Thinking Like a Modern...

Thinking Like a Modern Macroeconomist

Maybe there is in human nature a deep-seated perverse pleasure in adopting and defending a wholly counterintuitive doctrine that leaves the uninitiated peasant wondering what planet he or she is on.

— Robert Solow

CHAPTER 29

AFTER READING THIS CHAPTER, YOU SHOULD BE ABLE TO:

1. Distinguish between the standard macro model and the modern macro model.

2. Trace the development of the standard and modern models of the economy.

3. Discuss the advantages and disadvantages of the standard macro model.

4. Explain what assumption each of the letters in “DSGE” from the DSGE model represents.

5. State three policy implications of the dynamic stochastic general equilibrium (DSGE) model.

6. Summarize the complex systems approach to macro.

When I told my editor that I was going to write a chapter on think-ing like a modern macroeconomist, she screamed. Why? Because she knows what modern macro is; it is a highly theoretical, highly mathemati-cal set of models with no simple application to macro policy. Modern macro is all the things that the standard macroeconomics isn’t. To give you a sense of how technical modern macro is, consider this sentence from the introductory chapter of top modern economists Lars Ljungqvist and Thomas Sargent’s recent graduate introductory macro text. In it they discuss the “spectral density matrix and the Fourier trans-form of the covariogram of a covariant stationary stochastic process.” Yeah right—ain’t no way anyone is going to dare to put that into an in-

troductory economics text and hope to sell any. 1 Most introductory textbook authors shelter you from such terminological thuggery, believing it will scare you half to death. But, as you’ve gathered, I’m not the usual introductory textbook author. I jump in where only fools tread, and I bring my readers with me. I do so not because I am a sadist, but rather because I am a teacher who wants my students to have a sense of modern macro, and I can’t give you that sense without letting you know that it is highly technical.

Why It Is Important to Know about Modern Macro Theory There are three reasons why I believe that you need to know about modern macro theory. The first is that if you’re learning macro, you don’t want to learn outdated arguments and theories. You might be left thinking that macroeconomists are

1Well, I just put it in—so see, editors, I dare. But to students reading this chapter, don’t worry; I put it there just to show that I could. I don’t expect you to memorize it; it’s definitely not going to be on the test.

coL75888_ch29_678-701.indd Page 678 5/27/09 11:31:45 AM user-f501coL75888_ch29_678-701.indd Page 678 5/27/09 11:31:45 AM user-f501 /Users/user-f501/Desktop/Users/user-f501/Desktop

Chapter 29 ■ Thinking Like a Modern Macroeconomist 679

stupid, and that certainly isn’t the case. Modern economists are bright—super bright—and macroeconomists are among the brightest of the bright. You deserve to be intro-duced to their models. A second reason is that modern macro models are fundamentally different from standard macro models — the models (such as the AS/AD and multiplier models and their derivatives) used by most applied macroeconomists . Modern macro models provide differ-ent frames of analysis than those standard macro models. (Remember back to Chap-ter 6 where I discussed the importance of framing.) Were modern macro models simply a more technical version of the standard macro models that I have presented to you in the last two chapters, there would be no need to present them. But they aren’t. They are fundamentally different, and they provide you with a totally different frame for thinking about macro problems. That different frame provides important caveats about applying the standard macro model that led to the many addenda that I added to dis-cussions of the policy implications of the two models I presented above. This chapter helps you understand where those caveats come from. A third reason why it is important to include a discussion of modern macro models is that these models are being used for policy decisions. For example, modern macroecono-mists from the University of Minnesota, V. V. Chari and Patrick Kehoe write, “Over the last several decades, the United States and other countries have undertaken a variety of policy changes that are precisely what macroeconomic theory of the last 30 years [i.e., modern macroeconomics] suggests.” They further state, “To what particular policies should policymakers commit themselves? For many macroeconomists considering this question, quantitative general equilibrium models [another name for modern macro models] have become the workhorse model, and they turn out to offer surprisingly sharp answers.” Most economists agree that the modern macro model demonstrates some logical problems with the standard macro models, which should make any user think long and hard before basing policy on them. Many economists are not so sure it does more than that. There is a lively ongoing debate in the economics profession about the extent to which modern macro models can be fruitfully used for policy analysis. Advocates argue that, at a minimum, the modern macro model provides an important counterbalance to the policy activism suggested by the standard macro models. The chapter proceeds as follows. First, I review the distinction between engineering models and scientific models that I presented in Chapter 6, and relate that distinction to why modern and standard macro models differ. Second, I briefly review the history of macroeconomic thinking and explain why scientific economists moved away from us-ing standard macro models, while applied policy economists stuck with them. Third, I provide a beginning student’s guide to the dominant modern macro model (the dy-namic stochastic general equilibrium model), explaining what I believe it captures and what it doesn’t. And, finally, I discuss some work on the frontier of modern macro sci-ence that is working to reintegrate the scientific and engineering models, and how that work challenges the policy precepts of both the standard macro models and the domi-nant modern macro models.

Engineering Models and Scientific Models As I’ve emphasized throughout this book, modern economic thinking involves reducing a question to a model, and then using that model to analyze a problem or a question. In Chapter 6, we saw the debate between traditional economists and behavioral economists about which model to use, and how different choices led to different policy focuses. In macro, we have all those same debates about modeling, but we also have another debate—whether we should be using what is essentially an engineering model — a model with loose

Modern economists are bright—super bright—and macroeconomists are among the brightest of the bright. You deserve to be introduced to their models.

Modern economists are bright—super bright—and macroeconomists are among the brightest of the bright. You deserve to be introduced to their models.

Q-1 Name two standard macroeconomic models. Q-1 Name two standard macroeconomic models.

There is a lively ongoing debate in the economics profession about the extent to which modern macro models can be fruitfully used for policy analysis.

There is a lively ongoing debate in the economics profession about the extent to which modern macro models can be fruitfully used for policy analysis.


680 Macroeconomics ■ The Macroeconomic Framework

formal foundations whose primary purpose is to guide thinking about policy , or a deductive scientific model — a model with carefully specified formal foundations whose primary purpose is understanding for the sake of understanding. Policy relevance is a nice sidelight of a sci-entific model, but it is not the primary focus.

Standard Models as Engineering Models and the Modern Model as a Scientific Model The standard macro models that you learned in the past two chapters are the models used by most business and government macroeconomists. They are best seen as engi-neering models. They are very institutionally oriented models that provide a framework for discussing macro policy. Modern macroeconomic scientists use deductive scientific models of the macro-economy such as the dynamic stochastic general equilibrium (DSGE) model. (I will explain the DSGE model below.) Modern macroeconomists see their models as only indirectly relevant for policy. For example, when Robert Lucas, a Nobel Prize–winning modern macroeconomist at the University of Chicago, was asked what he would do if he were appointed to the Council of Economic Advisers, he said that he would resign. That doesn’t mean that modern macro scientific models don’t have policy implications—they do. At a minimum they raise serious concerns about naïvely applying the simple standard engineering models that you learned in the last two chap-ters to policy, and they point out several important addenda to those standard models. Some modern macroeconomists go further, however, and try to use their scientific models to provide specific policy advice. For example, some argue that modern macro models tell us definitively that discretionary monetary and fiscal policy won’t work. While textbook authors (me included) bend over backwards to be neutral and present all sides (it’s not that we’re naturally fair—it’s just that we don’t want to lose sales), let me say now that such a definitive view is simply wrong. As the famous London School of Economics economist Lionel Robbins put it: “In the present state of knowledge, the man who can claim for economic science much exactitude is a quack.” (Who said I wasn’t opinionated?) The different focuses of the standard macro model and modern macro model are reasonable. Government and business economists need their models to be quickly ap-plicable. They use a combination of empirical and deductive models that have evolved over time and provide a reasonable framework for talking about macro problems. The standard macro models are rough and ready models that seem to have worked in the past. The models are meant to be used by practitioners who care much more about whether a model works than how it works. Modern macroeconomic scientists care first and foremost about the logic of their models: knowing that they work without knowing how and why just doesn’t cut it for them. The scientific macro model is much more formal and precise than are engineer-ing models. Whereas the engineering models include lots of hand waving and loose discussion appealing to intuition, the scientific model carefully spells out all the as-sumptions. Economists can argue about whether or not the assumptions are reasonable, but the logical implications of making those assumptions are clear. (Well . . . the models are extremely formal and complex, so the implications are really only clear to someone who has studied lots and lots of math!)

