Construction Analytics

Home » Posts tagged 'starts'

Tag Archives: starts

In What Category is That Construction Cost?

Seldom do two sources present information the same way!

In the construction industry, a disconnect exists in the reporting of construction starts data and spending data.  Problems may arise when data is used to perform comparisons or forecasts. New starts and backlog may be listed in one category and spending for the same markets may be listed in another.

The U.S. Census Construction Put-in-Place (Construction Spending) Release follows these definitions. I adjust all other input/forecasting data that I use to conform to these Put-in-Place definitions. Some pitfalls to be aware of:

Residential spending data is about 35% renovations and improvements that has no units associated with the dollars, so cannot be included in a comparison to housing starts.

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.

Offices includes pubic buildings such as city halls and courthouses. Also includes data centers and bank buildings but excludes medical office buildings, offices at manufacturing sites and offices at educational facilities.

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, labs, food service and sports/recreation facilities at 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 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 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 institutional buildings or other public nonresidential buildings, but 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.

Similar discrepancies may exist when comparing starts or spending to indexes, such as the Architectural Billings Index, which broadly classifies projects as commercial, institutional or residential. Some resources classify Amusement/Recreation as institutional and some as commercial. Labs are sometimes classified as commercial but in many cases are included in educational or healthcare, both institutional.

As you can see, there are several 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 is required to make input and output match. For example: I move starts for terminals from nonresidential buildings to non-building infrastructure Transportation, so that really changes my totals by sector.

What does your source for data take into consideration? Know your data!



Backlog Construction Forecast Fall 2017


Total Construction Starting Backlog is at a record high, up 30% from the previous high in 2008.

Infrastructure and Residential sectors dropped to a decade low backlog in 2010. In 2013, nonresidential buildings hit the lowest starting backlog since 2004. Combined, total backlog hit a low-point in 2011, the lowest since 2003. Total Starting Backlog has been increasing since 2011, up 65% in 2018.

Nonresidential Buildings and Non-building Infrastructure backlog are both at all-time highs. 75% to 80% of all nonresidential spending within the year comes from starting backlog. Residential backlog is at a post-recession high, although as will be explained later, it is new residential starts that are more important and starts have tripled since the 2009 low. 70% of all residential spending in the year comes from new starts. Residential starts are still 20% below the 2004 peak.

Starts Generate Backlog

New Starts increased at an average rate of 11%/year from 2012 to 2016. 2017 starts slowed to less than half that pace.

Nonresidential Buildings starts, even though there was a 1% decline in 2015, averaged 13%/year growth for the last 4 years. 2017 will post an 8% increase. The 6 months from Aug 2016 to Jan 2017 was the highest starts since Jan-Jun 2008, also the year nonresidential buildings spending peaked. The 6 months Apr-Sep 2017 just surpassed both those previous peak highs.

Non-building Infrastructure starts were the highest in the 1st 6 months of 2015 than any 6-month period in history. Total 2015 starts increased 26%. 2016 is down just 2% from the peak 2015 starts and 2016 is the 2nd highest starts on record. Those early 2015 starts will still generate 10% of all spending in 2018. 2017 starts are level with 2016. After revisions, 2017 starts may set a new peak high.

Residential starts in 2016 posted the best year since 2005-2006. New starts in 2016 were revised up to show an increase of 10% over 2015. That follows five years of growth averaging 20%/year. New starts in Q1’17 reached an 11 year high.

Entering Backlog

New Backlog is added every month from New Construction Starts. When projects first enter backlog, the amount counted to backlog is the total value of project revenues under contract that are about to start construction, or the same as the new start values. For purposes of predicting future construction spending, as each month of project construction passes, work that has been put-in-place is subtracted from the total value to get the amount remaining in backlog.

Starting Backlog

Starting Backlog is the Estimate-to-Complete (ETC) value of all projects under contract at the beginning of a period. Projects in starting backlog could have started last month or last year or three years ago. The amount counted in backlog is the value of the project that remains or that has not yet been put-in-place. Backlog is the total amount of future spending that will be generated by the project, commonly referred to as the ETC. The sum of all ETCs represents current backlog.

  • Nonresidential buildings 2018 starting backlog is up 10%
  • Non-building Infrastructure 2018 starting backlog is up 12%
  • Residential buildings 2018 starting backlog is down 3%
  • Starting Backlog is at an all-time high for nonresidential buildings and non-building infrastructure.

Typically, starting backlog is a reference to the amount of work in backlog on January 1st. It is referred to as the Starting Backlog for the coming year. The sum of all ETCs as of December 31st represents Starting Backlog.

For any project that has a remaining duration going out past year end, backlog at the start of year does not represent the amount that will be spent within the year. Some of that project backlog will be spent in future years. For this reason, backlog is not representative of spending within the year and the change in Starting Backlog from year to year is not an indication of a change in spending from year to year.

Values and duration of projects that make up backlog help to better predict spending activity over time, particularly in the coming year.

A cash flow schedule of all ETC backlog and predicted new starts provides a tool to predict future spending. It is not enough to look at just the change in backlog to get an indication of the strength of the market. While continued growth in backlog is important, the predicted cash flow from backlog and cash flow from new starts is necessary for predicting spending.

