Overview
Q: John could you give us an overview of your project on presidential approval and economic performance?
A: The problem that has most concerned me over the years in my career been that the relationship between democracy and markets. The question whether markets and their operations undermine the workings of democracy and conversely whether the workings of democracy undermine markets. I have attacked this question from a number of different angles, in the project that we're talking about today I look at it much more behavioral, positivist perspective trying to integrate models or mathematical representations of social reality from both economics and political science. I believe also that the interdisciplinary pursuit of this question or pursuing this in a interdisciplinary way will give me insights that I couldn't obtain by studying them strictly from the point of view of political science or economics alone. This project takes this larger question and boils it down to a more specific one and tries to fuse the models of statistical methods of these two disciplines together in a way that gets us some purchase on this larger question on whether democracy and markets are compatible.
Research Questions
Q: Can you describe the state of the literature at the time you began this project?
A: I think the literature offers some four basic flaws. Again this vein of the literature that tries to model these things or represent them in mathematical and statistical terms. The first is there is an imbalance between the economic and political components of these approaches. One has very well developed political models without any kind of economy. Or one has a very underdeveloped polity or a very well developed market system without much of a polity. So we wanted to strike a better balance between economics and political science in terms of the model itself. Secondly, we'd say they're highly temporally aggregated. By that I mean the notions of equilibration and political-economic equilibration becomes the concept that really achieves a synthesis between economics and political science. It's highly temporally aggregated. By that I mean the notion of when markets and politics are in equilibrium is based on very long periods of time or rests on a very long term conception of balancing or equilibration, in some cases every election term. So the idea that politics and economic reach equilibrium every four years and this seems to be non-intuitive or simply inaccurate. Politics and economic must reach equilibrium at a more regular and temporally dis-aggregated way. Third, a lot of the work in economics and political science stress the stochastic elements polities in economics (???). Economies have stochastic elements in them. In economics these are things like technology shots. In politics it's something like approval shocks. And these notions of stochastic elements of markets and democracy haven't been integrated very well in the existing models that are out there. And last but not least, there's a very great disjuncture between what people do with models in analysis and statistics. One does very elaborate modeling in one form or another in politics or economics one comes to address the data and testing one's findings. One has a very weak sense on how these two things are connected. One simply throws statistical models against the data without asking 'are these statistical models well related?' to the mathematical models that I've worked with in my original analysis. These two things are very much unrelated and of course unpleasing and suggests inability to really get a handle at what's at work and the quality of the political-economic system.
Q: A central concept in your study is the notion of a political-economic equilibrium. Can you describe the importance of the concept of the equilibrium for us?
A: There is some idea. I think one of the things that is true of social science today is this idea of predictability. Or the ability to predict what's going to happen in terms of social behavior is best achieved by these notions of equilibrium. Some idea that there's some stasis or some recurring kind of behavior towards which social systems tend. And that one can discover this equilibrium and this state of behavior towards which people and agents tend or trend. That one can then predict behavior because the systems will continually return to this state. An example in economics is market clearing. The idea is that markets clear. There's supply, there's demand and in some sense they're in sync. That the markets clear. In political science it would be some idea of party positioning. Parties go to a particular policy position and once both parties are there they don't move again because that party position or policy position ensures them of a certain percentage of votes which in many cases there's enough to ensure at least a tie in the election. So if we know what that policy medium is and we know what the market clearing condition is, we can predict what the politicians will and what the market agents will do. In my paper the question is 'is there a link between these two things?' Does the fact that politicians go to a policy position or the governments try to achieve a particular level of presidential approval. Does that mean something about the way in which the market cleared or conversely if the market is clearing does that mean something about the government's ability to achieve its preferred level of presidential approval.
Q: What are the central features of the political equilibrium model that you specify? How does it differ from the models that have been used before?
A: The economic model comes to represent economics and the theory about optimal fiscal policies. So we graft that, actually we borrow it from economics. I should preface this by saying we don't so much create anything new is what we do. We take existing types of equilibrium conceptions and we put them together or fuse them together, we patch them together. And then in the end we have something that is novel way of fusion of these different concepts. We take the theory of optimal of fiscal policy and our conception in which households operate in the way which firms behave in a perfectly competitive market and we take the sense or the understanding of the way in which citizens dispense presidential approval and government seek presidential approval. And we put all these things together, bringing along with it these ideas of stochastic elements in economies and polities and bringing along with it the idea of technology shocks so the production of the economy is partly derived from stochastic production function that includes technology shocks. Presidential approval includes approval shocks, just spurts of presidential approval that comes from such things as international events or the government sending troops to Somalia or something. That creates these bursts of short term, seemly random bursts of approval and we pull all this together. And for the first time, I think it's never been done, have a household optimizing subject to the knowledge of the government is optimizing approval and the government maximizing approval subject to the condition of the households have to optimize. And that these other market conditions have to apply.
