This post is written by Ed Zarenski, an estimating executive at Gilbane Building Company, in Providence, R.I. Zarenski writes the quarterly report “Construction Economics — Market Conditions in Construction,” an in-depth analysis of economic trends in the construction industry available FREE at the company’s website.
Every month, the U.S. Census Bureau releases construction spending data. Two values are given for each category within the industry, the Not Seasonally Adjusted amount for the month and the Seasonally Adjusted Annual Rate. The SAAR shows what the yearly spending would be based on the SAAR factor for that month. All news reports reference the SAAR data.
Unless spending is perfectly seasonally balanced, individual monthly SAAR can and does vary from month to month and from the total spending for the year. The actual annual total is the aggregate sum of the Not Seasonally Adjusted actual spending from each month.
Spending occurs at a different rate every month. There is more spending in June through October and less spending in December through March. There are historical percentages that guide us in our expectations for spending each month. The Census Bureau developed a set of historical averages called the SAAR. My analysis of the actual historical monthly data, whether for short or long duration, shows slightly different monthly spending rates than depicted by the SAAR. Therefore, I have developed averages from the actual monthly data that vary somewhat from the SAAR.
We can look at the data two different ways: the expected for each month and the expected year to date. Statistical analysis shows that some individual months are much more reliable (least variation in monthly percentage from year to year). Residential data seems to be most reliable in July, August and September, while non-residential data is most reliable in May, June, July and August. Although the best and worst months vary by construction sector, generally November through March provide the least reliable data due to wider fluctuations from year to year. The year-to-date data obviously grow in reliability as the year progresses.
Predicting construction spending
So, how can I tell if my factors produce reliable results? From numerous examples, based on my factors using historical spending from 2001 to present, the expected year-end total spending can be predicted:
- from the September year-to-date spending, with a statistical average variance of +/-1.5%. In 11 years, once the variation was greater than 1.5%, but eight times it was 1% or less.
- from the June monthly spending, with a statistical average variance of less than +/-1.2%. In 11 years, only once was the variation greater than 1.2%, and six times it was less than 0.6%.
- from the June-July-August three-month total spending, with a statistical average variance of less than +/-1%. In 11 years, once the variation was 1.4%, but nine times it was 0.8% or less.
Interpreting the census construction spending data
News reports, always based on the Census SAAR, often vary from my analysis. For example, construction spending in August 2012 was reported as experiencing the “largest drop in a year.” Although that statement is supported by the Census SAAR, it is not supported by my data.
Including August, spending declined four times in the past year and both January and March declines are four to five times larger than August. To smooth the data, three-month moving averages compared on an ongoing basis shows that January, February and March 2012 was by far the worst three months since the three-month period December 2010, January and February 2011, with construction spending in January 2011 showing the worst monthly decline on record in my 12 years of data. Coincidentally, the bottom in spending was reached during that period from December 2010 to February 2011 and that agrees with the bottom in construction jobs.
Yes, August 2012 spending dropped from that in July. But it would be rare to see data climb continuously without dips and spending had just climbed four months in a row. However, August finished above the statistical average for the month. So, why then would August be down? Because July finished further above its monthly statistical average than August did. Also of note, August rate of spending almost always (nine out of 12 years, or 75% of the time) declines from July.
This post is written by Ed Zarenski, an estimating executive for Gilbane Building in Providence, R.I. Zarenski writes the quarterly report “Construction Economics — Market Conditions in Construction,” an in-depth analysis of economic trends in the construction industry, available free at Gilbane’s website.