The Backlog Number Has Never Described Small Contractors
Associated Builders and Contractors published its June backlog reading on July 14, from a member survey run June 22 to July 8. The Construction Backlog Indicator came in at 8.8 months, down 0.3 months from May and up 0.1 months from a year earlier.
That is the industry number and it is the one that gets repeated.
ABC also publishes the series behind it as a spreadsheet, broken out by firm revenue, region and segment. It runs quarterly from 2009 and monthly from January 2019. Almost nobody opens it.
Source: https://www.abc.org/Portals/1/CEU/CCI_CBI_Time_Series_7.9.2026.xlsx
I pulled the four revenue bands for all 90 monthly readings. Here is the thing that struck me, and it is not about this June.
In all 90 months, contractors with 100 million dollars or more in revenue carried more backlog than contractors under 30 million. Ninety out of ninety. The average gap is 4.57 months. The narrowest in that window is 1.33 months, in November 2023. The widest is 8.08, in July 2021.
Over the last three years the gap averages 4.35 months, with the smallest band sitting at 7.52 months on average and the largest at 11.87.
Two honest qualifications before I go further, because both are the kind of thing I would want to know if somebody handed me this.
The gap has closed exactly once in the file, and it is outside that window. Before 2019 the series was quarterly rather than monthly, and in Q2 2018 both bands read 8.09, dead level. That is one row out of 129. I am quoting the monthly window because it is the consistent one and because that is the data I actually worked with, not because the quarterly rows disagree, and one of them does.
And the four bands are not a clean ladder. All four line up in size order in only 53 of the 90 months. The middle two cross each other often. What does not happen, in any of the 90, is the largest band falling below the smallest.
The industry backlog figure is the average of those. It has never described the firms at the bottom of it, and it has never described the firms at the top either. It describes a middle that a lot of contractors are not in.
What June looks like inside the bands
June 2026, months of backlog by revenue band:
Under 30 million dollars, 7.69
30 to 50 million, 8.38
50 to 100 million, 10.19
100 million or more, 12.33
Total, 8.84
The published headline was 8.8. Three of the four bands sit more than a month away from it. Only the 30 to 50 million band is close.
What is driving the top
ABC chief economist Anirban Basu is direct about it, and calls the pattern a headwind for smaller contractors in those words.
His figures: "The 13% of ABC members under contract to work on data centers have significantly higher backlog (11.0 months) than the 87% that are not (8.5 months)." And on who is in that work: "just 8% of contractors with less than $100 million in annual revenues have data center work under contract, well below the 41% share of contractors with greater than $100 million in annual revenues."
Two figures to keep apart, because they are easy to run together. The 8.5 months is the backlog of everyone without data center work, at any size. It is not the small contractor number. The small contractor number is the 7.69 above. They answer different questions and averaging them produces something that is in neither.
Why smaller firms are not in that work, and it is not skill
Nothing about a hyperscale data hall requires better conduit bending than a hospital does. The trades are the same trades. What sorts these firms is procurement, and this part is my argument rather than anything ABC published.
If you have filled out a prequalification package for a large program, you know it is not a bid. It is a filter that runs before anyone is allowed to bid. It asks for bonding capacity, usually single job and aggregate. It asks for an experience modification rate. It asks for your largest completed project by dollar value, and often for references at that size. It asks how many people you can field, by trade, by month, across the schedule. It asks for named superintendents and sometimes their resumes.
A firm doing forty million a year can be excellent and still fail on the largest completed project line, which is a fact about their history rather than their work. They fail it before anyone opens a price. And the only way to fix that line is to have already done a project of that size, which requires clearing the filter the line exists to enforce.
That is an ordinary set of controls, and it compounds in one direction. A persistent gap is what a filter that never resets looks like from the outside.
The part I think actually hurts, and this is my argument too
Nothing in the ABC release measures labor competition. What follows is reasoning, not a finding.
ABC's own methodology puts data centers inside its Commercial, Institutional and Light Industrial segment, in a list that also contains K-12 schools, hospitals, warehouses, hotels and multifamily. By ABC's own classification the data hall and the school are in the same bucket. In practice they draw electricians and pipefitters from the same regional pool.
So when a campus lands in your metro, wage expectations move for everyone bidding there, not only for the firms on the campus. A contractor at 7.69 months is hiring into a labor market priced by a segment they are mostly not in.
Something in the same release is consistent with that, though I would not call it proof, and the source's own economist reads it differently. Confidence on sales and staffing both rose in June while confidence on profit margins fell to a seven month low, with all three still above 50. Basu attributes that to input prices generally: "The effect of rising input prices may be weighing on contractor profitability." That is his reading and it does not require my labor argument.
What is wrong with these numbers
The year over year band changes are noise, and they are the easiest thing here to over-read. Between June 2025 and June 2026 the total rose 0.15 months while the two bands under 50 million fell, by 0.34 and 0.44, and the two above rose, by 0.95 and 0.46. That looks like a break in the data at 50 million dollars. It is not. Check the neighboring months and the pattern inverts: the smallest band rose year over year in March and May, the largest fell in February and March, and in April the split fell at 100 million instead. Month to month, across the same 90 readings, the 30 to 50 million band moves by 1.20 on average and once swung by 5.09 in a single month, from 12.29 in January 2019 to 7.20 in February. Those moves are several times the year over year change itself. ABC publishes no respondent count per band, so there is no way to size the sampling error. One month's cross-band pattern is not a threshold.
The 90 month gap is a different kind of claim, which is why it is the one worth keeping. A pattern that holds in 90 of 90 readings is not a coincidence in the way that one month's ordering is.
The bands are not fixed sets of companies. A firm that grows past 30 million dollars leaves one band and joins another. So the persistent gap is a statement about size categories, not about the same companies being stuck. It is consistent with firms graduating upward over time, and the series cannot distinguish those stories.
Backlog is a ratio, and the denominator is last year's revenue. ABC publishes the formula: current month's backlog divided by prior fiscal year revenues, times twelve. For a single reporting firm that denominator is last year's number and does not move between its own fiscal year ends. It is tempting to conclude that this makes the published index a clean monthly read of contracted work. That is more than the source supports. ABC says it "has striven to encourage consistent reporting by individual member firms," which is an effort rather than a fixed panel, and the release says the index uses "the latest financials available." How much of a monthly move is composition rather than contracting is not determinable from what ABC publishes.
It is nonresidential only. ABC's methodology describes members "active in various nonresidential construction segments." Its commercial bucket includes multifamily and mixed use, but there is no single family homebuilding in it.
It is self reported, and it is one association. ABC is a merit shop association, its own term, and this is a survey of who answered rather than a census of contractors.
One note on the release's framing. Basu describes June as "still longer than any point from September 2023 to April 2026." The series puts September 2023 itself at 8.9887, above June's 8.8437, so the sentence works on a reading that excludes that month. Anyone opening the spreadsheet lands on the same row.