DataMethodSeasonality
How to seasonally adjust monthly billings with the Census method, where August construction runs 29% above January
Census divides each month of construction spending by a seasonal factor before it reports the headline rate. A controller can run the same division on monthly billings in a spreadsheet, and this page walks through it with an example firm.
Not seasonally adjusted
+31.8%
$155.5 billion in January to $204.9 billion in August
Seasonally adjusted annual rate
+1.4%
Same months, with normal seasonal swings removed
Total US construction spending per month, billions of dollars, September 2024 to September 2026. The adjusted line is the SAAR divided by 12. September 2026 is preliminary and August is revised. Source: Census Bureau via FRED, series TTLCON and TTLCONS. Download the data (CSV)
Census's unadjusted count of construction work put in place rose 31.8% from January to August 2026, from $155.5 billion in January to $204.9 billion in August. Over the same months the seasonally adjusted annual rate (SAAR), Census's estimate of what a full year would total at that month's pace once normal seasonal swings are removed, rose 1.4%. Both figures come from the November 2 release, in which September is preliminary and August is revised.
A controller who reviews monthly billings faces the same arithmetic. A sitework or paving contractor bills more every spring whether or not it is winning more work. The method below follows the Census Bureau's own explanation of how it adjusts construction spending, cut down to a spreadsheet that can sit next to the WIP schedule.
What a seasonal factor is
A January value of 100,000 with a factor of 0.80 adjusts to 125,000
A seasonal factor is one number per calendar month that says how far that month usually runs above or below the year's average, with 1.00 as average. Census's seasonal adjustment Q&A gives the arithmetic. Divide the month's unadjusted value by its factor.
Census also removes trading-day effects, which come from how many weekdays fall in a month, and moving-holiday effects such as Easter. It multiplies those factors by the seasonal factor and calls the product the combined factor. The SAAR is the adjusted monthly value times 12.
The survey methodology says Census estimates the factors with X-13ARIMA-SEATS software and re-estimates them every month as new data arrive. For series driven by weather, the Q&A describes the factors as estimates of average weather effects for each month. An unusually cold January or a wet spring stays in the adjusted number.
The national pattern
A normal August put in place 29% more work than a normal January
FRED carries both versions of each Census series, so the implied combined factor for any month is the unadjusted value divided by one-twelfth of the SAAR. For all construction in 2025, the factors ran from 0.857 in January to 1.107 in August, so a normal August put in place 29% more work than a normal January.
Implied combined seasonal factors for 2025: unadjusted monthly spending divided by SAAR / 12, where 1.00 is an average month. Highway and street includes public and private work. Computed from Census data via FRED, series TLHWYCON, TTLCON and PNRESCON with their adjusted versions.
Highway and street work had the widest range, from 0.595 in January to 1.341 in August. An August highway month carries about 2.25 times the work of a January. Private residential ran from 0.826 in January to 1.130 in August. Private nonresidential moved least, from 0.937 in February to 1.037 in October.
| Month | Total constructionTotal | Private residentialResid. | Private nonresidentialNonres. | Highway and streetHighway | Highway and street against 1.00 |
|---|---|---|---|---|---|
| Jan | 0.857 | 0.826 | 0.941 | 0.595 | |
| Feb | 0.860 | 0.831 | 0.937 | 0.632 | |
| Mar | 0.939 | 0.954 | 0.977 | 0.719 | |
| Apr | 0.990 | 1.005 | 0.998 | 0.901 | |
| May | 1.043 | 1.073 | 1.008 | 1.086 | |
| Jun | 1.085 | 1.113 | 1.019 | 1.205 | |
| Jul | 1.098 | 1.129 | 1.022 | 1.301 | |
| Aug | 1.107 | 1.130 | 1.033 | 1.341 | |
| Sep | 1.080 | 1.079 | 1.032 | 1.303 | |
| Oct | 1.054 | 1.042 | 1.037 | 1.212 | |
| Nov | 0.982 | 0.963 | 1.007 | 0.974 | |
| Dec | 0.903 | 0.861 | 0.989 | 0.707 |
Each factor is the month's unadjusted spending divided by one-twelfth of its SAAR. The bars run left of the center rule for highway months below 1.00 and right for months above. Computed from Census data via FRED, series TTLCON, PRRESCON, PNRESCON and TLHWYCON with their adjusted versions. Download the factors (CSV)
These national factors make a rough check on your own. A firm whose work looks like highway paving should expect a deeper winter trough than the total line suggests.
Building factors from your own billings
The example firm's factors run from 0.70 in December to 1.29 in August
The steps use only monthly billings from the accounting system.
- Pull monthly billings for at least two full years. Three to five years give steadier factors. The example uses two to keep the table short.
- For each year, divide each month's billings by that year's monthly average, which is the annual total divided by 12.
- Average the ratios for each calendar month across the years. The result is the seasonal factor.
- Check that the 12 factors sum to 12. If they do not, multiply each one by 12 divided by their sum.
The example below uses round numbers. A sitework contractor billed $12.0 million in 2024 and $14.4 million in 2025.
