What do we want a central bank to do? The initial mandate given to the Fed, in the Federal Reserve Act of 1913, was to "furnish an elastic currency." What the framers of the Act appeared to have had in mind was intervention by the central bank to smooth the effects of predictable seasonal fluctuations in the demand for currency, and to prevent the banking panic episodes that had occurred during the National Banking era after the Civil War. Intervention was to occur primarily by use of the discount window - there did not appear to be a good understanding of open market operations and their effects in 1913.
After World War II, with the growing influence of Keynesian Economics, Congress passed the Employment Act of 1946. This Act was passed in a context where the Fed had little to do. Between 1942 and 1951, the Fed was committed to pegging interest rates on government bonds, so the Act applied essentially only to the fiscal authority, setting up the Council of Economic Advisers, requiring the President to issue an annual Economic Report, and required the federal government to "promote maximum employment, production, and purchasing power." Of course, given the vagueness of that statement and the lack of any incentives, the Act had no teeth at all.
The Employment Act was amended in 1978, in the form of the Full Employment and Balanced Growth (or Humphrey-Hawkins) Act, with language that referred specifically to the Fed. The Chairman of the Board of Governors was required to testify to Congress on a regular basis, and was required by the Act to maintain long-run growth, minimize inflation, and promote price stability. The 1978 Act was somewhat more explicit than its precursor in setting goals for the federal government and the Fed, but it also has no teeth. However, both Acts enshrine the idea, presented to masses of freshman undergraduates in economics programs, that the government, through its monetary and fiscal authorities, is responsible for keeping us employed and subject to low inflation.
Though vague, the emphasis on real rather than purely nominal objectives for monetary policy is different from what is done in some other countries, particularly those with explict inflation targets, like New Zealand, Canada, or the UK. The Fed, while sometimes viewed as more independent than the average central bank in the world, seems somewhat more tied to Keynesian-type stabilization policy than some other central banks, through the above Acts of Congress.
Now, Old Keynesians of course had their opponents, the Old Monetarists, lead by Milton Friedman. Friedman thought that discretionary stabilization policy was generally harmful, due to the fact that we do not know enough about how the economy works, measurement is imperfect, policymakers work slowly, and policy itself works slowly, once implemented. Friedman's proposed solution was to bind the central bank to a policy rule, but unfortunately he picked the wrong one. Central bank experiments with money growth rules in the 1970s and 1980s were a failure, for well-known reasons. The Fed's current operating procedure - peg the policy rate between FOMC meetings - works because it optimally absorbs a variety of anticipated and unanticipated shocks to financial markets and payments arrangements that dominate over a short horizon.
The Taylor rule paper changed things dramatically for Keynesian-minded macroeconomists. The Taylor rule determines the current setting for the Fed's policy rate (in normal times, the fed funds rate) as a function of the observed inflation rate and the "output gap," specified in Taylor's original paper as the deviation of real GDP from trend. The Taylor rule permits one to be a Keynesian - Keynesian economics being a justification for putting the output gap in the policy rule - while also avoiding discretion, and therefore staying out of trouble with the Old Monetarists. The Taylor rule ultimately became enshrined in New Keynesian Economics, and is a very visible part of the policy narrative within all central banks currently.
Part of the attraction of the Taylor rule for central bankers was that it required no change in their thinking. Indeed, part of Taylor's original sales pitch for his rule was that it fit recent (in 1993) Fed behavior. Central bankers could then rest easily, given the notion that they had finally solved the monetary policy problem. Indeed, features of the US times series data after the early 1980s helped to promote this view. One of my favorite speeches by Ben Bernanke when he was still a Fed Governor is this one, from 1994, on the Great Moderation, the period following the 1981-82 recession until 2007. As we all remember, the Great Moderation was a period of relatively low real GDP volatility and low inflation in the United States.
Bernanke's speech makes the case that better monetary policy was an important (though not the only important) contributor to the Great Moderation, and uses the Taylor rule and the "Taylor curve" to organize our thinking about this. In public statements like this, and in the direction that research took in the Federal Reserve System and other central banks, the Fed set itself up for criticisms like this one, from Paul Krugman. Krugman simply takes the standard quadratic-loss objective function that we can use to derive a Taylor rule, and plots the implied losses over time. Of course, the loss grows substantially beginning in mid-2008 until the present, driven by a growing "output gap" and a decrease in inflation below an assumed 2% target.
On the Fed's terms, Krugman's criticism is quite reasonable. By buying into the Taylor rule language and in public statements by its officials, the Fed has implicitly accepted that:
1. The central bank should care about two things: the inflation rate and aggregate economic activity.
2. The central bank can determine what the inflation rate and aggregate economic activity should be at any point in time.
3. A fixed rule should determine a federal funds rate target as a function which is increasing in the observed inflation rate, and decreasing in the gap between observed economic activity and what aggregate economic activity should be.
On point 1: It seems well-recognized that long-run inflation, and an uncertain inflation rate are costly, and that inflation is a monetary phenomenon and therefore under the control of the central bank, at least over long periods of time. However, there seems to be no agreement among macroeconomists concerning the effects of monetary policy on real macroeconomic activity and the role of policy in affecting such activity. Even basic New Keynesian theory does not provide for such a role. In the standard New Keynesian framework, inefficiencies arise because of relative price distortions, which arise from price stickiness. The problem can be cured with price stability - there is no need in the New Keynesian world for the central bank to pay any attention to some "output gap" measure.
On point 2: What is the appropriate long-run inflation rate? We seem to have no idea. Here, I discussed what our models tell us, which is that some deflation (in some cases all the way to the Friedman rule) is optimal. Somehow, the Fed seems to have arrived (as far as we can tell) at an optimal long-run inflation rate of 2% per annum. But there is no good science to back this up, other than the lay-low principle of central bank political economy: 2% keeps you out of trouble. What about the output gap? Clearly no one has a clue how to measure this. The correct measure would be the difference between current real GDP and efficient real GDP, but typically the measure used is something like what Taylor originally proposed - the difference between trend and actual real GDP (or trend and actual unemployment rate). Under the current circumstances, I am not sure whether what I am seeing is close to efficient or not. Is the unemployment rate high because of some inefficiency arising from wage and price stickiness, an inefficiency due to malfunctioning credit markets, or has the matching process in labor markets become less efficient (due to mismatch - see here and here). Which it is determines whether monetary policy can or should do anything about it.
On point 3: Oddly enough, the Taylor rule is not an optimal policy rule that can be derived in any well-articulated monetary model that I know about. As I pointed out above, it is not optimal in basic New Keynesian models, though Woodford has some slippery ways of coaxing it out of his models (see here).
What is the bottom line? (i) Obviously we need to know more. More research is needed on the costs of inflation and the role of monetary policy. (ii) Until we know more, central bankers should be more humble (if they aren't humbled enough by recent events) and so should people like Krugman, who seem very certain about what they think the central bank should do. (iii) I think a superior approach to central banking in the United States would be to adopt explicit inflation targeting, as has been done in other countries. Why? The public (and indeed some well-known economists) seem to have the impression that central banking is an engineering problem whereby the Fed delivers the "correct" level of real GDP or the correct unemployment rate. Inflation targeting would help to dispel that notion. What should the target be? Good question - I'm groping for an answer here. In his younger days, Bernanke was clearly in favor of inflation targeting. Obviously he has been distracted of late, but maybe now is the time to get the ball rolling.