Construction spending is strongly influenced by long duration projects in backlog, more-so than normal monthly starts growth rate. The pattern of continuing or ending cash flows from the long duration backlog projects causes fluctuations in spending that supersede the balance cycle of one month of old jobs ending for every new month of jobs starting. This often can be responsible for some of the monthly fluctuations of construction spending.

The following table shows predicted cash flow from backlog on record as of October 1, 2017 and predicted starts that will generate future backlog in 2018.

Backlog Cashflow 2018 and 2017 11-10-17

Look Ahead to 2018

Buildings and Infrastructure will both hit new all-time highs for starting backlog in 2018.  For four years, from 2010 to 2013, all nonresidential backlog remained nearly constant. Since then, growth has been similar to the pre-recession construction boom of the early 2000s.

Nonresidential buildings 2018 starting backlog is 50% higher than at the start of 2014, the beginning of the current growth cycle. Starting backlog has increased for 5 years at an average 10%/year. Spending from starting backlog, up 10% in 2018, increased for 5 years at an average 9%/year. Buildings will reach a new high for spending in 2018.

Non-building Infrastructure 2018 starting backlog is up 35% since 2014 but spending from backlog is up only 10%. Infrastructure starting backlog has been increasing for more than 10 years, sometimes only a fraction of a percent per year. Since 2010, backlog increased only 3%/year for the first 5 years then it jumped 35% in the last 3 years. Spending within the year from starting backlog is up 8% in 2018. Infrastructure spending will hit a new high in 2018.

Backlog incld Res Starts 2005-2018 11-8-17

Due to the shorter duration of projects, nearly 70% of residential spending within the year is generated from new starts. Unlike nonresidential, backlog does not contribute nearly as much residential spending within the year. New residential starts in Q1’17 reached an 11 year high.

  • Cash flow models of construction projects in backlog are indicating substantial acceleration in nonresidential spending over next year, perhaps most notable in infrastructure.
  • 75% to 80% of nonresidential spending within the year comes from Starting Backlog.
  • 70% of residential spending within the years comes from New Starts. Residential starts are at a post-recession high.
  • Share of spending within the current year from backlog is at an all-time high for nonresidential buildings and non-building infrastructure.

Nonresidential Buildings

Nonresidential buildings experienced a decline in starting backlog as recently as 2013. Since 2013, nonresidential buildings starting backlog is up 60%. Backlog will hit a new all-time high for 2018, 5% over the previous high in 2009 . Not only is starting backlog higher coming into 2018, but also spending from backlog is predicted up by 10%. This will produce a new high in current dollar spending.

Revenues from starting backlog account for 75% of all nonresidential buildings construction spending within the year.

Backlog for Nonres Bldgs 2014-2019 11-12-17

Educational starts, backlog and spending has been increasing for 5 years or longer. 2018 starting backlog is up 16% from 2017. Starts for 2018 are predicted to go up 13% and this will push 2019 starting backlog even higher. This should produce good spending growth for the next few years.

Office construction starting backlog for 2017 was the highest in at least 8 years, more than double at the start of 2014 when the current growth cycle of office construction spending began. For 2018 it’s up 27% over 2017. Office starting backlog increased an average of 28%/year for the last 5 years. Actual spending increased an average of 17%/year. Backlog growth looks like it will support very strong spending increases into 2019.

Commercial Retail backlog will hold steady from 2017 into 2018. This should level off spending after 7 years of strong growth. 2018 backlog still produces a spending increase but current projections show a slight drop in 2019.

Lodging backlog increases slightly for 2018. Beyond 2018, spending will decline, but this is after 6 years of growth totaling 300%.

Manufacturing posted a 100% increase in new starts in 2014 that drove starting backlog to new highs for the next two years. With new starts slowing back to normal by 2016, starting backlog dropped 20% in 2017 and spending dropped 12%. That was expected. What was unexpected is that 2017 posted another very strong year of new starts and that pushed 2018 starting backlog again to a new high. This will support a spending rebound in 2018-2019 after a drop of 18% in the last two years.

Non-building Infrastructure

Non-building infrastructure backlog stood at $180 billion in 2008, the last pre-recession year. At the beginning of 2017, non- building infrastructure backlog hit an all-time high, $260 billion, up 45% from 2008. For the last three years, starting backlog is up 40%. In 2018, it’s up 13%, another new high.

Revenues from starting backlog account for 80% of all non-building infrastructure construction spending within the year.

Backlog for Nonbldg Infra 2014-2019 11-12-17

Power backlog has doubled since 2014. It’s up 11% for 2018. Starts are down 14% from the 2015 peak, but spending from backlog is increasing. Most relevant is that backlog increased much stronger than spending. Backlog is being driven higher by very long duration projects that started in 2015, 2016 and 2017.

Highway starts declined the last two years from the peak in 2015, but starting backlog increased the last 3 years and is now 25% higher than 2015.  The last three years of highway starts are still feeding spending in 2018. There is very little change in the amount of spending from backlog, so 2018 spending won’t change very much from 2017.

Transportation new starts shot up by 70% in 2017, pushing 2018 starting backlog to a new high, up 75% from 2017. That will help increase 2018 spending by more than 15%, but a larger spending increase could come in 2019.

