Russ Roberts

Rosenberg on the Nature of Economics

EconTalk Episode with Alex Rosenberg
Hosted by Russ Roberts
PRINT
Garett Jones on Stimulus... Bruce Meyer on the Middle Clas...

Alex Rosenberg of Duke University talks with EconTalk host Russ Roberts about the scientific nature of economics. Rosenberg, a philosopher of science talks about whether economics is a science. He surveys the changes in economics over the last 25 years--the rise of experimental economics and behavioral economics--and argues that economics has become more scientific and that economists have become more aware of flaws in economic theory. But he also argues that economics is unable to make precise predictions about the effects of various changes in policy and behavior. The conversation closes with a discussion of the role the philosophy of science can play in the evolution of economics.

Size: 26.3 MB
Right-click or Option-click, and select "Save Link/Target As MP3.

Readings and Links related to this podcast

Podcast Readings
HIDE READINGS
About this week's guest: About ideas and people mentioned in this podcast:

Highlights

Time
Podcast Highlights
HIDE HIGHLIGHTS
0:36Intro. [Recording date: September 19, 2011.] In 1994 in an essay titled "If Economics Isn't a Science: What Is It?" you wrote a very interesting essay. I want to quote a couple of quotes from it as a jumping off point. Here's the first one:
The ability to predict and control may be either necessary nor sufficient criteria for cognitively respectable scientific theories, but the fact is that microeconomic theory has made no advances in the management of economic processes since its current formalism was first elaborated in the 19th century.
You also wrote:
Euclidean geometry was once styled the science of space, but calling it a science did not make it one. We have come to view advances in the axiomatization and extension of geometry of events not in science but of mathematics. Economics is often defined as the science of the distribution of scarce resources. But calling it a science does not make it one. For much of their history, since 1800, advances in both these disciplines have consisted in improvements of deductive rigor, economy, and elegance of expression in better axiomatizations and in the proofs of more general results....
Very dramatic statements. I'm curious if you still feel that way. If you do, please elaborate on these ideas, and if you don't, why don't you? It's very interesting. I have to go back and say, 1st thing is that I said those things not in 1994 but in 1984. Thank you. By 1994 I may have already been prepared somewhat to qualify them. I think that they were very accurate as claims about economics in the 1970s and by the mid-1980s. I think, of course, there were exceptions and I think these exceptions have in the last 20 years drawn more and more attention and had a greater and greater impact on the nature of economics as a discipline. I think I'd have to say that on the one hand with qualifications I'm still sympathetic to the view about standards for science and scientific progress that stood at the background of those two points. And, on the other hand I have recognized changes in economic science since that time that make it at least somewhat less susceptible of that kind of analysis, not to say critique. I would be more specific and point to the increased thought about the nature of choice under uncertainty and the asymmetries of information characterized by the work, for example, of Akerlof, most famously. The interesting experimental economics that secured for Vernon Smith a Nobel Prize and most of all the impact of Kahneman and Tversky in cognitive social psychology on the way in which economists think about microeconomic processes, of which perhaps the most widely known example is the Vishny and Schleifer paper on the limits to arbitrage. All of these suggest that to varying degrees economics as a discipline has been increasingly sensitive to the kind of charges--not as I lay them because I think economists have generally been indifferent to what philosophers have said about the subject--but much more sensitive to these kinds of charges over the last 20 years and much more responsive to developments in other parts of social science that have made economics less guilty of the kind of hermetic insulation from the demands of application that both of those quotes that you started with articulate. That may be, but it is still true I think that formal economic theory is still subject to the critique that you made there. It's true that economics as "practiced by economists" has gotten a little less axiomatic. And I think that behavioral economics literature you mentioned is an example of that. But I guess my question would be: that behavioral literature, I don't see how it's made economics any more scientific in its ability to predict and measure. Or am I missing your point? It's clear that the discipline of economics is only now taking on and trying to figure out how to reshape its theory to take account of the factors that behavioral economics, evolutionary economics, and experimental economics introduced, and that at least in areas like financial economics, the tradition Walrasian, post-Walrasian, neoclassical paradigm continues to be the dominant force, if you look at the kind of intellectual cache that people like Eugene Fama have, or that other Chicago, fresh water school economists have in say, business schools or in, say, the theory of finance, financial economics. That, there, the paradigm still seems to be of the Walrasian sort that I was criticising in those quotes. The role of equilibrium thinking continues to be the same as it was back then, and although the word "Austrian" no longer has the kind of demonic possession--but I didn't mean possession--demonic halo that it used to have when economists would talk about people like von Mises and Schumpeter and even Hayek, still, the standard neo-classical approach does seem to me that it still holds the field; but it's at any rate much more defensive than it was when I wrote those things about it in the 1980s. And of course I did hold and develop much the same view in a book I published in the early 1990s called Economics, Mathematical Politics, or Science of Diminishing Returns, which in its title of course reflects the very theses that you quoted from the 1984 paper.
8:30Explain for the nonspecialist what you mean by Walrasian, pre-Walrasian, post-Walrasian equilibrium approach. What is the essence of the economist's view there? If we go back to Adam Smith, in the economics which owes its existence and its intellectual pedigree to The Wealth of Nations, it starts with the hidden hand, or the invisible hand, according to which we do not owe our well-being to the benevolence of the butcher and the baker but to his self-interest. And the idea that each individual pursuing their own narrow self-interest produces an outcome that is of benefit to the entire society. The market-price system and its allocative efficiency is an idea which though Smith originated it, economists set about trying to prove thereafter, formally; and it was only Walras who thought he had finally proved it by arguing that in an economy in which supply and demand in each market could be equilibrated such that there were the same number of equations in equilibrium as there were numbers of goods to be supplied and demanded; that under such circumstances there would be a unique, market-clearing allocatively efficient price. And the proof that Walras thought he had produced was not entirely adequate and it took another 70 years or so until Abraham Wald in the late 1930s produced another version of his proof. And still that was not entirely accepted in the profession; and nothing until the late 1950s, until Debreu and Arrow produced a formal proof of the Pareto-optimality of the unique and stable equilibrium in all prices that characterized a perfectly free market economy. And that unique stable equilibrium continues to be a kind of combination litmus test and touchstone of economic theorizing. Mathematical economics continues to be pursued on the understanding that the objective is to establish