1. Economics as a policy science
Part of the attraction and the promise of economics is that it claims to describe policies that will improve peoples’ lives. This is unlike most other physical and social sciences. Sociology and political science have a policy component, but for the most part they are concerned with understanding the functioning of their respective subject matters.
Physical science, of course, has the potential to improve peoples’ standards of living, but this is really a by-product of science as an intellectual activity.
In my view, many methodologists have missed this essential feature of economic science. It is a mistake to compare economics to physics; a better comparison would be to engineering. Similarly, it is a mistake to compare economics to biology; a better comparison is to medicine. I think that Keynes was only half joking when he said that economists should be more like dentists. Dentists claims that they can make make peoples’ lives better; so do economists. The methodological premise of dentistry and economics is similar: we value what is useful. None of the ‘‘policy subjects’’--- engineering, medicine, or dentistry---is much concerned about methodology, and economists, by and large, aren’t either.
When you think about it, it is quite surprising that there isn’t more work on the methodology of engineering or medicine. These subjects have exerted an enormous influence on twentieth century life, yet are almost totally ignored by philosophers of science. This neglect should be contrasted with with other social sciences where much time and energy is spent on methodological debate. Philosophy of science, as practiced in philosophy departments, seems to be basically concerned with physics, with a smattering of philosophers concerned with psychology, biology, and a few social sciences.
I think that many economists and philosopher who have written on economic methodology have not given sufficient emphasis to the policy orientation of most economic research. One reason for this is the lack of an adequate model to follow. There is no philosophy of engineering, philosophy of medicine or philosophy of dentistry---there is no model of methodology for a policy science on which we can build an analysis. The task of constructing such a theory falls to economists. This is, in my view, one of the most interesting problems for those concerned with methodological issues and the philosophy of the social sciences.
2. Role of theory in a policy science
Given my view that economics is a policy science, if I want to defend a practice in economics, then I must defend it from a policy perspective. So I need to argue about how economic theory is useful in policy. The remainder of the paper will consists of list of several such ways. The list is no doubt incomplete, and I would welcome additions. But perhaps it can help focus some discussion on why economists do what the do, and how theory helps them do it.
Theory as a substitute for data
In many cases we are forced to use theory because the data that we need are not available. Suppose, for example, we want to determine how a market price will respond to a tax. We could estimate this effect by running a regression of market price against tax rates, controlling for as many other variables as possible. This would give us an equation that we could use to predict how prices respond to changes in taxes.
We rarely have data like this; taxes just don’t change enough. But if people only care about the total price of a good, inclusive of tax---a theory---then we can use estimated price elasticities to forecast the response of price to the imposition of a tax.
This uses a theory about behavior---people will respond to the imposition of a tax in the same way that they respond to a price increase---in order to allow data on price responses to be useful. We can use the theory to forecast the outcome of an experiment that has never been done.
Here is another, slightly more esoteric example. Consider the assumption of transitivity of preferences mentioned briefly above. This theory asserts that if A chosen when {A,B} is available and B is chosen when {B,C} is available, then we can predict A will be chosen when {A,C} is available. This is certainly a theory about behavior; it may or may not be true.
If we had data on choices between all pairs of A, B, and C, then the theory wouldn’t be necessary. When we want to predict the choice out of the set {A,C} we would simply look at how the person chose previously---that is, we would just use brute induction. And we know why induction works---it has always worked in the past!
But we rarely observe all possible choices; typically we observe only a few of the possible choices. Theory allows us to interpolate from what we observe to what we don’t observe. In the case of the {A,B,C} example brute induction requires observing all choices the consumer could make from the various proper subsets available, which requires 3 choice experiments. But if the assumption of transitivity holds, then 2 choice experiments are all we need. The theory of consumer choice allows us to economize on the data.
Naive empiricism can only predict what has happened in the past. It is the theory---the underlying model---that allows us to extrapolate.
Theory tells what parameters are important and how we might measure them.
The Laffer curve depicts the relationship between tax rates and tax revenue. At some tax rates tax revenue decreases when the tax rate increases. It has been said that the popularity of the Laffer curve is due to the fact that you can explain it to a Congressman in six minutes and he can talk about it for six months.
The Laffer analysis demonstrates both good and bad economic theory. The bad theory is that inference that because the Laffer effect can occur it does occur. The good theory is that we can use simple supply and demand analysis to determine what magnitudes the elasticity parameters have to be for the Laffer effect to occur. We can then compare the magnitudes of estimated elasticities to estimated labor supply elasticities. In the simplest model a marginal tax rate of 50% requires a labor supply elasticity of 1 to get the Laffer effect. The theory tells us what the relevant parameters are; without the theory, one would have no idea of the relevant parameters are. Indeed, if one examines the rather sordid history of the use of the Laffer curve in public policy debates in the U.S. this becomes painfully clear.
