Home » Behind the Headlines (Page 4)
Category Archives: Behind the Headlines
During the period including 2011 through 2017, we had record construction spending, up 50% in 5 years, moderate inflation reaching as high as 4.6% but averaging 3.8%, record construction volume growth (spending minus inflation), up 30% in 5 years and the the 2nd highest rate of jobs growth ever recorded.
Residential spending was up 90% in 5 years, but real residential volume up only 50%. Residential inflation, at 6%/year, was much higher than all construction. Jobs increased only 33%.
Construction added 1,339,000 jobs in the last 5 years. The only time in history that exceeded jobs growth like that was the period 1993-1999 with the highest 5-year growth ever of 1,483,000 jobs. That same 93-99 period had the previous highest spending and volume growth. 2004-2008 would have reached those lofty highs but the residential recession started in 2006 and by 2008 spending had already dropped 50%, offsetting the highest years of nonresidential growth ever posted.
The point made here is the period 2011-2017 shows spending and jobs at or near record growth. Although 2017 slowed, there is no widespread slowdown in volume or jobs growth.
This 2011-2017 plot of Construction Jobs Growth vs Construction Volume Growth seems to show there is no jobs shortage. In fact it shows jobs are growing slightly faster than volume. But that just does not sit well with survey data from contractors complaining of jobs shortages. So how is that explained?
There have been cries from some quarters, including this blog, that the answer lies in declining productivity. There seems to be plenty of workers, but it now takes more workers to do the same job that took fewer in the past. As we will see, that is part of the answer, but doesn’t explain why some contractors need to fill vacant positions. To find data that might answer that question about a jobs shortage we must dig a little deeper.
The total jobs vs volume picture masks what is going on in the three major sectors, Residential, Nonresidential Buildings and Non-Building Infrastructure. A breakout of jobs and volume growth by sector helps identify the imbalances and helps explain construction worker shortages. It shows the residential sector at a jobs deficit.
7 years 2011-2017 – % Jobs growth vs % Volume growth
- Totals All Construction Jobs +31%, Volume +30%
- Nonres Bldgs Jobs +27%, Volume +19%
- Nonbldg Hvy Engr Jobs +21%, Volume +12%
- Residential Jobs +40%, Volume +54%
The totals show jobs and volume almost equal, data that supports the 2011-2017 totals plot above and what we would expect in a balanced market. But severe imbalances show up by sector. Both nonresidential sectors show jobs growth far outpaced volume growth. Residential stands out with a huge deficit, with jobs way below volume growth.
Just looking at 2017 growth shows the most recent imbalances.
2017 % jobs growth vs % volume growth
- Totals All Construction Jobs +3.4% Volume -0.8%
- Nonres Bldgs Jobs +3.3% Volume -1.6%
- Nonbldg Hvy Engr Jobs +1.7% Volume -6.0%
- Residential Jobs +3.5% Volume +4.2%
Census recently released initial construction spending for 2017, totaling $1.230 trillion, up only 3.8% from 2016. What is somewhat disconcerting is that 2017 construction spending initial reports growth of 3.8% do not even match the total inflation growth of 4.6% for 2017, indicating a -0.8% volume decline. However, as does always occur, I’m expecting upward revisions (estimated +2%) to 2017$ construction spending on 7-1-18. If we don’t get an upward revision, then 2017 will go down as the largest productivity decline since recession. Even if we do get +2% upward revision to 2017$ spending, 2017 volume would be revised up to +1.2% and jobs growth will still exceed volume growth.
Let’s look a little deeper at the data within the sectors. Each chart is set to zero at Jan 2011 so we can see the change from that point, the low point of the recession, until today. At the bottom of each chart is shown a Balance at start. That represents the cumulative surplus or deficit of jobs growth compared to volume growth for the previous 10 years prior to Jan 2011. If there are no changes in productivity, or no surplus or deficit to counteract, then jobs should grow at the same pace as volume.
There are slight differences between the data in the three sector charts and the total construction chart. The sector charts use annual avg data and the totals chart uses actual monthly data.
Nonresidential Buildings and Non-building Infrastructure, over seven years and the most recent three years, show jobs increasing far more rapidly than volume. Nonresidential Buildings started 2011 with a surplus of jobs after the recession, but Infrastructure started 2011 with a substantial deficit of jobs. Only in this last year did Infrastructure jobs reach long-term balance with work volume.
Nonresidential Buildings started 2011 with a 13% surplus of jobs and more than doubled it in the seven years following. I’ve suggested before it could be that a part of this surplus is due to companies hiring to meet revenue growth, and not inflation adjusted volume. Although nonresidential spending actually increased 43%, volume since 2010 has increased only 12%. Since 2010 there has been 30% nonresidential buildings inflation, which adds zero to volume growth and zero need for new jobs. A 43% increase in spending could lead companies to erroneously act to staff up to meet spending, or revenue, more than needed for the 12% volume increase.
This plot for residential work shows from 2011 to the end of 2017, we’ve experienced a 20% growth deficit in jobs. How many residential jobs does this 20% growth deficit represent? From Jan 2011 through Dec 2017, residential jobs increased from approximately 2,000,000 to 2,700,000. So the base on which the % growth increased over that time is calculated on 2,000,000. An additional 20% growth would be a maximum of 400,000 more jobs needed to offset the seven year deficit. But what about the imbalances that existed when we started the period?
