PLACES Lab

Overview

PLACES Lab is a research group in the Department of Earth & Environment at Boston University, led by conservation scientist Christoph Nolte, that studies the costs and effects of land-use and conservation policy by combining large-scale land-transaction data, remote sensing, and machine learning.[1] Its best-known output is a nationwide, high-resolution dataset of estimated fair market land value for private parcels across the contiguous United States, built by training tree-based ensemble models on roughly six million land sales.[2] Nolte's 2020 PNAS paper describing this work found that the cost proxies used in prior US-wide conservation-planning studies had underestimated the budgets needed for floodplain protection and species-conservation planning by factors of roughly 2 and 37.5 respectively, and that the new estimates predict actual conservation costs up to 8.5 times more accurately than the proxies they replaced.[2]

For this wiki, PLACES Lab is notable as a modern, non-Georgist demonstration that estimating land value separately and at scale — the core practical requirement of a land value tax — is technically achievable using contemporary data and statistical methods, complementing the assessor-focused mass appraisal methods already documented on this wiki.[2][3] The lab's motivating purpose is conservation-cost estimation rather than tax policy, so its work should not be read as an endorsement of LVT; it is cited here purely as evidence on assessment feasibility.

See Also

• Mass Appraisal Methods — the wiki's survey of statistical techniques (hedonic regression, CAMA) for separating land value from structure value, of which PLACES Lab's approach is a large-scale machine-learning instance
• Objection: Land value can't be assessed accurately — the objection this lab's nationwide dataset speaks to as practical counter-evidence
• David Albouy — an economist whose separate land-value estimation work (vacant-land sales) appears alongside PLACES Lab-style methods in this wiki's land-value-scale literature
• Land Value Tax

Sources

1. PLACES Lab, "People" and lab overview. placeslab.org — used for the lab's institutional home (Boston University, Department of Earth & Environment) and its focus on land policy costs, conservation, and machine learning; corroborated via Boston University's faculty profile of Christoph Nolte. bu.edu/earth/profiles/christoph-nolte
2. Christoph Nolte (2020), "High-resolution land value maps reveal underestimation of conservation costs in the United States," Proceedings of the National Academy of Sciences 117(47): 29577–29583. Free full text (PMC) — used for the paper's methodology (tree-based ensemble models trained on ~6 million land sales), its underestimation findings (factors of ~2 and ~37.5 for two prior studies), and the up-to-8.5x accuracy improvement claim.
3. This wiki's Mass Appraisal Methods concept page — used for the framing of PLACES Lab's work as one instance within the wiki's broader survey of land/structure value separation techniques.