"Standing Room Only" is not the phrase that leaps to mind when one thinks of lectures on monetary policy. But in San Diego this past winter, at a huge conference hall by the bay, no description could have been more apt.
As security agents guarded doors and TV cameras rolled tape, hundreds of economists packed the chairs and lined the walls. When the speaker approached the dais, these scholars craned their necks to catch a glimpse and held digital cameras aloft to snap his picture.
The speaker was Alan Greenspan, of course, and star power accounts for some of the excitement. The chairman is an almost mythically famous figure. But there was something more on this particular occasion. At the January 2004 meeting of the American Economic Association (AEA), Greenspan was set to review his years as chairman and to explain the art of making decisions in a world beset by doubt.
Because uncertainty is "the defining characteristic" of the monetary policy landscape, said Greenspan, "the conduct of monetary policy in the United States has come to involve, at its core, crucial elements of risk management." And because risks can be difficult to quantify, "risk management often involves significant judgment."
Critics, he acknowledged, have argued that such an approach "is too undisciplined—judgmental, seemingly discretionary, and difficult to explain." The Federal Reserve should attempt to be more formal in its operations, according to this argument, "by tying its actions ... to the prescriptions of a simple policy rule."
But Greenspan, speaking "from the perspective of someone who has been in the policy trenches," observed that while rules can serve as "helpful adjuncts" to policy, they don't permit the flexibility quintessential to coping with an uncertain world. "[A]t crucial points, like those in our recent policy history," he argued, "simple rules will be inadequate."
Greenspan thus delivered his verdict in the longstanding dispute between rules and discretion in monetary policymaking. It was a powerful message from a compelling messenger. And if the remarks of the prominent economists who followed him at the podium are any indication, his judgment carried both the day and much of the profession.
Former Fed Gov. Janet Yellen of the University of California, Berkeley, noted that the Greenspan Fed's "most important contributions probably came during episodes when policy departed from 'the rule'" in reaction to unusual conditions or large risks. The Federal Open Market Committee (FOMC) and Fed staff "earned their pay" by responding to such shocks, said Yellen, president of the San Francisco Fed as of June 2004. Harvard economist Martin Feldstein, president of the National Bureau of Economic Research, reinforced Greenspan's observations about the need for nuanced decision making under uncertainty.
And Mervyn King, Greenspan's counterpart at the Bank of England, observed that while constraints on his discretion—via explicit inflation targets—have worked well in England, the chairman's judgment was, truly, hard to fault. Joking that he'd love to play tennis with Greenspan (some 20 years his senior), King admitted, "I may move a little faster, but I expect him to make fewer mistakes."
But not all are convinced. In a November 2003 paper, Minneapolis Fed monetary adviser Patrick Kehoe, a University of Minnesota macroeconomist, has focused precisely on the question addressed by Greenspan: What is the proper amount of discretion in monetary policymaking? Should a monetary authority change direction and speed in response to economic winds and informed intuitions, or should it set forth on a track with a reasonable, tighter rein on its ability to deviate from a numerical target? Greenspan argued that Fed flexibility shouldn't be constrained. Kehoe has his doubts.
"What's the rationale for the Fed having unlimited discretion?" Kehoe asks, in an interview. "Well, in our theoretical model we argue that if the Fed doesn't have any special knowledge, any private information, then we, society, can just tell the Fed, 'here's what you should do.'"
In practice, Kehoe believes, the Fed should lay out a coherent quantitative rationale for its policy-including the model it's using; it should announce numerical targets and, most importantly, whenever it misses these targets by more than a pre-specified amount, it should have to explain to Congress in detail the reasons why.
On its surface, the paper, "The Optimal Degree of Discretion in Monetary Policy," written with Susan Athey of Stanford University and Minneapolis Fed adviser Andrew Atkeson of the University of California, Los Angeles, appears to offer a middle ground, an olive branch as it were. Optimal policymaking, it seems to say, should allow a limited measure of latitude to monetary decision makers, as long as the economy is operating within certain bounds. If conditions are more extreme, then decision makers should be constrained to implement a constant inflation rate. In practice, they show, this policy can be implemented with an inflation cap, similar to the constrained discretion that Fed Gov. Ben Bernanke advocates (see interview).
