Behavior of a New Median PCE Measure: A Tale of Tails

Abstract

We introduce two new measures of trend inflation, a median PCE inflation rate and a median PCE excluding OER inflation rate, and investigate their performance. Our analysis indicates that both perform comparably to other simple trend inflation estimators such as the trimmed-mean PCE. Furthermore, we find that the performance of the median PCE is related to skewness in the distribution of cross-sectional growth rates across categories in the PCE, and our results suggest that the Bowley skewness statistic may be useful in forecasting.

Finding a Stable Phillips Curve Relationship: A Persistence-Dependent Regression Model

“…First, because the core PCE price index simply excludes items from the basket, core PCE inflation can be subject to bias over prolonged periods. And as Carroll and Verbrugge (2019) indicate, this bias has also been highly unstable over time. For example, between 1995 and 2007, core PCE inflation was downwardly biased by 0.25 percentage points, while it was upwardly biased by 0.3 percentage points between 1980 and 1985.…”

“…Similarly, both the Reserve Bank of New Zealand and the Bank of Canada now shun the use of "core" (exclusion-based) inflation measures as measures of trend inflation. SeeCarroll and Verbrugge (2019) andVerbrugge, Zaman, and Nunna (2018) for additional evidence regarding the superior forecasting ability of the trimmed mean PCE and Prior studies, such as those ofClark and McCracken (2006),Faust and Wright (2013),Zaman (2013), andClark and Doh (2014), have shown that inclusion of an accurate trend estimate improves the accuracy of inflation forecasts. In the context of Phillips curve estimation, modeling inflation in terms of deviation from the trend * t  amounts to assuming that the Phillips curve relationship is silent about the long-run goals of monetary policy and, instead, pertains to fluctuations in inflation that are more closely related to business cycles.…”

Cyclical versus Acyclical Inflation: A Deeper Dive

“…The fact that cyclically sensitive inflation rates are heavily influenced by housing categories is not unique to these specific indicators. Housing tends to be an important category when computing median inflation rates, such as the median CPI or median PCE; see, for example, Bednar and Knotek (2014) and Carroll and Verbrugge (2019).…”

“…In general, inflation trend estimators such as the median CPI, trimmed-mean CPI, median PCE, and trimmedmean PCE have a strong connection to labor market slack (for example, Ball and Mazumder, 2019;Carroll and Verbrugge, 2019). Stock and Watson (2019) suggest that this strong association with economic slack comes from the large role that the highly procyclical housing component plays in their calculation.…”

“…Stock and Watson (2019) suggest that this strong association with economic slack comes from the large role that the highly procyclical housing component plays in their calculation. Figure 5 plots the unemployment gap and cyclical core PCE inflation alongside median PCE inflation (based on Carroll and Verbrugge, 2019) and trimmed-mean PCE inflation. The latter three series share similar trends and are highly correlated with each other, and they are strongly negatively correlated with the unemployment gap, even though the median PCE and trimmed-mean PCE inflation rates do not use the unemployment gap to inform their construction.…”