always the partial or the general equilibrium outcome of an economic process so that the way in which we understand the dynamics continually changing the economy is as a series of price and quantity vectors that, if only exogenous forces could be held at bay, would come to an equilibrium and are constantly adjusting themselves back in the direction of an equilibrium from the interference or the existence of these exogenous forces. So that economics--mainstream, neoclassical economics--continues to be invested in the intellectual importance, the explanatory role of general equilibrium even though it recognizes that that equilibrium never actually obtains. Kind of awkward, isn't it? It requires a certain intellectual schizophrenia. They would respond, and here I speak in defense of mainstream theorists, that if you look at markets the one thing you notice is that almost never will a change in one price result in a catastrophic explosion of all prices so as to produce chaos. The only time that we know of that this happens is in the hyperinflations of Weimar Germany or maybe Zimbabwe under Mugabe. And they will argue, therefore that the assumption of stability is one that's so well-established that no theory can be taken seriously unless it has stability as one of its implications; and that at least the market-clearing uniquely stable equilibrium has therefore a great deal to recommend it. There's a lot of stability in the real world; there's no doubt about that. I guess the question is how much does economics explain of that. I have a different perspective, I guess. I'm going to put you in the awkward position now of continuing to defend the neoclassical approach--ironic given that I'm a Chicago Ph.D. who was trained in that tradition. My challenge to that claim, the way you've phrased it, is the following: Adam Smith said that good things happen despite the fact that some people don't have the best of motives, or the purest of motives--they are self-interested. And I think that's a very deep insight. He wasn't the first person to have that insight. A version of it is in Mandeville. And others. But, he said it in a way that resonated and stuck; and he wrote well and understood it perhaps in ways that they did not. So, we have this insight that good things happen, sometimes despite the fact that people don't have the best motives. The fundamental question I have is: What have we gained through the sequence of Walras; I'd put Samuelson in there in 1948 with the Foundations of Economic Analysis, Arrow and Debreu, who took that to a much more theoretical level--I was going to say higher, not sure what the right adjective is, but to a different level, solved some of the problems, as you mentioned. We now have shown in some formal theoretical sense that yes, there is a set of prices such that under certain circumstances which never hold in the real world, letting people buy and sell freely ends up with a result that has some attractive characteristics. That's how I would describe what gets called Pareto-optimality or the various theorems of welfare economics.
15:38Now I think all that's fair. And you want me to defend that? Well, you could try. Or you could pile on. But my point is, and this is where I want to get into the philosophy of science: What have we gained--what are the dimensions of what we might have gained? In other words, I don't see anything that's more predictable about the world, confirmable, measurable, testable about the world from those theories relative to what we had before. Let me give you what I think of as the best defense I can. And I have to say frankly I've been searching for a defense for this character of the theory for a very long time, because I'd rather find a good explanation for the tremendous admiration that so many smart people have had for economics over the last 200 years, and particularly given its imperialism, it's methodological, substantive theoretical imperialism in the social sciences. I'm not impressed that there's a Nobel Prize for it; I know exactly why the Nobel Prize for it was created. It is a prize in Economic Sciences. Yes, exactly. There's a purely rhetorical role that it plays. But economics is a discipline in which a great deal of genius has been labored and the default position has to be that it's an important arena of intellectual achievement. So, let me see if I can at least give you part of an answer or as close as I can come to an answer that's satisfactory to the economist. First of all, if we think about mainstream economics, neoclassical economics, economics that comes from Smith through Walras, Wald, Samuelson, Debreu, Scarf, Hurwicz, Arrow, all of these great figures--if we think about it as institution design in the way that Jim Buchanan and Gordon Tulloch coined that term, then it has a really important role. Because the proof of the allocative efficiency of the perfectly competitive market which that equilibrium approach eventuates in is a blueprint for organizing complex human societies in such a way as to produce the largest quantity of what people really want. And we have of course in the history of the 20th century, the short 20th century, a great deal of very positive empirical evidence for the claim that no other alternative way of producing and distributing goods and services comes anywhere close to the free market in its allocative efficiency; and some of them are catastrophically bad at it. True. And to the extent that the proof of the existence, stability, and uniqueness of the general equilibrium is an important part of the argument for a certain set of institutions--economic institutions--then it is a powerful achievement. And when you add in that the proof and the assumptions of the proof--perfect divisibility, no asymmetries of information, avoidance of externalities, constant returns to scale--these assumptions in effect provide guidelines for government intervention in order to secure greater approximation to allocative efficiency, either by breaking up monopolies or prohibiting insider trading or forcing polluters to internalize their externalities; when you think about the way in which this proof provides guidelines to government intervention in order to secure allocative efficiency, you have still another powerful argument for the applicability of this intellectual achievement. But it's a very limited argument because, as I said, it's an argument that underwrites a certain set of institutional designs. And we want out of science something more than mere institutional design. We want actual prediction of events that we would like to control and to at least if not control, insulate or protect ourselves from. And there of course is where this version of economics has not done as well as it needs to do in order to attain the standards that the natural sciences and that biology have attained. Now, people are often likely to deny that a science like biology has great predictive power, and that therefore economics which certainly has less predictive power than biology nevertheless has a right to its place among the sciences. But I think in fact if you think about, spend as much time thinking about biology as I have, you'll see that there are a lot of reasons why it should be predictively considerably more powerful than economics; and you might also find a basis for concluding that economics can't, as a matter of fact, attain to a very high level of predictive accuracy; that there are limits to our very ability to frame a scientific theory of economics. I want to come back to biology in a minute, and I want to try to lead us there with a couple of questions about predictability. Before we do that, I want to register disagreement with your argument. I think it's an interesting argument; I think it's commonly held by most economists, so to be fair I don't know how much you accept it relative to how much you were standing in for the economists' viewpoint--which is that the framework for its assumptions of competitiveness, symmetric information, no information asymmetries, no externalities, etc., homogeneous goods, that this opens up the toolkit for government intervention in a way to lead us toward allocative efficiency. And given that you invoked Buchanan and Tullock in talking about institutional design--of course they were very important seminal figures in pointing out that what government does versus what economists think it should do are not the same thing. I think in many ways, economics, through its focus on efficiency and inefficiency--that is, through the focus on how well the price system works but it requires this assumption and if this assumption doesn't hold, therefore this intervention is justified--I think that's given government a frequently useful rationalization for intervention that doesn't move us toward efficiency but rather serves the politically powerful. What's called public choice. That's the great insight of the wildly misnamed "public choice" theory. The theory of public choice is the theory of how private choices deflect public choices or public decisions; and of course we understand all too well why governments fail too often to act in the interests of the governed, owing to considerations that Tullock and Buchanan, Geoff Brennan, and other important choice theorists have themselves identified.
