Bruce Hall seems to think that calculating 18-month inflation rates (whether on an annual basis or not) is a good thing. He. She he Fine. As long as you don’t do it with non-seasonally adjusted data. If you do, you should be really clear. An illustrative example of the Consumer Price Index is below.
Figure 1: The 18-month annual rate of inflation for the seasonally adjusted urban CPI for all consumers (blue), and for the seasonally unadjusted CPI for all urban consumers (brown), calculated using the variance of logarithms. The National Bureau of Economic Research has peak-to-trough recession dates highlighted in grey. Source: BLS via accounts by FRED, NBER, and the author.
My vision the last time I remembered the numbers 20-400 (instead of 20-20) was uncorrected, but even I can see the two strings are fundamentally different at different times.
In case the CoRev reader is going to accuse me of data manipulation, let me point out that I use the FRED series CPIAUCSL for seasonally adjusted CPI, and CPIAUCNS for non-seasonally adjusted CPI. All other calculations are pretty straightforward looking at the formulas in the legend box – unless one is unfamiliar with natural records, or doesn’t trust them. (Here’s Jim Hamilton’s post about using registers and register differences, if you don’t trust Menzie Chinn either because of his name or his worldviews.
I must say, 18 months is a rather strange choice. 1 month, 3 months, 12 months, 2 years, 5 years, ok. 1.5 years, well, one must have a reason.