For up-to-date Outlook
See Pandemic #13 – Midyear Construction Outlook published 8-14-20
See Pandemic #12 – Jobs & Starts Updated Aug 8 and
Pandemic #11 – June data Construction Spending Update Aug 3 and
Pandemic #10 – June New Construction Starts July 30
Analysis of a Recession Scenario 2020-2022
based on 6-1-2020 data
This analysis relies on the first available hard data reports to forecast the impact of a construction recession scenario caused by the current Covid-19 pandemic. This scenario does not assume a catastrophic failure of the U.S. economy, but does assume a large decline in construction activity in 2020 and 2021. Data always gets revised in the following months.
This analysis generates spending cash flows from current and assumed reduced new construction starts to then determine how new starts and spending may affect future construction activity.
6-1-20 What We Know Today
The revised construction spending report for March, released June 1st, showed no decline in spending from February. Census revised March construction spending from the initial report of up 1% to, still remarkably, level with February. https://census.gov/construction/c30/pdf/release.pdf
This was the estimated growth before the pandemic. Shutdowns began March 15th. This report seems to indicate construction was totally unaffected in March! This is hard to grasp. Just the shutdowns in Boston and San Francisco starting Mar 15. affected 300,000 jobs and would calculate to have erased over $2 billion in work from March total spending. By one estimate, we had already lost 300,000 to 400,000 construction jobs by Mar.31st. How spending could have increased is baffling.
The first hard jobs data reflecting Coronavirus impacts on the construction industry was the jobs report released in the first week in May which covers jobs from mid-March through mid-Apr. From March 15th to April 12th, construction lost 975,000 jobs, 13% of the workforce. Construction unemployment, which was below 5%, as low as 3.2%, is now at 16.6%. https://bls.gov/web/empsit/ceseeb1a.htm . I expect to see downward revisions to March spending. In addition to a 13% drop in the number of jobs, hours worked also dropped 3%, so the total worker manpower and hours worked output dropped 16%.
Ten states account for 80% of the 975,000 construction jobs lost between March 15th and April 12th. CA, FL, MA, MI, NJ, NY, OH, PA, TX & WA. Eight of those states are in top ten states for total number of construction jobs, MA and WA are not. https://www.agc.org/news/2020/05/22/construction-employment-shrinks-49-states-and-dc-april-new-association-survey-finds?
The next jobs report due out June 5th covers the period from April 13th through May 17th. Expect more downward movement in the jobs numbers. I expect most construction jobs were lost early and are included in the April report, so next report June 5th we may see only 400,000 more construction jobs losses for the period April 13 through May 17. (This report will be edited to reflect the June 5th jobs data.)
The loss of 975,000 jobs in a single month (if all jobs were lost for a full 30 days) at a rate of 60,000 jobs needed to put-in-place $1 billion in construction in one month, equates to a loss of $16 billion in spending between March 15th to April 12th.
In one of the most baffling reports I’ve seen, Census posted April Construction Spending down only 3%. It is beyond explanation that construction spending for March was flat and April declined only 3% from March, while we’ve lost 16% of worker output.
The change from March to April Construction Spending versus Construction Jobs varies by the most in 30 years. This is highly suspect! Jobs (and hours) are down 16% but spending is down only 3%, for a variance of 13%. The largest variances on record back to Jan 1991, most of which occurred in 2005-2006, are: in 360 months, jobs and spending growth varied by >4% only 10 times, >5% only twice. The 20 yr average 0.1%.
Variances like what the data show for this month have never occurred. A more likely explanation is there is an anomaly in the data. Either construction did not lose 975,000 jobs in a month or spending in April dropped a lot more than 3%. I suspect the later.
6-8-20 Another way of looking at the jobs data is this simple metric. The total number of construction jobs divided by the amount in $billions of work put-in-place each month = Jobs/$billion PIP. This value sometimes spikes or dips by 3% or 4%, but usually changes month to month by less than 1%. You can easily see the April construction jobs/spending data changed by what normally would take several years. The change in April spending data does not agree with the change in jobs data. It’s one thing to see a spike like this when jobs are increasing in excess of work put-in-place. But it is much more suspect to see jobs decline without a decline in work put-in-place. Who’s putting the work in place?
