3-19-19

Construction Spending for Jan was just released March 13th at $1.287 trillion saar, up 1.3% from Dec. But…

the most notable numbers in this March report are BIG REVISIONS DOWNWARD TO NOV (-2.3%) AND DEC (-2.2%), particularly to residential.

Residential revisions Nov -3.9%, Dec -4.4%.

Nonresidential revised down -0.8% in both Nov & Dec.

2018 total spending currently at $1.293 trillion, 2% below expected.

If these residential reports are correct, we have been in a residential construction spending downturn, now down 12% since last May. I predict these numbers are suspect.

Current total residential spending for 2018 in this most recent report is $545 billion, up 2.6% from 2017. But after deducting 4.4% residential inflation, that means real residential construction volume would be DOWN -1.8% for 2018. Yet, new starts in 2018 were up 6% and new residential jobs increased 3.5%.

However, in 2018 residential spending, the monthly variances from statistical rates of spending look very similar to 2006, up thru May then down 15% to year end.

Cash flow models of construction starts data from Dodge Data for residential spending spread over time indicate residential spending should be UP 3% since May, not down!

Just a reminder, we had a huge upward revision to 2015 residential spending data in July 2016. The 2018 data gets a big revision on July 1, 2019.

The monthly spending data gets revised three times after the 1st release. In the mean time, it can be compared to the statistical averages to determine if 1^st reports, or even 2nd reports of spending are in line with expectations. The current Nov and Dec residential construction spending data varies from the statistical average by 3 to 4 standard deviations. That’s highly unusual!

To understand just how unusual that is, let’s compare to how rare any monthly spending varies by 2 standard deviations (StdDev) from the statistical monthly average:

Non-Builiding Infrastructure has varied by 2 StdDev only 5 times in 18 years, only once in the last 15 years, even including recessionary years.

Excluding recessionary years, Nonresidential Buildings varied by more than 2 StdDev  only 12 times in 16 years and only 4 times in the last 8 years.

In both those data sets above, only twice ever in over 400 total months of data was the variance greater than 2.5 StdDev.

Excluding recessionary years, Residential Building monthly spending exceeded 2 StdDev from the average only 10 times in 14 years, all 10 times were in the last 8 years.

Recessionary years really skew the data. Non-building Infrastructure was the only sector not affected by the recession like all other construction. Residential spending was affected the most. In the four residential recessionary years, 2006-2009, residential monthly spending (in 48 months) exceeded 2 StdDev 28 times, 19 of those times 3 StdDev or greater.

In the residential data set, 5 of the 10 non-recessionary variances over 2 StdDev were in 2018. That simply does not occur in the historical data. That’s like having your teenage son grow an inch every year from 13 to 16 but then shrinking 2 inches in year 17. The average of those 5 unusual months in 2018 was 3.5 StdDev from the statistical average, with Nov and Dec posting the greatest variances since 2009.

The only time we’ve ever seen data like that was within the recession years of 2006-2009. So, either the 2018 residential data is foretelling the beginning of another recession, or the data, particularly the Nov and Dec data, will be subject to significant revisions, in this case upward.

Starts and cash flow expectations seem to indicate the 2018 data is currently being reported too low. I expect 2018 residential data will be revised up by $10bil to $15bil. We may not have that information until July 1, 2019.