Monday, June 29, 2009
For the Record: The BLS' NSA CPI circa June 2009
In a recent post [link to be added], Robert Wenzel spelled out something that had been implicit in my views. I'll recapitulate the argument here, and then I'll list the BLS' current figures so we have a convenient record later on.
For a while I've been warning that actual Consumer Price Index (CPI) inflation has been much higher than the official, "seasonally adjusted" numbers that the media report. For example, if actual CPI rose by 0.3 percent in May, the BLS would adjust it down to a mere 0.1 percent, and the media would dutifully say, "Prices rose 0.1 percent last month, showing deflation threat still looms."
Now if the "seasonal adjustment" for the first six months is to push down prices below their actual rates of inflation, then by symmetry in the latter half of the year, the BLS will need to bump up the actual numbers and report higher inflation rates to the media. This is where I thought the hanky panky would come in.
In this context, Wenzel clarifies that by next year, obviously the BLS numbers for 2009 have to show the same annual rates of inflation, over the course of 2009. In other words, the "seasonal adjustments" for the year 2009 will have to cancel each other out, over the four seasons.
So in practice, when I say the BLS is going to cheat, what they will do is revise the "seasonal adjustments" from January through June 2009. For example, they will say," Upon further review, back in May, seasonally adjusted prices actually rose by 0.2 percent. Gosh it's too bad we told the media the lower figure of 0.1 percent at the time. Our bad."
In this way, they can keep the upward seasonal adjustments very tame, from July through December. If the actual CPI rises, say, by 0.7 percent in August, then the BLS will report it as either a straight 0.7 (i.e. no seasonal adjusmtent) or a 0.8 percent, even though they were bumping down the numbers by bigger margins in the first half of the year. But they're not going to want to report a 1.0 percent price hike, so they'll contain the monthly inflation number by dumping some of it back into earlier months.
=============
Let's take a snapshot of the two series before the predicted hanky panky ensues:
MONTH....NSA CPI....SA CPI
Jan 09.....211.143......212.174
Feb 09.....212.193......213.007
Mar 09.....212.709......212.714
Apr 09.....213.240......212.671
May 09.....213.856......212.876
For a while I've been warning that actual Consumer Price Index (CPI) inflation has been much higher than the official, "seasonally adjusted" numbers that the media report. For example, if actual CPI rose by 0.3 percent in May, the BLS would adjust it down to a mere 0.1 percent, and the media would dutifully say, "Prices rose 0.1 percent last month, showing deflation threat still looms."
Now if the "seasonal adjustment" for the first six months is to push down prices below their actual rates of inflation, then by symmetry in the latter half of the year, the BLS will need to bump up the actual numbers and report higher inflation rates to the media. This is where I thought the hanky panky would come in.
In this context, Wenzel clarifies that by next year, obviously the BLS numbers for 2009 have to show the same annual rates of inflation, over the course of 2009. In other words, the "seasonal adjustments" for the year 2009 will have to cancel each other out, over the four seasons.
So in practice, when I say the BLS is going to cheat, what they will do is revise the "seasonal adjustments" from January through June 2009. For example, they will say," Upon further review, back in May, seasonally adjusted prices actually rose by 0.2 percent. Gosh it's too bad we told the media the lower figure of 0.1 percent at the time. Our bad."
In this way, they can keep the upward seasonal adjustments very tame, from July through December. If the actual CPI rises, say, by 0.7 percent in August, then the BLS will report it as either a straight 0.7 (i.e. no seasonal adjusmtent) or a 0.8 percent, even though they were bumping down the numbers by bigger margins in the first half of the year. But they're not going to want to report a 1.0 percent price hike, so they'll contain the monthly inflation number by dumping some of it back into earlier months.
Let's take a snapshot of the two series before the predicted hanky panky ensues:
MONTH....NSA CPI....SA CPI
Jan 09.....211.143......212.174
Feb 09.....212.193......213.007
Mar 09.....212.709......212.714
Apr 09.....213.240......212.671
May 09.....213.856......212.876
Subscribe to Posts [Atom]