Engineering and Scientific Models Are Different Models Ideally, the scientific and the engineering models would work in parallel, with the en-gineering models being simplified and less formal versions of the scientific models. In

It is important to distinguish engineering models from scientific models.

It is important to distinguish engineering models from scientific models.

Q-2 Is the standard macro model a scientific or engineering model? Q-2 Is the standard macro model a scientific or engineering model?

The different focuses of the standard macro model and modern macro model are reasonable.

The different focuses of the standard macro model and modern macro model are reasonable.



that case, it wouldn’t be necessary to introduce you to the modern macro model—the standard macro model would have been your introduction. Unfortunately, that isn’t the case in macro. The scientific models are completely different from the engineering models. This differentiation exists because modern macroeconomic scientists have come to believe that the standard macro models are fundamentally unscientific, so they’ve been working on developing more scientific alternatives. At the same time, most applied policy macroeconomists don’t believe that these alternatives actually work for answering most policy questions, at least not yet. That’s because, in order to be a model that can be solved, the current scientific model has had to leave out many complications, and most applied policy macroeconomists think that these complications play an important role in the workings of the economy.

From the Keynesian Revolution to Modern Macro Models Let’s now consider how we got to the development of these two vastly different kinds of macro models. As is discussed in previous chapters, macroeconomics began in the 1930s with the Keynesian model. Classical macroeconomists (all macroeconomists before the 1930s) didn’t have a formal model of the macroeconomy; instead they had a number of informal models—the most important of which was the quantity theory of money that posited a relationship between the quantity of money in a country and the price level: the greater the amount of money, the higher the price level. These informal models were based on loose empirical regularities that they had discovered over time. While Classical economists did not have a science of macroeconomics from which they could develop theorems, they did have a number of precepts about macro policy. These included

• The government should practice sound finance; the government should always balance the budget.

• The government should maintain a strong currency: The government should face restrictions that prevent it from printing too much money. Keeping the central bank independent and having a gold standard were seen as the best ways to do that.

• The government should be prepared for the central bank to be the lender of last resort; if you have a financial crisis, the central bank should make lots of credit available at a high interest rate.

These policy precepts guided Classical economists’ advice about macro policy and were supported by general reasoning, but no formal model.

The Emergence of the Keynesian Model The Great Depression of the 1930s challenged Classical economic precepts and led to the development of the Keynesian model and Keynesian (interpret “activist”) precepts about policy, which replaced the Classical model. This development was called the Keynesian Revolution because it was based on the work of Cambridge University econ-omist John Maynard Keynes. Through the 1950s learning macroeconomics meant learning the Keynesian model. The multiplier model in the last chapter developed from that early model. Keynesian economics, as it initially developed, was based on an idea known as the fallacy of composition —that what is true for the parts is not necessarily true for the ag-gregate. What this meant in terms of economics is that at least some features of the economy as a whole can’t be understood just by understanding the behavior of

Classical economic precepts included sound finance, maintaining a strong currency and the central bank being the lender of last resort.

Classical economic precepts included sound finance, maintaining a strong currency and the central bank being the lender of last resort.



individuals and firms. For example, in the supply/demand model, supply and demand are assumed independent of each other. In the aggregate economy, the assumption could not hold—when aggregate supply changes, aggregate demand will change as well—the two are interdependent. Microeconomics and macroeconomics were seen as distinct, much like the difference between quantum physics of the subatomic level and the standard physics of large objects. That’s why if you were taking economics 20 or more years ago, you could begin studying economics with either microeconomics or macroeconomics. They were different subjects with different models.

A Model without Microfoundations Appealing to the fallacy of composition, the standard macro model began as a top-down model that started with aggregate relationships, not with a deductive model of the aggregate economy built on the traditional building blocks of microeconomic reason-ing. Put another way, it had no formal microfoundations, and macroeconomists didn’t think it needed any—it was macro economic analysis. There is nothing wrong with a macro model without microfoundations. As I will discuss below, the modern scientific complex system approach to macro is trying to de-velop precisely such a model. An economist would use a macro model rather than a micro-founded model because he or she believes there are so many complex interactions in the economy that these interactions cannot be captured in a solvable micro-founded model, and that the aggregate system is going to behave quite differently than could be usefully predicted from a micro analysis of the components. An example from weather might help you understand: Meteorologists use “macro” properties like the jet stream (fast-flowing, relatively narrow streams of air about 6–10 miles from the ground that typically flow east across the continental United States) and hurricane patterns to pre-dict the weather. Trying to understand weather patterns by studying the micro-level motion of individual air molecules is simply not fruitful. The early macroeconomists’ vision of the aggregate economy was similar: macroeco-nomics, like the weather, is too complex to study from the bottom up. The subsequent history of macro can be viewed as a gradual but near-complete erosion of that vision.

The Movement Away from True Top-Down Macro Models Scientists don’t like to admit that a problem is too complex for them to analyze, so even though using macro models without microfoundations isn’t wrong in principle, macroeconomic theorists didn’t like it, and they worked hard to combine the micro models and the macro models. And as they did, the vision that the macroeconomy was too complex to have microfoundations began to fade. Starting in the 1960s, macroeconomists expanded many of the Keynesian models into more complicated models with selective microfoun-dations. They focused less on the fallacy of composition, and the strict separation of micro and macro disappeared. Macroeconomists continued to use the same basic Keynesian models but began to believe that the models were based implicitly on micro-foundations. The microfoundations were not fully understood, but macroeconomists believed they eventually would be. Microeconomics made its way into macro models when macroeconomists of this period built various sectors—a monetary sector, a labor market sector, a consump-tion goods sector, and a household sector, each of which could be described by an equation—into a composite model of the macroeconomy. 2 Many of these macro

The standard macro model began as a top-down model that emphasized the fallacy of composition.

The standard macro model began as a top-down model that emphasized the fallacy of composition.

Q-3 Are microfoundations needed for a macro model? Q-3 Are microfoundations needed for a macro model?

2It was at that time that it became standard to start the principles of macroeconomics course with a discussion of supply and demand and microeconomics, so that students had the tools to discuss the subsectors of the macro model.



models had over 300 equations. Engineering and scientific models at this time fol-lowed the same reasoning. The higher-level macro models were extensions of the sim-ple multiplier model with more sectors, and more sophisticated analysis of what went on in those sectors.

Problems Begin to Surface These multisector macro models provided the basis for policy analysis at the time—they were engineering models. But they were increasingly based on, and fine-tuned with, formal statistical tools that integrated real data into their equations, and by the mid 1960s these macro models were being presented not as engineering models, but as fully scientific ones. Unfortunately, the real-world data didn’t fit the models as well as macroeconomists would have liked, and some of the fundamental behavioral relationships (microfounda-tions) that their Keynesian macro models assumed began to be called into question. For example, Keynesian models had been based on a particular relationship between current consumption and current income (the consumption function). When economists started studying and empirically measuring that relationship, however, they found that it was not anywhere near as close a relationship as they had assumed. They discovered, for ex-ample, that expectations, wealth, and expected future income also played important roles in determining consumption.

Consumption Depends on Lifetime Income To correct the problem, economists set out to develop models that would reflect rational behavior in the traditional micro-economic sense discussed in Chapter 6. These models made current consumption based on lifetime, not just on current, income. In these models, changes in current income did not change consumption much. For example, say you earn an extra 10 percent in-come this year. The standard macro model assumes you will increase consumption by some percentage of that increase in income—say 60 percent. But, in these new models, unless the increase in income is expected to continue, a 10 percent increase in income represents less than half a percent of one’s lifetime income (depending on the person’s age). So if people base their consumption on lifetime income, as “rational” people would do, they would not increase their consumption much at all in response to an in-crease in current income. Thus, if people are rational, the underlying reasoning with the standard macro models was faulty.

