort of refer back to it with references. But I think, you know, on the left hand side is this empirical literature very focused on measurement. The applied policy work is kind of scenario driven in the middle, occasionally drawing from what's empirical and in a few new cases, bringing in work on uncertainty. In terms of equilibrium models, there's sort of focus on pricing and valuing uncertainty, thinking about public finance or spatial equilibrium and trade. And so, you know, this is for people to refer to as they're trying to like navigate what's been done out there. Thanks. Great. Thank you so much. So next we have Eric camp Benedict from the Stockholm Environmental Institute. And he's in person. I'll just wait for my slides. Okay, great. Thanks. So my goal with this presentation is to share to try to give a sense for people who might be using some outputs of macroeconomic models what what's the scope of such models and I've got to figure out how to do the next on my clicker there it is. So, and a major point I want to make that's already been conveyed here is that there's an enormous number of different kinds of models. And there are different ways to characterize them. I'm going to choose purpose structure and assumptions. And I'm going to use this by focusing on six widely used models dice which you've heard a ENV growth times macro FRB us Rami E3 plus and E3 M E. And by doing this, by focusing on widely used models. I'm going to rule out some that I think are relevant and interesting. One is a class of models called stock flow consistent models. These are being used much more in Europe. And they, they are particularly useful for looking at finance. Also context specific policy models so ones that are developed for a specific policy problem and that's actually the kind I normally work on. And then policy relevant models that have appeared in the literature that look pretty interesting and promising but they're not, they're not yet widely used. So purpose. This is crucial as as was emphasized in the first panel, why is a model being built. And here, here are the statements dice estimate the optimal path of reductions of greenhouse gases. Some unintentional duplication there and words. This is actually one of the ones those who know the SSPs this is used to generate some of the SSP trajectories to project future levels of global and country specific GDP and income times macro study the interconnections between economic development and energy demand FRB US forecast and to analyze macro economic issues including both monetary and fiscal policy and both remmy E3 plus and E3 M E with to look at energy and environmentally relevant sectors and see what happens to the larger economy. So that was purpose in terms of structure, as was also mentioned models differ in the number of economic sectors they're spatial detail time horizon number of household types environmental impacts of any, how they treat finance and so on. So the first three of my example models are long run one sector growth models times macro is medium run but long, longer run single sector growth models but then they have specifics dice. It's a global model coupled to a climate model. Envy growth is a single sector model that is applied to as many countries as possible in the world, and then those are used as SSP outputs times macro. It's a multi sector but it's got a bunch of households, and it's linked up to a physical energy model that is greatly detailed FRB US. It's a it's a short run business cycle model. If I got that right, I have someone in the room who could correct me if I got that wrong. But it's multi sector it's it's for policy analysis so it's very detailed many sectors in fact they look at multi firms multi household and detailed government policies. And both remedy E3 plus and E3 me are large simulation models multi sector multi region multi household with an energy and environmental impacts, and while they're long run, they also simulate short run fluctuations. Assumptions. Now this is key models differ significantly in their underlying assumptions. And what those assumptions are depend on the purpose of the model but also on on the prior knowledge and understanding of model developers which comes from their training. They're ongoing study their own research and the broader research in the in the broader field. So in Dyson times macro households optimize discounted future utility which depends on household consumption so it's a consumption based utility. Nothing about, you know, wanting to act on climate for example and as calculated by the model. Envy growth follows a different tradition and countries conditionally converge towards a long run frontier. For B us households and firms optimized but there's actually two options. You could optimize according to the simulation of the model or you could say it's based on imperfect understanding a possible future trends. You can have have wrong assumptions, and both remedy E3 plus and E3 M E is a simulation model in which households and firms respond to current conditions and this sort of model allows you to do those non linearities without a huge computational burden because you're not trying to do inter temporal optimization you're taking it forward step by step. E3 plus focuses on marginal changes whereas E3 me allows for path dependency and structural transformation. Unlike the others. E3 me is also a demand led model it is the only demand led model in this example. The models I built tend to be our demand led tend to be. So in a supply led model investment is constrained by available saving. So in a supply led models investment is planned to meet anticipated demand. So it's a forward looking bet, while banks largely accommodate demand for loans. And technical terms the supply led ones are often connected to exogenous money models and demand led to endogenous