The financial crisis that began in August 2007 has been the most severe of the post-World War II era and, very possibly–once one takes into account the global scope of the crisis, its broad effects on a range of markets and institutions, and the number of systemically critical financial institutions that failed or came close to failure–the worst in modern history. Although forceful responses by policymakers around the world avoided an utter collapse of the global financial system in the fall of 2008, the crisis was nevertheless sufficiently intense to spark a deep global recession from which we are only now beginning to recover.
Even as we continue working to stabilize our financial system and reinvigorate our economy, it is essential that we learn the lessons of the crisis so that we can prevent it from happening again. Because the crisis was so complex, its lessons are many, and they are not always straightforward. Surely, both the private sector and financial regulators must improve their ability to monitor and control risk-taking. The crisis revealed not only weaknesses in regulators’ oversight of financial institutions, but also, more fundamentally, important gaps in the architecture of financial regulation around the world. For our part, the Federal Reserve has been working hard to identify problems and to improve and strengthen our supervisory policies and practices, and we have advocated substantial legislative and regulatory reforms to address problems exposed by the crisis.
As with regulatory policy, we must discern the lessons of the crisis for monetary policy. However, the nature of those lessons is controversial. Some observers have assigned monetary policy a central role in the crisis. Specifically, they claim that excessively easy monetary policy by the Federal Reserve in the first half of the decade helped cause a bubble in house prices in the United States, a bubble whose inevitable collapse proved a major source of the financial and economic stresses of the past two years. Proponents of this view typically argue for a substantially greater role for monetary policy in preventing and controlling bubbles in the prices of housing and other assets. In contrast, others have taken the position that policy was appropriate for the macroeconomic conditions that prevailed, and that it was neither a principal cause of the housing bubble nor the right tool for controlling the increase in house prices. Obviously, in light of the economic damage inflicted by the collapses of two asset price bubbles over the past decade, a great deal more than historical accuracy rides on the resolution of this debate.
The goal of my remarks today is to shed some light on these questions. I will first review U.S. monetary policy in the aftermath of the 2001 recession and assess whether the policy was appropriate, given the state of the economy at that time and the information that was available to policymakers. I will then discuss some evidence on the sources of the U.S. housing bubble, including the role of monetary policy. Finally, I will draw some lessons for future monetary and regulatory policies.1
U.S. Monetary Policy, 2002-2006
I will begin with a brief review of U.S. monetary policy during the past decade, focusing on the period from 2002 to 2006. As you know, the U.S. economy suffered a moderate recession between March and November 2001, largely traceable to the ending of the dot-com boom and the resulting sharp decline in stock prices. Geopolitical uncertainties associated with the terrorist attacks of September 11, 2001, and the invasion of Iraq in March 2003, as well as a series of corporate scandals in 2002, further clouded the economic situation in the early part of the decade.
Slide 1 shows the path, from the year 2000 to the present, of one key indicator of monetary policy, the target for the overnight federal funds rate set by the Federal Open Market Committee (FOMC). The Federal Reserve manages the federal funds rate, the interest rate at which banks lend to each other, to influence broader financial conditions and thus the course of the economy. As you can see, the target federal funds rate was lowered quickly in response to the 2001 recession, from 6.5 percent in late 2000 to 1.75 percent in December 2001 and to 1 percent in June 2003. After reaching the then-record low of 1 percent, the target rate remained at that level for a year. In June 2004, the FOMC began to raise the target rate, reaching 5.25 percent in June 2006 before pausing. (More recently, as you know, and as the rightward portion of the slide indicates, rates have been cut sharply once again.) The low policy rates during the 2002-06 period were accompanied at various times by “forward guidance” on policy from the Committee. For example, beginning in August 2003, the FOMC noted in four post-meeting statements that policy was likely to remain accommodative for a "considerable period."2
The aggressive monetary policy response in 2002 and 2003 was motivated by two principal factors. First, although the recession technically ended in late 2001, the recovery remained quite weak and “jobless” into the latter part of 2003. Real gross domestic product (GDP), which normally grows above trend in the early stages of an economic expansion, rose at an average pace just above 2 percent in 2002 and the first half of 2003, a rate insufficient to halt continued increases in the unemployment rate, which peaked above 6 percent in the first half of 2003.3 Second, the FOMC’s policy response also reflected concerns about a possible unwelcome decline in inflation. Taking note of the painful experience of Japan, policymakers worried that the United States might sink into deflation and that, as one consequence, the FOMC’s target interest rate might hit its zero lower bound, limiting the scope for further monetary accommodation. FOMC decisions during this period were informed by a strong consensus among researchers that, when faced with the risk of hitting the zero lower bound, policymakers should lower rates preemptively, thereby reducing the probability of ultimately being constrained by the lower bound on the policy interest rate.4
Evaluating the Tightness or Ease of Monetary Policy
Although macroeconomic conditions certainly warranted accommodative policies in 2002 and subsequent years, the question remains whether policy was nevertheless easier than necessary. Since we cannot know how the economy would have evolved under alternative monetary policies, any answer to this question must be conjectural.
