Posts
![[Most Recent Quotes from www.kitco.com]](http://kitconet.com/charts/metals/gold/t24_au_en_usoz_2.gif)
![[Most Recent Quotes from www.kitco.com]](http://kitconet.com/charts/metals/silver/t24_ag_en_usoz_2.gif)
![[Most Recent USD from www.kitco.com]](http://www.weblinks247.com/indexes/idx24_usd_en_2.gif)
A math problem at the Labor Department?
Friday, February 19, 2010
This morning's inflation report is generating all sorts of headlines about core consumer prices falling for the first time since 1982. In looking at the screen shot of the detailed data from the Labor Department below, there is clearly some sort of a math problem associated with changes that were recently made to category weightings (note that only the relevant data is included below from a much larger table).
Looking at the weighting (Relative importance) and the indentation on the left to determine which categories are sub-categories of others it becomes clear that you can't get the large negative number of -0.5 percent circled in red from the data circled in blue.
It appears that they are mistakenly weighting the -2.1 percent decline for lodging away from home at a much higher level and, since housing is a major component of core inflation, the first negative reading in 28 years was the result.
14 comments:
So they figure that homeowners are paying less imaginary rent to themselves, so they can keep on printing with abandon. Pure Orwellian double speak.
That's funny - they mistakenly print the number they should have printed, given the joke that owners' equivalent rent is.......
yeah, looks like the 0.769 might actually have been 7.69 in the calculation. (turns the final division into roughly 20/40)
I ran the numbers and if the seasonally adjusted shelter subgroup numbers are correct, then the total change in shelter should be -.1 and not -.5. If so, CPI-U should be +.3 (+.2) and less fodd & energy +.1 (-.1).
Good catch. I wonder if an explanation will be forthcoming from BLS.
it might be the seasonal adjustment. i am not positive but the changes in the individual components is adjusted separately.
Well, I'm glad somebody else has looked at this and come away with the same conclusion. I've noticed some inbound traffic from the BLS and FRB over the last couple hours but haven't seen or heard anything else yet.
From Rosie today:
The U.S. consumer price data is hot off the press and while the headline came in below expected at +0.2% MoM (so much for the PPI being a leading indicator). The real key was the -0.14% print on the core index (which removes food and energy) — deflation in the core CPI is a 1-in-80 event and should be treated seriously in terms of what it means for bond yields and corporate pricing power in the broad retail sector (there were notable
This is interesting, Tim--good catch. Let's see if the BLS issues some kind of explanation for something that on the surface clearly looks wrong.
For what it is worth, I did the math on the non-seasonally adjusted data and at least it looks consistent.
The problem looks to be confined to the seasonal data, and since they just did an annual revision to the data who knows what is going on there.
I do agree that it sure looks like a mistake though, even if the data has been revised.
Excellent catch.
http://illusionofprosperity.blogspot.com/2010/02/shelter-and-cpi-mistake.html
Tim,
This was CalculatedRisk's comment regarding your error:
http://www.hoocoodanode.org/node/9132#comment-1151055
RATM, in the previous thread you linked to someone claiming the CPI number was calculated incorrectly. I checked ... the guy mixed using the NSA weightings for the SA CPI components. That is incorrect. Each component is individual seasonally adjusted and then added together to calculate the SA total. We can check the NSA math (it appears correct), but not the SA math.
best wishes
Thanks - I just saw that too, but it still doesn't make sense when you look at the numbers. They calculate individual SA numbers and the SA vs. NSA weightings can't be that different - dominated by rent and OER which are both far, far away from the -0.5 category total.
CR's explanation is bogus - very smart guy, but wrong this time. The shelter components and the total for shelter are all seasonally adjusted so you can just use ballpark weightings to see that the total published by the BLS is wrong.
Tim - the only reasonable explanation is the combination of seasonal adjustments and weighting - a lot of people stop paying mortgages and rent and stay in hotels while visiting relatives.
Even better is the item "Apparel": 4 positive sub-indices result in a negative number for the combined index.
Post a Comment