money. So we can determine differently in these different kinds of models with supply led prices and wages being assumed to clear markets whereas in demand led prices are set to cover costs whereas wages are socially influenced and so they, you can end up with some conflict over setting wages. One more minute. Perfect. Thank you. But this is my last slide. So, some questions, a potential model user might ask of a model or or a model, what's its purpose. What policy questions can it address. What should it not be used to address and at that point you just may say I'm not using this because it doesn't fit. Is it one sector multi sector for a lot of climate questions, you have to have certain sectors in order to make sensible decisions. There are multiple households if you care about distributional impacts, you have to look multiple regions. Does it include energy or environmental accounts. Does it include finance. And does it assume optimal behavior and if so, who is optimizing and what do they optimize. Do you allow for non optimizing behavior and if so what, and is it mostly demand led or supply led. Thank you. Thank you very much. And our last present also in person is Satya Gopala Krishnan from the Ohio State. Thank you. I'm actually online because I wasn't able to make it. And I think my slides are up there. Just wait one second for the slides to show up. Or should I, should I use my slides. Sorry just one second Satya. The problem with going last and I thought about it is, is you often have less, very less to add to the incredible set of comments that we've already heard. And so I'm going to try and come at this from just a sense of what you know how we define a coupled system, and how we might think of these coupled systems at different scales. And essentially because, you know, if we're thinking about what might be the potential challenges limitations and opportunities to do to figure out where we might want to couple human and natural systems or economic processes with with physical processes. I just want to think about what those questions are at different, at different spatial scales. So that's, that's really what, what my few minutes focus is going to be. So the next slide. You know what we mean by a couple of human and natural systems are the systems that evolve simultaneously, and that depend critically on the interactions or feedbacks across both, you know, geophysical socio economic political processes so to what extent can we understand these processes that are happening simultaneously and you have processes in the geophysical sphere so I have I have here an image of a coastal town, where you have geophysical processes that are driven by climate induced sea level rise sea level rise might lead to erosion increase storms. And that is in the geophysical world that is that might mean how sand moves around how waves move around. What might be going on in the ecological sphere with, you know, do you are we building dunes are we putting, are we putting more sand on the beach what's happening to the two aquatic species. And simultaneously, there are economic agents making forward looking decisions about where to live where to buy homes where to buy second homes where to go recreate. And that contributes together to the economy of say a tourist tourism centric coastal coastal town, and that then scales up into what might be going on at the at a regional scale and at a at a national scale. And we're when what we're interested in when we're trying to model these coupled systems is to what extent can we effectively consider the dynamics of these systems of systems that have interactions at across multiple processes and at different time scales. So what are the human relevant time scales at which we can incorporate some of these feedbacks to give us insight into how the system as a whole will evolve. And that will help us set those baselines for what is going on at these regional scales to potentially be inputs and informing any of the different models that that we have that we've discussed so far. So in the next day in the neck on the next slide I'm looking at these feedbacks and thinking about. So these are mainly open questions here for us for us to discuss. And the first set of questions which has already come up in in practically what everyone said and there's a few new things you can see here is that we're worse. We're essentially grappling with what are the temporal and spatial scales at which you can couple some of these models. So it goes back to the first questions that Jim started out with is how do you close the loop and where do you close the loop. So in my mind, you know, some of the work that I've been doing with with, you know, some of the interdisciplinary work with coastal geomorphologist is looking at this regional scale and thinking about, you know, what are the the feedbacks that we can incorporate, which are the feedbacks that are actually relevant at the time scales and the spatial scales that that are that we care about to answer the specific question at hand. We've already said there are different models for different purposes. So for the purpose for a purpose at hand, what are the types of feedbacks that that we that we should be considering. And each of those may involve processes at timescales and spatial scales that are very different. The second sort of big question at hand is how do some of these processes that that occur at at smaller could be really micro or it could be regional mezzo timescales cascade over space and time so as an example. Say we have a storm a storm affects the town, and then people decide to migrate so populations change, and then accounting for those population flows would then would would be an input into into determining what that growth path might be. In another town, we might be we might see behavior where individuals are