This is a follow-up to Tim Duy’s most recent Fed Watch. A recurring theme in the interpretation of monthly or quarterly data, and something noted in Tim’s comments, is to be careful about placing too much weight on the data for a particular month or quarter as they are released. First, the data are often revised and the revisions can have a substantial effect on the interpretation. However, even after the data are revised information concerning trend movements in many series measuring aggregate activity, prices, interest rates, and monetary aggregates is difficult to extract due to their high variability. Because of this, placing too much weight on the data for any one month or quarter can give a false impression about the underlying trend. As an example of this, here’s a graph of the monthly growth rate in industrial production (IP) over the last ten years. Superimposed upon the actual series is the smoothed value of IP (the average value over five months centered on the actual value).
Notice how noisy the monthly data are relative to the underlying value of the trend. This shows how difficult it is to extract information about the trend from any particular monthly release of the series (and these data have been revised, the variance is even higher with unrevised data).
Why is the trend important? Here is a graph of the smoothed value of IP and the target value of the federal funds (FF) rate:
The graph shows several important features of the relationship between the target FF-rate and the growth rate of IP (deviations of IP from a measure of its natural rate are highly correlated with deviations of unemployment from the natural rate of unemployment as estimated by the CBO so the use of IP as the measure of aggregate economic activity is not critical to the conclusions). First, once the smoothed value is extracted from IP growth it does appear to be correlated with the FF-rate. More sophisticated econometric work confirms this association.
Second, and importantly, in recent years policy has been persistent in the sense of only changing after output growth has clearly deviated from its recent history. Consider the period just before the May 2000 label in the graph. In late 1999 the Fed began raising the FF-rate (the black line measured on the right-hand axis) persistently. Even after the growth in industrial production (the blue line measured on the left-hand axis) began falling, the Fed continued raising the FF-rate for a time period. Then, around May 2000 the FF-rate levels out as the Fed pauses to assess the situation. Policy does not change until after the IP growth rate falls for several months and then becomes negative. At this point the Fed begins reducing the FF-rate rapidly until it bottoms out just after May 2003. Thus, there is a lag between the decline in IP growth and the change in policy as the Fed patiently assesses the data over several time periods to make sure there is an actual change in the underlying trend and not just a temporary blip in the data. A similar delay in changing the course of the FF-rate is in evidence when the FF-rate begins increasing around May 04 in the diagram. The Fed does not change its policy immediately, and only begins increasing the FF-rate after it observes several consecutive months of positive growth. This shows, as Tim noted, that the Fed does not change course easily and pays little attention any particular blip in the data. The Fed operates deliberately, and changes policy only after there is clear evidence that there has been a change in the underlying trend rate of output growth or, more generally, inflation.