It is a fact that when distributions have thicker tails than the normal Gaussian distribution, trimmed-mean and other robust estimators do a better job of estimating central tendencies than simple averages. It is a fact that food and energy sector prices, especially, have deviations from the overall inflation rate that have much thicker tails than the Gaussian normal. And it is a fact that over the past fifty years in America whenever the full CPI inflation rate has moved away, either up or down, from the core CPI inflation rate--the inflation rate excluding food and energy prices--the odds are better than 9-1 that the gap will then be closed by the full CPI moving back to the core, and not the core moving to the full CPI:
Thus our jaws drop when we read things like those below:
Robert Murphy: we are talking about CPI for all urban consumers, all items. In other words, this is old-school CPI, not "core" CPI (which excludes food and energy). And we're not doing anything fancy like looking at the middle 80% or whatever the newfangled techniques are.
David Henderson: We will use the standard CPI, not the nonsense CPI excluding food and energy.
Stephen Williamson: [W]hich prices are volatile and which are not will depend on the monetary policy regime and what the central bank is attempting to target…. Suppose that all prices are flexible, and that the Fed can control the price level, but only imprecisely due to errors related to measurement, the loose link between policy and prices, or even sheer stupidity. Suppose that there are essentially two kinds of goods, "volatile-price" goods, and "smooth-price" goods, with the relative price of these two types of goods fluctuating randomly…. [S]uppose the Fed chooses to target the price of the smooth-price goods, in dollars. As a result, the prices of volatile-price goods are indeed volatile…. [T]he prices of smooth-price goods, which are indeed smooth, are an excellent tool for forecasting the future price level…. But the Fed could choose instead to target the prices of volatile-price goods. Then, everything is turned on its head. The volatile prices are now smooth, the smooth prices are volatile, and the good forecasting tool for the future price level is the volatile price…. [Y]ou will now say: "But if the Fed attempted to target the non-sticky prices, we would have a problem, in that we would get inefficient fluctuations in real GDP. Further, the Bils and Klenow data, and Klenow-Malin work tell us something important about the stickiness of prices across goods and services." To which my reply is: "This evidence tells us only about the observed behavior of particular prices, and tells us nothing about how the pricing behavior depends on monetary policy. For that, we need a theory, and perhaps a structural model of price-setting that we can fit to the data. Indeed, I think I (or someone else) could write down a model with flexible prices where, under particular monetary regimes, the model delivers the features of the data. Indeed, this paper, by Head/Liu/Menzio/Wright does something like that." To say how a central bank should be responding to particular observed price movements, we need some solid theory concerning the welfare effects of inflation, how monetary policy affects inflation, and some solid measurement to tell us about the quantitative effects. I don't see anything solid that justifies the Fed's focus on core inflation measures. Indeed, one could, I think, make a better case for looking at headline inflation measures.
I simply do not understand why people would say things like this. IMHO, what we see here is simply the true weirdness of those who have not done their economic or statistical homework.