10 Years Of “Processing Errors” In Construction Spending Data Slamming GDP
Even as increasingly more parts of the economy, especially those with exposure to manufacturing and industrial production, sink into the recessionary quicksand, one sector that was seen as immune from the malaise gripping US manufacturing and was outperforming the overall growth rate of the US economy, was housing, and specifically spending on private and public construction: a direct input into the GDP model.
That all changed today when the US Census released its latest, November, construction spending data, which not only missed expectations of a 0.6% increase, but tumbled -0.4%, the most since June of 2014, while all the recent changes were mysteriously revised lower.
And then the source of the mystery was revealed: in the fine print of the release, the government made a rare admission: all the construction spending data for the past 10 years had been “erroneous.”
In the November 2015 press release, monthly and annual estimates for private residential, total private, total residential and total construction spending for January 2005 through October 2015 have been revised to correct a processing error in the tabulation of data on private residential improvement spending. An Excel file containg all of the revisions can be found here
The result of the “revision” of the processing error is shown below: every month starting with April and going through October, was “found” to have been a lower increase than according to the previous data. Not only that, but the October print which had been the strongest since May, confounding many data watchers as it did not fit with anecdotal evidence of a dramatic slowdown in energy-related construction, suddenly was barely positive, leading to the November sequential decline, the worst since the -0.7% drop in June of 2014.
And here is the big picture: what it reveals is that while spending data in 2013 was revised substantially higher, it proves what many have known, namely that the economy is now slowing substantially and that what until recently was seen as the strong annual increase in construction spending, namely the 14.3% increase of September 2015, was in fact substantially lower.
The result is that the October Y/Y% change of 10.5%, and declining, is not only the lowest increase since April, but matches the level first reported in December 2013. In other words, contrary to the previous narrative suggesting construction spending was solid and supporting a growing economy, it has in fact been declining since June!
And to think of the tons of digital ink spent by “strategists” and experts analyzing construction spending “data” in the past 5 years…
Sarcasm aside, what this exercise proves – which is clearly meant to lower the goalseeked glideslope of the US economy and make it easier to enter recession – is what many have already said, namely that Yellen clearly missed her window to hike rates with the economy now clearly slowing down, and instead of tightening monetary conditions, Yellen should be easing and preparing to lower rates.
To be sure, this is not the first time the US government has slashed historical data on a wholesale basis due to “revisions” and “errors” – recall our post from December 2014 “The Housing Recovery Remains Cancelled Due To 6 Months Of Downward Revisions” in which we showed how 6 months of New Home Sales were quietly revised materially and, of course, to the downside.
And since as noted above, this data feeds straight into the GDP “beancounts”, we expect substantial downward revisions of recent historical GDP data, which will once again confirm Yellen’s rate hike error.
Finally, we now await for even more government data (perhaps payrolls is next) to “unexpectedly” be shown as having substantial historical errors, and be revised, like in the cases above, materially to the downside because it will look silly if the US economy jumps from growth straight into recession with existing “data sets” which reveal that the bulk of what passes for “data” at the US government is simply double and triple-seasonally adjusted GIGO.