Brian Potter delivers a claim that cuts against everything we expect about industrial progress: construction costs almost never fall. Not occasionally. Not in downturns. Virtually never. His evidence spans 150 years of cost indexes, multiple countries, and granular task-level data — and the pattern is unbroken.
The Core Pattern
Potter opens with a finding that should unsettle anyone who assumes technology makes things cheaper. "Unlike manufacturing and agriculture, or the economy overall, which generally show improving productivity over time, in the field of construction we find that productivity tends to at best stay constant, and at worst decline over time." The cost data extends the problem further back than productivity metrics suggest. "U.S. construction costs have virtually never fallen with respect to overall inflation." That word — virtually — is the only hedge in a dataset that otherwise shows relentless upward pressure.
The argument works because Potter refuses to rely on a single index. He triangulates across output indexes (Turner Building Cost Index, Census Bureau's Single-Family Constant Quality Index, Craftsman estimates) and input indexes (ENR Construction Cost Index, RS Means, Boeckh). "All else being equal, I prefer output indexes to input indexes, because they should more closely track what we actually care about (the cost of finished buildings), and should be less subject to distortion." Yet both types track each closely — and both show the same stubborn trend.
This lands because it sidesteps the usual productivity debates. We don't need to argue about measurement artifacts or output mix changes. The cost of building a house, an office, a highway — these are concrete. And they rise faster than inflation almost everywhere, almost always.
The Exceptions That Prove the Rule
Potter identifies one major exception: 1975 to 1995. "The major exception is the period from 1975 to 1995, where most indexes show lower rates of increase or even declines against overall inflation." That two-decade window matters. It suggests the stagnation isn't inevitable — it's contingent on specific conditions. The Baumol effect offers one explanation: when productivity rises in other sectors (manufacturing, tech), wages rise across the economy, and construction must pay those higher wages even without productivity gains. Construction becomes relatively more expensive not because it's failing, but because everything else is succeeding.
Potter also notes that historical cost increases appear worse than modern ones. "For four of the five 10-year periods between 1915 and 1965, the Turner Cost index rose more than a percentage point faster than overall inflation, whereas for the periods from 1995 to 2025 it rose less than a percentage point." The mid-century boom in building codes and safety standards — what the Building Code deep dive traces — likely contributed. Higher standards cost more. Whether they delivered proportional value is a separate question.
Task-Level Evidence
The granular data strengthens the case. Potter examines individual construction tasks across RSMeans and Craftsman estimating guides spanning decades. "Some have risen in cost faster than overall inflation; others more slowly. But on average, the cost of these construction tasks has risen at the level of overall inflation." This matters because it rules out simple substitution narratives. If old tasks were being replaced by cheaper new methods, we'd see declines. We don't.
Potter acknowledges the adverse selection problem: "picking tasks that appear in many versions of the estimating guide might deliberately select for ones that have been difficult to substitute." Fair objection. But the experience curve effect — the idea that repeated production should drive costs down — should still operate on these persistent tasks. Nails got cheaper when manufacturing methods changed. Why don't foundations, framing, finishing?
Construction costs rarely fall. Not because we're building worse buildings. But because the entire system resists the pressure that makes other goods cheaper.
Critics might note that Potter's quality adjustment caveat undermines the conclusion. "A modern building might cost more per square foot, but be built to higher standards or otherwise have higher performance than an older building, which looking only at changes in costs won't capture." True. The Census Bureau's Constant Quality Index attempts this correction, but Potter admits most indexes don't. The counterargument holds: we may be getting more bang for our buck. But even if true, the bang is getting more expensive — and that's what matters for affordability.
International Confirmation
The pattern isn't American. Potter extends the analysis internationally, finding similar cost index behavior across wealthy countries. This rules out explanations rooted in U.S.-specific policy failures alone. The stagnation appears structural to construction itself, not to any single regulatory environment.
Bottom Line
Potter's strongest move is refusing to let productivity abstraction hide the real metric: cost. His evidence is comprehensive — 150 years, multiple indexes, task-level data, international comparison. The biggest vulnerability is the quality adjustment gap: we may be paying more for better buildings, not just paying more. But even if that's true, the affordability crisis remains. Watch for whether the 1975-1995 exception can be replicated — or whether the Baumol effect makes it impossible.