萨金特访谈录：金融危机与现代宏观经济学

摘自秋石经济学博客

Interview with Thomas Sargent, by Art Rolnick, Minneapolis Fed, June 15, 2010:

...MODERN MACROECONOMICS UNDER ATTACK

Rolnick: You have devoted your professional life to helping construct and teach modern macroeconomics. After the financial crisis that started in 2007, modern macro has been widely attacked as deficient and wrongheaded.

Sargent: Oh. By whom?

Rolnick: For example, by Paul Krugman in the New York Times and Lord Robert Skidelsky in the Economist and elsewhere. You were a visiting professor at Princeton in the spring of 2009. Along with Alan Blinder, Nobuhiro Kiyotaki and Chris Sims, you must have discussed these criticisms with Krugman at the Princeton macro seminar.

Sargent: Yes, I was at Princeton then and attended the macro seminar every week. Nobu, Chris, Alan and others also attended. There were interesting discussions of many aspects of the financial crisis. But the sense was surely not that modern macro needed to be reconstructed. On the contrary, seminar participants were in the business of using the tools of modern macro, especially rational expectations theorizing, to shed light on the financial crisis.

Rolnick: What was Paul Krugman’s opinion about those Princeton macro seminar presentations that advocated modern macro?

Sargent: He did not attend the macro seminar at Princeton when I was there.

Rolnick: Oh.

Sargent: I know that I’m the one who is supposed to be answering questions, but perhaps you can tell me what popular criticisms of modern macro you have in mind.

Rolnick: OK, here goes. Examples of such criticisms are that modern macroeconomics makes too much use of sophisticated mathematics to model people and markets; that it incorrectly relies on the assumption that asset markets are efficient in the sense that asset prices aggregate information of all individuals; that the faith in good outcomes always emerging from competitive markets is misplaced; that the assumption of “rational expectations” is wrongheaded because it attributes too much knowledge and forecasting ability to people; that the modern macro mainstay “real business cycle model” is deficient because it ignores so many frictions and imperfections and is useless as a guide to policy for dealing with financial crises; that modern macroeconomics has either assumed away or shortchanged the analysis of unemployment; that the recent financial crisis took modern macro by surprise; and that macroeconomics should be based less on formal decision theory and more on the findings of “behavioral economics.” Shouldn’t these be taken seriously?

Sargent: Sorry, Art, but aside from the foolish and intellectually lazy remark about mathematics, all of the criticisms that you have listed reflect either woeful ignorance or intentional disregard for what much of modern macroeconomics is about and what it has accomplished. That said, it is true that modern macroeconomics uses mathematics and statistics to understand behavior in situations where there is uncertainty about how the future will unfold from the past. But a rule of thumb is that the more dynamic, uncertain and ambiguous is the economic environment that you seek to model, the more you are going to have to roll up your sleeves, and learn and use some math. That’s life.

Rolnick: Putting aside fear and ignorance of math, please say more about the other criticisms.

Sargent: Sure. As for the efficient markets hypothesis of the 1960s, please remember the enormous amount of good work that responded to Hansen and Singleton’s ruinous 1983 JPE [Journal of Political Economy] finding that standard rational expectations asset pricing theories fail to fit key features of the U.S. data.1 Far from taking the “efficient markets” outcomes for granted, important parts of modern macro are about understanding a large and interesting suite of asset pricing puzzles, brought to us by Hansen and Singleton and their followers—puzzles about empirical failures of simple versions of efficient markets theories. Here I have in mind papers on the “equity premium puzzle,” the “risk-free rate puzzle,” the “Backus-Smith” puzzle, and on and on.2

These papers have put interesting new forces on the table that can help explain these puzzles, including missing markets, enforcement and information problems that impede trades, difficult estimation and inference problems confronting agents, preference specifications with novel attitudes toward the timing and persistence of risk, and pessimism created by ambiguity and fears of model misspecification.

Rolnick: Tom, let me interrupt. Why should we at central banks care about whether and how those rational expectations asset pricing theories can be repaired to fit the data?

Sargent: Well, there are several important reasons. One is that these theories provide the foundation of our ways of modeling the main channels through which monetary policy’s interest rate decisions affect asset prices and the real economy. To put it technically, the “new Keynesian IS [investment-savings] curve” is an asset pricing equation, one of a form very close to those exposed as empirically deficient by Hansen and Singleton. Efforts to repair the asset pricing theory are part and parcel of the important project of building an econometric model suitable for providing quantitative guidance to monetary and fiscal policymakers.

Another important reason for caring is that monetary policymakers have often been urged to arrest bubbles in asset markets. Easier said than done. Before you can do that, you need a quantitatively reliable theory of asset prices that you can use to identify and measure bubbles.