Questioning Theoretical Foundations As statistical techniques and computer power to use them developed, economists discovered that most of the close empirical relationships that earlier economists thought they had found were questionable. What they thought were empirically grounded relationships were, in fact, just common time trends between two variables. That is, over time, both increased and thus looked as if they were closely related, but in fact, when researchers analyzed the data with high-powered statistical techniques, they discovered that the variables were nowhere near as closely related as they first seemed. The mounting empirical evidence that their models were imperfect forced econo-mists to look more carefully at the theoretical foundations of their models. When they did so, they found that when they assumed individuals are rational in the traditional microeconomic sense and forward looking, the very logic of their macro models was flawed. They began to lose faith that these models could actually be micro founded, as they had tacitly assumed. As was the case with consumption, the relationships in the models assumed that people based their actions on current variables. Economists began to argue that rational individuals would base their actions on both current variables and expected variables.

Q-4 True or false? The standard macro models were abandoned despite the fact that they fit the empirical facts well.

Q-4 True or false? The standard macro models were abandoned despite the fact that they fit the empirical facts well.

coL75888_ch29_678-701.indd 83 5/27/09 11:31:49 AM user-f501coL75888_ch29_678-701.indd Page 683 5/27/09 11:31:49 AM user-f501 /Users/user-f501/Desktop/Users/user-f501/Desktop


This has important policy implications. Let’s say, for instance, that the govern-ment runs expansionary fiscal and monetary policy when the economy experiences a recession. If people are rational, they will come to expect that government will run expansionary monetary and fiscal policy in the future, and will build that expectation into their behavior. So whereas with the standard macro model a decline in aggregate demand leads to a falling price level as people cut their prices when they cannot sell their goods, if people expect government to boost aggregate demand, they no longer cut their prices—government will create sufficient demand through expansionary policy. The expectation of government policy will build an inflationary bias in the structure of the economy, which is what happened during this time. In fact, in the 1970s, the term stagflation —the combination of high unemployment and high inflation—developed. The recession raised unemployment while government policies and the expectation of government intervention kept inflation high. The standard macro model could not explain stagflation. If macroeconomists had been willing to accept that their models were ad hoc engi-neering models with no claim to being scientifically grounded, the standard macro models might have remained more popular. But macroeconomists of the time were un-willing to admit that, and the standard macro models lost favor among economic scien-tists. At that point, macroeconomists started working on a new model that avoided the problems of the standard macro model.

Microfoundations and New Classical Macroeconomics The first big step in the shift away from the standard macroeconomics occurred in the late 1970s and early 1980s, when macroeconomists, led by Nobel Prize–winner Robert Lucas, showed formally that you couldn’t derive the standard macro models from tradi-tional micro building blocks, and therefore the macro models were not based on deduc-tive reasoning (and weren’t even internally logical). This led to the development of New Classical macroeconomics — an approach to macroeconomics that studies macroeco-nomic questions using traditional microeconomic building blocks that emphasize rationality. New Classical macroeconomics changed the frame within which macro models were considered. The fallacy of composition was completely discarded as New Classical economists concentrated on getting the logic of the model right. The DSGE model, which I describe below, is an example of a New Classical macro model. In many ways, New Classical macroeconomics was not macroeconomics at all—it was the microeconomics of an entire economy. By that I mean that New Classical mac-roeconomics is a complete departure from standard macro. Instead of making the flows of aggregates the centerpiece of their model, New Classical macroeconomists made the decision process of an individual, using traditional assumptions of rationality and self-interest, the centerpiece of the model. For a New Classical economist, macro was simply micro writ large. In developing their macro model, the problem that the New Classical macroecono-mists had to face was the same problem that the earlier macroeconomists had to face—the complexity of the aggregate economy. New Classicals, however, approached the problem in a fundamentally different way than did earlier macroeconomists. Instead of thinking of the macroeconomy as involving too many interactions to formally model from the bot-tom up, they simply abstracted from those interactions by focusing on the decisions that a single individual —known as a representative agent —would make over multiple time periods. They set their goal as developing a model of the aggregate economy built on the solid traditional microfoundations of rationality, foresight, and self-interest. It was this New Classical version of “macro as micro writ large” that led to the modern DSGE (dynamic stochastic general equilibrium) model.

New Classical macroeconomics is an approach to macroeconomics that studies macroeconomic questions using traditional microeconomic building blocks that emphasize rationality .

New Classical macroeconomics is an approach to macroeconomics that studies macroeconomic questions using traditional microeconomic building blocks that emphasize rationality .

Q-5 How does New Classical economics differ from Keynesian economics?

Q-5 How does New Classical economics differ from Keynesian economics?

www Web Note 29.1 Recent Macro

Nobel Prizes



A Beginner’s Guide to the DSGE Model Probably the best way to introduce you to the DSGE model is with Nobel Prize–winner Robert Solow’s description:

[The DSGE model is] a model in which a single immortal consumer–worker–owner maximizes a perfectly conventional time-additive utility function over an infinite horizon, under perfect foresight or rational expectations, and in an institu-tional and technological environment that favors universal price-taking behavior.

That’s a verbal translation of the DSGE model down from the highly mathematical way that it is normally presented to the level of what might be called normal “professorese.” I agree; professorese still leaves it pretty incomprehensible to even the most conscien-tious beginning student. So I’ll translate it into something a bit more understandable.

A Single Immortal Consumer–Worker–Owner The DSGE model reduces the macro problem down to the problem facing a single indi-vidual whom we assume lives forever. (We are actually not worried about him dying, but if we don’t assume that he lives forever, the mathematics gets considerably more complicated.) Our representative individual is interested in consuming as much as he can and working as little as he can and is pondering how best to manage the trade-off between work and consumption over his lifetime. Thus, for example, he is asking himself questions such as: How much do I want to consume now as opposed to later? When do I want to retire? Should I borrow money now so I can consume more now, taking into account that I will have to pay it back later? The fact that he is making decisions that will guide him through his entire life-time rather than just decisions for a single point in time is where the d ynamic (in the D SGE) comes from; dynamic means taking “time” explicitly into account . Because the representative individual is looking forward and making decisions over time in the model, it is a dynamic model.

A Perfectly Conventional Time-Additive Utility Function To talk about our representative individual’s decision, we have to specify what he likes and what he doesn’t like, and how he makes his decisions. Solow captures the assumption about the representative individual’s likes and dislikes in the statement that the individ-ual maximizes “a time-additive utility function.” What that means is that our representa-tive individual is a fairly conventional person who sees trade-offs in life—he is willing to give up some pleasure now to get more later, and vice versa. Moreover, the specifications of the pleasure or utility he gets from those activities are not interconnected in compli-cated ways; for example, his decisions about what to do today don’t affect the choices he’ll want to make tomorrow, except possibly through their effects on his wealth.

Perfect Foresight or Rational Expectations The DSGE model also assumes that our representative individual is a very, very bright person, with a lightening-fast calculating mind. By that I mean that he is able to make all these decisions perfectly and instantaneously—his mind doesn’t get boggled as mine often does when faced with complicated decisions. Moreover, he is taking into account all possibilities that might happen. That’s where the perfect foresight and rational expectations — expectations that turn out to be correct in reference to the model —come from. He gets his decisions right, before he does anything. He makes no mistakes. This is what Solow means when he says the person has perfect foresight.

Rational expectations are expectations that turn out to be correct in reference to the model.

Rational expectations are expectations that turn out to be correct in reference to the model.

www Web Note 29.2 DSGE Applications



Actually, the concept of perfect foresight needs a bit of clarification: he doesn’t know what will actually happen; instead he knows all the things that might happen and the probability with which each of them will happen. That’s where the s tochastic in the D S GE comes in. Stochastic means that events happen with a certain probability that can be specified mathematically. So our representative individual doesn’t know what will happen, but he does know the probability that the various events will happen. (Think of throwing a die—you know that you will get a 1 through a 6 with a 1/6th chance of each.) Our representative individual figures out what to do for each of these possible outcomes (called contingent states), and for all the contingent states that are contin-gent on each of these. In fact, he has calculated his best strategy for every possible po-tential outcome that might occur. Doing so makes the stochastic model the analytic equivalent to a perfect foresight model (although it has a lot more variables), which is why Solow calls it a perfect foresight model.

