The AIA Midyear Consensus solicits the nonresidential buildings construction spending forecast from a number of firms and publishes those results and the Consensus average. The table posted here includes all the AIA forecasts and Construction Analytics 7-2-21 forecast.
Also included in this table is the year-to-date (ytd) actual spending through May. With 5 months of actual data, that ytd result should sway any forecast for any market estimate of year end result. A review of several years of history over all markets shows there are very few instances in the historical data where year end performance swings by more than 10% from ytd at month 4 or 5. Normal variances for about 80% of instances are in the range of 3% to 5%. So with few exceptions, at 5 months into the year, we could estimate year end will be within +/-5% of year-to-date. And yet, there are many instances in these forecasts that are outside that expected range.
The question is, can we determine, how accurate are these forecasts? Some rudimentary checks and balances, and some simple proportional math, provide the answer.
If you forecast a construction spending mrkt to finish 2021 at -30%, but the ytd after 5 months is -5%, the next 7 months would need to average near -50% to get to -30%. With the change in the yoy rate less than -3%/mo, it can’t happen.
If you forecast a construction spending market to finish 2021 at +3.5%, but the ytd after 5 months is -11%, the next 7 months would need to average +14% above Jun-Dec 2020 to get there. That’s a 25%/mo swing from the current rate that would need to hold steady for 7 months.
Likewise, If you forecast a construction spending market to finish 2021 at +11%, but the cum ytd after 5 months is -3%, the next 7 months would need to average +21% above Jun-Dec 2020 to get there. That’s a 24%/mo swing from the current rate for 7 months. Swings like that just don’t happen.
Another market with a glaring example, this time in almost every forecast. Lodging forecasts in the AIA Consensus range from -14% to -20%, with one wild estimate at -43%. Construction Analytics forecast for Lodging is -26%. The year-to-date is -27%. Well, from April to December 2020, spending fell at a rate of 4%/month. In the 1st 5 months of 2021, spending has been down slightly, still hovering near the December low. There are no indications that spending is poised for a rebound. In fact the forecast calls for spending to continue falling through 2021. The current monthly rate of spending averages -25%/mo from 2020. In order to hit any of the forecasts between -14% and -17%, the current rate of spending would need to flip by 15 to 20 percentage points for all of the remaining 7 months of 2021. Spending would need to increase at a rate of 2.5% per month for the next 7 months. This is a good time to remind everyone that Lodging construction starts last year dropped 45%, so the trend is down, not up. Current indications are that spending will decline 9 out of the next 12 months.
The forecasts in this Consensus report have numerous examples like those above. Nonresidential Bldgs actual ytd for the 1st 5 months is -10.5%. Consensus forecast for 2021 is -3.9%. The next 7 months each would need to avg +1% over 2020 to get there. The monthly rate of spending is currently -6% to -10% below 2020 and has fallen 13 of the last 15 months. That’s not going to flip to +1% immediately and stay at that level for 7 months.
The argument cannot be used that monthly data should not be compared to 2020 because of the rapid decline due to shutdowns skewing all the data. That did not occur in nonresidential buildings. Nonres bldgs spending declined 5% in April, but then it averaged a steady -1.5%/mo decline for the remainder of 2020. As of May 2021, spending is right where it was in December, still 16% lower than March 2020. There are no huge down months in 2020 to which 2021 spending would be compared resulting in a large increase to year-to-date percent.
At midyear, the ytd values give some indication of how the year will end. There are a few examples in historical data in which a market did swing by 10% or more from midyear to year-end, but there is less than 10% chance of a market varying by more than 10% and more than an 80% chance markets vary by only 3% to 5%. Rarely does -2% become +8% or +7% become -3%.
The averages suggest that a 5mo ytd value of -10.5% would end the year somewhere between -5.5% and -15.5%.
8-2-21 updated tabled added Here’s the same Midyear forecasts with year-to-date updated to June spending.