Environmental Public Works, Sewer/water/Conservation is experiencing declining starts, declining backlog and declining spending from backlog. All are at the lowest since 2014. We may not see any increase in construction spending until 2019.

Public Work

Public vs Private starts are not tracked separately, but the public share of markets is known. Therefore a projection of public backlog is possible. Highway and Environmental Public Works are 100% public.  Educational is 80% public, Transportation is 70%, Amusement/Rec is 50%, Healthcare is 20% and Power is 10% public, along with few other smaller shares. Starting backlog for 2018 is up 40% from 2014 due to the predominantly long duration projects that make up public work. This is a post-recession high and is nearing the all-time high of 2008. Increased backlog is indicating the best construction spending increases since 2008 for the next two years.

Residential Buildings

Residential new starts hit bottom in 2009 and starting backlog hit bottom in 2010. Residential on average has the shortest duration and new starts has a dramatic impact on the amount of available work. Both new starts and backlog are now 3x higher than the lows. New residential starts have increased every year since the 2009 bottom, but are still 25% lower than 2004-2005. Residential spending reached its peak of $630 billion in 2005. Current spending is still 15% below that peak. In constant $, spending is 30% below that peak.

Due to the shorter duration of projects, nearly 70% of residential spending within the year is generated from new starts. Unlike longer duration nonresidential projects, backlog does not contribute nearly as much. If no new work started within the year, within a matter of a few months there would be no backlog ETC left to support the residential construction industry.

New starts slowed in 2017 to only 4% growth and similar growth of 6% is expected for 2018. This is not unexpected after 5 years (2012-2016) of new starts growth at an average 20%/year. This leads to a prediction of 2018 spending up only 6%.

All construction starts data in this report references Dodge Data & Analytics starts data.

See this companion post for  Starts Trends Construction Forecast Fall 2017

Also see 2018 spending forecast Spending Summary Construction Forecast Fall 2017

Starts Trends Construction Forecast Fall 2017


It all starts here! Construction Starts Generate Construction Spending.

2017 construction starts through September total $557 billion Year-to-date (YTD), even with 2016. If/when 2017 gets revised as expected it will then show +3% to +4% growth over 2016, but we won’t see that growth in the revision data until next year.

  • Previous year starts always later get revised upwards. Therefore, current year starts ytd growth is always understated.
  • Revisions for the period 2012-2015 averaged +4%.
  • Revisions to 2016 year-to-date through September are +10%.
  • Starts have been increasing at an average rate of 11%/year for the last 5 years.
  • Nonresidential Buildings and Nonbuilding Infrastructure are at or near all-time highs.
  • Residential starts are at a post-recession high.
  • New starts will generate record high 2018 starting backlog for every sector.

Nonresidential Buildings starts, averaged 13%/year growth for the last 4 years, even though there was a 1% decline in 2015. 2017 will post an 8% increase. The 6 months from Aug 2016 to Jan 2017 was the highest period of starts since Jan-Jun 2008, the year nonresidential buildings spending peaked. The 6 months Apr-Sep 2017 just surpassed both those previous peak highs. This will help support increases in nonresidential buildings spending for the next two years.

Infrastructure starts posted a higher value of new construction projects in the 1st 6 months of 2015 than any 6-month period in history. 2016 is down just 2% from the peak 2015 starts and 2016 is the 2nd highest starts on record. Those early 2015 starts will still generate 10% of all spending in 2018. After revisions, 2017 starts may set a new peak high. This would set up infrastructure as the strongest growth sector for the next two years.

Residential starts in 2016 posted the best year since 2005-2006. New starts in 2016 were revised up by 5% to show an increase of 10% growth over 2015. That follows five years of growth averaging 20%/year. Initial values posted for 2017 show starts up by only 3.5%, however, the average revision for the past few years has been +2% to +4%, so 2017 will get revised higher. New starts in Q1 2017 reached an 11 year high.

All construction starts data in this report references Dodge Data & Analytics Starts data.

Starts CF 11-7-17

Nonresidential Buildings

Retail/Commercial starts may finish flat for 2017 but that is compared to peak starts in 2016. Starts for the 12 months Aug 2016 – June 2017 posted 10% growth over the previous 12 months. Commercial starts have been increasing every year since 2010.

Office construction starts have been increasing since 2010 with the strongest growth period of new starts in the 12 months July 2016 – June 2017, the highest 12 months on record, 60% higher than the previous 12 months. That high-volume period of starts is going to elevate spending in both 2018 and 2019 to come in higher than 2017. Office starts averaged year-over-year (YOY) growth of 20%/year for the last five years. Data centers are included in Office.

Educational starts are up 15% in 2017. Starts have averaged YOY growth of 8%/year for the last two years and have had slow but steady growth since 2012. The growth in starts will support growth in spending or the next three years.

Office, Retail and Educational markets comprise 60% of all nonresidential buildings. They are collectively responsible for 70% of the increase in 2017 nonresidential buildings starts.

Healthcare starts have quietly increased to a record high over the last 12 months, up 30% for the 12 months through August vs the previous 12 months.

Lodging starts will be down ~10% in 2017, but from 2013 to 2016 averaged over 25%/year for 4 years.

