Two Divergences, One Number: The Lopsided Reality of February's Construction Backlog

The February Construction Backlog Indicator says the market is stable. The data underneath says something more complicated.

The headline backlog number is up from January. Isn’t that good news?

Technically, yes. Directionally, it depends on where you're standing.

ABC's Construction Backlog Indicator (CBI) rose to 8.1 months in February, up 0.1 from January, but context matters here. January's reading was a four-year low. You don't pop champagne because you climbed off the floor. When you look at the year-over-year picture, backlog is down 0.2 months from February 2025. The correct read isn't that "the market is growing", but rather it's "the market stopped falling, for now."

This is a meaningful distinction if you're budgeting, hiring, or bidding work in 2026.

So is this still a "yellow light" market?

Yes. February's data makes that case more precisely than January did. The headline number is steady enough to feel reassuring. The composition underneath is where the story gets complicated. In construction economics, composition almost always matters more than the headline. That's true of inflation, it's true of employment, and it's very true of backlog.

You mentioned composition. What's driving the 8.1?

Two divergences are running simultaneously, and neither one shows up in the headline figure.

The first is by contract type. Contractors currently under contract for data center work are carrying 11.2 months of backlog. Contractors without data center exposure are sitting at 7.6. That's a 3.6-month gap and it's not narrowing. The data center moat is real, and for firms positioned inside it, the near-term outlook remains strong.

The second divergence is geographic. The Middle States, the industrial Midwest corridor from Ohio to Montana, is the only region posting higher backlog than one year ago. Every other region is flat or negative on a YoY basis. After years of post-pandemic struggle, the Midwest is seeing population and economic growth translate directly into construction demand. That's a reshoring and energy infrastructure signal as much as anything else, and it's worth watching.

What should owners and contractors make of the data center split specifically?

A few things. First, if your firm is on the right side of that 11.2-month figure, don't mistake the moat for a guarantee. ABC Chief Economist Anirban Basu notably flagged mid-survey that the conflict in Iran introduces real uncertainty around materials prices, borrowing costs, and owner decision-making. Data center work should continue at pace over the next few quarters, but the macro environment around it is getting noisier, not quieter.

Second, if you're on the 7.6-month side of that split, the composition of your backlog matters even more. Seven and a half months of mixed commercial work carries very different escalation and margin risk than seven and a half months of healthcare or industrial. Blending those together into a single backlog number, or a single escalation assumption, is where preconstruction analysis tends to go wrong.

Is there anything in the Construction Confidence Index that deserves more attention than it's getting?

Staffing expectations. ABC's Construction Confidence Index (CCI) series for staffing level expectations rose in February to its highest reading since March 2025. On the surface that sounds unambiguously positive. Contractors are planning to hire, which means they believe the work is coming.

But hold that alongside what's happening on the cost side. Profit margin confidence is under simultaneous pressure from rising oil prices and construction input costs. What you have is contractors staffing up into a potential margin squeeze. That's not a contradiction: it's a lag. Hiring decisions are being made today based on workload optimism, while the input cost environment that will determine whether that work will actually be profitable is still unfolding.

For owners, that lag has a direct implication. The labor availability window that contractors are signaling right now may be shorter than it appears. If input costs persist or accelerate, hiring momentum reverses, and it tends to reverse faster than it built.

You mentioned Iran twice now. How seriously should construction economists take that variable?

Seriously enough to track, carefully enough not to sensationalize. The reason it belongs in this analysis isn't geopolitical speculation: it's that Basu cited it explicitly as an active disruption during the survey period itself. That's not a tail risk footnote. That's the data's own author flagging a real-time signal he couldn't fully account for in the numbers being reported.

The transmission mechanism is straightforward. Oil prices affect diesel, transportation, and petrochemical-derived materials. If that feeds through to construction inputs in a sustained way, it puts additional pressure on the same margin confidence that was already softening in February. Watch the PPI data for nonresidential inputs over the next two months. That's where this either becomes a story or fades.

What's the takeaway for owners and preconstruction professionals?

Know which side of the divergence your portfolio sits on and build your assumptions accordingly.

If your work is in the data center, pharma, or healthcare corridor, the backlog picture is strong but not unconditional. Price your escalation to the input environment, not the CPI headline. A 2.4% CPI print, or even the Fed’s preferred PCE measure at 2.8%, means very little when your actual basket of copper, structural steel, and skilled craft labor is running at a materially different rate.

If your work is in the broader commercial market, February's stabilization is welcome but shouldn't be mistaken for recovery. Backlog is still below where it was a year ago, the regional picture is uneven, and the macro headwinds of input costs, borrowing costs, and owner sentiment haven't resolved.

8.1 months is the number everyone will quote this month. What it doesn't tell you is which 8.1 months you're actually holding.