But this interpretation of their paper is superficial and misleading, says Kehoe, like taking Swift's "Modest Proposal" at face value. A deeper understanding conveys a very different and severe message: Yes, a case can be made for limited discretion in monetary policy, but that case hinges critically on an assumption that economists generally find untenable—that the Fed relies on private information unavailable to others. If you deny that assumption—as most would—the case for discretion dissolves. Monetary policy will be optimal only if rules are followed.
A long tradition
Rules vs. discretion: The phrase has launched a thousand scripts. Reams of paper and decades of argument seem to have parsed every conceivable aspect of that dichotomy, and while policymakers tend to favor discretion, a substantial tradition in economics holds that rules are required for sound monetary policymaking.
Among the first proponents of monetary rules was University of Chicago economist Henry Simons who, in 1936, published "Rules versus Authorities in Monetary Policy." The paper was a broadside against giving monetary policymakers leeway. Writing in the midst of the Great Depression and fearful that socialism or fascism might arise out of economic chaos, Simons argued, "We should not balk at bold schemes for restoring the free-enterprise system to a securely workable basis." And paramount in that scheme: "[B]ring the appropriate [monetary] agencies more closely within the control of law, through the adoption of rules."
Stable monetary rules were critical to business, noted Simons. "We must avoid a situation where every business venture becomes largely a speculation on the future of monetary policy," he argued. It was essential to "avoid reliance on discretionary (dictatorial, arbitrary) action by an independent monetary authority" by defining "a statutory rule which might be enacted by the competent legislature."
A dozen years later, Milton Friedman took up the torch of his University of Chicago colleague. The "monetary framework should operate under the 'rule of law' rather than the discretionary authority of administrators," wrote Friedman. He called for reform that would "eliminate ... discretionary control of the quantity of money by central bank authority" and said "the existing powers to engage in open market operations ... should be abolished." Among other things, discretion encouraged incumbent politicians to pressure the central bank to boost employment and output right before elections, leading to the instability of the "political business cycle."
Friedman later pointed out that discretionary control of money supply was inherently flawed because monetary decision makers can't precisely determine the right magnitude or timing for policy interventions—measurements are inexact, relationships complex and time lags inevitable. For him, the best monetary rule was a publicly stated steady rate of growth in money supply, although "the precise rate of growth, like the precise monetary total, is less important than the adoption of some stated and known rate."
In arguing that rules were needed because monetary authorities were susceptible to political influence and inaccurate data, Simons and Friedman implicitly conceded that, in theory, a policymaker could do a good job if he or she had society's best interests at heart and had good information about the economy. A wise and benevolent central banker would, in effect, substitute the golden rule for a monetary rule: Do monetary policy as you would have others do policy unto you.
In 1977, that idea too was laid to rest. In their landmark paper, "Rules Rather than Discretion: The Inconsistency of Optimal Plans," Finn Kydland, now at Carnegie Mellon University, and the Minneapolis Fed's Edward Prescott showed that even thoroughly competent policymakers with good information and the best of intentions will—if they don't commit to a policy via a rule of some sort—inevitably make decisions that lead to suboptimal outcomes. The reason is time inconsistency, a powerful but less than intuitive concept.
We all suffer from time inconsistency: We make promises to ourselves ("eat less") but break them later on because what seemed like a great idea at one point seems less great later on ("that donut sure looks good"). In a classic 1956 paper, Northwestern University economist R.H. Strotz discussed the idea as it pertained to individuals. "The optimal plan of the present moment is generally one which will not be obeyed," he observed. "The individual's future behaviour will be inconsistent with his optimal plan." Strotz's alliterative phrase for efforts to resolve such inconsistencies: "the intertemporal tussle."
But as Kydland and Prescott explained, the tussle gets tougher when two or more parties are involved. If, for example, the government wants to discourage people from building homes in floodplains, it might announce that it won't provide flood insurance for such structures. But when a flood disaster hits and people lose their homes, politicians are likely to bail them out. That's time inconsistency—what once seemed optimal no longer is. The real problem is that because the public observes this inconsistency, subsequent government announcements discouraging floodplain settlement are unlikely to be believed. Policy becomes ineffective and unpredictable.
In essence, Kydland and Prescott show that because policymakers and the public are always watching one another, and making decisions based on their belief about what the other will do next, policymakers are prone—unless their hands are tied by a rule—to be time inconsistent. This temptation leads unavoidably to bad outcomes, either inefficiency or instability.
