How the Fed Works Today

Robert J. Barro, Harvard University

The triumph over inflation in the last 25 years has been one of the great successes of economic policy. In 1980-81, the inflation rate for the U.S. GDP deflator exceeded 10%. Fed chair Paul Volcker, supported in 1981-82 by President Ronald Reagan, committed to using contractionary monetary policy to bring inflation down. The main instrument was high nominal interest rates—the Federal Funds rate reached 20% in 1980 81.

In 1979-81, Volcker seemed to be focusing on the control of monetary aggregates as the instrument for guiding policy. Thus, it may be that a desire to contract monetary aggregates led the Fed in the early 1980s to its policy of sharply raising nominal interest rates in response to the high inflation that built up through the 1970s. However, the source of the high nominal interest rates in 1981-82 is probably a secondary issue. The key matter is that this regime proved remarkably successful, and inflation receded to 3% by 1983. Most importantly, by sticking to high rates despite the 1982 recession—the worst of the post-World War II period—the Fed established a reputation for raising nominal interest rates when inflation threatened and lowering rates only when inflation was tame.

In addition to responding to inflation, the Fed has tended to raise nominal interest rates when the economy is strong, and vice versa. This reaction turns out not to be to real GDP growth but rather to the labor market—rates rise when employment growth is high and when the unemployment rate is low and falling. Another way to say this is that the Fed does not tend to raise interest rates when high GDP growth reflects strong growth of labor productivity—increases in output per worker or per worker-hour. Unlike the response to inflation, it is unclear whether the response of Fed policy to the real economy, that is, the labor market, is useful.

The Fed adjusts the Funds rate only gradually in response to changes in inflation and other variables, rather than moving the rate more quickly to a desired value. That is, the Fed seems to include in its policy objective a penalty for adjusting short-term nominal interest rates a lot from month to month. It is unclear whether this gradualist method is useful. One reason the benefits are unclear is that smoothing of short-term nominal interest rates does not necessarily smooth the more important longer-term rates. That is, if the financial markets anticipate a period of changing short-term interest rates, the longer-term rates would react immediately to this expectation.

The Fed was an innovator in its monetary policy of the 1980s, but many other central banks have since become more sophisticated than the Fed. Beginning with New Zealand in 1989, a number of central banks successfully designed and followed formal rules for adjusting nominal interest rates to accord with an objective of inflation targeting. Examples from the early 1990s are Canada, the United Kingdom, Australia, and Sweden. The number of inflation-targeting central banks has grown over time to around 20, depending on the exact definition of whether a central bank has in place an inflation-targeting regime. Table 1 shows a customary classification. Economies that are notable exceptions from inflation targeting, aside from the United States, are the European Union (which may have an implicit inflation-targeting system) and Japan.

Despite increases in transparency, the U.S. system of monetary policy remains somewhat opaque and still relies a lot on the judgment and credibility of its chair, Paul Volcker from August 1979 to August 1987, Alan Greenspan from August 1987 through January 2006, and Ben Bernanke since February 2006. The occasion of a new chair is a good time to consider a shift to a more formalized procedure—particularly since credibility is more of an issue now that the chair is no longer a policy icon. Therefore, it is fortunate that Bernanke—at the center of policy through his previous positions as chair of the President’s Council of Economic Advisers and a member of the Fed’s Board of Governors—is also one of the main contributors to research on inflation targeting.

My view is that, although the Fed under Greenspan and so far under Bernanke has not formally embraced inflation targeting, the Fed’s reaction function for setting short-term nominal interest rates—specifically, the Federal Funds rate—comes close to inflation targeting in practice. Moreover, when Bernanke expresses uncertainty about the course of future Federal Funds rates, he is not expressing a lack of confidence about the way that the Fed will respond in the future to the state of the economy. Rather, he is saying that the Fed cannot perfectly forecast the variables to which it typically reacts, notably inflation and the strength of the labor market. I will illustrate the nature of the Fed’s regular policy response by isolating the form of the reaction function that has prevailed in practice and by using this analysis to forecast future Federal Funds rates.

Where are U.S. Interest Rates Headed?