investing in some sort of defensive expenditures either building a seawall putting putting dunes doing something to mitigate the impact of of of a storm, and that might lead to a different emergent attractor in that in that system right so so one of the questions that that is that is both interesting really at hand is how do we figure out which of these dynamic feedbacks cascade over time and space and which ones do we do we account for in building out these large models right and how do we how do we reconcile the differences in the in processes that occur at these different temporal and spatial resolutions. So that is one big question that I think has come up many many times. And then on the last slide, the next, the next slide. The second, I think in my mind, big challenge is how do we look at the trade off between what you know the completeness of everything that we want to incorporate into a decision into into what these investment patterns might look like, relative to how complex we want we can make we can make a model. So which parameters if we can empirically estimate some of these feedbacks and model what the, what, what the trajectories and the evolution might be, which of these feedbacks would actually affect something at a global scale or or a national scale so which parameters do we really need to identify that that have that that have an impact in terms of whether the system is is going to destabilize or whether there are ways to see how we would respond in the in the face of uncertainty and so on. And the second set of you know what I think are both is both an opportunity and and a potential challenge is how can we utilize data richness given that we have data now on what people are doing being able to downscale some of the models at these regional scales with some micro level richness understanding how these feedbacks play out. How can we use these data, these data to inform the X's right in our model, we're concerned about how do we estimate parameters, and what are those X's. And to what extent can we use data richness to inform how we should be aggregating the the X's in order to account for what's going on in the in environmental flows, as well as thinking about natural capital stocks and how we might measure measure those those feedbacks in in human behavior, and the impact on on natural capital in in our in our national and global accounting. With that, I will. I will stop. Thank you. Thank you very much. So those were wonderful presentations. So let me first ask the panelists to to respond to to to what has been said by the the other panelists and then after that we will have some time for questions. So maybe again in the order of the presentation. Jim, did you want to respond to other the other panelists. I'm not going to say very much. I think it is by love solves listing his impact his this is a wonderful literature review so I just want to see the whole thing. I think, I think, I think there's a challenge, a challenge in thinking about how to deal with, I'm going to follow up on just one comment that Lars was focusing on which is this this deep uncertainty. I think there's a real challenge about thinking about how to incorporate or how to address some of the transition risk uncertainties, these are ones that are just really difficult for us to quantify and they're going to be in our face in the short run they're not, they're not longer on issues. So, exactly how to approach that I think is a is an open question beyond just scenario analysis like is being used so the Fed is using scenario analysis to tackle these problems in that in their pilot study and that's, that's great that's getting started but it's a bit it's not ready to end. Thank you. Lars can ask if you want to respond to their panelists. So I read recently a paper in nature called the missing risk of climate change which is a very interesting paper. And it puts on the table these uncertainty notions which I'm very sympathetic with although it doesn't really address. It doesn't really connect to a bunch of recent decision theory, while even decision theoretic contributions will last few decades but it's, but it's certainly the case that if that that this is tendency that if we don't quantify the risk, you know with great accuracy we we either drop them but but the idea that we have to assign probabilities to everything and set a probability bounds I think is can be very problematic. This gets me back to Saul's table. I'm very glad to be included in his table but I'm sure, but I would have, but I could easily structure it differently and this is the motivation behind our research, which papers address uncertainty and which ones don't the North House work basically didn't touch uncertainty. Whiteman's papers all about risk and and which and and what our research has been about is showing these other sources of uncertainty could potentially have big impacts on policy. We've done this in highly stylized environments that I that I think are too stylized to be ready for for for final policy assessments, but I do think thinking about uncertainty in broader terms can have important impacts in, in terms of policy making. And so I hope that when you think about different categorizations of models we also think about the ways in which they treat uncertainty and very fundamentally different ways. Thank you. Thank you, Saul, would you like to respond to the panel. Yeah, I think, largely, I totally agree I think there's just lots of people thinking about uncertainty in different ways and different elements of uncertainty. I hope everyone took away from this is that there's lots of different models being used for different purposes. And in the policy making domain I think one thing that's really important to consider here especially as policies are actually being deployed and different models are being used. There