Inconsistent monetary policy
The phenomenon is ubiquitous—Kydland and Prescott provide examples of time inconsistency in patent and tax policy as well as flood control—but it is especially pertinent to monetary policy. Suppose, for example, that a labor union is negotiating a new contract and thinking about locking in a cost-of-living increase; if current inflation is 2 percent, the union might call for a 2 percent annual wage hike.
If some sort of structural problems (bad fiscal policies, for example, or monopoly practices) are causing unemployment to be higher than it should be, a well-meaning central bank might try to increase the money supply to boost inflation and—taking advantage of the short-run Phillips curve trade-off—decrease unemployment.
But the next time around, a forward-looking labor union would anticipate such central bank action and demand a wage increase to reflect its expectation of higher future inflation. Employers might agree to the wage demand but they'd have no incentive to offer more jobs. In the end, then, the central bank's actions would result in higher inflation but no gain in employment or output. Thus, time inconsistency creates a persistent inflationary bias. Still, the possibility of surprising the economy with a bit of inflation means that even benevolent policymakers are inexorably tempted to be time inconsistent—unless they're committed by rules.
"The implication of our analysis is that policymakers should follow rules rather than have discretion," wrote Kydland and Prescott. "The reason that they should not have discretion is not that they are stupid or evil but, rather, that discretion implies selecting the decision which is best, given the current situation. Such behavior either results in consistent but suboptimal planning or in economic instability."
Reputation and information
Economists have devoted substantial energy to exploring what policies or practices—short of rules—might mitigate time inconsistency. Robert Barro and David Gordon, for instance, studied the idea that by faithfully following through on announced policy, a central bank could build its credibility, and so reputation might substitute for a rule.
In 1985, Matthew Canzoneri, an economist at Georgetown University, contributed another idea. If the public and the central bank all have equal information about the economy, he said, then it should be fairly simple to solve "the precommitment problem" (that is, ensuring that the central bank sticks to its stated policy) either by the central bank building a credible reputation for adhering to its policy announcements or by Congress legislating a policy rule and policing the central bank's adherence to it.
But then Canzoneri introduced the confounding element of unequal information. "Credibility problems arise only when scope has been left for cheating," he wrote, and if one party knows more than another, that scope widens dramatically. What if the central bank—in his paper, the Fed—has private information about the state of the economy, information that the public doesn't have? "If the Fed's forecast ... is private information, then the Fed's adherence to the ideal policy rule cannot be verified directly, and efficient resolutions of the precommitment problem are much more difficult to find."
This is where Athey, Atkeson and Kehoe enter the discussion. If there is no private information, they point out, there is no way to cheat the system in Canzoneri's sense. Everyone knows what's going on and a monetary rule is optimal. But for the sake of argument, say the economists, let's give the advocates of discretion their strongest possible case: Let's assume that the central bank does have some private information about the economy.
This assumption instantly creates a tension between the rigidity of a rule that ensures no time inconsistency (but doesn't allow the bank to respond to information that might affect the economy) and the flexibility of discretion that enables the bank to use that information to society's benefit (but might also permit it to influence the economy in less constructive ways).
If society knew what the bank knows, it might wish the bank to use that information wisely in shaping monetary policy. But if society gives the bank too much freedom, it might abuse it by causing inefficient inflation. How much leash should the bank have? What is the optimal degree of discretion?
Games and mechanisms
To answer that question, the authors draw upon two distinct but related methods. The first is game theory—in which players try to do as well as possible in a game with other players by figuring out their best possible moves. The "prisoner's dilemma" in which two prisoners have to decide whether or not to confess to a crime they committed together, without knowing what their partner will do, is a well-known game theory problem, and its solution is called a Nash equilibrium, after John Nash, the mathematician portrayed in A Beautiful Mind.
The other method is mechanism design theory—in which the challenge is not to figure out the best plays given the rules of the game, but rather to discover the rules of the game that lead to the best possible outcome. Mechanism design theory was used to design airwave spectrum auctions, for example, so that the federal government could maximize revenues when it auctioned off radio frequencies.
Athey, Atkeson and Kehoe apply game theory and mechanism design theory to a monetary policy game in which one player (the central bank) has more information than the other players (the public: workers, firms and households). And as they blend these elements, they develop new techniques to address a very old problem.