Here’s what that April variance would mean. If construction jobs+hours drop by 16% but spending drops by only 3%, then productivity as measured by amount of work put in place per job increased by 13%, IN A MONTH, DURING A PANDEMIC.
If there is essentially no change in productivity, but due to restrictions and higher overhead, costs go up, then you must accept construction inflation just increased by 13% in one month. But there is no data yet showing unusual labor or material cost increases.
Or, if some portion of this discrepancy is due to a change in productivity and some part due to rapid inflation, the 13% variance is somehow split between these two issues.
6-5-20 The May Construction Jobs report issued today for the period April 13 through May 17 shows an increase of 464,000 jobs (and hours worked). I missed that estimate. So did everybody else. Jobs and hours worked output is now down net 13% for March and April. Spending is down only 3%. I still expect April spending to be revised down.
6-5-20 The May jobs data today is indicating an initial month (April) of stop work down 975,000 jobs (16%) is now down only 10%, net 600,000 jobs as of mid-May. That’s equates to a two-month average of jobs and hours worked down by 13%. My forecast was based on about 25% of work stopped for two months. If jobs is a better indication, the lesser extent of work stopped means the adjusted forecast will post higher spending in 2020 and will move less spending into 2021. The next important data date is July 1, when the May construction spending is released. With that May report every year, Census revises all previous months data back 5 years. We’ll see if April data gets revised down.
6-9-20 “In total, the construction sector recorded almost 1.3 million layoffs during March and April.” NAHB https://wordpress.com/read/blogs/68887359/posts/28427
6-5-20 There is a scenario in which jobs decline with little to no reduction in volume. But, it accepts that there was a very large number of nonproductive jobs in the workforce. I’d been writing about the increasing disparity for the last few years between volume growth and jobs growth Expect Construction Jobs Growth to Slow in 2020. Since 2017 jobs have been increasing but volume of work has been decreasing or level. It’s possible a large portion of the jobs losses in April are resetting this balance with little to no impact on volume output. These would be jobs that would probably not come back.
6-5-20 For Mar-Apr, 975,000 jobs were lost for a single month (if all jobs were lost for a full 30 days). For Apr-May, we regained 464,000 jobs. Net jobs loss for two months is about 600,000 jobs. April spending was down only 3% or $40 billion seasonally adjusted annual rate (SAAR) . At a rate of 60,000 jobs needed to put-in-place $1 billion in construction in one month, this equates to a loss of $10 billion in not seasonally adjusted (NSA) spending PER MONTH for both April and May. That would be a decline in the SAAR of spending of $120 billion. March posted no decline. April posted a $40 billion decline and we will get May spending on July 1st. The spending data remains suspect.
The pre-pandemic construction seasonally adjusted annual rate (SAAR) spending forecast was for a rate of $1,395 billion from March through July. Assuming no unusual changes in jobs, productivity or inflation, the spending can be calculated from the number of jobs available to perform work. For pandemic impact based on job losses, we should see March SAAR spending at $1,350 billion. April construction spending should drop to under $1,200 billion from a high of $1,386 billion in February. I expect to see March spending get revised downward. I expect April spending to be revised down 6% to 10%. For these reasons, my forecast is not using the reported April spending data.
Pandemic Impact on Construction
This was the baseline forecast pre-pandemic. It shows considerable strength in Nonresidential Buildings and Non-building Infrastructure starts and spending. There is weakness in residential.
This plot shows actual spending through January 2020 and forecast spending for 2020 through 2022 based on the predicted cash flow from existing starts and projected starts pre-pandemic.