Manufacturing is the only nonresidential building market that will NOT finish 2017 with new starts totals at or near post-recession highs. Manufacturing reached record high starts in 2014 and record spending in 2015. However, 2017 will post new starts 50% higher than initially predicted by Dodge.

Manufacturing spending was expected to fall in 2017 after peaking in 2015 from massive growth in new starts in 2014. Based on cash flows from starts, spending was expected to decline in 14 of the last 18 months. It did decline in 11 of those months. We are at the point of turn-around with only 1 monthly decline predicted in the next 3 months and no spending declines expected next year.

Non-building Infrastructure

Sewer/Water/Conservation, the three Environmental Public Works markets, posted declines in new project starts in 3 (sewer) or 4 of the last 4 years. Collectively, new starts in 2017 are the lowest in 5 years. Cash flow predicted from starts has been indicating spending declines since Q2-2016. In fact, spending has declined in 12 of the last 18 months. Cash flow still predicts more spending declines over the next 8 months.

Highway/Bridge/Street starts in the 2nd half of 2014 recorded the slowest rate of growth in the last 6 years. Starts that would normally be contributing spending through 2017 and into 2018 contributed a lower than normal volume of spending which will end in 2017. Had it not been for the extremely high volume of starts in the 1st 4 months of 2014, the most ever recorded in 4 consecutive months, 2017 spending would have dropped more than double the 4% spending decline now forecast.

Highway starts in the 1st 6 months of 2015 posted the next highest growth to early 2014. Spending in 2018 will benefit from those projects that started in 2015 but that have unusually long duration. They will contribute a higher rate of spending in 2018 beyond the duration that typical projects would have ended. It is not recent new starts but old backlog that is influencing 2017 and 2018 highway spending.

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. While this helped turn 2017 spending positive, 2017 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 spending but will post double digits gains in 2018. The high volume of 2017 starts has the most effect on 2019 spending. Terminal buildings is reported in Dodge Starts in Other Institutional Bldgs. However Census reports terminal spending in Transportation along with rail and Dock spending. I adjust the starts data in my reports conform to Census organization.

Power market starts peaked in 2015 at an all-time high, up 142% from 2014 and more than the prior two years combined. The Power market was the prime contributor to the abnormally high infrastructure starts in the 1st 6 months in 2015. Power spending was down 6% in 2015 and up only 3% in 2016 because Power starts were also at an all-time high in 2012, just below the 2015 level, and those starts drove 2014 spending to an all-time high, but then spending from those old jobs tapered off in 2015.

Power starts dropped 11% in 2016 and are down slightly in 2017. Recently, there has been an unexpected large volume of power plant and pipeline starts that are driving 2017 power starts to come in about 40% higher than initially expected.

Even though Power starts have been declining since the 2015 high point, Power had several periods with an exceptionally high value of new starts, some of these periods 2x to 3x the normal rate of growth and a year or two longer duration than typical; late 2014, Jan-May 2015, Feb-Jun 2016 and again in Feb-Jul 2017. A large share of the cash flow, or monthly spending, from all those exceptional starts will occur in 2018 and 2019 and will drive spending to 10%+ gains.

Although starts are not tracked for Public vs Private, Highway, Educational, Environmental Public Works and Transportation make up more than 80% of all Public construction. Only Environmental Public Works starts are down. Educational, Transportation and Highway all have a positive outlook in new starts and predicted spending for 2018 which pushes public spending to post-recession highs.

Here’s how to use the Starts data and how it affects spending Construction Starts and Spending Patterns 9-26-17

Also after New Starts, dollars are tracked in Backlog, Backlog Construction Forecast Fall 2017  11-10-17

See the Spending Forecast Spending Summary Construction Forecast Fall 2017


ABI – DMI – Starts – Construction Spending


The attached plot shows actual and predicted construction spending compared to several industry leading indicators. The ABI, produced by the American Institute of Architects (AIA) shows work on the boards at architectural firms. Values above 50 indicate work increasing, values below 50 = work decreasing. The DMI is a survey from Dodge that gives an indication of new construction momentum. Starts is the total cash flow growth from all nonresidential starts currently in backlog.

Both the ABI and the DMI have long lead times. For example, the ABI value posted by AIA today is an indication of what to expect 9 months from now. I’ve plotted the values for ABI and DMI out at the lead time dates (# of months) in the future so they would correspond to future cash flows from all starts and predicted spending. The Starts, DMI and Spending values on this plot are indexed so they could be plotted with the ABI while keeping growth trends in each index true.

  • ABI – Architectural Billings Index
  • DMI – Dodge Momentum Index
  • Starts – Aggregate Cashflows of Dodge Starts
  • Spending – Actual and Predicted Construction Spending


Overall Spending mostly correlates with Starts except that Starts showed a steeper growth rate in 2016 before a drop. Starts and Spending match well for all of 2014 and 2015. Both DMI and ABI are more erratic, however, the advances and declines in the ABI do correspond well with pickups and slowdowns in Spending. From mid-2015 through the end of 2016, the DMI was in a narrow range and that could possibly be said to be in synch with a slowed period of Spending.

Although they don’t match exactly by month, the ABI, DMI and Starts all show a drop sometime between 4th qtr 2016 and 2nd qtr 2017. That appears in Spending as a slight dip in 1st qtr 2017. The ABI gives an indication of a nice increase midyear. Both DMI and Starts are indicating substantial growth in spending by year end 2017.