During the residential recession from just 2005 through 2010, residential volume declined by 55%, but jobs were reduced by only 38%. For the entire period 2001-2010, total volume of work declined by 14% more than jobs were reduced. Some of the surplus jobs get absorbed into workforce productivity losses and some remain available to increase workload. It’s impossible to tell how much of that labor force would be available to absorb future work, so for purposes of this analysis an estimate of at least 5% seems not unreasonable. That would mean for 2011-2017, instead of a need for an additional 20% more jobs, the need could be reduced by 5% or 100,000 jobs.
This analysis shows a current deficit of 300,000 to 400,000 residential construction jobs. While it does also show nonresidential buildings jobs far exceed the workload and there are more than enough surplus jobs to offset the residential deficit, there would be several questions of how transferable jobs might be between sectors.
- Are there highly technical specialty jobs in Nonresidential Buildings that would not be transferable to Residential?
- What is the incidence of specialty workers engaging in work across sectors? i.e., job is counted in one sector but working in another sector.
- What has been the impact of losing immigrants from the construction workforce?
- Is the ratio of immigrant workers in Residential much higher than Nonresidential?
- Is the pay more attractive in Nonresidential construction?
- What, if any, percentage of the Residential workforce is not being counted? Day labor?
One thing is known for certain, high-rise multifamily residential buildings may often be built by a firm that is classified primarily as a nonresidential commercial builder. Therefore, some jobs that are counted as nonresidential are really residential jobs.
I think most of these would have a more negative impact on Residential jobs. However, there is some possibility that the overall deficit may not be quite as high as available data show (points 2 and 6). And there is always the possibility that we’ve crossed a threshold that has led to new gains in productivity, although to some extent, the stark differences between Residential and Nonresidential Buildings data might counter that proposition.
These two following report references both document that there is a large unaccounted for shadow workforce in construction. This workforce is probably mostly residential.
and these more recent reports adds volumes of data on immigrant labor
Unemployment and productivity includes only jobs counted in the official U.S. Census Bureau of Labor Statistics (BLS) jobs report. Both these reports document a large, unaccounted for shadow workforce in construction. By some accounts, 40% or more of the construction workforce in California and Texas are immigrant workers. Immigrants may comprise between 14% and 22% of the total construction workforce. It is not clear how many within that total may or may not be included in the U.S. Census BLS jobs report. However, the totals are significant enough that they would alter some of the results commonly reported.
The best way to see the implications that the available data do show is to look at productivity. The simplest presentation of productivity measures the total volume of work completed divided by the number of workers needed to put the volume of work in place, or $Put-in-Place per worker. In this case, $ spending is adjusted for inflation to get a measure of constant $ volume, and jobs are adjusted for hours worked.
As the Residential jobs deficit increases vs workload, this plot shows that $PIP is increasing. That makes sense. The workload continues to increase and the jobs growth is lagging, so the $PIP per worker goes up. For Nonresidential Buildings, the rate of hiring is exceeding the rate of new volume and therefore the $PIP is declining.
In boom times, residential construction adds between 150,000 and 170,000 jobs per year and has only twice since 1993 added 200,000 jobs per year. In the most recent several years expansion, residential has reached a high of 156,000 jobs in one year but has averaged 130,000 per year over 5 years. So it’s pretty unlikely that we are about to start adding residential construction jobs at a continuous rate of 200,000+ jobs per year.
If residential jobs growth were to increase by 50,000 jobs per year over and above current average growth, it would take 6 to 8 years to wipe out the jobs deficit in residential construction.
This problem is not going away anytime soon.
For more history on jobs growth see Is There a Construction Jobs Shortage?
For more on the imbalances of Res and Nonres jobs see A Harder Pill To Swallow!
For some hypotheses as to why nonresidential imbalances continue to increase see Construction Spending May 2017 – Behind The Headlines
The Producer Price Index (PPI) for material inputs to construction gives us an indication whether costs for material inputs are going up or down. The PPI tracks producers’ cost to produce the product and supply finished products to retailers or contractors. However, that is far from the total cost from the contractor.
A good example is steel. The producer price for steel from the mill might be $750/ton for long beams and columns. The only increases captured at the producer level might be the changes in cost for raw material, energy to manufacture and the producers labor and markup. But the structural steel contractor is then responsible for delivery to shop, detailing, shop fabrication, transport to construction site, load and unload, cranes and welding equipment needed to install, installation crews and finally overhead and profit accounting for at least eight more points of potential cost change. Finally the steel subcontractor must then assess the market conditions, whether tight or favorable to higher profits, to adjust the bid price or selling price. The final cost of steel installed could be $3000/ton.
The PPI for Construction Inputs IS NOT a final indicator of construction inflation. It is an input to construction inflation. It does not represent the selling price, nor does it give any indication of the trend, up or down, of selling price.
In 2009 PPI for inputs was flat but construction inflation, as measured by final cost of buildings, was down 8% to 10%. In 2010, the PPI for construction inputs was up 5.3% but the selling price was flat. Construction inflation, based on several decades of trends, is approximately double consumer inflation. However, from mid-2009 to late 2012, that long-term trend did not hold up. During that period, PPI ranged from 0% to +6.8%, but construction inflation/deflation ranged from -10% to +2.3%, lower than PPI for all four years, something which seldom occurs. Construction inflation/deflation was primarily influenced by depressed bid margins, which had been driven lower due to diminished work volume.
The following table shows the differences between the PPI Inputs from 2011 to 2017 and the actual inflation for the major construction sectors. This table shows clearly that PPI Inputs and Inflation not only can vary widely but also may not even move in the same direction.