is a emerging body of research showing that like when different risk assessments from different climate models are being used in different policy instruments but within the same economy. It creates arbitrage opportunities for like people bearing risk or underwriting mortgages or so we just need to be aware of that so in some cases we want to have different models for different purposes but if they're not harmonized properly you can sort of create new problems or just kind of unwind undo the impact that you're trying to impose on the marketplace. And so there just needs to be some awareness. Eric, would you like to respond. Sure, thank you I'd like to comment on Satya's excellent point about the issue of cascading impacts across scales. And just point to a few more I mean one was raised in the first panel that once you have an impact in one sector, or through in the financial markets it can it can cascade. Another is the technological regime literature talks about cascading scales from local innovation into larger transformations of the economy. And, and that also has ripple effects through through finance as well so I just wanted to, I guess, take the opportunity to make things even more complicated thank you. And finally Satya, would you like to respond. Thank you. I really really appreciate Sol's table like it put things in perspective in a way that that is very insightful. And, and I deliberately did not did not include uncertainty in in in what I was mentioning because I knew it would have come up earlier but I am. I do think one of the one really important point that has come up is how we think about uncertainty aversion or ambiguity aversion in the way people make decisions and is if there is a way to to consider that as an additional complexity in decision that would probably present an even more realistic way of modeling what what's going on especially when it comes to claim claim to risk. Thank you. Can I, we're now going to move on to some questions I'm going to prioritize questions from the panelists and then we'll move on to more more questions from the group. So, let me start with Heather. Okay. Thank you. What a terrific terrific panel. Thank you all. So I have a simple question, which goes as opposed to what Eric did adding complexity. I'd be very curious to hear from the panel what you would prioritize. And some of this was, you know, Jim's point about the real challenge about the, the near term transition risks, but very curious in terms of where we need to go next what you think the the most in him, the most important pieces to get done first star. So, can I just jump in. Yes, please do. I think I would point to two things one we saw it was really interesting in the in this morning's panel we saw two different numbers one from OMB where there's a scenario where there'd be a hit to something like if I looked at the paper right 7 to 14% on GDP by 2014 and from CBO where we had 2050 hit from climate being 1% those are pretty big differences we probably would like to get the baseline somewhere, you know, get some senses to what this baseline is and what's some senses to the distribution on this baseline. So I think a good place to start is where there's a ton of literature out there already, which is focusing on some of these, but what we think might be plausible hits to GDP so we can modify the baseline. Going out to let's say the 25 year 25 year horizon. That's not the most interesting economic research like for, you know, PhD students but I think it actually is a very useful practical step. Yeah, just a two finger on that sorry I wanted, I think the conversation from this morning to getting some of this literature into the baseline matters because in all subsequent decision making. If you don't have the baseline account for climate change, then all you're going to do is see the losses from costs but you won't see any benefits. And so it's decision makers will always be led to just not do anything so I think that that seems first order. Okay. So, I'm going to ramble a little bit so Jim had a present had a slide on the different models I think physical, then economic and then, Safia had a really good slide of the integration across models and I saw a disconnect there because Safia had like the engineered based model so they and Jim didn't right so it isn't just physical impacts is then the not the impacts on the natural system and then the impacts of that on engineered systems right. And so, I was wondering in all of these micro and micro models what is the role for incorporating these engineered based models to account for because the impacts of climate change on economic the causal, the causal effect is through these engineered systems very often, and we're dealing with impacts that we haven't seen before so all of the empirical based models may not be well, well calibrated to that so what's the role of those engineered based models in these analyses, and how they have been incorporated. And then the other thing I was wondering was, even in, like, so there's the impacts of climate change and then there's the impacts of climate action right and in, we're creating through, like, if you're doing mitigation you're creating new economic sectors with very different technologies that have different supply chains. So the additional models that are based on input past input output tables or they don't exist right so we've been doing some work for example with battery materials and the potential for supply chain disruptions and how that affects but we don't have any. We have good models of the flow of not even economic like material flows for those things and so how do you account for those technologies and when you're looking at, like the macro or micro impacts of mitigation. So, I can just start here and not not address all of it, but to say for the engineering based models. I