24:19Let's move on to this question of predictability, which I think is a subtle and difficult question. I struggle with it myself. I'm going to start with micro. We'll move on to macro at the end. I actually think macro is easier to think about. But let's start with micro. One of the standard "predictions" of microeconomic theory is that a binding price ceiling produces a shortage. I taught at Stanford in the mid-1980s and I had a research assistant who I found was telling my students very different things in his office hours very different things than I was telling my students. Which was fascinating. I asked him why he would do that. I was teaching that when price was below the market-clearing price that that would result in a shortage, and that in turn would usually--depends on the institutional rules--result in lines, queues, waiting. And that time would allocate the good along with money, and the full price would actually be higher if you included the money price and the time price. One of the standard results of microeconomics. His reaction was he had been told differently in his micro class. He was being taught by a game theorist; he told me with great assurance that he had learned that anything could happen. And I thought--well, I suppose that's true; for many reasons of course anything could happen: intervening events, all kinds of things. But he was taught what you might call a non-equilibrium theory of economics. And I said: let's make a bet. Let's suppose the government were to put on--let's say at that time the price of bread was $1 a loaf. If the government announced that tomorrow bread would be 10 cents a loaf, that no one could charge more than 10 cents a loaf, or 50 cents a loaf, something well below the current price. Do you think a line would form, and how long would it take; do you think you could be sure anything could happen and there would be no line, the same as it was before. He, of course, being a student and being told by somebody smarter than I was: he said, anything could happen and I wouldn't want to bet for sure what would happen. Now, when I said I'm confident, my confidence was partly based on the pictures that I drew for my students of supply and demand, and some kind of equilibrium. And today I tell my students this is a useful fiction; there is not one price of bread, many different kinds of bread; even for a particular kind there is price variation for many reasons; but I'm still pretty confident that this framework of thinking about competition, even though there are an infinite number of goods, even though all quality isn't the same. I agree with you. Now, here's the connection to prediction. Once we have established the generic prediction as I called it in a paper that I wrote right around the time I wrote that 1984 paper you started with--if we start with the generic prediction that price controls produce queues and shortages, the next natural step in the process of science is to quantify the prediction. That was my next question. How much of a shortage, when will the shortage obtain, how long will it take for the shortage to be remediated, will the result actually be a surplus and if so how much and how long will it take for that surplus to occur, and will the result of that surplus be a subsequent, another shortage, and how much will that be and when will it occur? The process whereby we move from generic to specific predictions, something characteristic of meteorology--something that moved in our own time--and agronomy, some biological phenomena: that process is one we have never seen in economics. If we get a lot more specific and say given the law of supply and demand and the shape of demand curves, there are elasticities and cross-elasticities and income-elasticity effects of the sort you might find in what used to be called the Slutsky equation, there's a natural research program that suggests itself of attempting to establish the values of those parameters and to actually determine the shape of the curve. And you and I both know as economists and students of economics two things: first that no economist has ever successfully establish the values of those for a reasonable range of commodities over a reasonable time period. And we now know, indeed we could have known, back in the 1940s when these kinds of projects may have suggested themselves, why they cannot be accomplished. Well, there was a large research program that try to estimate elasticities--and by that, for the noneconomists out there that's one particular way of measuring how responsive people's purchasing demands are to changes in price. It's like actually determining the exact shape of the downward sloping demand curve, exactly how quickly does it slope, what is the slope at each point, and how does you change over time. So thousands of hours maybe hundreds of thousands we spent on the enterprise, and I think one of the things we learned from that, perhaps--I'm going to give it the benefit of the doubt--is that some goods are more responsive than others. To which many people would reply: Yes, so what? But I'm curious why you think that project was a failure. Certainly they could've said in defense of it: it's approximate, it's close, it's in the ballpark. Why do you think it was a failure and why do you think we should have known that in advance? I think that the reason it was a failure was that the demand curves are intertemporally unstable to the degree that once you establish one with elasticities at a given time, for a given commodity, its usefulness in predicting consumer behavior in the next period is relatively weak and gets weaker as subsequent periods occur. I think the technology for establishing these values, the difficulty, the cost combined with the limited usefulness of them--literally useful maybe from a week to a week--made it a mug's game to actually do this work. And what's worse: the intervention of new products, new tastes, and various kinds of complementarities and substitutivities among consumer durables--anybody who understood that deeply enough understood how pointless it was to try to calculate elasticities. Because even if you got them right it was for a particular commodity, a particular time, in a particular market and nothing much of interest followed. Certainly nothing precise enough to enable you to make a dollar followed about the market demand and supply of even a closely similar commodity in the next week. Even gasoline prices or bags of wheat, it turned out that this entire enterprise was always overtaken by events at such a rate that it was never worthwhile doing the empirical work necessary. Now, what does that suggest? We just need better techniques. It could. It could suggest that you've got the wrong theory altogether--we don't need this theory, it's not the way in which to produce a predictively powerful science of human economic behavior. Or it might be the even more radical resolution that there is no such theory within reach of human beings. That's of course the Austrian view. That would be the critique of Hayek in 1945, "The Use of Knowledge in Society," a paper I reference here about every seven shows. Funny, because I think about it almost every other day. I think that's probably the single most important paper in the Austrian tradition. No doubt. And Vernon Smith confessed to having read it numerous times when I interviewed him about it. It's much more important than The Road to Serfdom. I agree. The reason Hayek should be celebrated as a great economist and not the silly claims made on his behalf by conservatives and Republicans to try to sandbag economic responses to our current political economic downturn. Well, I agree with the first part of that. I know you do. Not quite sure about the second. Quite provocative. Maybe we'll come back to that, because I'd like to hear some of those specifics. It's certainly true that Hayek and all great thinkers are sometimes used for political purpose, sometimes in ways that are not fair to their origin.