Q: Your general theory about the relationship between economics and politics is translated into a mathematical model. Can you describe, at least briefly, the nature of that mathematical model?
A: I think one things that its true in Social sciences is a belief, call it rational choice models, that have become many different groupings. That economic agents and political agents in some sense maximize their utility or their felicity it's sometimes called. The models basically the idea that households have an objective, that are a utility function in terms of consumption and leisure and they optimize their accrual of utility subject to certain conditions that exist with regards to the market. Things like the way which markets clear, the way which wages are set, the way which production occurs. And simultaneously the governments have a particular set of goals of what they want to achieve, a particular level of approval. But the households dispense approval on the basis of how well they do in the economy. So citizens are dispensing approval on the bases on how much they are consuming and how much leisure they can have. And so the model is basically a strained optimization. It's something that you learn in mathematics. When I was in mathematics in my second year of my math major I think now probably the first year which is strained optimization. Which is you optimize something subject to constraints. What makes the model very unique you have two agents optimizing subject to two sets of constraints and this poses a very difficult problem mathematically in brings us to this area of computation. It is not possible to derive a closed form solution. Most of us take mathematics classes and were given a problem and we derive the solution and there's an answer. And the answer is correct because a particular function solves a differential equation or a particular scale solves an algebraic equation. In this case it's actually is impossible to derive an analytic solution to this joint optimization problem. This puts us in the realm of computational equilibrium. Where we literally have to let the computer look at a variety of different decision rules that will optimize both the agents' economic welfare and the government's welfare. And we have to find that through a iterative process of searching through a set of possible decision rules and we find it through computational methods.
Design Sampling
Q: Critical to running your simulations is establishing some perimeters for key features of the model. How do you establish those perimeters?
A: Well there is two basic ways. One is to borrow from existing work. This is something that economists are prone to do. I think to be completely candid about it there is a tension here. Ideally what these models run into a problem of scale. When you get a model that's too complicated you literally lose your ability to estimate all of the perimeters. So the way this is handled is on the one hand and this is because of the problem of over determination or identification and the inability to conduct certain kinds of experiments. So one way perimeters are achieved by simply looking at existing econometric studies where the perimeters exist and other models and lifting the perimeters from one model done in a different study into our model. They are the same equations but they are done independently. Another approach is to simulate. And this is something that makes our study somewhat unique. The approach is not estimation, it's calibration. This borrows something, an approach that is very popular at the University of Minnesota. What we do basically is we build this model and derive the optimal decision rules in generic form and then we basically simulate an American economy. Once we draw the shocks and we simulate it. And we see how that economy behaves in terms of its very simple moments. The standard deviation of the growth time path de-treaded in a certain way, which we won't go in to. Under a certain assumption about perimeters. Where do those assumptions come from? They are just intuitive judgements about the value of the perimeters. How well does the American economy under these borrow (?) the perimeters and just guessing to its value. How well does it simulate. Then we simulate it again and again and again. We simulate it a thousand times. And then we say on average, does this simulated political economic system look like an American political-economic system. If it doesn't we change the perimeters and we simulated it again a thousand times. We keep iterating, keep repeating it until we get hypothetical model or a model of perimeters that mimics American economy well. That is admittedly an extremely subjective judgement as to what is good mimicking. And this is an art form. This is one of the things that has been immensely valuable to me in the interdisciplinary work is to go to the economics department. Where one would think that they have all these problems solved in a very definitive way. Learn the art of calibration, learn how to look at seven or eight results or statistics and say this is a model that mimics well, this is a model that doesn't mimic well. Because you can imagine invariably some of the variables will mimic very well and some don't mimic quite as well and that's a real art form. That's where working with Professor Howser was immensely valuable. I was saying this doesn't mimic the quality very well and does. The choice of the perimeters is partly then, borrow from existing studies and partly the result of calibration.
Q: Is the model in any sense tested?