Each month's factor averages its 2024 and 2025 ratios, and the 12 factors sum to 12.00
| Month | 2024 | 2025 | Seasonal factor | Against 1.00 | ||
|---|---|---|---|---|---|---|
| Billings | Ratio | Billings | Ratio | |||
| Jan | 700 | 0.70 | 888 | 0.74 | 0.72 | |
| Feb | 720 | 0.72 | 840 | 0.70 | 0.71 | |
| Mar | 860 | 0.86 | 1,056 | 0.88 | 0.87 | |
| Apr | 960 | 0.96 | 1,176 | 0.98 | 0.97 | |
| May | 1,080 | 1.08 | 1,272 | 1.06 | 1.07 | |
| Jun | 1,180 | 1.18 | 1,440 | 1.20 | 1.19 | |
| Jul | 1,260 | 1.26 | 1,488 | 1.24 | 1.25 | |
| Aug | 1,300 | 1.30 | 1,536 | 1.28 | 1.29 | |
| Sep | 1,200 | 1.20 | 1,464 | 1.22 | 1.21 | |
| Oct | 1,100 | 1.10 | 1,296 | 1.08 | 1.09 | |
| Nov | 940 | 0.94 | 1,104 | 0.92 | 0.93 | |
| Dec | 700 | 0.70 | 840 | 0.70 | 0.70 | |
| Year | 12,000 | 12.00 | 14,400 | 12.00 | 12.00 | |
Worked example, not reported data. Each ratio is the month's billings divided by that year's monthly average, and the seasonal factor averages the two years. Download the example (CSV)
This firm's factors run from 0.70 in December to 1.29 in August, closer to the national highway pattern than to the total.
Reading 2026 with the factors
Most of September's 8.0% drop in billings came from the season
Divide each 2026 month by its factor to get adjusted billings. Multiply by 12 for an annual rate on the same footing as last year's total.
Example contractor, thousands of dollars per month, January to September 2026, round numbers. Bars are billed amounts. The green line is billed amounts divided by the seasonal factor. Worked example, not reported data. Download the example (CSV)
Billings rose 66.7% from January to August and fell 8.0% in September. After adjustment, the annual rate fell in every month, from $15.0 million in January to $13.7 million in September. That September rate is 5.0% below the $14.4 million the firm billed in 2025.
| Month | Billed | Seasonal factor | Adjusted billings | Annual rate |
|---|---|---|---|---|
| Jan | 900 | 0.72 | 1,250.0 | 15,000 |
| Feb | 880 | 0.71 | 1,239.4 | 14,873 |
| Mar | 1,060 | 0.87 | 1,218.4 | 14,621 |
| Apr | 1,170 | 0.97 | 1,206.2 | 14,474 |
| May | 1,280 | 1.07 | 1,196.3 | 14,355 |
| Jun | 1,410 | 1.19 | 1,184.9 | 14,218 |
| Jul | 1,470 | 1.25 | 1,176.0 | 14,112 |
| Aug | 1,500 | 1.29 | 1,162.8 | 13,953 |
| Sep | 1,380 | 1.21 | 1,140.5 | 13,686 |
Adjusted billings are billed amounts divided by the seasonal factor, and the annual rate is adjusted billings times 12.
Adjusted billings fell 1.9% from August, so most of September's 8.0% drop came from the season. The 1.9% is the change a controller would want to explain at the monthly review.
Checking year-to-date billings against budget
Against the seasonal expectation, billings through September are 4.7% behind
The same factors spread an annual budget across the months. Suppose the firm budgeted $15.0 million for 2026.
A straight-line budget expects nine-twelfths of the year, or $11.25 million, through September. The factors for January through September sum to 9.28, so a normal year bills 77.3% of its total by then, or $11.6 million.
Example with round numbers. Expected billings through September 2026 against a $15.0 million annual budget, in thousands of dollars. Bars start at zero. Worked example, not reported data. Download the example (CSV)
Against the straight line, actual billings of $11.05 million are 1.8% behind. Against the seasonal expectation they are 4.7% behind, because a firm with this seasonal pattern should be ahead of the straight line by the end of the summer. The same factors can spread a cash-flow forecast or a line of credit draw schedule across the year.
Revisions and caveats
Limits of the method
Factors built from two or three years move with any one bad month. A January lost to storms lowers that month's factor and makes the next normal January look strong. The weather delay days post covers how NOAA climate normals set the expected weather for each month.
A change in mix changes the pattern. A firm that adds interior or service work is likely to have a flatter year than its old factors assume, so rebuild them when the mix shifts.
Census revises its factors when it revises the unadjusted data and as later years arrive, so the 2025 factors above will move slightly. The Q&A also says adjusted months will not sum exactly to the unadjusted annual total. Compare annual totals in unadjusted dollars.
Adjusted billings still include price changes. The inflation adjustment post shows how to divide by a producer price index to separate price from volume.
How to use it on your jobs
Add two columns to the monthly billings report
Add two columns to the monthly billings report, adjusted billings and the annual rate. The backlog in months post divides backlog by prior-year or trailing twelve-month revenue, and dividing by the annual rate as well puts backlog against the current pace of billing. For the national comparison, the September construction spending post covers the latest Census release.
To start, export monthly billings for the last three fiscal years into one sheet with a column for each year. Divide each month by its year's monthly average and average across the years.