The PPI tables published by the Bureau of Labor Statistics do include several line items that represent Final Trades Cost or Whole Building Cost. Those PPI items don’t give us any details about the producer price or retail price of the materials used, but they do include all of the contractors costs incurred, including markups, on the final product delivered to the consumer, the building owner. I would note however that those line items in the PPI almost always show lower inflation than final Selling Price inflation indices developed separately from the PPI. Follow this link to table of inflation values which includes the PPI final cost for trades and buildings.
Construction Managers responsible for working with the client to manage project cost, part of which includes preparing a full building cost estimate, should not rely on PPI values as an indication of inflation. Selling price inflation indices are more appropriate indices to use to adjust project costs.
It is always important to carry the proper value for cost inflation. Whether adjusting the cost of a recently built project to predict what it might cost to build a similar project in the near future, or answering a client question, “What will it cost if I delay my project start?”, the proper value for inflation (which differs by sector and differs every year) can make or break your estimate.
Contractors responsible for a particular building material, although the PPI Inputs will not track market conditions sale prices from producer to the contractor, can get some indication of whether material prices are rising or falling. Contractors should be aware of PPI trends to interpret the data throughout the year.
PPI TRENDS HELP TO INTERPRET THE DATA
- 60% of the time, the highest increase of the year in the PPI is in the first quarter.
- 75% of the time, two-thirds of the annual increase occured in the first six months.
- In 25 years, the highest increase for the year has never been in Q4.
- 60% of the time, the lowest increase of the year in the PPI is in Q4.
- 50% of the time, Q4 is negative, yet in 25 years the PPI was negative only four times.
So when you see monthly news reports from the industry exclaiming, “PPI is up strong for Q1” or “PPI dropped in the 4th Qtr.” it helps to have an understanding that this may not be unusual at all and instead may be the norm.
I’ve read several articles recently describing, Why 2018 could be a boom year for construction spending. Several reasons being given to support a potential boom, when we look a little deeper, actually may not be good indicators at all to predict the trend for a strong year in 2018. In my Fall Forecast I do predict 8% growth in 2018 construction spending, but let’s take a look at what gets us there.
Data that doesn’t tell us much about the future trend in construction spending.
Jobs increased in 2017 up 35% over 2016. In 2017 construction added 210,000 jobs, growth of 35% over 2016, but in 2016 jobs growth decreased by 55% from 2015. 2016 growth was the lowest in 5yrs. In 2013 jobs growth increased by 85% and in 2014 by 71%, but in 2015 and 2016 jobs growth slowed. Yet 2015 was one of the best construction spending years on record. And in 2017, jobs growth increased over 2016 but spending growth slowed. The direction of jobs growth is not an indicator of the future trend in spending.
Nov 2017 spending was higher than expected, and YTD is up 4.2%. This is a slippery slope. Actually we won’t know any particular monthly spending until several months after the initial release. All monthly spending values are subject to revision three times after initial release. However, residential spending is higher than expected for the YTD and nonresidential buildings spending is below expectations for YTD. But more importantly, construction spending normally fluctuates. For instance, in the 2nd half of 2015, spending was down 4 out of 6 months, lower than forecast three times, posting a total decline of 2.5%. Yet 2015 finished the year up 10%. Then, in the 1st half of 2016, spending was up 5 out of 6 months, far exceeding forecast 3 times, posting a total increase of 6% in 6 months. 2016 finished up 6.5% for the year. Neither half performance predicted final results within the year or the forecast for the future. Furthermore, after inflation, 2017 spending is currently flat with 2016$, so all we are seeing in the 4.5% spending growth in 2017 is inflation. Current and past spending is not an indicator of the future trend in spending.
What data does give an indication of the future trend in construction spending?
Construction Starts (Dodge Data & Analytics DDA), Backlog, Cash flow from Starts, the Architectural Billings Index (ABI), The Dodge Momentum Index (DMI) and New Residential Permits and # of Units Construction Starts all give an indication of the future trend in spending.
Residential Permits and # of new units started gives a fairly immediate indication of residential activity. The ABI gives an indication of nonresidential building to start construction about 9 months out and the DMI about 12 months out. The ABI and DMI give some indication as to whether future starts will increase or decrease. DDA Starts give an indication of the percent growth in future work, but not when the spending will occur, so cannot be used directly to predict spending. A good example is the new start for airport terminal work recorded as a new start in 2017 at $4 billion. But it may take 5 or 6 years to complete that $4 billion project and only cash flow will show the impact on spending.
Care must be taken to use Starts data properly. It is regularly misinterpreted in common industry forecasting articles. Starts dollar values represent a survey of about 50% to 60% of industry activity, therefore Starts dollar values cannot ever be used directly to indicate spending. Also, Starts do not directly indicate changes in spending per month or per year. Only by including an expected duration for all Starts and producing a forecast Cash Flow from Starts data can the expected pattern of spending be developed. Finally, it is the rate of change in Starts Cash Flows that gives an indication of the rate of change in spending.
Cash flow is the best indicator of how much and when spending will occur. Cash flow from DDA starts gives a prediction over time of how spending from each month of previous starts will occur from all projects in backlog. Cash flow totals of all jobs can vary considerably from month to month, are not only driven by new jobs starting but also old jobs ending, and are heavily dependent on the type, size and duration of jobs.
Of course, data highlighting demand, occupancy rates, labor and material trends and other economic factors affecting construction trends all weigh into determining future spending expectations. However, for nonresidential buildings and infrastructure approximately 75% to 80% of all spending within the year comes from starting backlog. Most economic factors that will have an affect on spending within the year are already captured in projects that have started and are in current backlog. On the other hand, new residential starts are more important. 70% of all residential spending in the year comes from new starts.
The following trend predictions are developed based on using this outline.