Rolnick: Before I interrupted, you had begun responding to those criticisms of modern macro. Please continue.

Sargent: I have two responses to your citation of criticisms of “rational expectations.” First, note that rational expectations continues to be a workhorse assumption for policy analysis by macroeconomists of all political persuasions. To take one good example, in the Spring of 2009, Joseph Stiglitz and Jeffrey Sachs independently wrote op-ed pieces incisively criticizing the Obama administration’s proposed PPIP (Public-Private Investment Program) for jump-starting private sector purchases of toxic assets.3 Both Stiglitz and Sachs executed a rational expectations calculation to compute the rewards to prospective buyers. Those calculations vividly showed that the administration’s proposal represented a large transfer of taxpayer funds to owners of toxic assets. That analysis threw a floodlight onto the PPIP that some of its authors did not welcome.

And second, economists have been working hard to refine rational expectations theory. For instance, macroeconomists have done creative work that modifies and extends rational expectations in ways that allow us to understand bubbles and crashes in terms of optimism and pessimism that emerges from small deviations from rational expectations. An influential example of such work is the 1978 QJE [Quarterly Journal of Economics] paper by Harrison and Kreps.4 You should also look at a fascinating paper that builds on Harrison and Kreps, written by José Scheinkman and Wei Xiong in the 2003 JPE.5 As I mentioned earlier, for policymakers to know whether and how they can moderate bubbles, we need to have well-confirmed quantitative versions of such models up and running. We don’t yet, but we are working on it.

Rolnick: And the other criticisms?

Sargent: OK. The criticism of real business cycle models and their close cousins, the so-called New Keynesian models, is misdirected and reflects a misunderstanding of the purpose for which those models were devised.6 These models were designed to describe aggregate economic fluctuations during normal times when markets can bring borrowers and lenders together in orderly ways, not during financial crises and market breakdowns.

By the way, participants within both the real business cycle and new Keynesian traditions have been stern and constructive critics of their own works and have done valuable creative work pushing forward the ability of these models to match important properties of aggregate fluctuations. The authors of papers in this literature usually have made it clear what the models are designed to do and what they are not. Again, they are not designed to be theories of financial crises.

Rolnick: What about the most serious criticism—that the recent financial crisis caught modern macroeconomics by surprise?

Sargent: Art, it is just wrong to say that this financial crisis caught modern macroeconomists by surprise. That statement does a disservice to an important body of research to which responsible economists ought to be directing public attention. Researchers have systematically organized empirical evidence about past financial and exchange crises in the United States and abroad. Enlightened by those data, researchers have constructed first-rate dynamic models of the causes of financial crises and government policies that can arrest them or ignite them. The evidence and some of the models are well summarized and extended, for example, in Franklin Allen and Douglas Gale’s 2007 book Understanding Financial Crises.7 Please note that this work was available well before the U.S. financial crisis that began in 2007.

Rolnick: I’ll come back to that in a second, but you haven’t said anything yet about what is to be gained in terms of understanding financial crises from importing insights of behavioral economics into macroeconomics.

Sargent: No, I haven’t.

Financial crises

Rolnick: OK then. Well, what useful things does macroeconomics have to say about financial crises, what causes them, how to manage them after they start and what can be done to prevent them?

Sargent: A lot. In addition to the formal literature summarized in the Allen and Gale book, I want to mention the example of the 2004 book by Gary Stern and Ron Feldman, Too Big to Fail.8 That book doesn’t have an equation in it, but it wisely uses insights gleaned from the formal literature to frame warnings about the time bomb for a financial crisis set by government regulations and promises. Indeed, one of the focuses of Gary Stern’s long tenure as president of the Minneapolis Fed was steadily to draw attention to financial fragility issues and what the government does either to arrest crises or, unfortunately as an unintended consequence, to incubate them.

Rolnick: Thanks for the nice words about Gary, but please elaborate further on macro scholarship and financial crises.

Sargent: I like to think about two polar models of bank crises and what government lender-of-last-resort and deposit insurance do to arrest them or promote them. Both models had origins in papers written at the Federal Reserve Bank of Minneapolis, one authored by John Kareken and Neil Wallace in 1978 and the other by John Bryant in 1980, then extended by Diamond and Dybvig in 1983.9 I call them polar models because in the Diamond-Dybvig and Bryant model, deposit insurance is purely a good thing, while in the Kareken and Wallace model, it is purely bad. These differences occur because of what the two models include and what they omit.

The Bryant and Diamond-Dybvig model starts with an environment in which banks can do things that are very worthwhile socially; namely, they provide maturity transformation and liquidity transformation activities that improve the efficiency of the economy. They enable coalitions of people, namely, the banks’ depositors, to make long-term investments—loans, mortgages and the like—while at the same time the bank’s depositors hold demand deposits, bank liabilities that are short term in duration, because they can withdraw them at any time. Banks thereby facilitate risk-sharing among people with uncertain future liquidity needs. These are all good things.