34:34Let's go to biology. I know you've spent a lot of time thinking about biology, and I've spent a little time. Which, as Pope said: a little knowledge is a dangerous thing. I want to make a claim and I want to get your reaction to it. My claim is that the value of economics--I don't know whether it's a science or not. To the extent that it is a science, it is like biology; and I will leave for you to talk about the science-ness of biology. But by that I mean that economics is a very powerful way of helping us think about very complex systems, where things are connected. And I've written--and been maligned occasionally by bloggers--that just as we would never ask a biologist to predict how many beavers would be alive in an ecosystem if we chopped down 20% of the trees in this region because that's something that biology isn't good at predicting, similarly we should not ask economics to do something it is incapable of doing: which is, if the government spends another $820 billion it will create 1.7 million jobs or 1.74 million jobs. And I say that because, one, I'm not sure of the underlying theory that leads to that prediction, and since I can't confirm it through observation and empirical tests--that is, 5 years from now when I go to check the beaver population, so many things have changed in the meanwhile that I don't have data on, I can't confirm if the original theory was correct. I have that same problem with macroeconomics. And therefore I'm left with the reality that economics is not good for the questions that many people demand it to answer; but that does not mean economics is useless, just like biology is not useless. What's your reaction to that? I think my reaction to that is that you underplay the predictive powers of biological theory and that the reasons why some biological processes are predictably recalcitrant are on a continuum, with reasons why many economic predictions are likely to be overtaken by events and to prove either useless or falsified. And there are some further variables that complicate economics's recalcitrance to predict. So, let me walk through that. First of all, it's certainly the case that biological processes have a kind of instability that is also to be found in economics. And we understand that instability very well. It's basically the arms race instability of strategic interaction. It happens that in the biological realm, the realm controlled by natural selection, that adaptations inevitably become part of the environment, which filter for new adaptations. The random variation that produces traits that then get filtered by the environment is the process whereby biological phenomena increase their adaptation, increase their fitness up to say levels close to perfection in a very stable environment. Give a biological system long enough in a stable enough environment and there will by natural selection alone be a very strong drive of traits of the biological system to optimality. The trouble is that environments are not universally stable. Just like economics. Over geological time, there's a huge amount of stability. But once organic, biological phenomena begin to be parts of one another, then each organism's trait becomes part of the environment of the other organisms that it's competing with. And the result is that it incites variation or at least selection among the random variants in the other organisms for ways of taking advantage of for itself. And this is of course what produces the arms race. Arms races don't become really serious--that is to say they don't become fast enough--that we need to worry about them in understanding biological processes probably until you get to the interaction of different species with one another: predator with prey. But even in the case of predators with prey, predictions are still possible, and that's why ecology is increasingly powerful a science. And we can make some predictions about population densities over seasons. I think you gave the example of beavers. It's just in ecology. And just as we've become reliably able to predict the weather over the last couple of years because of the combination of increased data points and cheaper supercomputers, similarly we can do this increasingly in biology. We can even predict in some cases what the likely DNA sequence that's going to be selected for by way of a mutant bacterium by a new pharmacology. So, at the level of microbiology, let's say, like microeconomics, there are some significant improvements in predictive power. And even at the level of macrobiology and ecology there are at least the prospects of similar improvements because the power of calculating devices are improving faster than the arms race environmental changes. And none of that is true in economics.
41:00Well, I don't know. It reminds me of our earlier discussion. So let me challenge you and see what I'm missing. With a small enough ecosystem and enough knowledge of the participants in the biological case, you are saying we could make some predictions about whether the population would grow or shrink in response to those exogenous changes like chopping wood. We might even be able to make predictions about the likely emergence of novel traits. So, that's cool. And that I would argue is not too dissimilar from saying if you put on a price ceiling you'll get a shortage. You've challenged that conclusion by saying: but how much of a shortage, how long would it take? Do you think we have made enough progress--and I'm going to pick a big enough ecosystem, because I don't want to talk about the microeconomics of my class; I want to talk about the market for gasoline. So I want to go micro-micro, then micro, then macro. And similarly for biology, I want to go lab, ecosystem, and then interacting ecosystems. Do you think for an ecosystem such as, say, Yellowstone Park, that we would have decent predictions about, say, the reintroduction of wolves, which was done in the last decade or so, and the impact on the elk population? Yes. We can make some vague, general, qualitative predictions about the elk population. Can we make some quantitative ones? I don't know whether at this point ecologists are actually making quantitative predictions and seeing them borne out. But I can certainly anticipate that they will be able to in the foreseeable future. That's what people say about economics, too, by the way. Really? Yes, they do. In fact, Ricardo Reis, a guest on this program, neo-Keynesian at Columbia, a first-rate scholar, said: We've mastered monetary theory but we need more time for fiscal theory and we'll get there eventually. I disagree. I don't think we'll get there. When I joked about being criticized in the blogosphere, I made that remark on this point about beavers: somebody mocked me and said biologists do that all the time; they predict how many beavers there are going to be in 10 years; we're required by government law often to make predictions about what's going to happen. And I wrote the guy back and said: What's the verifiability of that prediction? Very low. Very, very difficult. Ten years now and certainly you wouldn't get the answer right even if you could verify. And certainly ten years is probably too long. And my own view about economics, at least