Seldom do two sources present information the same way!
In the construction industry, a disconnect exists in the reporting of construction starts data and actual spending data. Problems may arise when data is used to perform comparisons or forecasts between starts and spending. New starts and backlog may be listed in one category and spending for the same markets may be listed in another.
Almost universally, reporting of actual construction spending data follows the U.S. Census Put-in-Place Spending format. I adjust all other construction starts input/forecasting data that I use to conform to these Census Construction Spending Put-in-Place definitions. Here are some pitfalls to be aware of:
Residential spending $ includes about 35% renovations and improvements that has no units associated with the dollars, so that portion of $ should not be included in a comparison to housing starts.
In census spending, MF dormitories is in educational and all types of MF healthcare related homes are in healthcare.
Demolition is not included in renovations/improvements. Partial repair of flood damaged homes is NOT included in residential improvements. Full replacement of flood damaged homes is included as improvements, not new single family. Here is the US Census definition of flood repairs
Offices includes pubic buildings such as city halls and courthouses. Includes data centers and bank buildings. Excludes medical office buildings, offices at manufacturing sites and offices at educational or healthcare facilities. Excludes Public Safety.
Commercial includes all retail buildings, warehouses, parking lots and garages. Excludes parking at educational/healthcare facilities.
Census DOES separate the costs for buildings that are mixed use retail/office/residential.
Educational, along with K-12, includes administrative offices, health centers, parking, residence halls, classrooms, educational research labs, food service and sports/recreation facilities at schools or colleges and universities and all associated infrastructure and maintenance facilities at the educational site. Also includes public libraries, science centers and museums.
Healthcare includes similar support and infrastructure to educational. Also includes medical office buildings, non-manufacturing and non-educational research labs.
Amusement and Recreation includes performing arts centers, civic centers, convention centers, sports and recreation facilities not located at schools or colleges.
Transportation includes air freight and passenger air terminals, runways, bus and railroad passenger terminals, light rail and subway facilities, railroad track, railway structures and bridges, docks and marine terminals and maintenance facilities and infrastructure associated with each.
Some sources of design or new construction starts data carry terminal buildings as commercial buildings, institutional buildings or other public nonresidential buildings. Census caries the building cost of all terminals grouped in with the non-building infrastructure costs of Transportation. Some sources carry public buildings such as city halls and courthouses as Public Safety but Census carries cost data for public buildings such as city halls and courthouses in Offices. Some sources classify laboratories as commercial and warehouses as industrial/manufacturing but Census includes warehouses in Commercial and Labs, depending on use, can be either Educational, Healthcare or Manufacturing.
Dodge Data New Construction Starts
Dodge includes monthly New Construction Starts for Terminals and Courthouses in Other Institutional Buildings, a Nonresidential Buildings category. The Census actual spending report includes Terminals in Transportation and Courthouses in Offices.
Although all of these still remain in Non-building Infrastructure, Dodge includes Rail, Mass Transit, Airport Runway and Pipelines in Other Public Works. Although not often mentioned by Dodge, it is assumed Communications is also included in Other Public Works. Census includes all mass transit in Transportation, Communications is listed separately and pipelines are included in Power.
Dodge does not identify Renovations in their residential starts data. Census reports SF and MF spending and Total Residential spending with the difference between Total and SF+MF being Renovations. Multifamily spending accounts for less than 15% of all residential actual spending. Dodge MF starts account for 30% of all residential starts dollars. Furthermore, Dodge totals for MF starts $ for the last 7 years exceed the actual total of MF spending for the year by 30% to 50%. Dodge MF data represents more than just MF starts. Which may mean it includes renovations starts. It might also include student housing.
Constructconnect (CC) Construction Starts Forecast
New starts for Transportation Terminals is in a line by the same name but subtotaled in Commercial (Nonresidential Buildings) starts. Census includes Terminals in Transportation.
CC lists Courthouse starts subtotaled in Institutional. Census carries Courthouses in Office (Commercial).
CC lists Military as a line item subtotaled in Institutional. This might include Office, Housing, Warehouse, etc., which would be carried by Census in Office, Residential, Commercial, etc., respectively.
CC lists Laboratories (Schools & Industrial) together and subtotals all labs in Commercial. Census separates labs by commercial, research and educational and carries spending in Manufacturing, Healthcare or Educational respectively which would subtotal spending in Manufacturing (industrial), or Institutional (Healthcare and Educational).
CC does not list rail or transportation separately, but does list Airport and Misc Civil (Power,etc.). This leads me to think rail is included in the line item with Misc Civil (Power, etc.). Also, CC does not list Communication, which I suspect is included in Misc Civil (Power, etc.) Already noted above is that Terminals is subtotaled in Commercial. Census carries rail, runway and terminals in Transportation and keeps Communication and Power separate from others.
CC provides an alternate table of new starts data that corresponds to a proprietary software, INSIGHT. This table of starts data reshuffles categories very far from anything that would resemble Census spending output.
The AIA publishes a twice annual Consensus Construction Forecast, comparing forecast of Nonresidential Buildings spending using inputs from seven or eight firms. Every firm but one follows a similar organization. The difference is FMI includes both Transportation and Communications in Commercial Nonresidential Buildings. I’m not aware of another other firm that reports these two categories of spending as Nonresidential Buildings. Both are typically carried as Non-building Infrastructure. That these categories include costs for projects such as rail beds, rail right-of-way civil structures, loading platforms, airfield runways and support structures, communication transmission lines and cell towers supports the more standardized inclusion of these items in Infrastructure.
