In an earlier post , I suggested
...the post 1970s period does seem to be comprised of of two distinct regimes -- the 80s and the 90s. The former was a big improvement over the 1970s, to be sure. But it seems that another step forward in monetary policy occurred in the 1990s as well.
I had in mind both the mean of inflation and the volatility. But reader Alzahr made an interesting observation. Here's the comment in its entirety:
The BLS changed the cpi methodology in 1983 which surely plays a part in inflation risk premium perceptions. Unfortunately, this can't really be judged from the time span in your chart.
Pre-1983 cpi calculations were based on an asset-pricing approach. Post 1983 they went flow-of-services. This meant cpi has not captured house price inflation for 20+ years.
There's also the hedonic adjustment the BLS make, begun in 1998, which flatters cpi. I think this now extends to over 30% of the cpi components (might be wrong on that guesstimate).
Toss these two back in and cpi looks less lovely.
Alzahr has a point. The series I showed in the earlier post -- and that we almost always look at -- is contaminated by methodological changes in the index. Fortunately, the Bureau of Labor Statistics does construct a special research series that is methodologically consistent. The series is explained here, but in essence the BLS constructs the series -- which it calls the CPI research series -- by recalculating past values of the index using today's procedures.
Here's how the methodologically-consistent series matches up with the one we usually see.
The research series is constructed for the period prior to 1978, but it does speak to my basic contention that, for some reason, the moments of the inflation process seem to have changed again in the 1990s. The following picture plots a time series of the five-year moving average of the standard deviation for the rates of growth calculated both the published CPI and the rate of methodologically-consistent CPI.
Again, the impression is that volatility declined in the 1980s, and then again in the 1990s. But interestingly, this "short window" standard deviation appears to have drifted back toward 1980s levels in the past ten years or so, as noted by pgl in a comment on the earlier post.
This is all, of course, very informal -- I'm in the process of assimilating more formal analyses of inflation risk premia and the inflation process, so there will be more to come. But this rough look at the data does raise some interesting questions. Did we just get lucky in the 1990s? And, if so, has the world (the financial world, in particular) caught on?
UPDATE: You can find Alzahr's blog here.
SECOND UPDATE: Waterdog and Calculated Risk have an interesting exchange below, emphasizing the "core" measure of inflation that excludes food and energy components. Here's what the pictures look like:
This latter picture is not so kind to my conjecture that the 1990s represented a distinct regime break in monetary policy. Chris Sims and Tao Zha have been promoting models in which we shift back and forth between a multiple regimes. Here is one of their examples (math alert!). Maybe that's the way to go.