NOTE: The first part of this essay presents my view BEFORE I looked into this "headline" and "core" inflation stuff. I say a lot of things I am now unhappy with. If you read it, be sure to read the rest of the essay, where I revise my thinking.
I use the CPI as my inflation measure. I use it because I remember it from the 1970s, and because news reports today still use it. The Fed prefers a different measure -- the PCE price index, something that sounds like it is mostly based on "Personal Consumption Expenditure" prices.
There is already PCE data on the volume of purchasing. Now we also have PCE data on prices. You should know not to mix them up. The PCE price data is an index, a "price index", so you can call it the PCE index and know that it shows a history of the price level, not the volume of purchasing.
What do they mean by "index"? I dunno. It hardly
matters. I just make sure to include the word "index" when I search for
the PCE price index data.
The Fed prefers to use the PCE price index in its work on inflation. I am not in the habit of using the PCE price index data. But for this essay it is relevant, so I spent half a day yesterday trying to find a news story that identifies the PCE index well enough that I can find the data. Everybody mentions the PCE price index and says the Fed prefers it, but nobody identifies the data.
I finally found something useful. From early in the 2021-2023 inflation, "Fed's inflation measure holds at record annual pace for fourth straight month" (October 29, 2021) at Fox Business, says:
The core personal consumption expenditures index, the Federal Reserve’s preferred inflation measure, continued to climb at the fastest annual pace on record in September.
They call it the "Core PCE index", and say it "excludes food and energy prices".
At this "record annual pace" article, the sidebar offers a link that sounds promising: "Inflation measure closely watched by the Fed rises 0.3% in April" (May 31, 2024). The "closely watched" article refers to a "closely watched" measure, and calls it the "personal consumption expenditures (PCE) index". They refer also to a second measure, the "core prices" measure, which strips out "the more volatile measurements of food and energy". This second measure is the one noted in the first article, the "record annual pace" article.
The "closely watched" article shows a graph of the two measures. The graph has the two measures labeled as "Core PCE" and "Annual PCE". They're both annual, actually (or what FRED calls "Percent Change from Year Ago"). In the paragraph below the graph, they say:
Both the core and headline numbers point to inflation that is still running well above the Fed’s preferred 2% target.
Here
they call the one measure "headline" instead of "annual". They use the
word "headline" twice in that paragraph. I'm going with that, with
"headline", for the one that is not "core". The headline measure
includes food and energy. The core measure excludes food and energy. I
remember better if I write it down.
At the Bureau of Economic
Analysis they figure the price level every month. The "closely watched"
article has it as "0.3% in April". That's the monthly version (or FRED's
"Percent Change" from the previous month). Then, going back a year, they
take a dozen of the monthly numbers and put them together (with
compounding, I figure, as banks do when interest accumulates) to get the
"Percent Change from Year Ago" number. "Closely watched" has the
percent change from year ago at 2.7%.
While the Fed is targeting the PCE headline figure as it tries to wrestle consumer prices back to 2%, Chair Jerome Powell previously told reporters that core data is actually a better indicator of inflation.
They always say the one measure is better than the rest. But they never say what makes it better. Headline PCE is "preferred" to the CPI, but Core PCE is "better". These excruciatingly insignificant opinions are endlessly repeated in news stories on inflation. But neither those stories nor the Fed's talkers ever say why the Fed prefers the PCE to the CPI. And nobody ever says why they think Core PCE is better than the Headline PCE.
But I think we know why. I think we know why Core PCE is "better". The primary job of the Fed is to keep inflation low. And Headline PCE runs lower than the CPI, so the Fed "prefers" it. But Core PCE often runs even lower than Headline PCE, so Core PCE is "better". The data that shows the least inflation is the data that best suits the Fed's purposes.
Go ahead: call me a cynic.
Hey, inflation is what it is. Having different ways to measure it, and picking the lowest number, does nothing to reduce inflation. It does nothing to solve the problem.
Let the corrections begin!
What's the difference between Headline
and Core PCE? Yeah, Core
doesn't include food and energy, I remember. But what's the difference
on a graph? What's the difference in the numbers? I thought I was going
to be able to say this:
In an economy where prices tend to go up, most of the price changes are increases. Few are price declines. Stripping out volatile price changes, or even changes at random, will on average remove more price increases than declines, and will reduce the inflation reported by the modified dataset. The Core PCE measure understates inflation by design.
But I cannot say it. My graphs refuse to
show it. I spent the whole afternoon trying to get them to show it, and
my graphs refused. Why? I dunno. But I trust my graphs.
This
graph compares the CPI to Headline and Core PCE. As a rule, CPI shows
inflation higher than the other measures. No wonder the Federal Reserve
doesn't like the CPI!
 | ||
| Figure N: Comparison of CPI (red), Headline PCE (blue), and Core PCE (black dots) Click Graph to Enlarge. Click This Text for the Graph at FRED |
On this graph, red shows the CPI, blue shows "Headline" PCE inflation, and the black dots show "Core" PCE inflation. When red and blue are going up, the black dots go up less. When red and blue are going down, the black dots go down less. The black dots move less because Core PCE is less "volatile". Core PCE is less volatile because they leave the volatile data out: Food and energy. Right. Who needs food and energy?
Time for a snack.
I get it now. Core PCE does not understate inflation. It understates changes in inflation. It works like the Hodrick-Prescott filter was thought to work before it fell out of favor: Core PCE smooths out the data. With Core PCE, the highs are less high and the lows are less low. It's right there on the graph.
When inflation is in the news, it is because the data is changing. It is going up. It is "volatile". At such times, if we look at it on a graph, "Core" inflation will always appear to understate the inflation, because the volatile data has been omitted.
It's just the opposite when inflation goes suddenly low. Headline inflation, the volatile one, goes down rapidly. Core inflation goes down only a little. At such times, Core inflation appears to overstate the inflation, for the same reason: the volatile data has been omitted.
I hope you can see why I like graphs so much!
After all the work was done, I had an idea. I went to FRED and searched their data for headline PCE. Just three results turned up:
- Real Personal Consumption Expenditures (PCEC96)
- Personal Consumption Expenditures: Chain-type Price Index (PCEPI)
- Personal Consumption Expenditures Excluding Food and Energy (Chain-Type Price Index) (PCEPILFE)
The first one is the volume-of-spending data. The other two are the headline and core PCE measures that I spent half a day looking for.
As an afterthought...
As I picture the process, they check prices once a month on a specific set of things people buy, goods and services say, and get the price of that "market basket" of stuff. After a few years, say, they have a set of prices, once price per month, for the same set of goods and services.
They can take those prices, or an "index" of those prices (which would follow the same path, on a graph), and figure "percent change" values to see how much prices changed, each month relative to the month before. That's when they get a small number like 0.3%.
Above, I said
Then, going back a year, they take a dozen of the monthly numbers and put them together (with compounding, I figure, as banks do when interest accumulates) to get the "Percent Change from Year Ago" number.
Yeah, they don't have to do that. They can just take the price (or the index number) from a year ago and figure how much the current price (or index number) has increased. This is when they get the bigger number, like 2.7%. This way there is no need to figure percent increase based on the monthly percentages. (That thought is what woke me up this morning, at 3 AM.)
No comments:
Post a Comment