The path from construction starts to spending is not direct and not quite as simple as you might think. Spending is the market activity measure that drives all construction economics, so that’s where we need to get too. With an appropriate modeling technique we can get from new starts to predicted spending in a few steps.
New Construction Starts (construction starts referred to here is Dodge Data & Analytics New Construction Starts) is excellent data for forecasting. The following forecast is entirely developed from starts data. No actual spending is incorporated into this forecast. The purpose is to show that using the data properly can produce an accurate forecast.
The starts data is a survey. As in any survey, starts represents a portion of new construction activity. Study shows the survey size varies with each market from about 40% to 70% of actual. Starts data captures a share of the total market or a portion of all construction, on average about 60% of all construction.
The easiest way to understand this is to compare total annual construction starts to total annual spending. National construction starts from 2016 to 2019 range from $750 billion/year to $800 billion/year, while spending in this period ranges from $1,200 billion/year to $1,300 billion/year. From this we see starts data captures a share of about 60% of the total construction market.
The total starts survey averages about 60% of the actual market. In this analysis every market is adjusted by its own individual market share factor. The adjusted starts represent the full amount of starts that would generate the full amount of spending.
To predict spending activity from new construction starts, the starts data must be spread over time using appropriate cash flow curves. On average about 20% of new construction starts gets spent within the year started, 50% is spent in the next year and 30% is spent in years three and four. The cash flow curves used in this model are specific to each market type and can vary from the average.
Applying a market survey factor to develop full magnitude of spending and an expected duration for all starts, depending on market type, to produce a forecast cash flow from starts data, the predicted pattern of spending is developed. The factors have been shown to produce a reliable prediction of total future market activity.
Backlog at the beginning of the year or new starts within the year does not give an indication of spending within the year. New starts within the year could contribute spending spread out over several years. Total cash flow in the year, or spending, could include cash flow from projects that started or entered backlog years ago.
Backlog increases if new starts during the year is greater than spending during the year. However, an increase in backlog does not necessarily indicate there will be an increase in market activity. An increase in backlog could represent a level rate of market activity, but for a longer duration.
Cash flow provides the best indicator of how much and when spending will occur. Cash flow from all previous starts gives a prediction of how spending will change monthly 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.
Total of all national construction starts increased every year since 2008. New starts slowed to +2% in 2018 and are forecast at a potential decline of 0.2% in 2019. Backlog is still up leading into 2020 but after that starts and backlog are forecast to remain flat or decline over the next few years. Total spending declines in 2022. However, as the next tables will show, work distribution is uneven with residential declining and nonresidential up.
Nonresidential Buildings starts (excluding Terminals) have reached a new high every year since 2009. The last three years starts are up 3% to 4% per year. Every market posted increases in 2017 and 2018. Only Commercial/Retail declined in 2019. The largest increases over the last two years were Manufacturing and Office Buildings. Spending is still strong in 2020 but then with the slowdown in starts forecast in 2020, backlog growth stalls and spending slows in 2021-2022.
75%-80% of all Nonresidential Buildings spending within the year will be generated from projects that were booked in starting backlog at the beginning of the year.
Nonbuilding Infrastructure markets total spending amounts to only about 70% of nonresidential buildings markets. The largest infrastructure markets are Highway/Bridge and Power but the largest increases in new starts recently is in Transportation. Transportation starts are up 30% in the last last three years and backlog is up nearly 100%.
Nonbuilding Infrastructure starts can be erratic with a long pattern of up then down years. Starts (including Terminals) gained only 2% in 2019 but that is only low because Power, the largest market overall saw starts decline by 7%. Total infrastructure starts are at an all-time high.
Infrastructure backlog peaks in 2020 and remains high into 2021. Spending increases are in the 6% to 8% range at least for the next two years. Infrastructure projects typically have the longest duration. Projects contribute spending sometimes up to 5 or 6 years. The largest spending increases in 2020 are in Transportation and Highway projects.
The Residential table shows that most of the spending in any year is cash flow from new starts. For short duration residential spending, single-family residential and renovations work, approximately 75% of the spending occurs in the year of the starts and 20% in the following year.
For long duration residential spending, typical of multifamily residential, approximately 50%-55% of the spending occurs in the year of the start, 35%-40% in the next year and only 5%-10% occurs two years out.