Universal Price-Taking Behavior To make his decisions, our representative individual has to know the structure of the markets. Will some firm create a monopoly and push up prices? How will the markets interact? To answer that question, we have to make assumptions about markets and here we see from Solow’s description that there is almost “universal price-taking behav-ior.” That means that almost all markets are assumed to be highly competitive, so that the forces of supply and demand are the primary determinants of price. All these mar-kets are assumed to be in equilibrium. So the model is a general equilibrium model , which means that it is a model of all the markets in the economy, not just a single market. A general equilibrium model is one in which you include interrelationships among mar-kets and do not hold “all other things constant” as you do in partial equilibrium supply/demand models. (That’s where the g eneral e quilibrium in the DS GE comes in.) Now these assumptions may seem highly restrictive, and they are. Macro scientists make them not because they think they perfectly, or even closely, describe reality, but because they are necessary if economists are going to be able to solve this macro model. Modern macroeconomists are working hard on extending the model—adding more than one representative individual, adding less than perfectly competitive markets, adding the possibility of more than one equilibrium, and many more variations. In the United States, most graduate macroeconomic dissertations are working on extending the DSGE model.

Policy Implications of the DSGE Model One reason to study the DSGE model is that it provides insight into policy and the way people will likely react to policy. Let’s consider three policy implications that modern macroeconomists have drawn from the DSGE model.

The Ricardian Equivalence Problem The first policy implication is the Ricardian equivalence problem — the problem that anything the government does to affect the economy will mostly be offset by countervailing ac-tions by private individuals as they optimize over the future. Say the government runs a deficit in order to expand the economy. Our representative individual will reason that the government will have to borrow to pay for the deficit, which will mean that he will have to pay higher taxes in the future in order to pay the loan back. Consider what that means for his decision. Since he has to pay higher taxes in the future, he will want to save more now, which will mean that he will consume less now. This means that the

Q-6 What does DSGE stand for? Q-6 What does DSGE stand for?

Three policy implications of the DSGE model are (1) the Ricardian equivalences problem, (2) the time inconsistency/credibility problem, and (3) the Lucas critique problem.

Three policy implications of the DSGE model are (1) the Ricardian equivalences problem, (2) the time inconsistency/credibility problem, and (3) the Lucas critique problem.



expansionary effect of the government deficit will be offset by the contractionary effect of the representative consumer decreasing consumption. The net effect is a wash and the supposed expansionary effect of fiscal policy evaporates.

The Time-Inconsistency/Credibility Problem The second policy implication of the DSGE model is the time-inconsistency/ credibility problem — the problem that the best government policy from today’s point of view can turn out to be a policy the government wants to change in the future, and that rational individuals can anticipate this. Consider, for example, the best government policy regarding prescription drugs. The best policy for a government that wants to encourage businesses to develop a new anti-cancer drug might be patent protection for new discoveries. But once an important new discovery has been made, the government’s best policy changes: govern-ment will want to allow competition so the drug will be more affordable. That’s the time inconsistency part of the problem. The credibility part of the problem is the observation that drug developers can anticipate that the government will have an incentive to renege on promises of patent protection. If the government cannot credibly commit to protect the patents, even when it doesn’t want to later, the government may find it hard to encourage businesses to develop the drugs at all. The same sorts of problems can arise with discretionary monetary policy to fight recessions. When the economy is falling into recession, standard macro precepts suggest running expansionary monetary policy. But let’s think about the consequences of run-ning expansionary monetary policy if our representative agent expects this to happen. He will build that expectation into his actions when deciding on his best dynamic plan of action. He will reason—whenever the economy goes into a recession, the govern-ment will step in and increase demand, making it unnecessary for me to lower my wage or price. The result will be that wages and prices won’t fall in a recession and there is an inflationary bias built into the economy. Now here’s the crunch. In order to prevent this inflationary bias from generating accelerating inflation, the Fed will have to run contractionary monetary policy. Because of this inflationary bias, the best policy might be for the government not to fight recessions so strongly so that the representative agent will lower his wage or price during recessions, reducing the need for government to implement contractionary mon-etary policy, allowing it to maintain a lower natural rate of unemployment. But even though the best policy rule is to refrain from fighting recessions too strongly, the best policy if and when any particular recession comes is to renege on the rule and fight it just as hard as if there were no inflationary bias problems. In this case, government can lower unemployment without risking higher inflation. In other words, the best thing for a government to do is to successfully convince people that it won’t fight recessions, and then, once they’re convinced, fight recessions anyway. But this means that people are unlikely to be convinced in the first place! One can get around the time-inconsistency problem by establishing credible rules—such as an explicit inflation target, or a “balanced budget rule” that guides policy rather than choosing the best policy on a case-by-case basis. Not surprisingly, the rules have become a central part of the discussions in modern macroeconomic policy.

The Lucas Critique Problem The final problem that the DSGE model highlights is the Lucas critique problem — the problem that because government policies can affect the behavior of individuals, historical data can lead to misleading predictions about the impact of a new policy. Because the standard macro model is based on historical relationships among the data , this undermines the logic of the standard macroeconometric model.

Q-7 What are three policy problems that the DSGE model has highlighted?

Q-7 What are three policy problems that the DSGE model has highlighted?

www Web Note 29.3 Still Problematic



The classic example of this problem has to do with inflation. Most economists now believe that unexpectedly high inflation can cause lower unemployment, but that (mod-estly) high inflation that is anticipated does not affect unemployment. Now imagine you didn’t know this, and you were examining an economy in which the government had long been taking a “hands off ” policy approach. In some years in this economy, just by random chance, inflation had been unexpectedly high, so that unemployment was low. In other years, inflation happened to be unexpectedly low, so unemployment was high. If you were making a model of the relationship between inflation and unemploy-ment by just looking at the historical pattern of data in this economy, your model would show that high inflation goes hand-in-hand with low unemployment. The model would “predict” that the government could lower unemployment by raising inflation. The problem with this model is that it does not take into account the possibility that indi-viduals are rational. If individuals are rational, they will know that government is im-plementing such a policy and will expect higher inflation. The underlying relationship between inflation and unemployment, which could not be discerned from the data alone, requires that inflation be unexpected. Because the model doesn’t take this into account, its prediction would be wrong. Anticipated inflation is unlikely to have much of an effect on unemployment. The basic problem with macro models based on patterns in historical data is that unless the model correctly captures why the patterns exist, changes in policies can re-veal shortcomings in the structural integrity of the model. This makes the models ir-relevant for providing policy advice. Notice that the model above based on the data worked perfectly well until it was used to provide policy advice—but then it stopped working completely. Without the ability to provide policy advice, you might as well throw out all the applied policy macro models.

How Relevant Are the Problems? As you can see, the DSGE model poses some significant challenges for the logic and policy implications of the standard macro model that you learned in the last two chap-ters. Those problems have been the subject of much discussion in economics over the past decade. Don’t worry; I am not going to go through that discussion. Let me just say that advocates of the standard macro model have answers for each of these problems. Essentially, the answer is that the standard macro model is an engineering model that requires care in application, and that in applying the models they take these problems into account. That’s a reasonable argument for an engineering model, but it does not save the standard macro model as a scientific model. The standard macro model is a highly imperfect and problematic model, and significant care must be taken in drawing any policy conclusions from it. When I presented the models, I tried to do that. If you read the chapters carefully, you noticed that all three problems were discussed as potential difficulties with the model in those chapters. Almost all economists agree with these cautions. Some DSGE advocates (macroeconomic theorists), however, go further than that and argue that the standard macro model is just plain wrong. That view is not universal, however, and my friends who are applied policy macroeconomists in business or gov-ernment tell me that standard macro models are much more useful in understanding what’s going on in the macroeconomy than are the modern scientific macro models. As you can see, there are some major differences of opinion among modern macroecono-mists about which model to use and what macro theory is relevant. You can probably tell where I stand from my discussion at the beginning of this chapter. While the DSGE model provides helpful cautions for the standard macro policy,

The standard macro model is a highly imperfect and problematic model, and significant care must be taken in drawing any policy conclusions from it.

The standard macro model is a highly imperfect and problematic model, and significant care must be taken in drawing any policy conclusions from it.



as it currently exists, in my opinion, the DSGE model should not be used for direct policy analysis. But more and more macroeconomists are arguing that the DSGE model is relevant for policy, so it needs some discussion. Central banks are now starting to use DSGE models for direct policy analysis in an attempt to be “modern.” I consider this use of the model a major mistake. To give you a better sense of what the advocates of using the DSGE model for policy argue, let’s consider some arguments of University of Minnesota economists V. V. Chari and Patrick Kehoe, who have been forceful advocates using the DSGE model for policy advice. In an article in the Winter 2006 issue the Journal of Economic Perspectives , they write:

Macroeconomics [by which they mean the DSGE model] is now firmly grounded in the principles of economic theory. These advances have not been restricted to the ivory tower [by which they mean graduate school macroeconomics professors]. Over the last several decades, the United States and other countries have under-taken a variety of policy changes that are precisely what macroeconomic theory of the last 30 years suggests. . . Examples of the effects of theory on the practice of policy include increased central bank independence and adoption of inflation targeting and other rules to guide monetary policy.

and

Macroeconomic theory has had a profound and far reaching effect on the institu-tions and practices governing monetary policy and is beginning to have a similar effect on fiscal policy. The marginal social product of macroeconomic science is surely large and growing rapidly.