in the case of macroeconomics, the impediments, the obstacles to a similar improvement in predictive power are merely ones of degree, not ones of kind. Meaning? Meaning that since I'm committed to the claim that biology either already does or will soon be in a position to make quantitative predictions about, say, predator/prey relations, similarly there is in principle, no philosophical, no logical, no causal obstacle to predictions of the same kind among macroeconomists. What would be an analogy there? Money supply and interest rate. Short-term Phillips' curve. Money supply and inflation. Inflation and employment. I can well imagine that insofar as these macro-relationships are symptomatic of traits of adapted institutions that are competing with one another, and that might be in some relatively short-term local equilibria over a long-enough time that those kinds of predictions are at least in principle possible. So, I used to believe that it was impossible to move economics from generic to specific predictions. I'm now inclined to think it's just extremely difficult. I don't know. It's an interesting question. I think the Great Moderation--so-called, this period of relative economic calm between, say, 1982 and 2008--that encouraged a lot of people to believe we had made fundamental progress in our ability to measure and steer the macroeconomy. Now, John Taylor, who is an important part of that argument and who has been a guest on this program a number of times, would argue that the reason that we can't do that any more is because we failed to learn the lessons of that time; but many others would argue that it was a game we could not really play. And my analysis is that it was a period of local equilibrium between competing economic forces which eventually broke down in an arms race. And the breakdown is the sub-prime mortgage denouement of 2008, and that is the result of the technological computational powers of traders exploiting the heuristics of non-financial economic agents looking for relatively safe places to put their investments. Reaching a point where the non-rational heuristic economic agents realized that they had start using the same tools as the people who were exploiting them. I look at it differently, but I don't want to go into that now other than to point out that I think government policy that made it harder for people to bear their losses and keep their gains--particularly bear their losses--removed the natural force that would have made those analytical tools that were dangerous less so. I agree with that, absolutely. Vernon Smith really is the right person here; and he's not won the day. But he's won my day, obviously. You and I both have the same analysis of what happened. It fits my bias, so I have to be careful. But the idea that people make mistakes all the time, they get overly excited, they are prone to exuberance; but markets have a natural way of punishing that when people get out of hand, and we've made it harder for markets to do that. And I see it as arms race; and arms races always end in some kind of a re-established equilibrium.
49:04You are a philosopher of science. You are interested in the philosophy of science. Most economists pay no attention--and I assume most scientists pay no attention--to these kinds of discussions. Absolutely. Most economists don't but many scientists do. But the most famous work in the philosophy of science would be Thomas Kuhn's The Structure of Scientific Revolutions, at least in modern times. I can't agree with that. Thomas Kuhn's book is of course an extremely important work in the history of science and it has been one that has been used as a stick with which to beat off the attack on non-scientific disciplines with; but I would hardly characterize it as the most important. I would accept it as the most influential. Maybe I should have said most famous, most well-known. My crude estimate of science's response to that is negative. Most scientists don't find it persuasive. Is that true? Most scientists do not find Kuhn's work on the nature of scientific progress correct. They certainly do not find the extreme version, the extreme interpretation of his view, which is common in the humanities and in some areas of interpretive social science as plausible. However, Kuhn himself rejected that interpretation and was appalled throughout the last 20 years of his life at the way in which his views had come to be interpreted. Interesting. But my general question is--not that it's decisive or perhaps even important--but I'm just intrigued by the fact that philosophers study science and scientists and they are not studying ants. They are studying human beings. And the human beings they are studying don't always pay much attention to it. What's your reaction to that? There's a famous quote of that great philosopher of science who didn't think he was one, Richard Feynman, who once said: The philosophy of physics is to physicists as ornithology is to birds. And that was his way of writing off the discipline of the philosophy of science. It was a wag and I wish I was he who said: In fact, physicists know about as much about the philosophy of physics as birds about ornithology. Good comment. I have to say: It's not the function of the philosophy of science to tell physicists what to do. It's not even the function of the philosophy of science to tell chemists or even biologists what to do. But there are problems in these disciplines and that biology is probably the best example that are conceptual, and that have attracted the interest of philosophers; and that philosophers have made considerable contributions. And that's one reason why the biologists are inclined to listen to us, or at least to talk with us. If you know much about the history of physics in the last 100 years, especially Einstein and quantum mechanics, you know about the influence of minds like Gottfried Leibniz's and Bishop George Berkeley's influence on the understanding of space and time. And you know about the difficulties about the interpretation, in making intelligible quantum mechanics. For the physicists to say: Shut up and do the computations at the back of a chapter, that's all you need, is probably safe and effective. Good career advice for graduate students. Right. And this is the point that Kuhn made about normal science that the discipline imposes. But science eventually and always finds itself faced with anomalies. And one it begins to be to treat those anomalies as central--and this is Kuhn talking. The philosopher has a role to play. Now ask yourself whether economics as a discipline, rolling on from Smith through Walras to Samuelson and Debreu and Alman and Scarf and Arrow has yet to accept or recognize that there is an anomaly. I suggest that there are some economists, and Paul Krugman is probably the best example, who are pounding their hands on the table and insisting there is a big bad bug in our discipline. And we have got to figure out what to do with it. And insofar as Kuhn provided us with an account of how normal science develops into a revolutionary period, Kuhn may help us understand that is one in economics. Now, are the economists going to solve their problem by paying attention to the philosophers? Well, it may concentrate their minds the way the gallows do; as said by Dr. Samuel Johnson to concentrate the mind wonderfully. But the solution isn't going to be found in the philosophy of science and we don't purport to provide it. That's not our function. And that wouldn't even count as good philosophy of science. Should economists pay attention to the philosophy of science? A bit. But not a lot. The solution to their problems is not going to be found in our discipline. But that they have a problem and what the broad outlines of that problem and what the blind alleys down which they shouldn't proceed in seeking solutions to that problem are, that's probably things we can help are. Probably not me, because I don't do much of this any more. But I think if you think about the work of, say, Dan Hausman or an economist like Kevin Hoover, or any number of others--we are likely to say things that may jog the complacency of the discipline.

