Pages

Wednesday, September 12, 2007

A Simple Metric for the Stance of Monetary Policy: Nominal GDP Growth Rate minus the Federal Funds Rate

I wrote up a note that briefly examines the role monetary policy may have played in the housing boom of 2003-2005. In that note I use a simple metric to measure the stance of monetary policy: the nominal GDP growth rate minus the federal funds rate. This metric is popular outside of academia and based on the idea that if the cost of borrowing is too low relative to the growth rate of the American economy, then excessive leverage and investment is encouraged and vice versa. The Economist describes it this way: "One way to interpret this [measure] is to see America's nominal GDP growth as a proxy for the average return on investment in America Inc. If this return is higher than the cost of borrowing, investment and growth will expand."

Using this metric, a neutral monetary policy would be one where the federal funds rate never wondered too far from the nominal GDP growth rate. Here, I calculate this metric as the year-on-year growth rate each quarter of nominal GDP less the nominal federal funds rate for the quarter. Since the early 1980s the average difference between these two series, called the policy rate gap, was about 0.50%. From the 1960s to the early 1980s the policy rate gap average almost 2.00%. If, in fact ,this is a reasonable measure of the stance of monetary policy, these two averages shed some light into the 'Great Moderation' debate in macroeconomics.

I used this metric in an earlier post where I looked at the role the Federal Reserve may have played in the housing sector. Now, I want to see how well it predicts the NBER recessions. To begin, take a look at the figure below which plots the policy gap rate for 1956:Q1 through 2007:Q2 and shades in those quarters that fall under the NBER Recession.
This figure shows that every NBER recession was preceded by a negative value for the policy gap rate, but not every negative policy gap rate was followed by a NBER recession. This problem also arises when using the yield curve spread to predict recessions, but my impression is that it is not as pronounced. This information can be used in a probit model to estimate the probability of a recession. Specifically, the policy rate gap is regressed upon a NBER recession dummy variable that is 1 if a recession is present and 0 otherwise. Below are the results from two forms of this probit regression. The first recession simply regresses the contemporaneous value of the policy gap rate on the recession dummy. The second recession regresses the 4 lags of the policy gap rate on the recession dummy.




These results look promising, but still need refining (e.g. need to account for serial correlation). Nonetheless, I took a first stab at the data by taking these estimates, the actual policy gap measure, and then plugging it into and standard normal cumulative distribution to get the following figures. These figures show the probability of a recession given the policy rate gap:

While these initial results look promising there are again some notable misses such as 1994. Of the two models, the non-lagged probit model appears to do better with the misses (compare 1984 in both models). Overall, the policy rate gap appears to be a promising way--in need of further refinement--to measure the stance of monetary policy.

Update

In response to a commentator's suggestion, I have enlarged the probit results and the last two graphs to make them more readable. I also went ahead and redid the analysis using a longer time series.

6 comments:

  1. Thoroughly fascinating posting. From your first chart, it appears that The Fed has plenty of room to cut up to one percent off the fed funds rate.

    The probit charts shows that at about 35%, the chance of a recession occuring is about 90%.

    As I recall in the late 80's and in 1994, interest rates were creeping up. Today the bias is down, so I would hazard a guess that the recession will not occur.

    ReplyDelete
  2. The last three tables/graphs are too small to read. (Not that I'd know what the numbers in the table meant anyway.)

    ReplyDelete
  3. Anonymous: I hope you are right concerning the recession...I moved to Texas from Michigan this summer and am still trying to sell my house.

    ReplyDelete
  4. Brian:

    Yes, they are small and I am not sure how to fix them. Blogger is free and usually does everything I want, but sometimes I find it challenging to use.

    ReplyDelete
  5. Have you compared the forecasting power of your simple metric with some of those other "real-time" forecasting methods like yield curve ?

    ReplyDelete
  6. Paul:

    No, I have not compared the two, but I have been thinking about doing it. I would also like to see if using both measures in a forecasting model reduces the forecast error. Any suggestions?

    ReplyDelete