Similar discrepancies may exist when comparing starts or spending to indexes, such as the AIA Architectural Billings Index, which broadly classifies projects as commercial, institutional or residential. Some resources classify Amusement/Recreation as institutional and some as commercial. In particular, the shifting of costs between Nonresidential Buildings and Non-building Infrastructure creates a particularly meaningful disparity between spending forecasts.
As you can see, there are numerous instances where the data are often mixed up. From the point of view of the forecaster, initial input data cannot always be used directly to forecast or match spending output. Some manipulation of the data may be required to make input and output match.
As an example, I move the Dodge data starts for Terminals from nonresidential buildings to non-building infrastructure Transportation, so that really changes my totals from theirs for Nonresidential Buildings to Non-building Infrastructure. My spending output conforms with most all others, most of whom also follow the Census PIP definitions.
What does your source for data take into consideration? Know your data!
Is Infrastructure construction spending near all-time lows? This question is raised because I saw comments to this affect recently posted on a major national construction professional organization twitter feed.
First, this raises several other questions:
- Exactly what construction markets are being referenced as infrastructure?
- Does this reference include public work only, or both public and private?
- Are educational and health care being included as infrastructure?
- Does this reference constant inflation adjusted spending?
The construction markets typically referred to as infrastructure, in order of largest to least volume, include; Power, Highway, Transportation, Sewage/Waste Water, Communications, Water Supply and Conservation. Sometimes also considered are Educational (3rd after Highway), Healthcare (after Transportation) and Public Safety (2nd smallest).
If only public work is included, everything changes. Most (90%+) of Power spending is private, so it represents less than 3% of public work. The largest contributors in this case are: Highway (32% of public work), Educational (25%), Transportation (11%), Sewage (8%) and Water Supply (4%). No other market is greater than 3% of public work.
And finally, is the reference to current dollars as originally spent within each year, or to constant inflation adjusted dollars, adjusting all historical expenditures to constant 2017 dollars? Any comparison to determine if real growth has occurred should be in constant dollars, in this case all adjusted to 2017.
Typical infrastructure, not including educational, healthcare or public safety, but including all public and private sector work produces this result:
However, the most likely reference is to typical public infrastructure, not including educational, healthcare or public safety. This scenario includes only the public sector work of typical infrastructure and eliminates private spending. This eliminates 90%+ of all power work, 30% of transportation and 100% of communications, in total, more than $100 billion in current dollars. This is the result:
In both instances, the lows, whether using current or constant dollars, occurred between 1993 and 2004. The highs are recent, all occurring from 2007 to 2016. 2017 spending dropped somewhat from 2016, but this is still prone to revision, which is always up.
To answer the question, Is Infrastructure construction spending near all-time lows? NO! Infrastructure construction spending is not at or even near all-time lows. Public sector infrastructure is lower than All infrastructure, but All infrastructure is not even near recent lows. It is near all-time highs!
Infrastructure construction spending in June-August dropped to the lowest since November 2014. However, this was not unexpected. Cash flow models of infrastructure starts from the last several years show monthly spending dips and peaks. Current dips in spending are being caused by uneven project closeouts from several years ago. The actual current backlog is at an all-time high and spending will follow the expected cash flow.
Infrastructure starting backlog hit a new all-time high in 2017 and will again in 2018. Public Infrastructure new starts reached all-time highs in 2013 and 2015 and are on track to go higher in 2017. 80% of infrastructure spending within the year comes from backlog at the start of the year and that backlog may be comprised of jobs one, two, three and even four years old.
Infrastructure spending in 2017, although down slightly from the all-time high reached in 2015 and nearly equaled in 2016, will reach a new high in 2018.
(This analysis does not include any spending projections from an infrastructure investment bill).
Highway spending is currently benefiting from projects that started in 2015 but that have unusually high value and long duration. They contribute spending well into 2018 beyond the duration that typical projects have ended.
Transportation Terminal starts in the first three months of 2017 were more than three times higher than any three-month period in the previous five years. However, 2017 spending is still affected by uneven starts from two to three years ago, holding down gains in the 2nd half. Transportation will show only a 1% gain in 2017 but produces double digit gains in 2018.
Infrastructure construction spending is near all-time HIGHS and has been for the last several years. That is not meant to indicate there is no need for infrastructure investment. I think the need is well established, particularly for public infrastructure. However, I’ve been writing about infrastructure for more than a year, pointing out the level of activity in this sector and the difficulty that will arise when we try to increase work volumes. The approach to adding new work and the discussions surrounding this approach should reference accurate data, and that should include an accurate representation of current workload and future ability to absorb more work.
For much more in-depth related to infrastructure construction see this post Infrastructure Spending & Jobs
You know those articles you’ve been seeing, “Worst year for construction spending since 2010″, well there’s some truth to that, BUT
2017 is the 6th year of the expansion. It has slowed, but… Here comes the BUT!
10-4-17 – Construction numbers are at all-time highs! Slowing or not, activity is very strong. Looking behind the headlines, here’s what we see;
Residential construction spending is slowing the most, from +11% in 2017 to only +2% in 2018 after six years averaging 13%/yr. Nonresidential buildings spending this year just kept up with the rate of inflation (4%), none-the-less, it’s at record highs. It doubles that rate of growth to 8% in 2018. Non-building infrastructure, down 2% in 2017, next year expect growth of 10%+, coming from long duration jobs.
The real performance numbers in Infrastructure are completely hidden. Spending was near flat for three years. But during that time, contrary to every other sector which experienced inflation of 15%, Non-building Infrastructure experienced deflation of 7%. (Gee, didn’t I read somewhere that activity within a sector is a primary driver of inflation?) Anyway, flat spending means volume really increased by 7% during that time. Spending by itself never tells the whole story!
