Adjusting Median and Trimmed-Mean Inflation Rates for Bias Based on Skewness – A Review

To keep in mind

Research has shown that the Cleveland Fed Median PCE index and the Dallas Fed Trimmed mean index are biased indicators of “core” inflation. The bias is due to the time varying nature of the skewness of price changes. Correcting for this bias is very important because over some sub-samples it can be larger than 1 percentage point. Therefore, relying on the unadjusted Cleveland Fed median PCE index (or on the Dallas Fed Trimmed-mean index) can be seriously misleading.

At the moment, both indexes are downwardly biased. The bias-adjusted Median PCE is running at 3.75% and the bias-adjusted Trimmed-mean is running at 3.4%.

Our own measure of median PCE (which you receive every month) is unbiased and it is running at 3.45%, in line with the bias-adjusted alternatives.

What the paper does

The authors examine the bias that arises when constructing alternative measures of “core” inflation from the skewness of price changes. The paper documents time variation in the biases of median PCE inflation and trimmed-mean PCE inflation and notes that this time variation is -in fact- related to changes in skewness. Based on this, the paper constructs unbiased measures of median and trimmed-mean PCE inflation rates that removes the biases generated by the observed time-varying skewness of the underlying distribution.

The idea behind the paper

The idea of the paper is based on the fact that the distribution of monthly (PCE) price changes has been moderately left-skewed most of the time. For convenience, we report here below the authors’ computed time-varying measure of skewness of PCE price changes.

Note: the figure shows the MA(12) of a measure of price changes skewness computed by the authors. The authors use the average of the Bowley coefficient of skewness and the Kelly coefficient of skewness (both measures are bounded between -1 and +1, with negative and positive estimates indicating left- and right-skew, respectively).

The left-skew of the distribution mostly reflects the behavior of goods prices that not only tend to increase at a slower pace compared with services prices, but also exhibit deflation in some components. The presence of skewness in the price change distribution has important implications. In fact, when a distribution is left(right)-skewed, the mean lies below(above) the median, and, hence, the median PCE inflation rate will be upward(downwardly)-biased with respect to its headline counterpart (indeed, the authors show that the bias can be as high as 1½ percentage point). Because of the shift from left (negative) skew to right (positive) skew during the pandemic, the median or symmetric trimmed-mean is now downward biased with respect to the mean. Consequently, the median PCE inflation rate has now become a downward-biased estimate of the mean. Not only but compared to the median PCE inflation rate, the Dallas Fed’s trimmed mean PCE inflation rate currently suffers from more downward bias owing to its asymmetric trimming. This occurs because the current configuration of the price-change distribution would call for more trimming of the weight from the lower tail and less trimming of the weight from the upper tail.

Results of the paper

Using regression analysis, the authors estimate the bias introduced in the median PCE index and the Trimmed-mean PCE index by the (shifting) skewness. According to the authors, at the end of their sample (December 2021), the bias-adjustment procedure boosts the final reading of median PCE inflation by 14 basis points to 3.75 percent and the trimmed-mean PCE inflation by 34 basis points to 3.39 percent.

The figure below shows the authors’ results comparing YoY of headline PCE to Median PCE, bias-adjusted Median PCE, Trimmed-mean PCE, and bias-adjusted Trimmed-mean PCE.

Comment

Since the start of our business, we have introduced and circulated our own unbiased measure of median PCE (please, see our “Monthly distributions” email). The bias in the median PCE index and the Trimmed-mean index is well known in the literature. However, it was less understood that the bias is time-varying due to the shifts in the skewness. Correcting for this bias is very important because over some sub-samples it can be larger than 1 percentage point. Therefore, relying on the unadjusted Cleveland Fed median PCE index (or on the Dallas Fed Trimmed-mean index) can be seriously misleading. The latest reading (YoY) of our own median PCE index was at 3.44 percent (in January), in line with the authors’ unbiased estimates. Going forward, you will continue receiving our median PCE index updates as soon as new data will become available.

Contact

Want something more tailored?

We provide tailored consulting on ad-hoc projects.