Comments and Sharing



TWITTER: Follow Russ Roberts @EconTalker

COMMENTS (19 to date)
Dan H writes:

Many years ago when I was a young MBA student in China, I visited the main bookstore in Beijing to look for Chinese books on economics. I was shocked to find that not only was there no Economics section in the store, but further, the econ books were kept in the Politics section. At the time I thought that was terribly wrong, since economics was a objective science that should never be mixed with politics. Many years later, I know better.

Similar to economics, astrology is also supported by complex mathematical equations, but most would not imagine that set of belief systems to be a science.

BZ writes:

This was an interesting discussion as usual. I think I caught a disconnect between Dr. Roberts and his guest on what exactly is wrong with economics. Dr. Roberts has usually poked fun at it's pretended precision, a point he reiterated with the beavers example. Mr. Rosenburg gave me the impression that his skepticism is far deeper -- that it goes to the very assumptions of economics itself. Thinking specifically of his talk about the attempt to prove the relationship between incentives and outcomes, and how it led to perfect competition models, which he explicitly dismisses. It makes me wonder whether Mr. Rosenburg believes that incentives matter.

@Dan: I seem to recall that, in the 19th century, economics was commonly known as "political economy". :)

Schepp writes:

A thought experiment. May be economics is like probability densities of particles in physics.

Physics, to a great extent, is people making up rules to best describe phenomena. It is true that this has some pretty good predictive power, but in limited circumstances. Ask the best physist to write an equation of motion of 5 objects interacting with each other in space. She will tell you that there is not a simple equation that the interaction are to complicated.

The simple task of determining the location of an electron in an orbit around a nucleus is not provided by time and location trajectory but rather by a probablity density of location and momentum.

May be economics is like the electron. In many cases we cannot determine the exact effect or predict the future outcome, but over time we can provide a better projection compared to not using the methods of economics. In many cases economics such as those observations by Bastiate provide significant insight in the mechanics of choices we make and those that are made for us.

I acknowledge the skeptism of many definitive detailed findings of economics, but I am not ready to not think of it as a analytical assessment of conditions that can provide improved results as compared to not using economic methods. To me, well done economics has great value. I think that is why Dr. Roberts does this podcast and so many us of listen and participate.

marris writes:

I liked Professor Rosenberg's use of the phrase "arms race" to describe the instability of strategic interactions. I wish he had elaborated a bit more on why he's personally moved from an "economics can predict very little" to an "economics can predict a reasonable number of things" view.

[Maybe he would be kind enough to do this in another interview]

I tend to think of "biology vs. economics" as a difference in kind [not degree]. Birds don't understand ornothology or ecology, but *we* can understand the latest and greatest economic theories. So why can't we quickly diagonalize against them?

David Taylor writes:

I should start by saying that there is a huge disconnect between Rosenberg's claims about a first order predator prey nonlinear differential equation and the very complicated (systems of nonlinear differential delay?) equations that are likely needed describing economic problems discussed, even without a discussion on initial conditions and boundaries. The comment about a supercomputer winning an "arms race" with this problem is far more silly than the comments made about solving the information problem in economic distribution within a socialist state by computing a system of linear equations.

Thanks in great part to the efforts of Russ Roberts in the EconTalk podcast, I have some sympathy with the problems of economists and the study of the political economy. However, I spend enough time as a lab rat that I do not much appreciate what seems to be a nearly random series of uncontrolled experiments run on the national and world economy. If particle physicists are wrong about their theory and fail to find the Higgs Boson at both FermiLab and CERN, we will have wasted a good bit of time and money. If - as seems even more likely - the economists are wrong (along with the politicians listening to them) we will be FAR WORSE off.

---

Let's go back to some more Richard Feynman quotes (since this use of the above Feynman quote has probably got him spinning in his grave) to answer the question, is economics a science?

On the pervasive lack of critical thinking:
RPF: “Science is the belief in the ignorance of the experts”

On Russ Robert's often discussed problem of confirmation bias in economics:
RPF: “The first principle is that you must not fool yourself and you are the easiest person to fool.”

On Russ Robert's often discussed problem of falsifiability in economics:
RPF: “It doesn't matter how beautiful your theory is, it doesn't matter how smart you are. If it doesn't agree with experiment, it's wrong.”

And a paraphrase of a Feynman quote on the soft sciences from a interview video:
I know what it is to know something. I look at what they say they know, and they don't KNOW that.

Praxeologue writes:

If I understood Rosenberg correctly, he compareed economic forecasting to biology forecasting.

Superficially this is interesting but any reader of Hayek would know that in the field of economics we are dealing with other human minds, not numbers of squirrels. This is not a flippant point. Let's make the distinction greater, take water... H20. It makes no difference who knows H2O is what makes water. However, if we look into the Rosenberg future prediction machine and ask where the stock market will be in 10 years from now, clearly, this is going to change the behaviour of people today because of the output of the machine. That is to say the machine itself will start affecting the future. It's not plausible. Take it to another level, given that economic outcomes are really predictions of how men's interactions with nature and each other play out, why not speculate that we can predict the output of each man in the future, e.g predict and print the text of the Shakespeare of 2150 today...except, if we did that, would the future Shakespeare bother to write it... no, but then he would write something else or nothing at all, except, that would contradict the machine.. and so on.

muirgeo writes:

Excellent discussion by two very brilliant people.

Laymans summary: Economics is hard.


What also comes to mind is the contrast between theory and reality.

Think of the economic theory of price caps then think of economics happening in the real world with large masses protestors collecting on Wall Street while bankers symbolically drink expensive champagne from their balconies above as they LOOK DOWN on the protestors.... some who unlike the bankers are being arrested for supposedly violating an old law prohibiting the wearing of masks in public.