There were some expected dips in spending recently, Manufacturing, Power, Highway, and there will be more in early 2018. BUT, there are also expected boosts in spending, Office, Commercial/Retail. Some of these already have matched up with the forecast, and there are more to come in 2018, Power, Transportation.
All Nonresidential Backlog is at record highs.
Buildings and Infrastructure will both hit new all-time highs for starting backlog in 2017 and again in 2018. For four years, from 2010 to 2013, all nonresidential backlog remained fairly constant. Since then, backlog for infrastructure is up 30% and for buildings it’s up 60%. (75% to 80% of nonresidential spending within the year comes from backlog at the start of the year. For residential, 70% of spending comes from new starts within the year.) Buildings will hit spending records in both 2017 and 2018. Infrastructure spending will hit a new high in 2018.
Ignoring for the moment that comparing any month to the same month last year can be grossly misleading as to the direction the markets are headed (for reasons explained in other recent posts on this blog), 2017 total spending growth is the lowest % yr/yr growth since 2011 (not 2010). Does that make it “worst”?
Spending will gain +5.6% in 2017, the least gain in six years. Last year was +6.5%, 2013 was +6.6%. The average for the last six years is +8%. So 2017 is the worst. Pretty damn good worst!
Construction spending for July was released yesterday, posted at $1.211 trillion, down 0.6% from an upwardly revised June. This is the sixth time in seven months of 2017 in which the initial release for monthly spending is down from the previous month. This is actually a very normal occurrence.
The 1st release of monthly spending vs the previous month has been down 15 times in the last 21 months. This may be what leads some analysts and pundits to write that construction spending is heading to recession. Nothing could be further from the truth!
For the last 21 months, in which 15 first reports showed a decline vs the previous month, 18 of the monthly values were revised up. After revisions, only five months remain down vs the previous month. Seven months are still pending further revisions, almost always up.
Construction spending is highly prone to revisions. After the 1st release it is revised each of the next two months and once again the following year. Spending has been revised UP 48 of the last 52 months, 92% of the time. The average upward revision for the last five years is +3.2%/month. In the last 52 months the upward revision averaged 3.7%.
Construction spending revisions after first release of data:
- Total Construction UP 48 of last 52 months, avg 3.7%/mo
- Total Construction UP 16 of last 18 months, avg 2.6%/mo
- Residential revised UP 29 of last 30 months, avg 7.0%/mo
- Residential UP 17 of 18 avg 3.8%/mo
- Commercial UP 17 of 18 avg 6.0%
- Educational UP 14 of 18 avg 2.2%
- Power UP 18 of 18 avg 12.0%
- Commercial/Retail May +3.9%, June +2.6%
- Lodging May +3.8%, June +1.1%
- Educational May +2.8%, June +3.6%
- Transportation May +3.6%, June +2.3%
January through May values have already been adjusted twice in these reports. June has one more revision next month and July gets revised twice. It’s quite likely both June and July values go up. All 2017 months still get one more revision next year when the May data is released (July 1). The post-annual total revision for the last 15 mo averages +2%, close to the long term average. First release values are ALWAYS being compared to previous values that have already been revised, 92% of the time UP. So first release values almost always understate performance. Since July 1st 2017, all 2016 monthly values have been revised three times so monthly releases this year starting with May have the most understated initial % comparison year-over-year because an un-adjusted release is being compared to a 3x-adjusted value.
When judging performance of monthly spending, it is reasonable to predict spending will get revised UP from the first release. Therefore, the most immediate monthly analysis you read, if based on initial release, 92% of the time is under-stating the performance of construction spending.
Construction spending forecasting not only must rely on performance year-to-date, but also on predictive analysis of how much revision there may be to current values. As an estimate, if monthly spending is initially posted as 2% down, 18/mo.averages indicate it will end up at least +2.6% higher after revisions, so would be a positive 0.6% growth month.
A few closing points:
Construction Spending 1st release for July is $1.211 trillion. Expect this to be revised up. YTD Jan-Jun revisions are UP 1.8%. Historical revisions last 5 years predict the final July value will be up 3% from the 1st release.
Construction Spending AVG 2017 Jan-Jul YTD ($1.226tr) has reached an all-time high. We’ve now posted three consecutive quarters of spending all averaging above $1.220 trillion. Spending is on track to total $1.250 trillion for 2017, up 5.5% over 2016.
Construction Spending avg YTD = $1.226tr, is up YTD 4.7% with revisions through May. Without revisions, the 1st releases would have averaged only $1.208tr, up only 3%.
Commercial Retail, Office and Residential lead 2017 construction spending gains, all over 10%. Office spending is at a record high.
After 5 months of stalled construction jobs growth, August added 28,000 jobs and put 2017 growth back on track towards 250,000 jobs. YTD is up 135,000. March thru July added only 19,000 construction jobs. Jan+Feb added 88,000, ending a six-month period, Sep16-Feb17, that added 167,000 jobs.
Harvey related jobs will be muted by jobs lost, I suspect for at least two months. There will be a period of slack records that will take some time to see the real effects of Harvey.
Further reading on this topic
Here’s some headlines this month on the June Construction Spending release: Plummets in June; Largest one month drop in 15 years; Clearly Decelerating; US Construction Spending Just Collapsed; and my personal favorite, Construction Spending Plummets to Economic Crisis Levels.
Frankly, I have much more trust in my data than to suggest we are at crisis levels.
