The Bidder Pool Is Shrinking

The short version

In July 2024, researchers at Yale, Columbia and UC Berkeley published a study of what actually drives highway construction costs. It was funded by the U.S. Department of Transportation and the National Bureau of Economic Research. The authors surveyed 123 procurement officials across all fifty states and the District of Columbia, surveyed 211 contractors, and matched the responses against cost data from 250 resurfacing projects.

Three findings sit at the center of it.

First: the practice most strongly associated with lower project costs is bidder outreach — the deliberate work of telling qualified firms a solicitation exists. A one-standard-deviation increase is correlated with a 17.6% decrease in costs: roughly $65,000 per lane-mile, $1 million on a project.

Second: 70% of states rarely do it.

Third, and least discussed: the contractor base itself is contracting. Between 2007 and 2017, almost 70% of states lost highway construction establishments, with the median state down 13% — a loss of roughly seventeen firms.

This is not an outside critique. It is DOT-funded research describing state DOT practice. This paper is about the gap between what it found and what most states currently do.

The finding, stated precisely

From the paper:

"States that do outreach to increase the bidder pool have significantly lower costs, highlighting both the importance of competition and the role the DOT can play in order to increase competition. A one standard deviation (12 percentage point) increase in bidder outreach is correlated with a 17.6% decrease in costs. At the mean, this translates to a decrease in costs of $65,000 per lane-mile and $1 million at the project level."

And the gap:

"Despite the focus on competition and contractor availability in the free response, 70% of states rarely do bidder outreach."

This result is unusually durable. When researchers test many variables at once, some appear significant purely by chance; the correction is a multiple-hypothesis adjustment, which most findings do not survive. This one does — Romano-Wolf adjusted p = 0.01. Among everything the study examined, outreach is one of the most statistically robust relationships in it.

The authors also identify where in the process this lever sits. Their own description of the bid-letting stage:

"The DOT has discretion over when and where to post the call for bids, and over any bidder outreach efforts."

Discretion, not statute. No legislature has to act. No rule has to change. It is already inside the agency's authority.

What officials said in their own words

The researchers asked an open question — what raises construction costs? — and let procurement officials answer freely. Two responses, verbatim:

"Competition: Costs tend to rise when the number of bidders falls (e.g., a single bidder can 'try to name their price). Number of bidders tends to fall as the market reaches capacity."

"Limited funds cause limited projects cause limited contractors cause limited competition. Years of limited work has caused many contractors to get out of the business. Now we have very few contractors. Limited competition causes higher prices."

That second answer is a description of a doom loop, written by someone inside it. Fewer projects thin the contractor base; a thinner base means less competition; less competition means higher prices; higher prices buy fewer projects.

The study's establishment data confirms it is not a perception. Almost 70% of states lost highway construction firms between 2007 and 2017. The market these officials are describing has, in most states, measurably shrunk.

Why the fix goes undone

The study offers two possible explanations for the 70%, and is careful not to choose:

"This could be because the DOT has low capacity or willingness to do outreach, or because they know the market well and no other capable firms exist in the area."

The second reading is genuinely possible in some markets and should not be dismissed. But the paper's other findings point mostly at the first.

A one-standard-deviation increase in DOT employment per capita is correlated with 16% lower costs — drawn from Census administrative employment data rather than the survey, which makes it methodologically stronger than most of the study. Capacity and outreach tell the same story twice. An office without people cannot perform the practice that saves the money.

There is a structural reason outreach loses that fight. It is unbudgeted, unstaffed, and invisible in every metric a DOT reports. Nobody is measured on how many firms heard about a letting. They are measured on whether the letting happened on schedule. Outreach is the work that gets cut first and shows up in the cost data last — a year later, as a bid that came in high because only two firms showed up.

The mechanism is friction, not persuasion

It is worth being precise about what outreach means here, because it is easy to imagine a sales campaign. It is closer to removing obstacles.

The study describes several places where a DOT's own process narrows the field:

• Where and when the call for bids is posted — explicitly at the agency's discretion
• Pre-qualification requirements, which a firm must complete before it can bid
• Fees to access project plans — the paper notes bidders "may have to pay a small fee in order to receive access to the project plans"
• Subcontracting limits. The study found that caps on how much work may be subcontracted are positively correlated with costs, because they shrink the set of prime contractors capable of taking the job at all

Each is defensible on its own terms. Together they form a filter, and the filter is invisible to the agency operating it — because the firms it excludes never appear in any record. A DOT sees the bids it receives. It never sees the firms that looked once, encountered a barrier, and stopped looking.

There is also precedent that outreach works as an intervention rather than a theory. The study cites work in St. Paul, Minnesota showing bidder outreach achieving exactly the effect the auction literature predicts.

Timing

Two conditions make this more consequential now than when the research was published.

Infrastructure funding is moving through the same constrained agencies. More projects are being let by DOTs the study found to be capacity-limited and largely not performing outreach. Volume rising against a shrinking contractor base is the precise condition under which thin-bidder pricing shows up — and at scale, a percentage of a much larger program.