mean there are engineering based models linked to macro models times macro as an example, they tend to be in the energy sector. I work on ones in energy and water, much less on the water side, but I would just say. Engineering models and this is a way to deal with uncertainty is to design against a return period and that that is you design against an exceedance probability of some particular level of damage. So you choose that that's a policy choice and then you design it and that influences your costs. I think that the way that engineers deal with uncertainty, it can be really helpful. It then becomes extremely complicated under non stationary conditions where what the return period is is changing. So there the question is, how are those systems taking the information about the changing climate into account. And during the first panel, I was wondering, you know, the distributions that you are finding can you use those to inform actual planning. And if you can that would then impact the costs. But it would be more from an engineering perspective and basically an engineering economics approach where you're looking at those costs. But that's that's on the mitigation side right then there's the physical impacts on the infrastructure. No, it's, it's actually on both the model I have in mind is my own, in which it was it was your, your, it took both the impact the cost of the impacts and the anticipated impacts and therefore the adaptation into account so that you could look at the combined output. So I think it's a way to do it to to and to improve, but we can discuss later. Lars, I heard that you had a response to the earlier questions to, did you want to respond now or should I go on. So I mean some very interesting points were raised. First, just a follow up to this engineering. There's, there's a real sense of which engineers have been better at thinking about some of these various different robustness issues than economists and and and indeed some of the decision theory stuff has been been building on constructs they are modifying and adapting them to ways that a little bit better economic models but I put there's been some important insights there and indeed my own work has drawn on those as well. And I think that we do have to remember that it's kind of the, we want to be able to take these feedback effects into play and so, and so we get loading in these engineering systems is, you know, it's certainly a very important part of things and I hope we can do better jobs that that going forward, but there are these kind of interaction effects or feedback effects like that and so I think these are particularly challenging for us to deal with and but but at the end of the day very important ones. So, I, I, I agree with the point that it's going to be very hard to just treat this as empirical challenges that a lot of this stuff going forward, it's going to require us making hunches about what are good models as we change the underlying as we push the economy into places which we haven't visited. And I think we're going to have to rely even more heavily on kind of conceptual models than on just straight empirical evidence going forward and in order to address these questions just because I mean, yes, historical evidence can be used but it's limited that there's big limitations in terms of what it can tell us as we push economies into new places. Thank you, Bob. Did you have a question. Yeah, thanks everybody. Great panel. So, I'm going to ask a question that hoping to get people's initial thoughts I think the question might be one that requires our discussion over multiple workshops, but just look thinking about the panel this morning and thinking about this talk which began with this great slide showing time horizon versus a sort of task access, looking at the different tasks, macroeconomists do it federal agencies. And right now, one of those tasks really says social got to carbon and, you know, clear how climate integrates into that but then also guiding monetary policy ensuring financial system stability effects of counter cyclical and the decadal scale economic consequences and multi decadal fiscal policy. To what extent in your opinion are, you know, we've had the term baseline come up a lot so so one one clear thing that I think it seems like we agree that there's a need to improve the baselines and these models in a way that incorporates climate change. Are there other ways in which the sort of climate models we've talked about in the second talk can plug into the tools that are being used now without major sort of epistemological revisions to have these agencies are thinking about things at risk, or is a lot of the emperor going to be at the level of well actually like the tool is not really fit for the purpose once you start dealing with climate change and we need to have sort of broader level conceptual provisions. Can I jump in on that. So, so I think I have a moderate familiarity both with the model that the Troika uses, which is the OMB CBA Treasury model and then also the CBO model. So those can take inputs in in terms of those can take inputs into convenient fashions and one of them is if you want to change things like productivity growth rate assumptions modify that for climate reasons that will give a hit to the capital stock or something on those lines you can just put that right in and that's actually the sort of mechanism that CBO used to get its 1% hit and that's the sort of mechanism that OMB used to get it to get it's much larger hit and it's more extreme scenario. So, one could even because those, because the feedbacks are so small, you could actually iterate on that and essentially solve the system in a couple of steps, and that would be that would be perfectly adequate. That's not in any sense a deep solution to the sorts of questions that Mars is raising or