reader Eric Poole comment on this number (at this postal) wrote:
Assuming all inflation forecasts in the chart above are one-year forecasts, do we have any inflation expectations data for a shorter time horizon (eg 6 months)?
The above is a fancy way of asking the question: are financial markets and professional forecasters really bad at predicting inflation?
Taking a look at the chart above, it seems like a pretty good argument adaptive expectations Promote inflation expectations of economic entities.
I agree with this point of view in a year’s time. Adaptive expectations are something I learned as an undergraduate in my first macro course in 1982 (with uniform autoregressive coefficients, adaptive methods supporting accelerationist assumptions).I also use this as initial point For pedagogical reasons, I’m talking about dynamics in my macro course this semester.However, I quickly moved on to a more complex approach due to the Coibion et al. (Nippon Economic Academy 2018).
In fact, adaptive expectations do not seem to be the best way to assess informally if we are to evaluate these measures in terms of predicting rather than explaining behavior. To see why, see this graph of forecast errors.
figure 1: Year-over-year CPI inflation forecast error due to lagged year-over-year inflation (black), professional forecaster survey median (sky blue+), and Michigan consumer survey (red), all expressed as a percentage. Dates of peak-to-trough recessions as defined by NBER are shaded in gray. Source: BLS via FRED, Philadelphia Fed, University of Michigan Consumer Survey via FRED, NBER, and author’s calculations.
Applying a common sample (SFP reports quarterly from 1982, Michigan from 1979), we find that adaptive expectations have a smaller average error, but a much larger RMSFE – 1.86 versus 1.59 for SPF and 1.54 for Michigan .
Note that this pattern of RMSFE is even more prominent if we ignore the Biden period; then the adaptive method yields an RMSFE of 1.68 compared to 1.19 for SPF and 1.37 for Michigan.