The contractor base has not recovered. The establishment decline runs through 2017. Nothing in the study suggests it reversed, and the officials surveyed describe firms that left the business and did not return. Capacity, bonding relationships and estimating staff are slow and expensive to rebuild. A firm that exits highway work does not re-enter because a letting appeared.

What this paper does not claim

We are not claiming more bidders is always better. There is credible research showing that past a point, additional bidders anticipating a winner's curse bid less aggressively, and gains flatten. The money is at the bottom of the range, not the top. Moving a letting from one or two bidders to four or five is where the evidence is strongest — and given that public procurement typically draws two to three bids, that is the range nearly every agency is operating in.

We are not claiming causation. The authors are explicit: "we can correlate specific procurement practices with cross-state data on resurfacing costs, with the acknowledgment that correlation is not causation." We use "associated with," as they do.

We are not using the study's per-bidder figure. The paper also reports that one additional bidder is associated with 8.3% lower costs. We have deliberately not built on it: that coefficient appears only in a specification the authors themselves describe as "over-controlling," drawn from 94 projects in 19 states, and in their cleaner specification the same relationship is statistically indistinguishable from zero. It is the number most often quoted from this paper. It does not survive its own source, so we leave it alone.

Scope, stated plainly: state and U.S. highway resurfacing, non-interstate, one to twenty miles, most contracts under $5 million, projects begun 2018–2019. The outreach finding is about resurfacing lettings — not bridges, not new construction, not design-build.

What follows

If outreach is the highest-return lever, and the constraint is capacity rather than authority, then the practical question is what makes outreach cheap.

It fails for a mechanical reason before it fails for a budgetary one: a letting has to be findable before anyone can be told about it. An agency cannot conduct outreach to firms it cannot identify, and a firm cannot respond to a letting it never sees. Every barrier between the two — a portal that requires an account, a plan set behind a fee, a notice that lives only where existing bidders already look — narrows the pool before outreach begins.

The lowest-cost interventions are the ones that reduce that friction:

1. Make lettings findable outside your own system. A qualified firm in an adjacent state should be able to find your work without knowing your portal exists.

2. Remove cost between the notice and the bidder. Where plan access carries a fee, that fee is a filter — and it filters hardest on exactly the firms that would widen your pool.

3. Re-examine subcontracting limits. The study found these positively correlated with costs. They are usually adopted for good reasons; they should be re-examined against what they do to the prime pool.

4. Then do targeted outreach. This is where the measured savings sit, and it works far better when the underlying notices are already easy to find and share.

5. Measure the pool, not just the letting. Bids per letting. First-time bidders per year. Firms that pre-qualified and never bid. Most agencies do not track these, and what is not tracked cannot be managed.

Where we fit

Mindy publishes public solicitations on a free public map — including state, county, municipal and district work. Agencies post at no cost. Contractors search at no cost. No subscription, no per-document charge, no login between a public notice and the firms that would bid on it.

Public bid notices posted to Mindy stay free to find, free to search and free to receive. An access toll between an agency's letting and its market is a cost the agency pays without receiving the revenue.

Sources

Primary Liscow, Z., Nober, W., & Slattery, C., Procurement and Infrastructure Costs, working paper, July 11, 2024. Funded by the U.S. Department of Transportation and the National Bureau of Economic Research. — Outreach 17.6% / $65,000 per lane-mile / $1 million per project: pp. 4 and 25; Appendix Table E.1, Romano-Wolf adjusted p = 0.01. — 70% of states rarely conduct outreach: p. 17, Appendix Figure C.2. — DOT employment per capita, 16%: p. 4 (US Census Bureau public-sector Highways employment, 1997–2021). — Establishment decline: p. 17, Figure 7 (US Census Bureau 2007, 2012, 2017; NAICS 2373, "Highway, Bridge, and Street Construction"). Almost 70% of states lost establishments; median state −13%, approximately 17 firms. — Subcontracting limits positively correlated with costs: pp. 4 and 25. — DOT discretion over posting and outreach; pre-qualification and plan-access fees: p. 8. — Procurement official quotations: p. 17. — St. Paul, Minnesota outreach precedent: p. 40, citing Government Performance Lab (2016) and Liebman & Azemati (2016).

Scope. State and U.S. highway resurfacing, non-interstate, 1–20 miles, most contracts under $5 million, projects begun 2018–2019. 250 projects with cost data; 123 procurement officials and 211 contractors surveyed.

Deliberately not cited. The paper's 8.3%-per-bidder figure, for the reasons stated above. The paper's employee-quality finding (28.6% higher costs) is not statistically significant after multiple-hypothesis adjustment — unadjusted p = 0.09, Romano-Wolf p = 0.21 — and we do not use it.

Counter-evidence considered Hong, H., & Shum, M. (2002). "Increasing Competition and the Winner's Curse: Evidence from Procurement." Review of Economic Studies, 69(4), 871–898. Open Contracting Partnership (April 2025) — two-to-three bids observation; cross-jurisdictional, not a U.S. statistic.

This working paper should be checked for a later published version before citation in formal documents.

GovCon Giants · Mindy — getmindy.ai