The Federal Reserve raised the Federal Funds rate at 17 consecutive Federal Open Market Committee (FOMC) meetings from June 2004 to June 2006—going from the extremely low value of 1% to 5-1/4%. Then rates were held fixed at the next two meetings (August and September 2006). A key question (asked at the time of the conference in October 2006) is where will the Fed go next in rates?I have been using three approaches to attempt to answer this question. First, I use a simple reaction-function model to predict what the Fed will do. This approach is straightforward for the very short run but requires forecasts of right-hand-side, explanatory variables, such as inflation and employment growth, for longer-term projections. Second, I use the Chicago Board of Trade (CBOT) futures contract on the Federal Funds rate to assess the views of financial markets. Third, I look at the term structure of yields on U.S. Treasury bonds to get another way of inferring financial-market projections of future short-term interest rates.

A Reaction Function for the Federal Funds Rate

At least since August 1987, when Alan Greenspan took over as chair, the Fed’s monetary policy has approximated a formula in which the Funds rate reacts to several variables. First, rates tend to rise in response to inflation, measured particularly by a broad index, such as the GDP deflator. Although popular commentary often focuses on the deflator for personal consumer expenditure (PCE), the GDP deflator turns out to have a little more explanatory power (and the PCE deflator has no incremental explanatory power). There is a moderate increase in explanatory power from the inclusion of inflation measured by the consumer price index (available monthly in contrast to the quarterly series on the GDP deflator).

Second, the Fed tends to increase the Federal Funds rate when the economy is strong, gauged by the labor market—especially by the growth rate of payroll employment but also by the levels and changes in the unemployment rate. Real GDP growth, per se, seems to play no independent role. There is also some reaction to the stock market, particularly important during major crashes. I have not found any significant, separate reaction to exchange rates. I use a regression approach to estimate a simple reaction function—amounting to a glorified Taylor Rule (named for John Taylor)—that describes the Fed’s behavior with respect to the Federal Funds rate from August 1987 through September 2006. The result, based on an ordinary-least-squares (OLS) regression technique with monthly data, is

(1)  DFt = 0.0003 + 0.253∙DF_t-1 – 0.032∙F_t-2 + 0.060∙DP_t-3 + 0.008∙DCPI_t-2
                   (0.0007)    (0.061)          (0.006)            (0.015)               (0.004)

+ 0.054∙DL_t-1– 0.245∙DU_t-1 – 0.027∙U_t-2 + 0.0014∙DSP_t-1,
(0.009)         (0.077)          (0.012)            (0.0004)

standard errors of coefficients in parentheses, s.e. of regression = 0.00145, R2 = 0.51.

In equation (1), F_t is the Federal Funds rate (average of daily data) for month t, and the dependent variable, DF_t, is the change in the Funds rate between months t and month t-1. The right-hand side variables in equation (1) are constructed so as to be available prior to month t. For example, DP_t-3 is a moving average of the inflation rate for the GDP deflator computed over the previous three quarters. DCPI_t-2 is the inflation rate for the overall CPI index for urban consumers between months t-2 and t 3. DL_t-1 is a moving average of the monthly growth rate of payroll employment for the total non-agricultural economy, computed between months t-1 and t-4. DU_t-1 is the change in the civilian unemployment rate between months t-1 and t-2, and U_t-2 is the level of the unemployment rate in month t-2. DSP_t-1 is a moving average of the growth rate of the S&P 500 stock-market index, computed between months t-1 and t-5.

In equation (1), the positive coefficient on the first lagged change of the Funds rate, DF_t-1, picks up the time-averaging in the data (that is, the calculation of Ft as an average of daily data). The significantly negative coefficient on F_t-2, -0.032 (s.e. = 0.006), indicates that the Funds rate tends to revert to a level determined by the other right-hand side variables in the equation.

The response to inflation from the GDP deflator, 0.060 (0.015), is significantly positive. This inflation rate was measured as an average over three prior quarters—entering two separate values of the inflation rate does not significantly enhance the explanatory power. The CPI inflation rate with a two-month lag has a marginally significant additional positive effect, 0.008 (0.004). Surprisingly, the first monthly lag of CPI inflation has essentially a zero effect if added to equation (1).