others have raised, but as a mechanical step going forward there are definite points of entry in these models and that's what that's what's being exploited. And that's one in terms of figuring out how best to agree on what those points of entry should be and how you do it, but there's definite points of entry. Anyone else like to respond to that question. We've been given kind of extreme outcomes in the sense of this is like a minor tweak on our existing modeling versus we have to throw everything away and start over again well. There's probably neither of those. I do think there's going to be some big changes relative to a lot of other macroeconomic problems here but but you know, our existing toolkit continues to have some value here and we just have to figure out how to build on it and and and and modify it in ways that are really well suited for climate change. Sonya, you have a question. I do thank you actually have to thank you for an excellent panel and as well as the one before. My first question is about human behavior and you know, Eric you did a great job of saying if you know one model, you know one model but when it comes to human behavior I think you know them all. I'm pretty terrible at predicting human behavior and that the cycles, the implications the indigeneity of human behavior so my first question is how well do our current macro models include the individual the consumer how can they be improved, and is there opportunity here actually extending Paulina's question of thinking about other modeling techniques such as agent based modeling that can better incorporate how humans respond to macroeconomic shifts. And the second one is actually to set a baseline for us and our understanding just to make sure that we all know the current state of affairs with the models that are out there. Solomon this question is particularly for you but others that have built these models that also incorporate climate. Can you give us a sense of how you've done it so far, and your own critiques of these modeling platforms how if if we were to pick up what you've done, how can we extend it going forward. Thank you. Can I respond. In terms of the first one. I do want to say, I mean it's, it's very easy to get that impression that there is only one behavioral assumption in the economic models. There are in fact multiple traditions. There's a dominant style right now that is based on the welfare theorems. That is not the kind that I work with there are alternative sources of behavioral assumptions. For example in postkinsian models there's a lot in the literature on evolutionary economics that I think could be relevant here. And, and this, this approach of, of starting with behaviors and then asking what the implications are in terms of exchange physical activity and so on. There, we're not starting from scratch there. I'd like to talk a little bit about Once we're in the situations where the uncertainty gets more complex exactly what exactly is rational is even a bit of an open question here so it's not as if there's one's one size fits all there anymore. On the other hand, I find agent based models as taking a shortcut but some some but but ones that rather mechanical it can get us maybe interesting places. But at the end of the day if we want to do policy analysis there has to be some kind of for looking aspect of things because if we change the environment. The adaptation that the backwards looking adaptation approach may eventually get there but it seems to me like there's going to be some type of looking forward which we have to take into account. Now there's other macroeconomic literatures all the way that that's four different aspects of bounded rationality all the way from like limited information to limited to limited attention and other stuff like that that can also provide insights. I don't think we're at this point point where one size fits all here I think this is as we want to entertain these complex uncertainty environments. There's a non trivial issue about how best to model the people both insider models and and you know what, as well as external but what's the best way for external accommodations to proceed. Thank you, Adele Morris, you had a question. Yeah hi thank you so much for this very interesting panel got a lot out of that. So following up on Jim's idea that a priority should be incorporating climate factors into the baseline and thinking through maybe a horizon of around 25 years just to play it out for a second. Okay, let's say you did that. And then you're contemplating a major emissions mitigation policy. So you've got a new, a new with policy trajectory that one challenge. And I love the panelists thoughts on this is that you know greenhouse gas is being a stock pollutant. What you're doing is you're putting a decrement down in the emissions that are added to the stock. So, in principle it might take quite a while before you see a sufficient reduction in the stock, relative to the counterfactual to see, you know, significant reductions in the damages as reflected in us GDP, right, over a 25 year time frame. And just at an empirical level, you know, how, what do we what do we know about that. And then I guess the alternative framework is your standard circular a for, you know, we're going to do costs and benefits and then you've got the social cost of capital. The social cost of carbon, excuse me, which has a global measure of damages in a present value tense. And it also has potentially added into it the ancillary benefits from reductions in criteria air pollutants and other environmental damages associated with fossil fuel production and consumption. I'm just wondering from an empirical standpoint, like, kind of what you what what you might guess is is is going to show up if we did this exercise. Can I just just clarify one, one thing about my suggestion of the