One approach used by many who have addressed this question is to compare Federal Reserve policies during this period to the recommendations derived from simple policy rules, such as the so-called Taylor rule, developed by John Taylor of Stanford University (Taylor, 1993). This approach is subject to a number of limitations, which are important to keep in mind.5 Notably, simple policy rules like the Taylor rule are only rules of thumb, and reasonable people can disagree about important details of the construction of such rules. Moreover, simple rules necessarily leave out many factors that may be relevant to the making of effective policy in a given episode–such as the risk of the policy rate hitting the zero lower bound, for example–which is why we do not make monetary policy on the basis of such rules alone. For these reasons, even strong proponents of simple policy rules generally advise that they be used only as guidelines, not as substitutes for more complete policy analyses; and that, to ensure robustness, the recommendations of a number of alternative simple rules should be considered (Taylor, 1999a). That said, as much of the debate about monetary policy after the 2001 recession has made use of such rules, I will discuss them here as well.
The well-known Taylor rule relates the prescribed setting of the overnight federal funds rate–the interest rate targeted by the FOMC in its making of monetary policy–to two factors: (1) the deviation, in percentage points, of the current inflation rate from policymakers’ longer-term inflation objective; and (2) the so-called output gap, defined as the percentage difference between current output (usually defined as real GDP) and the “normal” or “potential” level of output. In symbols, the standard form of the Taylor rule is given by the equation shown in Slide 2. In this equation, it is the prescribed value of the policy interest rate in a given period t; pt — pt* is the deviation of the actual inflation rate p from its target p* in period t; and yt — yt*, the “output gap,” is the deviation of actual real output y from potential output y* in period t. The parameters a and b are positive numbers that describe how strongly the policy rate should respond to deviations of inflation from its target and of output from its potential.
As we would expect, the Taylor rule tells policymakers that interest rates should be higher when inflation is above target, (pt — pt*) > 0 , or when output is above its potential, (yt — yt*) > 0. Taylor (1993) estimated the long-run real value of the federal funds rate to be about 2 percent. The equation for the Taylor rule accordingly shows that when inflation and output are equal to their targets, the federal funds rate–which is expressed here in nominal terms–should equal 2 plus the rate of inflation. Equivalently, when inflation and output equal their targets, the real value of the federal funds rate should equal 2 percent.
To make the Taylor rule equation shown in Slide 2 operational, one needs to specify numerical values for the coefficients a and b, choose appropriate indicators of inflation and output, and specify a target rate for inflation and a measure of potential output. In his 1993 paper introducing his eponymous rule, Taylor suggested setting both a and b equal to 0.5. So, for example, according to the original Taylor rule, if output rises 1 percent relative to its potential, then, all else equal, the Federal Reserve should raise its policy rate by 0.5 percent, or 50 basis points. Following Taylor’s suggestions for parameter values, in Slide 3 we show by the dashed red line the values of the federal funds rate implied by the Taylor rule for the period from 2000 to the present, with inflation measured by the consumer price index (CPI), the Fed’s assumed inflation target set to 2 percent, output measured by real GDP, and the output gap as estimated retrospectively by the Federal Reserve’s primary forecasting model, the FRB/US model. The Taylor rule prescription is juxtaposed with the actual path of the policy rate taken from Slide 1, again shown in blue.
The comparison displayed in Slide 3 provides the most commonly cited evidence that monetary policy was too easy during the period from 2002 to 2006, as the actual federal funds rate is below the values implied by the Taylor rule–by about 200 basis points on average over this five-year period (Taylor, 2007).
Of course, the validity of that conclusion depends on whether the specific assumptions and measurements used to construct the Taylor rule’s policy prescription are appropriate. Room for disagreement exists. For example, some empirical and simulation evidence suggests that the responsiveness of policy to the output gap, given by the parameter b in the Taylor rule equation, should be higher than the value of 0.5 originally chosen by Taylor.6 Higher values of b lead the Taylor rule to recommend somewhat lower policy rates during recessions and their aftermaths.
