Inflation Detour: Trimmed Mean PCE
Today’s release by the Federal Reserve Bank of Dallas of October’s Trimmed Mean Personal Consumption Expenditure gives us a chance to check this “alternative measure of core inflation.”
The clearest thing is that it does what the FRB Dallas intends: generally reduces the measure of inflation:
For the graphic above, any value above the line shows where the 12-month Average of the Trimmed Mean PCE is greater that the Annualised CPI for that same period. With few exceptions, those points are places where the actual CPI is negative for the period. (Note also that all of periods where CPI is over 5.0-5.5% are below the line. While the 12-month average of Trimmed Mean PCE has a maximum of 8.7%, while CPI reaches slightly over 14.75% in the same time period.)
So the natural next step is to compare it to a measure of Consumer Sentiment. Let’s do that below the fold
Comparing Trimmed Mean PCE to the University of Michigan Index previously referenced:
Again, the preponderance of data points are to the right of the line, indicating that the Michigan Consumer Inflation Expectations is higher than the monthly Trimmed Mean PCE. But there is much more balance: the largest cluster of Expectations Dominance is between 2 and 4%; that is, periods of normal inflation.
The two measures correlate rather well with each other (86.13%) while a simple fitted regression that excludes a constant term has an adjusted R-Squared of 94.1% and yields MICH = 1.0416*Trimmed Mean PCE.
Trimmed Mean PCE may well understate inflation, but it appears to compare fairly well with what people think of when they think about inflation.
Slightly off topic. With “Trimmed Means Analysis” the proponents of it are whacking the outliers to improve Kurtosis and minimize Skewness of the curve rather than explain the outliers? One of the issues in the late nineties and around 2005 was the failure of the Fed to adequately read the economy and take appropriate actions which “may” be traced back to the hedonics used to calculate inflation. I kind of wonder whether we are failing to capture the indicators of a down turning or an up turning economy and are prone to the “awww crap” knee jerk reactions of the Fed when they realize they are late to the economic party. I do not like the transformation of the data curve by eliminating data points to achieve a particular outcome.
This also brings up another thought. As more and more items become globalized by being outsourced, we lose control of the pricing mechanism. If we do not buy the overseas goods, it can be sold elsewhere for a similar price. In the end, the Fed may lose the ability to control inflation with Fed Rates.
The FED is not trying to use the trimmed mean to estimate risk; they are trying to exclude outliers that leverage the mean. There are plenty of good “common sense” reasons for doing this. First, as Ken’s analysis suggests, this is what actual consumers seem to do and in formulating inflation expectations it’s always a good idea to try and mimic what actual consumers. Another reason for excluding outliers is that the CPI includes imports, which are subject to extreme overshooting and undershooting.
Ken’s analysis did surprise me though. I would have expected consumers to react more strongly to outliers because people seem to misremember things. For example, people were still thinking of $4 gasoline long after gas prices had crashed through the $3 floor.
“The clearest thing is that it does what the FRB Dallas intends: generally reduces the measure of inflation:”
Is this what the Dallas Fed says is it’s goal? I’m just asking, bacause I don’t remember that from the release or the Dallas Fed’s explanation of its motive for creating the index. My recollection is that the Dallas Fed recognized that outliers could skew a series far from trend. A look at the performance of headline and core inflation measures over the period of rapidly rising and rapidly falling commodity prices shows this very clearly. My recollection is that the Dallas Fed, and the Cleveland Fed in its reworking of CPI, aim at another version of core inflation. But correct me if I’m wrong.
Otherwise, my impression is that Ken has introduced his own bias regarding semi-official inflation measures as if that bias were fact.
I’d suggest (but will not assert as fact) that the Fed, charged with maintaining price stability, wants to know as much as possible about the nature of price movement, so slices and dices all over the place. That could, hypothetically, help Fed officials get past their own biases about the pace of inflation when making policy. That seems like it could be a good thing.