Housing Starts vs Residential Construction Spending


Housing Starts (# of units started as reported by U.S. Census) can be erratic from month to month and short term changes in growth can sometimes be misleading. Trends should be looked at over longer term periods. New monthly starts on a seasonally adjusted annual rate (SAAR) basis for the last eight months through January 2017 have now averaged over 1,200,000. For the last four months starts have averaged 1,250,000. Permits have been following a similar pattern. Although starts versus permits varies considerably in some months, statistically they follow the same growth pattern. Growth in the number of new starts has been 5% to 25% per year due to erratic movement but in the longer term has averaged 18%/yr over six years since January 2011. We experienced an un-sustained start to recovery in 2010, but essentially we went through a protracted bottom between 500,000 and 600,000 new starts that lasted all throughout 2009-2010.


Dodge Data reports SAAR new residential construction starts by contract value in current dollars (not inflation adjusted). Unadjusted growth for the same six-year period increased from $120 billion SAAR to over $300 billion SAAR, or at an annual rate of over 25%/year. However, there was 25% residential cost inflation during that period. In constant 2016$, Dodge new residential starts growth averages 20%/year for six years since January 2011.


Now let’s look at construction spending, actual dollar value of work put-in-place. Here’s where the data has a disconnect.

At the start of 2011, total residential spending  had a monthly SAAR of $240 billion and at the end of 2016 was $470 billion, an increase of 16%/year for 6 years. To find real volume growth those values must be adjusted for inflation. After adjusting for inflation, the actual spending volume growth in 2016$ from 2011 through the end of 2016 increased from $305 billion to $465 billion, an increase of 52%, or an average increase of 9%/year for 6 years.


Furthermore, the number of residential construction jobs reported by BLS increased only 33% over that time, an average growth rate of less than 6%/year.


What could explain these differences?

The low rate of jobs growth compared to spending growth is partially explained by the fact that in the preceding few years, even though about 1.5 million jobs were lost, 40% of the workforce, staff was not reduced nearly at the same rate that residential construction volume declined (55%). There remained significantly more staff on payrolls than was needed to complete the amount of volume that was being built during the residential recession. When growth resumed, spending increased at a much faster rate than new jobs were added and the excess labor slack was reduced. I suspect also that a portion of the labor vs spending difference is explained by the fact that not all jobs are captured by BLS. It has been suggested that a large percentage of residential workforce in some southwestern states is undocumented.

The variance between starts and spending is a bit more complicated. We need to look at completions vs starts, the mix and size of housing units being built and the amount of spending related to renovations.

The most commonly reported housing statistic is housing starts. Also in that data series is housing completions. Housing completions are always lower than starts. For the last five years completions have averaged almost 15% less than starts. While the growth in starts averaged 18%/year, growth in completions from 2011 through 2016 averages less than 15%/year. 

From 2011 to 2016 the average number of new single family (SF) units started increased from about 450,000 to 800,000. During that same period multi-family (MF) starts increased from 100,000 to 440,000. The percentage of MF units in total construction grew from 18% to 36% of total.

On average MF units are about half the size of SF units. Although the average size of SF homes increased about 10% during this period, the growth in the number of smaller MF units exceeded that of larger MF units by a factor of 2x. The ratio of smaller MF units doubled.

The share of MF units as a percent of all units doubled and the ratio of smaller vs larger MF units doubled. The total square feet of housing being built increased but did not grow at the same rate as the number of units. The average size of all units is getting smaller and therefore the constant cost per unit went down.

I suspect the increased ratio of smaller MF units and the percent increase of MF within the total number of all housing units has a big influence on the overall average cost per unit of total housing. That with the lower growth rate in completions helps explain why spending is not increasing at the same rate as overall number of housing unit starts. We are building more units per dollar spending because average unit size is smaller.

There is one more hidden factor to look at. That is, residential construction spending includes renovations. From 2009 through 2012 renovations totaled 45% of all residential spending. It began to decrease in 2013. For the last three years, renovation spending accounts for only 33% of all residential construction spending.  Renovation spending has no comparable # of units or total square feet associated with it.


The impact this has, since the share of renovations spending is declining, is to increase the percent growth in residential spending attributable to housing units to greater than the 9% calculated above. Removing renovations work from total spending shows growth in real inflation adjusted spending specific to housing units averaged about 13%/year for 6 years.

Summarizing everything from above, since 2011:

On the surface it looks like this:

  • Housing Starts # of units increased at 18%/year
  • Residential new starts in unadjusted dollars increased 20%/year
  • Residential construction spending increased 16%/year

After adjusting both units and spending we get:

  • Inflation adjusted total residential spending increased 9%/year
  • Inflation adjusted spending on units (excluding renovations) increased 13%/year
  • Growth in the # of housing units completed increased 15%/year
  • Share of Multifamily units has increased
  • Average size of multifamily units has decreased
  • Average size of all housing units being completed has grown smaller
  • The growth in the number of units completed can exceed the growth in spending because the average constant value cost per unit has decreased

The growth in the number of housing unit starts is NOT an indicator to use for forecasting growth in residential construction spending or constant volume. Increases in the number of units alone will not give a realistic indication of growth in residential jobs or spending. The rate of growth in completions, combined with the ratio of the sizes of units, not just size of SF homes but average size of all SF and MF units, has a significant influence on the spending volume and can only be compared to inflation adjusted spending specific to units, that is, total spending minus renovations.