In the latest Census construction spending report, June spending dropped 1.3% from May, but May was revised down -0.7%. The consensus of economists predicted spending would be up +0.5% (from the original May value), so the data posted is actually 2.5% below consensus estimates.
I expected May to get revised up 0.6% and the initial June release would be flat vs the revised May value. So the actual came in 2.6% below my expectation.
June construction spending was posted at $1.205 trillion, down 1.3% from May and down 2.7% from March. With the revised data, the May Year-to-date (YTD) vs 2016 was only +5.5% (not +6.1% as initially reported) and for June it’s now +4.8%.
My opinion is this preliminary June value appears suspect. This is sort of like driving a well maintained car that gets 30 mpg and all of a sudden the gauges indicate 20 mpg for the latest tankful of gas. Although the road may be a little bumpy, there does not seem to be any serious mechanical problems, so we have to ask, why did gas mileage drop so much?
The April decline and the Apr-May-June decline are the single largest monthly and 3-month total non-recessionary declines on record. We would need to look at recession data to find similar declines. Spending drops like this just don’t normally occur, especially when cash flow patterns from starts predict 4% growth during the 3-month period. That’s a 6.7% miss over 3 months.
The largest declines in the June Seasonally Adjusted Annual Rate (SAAR) construction spending were Highway and Educational, together 60% of the total monthly decline. (There are other markets with greater mo/mo% declines, however most of those markets have a very small share of the total spending so don’t amount to much). Almost all of the largest declines are public work. In fact, the initial June release shows every public market declined. However, all ten other public markets together don’t equal half of the declines generated by these two major markets. Furthermore, for the past 3 months Highway spending shows a decline of 12.5%, and Educational spending is down 7.6% in 4 months. A review of data back to 2005 shows neither of these markets have ever had any periods where they’ve experienced declines of this magnitude. These would be record declines if they stick. Market trend data simply is not indicating to expect record declines at this time. So I consider these data suspect.
Construction spending initial release is always preliminary data. The June value, released August 1st, will be revised in each of the next two reports and then once again next year when all 2017 data is reviewed. The average revision to June spending data over the last 4 years (similar growth years to current expectations) is +4.8%.
There are three more opportunities for revision to the June data and two more to the May data. We will have a much better idea what really happened on October 1st, but we won’t know the final outcome until the final 2017 revision on July 1, 2018.
So, what data seems to indicate a trend contrary to current declines? The last 12 months of Dodge Data new starts for nonresidential buildings are the highest since 2008 and they peaked from August to October. Residential starts, at their highest since 2006, peaked from December’16 to March’17. Backlog is at an all-time high. There is no indication here that spending will plummet.
Also, one month of Educational or Highway new starts each generate about $250 to $300 million per month in spending, for the next 24 to 36 months. Normally, with some variation, we have the current month of new starts coming into backlog and one month of old starts ending. Since starts have been normal or high recently, the spending declines posted in June would imply that we’ve lost two to three months of backlog from current spending. Again, there are no indications that we have an extreme imbalance or a canceling of backlog.
Most of the nonresidential spending occurring right now is from projects that started between mid 2015 and the end of 2016. Nonresidential buildings projects that started in 2015 or earlier still make up one third of the spending in the 1st half of 2017. Non-building infrastructure projects that started in 2015 and earlier contributed 50% of spending in the 1st half of 2017. Residential projects have shorter duration so most spending is from more recent jobs, but we hit a 10 year peak in new residential starts just a few months ago. All sectors have fluctuations in spending and have down months but the index of long term cash flows out to completion shows normal backlog and spending growth across every sector.
I’m inclined to expect substantial upward revisions to June construction spending in the next two releases. No other data supports a big June drop.
Keep in mind, current construction spending is always being compared to previous months revised spending and growth is almost always being understated. Monthly spending has been revised UP 45 times in the last 48 months. All previous months and all 2016 data have been revised several times. The average revision to ALL spending data over the last 4 years is +3.9%/month. Since January 2016, the average revision is +3.0%/month. The average revision to June spending data over the last 4 years is +4.8%.
June data is un-adjusted preliminary data. Many of the news articles declaring construction spending was a miss are based on this preliminary data which very often gets revised away in following months. For example, The 1st 6 months of 2016 have already been revised up, three times each, by a total of 2.5%. All the months YTD in 2017 still have pending revisions. June 2017 vs June 2016 shows a percent growth of only +1.6%, but June 2016 has already been revised up by 4.7% and June 2017 has not yet been revised at all. June 2017 has a 90% chance of being revised up.
I predict after all the revision are in we will see that June spending did not drop to a low of $1.205 trillion, but that it was closer to $1.250 trillion.
7-6-17 Construction Spending May 2017 – Behind The Headlines
Headline – Construction Spending for May came in flat compared to April, up 4.5% vs May 2016.
In this latest May report, April spending was revised up by 1% and May 2016 was revised up by 3%. The average revision since Jan 2016 is 3%/month. May 2017 will be revised in each of the next two reports and again with the May report issued in July 2018.
Current unadjusted construction spending is always being compared to previous months revised spending and growth is almost always being understated. Spending has been revised UP 45 times in the last 4 years.
In 2016, the 1st report indicated monthly spending declined 8 times from the previous month. After revisions, spending declined only twice from the previous month. Most MSM articles declaring construction spending was a miss are revised away in following months.
Nonresidential Construction Spending Remains Stagnant in May.
I’ve said this before many times, spending predictions are best tracked based on cash flows from all projects that have started. This is not simply tracking total backlog, nor is it tracking new construction starts. New starts (new backlog) represent only 20% to 25% of total spending within the year. Most spending comes from projects that started in previous years.