PRE-RECESSION SCENARIO FORECAST SPENDING
Backlog leading into 2020 was up 30% in the last 4 years, at all-time high. Although spending was forecast up only 4%/yr. the next two years, spending is at an all-time high. 80% of all nonresidential spending in any given year is from backlog. If new starts drop by 10%, that has only a 1.5% to 2% impact on total spending in the first year. The following year spending would be down 4% to 5%. Residential spending is far more dependent on new starts than backlog. Only about 30% of residential spending comes from backlog and 70% from new starts. If residential new starts drop 10% that impacts total spending by 7% in that year.
Residential construction starts peaked in 2018. Starts in 2019 are level yoy, but have been flat or in moderate decline since mid-2018. Spending was forecast up 5% in 2020 but down 1% in 2021.
Nonresidential Buildings starting backlog increased 10%/year for the 4 years 2017-2020. Starts have moved sideways or in slight decline since mid-2018. 2019 starts are down 9% from 2018. Spending was forecast up 3% in 2020 and 2021.
Infrastructure starting backlog, by far the most robust, has increased 15%/year for the 3 years 2018-2020. Spending was forecast up 6% in 2020 and up 8% in 2021.
Regardless what may lead to a construction recession, in this case a global pandemic, it is the current high amount of work in backlog that will work hard to mute its effect.
When a recession occurs, new construction starts would be substantially reduced. Although some projects will be canceled or delayed mid-schedule, most projects already in construction would move on to completion.
Construction projects will most likely experience delays. Potential product shortages, delivery delays and shutdowns will drive up costs and extend project schedules.
Projects in planning may be canceled due to drop in demand, decline in capital or slowdown in economy. Retail stores may cancel expansion, educational facilities may delay starting new construction, transportation facilities may postpone later phases of long planned growth.
This recession scenario does not assume a catastrophic failure of the economy.
This pandemic recession forecast is based on the following:
- April spending is forecast and not carried as reported due to reasons cited.
- New Construction Starts in 2020 canceled, Residential -15%, Nonresidential Buildings -8%, Non-building Infrastructure -11%, Total Construction starts canceled -11%.
- Work in backlog that has been delayed, minimum 2 month delay, restart build up over a period of 8 months; Residential -30%, Nonresidential Buildings -28%, Non-building Infrastructure -22%, Total Construction delays -25%.
- Work in backlog that has been canceled, Residential -3%, Nonresidential Buildings -3.4%, Non-building Infrastructure -2.2%, Total Construction backlog canceled -2.8%.
This plot shows the resulting change in spending. Only the estimated spending to the right of the dateline changes.
RECESSION SCENARIO FORECAST SPENDING
6-20-20 Data on most recent Construction Starts
Dodge Construction Starts average for Apr+May 2020 compared to Avg Jan-Feb-Mar 2020 Nonresidential Bldgs 66%, Nonbldg Infra 96%, Residential 73%.
Dodge Construction Starts average for Apr+May 2020 compared to Avg Apr+May 2019 Nonresidential Bldgs 82%, Nonbldg Infra 85%, Residential 97%.
So, while Nonresidential buildings starts were down 34% in Apr+May compared to Q1 2020, that’s down 18% compared to the same months 2019. Residential Apr+May starts are down 27% compared to the avg in Q1 2020, but that’s down only 3% compared to Apr+May 2019.
Nonbuilding Infrastructure starts in Apr+May are down only 4% from Q1 average, but Q1 is already 15% lower than last year.
What impact does that have on the year? if Nonres Bldgs starts are down 34% for 2 months, that reduces starts by 6% (34 x 2/12) for the year (from the prior trend). if Residential starts are down 27% for 2 months, that reduces starts 4.5% (27 x 2/12) for the year. Most of the impact in nonresidential occurs from 12 to 24 months out. Most of the residential impact occurs in the first 12 months.
After the resumption of work that had been halted, which for various reasons cited will take several months, and which is not all expected to return to a full level of pre-pandemic spending, a pull-back in new construction starts will hold spending nearly flat from Q3 2020 through Q2 2021. During that time nonresidential work will reach a post-pandemic peak but residential work will hit a post-pandemic low. Non-building Infrastructure work is not affected nearly as much and still shows spending growth leading into 2022-2023.