Starts Point to Robust 2017 Spending


Starts Point to Robust 2017 Spending

Construction Starts for September were released 10-18-16 from Dodge Data and Analytics. Here’s some of the major points that can be developed from the data:


The six Nonresidential Buildings markets, Office (+30% YTD), Lodging (+50%), Educational (+10%), Healthcare (+20%), Commercial Retail (+15%) and Amusement/Recreation (+15%) make up 80% of all nonresidential buildings spending and account for combined growth of 16.5% in YTD new starts. Office and Lodging in 2016 will reach the 5th consecutive annual increase. Educational Markets, Commercial Retail and Amusement/Recreation will each record the 4th consecutive annual increase in total value of new starts. Spending combined for these six markets peaked in 2008 and dropped 37% to a bottom in 2012. For the last 3 years spending growth has ranged between 9%/yr and 12%/yr. For 2017, expect spending growth of 8%.

Manufacturing makes up 18% of nonresidential building market share. New starts 2016 YTD are down 54% from 2015. However, in 2014 and 2015 this market posted the fastest growth of any market in a decade and posted the two highest years on record for this market. It is currently settling back to a normal growth range. In 2014 starts increased 90%. In 2015 spending increased 33% to the highest ever recorded for manufacturing buildings. Spending will be down 2% to 3% in 2016 and down another 13% more in 2017, but 2017 will still be the 3rd highest year of spending on record.

Non-building Infrastructure starts will be down nearly 10% in 2016 but were up 25% in 2015. Power and Highway/Bridge/Street make up 2/3rds of non-building infrastructure spending. In 2015, Power starts increased 150% to an all-time high and Highway/Bridge/Street finished just shy of a 6-year high. It is not unexpected that starts in these markets will be down for 2016. The volume of monthly spending from projects started in 2014 and 2015 in this sector will contribute to spending for several years to come. Spending in 2017 will be the highest ever in this sector, up 7% from 2016.

Residential starts are having the best year since 2005-2006. Residential starts bottomed in 2009 and are now in the 7th consecutive year of growth. Although new starts will increase only about 7%-8% for 2016, that follows 4 years of growth averaging more than 20%/year. Spending peaked in 2005-2006 and dropped 60% to a low in 2009-2010. Spending has bounced 90% off the bottom in large part due to 17%/year average growth in 2013-2014-2015. Both starts and spending slowed in 2016 but still expect 7% to 8% spending growth in both 2016 and 2017.

Starts are recorded in full in the month a project starts but the total project budget gets spent over a long duration, so the effects on spending are spread over the next 2 to 3 years. Total starts are Up 10%/yr to 12%/yr for the last 4 years. The current forecast for 2016 is growth of only 3.5%, but that now leads us to a very important factor that must be considered when using starts data to predict future spending.

There is a major factor that keeps new starts in the current year from appearing as good as they should. Dodge Data continually revises starts. In every monthly release, the previous month is revised AND the last year’s year-to-date is revised. Dodge does incorporate other (usually minor) revisions at a later date, but the “12 month” revision to the previous year-to-date values captures a large part of all revisions.

So this September report includes revisions to the total 2015 YTD values through September 2015. None of the 2016 values yet include that equivalent “12 month” revision and won’t until next year. But the current year YTD not-yet-revised values are being compared to the previous year YTD revised values which has the affect of making current year growth appear lower than it should.

In the last 10 years the YTD revisions have never been down. Usually, most of the revisions occur to nonresidential buildings, about 5% to 6% per year, with only a 2% to 3% revision each to infrastructure and residential.

For total nonresidential buildings, so far year-to-date 2015 values through September have been revised UP by 9%. So while the 2016 year-to-date nonresidential buildings value this month is noted as down 2% compared to last year, much of the reason it is down is because 2015 values have had revisions applied that increase the 2015 base by 9%. We won’t get those equivalent “12 month” revisions applied to 2016 values until next year. When all the revisions are in, new starts for nonresidential buildings (typically revised up by 5% to 6%) in 2016 are on track to equal or exceed 2015 and perhaps record the third consecutive year of over $220 billion. We are within easy striking distance of the all-time high for nonresidential buildings starts reached in 2007!

For residential starts, if 2016 values get revised up next year by only 2%-3%, then 2016 will have grown by nearly 10% over 2015. Unless we experience a severe downward trend in new residential starts, which is NOT predicted, 2016 will post an all-time high for new residential starts.

(Year-to-date by market and month/month values by market are not published.)


See also this post on Construction Spending Sept 2016

New Construction Starts Much Better Than Might Appear


Dodge Data and Analytics yesterday released August new construction starts. The August number came in right about where I expected it, just over $700 billion. August starts are 21% higher then July which was an 8 month low. However, year-to-date through August totals $439 billion, down 7% from the same period 2015.

August came in at a seasonally adjusted annual rate (SAAR) of $711 billion, the highest since May 2015. In fact, this is only the fourth month since January 2008 that registered new starts over a SAAR $700 billion. The other three were in the 1st half of 2015.


