Bruce Corridor appears to suppose calculating 18 month inflation charges (both annualized or not) is simply nice. It is simply nice. So long as you don’t do it utilizing not-seasonally-adjusted information. For those who do this, you actually must be clear. Illustrative instance for CPI under.
Determine 1: 18 month annualized inflation for seasonally adjusted CPI-urban all customers (blue), and for not-seasonally-adjusted CPI-urban all customers (tan), calculated utilizing log differenes. NBER outlined peak-to-trough recession dates shaded grey. Supply: BLS by way of FRED, NBER, and writer’s calculations.
My imaginative and prescient final time I recall the numbers is 20-400 (quite than 20-20) uncorrected, however even I can see the 2 sequence differ considerably at totally different instances.
In case reader CoRev goes to accuse me of manipulating the info, let me word that I take advantage of FRED sequence CPIAUCSL for seasonally adjusted CPI, and CPIAUCNS for seasonally unadjusted. All different calculations are fairly simple trying on the formulation within the legend field — until one is unacquainted with using pure logs, or distrusts them. (Right here is Jim Hamilton’s publish on using logs and log-differences, should you don’t belief Menzie Chinn both due to his identify or worldviews.
I have to say, 18 months is kinda odd selection. 1 month, 3 month, 12 month, 2 years, 5 years, okay. 1.5 years, effectively, one ought to have a purpose.
