The Atlanta Fed's macroblog provides commentary and analysis on economic topics including monetary policy, macroeconomic developments, inflation, labor economics, and financial issues.

Authors for macroblog are Dave Altig, John Robertson, and other Atlanta Fed economists and researchers.

« Should the Fed stay in regulation? | Main | Exit strategy, tactics, and decision makers »

February 23, 2010

Looking behind the core inflation numbers

Last Friday's consumer price index (CPI) report showed headline inflation in January remaining stable from the previous month at a 2 percent annual rate, a bit above most private forecasts, boosted by higher fuel costs. But the show was stolen by the core measure. Excluding food and energy, consumer inflation saw the largest monthly drop in more than 27 years and its third largest decline in 47 years.

Several factors were behind the decline in the core index (such as falling airline fares, a dip in new car prices, and ongoing declines in prices for household furnishings and operations), but a sizeable drop in shelter prices, which account for more than 40 percent of the core CPI index, was a significant factor in January's dip. A concern that decelerating shelter prices could skew the core inflation measure was noted in the minutes of January's FOMC meeting:

“Though headline inflation had been variable, largely reflecting swings in energy prices, core measures of inflation were subdued and were expected to remain so. One participant noted that core inflation had been held down in recent quarters by unusually slow increases in the price index for shelter, and that the recent behavior of core inflation might be a misleading signal of the underlying inflation trend.”

Chart 1 illustrates the recent decline the in shelter component of the CPI and how, excluding shelter, core inflation has been growing at a more robust pace than is indicated in the overall number.

Chart 1

However, once we've opened the door for pruning sectors that have displayed unusual price behavior in recent months, we can find a slew of outlying components to pare. Take, for example, vehicle prices. With the impact of the ailing state of U.S. automakers, the federal government's “Cash for Clunkers” program, and the major recalls from Toyota, auto prices have been particularly volatile lately. Used car and truck prices have grown at annual rates between 20 percent and 44 percent in each of the past six months, skewing, some could argue, the core measure upward.

Chart 2 shows core CPI after subtracting both shelter and used vehicle prices, a picture that shows a lower inflation rate—more in line with the numbers in the overall core CPI.

Chart 2

My point here is not to advocate lopping shelter and vehicles, along with the already excluded food and energy prices from inflation—which, by the way, would leave us with less than 45 percent of the overall CPI index. In fact, my argument is the opposite. There are always some components of the index that seem anomalous—on either side of the distribution. Discriminately cropping entire sectors from the CPI may not be the best method for teasing out true underlying price pressures.

Chart 3

The Cleveland Fed uses a more methodical approach to exclude the CPI components that show the most extreme price changes each month. Their calculations show a more subdued underlying inflation environment, with median and trimmed mean CPI hovering around 1 percent for the past several months (chart 3). I'm not endorsing this method as a perfect estimation of “true” underlying inflation, but it does provide an example of indiscriminately trimming the outliers to see what's beneath.

By Laurel Graefe, senior economic research analyst at the Atlanta Fed

February 23, 2010 in Inflation | Permalink


TrackBack URL for this entry:

Listed below are links to blogs that reference Looking behind the core inflation numbers :


A couple of blogs found that BLS made an error. The actual core reading was a positive .1%, not the -.1 as reported. So much for colas.

Posted by: flow5 | February 23, 2010 at 02:24 PM

Another thing that Cleveland (CPI) and Dallas (PCE deflator) statistical rejiggering efforts do is to break out the share of components for which prices fell in a month, rose less than 3%, rose from 3% to 5%, more than 5% and so on. There has been a fairly dramatic shift toward falling prices and away from rapidly rising prices.

Posted by: kharris | February 23, 2010 at 02:49 PM

The median CPI is also very flawed.

How does one justify removing the impact of components with very large weights, such as housing, food, or energy? By doing so, one ends up with a CPI that's representative of as little as 50% of the original index.

You do also realize that the BLS adjusts other CPI components for changes in the energy prices. If there's a large spike in fuel prices, the BLS assumes that the lack of adjustment in the OER amounts to a subsidy and makes a corresponding change to the heating costs component in OER. If one is to remove energy costs because it's an outlier, then one must also back out all of its effects on the other CPI components.

Posted by: Les | February 23, 2010 at 03:00 PM

This post is well constructed. I enjoyed reading it.

There is reason, however, to omit shelter as opposed to cars or air fares. Shelter is different from other components of the CPI because of the presence of Owner’s Equivalent Rent (OER). Although the BLS does a masterful job of attempting to impute housing services to homeowners, these rents are inherently impossible to measure. The BLS makes every effort to sample rental prices of houses equivalent to the owned stock of housing. But the samples are small and biased. In any neighborhood, rental properties are inherently different from owned properties. This is likely particularly true now when many homeowners are unable to sell their houses and opt to rent, often below market rates, instead of selling at a steep discount.

Policy makers should not automatically exclude OER from measures of inflation; neither should they take it at face value. The good news is that at the moment it does not matter much. Although core and core ex shelter are evolving quite differently, both measures point to contained inflation and neither measure is accelerating.

Posted by: The Secret Economist | February 24, 2010 at 08:11 AM

I'm afraid before this is all over, wages will look like chart 1.

Posted by: FormerSSResdient | February 24, 2010 at 11:48 AM

I would point out to Les and to Secret E that policy makers have a fairly sophisticated knowledge of inflation measures, and make use of a number of them. If median CPI is flawed, well that's OK. In fact, its flaws are also its virtues, if used correctly. It is flawed in a particular way that is not entirely shared with other inflation measures. Tossing out the outliers does mask their impact in the single index from which they are tossed out, but conventional CPI measures, the PCE deflator, market based PCE deflator, Dallas Fed rejigger of the PCE deflator, GDP deflators of various kinds are all available to policy makers. Toss out food and energy, use a different, smaller housing component in the PCE deflator, toss out the most volatile components in a given month, look at a number of headline measures over a variety of periods, and then compare the trends among these many measures, and pretty soon, you have a detailed understanding of what is going on with consumer prices. Knowledge of how those measures are constructed, what their characteristics are and what drives them bolster their use. Given that range of information, we should welcome the "flaws" of each individual measure.

Posted by: kharris | February 24, 2010 at 01:18 PM

I really don't think food and energy should be removed. At best some type of rolling average should be included in the numbers. As food and energy make up such a huge part of peoples monthly expenses, it's misleading to exclude them.

Posted by: Ron Stone | June 02, 2010 at 11:55 AM

If there's a large spike in fuel prices, the BLS assumes that the lack of adjustment in the OER amounts to a subsidy and makes a corresponding change to the heating costs component in OER. If one is to remove energy costs because it's an outlier, then one must also back out all of its effects on the other CPI components.

Posted by: Homework Help | July 04, 2011 at 02:37 AM

Who doesn't like lower prices? Economists, for one. The good news is that many economists think the biggest deflation risk has passed.

Posted by: VMware course | October 12, 2011 at 01:15 AM

Post a comment

Comments are moderated and will not appear until the moderator has approved them.

If you have a TypeKey or TypePad account, please Sign in