They draw the following implications from the DSGE model:

. . . discretionary (macro) policy making has only costs and no benefits, so that if government policymakers can be made to commit to a policy rule, society should make them do so.

Is such a strong position justified and should we accept the “sharp answers” of the DSGE models? My view is definitely not, and the mistake that Chari and Kehoe make is not distinguishing theorems from precepts. One draws theorems—logical deductions that follow from the model, given all the assumptions of the model—from a model. To move from theorems to policy precepts you have to make a decision about how well the assumptions of the model fit reality. For example, what happens if the economy isn’t perfectly competitive (as it surely is not)? What happens if people cannot make infinitely fast calculations or aren’t per-fectly rational? What happens if the system does not have a single equilibrium but many equilibria? What happens if people’s pleasure depends on complicated interrela-tionships over time, so that we can’t use additive utility functions? And many, many more. Once one moves to precepts, most economists believe that the implications about policy that Chari and Kehoe draw from the model are incorrect. The DSGE model is just too far removed from anything resembling the real-world economy for anyone to draw definitive policy conclusions from it. It provides cautions and addenda, not policy conclusions. Consider the bank crisis of 2008. Ben Bernanke, an advocate of inflation targeting and central bank independence—both policies that follow from the DSGE model— directly violated the precept of inflation targeting; he also asked Congress for a stronger connection between the central bank and the government than had existed before. Why? Because he felt that the crisis warranted it, and the precepts that followed from

The DSGE model is just too far removed from anything resembling the real-world economy for anyone to draw definitive policy conclusions from it.

The DSGE model is just too far removed from anything resembling the real-world economy for anyone to draw definitive policy conclusions from it.



the DSGE model did not fit the circumstances. He was most worried about preventing a major depression. Most applied policy macroeconomists agreed with him. All macroeconomists agree that views on macro policy have changed over the past decades, but most applied macroeconomists argue that events shaped that change more than developments in macro theory, and that the arguments for and against central bank independence, inflation targeting, and rules over discretion were well known long before the modern DSGE model developed. The primary contribution of modern macro theorists has been to translate well-known policy insights into formal mathematics, and remind economists of their importance.

Modern Macroeconomic Policy and the Collapse of the Tacoma Narrows Bridge Let me give you an example of when too heavy reliance on a formal scientific model can lead engineers astray. In the late 1930s, using the best science of the time, engineers built a bridge over the Tacoma Narrows in Washington. The formal scientific models of the bridge concluded that the bridge was solid and safe. Unfortunately, the modelers were not able to include something that we now call “aerodynamic flutter” that can cause tortional disturbance. (Don’t ask.) This aerodynamic flutter meant that even

mild winds caused the bridge to sway. When the bridge was built, people noticed that it swayed a bit and initially the bridge architects said the swaying presented no problem—their models said the bridge was solid and safe. On November 7, 1940, they were proved wrong when the swaying increased, causing the bridge to sway like a blanket being lifted up and down on both sides. (You can see the col-lapse of the bridge on a variety of sites on the Web; it is worth seeing them, and thinking of the economy as you watch the bridge sway.) This footage is still shown to engineering, ar-chitecture, and physics students to teach them about the po-tential of interactive positive feedback effects to undermine a model’s results that do not include them. What’s the collapse of a bridge got to do with modern macroeconomics? Something similar to aerodynamic flutter is one of many second- or third-order dynamic effects that can occur when individuals interact in the economy. It is the type of effect that the DSGE model cannot yet take into ac-count. What applied policy macroeconomists worry about is

that using a scientific model that does not take such effects into account could lead to a collapse of the economy similar to the collapse of the bridge. It happened before—in the 1930s. The concerns about such a catastrophic collapse of the macroeconomy were espe-cially strong in 2008, when the U.S. Federal Reserve Bank and the U.S. government intervened in the macroeconomy in ways that previously would have seemed impossible. (We will discuss this intervention in later chapters.) One of the reasons they intervened so quickly was that the chairman of the Federal Reserve was Ben Bernanke, an economic historian who had studied the Great Depression. He could see parallels between the U.S. economy in 2008 and the economy in the 1930s during the Great Depression. When he saw what was happening to financial markets in 2008, he got scared, not because of the scientific model, but because of his knowledge of economic history. It looked and felt like the economy was really swaying.

The primary contribution of modern macro theorists has been to translate well-known policy insights into formal mathematics, and remind economists of their importance.

The primary contribution of modern macro theorists has been to translate well-known policy insights into formal mathematics, and remind economists of their importance.



It was the Great Depression that led to the development of the standard macro model, a model that incorporates positive feedback effects. But as we discussed above, modern macroeconomists have shown that the standard model doesn’t follow from tra-ditional microeconomic assumptions. The modern DSGE macro model is more useful as the beginning of a scientific model, but it remains far too simple to use as a guide to policy. Its flaw is that in order to be a solvable analytic model, it has to assume away interactive effects of one individual’s decision on others’ decisions. That essentially rules out any of the multiplier effects that were central to the standard macro model by assumption, not by analysis. Applied policy macroeconomists believe that in many instances, these interactive effects are central to the real-world problems that macroeconomies face, and can cause the type of dynamic feedback problems that brought down the Tacoma Narrows Bridge. This is why the fallacy of composition is central to the standard model. So the engi-neering approach justifies its difference from the formal scientific model by the pre-sumed existence and importance of complex interaction effects that the scientific model does not yet incorporate.

The Complexity Approach to Macro: The Future of Modern Macro By now you are likely asking—if these interactive effects are so important, why don’t modern scientific macroeconomic models deal with them? The first answer to that question is that doing so is really, really, hard, especially in the formal DSGE model. The second answer is that other macroeconomists are, in fact, working on developing new types of models that can deal with them. These models are called complex systems macro models — macro models of the economy that take into account dynamic interactions of agents in the models, where agents have less than full information and can be less than infi-nitely rational. Complex systems can only be understood by considering the interaction of the parts simultaneously with the analysis of the parts themselves. Economists have found that, because of these interactions, these complex system macro models can have emergent properties — properties of the system that could not have been predicted from a deductive analysis starting from the components of the system, thus justifying the basic macro approach of the standard macro models.

The Underlying Dynamic Assumptions of the Standard, DSGE, and Complex Systems Models To give you a sense of how the complex systems model differs from the standard model and the DSGE model, let’s consider a model that highlights the differences in assump-tions about expectations in the various models. It is a variation on what is called the cobweb model, which played an important role in developing modern macro models. This very simple model is a standard supply/demand model with one difference—suppliers don’t know what price they’ll be able to sell their product for when they decide how much to supply. The model starts at equilibrium and then demand shifts up. The implied assumptions of the standard model, the DSGE model, and the complex systems model can then be used to make predictions about what will happen.

The Standard Macro Model: Backward-Looking Expectations Let’s begin with the assumption of the standard macro model. In that model, individu-als and firms predict the future using historically based expectations — expectations

Q-8 What does the Tacoma Narrows Bridge have to do with the fallacy of composition problem?