baseline would say 25 years I'm taking the, you know, the task at hand is being thinking about CBO budget exercises and building the economics around that or thinking about your stress test and stuff like that. I did not mean to say that if we were doing, suppose you want to do an evaluation of the EV tax credits in the IRA, that would be a perfectly legit policy analysis. I'm not actually thinking about that. Clearly, those benefits if you're going to do the CBA on that those benefits go way beyond any 25 year window because of the exact reasons that you said, it's also the case that those effects would be sufficiently second order that you could iterate really quickly and get, you know, there's going to be very little feedback effect, but I do. I just want to, I want to make it clear that my suggestions of trunking were based on what I viewed as the red here. If I could just jump in and say, if you're looking at regional impacts, then climate uncertainty aside from anything we do with emissions is going to trump anything else that would be my, my guess. It depends on the scale at which you're looking. I think, I think it's a, it's a really good point that some of those benefits will emerge, like pretty slowly warming does occur very quickly in the atmosphere but like the US is not the dominant emitter is not the only emitter not the majority of emissions. What I do is second order on a small fraction of global emissions. And so I think you are correct that it will show up slowly I was, your question was really interesting I was starting to think about it. I know many people here are involved in like the interagency working group on social cost to greenhouse gases. And I guess, thinking about the metric of GDP is interesting. GDP is just like prices times quantities of everything we produce. So if we were to. The big problem is that we're producing a lot of carbon at a price of zero, and so it doesn't enter GP, but if you treated carbon is just like a good. It would just show up in GDP instantly. So I guess if people are looking for a measure that's an adjustment that would cause these measures to be sensitive to what we're doing to the economy. That might be one thing to think about. I mean in the short run within the 24, 25 year horizon, just to pick up on the last question on how different models fit. One thing that's worth lagging here is that like all the models can be fit to data, and then they look really good, because that's just a model fitting exercise. But a challenge is like what's left out of the model in the fitting exercise so if you don't have climate in the model, it's fit very well to the world, assuming a stationary climate. And then when you plug in the climate you haven't fit that component of it properly so having the model be complete and accurate in the fitting step seems to always be important. And then once you fit it like all of these models can reproduce certain things it's just what, and I think the out of sample forecasting is not like what you do in machine learning is not super widespread and economics. And so we don't have really great cross validation metrics for out of sample fit. And so that's something I think our community has to start doing better. There was a question on how to pick up and move forward with some of the work that's been done I think there is a lot of literature that's out there it's not just our group. I think it would be really helpful to basically I heard folks in the first panel talk about the models that they're using, and the inputs and the components I think it would be really helpful to just like compile a list of all those inputs. Jim seems like he's familiar with some of them, and then compile a list of all the studies or we understand something about the climate and try to figure out like where there's overlap and where there's important things missing. I feel like that's just sort of a reflexive exercise that would be like just helpful for bringing the models up to date. Peter Wilcoxson, you have a question. Thanks very much. Thanks for the terrific talks to everyone on the panel. Coincidentally, Saul just touched on the issue that I want to raise the two words that that I'd like to hear the panelists react to our validation and transparency. What we're talking about here is making the standard practice of the government agencies, maybe an order of magnitude more complex in an environment that's frankly adversarial. And so we'll not only have to have models but we'll have to have defendable models. And that's going to be a challenge if we're going into new territory where we've decided that statistics are not available. So I'm not sure that this is a quick question, but any thoughts on that be very welcome. Thanks. Quick thought. And I don't know if it's actually possible in a US policy context, especially in the policy focused scenario literature there's this approach where you look at robustness. The way that you defend your policy is that it should be robust against different possible future outcomes. So you're not saying that there is going to be a specific future that you're planning against. And, and that could be one way of kind of stepping back from the unquantifiable or hard to quantify complexities is is to ask about robustness against different possibilities. So kind of robustness is what I was talking about in terms of this decision theory literature is all about the robustness question and how to deal with it very coherent ways. So, so I guess I would echo on that echo that I think validation is a serious problem. In this context, I don't have cheap solutions to it. Yes, we need to think harder about it. I think transparency is also an interesting question. Transparency is generally a problem. As we kind of build more complicated models become some somewhat less transparent if we build two simple models, we