The prescriptions of the Taylor rule may also depend sensitively on how inflation and the output gap are measured. The difficulties in measuring the output gap, particularly in real time, are well known. The choice of inflation measure may also be consequential. In his original 1993 paper, Taylor chose to measure inflation using the GDP deflator. As noted, the Taylor rule policy prescription shown in Slide 3 is based on the familiar CPI measure of inflation. For its part, during the past decade, the FOMC has typically focused on inflation as measured by the price index for personal consumption expenditures (PCE), because that measure is less dominated than is the CPI by the imputed rent of owner-occupied housing, and for other technical reasons. As it happens, the choice of inflation measure matters for the interpretation of this episode, as alternative measures gave policymakers somewhat different signals. Notably, core PCE inflation for 2003 was initially reported, in the first quarter of 2004, as having slowed to about 1percent, and it appeared to be on a steep downward trajectory.7 These data heightened concerns about deflation on the FOMC. In contrast, the CPI data released at the same time showed core inflation for 2003 of about 2 percent. In this case, data revisions ultimately raised estimates of PCE inflation for that period, implying that deflation was less of a risk than was thought at the time. But that such revisions would occur could not be known in advance, and policy decisions, of course, must be made based on the information available at the time.
For my purposes today, however, the most significant concern regarding the use of the standard Taylor rule as a policy benchmark is its implication that monetary policy should depend on currently observed values of inflation and output. In particular, the Taylor rule recommendation shown in Slide 3 relates the prescribed policy interest rate to the inflation rate and output gap that correspond to the same quarter in which the policy decision was made.8 However, because monetary policy works with a lag, effective monetary policy must take into account the forecast values of the goal variables, rather than the current values. Indeed, in that spirit, the FOMC issues regular economic projections, and these projections have been shown to have an important influence on policy decisions (Orphanides and Wieland, 2008).
The distinction between current and forecast values does not always matter much, as (for example) high levels of inflation or output today may signal high levels of those variables in the future. However, over the past decade, the distinction between current and forecast inflation has been an important one. On several occasions during this period, surges in energy prices led to increases in overall inflation. According to the standard Taylor rule, whose policy prescription depends on the current value of inflation, these episodes should have led to a significant tightening of monetary policy. However, both the FOMC and private forecasters expected these increases in energy prices to subside–correctly, as it turned out–and therefore did not much adjust their medium-term forecasts for inflation. Consequently, policy was not tightened as much as would have been called for by the standard Taylor rule. Put another way, the standard Taylor rule makes no distinction between increases in inflation expected to be temporary and those expected to be longer lasting. In practice, however, policymakers have responded less to increases in inflation that they expect to be temporary, a reasonable strategy given that monetary policy affects inflation only with a significant lag.
Slide 4 shows the quantitative implications of this point. The actual paths of the policy rate, in blue, and the policy prescription implied by the standard Taylor rule, the dashed red line, are the same as in Slide 3. Also shown, as a dotted green line, is the monetary policy path prescribed by an alternative version of the Taylor rule that replaces the current rate of inflation on the right-hand side with a forecast of inflation over the current and subsequent three quarters. Forecasts are those that were actually made in real time, that is, at the time at which the corresponding policy rate was chosen. For the period through 2004, these forecasts are the staff forecasts (the so-called Greenbook forecasts) that were prepared for each policy meeting. Because Greenbook forecasts for the period after 2004 are not yet publicly available, from 2005 on the forecasts are constructed from the publicly released, contemporaneous projections of FOMC participants, using methods developed by Athanasios Orphanides and Volker Wieland (2008).9 In addition, consistent with the practices of the FOMC, inflation is measured by the PCE price index as was available in real time, instead of by the CPI.10
As Slide 4 shows, the alternative Taylor rule prescribes a path for policy that is much closer to that followed throughout the decade, including recent years. In other words, when one takes into account that policymakers should and do respond differently to temporary and longer-lasting changes in inflation, monetary policy following the 2001 recession appears to have been reasonably appropriate, at least in relation to a simple policy rule.
Which version of the Taylor rule–the standard version, that uses current values of inflation, or the alternative version, that employs inflation forecasts–is the more reliable guide? I have explained my preference for using inflation forecasts rather than actual inflation in the policy rule: Monetary policy works with a lag, and therefore policy decisions must be forward looking. One might still prefer the simplicity of the standard Taylor rule that uses current inflation values. However, note from Slide 4 that a proponent of the standard rule would have recommended that the FOMC raise the policy rate to a range of 7 to 8 percent through the first three quarters of 2008, just after the recession peak and just before the intensification of the financial crisis in September and October–a policy decision that probably would not have garnered much support among monetary specialists. In contrast, Slide 4 shows that the version of the Taylor rule based on forecast inflation (in green dots) explains both the course of monetary policy earlier in the past decade as well as the decision not to respond aggressively to what did in fact turn out to be a temporary surge in inflation in 2008. This comparison suggests that the Taylor rule using forecast inflation is a more useful benchmark, both as a description of recent FOMC behavior and as a guide to appropriate policy.
Although monetary policy from 2002 to 2006 appears to have been reasonably consistent with the Federal Reserve’s mandated goals of maximum sustainable employment and price stability, we have not yet addressed the possibility that accommodative policies–though perhaps appropriate for achieving medium-term inflation and output goals–inadvertently contributed to the housing bubble. I turn now to that question.