Only 25% (for short duration SF and Reno) to 50% (for longer duration MF) residential spending within the year comes from work that was booked in backlog at the beginning of the year. The performance of residential spending in the year is very much dependent on new starts.
The level of activity has a direct impact on inflation. When the activity level is low, contractors are all competing for a smaller amount of work and therefore they may reduce bids. When activity is high, there is a greater opportunity to bid on more work and bids can be higher.
Residential construction saw a slowdown in inflation to only +3.5% in 2015. However, the average inflation for six years from 2013 to 2018 was 5.5%. It peaked at 8% in 2013. Residential construction spending dropped an unexpected 6% in 2019 and after adjusting for inflation that is a 10% decline in construction volume. Typically, large declines in volume are accompanied by declines in inflation. National average residential construction inflation for 2019 is now at 3.8%. 2020 is forecast at 3.75%.
Nonresidential Buildings indices have averaged 4.4% over the last five years and have reached over 5% in the last three years. But spending slowed dramatically in 2019. This forecast indicates spending in most nonresidential buildings markets will gain little in 2019, the slowest rate of growth post-recession. However, new starts in 2018 and 2019 boosted backlog and 2020 spending will post the strongest gains in four years. Strong gains in spending historically has led to accelerated inflation. National average nonresidential buildings construction inflation for 2019 is now at 4.8%. 2020 is forecast at 4.2%.
Examples of how commonly reported construction data can often be misused – Construction Spending and Construction Starts
Construction Grew $41 billion, 3.3%, from 2017 to 2018
An increase in construction spending is often referred to as growth for the industry, but that is incorrect. Construction spending measures the change in the dollar value of work performed, not the volume of work performed.
The difference between spending (or revenue) and volume can be explained by a simple example, the Crate of Apples. A farm stand sold a crate of apples last year for $100. Costs have gone up. Today the same size crate of apples sells for $110. Farm stand revenues increased 10%, but the amount of business volume did not increase. Volume of sales is still one crate of apples. All the increase in revenue was inflation.
The $41 billion increase in construction spending from $1.266 trillion in 2017 to $1.307 trillion in 2018 is a 3.3% increase. However, construction inflation for that period averaged 4.7%. Construction inflation adds only cost, not volume, to the amount of work. Construction spending is measured in current dollars, actual dollars spent within the year in the value that year. Construction volume is measured in constant dollars, adjusted for inflation, so any and all years can be compared to each other.
Real construction volume adjusted for inflation actually decreased 1.4% from 2017 to 2018.
Total Construction volume, after accounting for inflation, has been down for five of the last six quarters. Construction volume peaked from Q1 2017 to Q1 2018, is now down 6% from the 2018 peak.
Construction volume is not directly known or reported. It is not a commonly referenced industry measure reported in the news. But it is a more important indicator of activity in the industry than spending. Volume is found only through analysis of spending and inflation data.
Another common misrepresentation using spending data relates to jobs growth. Jobs growth is often compared to spending growth where a 3% to 4% increase in jobs from year to year is substantiated if we have a similar 3% to 4% growth in spending. However, current $ spending is not yet adjusted for inflation and does not represent growth in real volume of work. Jobs must be compared to volume. Real volume increases are represented by constant $, or construction spending adjusted for inflation.
In the last 3 years jobs have increased by about 8% but real construction volume has decreased by about 6%. In recent years, construction volume has not supported jobs growth.
Construction Starts Predict Changes in Spending
Two very important criteria must be known about new construction starts in order to properly predict spending.
1st – To predict spending from new starts, the starts data must be spread over time using an appropriate cash flow curve. A simple illustrative spending pattern for nonresidential buildings starts, or a typical cash flow curve, for total starts within a year is: 20% of the revenue gets spent in the 1st year, 50% in the 2nd year and 30% in the 3rd year. This shows predicting spending in any given year is dependent on several previous years of starts.
Multi-billion $ highway projects, manufacturing facilities, power projects and transportation terminals often have much longer duration cash flow curves. In other words, if your intent is to predict construction spending in 2019, you need to know what starts were at a minimum in 2017 and 2018, and in many cases back to 2016 or even 2015.
Starts spread over time with cash flow curves predict spending.