Nick writes:

One of the best guests recently, any chance you can have him on again to discuss Hayek. I was intrigued by his comments about the political right holding up Hayek for the wrong reasons and you said you would get back to it but never did.

George Gantz writes:

The exchange between Rosenberg and Roberts was wide-ranging, interesting and very perceptive, but I was disappointed that both missed the mathematical dimension to the issue of predictability. In fact, all of the disciplines they discussed, from biology, to meteorology to economics, as well as other scientific inquiries, are subject to the same limitations due to complexity. As the number and nature of interactions in any system increase, the calculations required to predict outcomes deterministically escalates. Eventually those calculations become infeasible, and outcomes can only be postulated as indeterminate probabilities. Rudy Rucker theorizes that the interactions in complex natural systems are actually calculations – some of which are “irreducible” – meaning the outcomes can only be determined by doing the calculation, or letting the natural phenomena proceed to its conclusion. John Conway’s Game of Life demonstrates that even a simple system with simple rules (e.g. cellular automata) can be irreducibly complex and impervious to prediction.

This is small comfort to scientists or economists who are trying desperately to model such complex systems. Indeed, the stakes can be quite high.

I do take issue with Rosenberg classifying certain disciplines as “science” on the basis of their predictive capabilities. Meteorology is a good example. While many improvements in weather forecasts have been possible, certain problems such as predicting where tornados will touch down or how long term climate change patterns will manifest locally, are likely to remain intractable. Predicting behavior of economic systems is no different.

Praxeologue writes:

As luck would have it, Greg Ransom just posted a blog which touches the issues raised in this podcast re predictability.

'Eventually, Hayek convinced Popper that his own “falsifiability” criterion of “science” was falsified by the example of Darwinian biology, economics, global brain science, and by other sciences involving what Hayek called “essentially complex phenomena” where only patterns could be explain via causal mechanisms involving innumerable, functionally-defined, or open-ended causal components, and no “falsifiable” particular predictions at a particular places in space and time could be made.'

http://hayekcenter.org/?p=4999


Jonathan writes:

I agree with Nick (Posted September 27, 2011 10:13 AM). As an undergrad I studied both philosophy and economics, and this talk has been a dream come true. Do have Rosenberg back on the show.

rovescaito writes:

I often wonder if the only science/social science we should teach that involves human behavior is the Laches, the socratic dialog that demonstrates the inability to define the terms of the supposed dichotomy courage/foolhardiness. Aristotle then proceeded to unwittingly institutionalize the problem by stretching it into a continuum what was simply indefinable using the two concepts, such that today all theories based on language (and for all i know all mathematical theories may be ultimately based on dichotomy as well) are at least 3 removes from what they are trying to explain (the arbitrary selection of meaningful attributes from the whole that embodies them, the imperfect nature of those attributes themselves or the inability to define them in a single useable concept, and the movement away from their naturally occurring condition via an imposed dihcotomy, or reason) setting up the opportunity for a Richard Feynman to sound wise ticking off rather dull imitations of Oscar Wilde.

The dedicated pursuit of scientific/theoretical knowledge is never wasteful, although perhaps economically a waste of resources, but it seems that so much more is gain from it in the tearing down rather than the building up of bodies of knowledge. The Laches itself is removed at least a step from the embodiment of human qualities in the immortals where the precise departure of foolhardiness from courage is not necessary to determine as the act of courage/foolhariness is always embodied within the context of a specific action. I find that my understanding of the price mechanism is much richer the less external restrictions and the more i just imagine it working itself out in real transactions. Similarly I find that the conclusions of reason and experiment are much more potent and personally useful when embodied in a novel of the quality of a Middlemarch, much as courage might be embodied in the figure of Appolo. Such literary/utalitarian theory of knowledge would necessitate abandoning the belief in the ability of transmitting precise knowledge from person to person and between the generations however.

ric caselli writes:

Why oh why Academia cannot learn from meteorologists?

In South Florida, we deal with hurricanes.

To protect life and property, we rely on hurricane forecasts, which are based on spaghetti models, each predicting trajectory and strength based on different mathematical models. It works!!!

It separates what is possible from impossible and even tries to provide a humble idea of what might be likely.

It is self evident (excluding tenured professors) that since at least the 1500's different economies at different times behaved following Schumpeterian, Keynesian or even Marxist models for a period.

In the last 2 decades or so, engineering has worked out ways to handle complexity. In science still prevails the arrogant, over simplifying approach of the 70's

Keep searching for a better way Ross.

John Berg writes:

Perhaps Economics needs that period of exploration, collection, organizing, taxonomies, and cataloging that the 18th century produced for the natural sciences. Clearly that period never really ends.
Another suggestion, perhaps Economics needs that kind of mathematics that helped make Computer Science a science. DBMS work needed a basis in Shannon's Information Theory, and the use of Logic to permit publications that allowed peer reviews. I recall that Ted Codd, creator of one form of DBMS theory, began as a Cellular Automaton theorist.

John Berg

Dmitry writes:

For me it is also "a dream come true" podcast. Thank you Russ.

Let me put some thoughts two.

I think that two essential features of physics-related sciences are:

1) when physics deals with something habitual to our day-to-day experience (like simplest motions of bodies like cars, trains, etc), it quantifies it to the highest precision (even the speed of light is known to a very high accuracy);

2) the mathematical apparatus of physics helps us understand nontrivial phenomena, which otherwise would have been misunderstood by our mind accustomed to everyday events (relativity, the existence of quantum of energy, etc).

And I think that neither of these two properties one can find in economics: 1) the precision of economics was discussed at length in the podcast and 2) through my undergraduate courses I have never seen the necessity for cumbersome mathematics; and I haven't ever noticed any nontrivial implications of economic theory - "the elasticity of demand varies between goods, so what? =)" - exactly (maybe I can mention Keynes, who tried to show us some paradoxes and tried to explained them, but that's all that impressed me).