For another example, consider the theory of investment in risky assets. I take it as given that risk is a ‘‘bad.’’ Therefore when wealth goes up, people may want to purchase less of it. On other hand, you can afford to bear more risk when you have more wealth. So an argument based on intuition alone shows that investment in a risky asset can go up or down when wealth increases. A systematic theoretical analysis shows what the comparative statics sign depends on: how risk aversion changes with wealth. So the risk aversion parameter is the one you want to estimate in order to predict how investment in risky assets changes with wealth. Conversely how investment changes with wealth tells you something about how risk aversion changes with wealth.
Theory helps keep track of benefits and costs
I indicated above that the sorts of optimizing models used by economists serve the purpose of providing guidance for policy choices. Indeed one of the important roles of economic theory is to keep track of benefits and costs. The idea of opportunity cost is a fundamental one in economics, and would be very difficult to use without a theoretical model of economic linkages.
This brings up the important point that the correct way to measure an economic benefit or cost can only be determined in light of a theoretical model of choice: a specification of what objectives and the constraints facing an economic agent.
Consider for example, the practice of computing present value or risk adjusted rates of return. These computations are only meaningful in light of a model of choice behavior. If the model of behavior does not apply, the policy prescription cannot apply either.
Benefit-cost analysis is only one small field of economics. But the idea behind benefit-cost analysis permeates all of economics. If economic agents are making choices to maximize something, then we can get an idea of what is being optimized by looking at agents’ choices. This objective function can then be used as an input to making policy decisions. In some cases, one may need a quantitative estimate of the objective function. In other cases, one may want to show that one kind of market structure, or tax structure, may do a better job of satisfying consumers’ objectives than another. But the basic framework of moving from individual objectives, to individual choice, to social objectives and social choice is common to many, many economic studies.
Theory helps relate seemingly disparate problems
If one describes a model in a purely mathematical way, it often happens that the underlying equations will describe a rich set of economic phenomena. The classic example of this phenomenon is the Arrow-Debreu general equilibrium model. The concept of ‘‘good’’ can be interpreted as a physical commodity available at different times, locations, or states of nature. One theoretical model can thereby provide a model of intertemporal trade, location, and uncertainty.
Another example from general equilibrium theory is the First Welfare Theorem. This result shows the intimate relationship between the apparently distinct problems of equilibrium and efficiency.
A third example is that a formal analysis of the problem of second-degree price discrimination shows that it is equivalent to the design of an auction or the determination of optimal provision of qualities. Quality discrimination, auction design, and nonlinear price discrimination are essentially the same sort of problem.
Each of these insights came from examining an abstract theory. Once the the ‘‘irrelevant’’ details are stripped away, its becomes apparent that the same essential choice problem is involved.
Theory can generate useful insights
Let me illustrate this role of economic theory with an example. In the U.S. most interest receipts are taxable income, but many kinds of interest payments are tax deductible. This policy has been criticized as ‘‘subsidizing borrowing.’’ Does it?
The answer depends on the tax brackets of the marginal borrowers and lenders. If the tax brackets are the same, for example, the policy has no effect at all on the equilibrium after-tax interest rate. The supply curve tilts up due to the tax on interest income, but the demand curve tilts up by the same amount due to the subsidy on interest payments. This is a simple insight, but it would be very difficult to understand without a model of the functioning of the market for loans.
A theory that is wrong can still yield insight
Pure competition is certainly a ‘‘wrong’’ theory many markets; pure monopoly is a wrong theory for other markets. But each of these theories can be very useful for yielding significant insights for how a particular market functions. No theory in economics is ever exactly true. The important question is not whether or not a theory is true but whether it offers a useful insight in explaining an economic phenomenon.
In my undergraduate textbook I examine a very simple model of conversion of apartments to condominiums. One result of the model is that converting an apartment to a condominium has no effect on the price of the remaining apartments---since demand and supply each contract by one apartment.
This result can hardly be thought of as literally ‘‘true.’’ There are a host of reasons why converting an apartment to a condominium might influence the rent of remaining apartments. Nevertheless, it focuses our attention on a crucial feature of such conversions: they affect both the supply and the demand for apartments. The simple supply-demand framework shows us how to start thinking about the impact of condominium conversion on apartment prices.
Theory provides a method for solving problems
I take the method of neoclassical microeconomics to be 1) examine an individual’s optimization problem; 2) look at the optimal equilibrium configuration of individual choices; 3) see how the equilibrium changes as policy variables change.
This methods doesn’t always work---the models of behavior or equilibrium may be wrong. Or it may be that the specific phenomenon under examination is not fruitfully viewed as an outcome of optimizing, and/or equilibrium behavior. But any method is better than none. In the words of Roger Bacon: ‘‘More truth arises through error than confusion.’’
Methodological individualism is a limited way of looking at the world, no question about it. It probably doesn’t do very well in describing phenomenon such as riots or class loyalty. Certainly this sort of individualistic methodology works better for describing some sorts of behavior than others. But it is likely to add insight to all problems.
Theory is an antidote to introspection
Most people get their economic beliefs from introspection and their personal experience- --the same place that they get their beliefs about most things. Economic theory---and indeed science in general, can serve as an antidote to this kind of introspection.