Construction Spending Forecast 2020 – Residential -2%, Nonresidential Buildings -4%, Non-building Infrastructure <-1%, Total Construction Spending 2020 -2.3%.
Residential construction spending would drop 11% from $566 billion to $506 billion in 2020 and then drop 18% from $550 billion to $453 billion in 2021. Residential is far more dependent on new starts within the year for spending than on backlog.
Nonresidential Buildings spending drops 8% in 2020 from $469 billion to $434 billion and then drops 2% from $475 billion to $468 billion in 2021.
Non-building Infrastructure spending drops 5% in 2020 from $348 billion to $330 billion and then drops <1% from $378 billion to $376 billion in 2021.
By 2022, nonresidential buildings and infrastructure are back within 1% to 2% of baseline pre-pandemic spending. However, residential spending is set back $100 billion, back to the level of 2016.
About 300,000 of the jobs lost do not return. While about 75% of the initial jobs lost have returned by the end of 2020, we then slowly lose jobs for the next two years.
Total all spending would drop from the current 2020 forecast of $1.380 trillion to $1.270 trillion. In 2021 and 2022, instead of baseline spending of $1.400 trillion, spending would drop to near $1.300 trillion, back to the level of 2018. The losses in the Great Recession, a total drop of almost $400 billion, set construction spending growth back 12 years.
The difference with shutdown vs a reduction in new starts is that work shut down is delayed. It will reduce total spending in that month in 2020 but will simply shift all remaining months and the end of the projects, which could occur in 2020, 2021 or later. If 50% of all U.S. construction shut down for two months, it would delay $100 billion worth of work, most of it from 2020 into 2021.
7-3-20 updated spending data on 7-1 Census released spending data for May and revised all monthly data back two years. 2018 and 2019 spending was revised up by 2% and 4%. But more important, the data for Mar-Apr-May 2020 shows that shut down delays were far less than anticipated. More than 90% of all work continued unabated through April and may. The result is fewer reductions in spending in 2020 and less delayed work pushed out into 2021. This shifts the balance of work to now show slight growth in 2020 and only a minor increase in 2021. SEE PANDEMIC #9
When we see spending increasing at less than the rate of inflation, the real work volume is declining. With typical construction inflation between 3% and 5% annually, a spending drop of -2.8% in 2020 may reflect a work volume decline of 6% to 8%. Spending growth of 2% when inflation is 3% is really a decline in volume of work by 1%. The extent of volume declines would impact the jobs situation.
Historically jobs declines of the same magnitude do not follow immediately after volume of work declines, therefore after the initial delayed job losses return to work, we would not expect to see much reduction in workforce in 2020.
What this will do to the construction inflation rate is hard to predict. Typically, when work volume decreases, the bidding environment gets more competitive and prices go down. However, if materials shortages develop, that would cause prices to increase.
There have been reports that scrap steel shortages may result in a steel cost increase and numerous imported products are not available or in short supply. Some firms that manufacture goods used in construction were closed temporarily, so their production was disrupted.
Steel Statistics and Steel Cost Increase Affect on Construction? 2/3rds of all steel manufactured in the U.S. is EAF steel made from scrap steel. Almost 100% of steel used in construction is EAF steel. The Pandemic is causing scrap steel shortages.
U.S. Steel production through May is down 13% compared to the same period 2019. Steel imports are down 20%. The U.S. imports about 30% of all the steel it uses. 40% of all steel used in the U.S. is used for construction. So through May, total steel available in the U.S. is 13% + (20% x 30%) = 19% less than 2019.
Add to these issues the fact that many projects under construction may have been halted for a period of time and many more may have experienced disruption. The delays may add several weeks to perhaps a month or two to the overall schedule and management cost goes up. I think in this case the materials availability issues and schedule delays will outweigh any decline in work available for bid.