A: I would argue that it is tested, I just wouldn't use the word tested. I would argue it is evaluated. I think personally it's evaluated in a more meaningful way then it is when you take a very artificial slice of data and fits a line to a set of points. I think this in some sense is more honest about the problems we face in matching, identifying the structure what is in fact a very complex simultaneous kind of social system. Much of what passes as 'good tests' social science is a reductionist enterprise. It's taking a single piece of complex reality and assuming you can isolate that one piece, that one equation. And you run that and fit that and you get very good statistics and you sat I've got a good fit. But we all know that you are engaged in a reductionist enterprise. You have lifted one equation about what probably is a 100 equations or 50 equations. You have taken one relationship out of what is always complex social relationship. In that sense I would argue most of what passes as good fit is an illusion. And so, this would be my refuge. If we are really honest about what we are dealing in terms of complex social system this is a better evaluation than what passes as good fit. But I also admit it is a highly subjective and artistic kind of enterprise. When you get this model fit then, calibrated I mean. When you have this model calibrated, you don't really report that as your result. You don't say 'I've got my result' from a calibrated model. Anybody can play with a model long enough to make a god calibrated model. The value of this whole enterprise is what you can do with it counterfactually. It allows you to do social experiments that you couldn't do otherwise. That is the real payoff.
Q: What is the importance of accurately specifying the time horizons of the agents of the model?
A: This is actually a very big area of economics, econometric statistics. The question of temporal aggregation. When one reads that literature and I actually have published an article on this. One finds reference to something called the natural time unit. There is an idea that theories ought to specify the correct time unit in which agents operate or which social reality is structured. Example is writing about fiscal policy that argues that government officials and economists think in yearly units of analysis. When one talks about budget making. If you have a theory of the budget that people think in yearly chunks of time. But when one talks about financial markets there's writing that suggests that somehow social reality is structured so that things incorporate on hours, seconds and minutes. This becomes a dicey or interesting conceptual issue. What is the natural time unit? What unit of social reality structured? One is going to argue that concepts of equilibrium are the key of making predictions of social behavior. One has to address the thing. Has to address this question of equilibrium and what timing. Because if it the time unit is inaccurate then you're conceptual equilibrium may be inaccurate or ill conceived and consequently unpredictably. In statistics, it turns out if the natural time unit is very temporally disaggregated one, so the relationship between inflation and unemployment is based on weekly units of analysis. You aggregate those up to months to years you can destroy the causal relationship that exist in the original state. This temporal diaggregation can just destroy causal relationships. So the act of this measuring the temporal unit could prevent you from finding causal relationships. The problem is we don't collect on a regular basis so our very data collection enterprises may prevent us in this sense from finding causal relationships. Not because we have bad theory, but we literally we collect the data on an inappropriate level of aggregation.
Q: Isn't possible that the agents in your model live in a variety of temporal worlds with different time horizons? And if so, what theoretical and methodological problems does that create for you?
A: Absolutely. Indeed some of the disequilibrium may derive from exactly the point you're making. From the fact the actual markets equilibrating at a much faster pace than the political reality. In our model to be honest we force or shoved both economic-polity into the same temporal unit. Our idea of an equilibrium is based on the idea of one political economic equilibrium as the shocks occur. There is a shock in technology, a shock in approval and shock of government spending which we didn't talk about. The government chooses a policy tax event, interest rate policy. Households optimize it, government optimize it. All of that activity is one time unit. We assume that. That's how we simulate. We draw the shocks, we do the optimizing (???), do the government policies, all the variables each their values and that's one time unit. Then we draw another set of shocks same thing is repeated. And we do that a thousand times, keeping track what we've done in the previous draw. We are forcing all these events into one time unit, when in reality there's probably all these things are out of sync in that sense and maybe why we can ????? in that case. That is not in our model and is an important challenge for future work.
Q: What period of time is covered by your empirical analysis?
A: The period of time is 1980-1990 and that is period of continuos Republican incumbency. And that is consistent in the way the model was conceived that is as a government of one partisan identity.
Measurement & Data Collection
Q: Can you describe the nature of the data you used to examine the performance of the model. What are some of the problems in using data of those kind?
