Do Pilot and Demonstration Projects Work? Evidence from a Green Building Program

Authors: Christopher J. Blackburn, Mallory E. Flowers, Daniel C. Matisoff (Georgia Tech), Juan Moreno-Cruz (Erasmus / U. Waterloo). Journal of Policy Analysis and Management, 2020.

Method

Difference-in-difference-in-differences design applied to USGBC LEED-Pilot data, isolating the causal effect of a completed pilot & demonstration (P&D) project on subsequent local green-building certification rates.

Key findings

Local quarterly green-building adoption rates roughly double in the period following completion of a nearby P&D project — a nationwide stock increase of 0.5–1.4%.
Distinguishes pilot projects (experimental, “learning-by-searching/doing” — reduce internal transaction costs for the piloting organization itself) from demonstration projects (showcase verified value to outside/inexperienced adopters — “learning-by-interacting,” reduce information asymmetry for non-participants).
Spillovers reduce non-participants’ certification/implementation timeframes by ~9–12% (about 2.5 months faster) — evidence the effect runs through genuine knowledge transfer, not just visibility/marketing.
Names a failure mode — “herding”: demonstrations that get copied without real knowledge transfer can lock in under-performing technology rather than diffusing genuine learning. Effectiveness depends on whether the P&D actually builds a knowledge-sharing network, not just visibility.
Cites prior work (York, Vedula & Lenox 2018) on P&D projects overcoming “psychological and social barriers” to green-tech uptake.

Why it matters for this archive

This is the closest thing to a controlled experiment on whether demonstration buildings actually drive market adoption, with a quantified effect size and an explicit mechanism. Directly relevant to the proposed Double Play container demonstration at Los Ranchos: the effect is driven by local network learning, not national press — meaning siting and community connection may matter more than media reach.

Full text of the published version is paywalled; findings above are drawn from the freely available accepted manuscript.