Nonresidential Buildings new starts for August came in at a seasonally adjusted $267 billion, the second highest month since early 2008. The year-to-date is down compared to last year because 2015 had some very high months that helped the first half of 2015 reach an average of $214 billion, but the first five months of 2016 had some soft months that averaged only $189 billion. New starts for the last three months average $212 billion and starts have been increasing since May.

Residential new starts reached a SAAR of $291 billion in August, the third time this year over $290 billion, averaging over $280 billion so far for 2016, the highest since 2007.

Non-building Infrastructure starts for August total SAAR of $153 billion. Infrastructure starts fluctuate much more than any other and this year have ranged from $121 billion to $200 billion. Last year they ranged from $127 billion to $261 billion. Since 2006 Infrastructure starts annual totals have been between $140-$160 billion, except for last year when they shot up to $180 billion. So even though 2016 is coming in near the high end of the average from 2006-2014, it’s still well below last year because last year was so unusually high.

But there is another major factor that keeps new starts from appearing as good as they should look. Dodge Data continually revises starts. In each monthly release we can see not only the previous month revision but also the previous year-to-date revision. They do incorporate other revisions at a later date, but the “12 month” revision to the previous year-to-date values captures a large part of all revisions. So this August report includes revisions to 2015 values through August 2015. None of the 2016 values yet include that “12 month” revision. In the last 10 years the revisions have never been down. Usually, most of the revisions occur to nonresidential buildings, about 5% to 6% per year, with only a 2% to 3% revision to infrastructure and residential.

For nonresidential buildings, so far year-to-date 2015 values have been revised UP by almost 8%. So while the 2016 year-to-date nonresidential buildings value this month is down 10% compared to some very strong starts in early 2015, part of the reason it is down is because 2015 values have had revisions applied that increase the 2015 base by 8%, but we won’t see those equivalent “12 month” revisions applied to 2016 values until next year. When all the revisions are in, new starts for nonresidential buildings in 2016 are on track to equal or exceed 2015 and perhaps record the third consecutive year over $220 billion. We are within easy striking distance of the all-time high for nonresidential buildings starts reached in 2007!

For residential starts, if 2016 values get revised up next year by only 2%-3%, then 2016 will have grown by nearly 10% over 2015. Unless we experience a severe downward trend in new residential starts, which is NOT predicted, 2016 will post an all-time high for new residential starts.

Construction Spending vs Dodge Starts vs New Housing Unit Starts

Read my last few blogs and all of this is detailed, but this is worth a look.

Dodge Data Construction Starts cash flowed shows a predicted spending pattern.

Actual spending is shown to compare to the prediction.

For another residential input we have new housing starts. Here I’ve spread activity out from start to completion like a cash flow to get monthly activity. History compares to actual spending and future compares to Dodge New Starts cash flow.

The time flow of activity generated by housing starts is much more important than the monthly starts themselves. It prompts us to look at a much longer term trend of housing starts than just whether they have moved up of down in the last month or quarter.

Modeling for nonresidential buildings and non-building infrastructure appears more accurate than residential. It looks like my prediction of cash flow from Dodge residential starts needs to move 2-4 months to the left.

5-4-16 The cash flow plot for residential has been revised to use a different duration for SF vs MF vs Reno.

Starts Cashflows to 6-17 vs Spend 3-16

Housing Starts Workflow 4-16 SAAR


Residential Work Flow From Housing Starts

Housing starts can be erratic. It’s not unusual to see monthly housing starts fluctuate up or down by 10%, sometimes 20%. But what affect does this have on the flow of housing work? Not as much as you might think.

Although housing starts is in units, not dollars, we can create a “cash flow” to see how the new starts generate activity over future months. To see the flow of work I’ve created a simple time flow of starts to show the activity generated for new housing starts.

About 2/3rds of housing starts are single family units. These might have a construction duration ranging from 6 to 9 months. The remaining 1/3rd of starts are multifamily units. Those could have construction duration of anywhere from 8 months to 16 months and in some cases longer. For this simple analysis I’ve used a work flow duration of 2/3rds at 7 months and 1/3rd at 17 months. Varying the duration longer or shorter by a few months will not have a big effect on the outcome. It changes the slope of the growth rate but does not change the consistency of the growth pattern.

A time flow of housing starts shows growth rates of; 2013 +13%; 2014 +10%; 2015+12%. Actual construction spending shows growth of 2013 +19%; 2014 +14%; 2015+13%.

Housing Starts Actuals for Workflow 4-16

The chart above, “Housing Starts Monthly and Trend” shows the actual monthly starts values and a three month moving average. Monthly starts periodically peak and dip erratically.  Look at February 2015, the biggest dip in 5 years. The 1st quarter 2015 was down 7% qtr/qtr. But then notice it took less than 4 months for starts to come right back to the trend line and the trend remained intact.  2015 finished up 11%. This is how the monthly housing starts (# of units) data goes.

Housing Starts Workflow 4-16 SAAR


The “Work Flow” chart plots the actual work load out over time from the month the work started to completion. The total work flow in any given month is the sum of the work contributed from starts in previous months that have yet to be completed. Residential work flow has averaged +12% for the last 3 years. In 2015, growth was 14%. The very steep climb in early 2013 activity reflects work generated from the 28% rise in new starts in 2012, the largest % increase in new starts in 30 years.