Big monthly changes in spending come from unusual fluctuations in starts. Very large projects ending (spending ending), compared to new projects starting, would cause a monthly drop in spending. The reverse would cause an increase. If a record volume month of construction projects that started two or three years ago are now reaching completion, and new starts today are experiencing normal growth not at record levels, then spending will most likely decline temporarily. Most monthly construction spending predictions are predetermined months ago.
Also, Nonresidential construction is comprised of two very different sectors, nonresidential buildings and non-building infrastructure. Infrastructure is quite erratic while buildings spending has been climbing at a steady strong rate for several years. Buildings spending is up 2% from Q2’16 and up 6% YOY. In the 2nd half 2017 YOY spending is expected to reach 8%.
Most infrastructure projects that started in 2015 and 2016 are still ongoing so do not effect much change in current monthly spending. It is projects from late 2014/early 2015 that are finishing that are resulting in the largest share of current spending drops. Worthy of note is that non-building infrastructure spending just experienced two years of record highs, so even though spending is down slightly we will still see 2017 finish near record highs.
Construction Companies Continue to Face Labor Shortage Challenges
Construction Spending for the last 24 months increased +13%, but after inflation actual volume during that period increased only +5.5%. Construction jobs output, (jobs x hours worked) for that same period increased +7.6%. Overall, jobs output is exceeding the growth in volume put-in-place. Most of this is being driven by imbalances in Nonresidential Buildings, for which jobs output grew by 7% in two years but volume growth measured only 2% after inflation.
Why is it that jobs output is growing faster than construction volume? Could it be that shortages are localized, not as widespread as thought? Or perhaps it’s that contractors can’t get skilled workers, so they are hiring more workers with less skill? Maybe contractors anticipate growth, so they are hiring more now to prepare for the future? Whatever the case, jobs are growing faster than construction volume and that is not what should be expected in a labor shortage.
Are contractor’s responses to survey questions about filling job positions based on an anticipated need to staff up to meet revenue growth? If so, that is a major miscalculation to determine staffing needs. This is not as far-fetched as you might think. I’ve talked with numerous contractors in the past who were doing this. As I tried to explain in several previous articles, growth in revenue (or construction spending) doesn’t address how much of the growth is due to inflation. Right now, in fact for the last 24 months, the largest portion of spending growth is inflation, not real volume growth.
If you are hiring to match your revenue growth, you are part of the reason jobs are growing faster than volume. INFLATION!
Is there a Residential Construction Spending slowdown? If so, how significant?
YTD Residential Construction spending for the 1st 5 months 2017 is up 12.2% from 1st 5 months 2016. YTD has been above 12% since January.
Average spending for the last three months is up 4.0% from the average in Q4 2016. That’s a ~10% annual rate of growth. Starts cash flows are indicting flat spending for the next few months but then accelerated spending from late Q3 into the end of the year. Current projected spending for 2017 is $523 billion, +10.5% higher than 2016.
May vs April residential construction spending shows a 0.5% decline. However, April has been revised up once and May has not yet been revised. All months are revised twice after the first release of data. The average revision (to residential data) for the last 16 months is up 4%, the average revision for the last 28 months is up 7%. All revisions for the last 28 months were up. After revisions, there were only two monthly declines in the last 28 months, and both of those were slight.
If new starts collapse to show no gains for the remainder of the year, then based on starts already in backlog and reduced starts for the remainder of the year, spending would be reduced to $513 billion. That’s still 8.5% higher than 2016. Of course, this would be an extremely unlikely scenario. The last time residential construction starts declined for three or more consecutive months was 2010, and the last time there were no gains for six or more months was 2008.
April construction spending 1st release was issued on 6-1-17 by U. S. Census. The initial release shows April DOWN 1.4% from March, a value many news sources have reported as “construction spending is slowing”, “one of the largest drops in six years”, “an unexpected slump”, “spending continued to demonstrate substantial weakness.” I’ve written about this numerous times but it’s worth repeating again. Construction spending almost always gets revised UP in the following month after 1st release. Average revision so far in 2017 is +1.8% and for the last 18 months +1.3%. Monthly construction spending has now been revised UP every one of the last 43 consecutive months.
Headlines of construction spending declines are almost always premature.
Construction spending is almost always a miss when first posted, until it gets revised up in the following monthly report to show is it almost never a miss.
The 1st release of March construction spending came out May 1. This initial release indicates a decline of 0.2% from February. Keep in mind, all 12 monthly reports in 2016 were subsequently revised up. Nine times in the previous 14 months, the 1st report of spending was down vs the prior month. After revisions, only three months were down compared to the prior month.
In the last 48 months, the 1st report of spending was down vs the prior month 20 times. 47 times the initial value was revised UP. After revisions, only nine months were down compared to the prior month.
Monthly construction spending has been revised UP every one of the last 42 consecutive months.
The 1st release of spending is almost always being compared to a previous month and a previous year that have been revised up. Upward revisions to monthly construction spending in 2016 have been as high as 3.4% and for the year average 1.1%/mo. So, a 0.2% mo/mo decline s probably not a decline at all after revision, and there will be a revision, most likely UP.
After spending is first published it is revised in each of the two following months. Then all the values for the entire year are revised with the May data release the following year.
Some specific markets construction spending revised after 1st release (2016 data). These markets represent almost 50% of nonresidential data.
- Office revised UP 8 of 12 months (average of all 12 +1.1%)
- Commercial UP 9 of 12 avg 1.8%
- Educational UP 10 of 12 avg 1.8%
- Power UP 12 of 12 avg 3.6%