First they develop a mathematical representation of the monetary policy question. The central bank's objective, they say, is to maximize social welfare, and that depends on nominal wages, the rate of growth in money supply (roughly equal to inflation), plus the Canzoneri element: the private information the central bank has about the impact of its policy stimulus on social welfare (similar to the time consistency phenomenon of the Fed boosting the economy with surprise inflation).
Given this model, the economists explore how its elements interact over time, with money supply growth, private information and wages all interdependent. "The mechanism design problem," they write, "is to choose a monetary policy, a [private information] reporting strategy, and a strategy for average wages the outcomes of which maximize social welfare subject to the constraint that these strategies are incentive-compatible [that is, no one is tempted to cheat]."
The objective is to figure out what rules of the monetary policy game will result in the best outcome. In other words, if society (via legislation) were to draw up a contract with the central bank, how much flexibility should that contract allow? The optimal contract will give the bank enough leeway to use its private information for society's benefit—stimulating the economy to avoid recession—but not so much that the bank creates the inefficient inflation or instability inherent to time inconsistent policymaking. "That's what the whole paper's about," says Kehoe. "And then there's a whole bunch of math."
Indeed. The paper is exceptionally complex, replete with esoteric mathematics that even fellow economists find daunting. With four technical appendices, three lemmas, five propositions and 59 Greek-laden formulas, the entirely theoretical piece is a virtuoso display of advanced methodology. Impressive. But how does it solve the problem?
Dynamic or static?
The most difficult part of the puzzle is determining whether this dynamic game—in which the Fed makes a move, followed by society's move, followed by the Fed's, ad infinitum—requires a dynamic or static solution.
A dynamic solution, explains Kehoe, is one in which "your future ability to do something is a function of what you did before. And the more you do of what I don't want today, the more I take it away from you in the future." Relative to monetary policy, then, a dynamic mechanism would be a contract with an incentive structure that, round by round, tightens the central bank's range of freedom if it exceeds certain limits. A static solution would be a single rule determined without reference to the central bank's previous actions.
In essence, then, the economists' intricate math determines whether society's welfare is maximized by a contract that just sets a monetary rule for the Fed and sticks to it, or one that establishes incentives for the central bank, telling policymakers that what they can do now depends on what they did last time around.
A dynamic game seems to call for a dynamic solution: Most repeated games in which one player establishes a contract with another need changing incentives to generate best outcomes. But unexpectedly, says Kehoe, "the technical work in our paper shows that it doesn't work like that. Our mechanism is static, a flat number," not a dynamically tightening band. A partial reason for the surprising result: In this game—unlike many—the conflict between the players has to do with time inconsistency, not competition over resources. The central bank is trying to maximize society's welfare, not compete for its resources, so society doesn't directly benefit by progressively restricting the bank's authority.
The character of discretion
Mathematically, once they've found that a static mechanism is optimal, the hardest work is done. They've dramatically narrowed the range of possible mechanisms. The next step is to describe the optimal static mechanism, the shape of discretionary power the central bank should be granted so that it can achieve the best societal outcome.
With another series of equations and two more proofs, they determine that giving the central bank total discretion never provides monetary policy that maximizes social welfare. Rather, welfare can be maximized if the central bank contract permits it discretion up to a limit. The bank is free to vary its policy in response to its private information (as well as public information) about the economy as long as it keeps below this explicit upper bound on monetary growth. But beyond that level, the bank is given no discretion: It must choose a constant inflation rate regardless of what it knows.
The economists then demonstrate that this optimal policy of bounded or no discretion can be implemented simply by setting an inflation cap—an upper limit for the central bank's inflation rate. Again, the bank has freedom to implement whatever inflation rate it wishes as long as it doesn't exceed that cap: Such a mechanism can maximize societal welfare.
"One interpretation of our work," write the economists, "is that we solve for the optimal inflation targets." And indeed, inflation targeting—in which a central bank publicly announces a numerical target for inflation and conducts policy so as to hit that target—is an increasingly popular monetary mechanism, adopted in perhaps a dozen countries. As the Bank of England's Mervyn King noted in San Diego, it works for him.