Q-8 What does the Tacoma Narrows Bridge have to do with the fallacy of composition problem?



about the future that are based on past events. Let’s keep the simplest type of historically based expectations and say suppliers expect the price today to be equal to whatever it was yesterday—that is, suppliers expect the most recent price to continue. 3 Let’s first see how this assumption works out if everything starts out in equilibrium and neither supply nor demand shifts. The situation is shown in Figure 29-1 (a): suppliers expect price P e and produce quantity Q e . Because the quantity demanded at P e is also Q e , the market clears—the suppliers’ price expectations are exactly correct. Now suppose the market doesn’t begin in equilibrium, as shown in Figure 29-1 (b). Say that suppliers expected price P 0 , which led them to supply quantity Q 1 . The market price at which buyers are willing to buy Q 1 units is P 1 . So instead of getting the price P 0 they expected, suppliers end up getting the higher price P 1 . Suppliers like getting a higher price, but if they had correctly anticipated the higher price, they would have wanted to produce more than Q 1 : if they had known the price was going to be P 1 , they would have supplied Q 2 instead of Q 1 . Suppliers now face the problem of deciding what to produce for time period 2. Since suppliers are backward-looking (they assume that the price in the coming period is the price in the market during the last period), suppliers expect price P 1 to continue and now decide to supply Q 2 . But as you can see in Figure 29-1 (b), when Q 2 is supplied, the price falls to P 2 , much lower than expected. Again, suppliers are unhappy—this time, they wish they had produced less. Because price is so low and suppliers assume that the low price P 2 will prevail, many suppliers decide not to produce in the next period. Specifically, suppliers reduce their supply to Q 3 . As you can see in Figure 29-1 (b), in doing so they actually make the

3We can change the assumption to be a weighted combination of past prices, but it is far easier to assume that they base it only on the last price—the same issues carry through to more complicated variations of historically based expectations.

Supply

Pric

e

Demand

Quantity Quantity

Pe

Pric

e

P4

P2

P0

P1

P3

Qe Q5 Q3Q1 Q2 Q4

Demand

Supply

(a) (b)

FIGURE 29-1 (A AND B) Backward-Looking Expectations

With historically based expectations, a small gap in expected and actual prices can become larger and larger. This model of continually getting it wrong isn’t particularly believable.



situation worse. The price suppliers get rises to P 3 , which is even higher than P 1 , the price at which the market began. Suppliers wish they had produced more than Q 3 . The price dynamics of the market are shown in Figure 29-2 . Given the assumptions about backward-looking expectations, this process of con-tinually getting it wrong will persist and, given the nature of the supply and demand curves drawn, will lead to ever-greater fluctuations in price and quantity. Other possible outcomes, with different supply and demand curves, include fluctuations that dampen and eventually converge and persistent fluctuations that oscillate between the same values. Of course, this model isn’t terribly believable: suppliers never seem to learn from their mistakes, and never seem to notice that “too high” prices in this model are followed by “too low” prices. In other words, in this simplistic model, suppliers are just plain dumb. The standard macro models rely implicitly on similar sorts of assumptions involving “dumb” expectations. For example, the expansionary effect of tax cuts in the multiplier model is based on the assumption that people don’t pause to think that lower taxes today will typically require higher taxes tomorrow. The DSGE model was developed to help free macroeconomic models from making the “people are dumb and don’t learn” assumptions that the standard macro models implicitly embodied.

The DSGE Model: Rational Expectations New Classical economists rightly strongly objected to these sorts of “dumb expectations” assumptions of the standard macro model—people would learn from their mistakes. The question was how to model this “learning” process. A group of economists got together to figure out how, including two future Nobel Prize winners, Franco Modigliani and Herbert Simon, and a third economist, John Muth. Muth proposed a very simple alter-native assumption— people are smart and would expect a price predicted by the model. After all, all they had to do was to talk to an economist, and he or she would tell them what the equilibrium was. Muth called these expectations rational expectations . Say the market had been in equilibrium with a price P 5 39 for a long time when, all of a sudden, demand shifted out, raising the new equilibrium price to 40. Figure 29-3 shows the price dynamics for rational expectations. When the demand shifted out (on day 50), the price instantaneously shifted to the new equilibrium. That’s a really nice property if you are trying to analytically solve and model, and because it led to solvable models, rational expectations became the center stone for New Classical models, including the DSGE model. So in the DSGE model, all markets are assumed to immediately move to equilibrium—the adjustment to equilib-rium is instantaneous.

Q-9 Why does the assumption of rationality cause problems for the standard cobweb model?

Q-9 Why does the assumption of rationality cause problems for the standard cobweb model?

Day1 2 3 4 50

Pric

e

P5

P3P1P0P2P4

FIGURE 29-2 Price Dynamics with Backward-Looking Expectations

With backward-looking expectations, fluctuations in price can explode.



Complex Systems Models: Smart People with Less-Than-Perfect Foresight Herbert Simon, who would go on to pioneer work in artificial intelligence, was un-happy with Muth’s “rational expectations” solution—he said it made things too easy—and, more importantly, did not describe reality. He argued the economists didn’t know precisely what the underlying model was, and thus we needed a model of how reason-ably smart people interact when they don’t know the precise underlying model. His work became the foundation for complex systems models. In these models, people are smart but do not have perfect foresight or infinite amounts of computing power, thus making the macroeconomy too complex to model. In a complex system, people don’t know what the correct model is because the scientists don’t know what the correct model is. They are both continually learning. In a complex system, there is no “correct” model on which to base price expectation, so everyone has to make his or her best guess about what the price will be. The analysis focuses on how people make reasonable guesses. Modern complex systems macroecono-mists have been struggling to create a reasonable guess model, and have found some interesting results, but everything is still very tentative. Specifically, economists have found that learning procedures — the methods by which people learn about the system— are central to the expectations process, and that people will likely use historical data to learn from in highly sophisticated ways. They have found that people learn by discovering patterns in the past data and tentatively check to see if those patterns are likely to continue. To the degree that they believe that the pattern will persist, they will start basing their actions on the expected pattern. But there is a catch—in doing so they will often change the pattern. This will lead them to start looking for patterns of patterns, which means they will be learning how patterns change over time. If they find such relationships, they will start basing their actions on patterns of changing patterns, but when they do, they will change the patterns of the change in patterns, which leads them to higher levels of pattern-matching analysis. Thus, the fluctuations in the aggregate economy are driven by people who are trying to find pat-terns of patterns of patterns . . . , and they get so caught up in those patterns that they are no longer concerned with the underlying realities of the economy. For most historians of economic thought, that idea was precisely the idea that Keynes was trying to convey in his work that marked the beginning of macroeconomics. For

In a complex system, there is no “correct” model on which to base price expectation.

In a complex system, there is no “correct” model on which to base price expectation.

Q-10 What are complexity economist arguments against the rational expectations model?

Q-10 What are complexity economist arguments against the rational expectations model?

0

41

40

39

3850 100

Day

Pric

e

150 200

FIGURE 29-3 Price Dynamics with Rational Expectations

With rational expectations— people expect the price predicted by the model—price adjusts to its new equilibrium immediately.



example, he made an analogy to a beauty contest with the way the market works. He wrote: “It is not a case of choosing those [faces] which, to the best of one’s judgment, are really the prettiest, nor even those which average opinion genuinely thinks the prettiest. We have reached the third degree where we devote our intelligences to an-ticipating what average opinion expects the average opinion to be. And there are some, I believe, who practice the fourth, fifth and higher degrees.” He argued that as people’s expectations become so interdependent, no one focuses on the fundamentals of the economy, and the result can be a serious depression such as happened in the Great Depression. The problem lies in the dynamics of the system. Many analysts believe that is precisely what happened in the recent credit crisis and housing bubble—where housing prices kept rising because more and more people started basing their expectations of future housing prices on patterns of ever-rising prices. Let’s consider the implications of one such learning model that combines the as-sumptions of the standard model and the DSGE model. 4 In this model, suppliers have to pay a small cost to make better predictions about the future price. To make things simple, we’ll assume that if they pay the small cost, they’ll be able to perfectly predict the future price—the assumption in the DSGE model. Alternatively, they can forgo paying that small cost and rely on their “naïve” predictions that last period’s price will be the price in the next period—the assumption in the standard model. Figure 29-4 plots the price dynamics that can emerge from this model in response to the same demand shock I showed for the standard and DSGE models above. As you can see, prices behave very oddly (technically speaking, prices behave chaotically ) after the demand shock on day 50. In the early days after the demand shock, the prices bounce around a bit, but then they seem to settle down close to the new equilibrium price of 40. Prices stay close to equilibrium for a long time, until suddenly, without any external shock, prices begin to fluctuate wildly again. Again, prices settle down only to fluctuate wildly again. This chaotic pattern continues. What we see, in short, is extended periods where the market price is the price predicted by the DSGE model, but with cyclical—but aperiodic and unpredictable—periods of significant instability.

4This model was developed by William Brock and Cars Hommes in 1997. It appeared in the journal Econometrica.