2nd – For new construction starts survey sample to be used to compare to itself from year to year to predict growth in spending, sample size must be known. Starts data captures a share of the total market or a portion of all construction, on average about 60% of all construction. The easiest way to see this is compare total construction starts to total spending. Starts from 2016 to 2019 range from $750 billion to $800 billion while spending in those years ranges from $1,200 billion to $1,300 billion. From this we see starts capture a share of the total market. Any time a survey of a total population is used to forecast the total, the survey share of total must be considered. If sample size is not constant, the apparent growth in starts does not all reflect real growth in spending.
Amusement/Recreation is an example that shows starts that generate a predicted cash flow pretty well balanced with actual spending from year to year. The share of starts in the survey is fairly consistent never varying from 63% to 66% from year to year.
Office provides an example of variation in sample share of total. Starts generate predicted cash flow that increased substantially from 49% to 57% and remained higher compared to actual spending.
Office starts increased from $21 billion in 2013 to $50 billion in 2019. This data generates the predicted cash flow that is compared to actual spending. To predict total spending from unadjusted starts, unadjusted starts CF$ are factored up (divided by) share of total market. If the share of market captured in the survey remained constant then the predicted spending would remain close to 50%.
CF$ Predicted/Actual shows that Cash Flow Share of actual spending was 48%-50% for several years but then jumped to 56%-57%. The predicted cash flow generated from the increase in starts is not entirely representative of an increase in spending but represents the combined value of the expected increase in spending and an increase in share of market data captured.
Starts cash flow and starts survey share of total spending are never directly known or published. These factors are found only through analysis of the data.
The Educational market data shows a similar situation as Office data. Starts generate predicted cash flow that increased substantially compared to actual spending.
Starts data generate predicted cash flow to forecast spending. This requires tracking share of total market captured in the starts survey data to account for any growth in the market share captured vs growth in predicted spending.
Construction Spending in August is down slightly from July and down 2.3% year-to-date (ytd) from 2018. Spending for the last three months has remained flat. Qtr/Qtr spending has ranged +/- 2% for the last five quarters. Total 2019 spending will be down 0.5%. I’m expecting 2019 Nonresidential Buildings spending up 1%, Non-building Infrastructure up 9% and Residential spending down 6%.
Residential construction spending, down for six consecutive quarters, is now down 11% from Q1 2018. Residential volume (spending minus inflation), also down for six quarters, is down 16% from the Q1 2018 peak.
Some markets spending totals for 2019: Lodging +11%, Office +10%, Amusement and Healthcare both +5%, Commercial/Retail -14%, Highway +11%, Power +7%, Transportation +6%, Environmental Public Works (combined) +12%.
2020 forecast Starting Backlog for Nonresidential Buildings is currently up 6% and for Non-building Infrastructure is up 9%. Strong backlog leading into 2020 will increase spending in most nonresidential markets. Exceptions are: Commercial/Retail and Power backlog will decline. Residential spending is 65% dependent on new starts but Nonresidential spending is 80% dependent on backlog.
Forecast growth for 2020 is welcome since real construction volume, after accounting for 4% to 5% inflation, has been down for five of the last six quarters. Annual construction inflation since 2011 has been as high as 5.8%. For the last 3 yrs it has averaged 4.6%/yr. Construction spending for the last 3 years avg. annual growth is only 2.4%. When construction spending is lower than inflation, real volume is declining. Jobs must be compared to volume.
Total construction volume after inflation (quarterly avg) reached a peak in the 1st quarter of 2017 (which was then matched again in Q1 2018) and is now down 6% from the peak. Most of the volume decline was in Residential. Only Infrastructure has seen volume gains in the last two years. We have seen jobs growth slow in the last year, but the disparity between construction volume and jobs growth is the greatest ever. I expect to see a much more significant slow down in jobs growth.
Backlog growth for over the past two years will provide the base for Nonresidential construction spending increases in 2020. Major backlog increases from 2018 to the start of 2020 are: Manufacturing +20%, Educational +13%, Office +25%, Commercial/Retail -12%, Power -9%, Highway +20%, Transportation +50% and Environmental +23%.
The forecast for 2020 spending is total $ up 5%, but Residential spending will be flat to down slightly.
Construction Spending Forecast strength over the next 18 months is all nonresidential. Current spending seasonally adjusted annual rate (SAAR) vs SAAR at the end 2020 shows Nonresidential Buildings now at $450bil will end 2020 at $500bil and Non-building Infrastructure, now at $340 billion, will end 2020 at $375bil. Residential is now at $510bil but will move up slightly then down to finish 2020 at $500bil.