The Greenspan Fed has resisted the trend, but some FOMC members favor it. Fed Gov. Bernanke, for instance, has suggested that a variation of inflation targeting that he calls "constrained discretion" would improve Fed policymaking. The connection isn't lost on Athey, Atkeson and Kehoe. "Indeed," they write, "our work here provides one theoretical rationale for the type of constrained discretion advocated by Bernanke and [his collaborator, Frederic] Mishkin."
When sirens beckon
The final step is to explore how the level of allowable discretion varies relative to the time inconsistency problem faced by the economy. Some economies—those with structural problems like bad tax policies, rigid labor unions or entrenched monopolies that inhibit investment, labor mobility and price flexibility—face greater temptation than others to use monetary policy to inflate away unemployment or debt problems.
And as expected, Athey, Atkeson and Kehoe (AKK) find that the more temptation a monetary authority faces to create "surprise" inflation, the less discretion it should have. It's like that resolution about eating less. Your dietary leash should be a lot tighter if you work in a bakery. As Strotz put it, the more enticing the sirens, the more tightly Ulysses should be bound to the mast.
Kehoe explains it with an international example: Take two countries, let's call them Argentina and Chile. Think of Argentina as having such an inefficient tax system that trying to pay its national debt by raising taxes is politically very costly. "So it has a huge incentive to pay for things by using inflation rather than taxes," says Kehoe. "But think of Chile as having a better tax system so it's not so costly to pay for its debts by raising taxes." In this hypothetical situation, Argentina has a worse time inconsistency problem, a greater temptation to use monetary policy in a way that would be expedient but ultimately harmful.
"What would our model say about the way their central banks should operate? Well, in Argentina they should have a really low inflation cap or really tightly constrained rule because as soon as you give the Argentine central bank a little bit more leash to do what they want, boom, inflation will skyrocket." Chile, on the other hand, faces less temptation to inflate its debt problems away. "So if you give them a little bit of room, they won't abuse it so much. You can give them a wider band. That's the idea."
Last November, not long before Greenspan's AEA speech, Kehoe presented his paper at a high-level conference on monetary policy in Washington, D.C. Given the venue—the Federal Reserve Board—perhaps it was predictable that reactions were mixed.
Bennett McCallum, a Carnegie Mellon University macroeconomist, was one of the paper's designated discussants at the meeting, and in an e-mail to The Region he describes the paper as "extremely skillful but not highly relevant." In his conference remarks, McCallum expressed several reservations. "I find unpersuasive the notion that a CB [central bank] cannot commit its actions, but someone else can constrain it to behave in a committed manner," he writes. (Kehoe responds that society often creates constraints by imposing costs, financial or otherwise. "Our Constitution is a perfect example," he notes. "It's a device created by society to define and constrain the powers of government.") In addition, the AAK model doesn't include "forward-looking behavior" by individuals, a feature that McCallum has developed in some of his recent work. (Kehoe says that its inclusion would complicate the math but not change the results.)
But it's most telling that McCallum and the paper's other discussant, Kenneth Rogoff of Harvard University, seem especially concerned with another element of the AAK model: It hinges on the tenuous assumption that central banks rely on private information.
"Overall, it is a terrific paper," observes Rogoff by e-mail. "In my remarks, however, I did question the emphasis their paper (and Canzoneri's earlier paper) puts on the FOMC knowing more than the private sector does." The Fed may be better informed than others about its own intentions and preferences, Rogoff acknowledges, but he doubts it has a real information advantage about the state of the economy. McCallum agrees: "To me it seems doubtful that such information is of first-order importance," he remarks. "My position seems to be rather standard among monetary economists."
Hearing these comments again, months later, Kehoe appears frustrated. It's clear that the paper's underlying message hasn't come across. He doesn't believe the Fed depends on private information any more than McCallum or Rogoff do. Rather, he says, the paper shows that only by embracing such a questionable assumption can one make a case for central bank discretion. And even then, it's constrained discretion.
"The deep message of our paper is really a critique of the whole idea of discretion," he explains. "The real purpose was to say, 'you keep telling me there should be discretion. Well, for you to believe that, you have to believe the Fed knows all this information that the private sector doesn't know. And if you don't believe that, then you'd better argue with me that the Fed should indeed not have much discretion. It should have tight rules.' That's the real meaning."