But there is a potential problem here because for the long-term investments to come to fruition, enough patient depositors must leave their funds in the bank to avoid premature liquidation of a bank’s long-term investments. Without deposit insurance, situations can arise that induce even patient depositors to want to withdraw their funds early, causing the banks prematurely to liquidate the long-term investments, with adverse affects on the realized returns.

What triggers a bank run is patient depositors’ private incentive to withdraw early when they think that other patient investors are also choosing to withdraw early. Technically speaking, that amounts to multiple Nash equilibria. There are situations in which I run (i.e., withdraw from the bank early) because I expect you to run, and when you also run because you expect me to run. But there are other situations in which we both trust that the other person isn’t going to run and we don’t run. Which equilibrium prevails is anyone’s guess, or something resolved only by an extraneous random device for correlating behavior, a device that economists sometimes call a “sunspot.”

So without deposit insurance, the economy is vulnerable to bank runs. The situations where depositors don’t run lead to good outcomes, but when there are bank runs, outcomes are bad. The good news in the Diamond-Dybvig and Bryant model, however, is that if you put in government-supplied deposit insurance, that knocks out the bad equilibrium. People don’t initiate bank runs because they trust that their deposits are safely insured. And a great thing is that it ends up not costing the government anything to offer the deposit insurance! It’s just good all the way around.

Rolnick: Do you think that an abstract model like this ever influences policymakers?

Sargent: I believe that the Bryant-Diamond-Dybvig model has been very influential generally, and in particular that it was very influential in 2008 among policymakers. A perhaps oversimplified but I think largely accurate way of characterizing the vision of many policy authorities in 2008 was that they correctly noticed that a Bryant-Diamond-Dybvig bank is not just something that has “B A N K” written on its stationary and front door. It’s any institution that executes liquidity transformation and maturity transformation, thereby offering a kind of intertemporal risk-sharing.

So in 2008, there were all sorts of institutions that were really banks in the economic sense of the Bryant-Diamond-Dybvig model but that did not have access to explicit deposit insurance, institutions like money market mutual funds, shadow banks, even hedge funds that were doing exactly those maturity-transforming and risk-transforming activities.

When monetary policy authorities, deposit insurance authorities and others looked out their windows in the fall of 2008, they saw Bryant-Diamond-Dybvig bank runs all over the place. And the logic of the Bryant-Diamond-Dybvig model persuaded them that if they could arrest the runs by effectively convincing creditors that their loans—that is, their short-term deposits—to these “banks” were insured, that could be done at little or no eventual cost to the taxpayers. You could nip the run in the bud and really prevent the next Great Depression. This is a very optimistic view of those 2008 interventions enlightened by the Bryant and Diamond-Dybvig model.

But Diamond and Dybvig themselves were cautious about promoting such optimism. In the last part of their 1983 JPE paper, Diamond and Dybvig recommend that their readers take seriously the message of a 1978 paper (written at the Minneapolis Fed, as I mentioned earlier) by Kareken and Wallace. That paper includes something important that Diamond and Dybvig recognize that they left out: moral hazard.

Rolnick: And the Kareken-Wallace story?

Sargent: The main idea is that when a government is in the business of being a lender of last resort or a deposit insurer, depending on how it regulates banks, it affects the risk that banks take and the probability that the government is actually going to be required to exercise lender-of-last-resort and bail out facilities. Neil and Jack call it the “moral hazard” problem, which is the idea that when you insure a bank, you alter its incentives to undertake risks.

In the Kareken-Wallace model, deposit insurance is purely a bad thing. Kareken-Wallace envisions a different economic setting than Bryant and Diamond-Dybvig. Of course, like all models, it’s an abstraction; it simplifies things in order to isolate key forces. The Kareken-Wallace setting has complete markets. There are markets in all possible risky claims. There are also some people who wanted to hold risk-free deposits.

Kareken and Wallace compare two different situations. In one, there is no deposit insurance; depositors are on their own and know that their deposits are uninsured. If they want to hold risk-free deposits, they’d better hold them in banks that are holding risk-free portfolios. Some very conservative banks emerge that can issue safe deposits because the bank portfolio managers themselves hold assets that allow these banks to pay depositors in all possible states of the world.

Kareken and Wallace compare that no-deposit-insurance situation to another situation in which a government agency provides deposit insurance that is either free or is priced too cheaply, meaning that it’s not priced with a proper risk-loading. Kareken and Wallace show that in that situation, banks have an incentive to become as risky as possible, and as large as possible. Therefore, with a positive probability, banks will fail and taxpayers will have to compensate banks’ depositors. It is in banks’ shareholders’ interest that he banks organize themselves this way. This lets them gamble with the insurers’ and depositors’ money.