Only twice in 50 years have we experienced construction DEflation, 2009 and 2010. That was at a time when business volume was down 33% and jobs were down 30%. I expect inflation to range between 4% and 5% for 2020 and 2021, lower for residential work.
You can visit Ed Zarenski’s website,
Construction Analytics – Economics Behind the Headlines
Ed….I love your analysis and have a quick question: Why does Nonresidential spending increase 7.8% in 2021? Is it the backlog you refer too?
I am the Director of Sales for MSA Safety and a lot of our construct business is tied to the OGP/LNG build out across the GC….do you have any insights you could share on that specific market?
Thanks in advance for your great insight into this very difficult to measure and predict market.
Bill, my forecast has been changing rapidly as more data comes in. A lot of the Nonres Bldgs increase in 2021 was because 2020 was predicted to drop ~5%. New data shows the expected drops in Mar-Apr-May did not materialize as expected. Shut downs did not lead to extent of drops in spending that were predicted. See article update Pandemic #9. Now forecasting Nonres Bldgs 2020 up ~1%, 2021 up ~1%.
OGP/LNG? Pipeline? Liq Nat Gas? Spending in Power markets declined only 7% from Jan-Feb to Apr-May, but Census does not break out the components of the Power markets. Dodge suggested new 2020 starts could drop by ~30%. Spending increases through year-end because projects started from 12 to 24 months ago approach peak spending. The large reduction in new 2020 starts will impact 2021 spending.
can you clarify the non-res forecast….this statement…’Nonresidential Buildings spending drops 8% in 2020 from $469 billion to $434 billion and then drops 2% from $475 billion to $468 billion in 2021′ does not match the numbers in the table just below it.
The resultant that spending drops “to” matches rounded to nearest whole billion. The initial value in the text that spending dropped “from” is the forecast prior to the pandemic.
Hi Ed: its interesting that the inverse (to a point) of this move is showing in 2011 when the volume drops and jobs recover. I wonder if there is a relationship there. Is part of the 2011 volume drop a de-escalation effect?
I suspect there is an anomaly that began in 2011 and the cynical side of me thinks that the long term divergence your chart indicates was starting to be seen as a problem, and someone used the opportunity to correct the data.
I’ll put my tinfoil hat away for the moment.
Chris, have a look at a plot in this link. The plot is 2006 Modified Jobs vs Spending. I think it shows the anomaly start. These charts are the same data, one chart is plotted monthly, the other annually.
Interesting, that’s when I started with GBCO. It must be my fault!
One does wonder if there is some large unmeasured cash flow that moves through construction books and is unrelated to actual activity.
I think there is also some impact in the payment delay, work in place is never paid on less than 30 days, and is probably closer to 60 days in reality.
Census measures PIP$ , not by paid$, but by requisitioned$. Also in the case where the req$ are not available, they use amount PIP forecast by cost loaded schedule. So there is no delay. But it is this 2nd part that could lead to misreporting data. If jobs are shut down and no one is there to report monthly data to Census, then Census would provide an estimate of the expected spending based on cost loaded schedule. It’s possible in this case some projects could actually be shut down and the project actually recorded no progress during the month. Census will revise these numbers when projects provide actual reports again.
Got it. It would have been a small piece at any rate, the trend lines don’t seem to show a payment delay. Other than anomalous problems like data corruption, or incorrect inputs, it kind of leaves us with nefarious possibilities.
Back to the tinfoil hat.
Didn’t see the second half. Corrections in the data for the last 2 months won’t answer the long term trend.
Another great analysis Ed. A stunning turn for our industry. Thank you. Dan
Ed, could the disconnect in data between work volume and job loss be attributable to the portion of the construction industry that may skip pulling building permits (remodelers) but who’s workforce was most likely forced to apply for pandemic unemployment benefits?
I don’t think so. There are varied sources surveyed to collect the monthly work put-in-place values, not just permits. Also the magnitude of the disparity is many times greater than just that portion of the industry.