Starts in any given month have only a small % impact on the slope of change in every succeeding month until completion. This is the same concept as cash flow. Construction spending in any given month is the sum of all the ongoing projects from all previous months.

This next plot shows the same workflow, only Not Seasonally Adjusted, so it shows the winter dips in activity and the steeper rate of growth during the more productive months. Although the average slope of growth is similar to the SAAR plot, this shows the real total work activity in any given month varies from that shown by the SAAR plot. However, it is not erratic like the starts plot, it is smooth and repetitive year after year.

Housing Starts Workflow 4-16 NSA

It would take a dramatic change in housing starts to significantly alter the progress of work flow and it would need to be a sustained change in starts. If a 20% decline is offset by an corresponding increase in the following month or months, then the future months of work flow will show little affect from the decline.

Construction Expectations 2016


What should we expect in 2016 for construction spending, jobs and cost?

Nonresidential buildings starts (as reported by Dodge Data & Analytics) were well above average from March 2014 through May 2015 but since then have been below average. It takes about 24 to 30 months for nonresidential building starts to reach completion. The effect of below average starts will kick in at the end of this year after strong spending growth.

Non-building infrastructure starts jumped 50% above average from November 2014 to peak in February 2015, then settled back to average in July of 2015. Those very strong starts in early 2015 will be spread out over 4 to 6 years so will not cause spending to spike. They will help support a slow steady increase in spending over the next two years.

Residential starts averaged near 20%/yr growth for 3 years but dropped below average for the entire 2nd half of 2015. That late 2015 dip in starts may not slow residential spending too much until the end of 2016. Overall, the data shows another repeat year of growth similar to the last three years.

Construction Spending by Sector 2013-2016 Feb2016

2015 Construction spending finished the year up 10.6% over 2014. After 3 years of growth averaging 9%/year, 2016 total construction spending could climb 11% above 2015, the largest percent gain in over 10 years. Any construction spending slowdown is temporary, baked in from old uneven starts causing uneven cashflow, soon to be ending. By the 2nd quarter 2017 all sectors return to positive growth for strong spending in 2017.

Nonresidential buildings construction spending went from zero growth in 2013 to 9% in 2014 and took off to hit 17% growth in 2015. Nonres bldgs spending could reach 12% growth in 2016 and 7% in 2017.

Infrastructure spending will increase a little in 2016 but we won’t see a sizable increase of 8% until 2017.

Residential spending averaged over 15%/year for the last 3 years and could go over 15% growth in 2016, combining for the best four years of spending growth since 2002-2005.

Don’t be mislead by news that construction spending is close to reaching the previous highs. That may be true of spending, but spending is not the measure of expansion in the construction industry. The measure of expansion is volume, spending minus inflation.

Construction spending is up nearly 40% off the 2011 lows and within 5% of the 2006 highs. But after adjusting for inflation, volume is up only 22% from the 2011 lows and is still 17% below 2005 peak volume.  We still have a long way to go. While spending is predicted to reach over 11% growth in 2016 and may do the same in 2017, volume will increase only 5% to 6% each year. The rest is due to inflation.

Spend current vs constant2015 22pct volume growth

March 2016 construction jobs increase 37,000 from February and although up and down, have averaged 37,000 jobs per month for the last 6 months.  That is the highest 6 month average growth rate in 10 years. That certainly doesn’t make it seem like there is a labor shortage. However, it is important to note, the jobs opening rate (JOLTS) is the highest it’s been in many years and that is a signal of difficulty in filling open positions.

To support the expected 2016 volume growth we need an average 25,000 new jobs per month in 2016, 300,000 new jobs, reaching a three-year gain of nearly 1 million jobs for the period 2014-2016, the highest three-year total jobs growth since 1997-1999. The labor force hasn’t expanded this fast in over 16 years. That can have some undesirable consequences.  Rapid jobs growth may result in accelerating wages and lost productivity, compounding the cost to labor.


If we get a construction jobs slowdown in the next few months, it’s not all due to labor shortages and not being able to find people. Construction volume has been growing faster than jobs for more than a year. It means productivity in 2015 is up after several down years. But, while we’ve recorded consecutive years of productivity declines many times, we have not had two consecutive years of productivity gains in the last 22 years. So historically we should expect a decline,  not gains this year.

Material input costs to construction are down over the last year, but that accounts for only a portion of the final cost of constructed buildings. The cost of new residential construction is up 5% to 6% in the last year. Several nonresidential building cost indexes are indicating construction inflation between 4% and 5%. The Turner non-residential bldg cost index for 2015 is 4.6%. The 1st qtr 2016 is up 1.15% from the 4th quarter 2015. The Rider Levitt Bucknall nonresidential building 2015 cost index is 4.8% and the Beck Cost Report has 5.0% for 2015. I recommend an average 5.5% cost inflation in 2016 for residential and nonresidential buildings. Non-building infrastructure costs are unique to each individual infrastructure market, so average building cost indices should not be used for infrastructure.

Building Cost Inflation Index

%d bloggers like this: