Compiled in one neat table, here are 2015 predictions from eleven construction data firms for spending growth in nonresidential markets. Several firms provided mid-year estimates and recent estimates. Some provided only mid-year and some just recent estimates. Midyear estimates are separated so changes can be seen to current estimates.
Actual spending put-in-place for September year-to-date (YTD) became available November 2nd and new construction starts for October became available November 23rd.
There is a wide range of variance in predictions with the closest spreads at 9% and the widest spreads in lodging and manufacturing markets. It will be interesting to look back at this chart when the final numbers for 2015 become available in February 2016 to see how we did.
- One recent estimate published in Engineering News Record (ENR) magazine 11-16-15 lists 7% growth for manufacturing buildings. Each of the first nine months in 2015, the year over year growth has ranged between 40% and 60%, so the huge growth expected has been apparent for some time. Even if the last three months drop 15% below the current average we will still finish the year up 40%.
- For growth in educational buildings to fall to only 3%, the last three months would need to drop 15% below the current six month average, a change we will not very likely see.
- The spread on lodging is 18%, from the low estimate of 15% to high of 33%. YTD lodging through nine months is up 31% over last year. To finish at less than 25% growth in 2015, spending for the next three months would need to drop 20% from current levels.
We get a chance to tweak these numbers a little tighter when October spending gets released on December 1st.
Common comparisons published in news reports for construction spending are, from best to worst:
- Current year-to-date vs same period previous year. YTD
- Current month vs previous month. Mo/Mo or MOM
- Current month vs same month last year. Yr/Yr or YOY
- Number of months to current value since last time that value achieved.
In some cases a comparison uses Not Seasonally Adjusted (NSA) dollars and in other cases Seasonally Adjusted Annual Rate (SAAR) dollars. NSA dollars is the actual amount spent within the month. SAAR dollars represents the annual rate that monthly amount would generate based on the normal proportion typical spent within that month. Typical spending is always much higher in summer months than winter to produce the same SAAR.
Year-to-date is the best comparison. It increases in strength as more months are added to the YTD. It is a value that gives a strong indication of growth over the previous year. Comparisons must be made using Not Seasonally Adjusted (NSA) dollars. Although it does lack adjustment for inflation, only one year of inflation is involved. Not adjusting for inflation is explained as the difference between current dollars and constant dollars.
Current month to month (MOM) comparisons are not generally affected by inflation but may not give a clear indication of movement due to monthly fluctuations. Comparisons absolutely must be made using Seasonally Adjusted Annual Rate (SAAR) dollars. It is a gross error to make month to month comparisons using NSA dollars, since there is a normal spending curve that shows the percent of total annual spending can vary considerably from month to month, sometimes by as much as 10%. This variation is accounted for in the SAAR.
When comparing to the same month last year (YOY), the question arises, “Is any big change in YOY caused by the current month performance or by the performance in the same month last year?” Again, YOY is missing adjustment for one year of inflation. Comparison can be NSA or SAAR dollars.
Number of months/years since current value was last achieved is almost always presented as a current dollar comparison. When dealing with cost, because of the long duration often involved it would be much better if it were a constant dollar comparison to account for construction inflation. However, construction inflation may not be readily available and this type of comparison is rarely if ever published using constant dollar comparisons. Comparison should use either entire year total dollars or should use SAAR dollars.
Current dollars = dollars are reported in the value of the year reported 2008=2008$, 2015 = 2015$. 99% of news reports use current dollars and therefore do not account for the influence of inflation.
Constant dollars = all dollars adjusted to represent dollars in the year of comparison. Adjusts for inflation so 2008$ in this case are converted to equivalent 2015$.
It’s not to hard to understand why we need to use constant dollars when you think of it in terms of buying products like food or heating oil. Today heating oil costs $1.90/gallon. In 2008 heating oil cost $3.50/gallon. So, with respect to oil, $350 in 2008 dollars is no different than $190 in 2015 dollars.
In addition to the Year-to-date growth values, here’s two more less common stats for looking at the same information.
Percent change from the last cycle high current$ (contant 2015$)
- Residential – Q1 2006 $390b vs $$680b -43% (-50%)
- Nonresidential Buildings – Q1 2008 $392b vs $$439b -11% (-22%)
- Nonbuilding Infrastructure – Q1 2008 $298b vs $$286b +4% (-10%)
Percent change from the recent recession low current$ (contant 2015$)
- Residential – Q1 2011 $390b vs $$239b +64% (+48%)
- Nonresidential Buildings – Q1 2011 $392b vs $$267b +49% (+35%)
- Nonbuilding Infrastructure – Q1 2008 $298b vs $$243b +22% (+13%)
Constant dollars makes a huge difference in the statistics. Just take a look at Nonresidential buildings. Current dollars would indicate we are now only 11% below the previous high and we’ve had growth of 49% from the recession low. Constant dollars adjusting for inflation shows we are still 22% below the previous cycle high and we’ve had growth of only 35% since the recession lows.
You can find a complete section providing constant dollar cost comparison in my quarterly report. Access the report through the Featured Economic Report tab at the top of this blog
October housing starts released Nov. 18th didn’t come in as expected. The annual rate for October is 1,060,000 new starts vs 1,191,000 in September and 1,079,000 in October last year. BUT look a little deeper than just one month.
The last 4 months of starts have been pretty high, averaging 14% higher than the previous 4 months and 16% higher than the same 4 months last year.
Take a look at this chart. Monthly starts periodically peak and dip erratically. Look at February 2015, the biggest dip in 5 years. But then notice it took less than 4 months for starts to come right back to the trend line and the trend remained intact. This is how the monthly housing starts data goes.
So don’t get too alarmed over one month of data. Now if this downward trend were to continue for several months, go ahead get concerned, but that hasn’t been the pattern.
( Also See 1-31-2016 comments and chart on inflation )
Over the last 24 months work volume has increased and short-term construction inflation has increased to more than double consumer inflation. It appears construction inflation is already advancing faster than and well ahead of consumer inflation, which supports that consumer inflation is not an indication of movements or magnitude of construction inflation.
It is always important to carry the proper value for cost inflation. Whether adjusting the cost of a recently built project to predict what it might cost to build a similar project in the near future or answering a client question “What will it cost if I delay my project start by one year?”, whether you carry the proper value for inflation (which can differ every year) can make or break your estimate.
- Long term construction cost inflation is normally about double consumer price inflation (CPI).
- Since 1993 but taking out 2 years of recession (-8%), the 20-year average inflation is 4.2%.
- Average long term (30 years) construction cost inflation is 3.5% even with any/all recession years included.
- In times of rapid construction spending growth, construction inflation averages about 8%.
- Although inflation is affected by labor and material costs, a large part of the change in inflation is due to change in contractors/suppliers margins.
- When construction volume increases rapidly, margins increase rapidly.
- Construction inflation can be very different from one major sector to the other and can vary from one market to another. It can even vary considerably from one material to another.
In the 5 years of rapid growth in spending for nonresidential buildings from 2004 through 2008, nonresidential buildings cost inflation totaled 39%, or averaged ~8% per year.
In the 6 years of spending during the residential construction boom from 2000 through 2005, residential building cost inflation totaled 47%, or averaged ~8% per year.
Neither the producer price index (PPI) for construction inputs nor the CPI are good indicators of total construction cost inflation.
Some construction cost indices include only the cost changes for a market basket of labor and materials and do not include any change for margins. Those indices are not a complete analysis of construction cost inflation.
Construction cost inflation must include all changes related to labor wages, productivity, materials cost, materials availability, equipment and finally contractors margins. Margins are affected by the volume growth of new work and demand for new buildings. So be sure to verify what is included in any cost index you reference for real construction cost inflation.
For the last three years residential construction inflation has averaged 5.7% and nonresidential buildings inflation has averaged 4.2%. Nonresidential buildings cost inflation has increased for five consecutive years. Both are likely to increase next year since anticipated volume in both sectors will grow next year.
In my construction spending data set, which goes back to 1993, there were six years with greater than 9% spending growth. By far the largest spending growth years were 2004 and 2005, 11.2% and 11.5%. We are about to repeat that historic level of spending growth. I am predicting 2015 will finish with growth of 11.6% and 2016 will experience 11% growth.
(8-12-16) 2015 finished at 10.6% because 2014 was revised up. Construction spending for 2016 will probably finish closer to 8%.
I expect historic levels of growth in spending will be accompanied by inflation relative to historic high growth periods. Don’t expect long term average inflation in high growth periods. Don’t be caught short in your construction cost budgets!
Graphic updated 1-8-16
The chart shows the low and high range of various independent nonresidential buildings construction actual cost indices. In 2015, the range of estimates was from 2% to 5%. The actual inflation came in at 4%. The plotted line is my result of where inflation actually ended up. A chart for residential construction would show much different values.
( Also See 1-31-2016 comments and chart on inflation )
The good news is 2016 construction spending will be up across the board, although growth will vary between the three major sectors, residential, nonresidential buildings and nonresidential infrastructure. This post is about how we will get there.
This plot shows a three year pattern of spending for nonresidential construction. The seasonally adjusted annual rate is plotted every month and the horizontal bars show the total average spending for each year.
I often include a linear trend line in my plots because it’s good to be reminded of the long term direction and rate of growth rather than the monthly fluctuations. I’ve removed the trend lines from this plot to make a point. A fairly typical growth year shows spending finishing the year at a higher rate than when we started. The total spending for the year (shown by the horizontal bars) is the sum of the unadjusted monthly values, which is the same as the average of the monthly adjusted rates of spending. The plot above shows spending for buildings totals of $330 billion in 2014, $390 billion in 2015 and $430 billion in 2016.
If spending always occurred evenly we would see a smooth constant slope plot line indicating the rate of growth, sometimes punctuated with minor monthly declines but over the long term up at a constant rate. The plot for buildings shows a pretty constant growth pattern from March 2014 to March 2016. We see a few slight monthly dips and a flat spot in mid-2015, but overall fairly constant growth.
That will all change in 2016.
The plot shows both buildings and infrastructure will experience multi-month declines in 2016. By midyear both will dip so low that the year over year comparison will drop to near zero percent. Spending on buildings will drop more than 10% from the 1st quarter 2016 high to the midyear low. (edit 11-16. Most recent analysis is showing the nonresidential buildings peak and drop is moved out about 4 to 6 months. The plot above does not reflect this most recent analysis.) Infrastructure will drop in a similar pattern but not as dramatically. Both will resume growth and finish the year at or near the yearly high. Together these major sectors make up 60% of total construction spending with residential being the other 40%. The magnitude of these declines will be large enough to drag total construction spending down for the period, but it too will resume growth and finish the year well above where it started.
What causes that pattern?
Spending in any given month is the sum total contributed by all the projects that started and are currently underway. That includes spending from projects that started recently with foundations just coming out of the ground and also projects that started 18 to 24 months ago that are near completion and near ready for occupancy. Spending is affected by the pattern of starts recorded over the previous 24 months. Only if that pattern is even in growth will spending be even in growth.
Let’s make some assumptions for an example that should be easy to understand.
- All projects take 20 months to build.
- The construction budget gets spent at a constant rate of 5% per month.
Of course, reality never occurs as simple as that, but it makes for an easy to understand example. With these simple assumptions we know this about spending;
- Every month includes some spending from all starts over the previous 20 months.
- New starts in any given month contribute only 1/20th or 5% to total spending that month.
- Predicted spending 3 months out includes only 15% from new starts, 4 months out 20%, etc.
Now, back to reality.
Nonresidential buildings spending in 2016 includes projects that started all the way back to mid 2014. From April 2014 to April 2015, concentrated mostly in mid to late 2014, there were several instances where monthly starts exceeded the previous month by 50%-60%, then settled back. Those were great months and indicated a huge growth spurt in nonresidential building construction. We see that growth from early 2014 all the way up to a peak in early 2016. Nonresidential buildings spending experienced growth of 9% in 2014 and 19% in 2015, with 9% growth predicted for 2016.
We don’t typically see a seesaw pattern in spending from the starts when all the projects are ongoing at the same time, we see flatter and steeper rates of spending growth. The six biggest months from that period averaged more than 30% higher than the current rate of new starts growth. So we’ve had a pause in the rate of increase for new starts. We see the big affect on the spending pattern when a very large volume of spending from starts 18 or 24 months ago reaches completion and drops out of the current month spending. That is what will happen in 2016.
New 2016 nonresidential buildings starts is not the cause of a 10% dip in spending in 2016. It is the completion of a very large volume of starts from 2014 that will no longer be contributing a large share to monthly spending. That will work itself out by year end. But for those who do not look at the patterns that contribute to current spending it will create quite a stir. Three to five months of consistent declines in spending will look like the end of recovery in nonresidential construction. It is not the end.
Here’s a peak at expected annual Construction Spending for 2015 and 2016.
My predictions (GBCo) include the latest actual construction spending data released Nov. 2nd for September spending. This updated projection also includes revised future spending based on Dodge Data & Analytics construction starts released at the DDA Outlook 2016 conference Oct. 30th. My prediction for total spending in 2015, now at $1.075 trillion, hasn’t changed much (up 0.7%) since August. However, for next year my projection has increased from $1.150 to now expecting total spending of $1.190 trillion in 2016. Both projections will be further refined in my winter economic report.
Why are my predicted values so much higher than other estimates?
I put emphasis on using the cash flows from all previously recorded construction starts to predict future construction spending. I’ve talked about and documented in past reports the correlation between these two data sets. For 2015, with only three months left to go, 80% to 90% of all starts that will generate spending in the final three months are already in place.
Very little affect on total 2015 spending will be brought about by new construction starts in the 4th quarter. New starts could crash to a level less than one half of current trends and that would still not affect total spending enough to get below $1.050 trillion for the year. In order to have the final total spending come in less than $1.040 trillion, the rate of spending for each of the next three months would need to drop off to the level of a year ago. That is not what the cash flows are indicating. In fact, cash flows are indicating spending will increase in the final three months of 2015. The cash flow plot provides us with the direction and the rate of change, but not the actual value of spending.
With the September actual spending values included in the data, the statistical average of predicted spending for 2015 is $1.073 trillion. My cash flow analysis by sector predicts 2015 will finish at $1.075 trillion. In 11 out of 14 years, the actual final value has been within 0.5% of the predicted.
The statistical analysis gives a predicted range for total 2015 annual spending between $1.066 trillion and $1.086 trillion. The actual spending total has not fallen outside the statistical range since 2001, as far back as I’ve been tracking the data. I’m confident that total spending for the year will fall within this predicted range.
This can be a confusing set of housing numbers and I thought needs some clarification. Doesn’t help any that I misquoted my housing statistics in Joe’s interview. Here’s the correct numbers.
New Housing Starts (# new units started) from U S Census.
- 2012 & 2013 up 28% & 18%. 2014 up 8.4%. 2015 expected up 12.6% and 2016 predicted up 15%.
- 2012 added 172,000 new units to total 781,000 for the year
- 2013 added 144,000 new units to total 925,000
- 2014 added 78,000 new units to total 1,003,000
- 2015 expect 127,000 new units to total 1,130,000
- 2016 predict 170,000 new units to total 1,300,000
It’s worth noting here that we would need to go back to 1992 to see another year where the number of new units started in the year exceeded 170,000 units. In the 1970s and early 1980s when total housing units started in a year were near two million units, we see growth years of 400,000 to 600,000 new units in a year. After 1984, only three times have we reached new starts over 170,000 unis in a year, 2012 being one of those years. I anticipate we will reach that mark again in 2016.
New Residential Construction Starts $ from Dodge Data.
- 2012 & 2013 up 32% & 27%. 2014 up 10%. 2015 expected up 16% and 2016 predicted up 16%.
In the following chart of Dodge residential Starts $ we can see the dollar volume of new residential starts stalled from about Q2 2013 through Q4 2013 and then again in early 2014. That slowed spending.
- 2012 & 2013 up 13% & 19%. 2014 up <1%. 2015 expected up 13% and 2016 predicted up 18%.
The 2014 drop in spending is influenced by starts that occurred in the later half of 2013 and through 2014. New units starts monthly were low from May to September 2013 and then again in the 1st quarter of 2014. In the Housing Starts chart above, Jan. 2014 starts 3mo move avg are about the same as Jan 2013, showing the slowed growth. The result is spending dropped from Q4 2013 to Q3 2014. Since then new starts have resumed fairly strong growth and spending for 2015 is expected to finish up 16%. See the period from Aug’14 through Apr’15 when spending increase by 20% in 8 months. I think we will continue with 2016 repeating the same growth, although not without some dips in the monthly readings.
Speaking to Joe’s point on when does this affect GDP, we can see in these charts that the actual spending gets spread out over time, such that any slow down, or in more recent data any acceleration, gets reflected later in the spending numbers, perhaps over the next 9 to 12 months for residential work. Up to the current quarter where we see a dip in new $ volume of starts, prior to that we recorded 6 consecutive quarters of growth in starts. After a flat year in 2014 we are poised to see residential construction spending contribute 13% growth & represent 36% of total construction spending in 2015. For 2016 I expect similar growth at a very substantial pace up to 18% growth.
Here’s a comparison of projections for total construction spending in 2015.
My numbers (GBCo) include the latest actual construction spending data released Nov. 2nd for September spending. This updated projection also includes revised future spending based on Dodge Data & Analytics construction starts released at the DDA Outlook 2016 conference Oct. 30th. My prediction for total spending in 2015, now at $1.075 trillion, hasn’t changed much (up 0.7%) since August. However, for next year my projection has increased from $1.150 to now expecting total spending of $1.190 trillion in 2016. The 2016 projection will be further refined in my year-end report.
With the September spending values in the data, the statistical average predicted spending for 2015 is $1.073 trillion. My cash flow analysis by sector predicts 2015 will finish at $1.075 trillion. In 11 out of 14 years, the actual final value has been within 0.5% of the predicted.
The statistical analysis gives a predicted range for total 2015 annual spending between $1.066 trillion and $1.086 trillion. The actual spending total has not fallen outside the statistical range since 2001, as far back as I’ve been tracking the data.
Here’s a summary of predictions for several of the major markets. Again, my predictions from earlier in the year haven’t changed too much.
In today’s data release from U.S. Census, spending for manufacturing buildings was lowered in both July and August, and September came in lower than I expected. That is the primary mover in the lower prediction for nonresidential buildings. Spending for manufacturing buildings is at an all-time high. Through September, spending on new manufacturing buildings has already reached an all-time annual high. Manufacturing buildings helped 2015 spending for nonresidential buildings reach 19% growth but this won’t continue and I expect 2016 growth of 10%.
Residential spending has been a nice surprise to the upside. The current rate of growth for the last 12 months is 17%/year and this rate of growth is expected to continue again in 2016.