The estimated relation in equation (1) shows a clear positive response of the Funds rate to lagged employment growth, DL_t-1—the coefficient is 0.054 (s.e. = 0.009). This employment growth rate was entered as an average over the four prior months—entering three separate values for the employment growth rate does not significantly enhance the explanatory power. There is also a significant tendency to cut rates if the unemployment rate over the two prior months is high and rising—the coefficients are 0.027 (0.012) on U_t-2 and -0.245 (0.077) on DU_t-1.

Given the indicators for labor-market activity (employment growth rates and unemployment rates), there is no tendency of the Federal Funds rate to respond to real GDP growth. For example, if the real GDP growth rate over the prior two quarters is added to equation (1), the estimated coefficient is 0.0003 (0.0061). This result means that, for given employment growth, higher real GDP growth does not induce the Fed to raise interest rates. That is, the Fed does not tend to raise interest rates in response to higher growth of labor productivity.

Equation (1) shows a significantly positive response (coefficient of 0.0014, s.e. = 0.0004) of the Funds rate to growth of the Standard & Poor’s 500 stock-market index, DSP_t-1. This growth rate was computed as a moving average of the growth rates over the five prior months—entering these growth rates separately does not contribute significantly to the explanatory power.

As an example, over the period since August 1987, the largest stock-market decline (computed over a five-month interval) was at the rate of 0.97 per year for the five months ending in November 1987. This collapse was associated with the global stock-market decline of October 1987. The estimated coefficient of 0.0014 on DSP_t-1 implies a cut in the Funds rate by 0.0014—that is, somewhat more than half of the common quarter-point rate cut. In fact, the Funds rate fell by much more—by one-half of a point from September-October 1987 to December 1987. (Chairman Greenspan has often been applauded for this aggressively expansionary policy in the face of the stock-market crash.)

As another example, the stock-market decline over the five months leading up to September 2001 (including the market response to the September 11th attacks) was at the rate of 0.56 per year. Again, the rate cut—by more than one full percentage point from August to September—was much more than predicted. However, since the Fed was already on a path of cutting rates in response to the weak employment market, it is hard to attribute all of these rate cuts to the stock market.

Since the estimated coefficient on the stock-market variable in equation (1) reflects the average partial effect on the Funds rate, there are other examples in which the response of the Fed to the stock market was weaker than that implied by the estimated coefficient. For example, rates did not rise in response to the strong stock market in early 1991, late 1996, mid 1997, and late 1998. Possibly, the analysis of Fed response to the stock market would be improved by distinguishing reactions to weak markets from those to strong markets. A more interesting analysis, not attempted here, is whether the Fed’s reactions to stock-market fluctuations actually serves to lessen the probability of severe stock-price declines.

Using the Fed’s Reaction Function to Forecast Federal Funds Rates

Equation (1) can be used directly to forecast the Federal Funds rate for the next month—for example, for October 2006 based on data through September 2006. The result in early October 2006 was that the Funds rate seemed unlikely to change in October (when an FOMC meeting was scheduled toward the end of the month).

More distant forecasts of the Funds rate require projections of the right-hand-side variables in equation (1). One way to proceed, not undertaken here, would be to construct and estimate a full system (a vector autoregression model) that could be used to predict simultaneously future values of all the independent variables, along with the Federal Funds rate, which is the dependent variable in equation (1). This system would allow for effects of the Federal Funds rate on right-hand side variables, such as inflation rates, as well as for the responses of the Funds rate to inflation and other variables, as in equation (1). In practice, the two most important determinants of future Federal Funds rates are future values for inflation (particularly for the GDP deflator) and employment growth. Therefore, in the present analysis, I focus on ways to project future values of these two variables.

The remainder of Section B, heavily dependent on graphs and tables, is available in this document’s PDF version, which may be downloaded via the link at the top of this page.