A: There are various data sets from macro economic time series. It's not particularly difficult to get those. Citibank publishes, Federal Reserve bank has them, it isn't particularly difficult to get those. There are college department that publishes data on I believe on number of hours worked. I believe I got data from the bureau of labor statistics. What can actually calculate for example, the amount leisure time for people, it is a percentage they can spend in number of hours. This perimeter of percentage of time spent on leisure is something that economists believed that we have 139 hours of waking time. That actually have numbers of this kind. You can calculate the number of hours we work out of that. That would give you the percentage of time at leisure instead of work. There is a wealth of data, macroeconomic data. Political data is a little bit more difficult to secure. Especially if one is looking for fine-grain time series. Remember my point earlier on about temporally disaggregating notions of political-economic equilibrium. If we believe that markets and politics are in equilibrium on a regular basis, weekly or monthly. We can get some economic data although not all economic data things like output, gross national product for example are not available on a weekly basis. You can get stock data down to the minute or second. We can't get output like GNP on a hourly or weekly basis. Public opinion data is available on a monthly level, but you can't go much lower than that. In terms of temporal disaggregation. So we took macro economic data from a number of sources. Our political data came from public opinion polls on approval management and the level of approval of the president. But we couldn't go much lower than a monthly level of temporal aggregation. And I think this raises a really interesting question about this point of political-economic equilibration. At one level are politics and economics in equilibrium. Are they equilibrium or don't think they're in equilibrium only every 4 years. But do we believe they are in equilibrium every day, or every second, or every hour. There is a lot of writing financial economics I believe that says financial markets are almost in continuous equilibrium. Perhaps there are spurts of volatility. This is a big debate in financial literature. But is there anything remotely related to that in the realm of politics? That would suggest that politics even on a weekly basis is an equilibrium even in itself in isolation. Politics indeed just may be just the opposite, constant state of disequilibrium.
Data Analysis
Q: Once you are confident that your model captures important features of economic and political processes fairly well, how does the model get used to extend our knowledge of politics and economics?
A: Again I apologize I have myself, once the model mimics well. It mimics the American model, then you can ask question counter factually that you couldn't have in other models. Because you can then go in to this very complex system and you can change in a counter factual way key perimeters to see what effect that has on operation of the American economic-political system. There are two examples in our paper of approval volatility and the pursuit of high levels of approval. Because we wrote this for a political science journal we stressed the political counter factually. The former is the idea is that when we create the model we have to assume a certain variance in approval shocks. The president is above my approval shocks like Monica Lewinsky, like invasions overseas in armed intervention. This causes shocks of approval we to positive variance for that shots to get the model to mimic well. So we positive variance so we get good mimicking. Now the question is what if the variance of shocks becomes much greater? Example, there is much writing in international relations that suggests that as the cold war winds down as the loose bi-polarity that define the cold war disappears. We will enter a world of increased chaos and increased disorder and that American intervention around the world such as Somalia or other places. Or perhaps the United States in being dragged into Bosnia and other places around the world. This will become much more prevalent. This will cause enormous increases in the variance of approval shocks that the president suffers. As the president is buffeted about by increasing shocks by deriving from his interventions in international affairs. This will have deleterious consequences on the American economy. We can evaluate that because we now have a model in which these stochastic shocks in approval is one of our perimeters. We can go into that model and increase the variance by 300 percent. We can now say what if we have a president managing approval and the economy in which the households are optimizing their consumption but approval variance three times what it is in reality. What happens to the American economy in that situation? And we then simulate the model a thousand times thanks to computing power. Average out the results and say this impact that this is going to have on the American economy. In the paper we trace out the implications on average that this will have on the growth for debt, for a number of macro economic variables. As a consequence of approval volatility. What we found is there is a slight decrease in social welfare that comes from increased approval volatility but not particularly great. So that is reassuring in some senses. The second example, which in some sense is perhaps more interesting. Is the question of 'loving dangerously'. The economists believe it or not three of four years ago published an article with Bill Clinton's picture on it entitled 'The art of living dangerously'. This was five full years before the impeachment hiatus. It had to do not with president Clinton's alleged or affair with certain individuals but rather with his very great concern with his approval ratings. Clinton was very much captivated by public opinion polls, by being loved by the American people. So we ask the question, what if the president is not seeking 55 percent approval? Which goes back to this question of winning a minimum number of votes or securing a minimum number of votes. What if Clinton is driving 70% percent approval and he manages a tax and interest rate policy to achieve a 70% approval instead of 55%. This is counter factual because we believe no president has actively done this. But a president might do this. And what we find is that this actually enhances social welfare. The level of consumption and leisure that households accrue actually increases to some degree, as I recall from the article, a minor degree. And at least again it doesn’t have the social consequences that economists and other lay people have thought it would. In terms of a president that engages in various kinds of economic policies which have all kind of harmful effects on the economy. We could not have done this experiment if we had a model that mimicked the US economy well. A complex model with a lot of variables could we evaluated this?
Interpretation & Dissemination
Q: What theoretical directions is the work taking now?
A: The thing we did next following from this paper, this is a paper which is now finished and is under review. Is try to understand how different kinds of political institutions create different political-economic equilibrium. Our model in this paper is based on the American political system. Which is a roughly pluralist majoritarian with a presidential system. We have moved to studying a Parliamentary system. One based roughly on British political economic system and also to some degree have thought and talked a great deal about Parliamentary systems and how those type of political systems would produce different types of political economic equilibrium then the ones that we charted in this paper.