Indeed, for all its algebraic obscurity, the paper is quite lucid on this point. "Here the rationale for discretion depends in a critical way on the monetary authority having some private information that the other agents in the economy do not have," write the economists. "Of course, if the amount of such private information is thought to be very small in actual economies, then our work argues that in such economies the logical case for a sizable amount of discretion is weak, and the monetary authority should follow a rather tightly specified rule."
At the Washington conference, Kehoe recalls, those participants who tried to sustain these conflicting ideas—that private information isn't important but the Fed should have discretion—experienced "a dissonance, a cognitive dissonance" and they reacted by invoking "Knightian uncertainty."
As they used the term (inaccurately, says Kehoe) it refers to unforeseen events (a 9/11 disaster, for example, or an October 1987 stock market crash). Given that such events will occur—unpredictably—the Fed shouldn't be held to a strict rule; it needs to be able to respond. Without flexibility, central banks become stones that sink a weakened economy.
Kehoe responds that Knightian uncertainty is a rather vague concept. "To many," he says, "it means events which are ex ante zero probability, but which actually happen. And I mean truly zero probability. Not small; literally zero. They happen and you could never have even dreamed them up. Those kinds of events are so rare they're irrelevant." (Irrelevant, at least, to making decisions.)
By contrast, terrorist attacks and stock market crashes are, regrettably, not so rare. We've experienced them before, can imagine them happening again, and, to some degree, we've actually prepared for them. It isn't that difficult to estimate the probability that such things might happen. Kehoe refers to these unusual (but not zero-probability) occurrences, as "exotic events." If they happen and the Fed wants to raise inflation above the inflation cap to accommodate them, says Kehoe, that's reasonable. But exotic events don't vitiate the argument for limiting discretion. They're exceptions that prove the rule.
And in such cases, the Fed should be required to publicly and clearly explain why it needs to exceed the limit. "Anything above, say, 4 percent," Kehoe says, hypothetically, "the Fed chairman should have to trot up to Congress and explain clearly why the Fed deviated." Alternatively, policymaking rules could require a unanimous vote of the FOMC to implement a policy that exceeds the rule. "Something that ups the ante for breaking it," says Kehoe, "so there's a cost to deviating."
There are other views of Knightian uncertainty, Kehoe remarks. "Some envision events that have high ex ante probability, but which we could never hope to describe before they happen. Society could deal with these in the same way: Allow the Fed to deviate but make sure there is a clear cost to doing so."
A structured debate
Kehoe has spent years studying optimal monetary policy. Often collaborating with Atkeson or V.V. Chari at the Minneapolis Fed, he has developed, tested and refined mathematical models to better understand central bank policies and the macroeconomy they influence. In that sense, "The Optimal Degree of Discretion in Monetary Policy" is not unique. His goal in this and a dozen other papers has been to clarify the links between policy and outcome—or as he might put it, the mapping between actions and objective functions.
Fed decision making could be improved if these theoretical links were clearer, Kehoe believes, but the clarity now lacking has more to do with the Fed's methods than with its models. "Most economists who interact with the Fed, academic economists, are somewhat surprised by ..."
He pauses—uncharacteristically—looking for a phrase both accurate and diplomatic.
"... by the lack of coherent statement of what exactly the Fed's objective function is, in an almost quantifiable way, and how current policies are chosen so as to maximize that objective."
Economists are trained in that language, he notes. State your objectives; write down a model describing the relationship between economic variables and those objectives; estimate the relevant parameters. And then, focusing on the variables that can be influenced by policy, set them so as to maximize the objectives.
"Almost all economists I know would be much more comfortable if the Fed moved toward that language and method because then we would know what we're really arguing about," he explains. "We could say, 'that's not the right objective function.' Or 'I disagree with that coefficient.'"
Current Fed decision making is too hazy, in Kehoe's view. And discretion is part of the problem—it limits the transparency of the decision-making process. "The Fed does what it wants and then crafts some words about why it makes sense," he observes. "There's a much less tight connection to economic theory, models and quantitative data than most of us are comfortable with."
Kehoe's objective function, then, is not only to improve theory, but to persuade policymakers that they'd be better served by dropping the veil that obscures the decision-making process. "The Fed should be saying, 'Here's our objective function. Here's how we think our policy maps into outcomes,'" he argues. "If it's in that language, then at least, economists can have a structured, concrete debate. The way it is now, the profession could go on for years and just run in circles."