Federal Funds Rate Forecasts from the CBOT Futures Market

My forecasts for the Funds rate can be checked against predictions implied by the Federal Funds Futures contracts traded on the Chicago Board of Trade (CBOT). On October 6, 2006, the CBOT market predicted no change in the Funds rate for the months corresponding to the next two FOMC meetings: October and December 2006. The implied probability of a quarter-point rate increase (to 0.055) at the January 2007 meeting was about 20%. Hence, these predictions accord reasonably well with those from my baseline scenario in Table 3, column 1. The difference is that my projection gives higher probability to a quarter-point rise to 0.055, with this chance shared roughly equally between the December 2006 and January 2007 FOMC meetings.

Implications of the Treasury Yield Curve

The yield curve for U.S. Treasury securities also implies a forecast for Federal Funds rates. In early October 2006, when the Funds rate was 5.25%, the 5-year nominal Treasury yield was 4.63% and the 10-year yield was 4.69%. Thus, unlike the financial environment in 2002-04, the yield curve was inverted, with the short rate well above the long rates (for 5-year and longer maturities). The more typical pattern is an upward-sloping yield curve. For example, from July 1954 to September 2006, the average of the Federal Funds rate was 4.81%, the average 5-year Treasury yield was 5.86%, and the average 10-year yield was 6.24%. When the short rate (Federal Funds rate) is much higher than the 5-year (or 10-year) rate, the bond market is predicting that the Funds rate will fall over time—in particular, sometime over the next five years.

Suppose, for example, that the Federal Funds rate returns fairly soon to a normal spread from the 5-year rate—that is, the spread of 1.05% per year seen from July 1954 to September 2006. In this case, the Funds rate would have to fall “fairly soon” from 5.25% in September 2006 to 3.5-3.75% (since the 5-year yield in September 2006 was 4.63%). This prediction accords in a rough way with the pattern shown by my baseline scenario in Table 3, column 1. Thus, my conclusion is that consistent predictions for Federal Funds rates emerge from three approaches: the reaction-function for the Federal Funds rate given by equation (1), which led to the results shown in Table 3, column 1; predictions over the near term implied by CBOT futures contracts on the Federal Funds rate; and predictions embedded in the term structure of nominal yields on Treasury bonds.

Inflation Stability and Economic Growth

Over the last 20 years, the Federal Reserve’s “formula” for adjusting nominal interest rates has helped to promote economic stability. In particular, the upward reaction of interest rates to inflation has been critical in achieving low and stable inflation rates. A deeper question is how much this success of monetary policy—in the United States and elsewhere—contributes to broader macroeconomic objectives, such as strong economic growth.

The best empirical evidence—albeit imperfect evidence—comes from examination of data from the experiences of many countries observed over time. Cross-country growth relations show that higher inflation goes along with lower growth of real per capita GDP. Moreover, there is some evidence that the negative association between inflation and growth reflects causation from high inflation to low growth, rather than the reverse. However, the main statistical evidence on the inflation-growth relationship comes from the observation that countries with very high inflation have low economic growth.

The empirical results, exemplified by Figure 8, do not clearly demonstrate that inflation matters a lot for growth when inflation is moderate, say, less than 10% per year. And the overall point estimates imply that an increase in the average inflation rate by 10 percentage points—a very large change—would reduce the average growth rate of real per capita GDP by only around 0.2% per year. On the other hand, there is no evidence that countries gain anything in terms of higher growth by allowing the inflation rate to be higher, even in the range of low inflation rates. Plus a change by 0.2% per year matters a lot for standards of living in the long run.

The bottom line is that low and stable inflation contribute something to economic growth, and surely the achievement of this inflation stability has been highly worthwhile. However, one should not overestimate the growth effects of this accomplishment. The cross-country history suggests that other factors—such as strong institutions that foster property rights and limit corruption, well-functioning domestic and international markets, and efficient systems of schooling and health—are more significant than macroeconomic stability, including low inflation. Also, as is clear from the record of low inflation and anemic economic growth in Switzerland from the 1970s through the 1990s, inflation stability does not ensure a robust economy. Nevertheless, central bankers of the last 20 years can justifiably congratulate themselves on playing more than their part in fostering strong economic performance.