Consider, for example, the widely held belief that all demand curves are perfectly inelastic. If the price of gasoline increases by 25%, a layman will argue that no one will change their demand for gasoline. He bases this argument on the fact that he would not change his demand for gasoline.
Indeed, it is perfectly possible that most people wouldn’t change their demand for gasoline...but some would. There are always some people at the margin; these people would change their demand. At any one time, most people are infra-marginal in most of their economic decisions. The marginal decisions are the ones that you agonize over. If the price were a little higher or a little lower, the results of your agonizing might be different, and this is what causes the aggregate demand curve to slope downward.
Another nice example of this phenomenon is free trade. It’s hard to convince a layman of the advantages of free trade since it is easy to see where the dollars go, but difficult to see where they come from. People have personal experience with imports of foreign goods of foreign goods; but they rarely encounter their own country’s exports unless they travel abroad extensively. Only by abstracting from introspection can we see the total picture.
A third example is the bias in perceptions of inflation: price moves are perceived to be exogenous from the viewpoint of the individual, but wage movements are personalized. Even if prices and wages move up by the same amount, people may feel worse off since they think that they would have gotten the wage increases anyway.
Verifying that something is obvious may show that it isn’t
One of the criticisms that economists have to deal with is that they spend a lot of time belaboring the obvious. Isn’t it obvious that demand curves slope down and supply curves slope up? But many theories that seem to be obvious turn out not to be. It may be obvious that demand curves slope down---but as the theoretical analysis shows, it is possible to have demand curves that don’t.
Economic theory shows that a profit-maximizing firm will decrease its supply when the output price decreases. But farmers often claim that removing milk price supports will increase the supply of milk since farmers will have to increase output to maintain the same income. The second effect sounds like it might be possible---after all, farmers wouldn’t advance the claim unless it had some plausibility. However, theory shows us that this particular claim cannot be true if the farmers attempt to maximize profits.
Strategic interactions are a good source of counterintuitive results. A simple analysis of a two-person zero-sum game shows that improving your backhand in tennis may lead to your using it less often.
It would seem that a public offer to match any competitor’s price is a sign of a highly competitive market. But when you think about the problem facing a cartel it is not so obvious. The prime problem facing a cartel is how to detect cheating on the agreed-upon prices and quotas. Offering to match a competitor’s price is a cheap way to gain information about what your competitors are doing. What appears to be a highly competitive tactic can easily be viewed as a device to support collusion.
Theory allows for quantification and calculation
According to Lord Kelvin, ‘‘When you cannot measure it, when you cannot express it in numbers, your knowledge is of a meagre and unsatisfactory kind.’'[1]
Theoretical economics gives us a framework to calculate and quantify economic relations. Consider the Laffer curve mentioned above. Laffer gave the existence proof, but it took some theoretical calculations to see what magnitudes were important.
In fact, one of the major differences between economics and the other social sciences is that in economics you can compute. There is very little computation in sociology, political science, history or anthropology. But economics is filled with computation.
Economic theory is useful since you can use it to compute answers to problems. They aren’t always the right answers---that depends on whether the model you have is right. (Or, at least, whether it is good enough for the purposes at hand.) But a desideratum of a good model is that you can compute with it: the model can be solved to determine some variables as a function of other variables.
In my view, it is impossible to learn economic theory without solving lots of problems. Richard Hamming, a highly prolific electrical engineer, once gave me some excellent advice about how to write a textbook. He told me to assemble the exams and problem sets that you want the students to be able to solve by the time they had finished the course, and then write the book that would show them how to solve them. In general, I have tried to follow this advice, with, I think, some success.
Economics is amenable to experimental verification
Because neoclassical economic models enables one to compute answers to problems, it is possible to compare the answers you get with the outcomes of controlled experiments. In my view, experimental economics has been one of the great success stories of the last 20 years. We now have rigorous ways to test models of human behavior in the laboratory. Some standard models, such as supply and demand, have turned out to be much more robust than we would have thought 20 years ago. Other models, such as expected utility, have turned out to be less robust.
But this is to be expected---if there were no surprises from experiments, they wouldn’t be worth doing. The growth of experimental economics has led many theorists to construct theories that simple, concrete and testable, rather than theories that are complex, abstract, and general. And experience in observing human subjects in the laboratory has no doubt contributed to the current emphasis on investigating models of learning. Laboratory observations have also been instrumental in alerting us to theoretical dead ends, such as some of the more convoluted refinements of game-theoretic equilibrium concepts.
I expect that the interaction between theory and experimentation will continue to grow in the future. As economists become more comfortable with experimentation in the laboratory, they will also become better at identifying ‘‘natural experiments’’ in real-world data. Such developments can only lead to better models of economic behavior.
3. Summary
I have argued that in order to understand why economic theorists behave in the way they do one has to understand the role of economic theory’s contribution to policy analysis. The fact that economics is fundamentally a policy science allows one to explain many aspects of economic theory that are quite mysterious otherwise.
___________________
1 However, the same poet whose praise for abstraction and synthesis I quoted in the introduction also once said: ‘‘...High Heaven rejects the lore of nicely calculated less or more.’’
Posts
