“In other words, even with a really strong recovery (which almost nobody expects), the Fed should keep rates on hold for at least two years… It’s the people saying that the Fed should start tightening in the near future who are inventing some kind of new, unspecified framework to justify their views.” (Krugman 2009).

With the deepest US recession in the post-World War II era coming to a close, attention is turning to reversing some of the emergency policy measures taken to fight the crisis (Blanchard et al. 2010). This is reflected in ongoing discussions of a budget freeze for the Federal government and the exit strategy of the Federal Reserve. A key element of the exit strategy is when to start raising interest rates.

One indication of when interest rates are expected to start rising comes from the Fed Funds Futures market. The latter is pricing in Fed rate hikes that will take the Fed funds rate to around 0.75% by the end of 2010 (Rosenberg 2010). Similarly, professional forecasters are predicting that the 3-month T-bill rate (which moves almost one-for-one with the Fed funds rate) will be nearly 1% by the end of 2010 and over 2% by the end of 2011 (Blue Chip Economic Indicators 2010). The market view is thus that the zero bound on interest rates will stop binding by the end of this year. An alternative approach is to make use of Taylor (1993)-type rules, which model the Fed’s interest rate decisions as a function of deviations of macroeconomic variables from their target rates. Paul Krugman, Michael Rosenberg and others have recently employed this approach to predict the future path of interest rates. They conclude that interest rates are likely to remain at zero until the end of 2011. Apart from contrasting sharply with the financial market’s expectations, this finding suggests an extraordinarily long period of expansionary monetary policy which could conceivably generate a new bubble in asset prices and further complicate the Fed’s exit strategy.

# An enhanced Taylor rule

We argue that there is, in fact, no inconsistency between forecasts based on the Fed’s historical reaction function and the market’s forecast of future interest rates. They key to reconciling the two comes from the fact that the specific Taylor rule used by Krugman and Rosenberg is not an adequate description of how the Fed has been setting interest rates since Paul Volcker became chairman, as documented by much of the empirical literature on estimated Taylor rules (Orphanides 2003, Boivin 2006, Coibion and Gorodnichenko 2009). The primary distinction is that recent estimated reaction functions emphasise the fact that the Fed responds strongly to output growth, while the traditional Taylor rule is expressed in terms of the output gap – the deviation of real GDP from potential GDP. This often unnoticed distinction matters significantly for the timing of interest rate decisions because the output gap evolves slowly while the growth rate of GDP changes much more rapidly.

To illustrate this, we estimate a standard reaction function from the empirical literature that allows for forward-looking behaviour on the part of the central bank, interest smoothing, and endogenous responses to expected inflation, the output gap, and output growth. We then generate fitted and forecasted values of the interest rate from this empirical specification, as well as from a specification which does not allow for an endogenous response to the growth rate of output (as assumed by Krugman and the original Taylor (1993) rule (see Technical appendix)).

# Contrasting predictions

The results from both specifications, as well as the actual Fed funds rate, are plotted in Figure 1. Both Taylor rules imply that the Fed funds rate should currently be below zero, so that the zero-bound is binding. The estimated reaction function excluding output growth yields the same prediction as made by Krugman, namely that interest rates should stay below zero well past 2011. This reflects the fact that inflation is expected to remain low and that the output gap will only slowly be closing. The forecast made by the rule which allows for a response to output growth, however, predicts that interest rates should start to rise around the last quarter of this year or the first quarter of next year, reflecting the fact that output growth has already accelerated. In fact, were the zero-bound not binding, our estimated Taylor rule indicates that interest rates would have started to rise in the third quarter of last year, which coincides with evidence that the recession likely ended in the summer of 2009.

**Figure 1.** Two Taylor rules

We should emphasise that this result is not driven by unduly optimistic forecasts of future growth in real GDP. The consensus forecast from the Blue Chip Economic Indicators (2010) used in our interest rate predictions is for a growth rate of real GDP of 2.8% in 2010 and 3.1% in 2011. The 2010 forecast is almost identical to forecasts from the Administration and the Federal Reserve (Romer 2010), while the 2011 forecast is distinctly more pessimistic than the 4.3% growth rate assumed by the Administration or the 3.4%-4.5% range predicted by the Federal Open Market Committee.

# Conclusion

The Taylor rule with a response to the growth rate of output generates interest rate predictions which are quite consistent with financial and professional forecasts of future interest rates. The key to this consistency is that the historical behaviour of the Fed (since Volcker) has been characterised by a strong response to output *growth*, which returns to normal levels quickly, rather than solely the output gap, which recovers only gradually. Hence, there is a broad consensus that, given current information and barring political or populist pressures, one can reasonably expect the Federal Reserve to start raising interest rates toward the end of this year in its attempt to balance the risks of higher inflation against prolonging the current economic downturn.

# References

Bernanke, Ben S (2010), “Monetary Policy and the Housing Bubble [1],” speech given at the annual meeting of the American Economic Association, January.

Blanchard, Olivier, Giovanni Dell’Arricia, and Paulo Mauro (2010), “Rethinking macro policy [2]”, VoxEU.org, 16 February.

Boivin, Jean (2006), “Has U.S. Monetary Policy Changed? Evidence from Drifting Coefficients and Real-Time Data [3]”, *Journal of Money, Credit and Banking*, 38(5):1149-1173.

Coibion, Olivier and Yuriy Gorodnichenko (2009), “Monetary Policy, Trend Inflation and the Great Moderation: An Alternative Interpretation [4]”,* American Economic Review*, forthcoming

Krugman, Paul (2009), “When should the Fed raise rates? [5]”, *New York Times*, 11 October.

Orphanides, Athanasios (2003), “Historical Monetary Policy Analysis and the Taylor Rule [6]”,* Journal of Monetary Economics*, 50(5):983-1022.

Romer, Christina (2010), “The Economic Assumptions Underlying the Fiscal 2011 Budget [7]”, 1 February.

Rosenberg, Michael R (2010), “Fed Funds Rate Outlook – A Taylor Rule Perspective”, Bloomberg Financial Conditions Watch, 27 January.

Taylor, John B (1993), “Discretion versus Policy Rules in Practice [8]”, *Carnegie Rochester Conference Series on Public Policy* 39:195-214.

# Technical appendix

Specifically, our regression equation follows Coibion and Gorodnichenko (2009) and is given by

which relates the Fed funds rate (the current interest rate set by the central bank *i*_{t}) to the Fed’s forecast (denoted by* F*_{t}) of macroeconomic variables at horizon *h*, such as inflation (*π*_{t}), the output gap (*x*_{t}), and the growth rate of real GDP (*gy*_{t}). Our estimation procedure can be summarised as follows. First, we estimate the equation above using the Fed’s Greenbook (GB) forecasts from 1982Q3 until 2003Q4, when the GB forecasts are no longer available. We then generate fitted values of this reaction function and out-of-sample forecasts using the Survey of Professional Forecasters’ mean forecasts of inflation and GDP growth in place of the GB forecasts until 2009Q4 then use forecasts from the Blue Chip Economic Indicators out to 2011Q4. Measures of the output gap are constructed as deviations of forecasted real GDP relative to real-time estimates of potential GDP from the Congressional Budget Office. Following Coibion and Gorodnichenko (2009), we let the Fed respond to expected inflation over the next two quarters as well as the current forecast of the output gap and output growth in the current quarter.