build simple models become transparent but then they look too simple and where the middle ground and all this is really is really challenging now. I personally also think that we need to do sensitivity analysis looking across different model specifications and and and trying to go through and sort out different outcomes as well as I say you know beware the person of one model. And so that that makes the transparency challenge somewhat more somewhat more severe but I think it's also important so I believe I believe going forward it's it's kind of handy to have two things one of has very simplified models that illustrate all the mechanisms and and and get it some of the core basic results, coupled with the models that have all the additional rich richness that to some will look like black boxes. Let me echo the comment about transparency when it comes to national statistics we've talked a lot about statistics like GDP and having seen the debates relating to GDP and inflation statistics in relation to other issues. And there continually are conspiracy theories relating to how inflation for example is constructed. And there are cases where we know we can make the statistic better in in a conceptual sense like for example hedonic modeling in some areas, but there's a there's a trade off because it would also make the construction more subjective and less transparent. So that's something that one struggles with even outside the question of climate change modeling with these national statistics which have taken so many decades for people to become confident in. So, sorry, we have time for one more question, fitting me from Wendy Adler. My question maybe comment is follows on very nice I think from from Saul and Pete's which is interesting as maybe it's kind of where we're ending up. More than where I started I'm realizing that empirical models aren't going to get us very far. I think that the relationship between the aggregate economy and climate in the future is that those aren't relationships that that history is going to give us a fantastic guide for. This is then generally a place where structural models and economics are pretty darn useful. But then I want to say two things as to what the structural models are useful for first we need to be really really careful about the evil of false precision. We're going to build these models. They're going to come up with numbers and, you know, we need to be very, very humble about about not trusting these estimates. But what structural models are really good for is making sure that we haven't missed channels and avenues that are really important that perhaps are even counterintuitive. So, I don't know, I'm just now and now I am thinking that maybe maybe what we need to do more going forward in terms of improving our macroeconomic modeling is come up with, you know, structural ways of putting climate into our models. Maybe it's, I don't know, I don't know what that means for the hard work that like CBO or OMB has to do where you like you need a number. But, but we do have to be awfully careful about false precision. Oh, sorry, go ahead. I was going to say since we are basically out of time, perhaps the the other organizers in this room can comment on this but my thought was perhaps we can take that as a broad challenge to the group. And, and if we want to continue on schedule then we would need to make one point because I just want to make sure that when it comes away. I think I disagree very strongly with both what you suggested and I think some things Laura said and that's okay we're allowed to disagree. I absolutely think we should be using a lot of empirics throughout all of our structural modeling and we should be the long history of writing structural models that were totally wrong and had nothing to do with the past and the other risky run is that when it becomes too subjective, the next elected official just has a different perspective on what the discount rate should be or something else. And with the flip. And so, like the real world data is the only thing that pins us down to reality, and make sure that we are doing science and that it is that it is not just like people's opinions. I'm going to probably side in with Saul on this one. If I could add one quick thing. Of course we do want to look at all that historical data available but it still may leave us with only bounds on things and not sharp answers. I'm the moderator for the next time I just make a comment on this. There was the discussion about validation that I think the round table needs to consider because there are four conditions for validatable models. And I'm just looking them up, being able to observe and measure the situation being model. The model must exhibit a consistency of structure in time. It must exhibit consistency across variations in conditions not specified in the model, and it must be possible to collect ample data, which with which to make predictive tests of the model. So I think in climate, like when we're looking at the impact of climate change and models that release would probably don't meet most of those conditions which means the models are probably available to the conventional thinking of what validation. And so what do we do, given that constraint. I think it's an important tool. This is something we deal with in energy system models all the time. And so the question is, and the answer we have in that context is, what's the use of the model. Are we really making predictions or is it more of an experimental process. Great. I think we can move on. I wasn't sure if you were taking responses. Yes, there's also a literature in the environmental sciences about right dealing with extreme uncertainty. Yeah. Okay. All right, thank you very much, any for managing the many, the many voices that was that was super. We are now going to hear of, I just lost it.