Schepp, all of what you are saying about complex systems in physics and economics is true, but the eventually physics shows its power in engineering (airplane is a rather complex mechanism).

For me the saddest part is how much intellectual effort is wasted in tackling toy models in economics for no good reason (strangely enough, economists think that long arithmetic expressions, some FOC's and first order differential equations are higher level mathematics. Then what is topology, differential operators, Galois theory?).

Charlie writes:

As best I can tell, Rosenberg's definition of science was that it be able to predict something to Rosenberg's arbitrary sense.

If Rosenberg concedes that we can say something predictive about a price floor, he's really conceded everything. He says we can't say enough about elasticities, but can't we? Would he also concede that the response for a price floor on crude oil would be different than on Prada hand bags? I would have liked Russ to just go through intro micro text book and see what else he concedes. Can I say something about a hurricane knocking out several oil rigs (a supply shock)? Can I say something about the Pope encouraging followers to eat fish on Fridays (a demand shock)? Can I say something about complements and substitutes?

Most people are mad at economists, because the questions they ask are too hard. A physicist can predict where a projectile shot from a distance with no wind will land, but if the wind blows unexpectedly and knocks the projectile of course, people don't blame the physicist. They understand there are limits to the physicist's knowledge. That the physicist was using a model that only held under certain assumptions. Economists can say many useful things, which is one reason it is becoming more and more popular.

Maybe there is something philosophy of science has to offer. I have no idea. I have not yet heard a rigorous definition of what a science is. And if economics is, in fact, not a science under this definition, it is not clear whether economics should strive to be a science.

The guest liked experimental and behavioral economics. Why? We have gotten much better at predicting what people do under very controlled circumstances, but they haven't helped much in predictions of less controlled environments. We now know many ways in which people behave irrationally, but the size and type varies quite dramatically with context. This line of research has done little to help us make more precise predictions that Rosenberg seemed to require. Though, I guess if a science is a discipline where practitioners do things that seem sciencey, then experimental economics has helped.

Just curious, is medical science a science? As best I can tell, it is just as bad or worse at making predictions. Most published studies cannot be replicated. Even "successes" have only limited usefulness. "I know if I give this drug to 100 it will cause a decrease in symptoms in 25, no effect in 70, and make 5 very sick. We aren't really sure why. And this probably won't be replicable."

David Taylor writes:

To Praxeologue:

I believe that falsifiability is a crucial criteria for a science. Let's stay with physics as a prototype "science." There is (even now) an argument about a Popper type falsifiability in physics theory of string theory. To stay with the Feynman thread, he was quite vocal in criticizing physical theories without quantifiable experimental predictions. The search for the Higgs Boson is a case in point as we would learn a great deal from not finding it and falsifying the theories that predict it.

I think that an interesting question involves the tie between a science and mathematics. The example of evolutionary biology is tied to the mathematics of genetics and, in this context, some theories are clearly falsifiable. (Also, the math and logic does not have to look like algebra or calculus.) Can we recast parts of physics to not be a science if we use a non-mathematical circular statement of Newton's laws without predictive power? We should know enough to know what we don't know: especially basic things like the long time solutions of the three body problem. This mathematical foundation for science allows some hope of understanding the limits of what is predicable.

Falsifiability is key to understanding the limits of what can be predicted.

Further, even though I think an underlying mathematics formalism is necessary to have a science, it is not sufficient without falsifiability. In responding to a criticism of part of an economics text, Paul Samuelson pulled out a quote from Fermi's short book "Thermodynamics" that was at least as muddled as the statement of the economics principle. Both statements were correct, but like Newton's Laws they required the mathematics. Thermodynamics theory is falsifiable. I am not so sure that economics is falsifiable even when it is correct.

John Arkwright writes:

1. What Charlie said.

2. Given the sad state of reasoning on the economy that is put forth by those in power, economics is an important discipline to understand. The Nixon wage-price controls were foreseen by the Chicago School to be a disaster. Milton Friedman's permanent income hypothesis accurately predicted that Bush II's stimulus--sending checks--would result mostly in savings and debt reduction. Fredric Bastiat could have advised Christina Romer on Obama's stimulus package Alas, Christina Romer, the researcher, could have advised Christina Romer, the person who was paid by the Obama administration to favorably represent his policies, on Obama's stimulus package.

The legion's commanders need to understand the principles of warfare. But that does not mean that they can predict where Hannibal is hiding the troops or whether the troops can negotiate the swamp.

Understanding the traffic flow helps me to plan to be at the meeting on time. An accident on the highway may undo my planning, though.

Scott M writes:

I enjoyed listening to Prof. Rosenberg's very careful use of language -- it made me nostalgic for my days as a philosophy undergrad.

I'm puzzled about the claim Rosenberg made that the theorists of the 40s could and should have known a priori that their efforts at prediction were doomed to failure. About 25 minutes into the podcast he says:

And what's worse: the intervention of new products, new tastes, and various kinds of complementarities and substitutivities among consumer durables--anybody who understood that deeply enough understood how pointless it was to try to calculate elasticities. Because even if you got them right it was for a particular commodity, a particular time, in a particular market and nothing much of interest followed. (emphasis added)

But later in the podcast he says there is no reason why economics shouldn't be able to make the kind of quantitative predictions that are longed for (around 43 minutes in):

Meaning that since I'm committed to the claim that biology either already does or will soon be in a position to make quantitative predictions about, say, predator/prey relations, similarly there is in principle, no philosophical, no logical, no causal obstacle to predictions of the same kind among macroeconomists.

What has changed since the 40s that makes him more sanguine about current economics's prospect for prediction? Simply better computing power? If the researchers of the 40s ought to have known better than to waste their time, how does an economist today know differently? In other words, what are Rosenberg's reasons that economics should, at some point, be able to make quantifiable predictions?

P.S.
You folks at econtalk do a tremendous work with this podcast: new podcasts every Monday, the detailed transcripts on this web page, the thoughtful links posted for each episode -- bravo! and thank you!

Comments for this podcast episode have been closed
Return to top