The Kareken and Wallace model’s prediction is that if a government sets up deposit insurance and doesn’t regulate bank portfolios to prevent them from taking too much risk, the government is setting the stage for a financial crisis. On the basis of the Kareken-Wallace model, Jack Kareken wrote a paper in the Federal Reserve Bank of Minneapolis Quarterly Review referring to the “cart before the horse.”10 He pointed out that if you’re going to deregulate financial institutions, which we in the United States did in the late ’70s and early ’80s (deregulation is the cart), you’d better reform deposit insurance first (that’s the horse). You’d better make it clear that financial institutions that take these risks are not allowed to have access to lender-of-last-resort facilities. But the U.S. government didn’t do that.

So, of those two models, the Kareken-Wallace model makes you very cautious about lender-of-last-resort facilities and very sensitive to the risk-taking activities of banks. The Diamond-Dybvig and Bryant model makes you very sensitive to runs and very optimistic about the ability of insurance to cure them. Both models leave something out, and I think in the real world we’re in a situation where we have to worry about runs and we also have to worry about moral hazard. As you know, an important theme of research for macroeconomics in general and at the Minneapolis Fed in particular has been about how to strike a good balance.

Rolnick: Jack and Neil concluded their 1978 paper with a proposal for dealing with this tension, and that was to require much more capital than was required at the time. Now the government actually requires even less capital than it did when Jack and Neil wrote. If you go back prior to FDIC insurance, turn-of-the-century banks were holding, by some estimates, 20 percent, maybe 30 percent, capital. Capital-equity ratios were that high.

What would you recommend? You just observed that if deposit insurance isn’t priced properly, that leads you in one direction. And Jack and Neil had this idea of making sure there’s a lot more skin in the game, meaning much closer to what banks used to hold when there was no deposit insurance, no too-big-to-fail.

Sargent: The function of capital is exactly to protect against making risky loans. Another proposal is the narrow banking proposal of Milton Friedman and [other economists at the University of] Chicago, which is a proposal to force deposit banks to hold safe portfolios.

Rolnick: Well, with large banks, too-big-to-fail concerns and deposit insurance, I would make the case to tier it based on size. Jack and Neil made the point, I believe, that shareholders of large banks can diversify, but shareholders of smaller banks find it harder to diversify, so they tend to be more risk-averse. Their prediction would therefore have been, I think, that moral hazard is more likely to manifest itself in larger banks—and I think that’s what we saw in the 2007-09 financial crisis. How seriously would you take the relevance of the historical evidence that I cited?

Sargent: I would take it very seriously. I recommend a very interesting paper by Warren Weber presented at the Minneapolis Fed conference in honor of Gary Stern this past April in which Warren compared different private insurance arrangements for managing banks’ risk-taking before the U.S. Civil War.11

The 2009 fiscal stimulus

Rolnick: A January 2009 article quotes you as saying, “The calculations that I have seen supporting the stimulus package are back-of-the-envelope ones that ignore what we have learned in the last 60 years of macroeconomic research.”12 What calculations had you seen?

Sargent: I said something like that to a reporter. I had just read an Obama administration’s Council of Economic Advisers document e-mailed to me by my friend [Stanford University economist] John Taylor.13 I agreed with John that the CEA calculations were surprisingly naive for 2009. They were not informed by what we learned after 1945.

But I suspect that the council was asked to do something quickly, and

they did what they thought was “good enough for government work,” as some of us said during my days at the Pentagon in 1968 and 1969. Back-of-envelope work can be a useful starting point or benchmark. But it does mischief when it is oversold.

In early 2009, President Obama’s economic advisers seem to have understated the substantial professional uncertainty and disagreement about the wisdom of implementing a large fiscal stimulus. In early 2009, I recall President Obama as having said that while there was ample disagreement among economists about the appropriate monetary policy and regulatory responses to the financial crisis, there was widespread agreement in favor of a big fiscal stimulus among the vast majority of informed economists. His advisers surely knew that was not an accurate description of the full range of professional opinion. President Obama should have been told that there are respectable reasons for doubting that fiscal stimulus packages promote prosperity, and that there are serious economic researchers who remain unconvinced.

Rolnick: Do any New Keynesian models provide any support for the CEA numbers?

Sargent: Some do; some don’t. I recommend looking at calculations by John Taylor and his pals.14 Based on that work, John remains very skeptical of the 2009 CEA calculations. But Christiano, Eichenbaum and Rebelo have used variants of a New Keynesian model together with particular assumptions about paths of shocks to create quantitative examples of situations in which fiscal multipliers can be as big as those assumed by the CEA.15