036

37

38

39

Pric

e

40

41

42

43

44

50 100Day

150 200

FIGURE 29-4 Price Dynamics in a Learning Model

In a learning model, prices are relatively stable but experience periodic fluctuations.



What’s going on here? The answer is that when prices remain close to the equilibrium price for a long time, suppliers think they have found a new pattern—prices stay roughly constant over time—and they figure that they won’t waste the money to make better predictions telling them what they already know. They choose to “free ride” rather than pay the cost of predictions that are only marginally better than their naïve predictions. But when more and more people start relying on the pattern, the equilibrium becomes unstable and prices tend to spiral away from it—just like in the standard cobweb model. When prices get sufficiently far from equilibrium, suppliers suffer sufficiently from their bad predictions that it becomes worthwhile to pay the cost of becoming better informed. These better expectations restabilize the market, just like in the DSGE model. What is interesting about this model is that the periods of instability occur at seemingly random times (and, in fact, the times are actually random in the formal sense). Complexity economists argue that this type of price movement is typical in competi-tive markets in a complex economy where people independently form their expectations. Such random price movements make macro policy much more an art than a science.

Macroeconomics, Learning, and Behavioral Economics The pressures for instability in markets become even greater when behavioral economic insights are added to the model. Modern complex systems macroeconomists rely heavily on insights from behavioral economics to provide the key to how people learn. In these models, they have found that people tend to herd — to copy other successful behavior even though that successful behavior may have been just luck . Herding can result in even larger fluctua-tions away from equilibrium, keeping the market from equilibrium for long periods of time. People’s tendency to herd is well documented in numerous experiments. For example, in one experiment, economist Matthew Salganik and his coauthors created an artificial music market from individuals who visited a popular Web site. They were given a list of songs from unknown bands and asked to download them and assign a rating to the songs. Half the participants were not given any information about what other people had chosen. The other half were shown how many times each song had been downloaded. If people’s choices were not influenced by others, then the two groups should have made approximately the same selection. If not, then people were basing their selection on what others had chosen. They found that people downloaded what others had downloaded, making the model a path-dependent tipping-point model I introduced in Chapter 6 and consistent with the standard macro model, rather than the DSGE macro model. This means that for music choices, one cannot use a unique equilibrium rational choice model, and that the choice of a group has to be analyzed in the context of the group. The same has been found to be true in many other cases—individuals in groups decide differently than do individuals outside groups. Translating that behavioral in-sight into macroeconomics provides a likely foundation for the standard macro models with their emphasis on the fallacy of composition.

Agent-Based Computational Models of the Macroeconomy The problem of these multiple-level pattern-finding learning models that are key to complex systems models is that they are almost always too complex to solve analytically. They are far more technically complicated than the dominant modern macro DSGE models, which are already highly technical. If we had to rely on analytic solutions to these models, it would be a long time before we get any reasonable results from them.

Translating behavioral insights into macroeconomics provides a likely foundation for the standard macro models.

Translating behavioral insights into macroeconomics provides a likely foundation for the standard macro models.

www Web Note 29.4 Herding Tendencies



However, because of increasing computational power, modern economists such as Iowa State economist Leigh Tesfatsion, George Mason economist Rob Axtell, and Brandeis economist Blake LeBaron believe that they may be able to gain insight into these complex systems models by creating computer models of miniature macroecon-omies. They hope to be able to study real-world macroeconomies through their study of agent-based computational (ACE) models — virtual computer macroeconomies of rational agents with less-than-perfect information. Having created such virtual macro-economies, they experiment with different learning patterns of the agents in the model. The hope is that useful insights can be drawn from these computer models without having to fully understand the underlying technical complications. (You can see a discussion of ACE modeling along with many examples at: www.econ.iastate.edu/ tesfatsi/ace.htm.) In these models, enormous price fluctuations occur as markets dynamically adjust. These fluctuations are not caused by people being dumb or not using all available infor-mation; people are assumed to be fully rational. Instead, the fluctuations are caused by people trying to learn in an environment that is continually changing. Everyone is try-ing to figure out what everyone else is doing when the process of figuring out changes the way the system works. One learning procedure that people use is to ask a trust-worthy friend what he or she thinks will happen and to copy the strategy that others use, especially others that they consider especially bright. They also try to get whatever information they can about future demand and supply, but that is likely limited and often contradictory. Ultimately, they make some prediction, but they will not be too confident in that prediction. And there will be lots of different predictions out there. The people in these complex systems will rely on estimates of various groups, their knowledge of what has happened in the past, and their gut instincts—what Keynes called animal spirits. The resulting system can never be perfectly specified.

The Limits of ACE Models These complex systems models are really neat and likely are the future of modern macro-economic theory. But were it not for some modern macroeconomists’ attempts to apply the simpler DSGE macro model to reality (and to some policy makers actually listening to them), I wouldn’t be presenting these models at all because the standard macro models that you have learned captured many of the insights found in complex systems models. As impressive as these scientific complex systems macro models are, they are not especially useful to policy makers as direct guides to policy. They are primarily useful in letting policy makers know how little we actually know about the macroeconomy. Just about their only concrete result is to show that models that don’t take these interactive effects into account may well miss important elements of the macroeconomy, and that when these elements are added, just about anything can happen. Applied macroecono-mists already knew that.

Choosing the Right Model The important question for policy makers is which of these many models of the macro-economy to use. Will people on average get it right, as the DSGE model assumes? In other words, will the mistakes that people make cancel out? Those who make poor estimates will go out of business; those who make good estimates will win, and eventu-ally only those who make good estimates will remain. If, on average, suppliers get it right, then the DSGE model may be a reasonable one to use. Or will people get it wrong— creating bubbles and large deviations of the actual price from the fundamental price? There is no simple answer to this question.

In complex systems models fluctuations are caused by people trying to learn in an environment that is continually changing.

In complex systems models fluctuations are caused by people trying to learn in an environment that is continually changing.

The standard macro models capture many of the insights found in complex systems models.

The standard macro models capture many of the insights found in complex systems models.

www Web Note 29.5 Reassessing Macro Models



To figure out which model to use, modern macroeconomists conduct experiments about how people respond in periods of uncertainty and do historical studies of what people have actually done. They have found that suppliers develop highly changeable rules of thumb that often, in normal times, get the prediction right, so that the DSGE rational expectations model is often a reasonable description of the macroeconomy. But they also have found that sometimes suppliers get it wrong—really wrong. In this case, there are serious problems for the market and the appropriate model can involve explod-ing cycles that occur when people are basing their decisions on past prices. In these cases, the standard macro model is the appropriate model. Researchers have found that it is easiest for people to get it right when markets are relatively stable, but are less likely to get it right when markets become slightly unstable. Moderate instability induces more instability. UCLA economist Axel Leijonhufvud, whose work I discussed above in the discussion of whether to use the AS / AD model or the multiplier model, has pushed this argument the furthest. He argues that there tends to be a corridor in which people’s rules of thumb operate well and markets lead to reasonable macro results. When the market is in the corridor, the DSGE model is a reasonable one; the market will coordinate the economy. But when the economy moves outside the corridor—when it faces big shocks, especially ones that have not been experienced before—markets can have big problems and might not coordinate the economy; they might actually make it worse. In 2007, the U.S. economy was getting to the edge of the corridor and in 2008, it got outside it. The result was the swaying of the economy, and the fears that the economy was headed toward a type of Tacoma Narrows Bridge collapse. So the answer that the complex systems approach comes to on whether people’s expectations will stabilize an economy or not is that it depends . Sometimes they will and sometimes they won’t. Unfortunately, that’s not a lot of help to policy makers since they can’t tell them when it will and when it won’t.

Conclusion So there we have it—a whirlwind tour of modern macroeconomic theory. What are you supposed to take away from it? First, that macroeconomics is complex and a hard subject; there are no easy answers. Second, all macro models have problems, and we must put together insights from all of them to come up with useful guides to policy. In terms of policy, the DSGE model has some important insights that you had better add as addenda to your standard macro models that you learned in previous chapters. These include:

• People are smart, and your policy had better take that into account. • Credibility is important. • Rules have certain advantages over discretion.

Complex systems macro also has important insights to keep in mind too. These include: • In complex situations, groups of individually rational smart people can

sometimes act in ways that seem collectively quite irrational. • All models may leave out important elements that can cause a complex system

to collapse, so you should always look at reality as well as the model when conducting policy in a complex system.