Q: So the empirical work is now extending to other political and economic systems.
A: Yes. So now we end up building a political-economic system of Britain. Simulating the British economy and looking at hypothetical conditions in the British political-economic system in comparison to those in the United States. My next question is approval management is different by virtue of the fact that the Prime Minister of Britain occupies a different position in the political institutions then the President of the United States. Does that mean that the approval management of Britain has different political and economic and social consequences than political economic equilibration in the United States.
Q: What are some of the difficulties of studying another system, other than the US as you did in this first study?
A: One of the difficulties is the Parliamentary kind of set up. In the United States we can talk about the president and approval management as something that derives from the presidential administration control over certain policies. The parliamentary system is a much more fluid and complex coalitional kind of government that one is dealing with. One can certainly argue that there is coalitional politics in the United States but one deals with the Parliamentary systems one has a little bit more of a challenge because the Prime Minister is not necessarily has complete control over policy decision. One gets to countries like Germany where one has a possibility of votes of confidence and so on and it gets much more complex. And then with the possibility of political disequilibrium in a form of coalition following and a new coalition coming in which is not in our model at all. We assume the president is in office and we look at the president as the permanent incumbent. Whereas in these other systems have the possibility the Prime Minister being thrown out of office by virtue of their being a change in coalition. That's not in the model.
Q: When you are comparing the application of model to a variety of economic and political systems, how can you be certain which of the several possibilities are causing the differences? The model, calibration of the model, the quality of the data available for each system?
A: I think you used the correct word is confident, its always a matter of confidence. Always a subjective judgement. I think one of the problems we face is a lack of variety. As one moves to a higher level than everyone else, one starts talking about macro political economic systems. We simply don't have enough countries. That's a problem is that is my friends in across the river tell me. It is akin to Astronomy in some sense. We engage in some kind of social astronomy. Not enough big bangs, not enough big comets coming through the solar system on a regular basis. It's a very crude analogy, I haven't developed it fully yet. But you know we need more countries, we need more systems. We have one good one with a stylized model of Britain and we can simulate it but as you point out, we are not really sure if we got a bad model, we're not really sure if we have the model appropriately calibrated. If we had 30 countries that were identical to Britain and thirty countries that were identical to the United States then we could build 30 models of each and get 30 histories of both countries, we'd be on much stronger footing. That's one of the problems we face as one gets to higher levels of analysis in terms of doing political economic or any other kind of social science one runs out of cases very quickly. These questions of whether it’s the model or the perimeters or the analysis become almost impossible to answer. So one person has to fall back on these questions and arguments that are highly subjective in nature. One thing about the computable equilibrium is that this idea of being able to do synchronous simulations does allow you to check for robustness. You can go into a model and suppose we have this perimeter incorrectly and at an incorrect value. Let's let this perimeter change, let's give it 7 values and do a thousand of all seven values. We change the perimeter to 7 different values and run a thousand histories of Britain under each one. We look at the average behavior of approval in each of those experiments. If we find under all 7 experiments that is 7 different perimeters that the mean and standard deviation of growth is basically the same. That gives us a little more confidence that that perimeter is not and that our results are not sensitive to that perimeter. That our results are robust against a possible mistake in the calibration of perimeters. But we have to do next what if we got that perimeter incorrect and these other three incorrect. At some point this becomes very unwieldy. The computer does give some handle on this at least in sense of gauging our confidence.
Q: What reactions have you received from your colleagues in economics and political science?
A: I think my close friend at the University of Chicago that this interdisciplinary work is like standing in the middle of a highway, middle of the freeway. You can get hit by cars going both directions. The economists don't like it because they feel economy should be more complex, the political-scientists don't like because they don't feel the political system is not complex enough. So the first reaction is the model is too simple. That is in my opinion grossly unfair. This is an immensely difficult task. Joint optimization subject to stochastic constraints is extremely difficult problem. Economists have never done anything at this all. Very few have done anything remotely like this. People that do statistics in political science don't like it because it's not estimation it's calibration. In fact there are many statisticians in economics don't like this kind of work because it's not estimation it's calibration. So again, to some extent to do interdisciplinary work, work that's innovative in terms of the way the data is used to evaluate or build confidence towards one's models. Always raises the eye of other people. But on balance, the people who are really interested in the political economic analysis and building a bridge to economics have liked it a great deal and feel its genuine process and a real advance. So we are very happy that we have appeared in this journal because. In my opinion probably the toughest critics in political science are at the American Journal of Political Science. We are very happy about that. The other paper we will say what the reviews say. But one does interdisciplinary work, one has to be ready for some very severe criticism.