Former senior Fed researcher David Wilcox has a very informative Bloomberg Economics Piece on the difference between the University of Michigan and Federal Reserve Bank of New York measures of inflation expectations (HERE). One real oddity in the NY measure is that the 3- and 5-year inflation expectations series relate to the one-year inflation rate in the third and fifth years. They are effectively the 2-year and 4-year forward 1-year inflation expectation. I did not know that. So, thank you David Wilcox.
The real insight here, though, relates to the 1-year inflation expectation as measured by the NY Fed. Their survey involves asking respondents to assign probabilities to inflation falling into various buckets. (This is true for the 1-, 3- and 5-year rates, but it is most relevant in the case of the 1-year.) NY then calculates a mean of the probability density function (PDF) for each respondent and then reports the median result as the inflation expectation. I assume, without knowing, that the researchers at NY want to steer respondents away from answering with the single most likely rate. They want each respondent to offer their mean, not their mode. And then NY reports the median of those individual means. That is clever, although possibly also a “bridge too far,” as Wilcox observes.
What is super interesting here is that the NY Fed also runs an alternative survey that asks each respondent to report their single inflation expectation directly, just as the University of Michigan does. And this alternative survey has picked up pretty much in line with what the University of Michigan is reporting, as indicated in the chart below. The University of Michigan reading for March is higher than the NY Fed reading for February. But that might be just about a reporting lag, in which case the NY Fed measure is headed higher. Wilcox’s desire to highlight that point explains the white box in the chart.
The point here is that these two surveys are less inconsistent than I had assumed, at least for this current episode. They both show short run inflation expectations starting to drag a bit. An interesting question remains, though. Should we be using the University of Michigan approach, which arguably starts by polling the modes for each individual. Or should we insist on the PDF which effectively forces respondents to cough up a mean, as in the main series produced by the NY Fed. I don’t know. But the practical point here is that there is some evidence that inflation expectations have begun to drag in the real economy.
Fed types, such as John Williams, will occasionally tell us that they prefer market-based measures of inflation expectations over the longer haul. But there are two issues with that. Inflation expectations in the financial markets are not likely to be self-fulfilling unless they are a more accurate measure of inflation expectations in the real economy than surveys of the real economy themselves suggest. Please pause to internalize this, because I think it is an important point that is not widely circulated. If some bond investors think inflation is going up but nobody agrees, then who cares?
Second, there is little reason to suspect that longer-term inflation expectations, wherever they reside, are likely to be self-fulfilling, at least as interpreted through the main analytical model applied at central banks, including the Fed. It is the short run expectation that matters there, as Fed researcher Jeremy Rudd has forcefully and helpfully (in my view) pointed out. In his own piece, Wilcox lets on that he does not see this issue the way I do. That is fine. Maybe he is right, and Rudd and I are wrong. I suspect that Fed types might take personal offense at longer-term inflation expectations being high, simply because that is an objective vote of no confidence in the central bank. But that does not mean that central banks should respond to long-term inflation expectations. Unless they pollute the short term, they are more likely to be occasionally insulting than practically relevant. But Fed types will never tell you that, even if it is correct, which I concede is debatable.
Source: David Wilcox, via Bloomberg