• Macroeconomics is complex, and we have no definitive theory to guide us. Thus, policy must be made based on an educated common sense—not on any single theory. There are no unchanging rules of macro policy.

If you remember these, and modify your use of the standard macro models with these insights, you will have all you need to know about the implications of the DSGE and complex systems models for macro policy—at least for my course.

As addenda to the standard model you should remember that (1) people are smart, (2) credibility is important, and (3) rules have advantages.

As addenda to the standard model you should remember that (1) people are smart, (2) credibility is important, and (3) rules have advantages.



Summary

• The standard macro model is a top-down model of the economy based on historical relationships among vari-ables. It is an engineering model suitable for policy.

• The modern macro model is a model built from micro-foundations. It is a scientific model designed to under-stand how the economy works.

• The standard macro model is based on the fallacy of composition and while the interrelationships among variables in the economy are sometimes too difficult to understand, the standard macro model provides general guidance for policy makers that take these interrelationships into account.

• The standard macro model was developed as a result of the experience of the Great Depression. The modern models developed as economists had access to more data and computational power.

• The disadvantage of the standard macro model is that it assumes that people are not rational or forward thinking.

• The advantage of the DSGE model is that it considers that people are not stupid when making their deci-sions. The disadvantage is that it makes highly restric-tive assumptions that do not hold in the real world.

• D stands for dynamic , S for stochastic , G for general , and E for equilibrium .

• Three policy prescriptions of the DSGE model are (1) people are smart, and your policy had better take that into account; (2) credibility is important; and (3) rules have certain advantages over discretion.

• The complex systems approach to macro assumes that people cannot predict the future with perfect accuracy (as does the DSGE model) nor does it assume that people rely only on historical information (as does the standard macro model). It assumes that people learn and that the process is complex.

• The standard model assumes people base their expec-tations on history. The DSGE model assumes people can predict future events. The complex systems mod-els assume that people learn from past mistakes.

• Complex systems models are too complicated to solve mathematically. Agent-based computational models are microeconomies with agents who are given learn-ing mechanisms. The results of the model emerge as the agents interact and make decisions.

agent-based computational models (697)

complex systems macro models (691)

deductive scientific model (680)

dynamic (685) emergent properties (691)

engineering model (679) general equilibrium

model (686) herd (696) historically based

expectations (691) learning

procedures (694)

Lucas critique problem (687)

New Classical macroeconomics (684)

rational expectations (685)

representative agent (684)

Ricardian equivalence problem (686)

standard macro models (679)

stochastic (686) time-inconsistency/

credibility problem (687)

Key Terms

Questions and Exercises

1. Are modern macro models better characterized as engineering models or deductive scientific models? Explain your answer. LO1

2. Are standard macro models better characterized as engi-neering models or deductive scientific models? Explain your answer. LO1

3. When did macroeconomics emerge as a separate discipline? Why did it happen at that time? LO2

4. What did early macroeconomists believe about the economy that justified a separate macroeconomics? LO2



5. How did the fact that the modeling of consumption within the standard model fell short of fitting real-world data lead to changes in the model in the 1960s? LO2

6. What changes did macroeconomists make to the con-sumption portion of the standard model in the 1960s to make it better fit real-world data? LO2

7. What is New Classical macroeconomics? LO2 8. What distinguishes New Classical macroeconomics from

Keynesian economics? LO2 9. Explain the meaning of each of the letters in DSGE. LO4 10. What is the Ricardian equivalence problem? LO5 11. Why might a government want to establish credibility

with regard to fighting inflation? LO5 12. Why might a government break its promise of fighting

inflation? LO5 13. What is the time-inconsistency/credibility problem? LO5

14. What is the Lucas critique problem? LO5 15. Why does the author believe that it is problematic to

draw precepts from the DSGE model? LO5 16. What does the collapse of the Tacoma Narrows Bridge

have to do with modern macroeconomics? LO6 17. Demonstrate how historically based expectations can lead

to increasingly greater fluctuations in prices. LO3 18. How is the complex systems approach to macroeconomics

an improvement to both the standard and neoclassical macro models? LO6

19. Why is learning in the complex systems approach an ongoing process? LO6

20. How does the learning model by Brock and Hommes explain unexpected price fluctuations? LO7

21. How does herding lead to price fluctuations? LO7 22. How can instability create more instability? LO7

Questions from Alternative Perspectives

1. Consider the quotation in the chapter from Chari and Kehoe: “Discretionary [macro] policy has all costs and no benefits, so that if government can be made to commit to a policy rule, it should do so.” Is it true that using expansionary fiscal policies in times of recession has had “all costs and not benefits?” Can you think of anyone who might disagree with this? (Radical)

2. The “general equilibrium” part of the DSGE model seems to suggest that there is a unique equilibrium for a given economy, and indeed, most modern scientific macro models have a single equilibrium. (In fact, the existence of a unique equilibrium is regarded as a desirable feature of such models, since that’s when it gives the sharpest predictions.) Contrast this with the multiplier model, and explain why such a feature is likely to bias the model towards suggesting non-activist macroeconomic policies. (Post-Keynesian)

3. Consider the end of Bob Solow’s quote describing the DSGE model: “. . . in an institutional framework that favors universal price taking behavior.” Do you think this is an accurate description of the economy you live in? In particular: a. Are you a “price taker” in most of your economic

interactions?

b. Can you think of any institutions that are not well described as price takers?

c. Do you think these institutions play an important role in the modern economy? (Institutionalist)

4. This chapter contained a quick-and-dirty history of macroeconomic thought. It started with Keynes, worked through Robert Lucas and the New Classicals, as well as modern DSGE theorists like V. V. Chari and Patrick Kehoe and one of their critics, Robert Solow. What do all of these scholars have in common, and do you think this is just a coincidence? (Feminist)

5. The “complex systems” approach to macroeconomics starts from the idea that the economy involves incredibly complex interactions between many individuals and institutions. One of its important insights is the idea of emergent properties—properties of the system that couldn’t have been predicted from a deductive analysis starting from the components of the system. Does the complexity of the economy and the existence of emergent properties make you more of less confident in the ability of the government to successfully manage the economy? (Austrian)

Issues to Ponder

1. The book contrasts scientific and engineering models. Loosely speaking, a model is “scientific” if it is designed to understand how the economy works in the abstract, while engineering models are those that are designed

to capture how the economy works in practice, even though we don’t really understand why. Which do you think is more important? LO1



Answers to Margin Questions

1. Two standard macroeconomic models are the AS / AD model and the multiplier model. (679)

2. The standard macro model is an engineering model. (680)

3. No, there is nothing wrong with a macro model without microfoundations. Whether microfoundations are wanted, or needed, depends on the nature of the system being modeled and the use to which the model is being put. (682)

4. False. One of the reasons they were abandoned is that they did not fit the empirical data well when modern statistical techniques allowed a careful consideration of those data. (683)

5. New Classical economics uses a representative agent approach and assumes agents are fully rational; it uses a microfoundations approach. Keynesian economics uses a macro approach, does not rely on microfoundations, and does not necessarily assume that we can specify what rationality is. (684)

6. DSGE stands for dynamic stochastic general equilibrium. (686)

7. Three policy problems that the DSGE model has highlighted are the Ricardian equivalence problem, the time-inconsistency problem, and the Lucas critique problem. (687)

8. In designing the Tacoma Narrows Bridge, engineers did not take account of certain small feedback effects, assuming they would cancel each other out. They did not; the fallacy of composition problem is precisely this—in the aggregate, a system might operate quite differently than in a simple model that assumes away small feedback effects. (690)

9. The standard cobweb model assumes people are dumb in a way that does not make sense. Rational people will learn from their past mistakes. (693)

10. The rational expectations model assumes that people know the true model. In fact they don’t and there is no knowable single model. People are continually learning and economists need to use models that incorporate that learning. (694)

2. How is the “fallacy of composition” related to the difference between micro demand and supply curves and macro aggregate demand and supply curves? LO2

3. The three big “problems” identified by modern macroeco-nomics were the Lucas critique problem, the Ricardian

equivalence problem, and the time inconsistency/credibility problem. How important do you believe these are? LO5


Thinking Like a Modern...

Documents

Transcript of Thinking Like a Modern...