Does Georgism Work? (Doucet, ACX series)

Summary

Lars Doucet's "Does Georgism Work?" series, published on Astral Codex Ten in 2021–22, is the most widely read modern empirical examination of land value taxation. Written for a general but analytically demanding audience, it grew out of Doucet's ACX Book Review Contest-winning review of Progress and Poverty and launched the contemporary Georgism revival in rationalist and effective-altruist circles.

Doucet is a self-identified Georgist and a practitioner-advocate (games-industry analyst, later co-founder of the Center for Land Economics), and states his bias openly at the outset: "I'm a Georgist, and I would be happy to find this philosophy true and sad to find it false" (Part 1). This page therefore treats the series as a navigation source — a well-organized synthesis whose evidential weight rests on the primary studies it cites, not on Doucet's own authority. The "citation ledger" sections below de-reference the actual sources behind Parts 1 and 3, routing each to the primary study (with its own wiki page where one exists) and flagging those not yet ingested. ATCOR and the Henry George Theorem appear in the series as attributed theory, not established results, consistent with Doucet's own caveats.

The Three Parts

1. Is land really a big deal? Examines whether land values are large enough to matter — concluding they are, consistent with Rognlie and Bonnet et al.
2. Can landlords pass LVT on to tenants? Works through tax incidence and argues that, because land supply is fixed, an LVT is borne by the landowner and cannot be passed to tenants.
3. Can unimproved land value be assessed? Tackles the assessment objection and concludes land can be assessed with accuracy at least comparable to existing property assessment.

Part 1 — Is Land Really a Big Deal?

Doucet frames Part 1 as a reply to Paul Krugman, whom he quotes conceding the theory but dismissing the magnitude: on financing a modern welfare state by taxing land, "I don't think you can raise nearly enough money... It's just not a big enough thing" (Krugman, quoted in Part 1). Against this he tests five propositions — most urban real-estate value is land; US land rents equal a sizable share of government spending; land is a large share of bank loans and of gross personal assets; land ownership is concentrated among the wealthy.

Key findings, in Doucet's framing (magnitudes are the cited studies', not his):

• Aggregate US land value spans ~$19–65 trillion (2020) across twelve estimation methods, which Doucet narrows using the Federal Reserve method (~$24T) as a conservative lower bound and Jeffrey J. Smith's Counting Bounty estimate (~$44T) as an upper bound. He argues cost-approach methods (Fed, Lincoln Institute, Larson 2019) systematically understate land because they value depreciated structures at inflated replacement cost, and prefers Albouy's vacant-land-sales method as the most convincing.
• Land share of urban property value is high — his San Francisco "spot check" (a vacant lot vs. an adjacent sold townhouse) yields ~76% land, congruent with the AEI Land Price Indicators 70.9% county figure for SF (2020).
• Land rents cover a sizable share of budgets. Converting land values to rents at urban cap rates of 5–8%, Doucet argues even the low Fed/5% estimate is "enough to cover any one of Defense, Social Security, or Medicare + Medicaid, all by itself," while Smith's figure at 8% approaches a federal "Single Tax" (Part 1).

Doucet is explicit that the single-tax question turns on which estimate and cap rate one accepts, and that his bullish lines are back-extrapolations "just to set an upper bound" — a caveat that should travel with any citation of the $44T / single-tax figures.

Part 1 citation ledger (de-referenced)

Doucet's Part-1 evidence is largely already ingested on the wiki as primary-source pages:

• Land-value estimates: Albouy, Ehrlich & Shin, Metropolitan Land Values (2018) (incl. the "online appendix" showing dense pure-land sales in city centers); Larson, New Estimates of the Value of Land of the United States (2015) (hedonic); Davis, Larson, Oliner & Shui, FHFA (2019) (cost approach, vacant-land sales excluded); Yglesias's Federal Reserve method (2013); Counting Bounty (Jeffrey J. Smith, 2020) — the source of the Foldvary, Tideman, Gaffney and Cord data points.
• Rent-share-of-income anchor: Dwyer, The Taxable Capacity of Australian Land and Resources (2003) — the paper behind Foldvary's "one-third of national income" claim, which Doucet recomputes to ~21% and flags Foldvary as "contradicted by his own source."
• Theory: ATCOR (Gaffney, The Physiocratic Concept of ATCOR, 2005 — attributed hypothesis); Henry George Theorem (Arnott & Stiglitz origin paper; practical guide). Doucet relays Tideman's private-correspondence gloss that a non-pure-public-goods variant holds "the combination of land value increases and charges equal to marginal cost will finance these expenditures," but "neither theorem applies if people have different tastes or if benefits do not decline with distance" (Tideman, quoted in Part 1).
• Policy synthesis: Tideman, Kumhof, Hudson & Goodhart, Post-Corona Balanced-Budget Super-Stimulus (2021), which Doucet points readers to as the worked-out policy paper.

Part-1 citation ledger (updated 2026-07-11): Kuminoff & Pope (2013) and Michael Hudson's Where Did All the Land Go? (2001) now have de-referencing pages, as do three more Part-1 anchors: Barr, Smith & Kulkarni, What's Manhattan Worth? (the vacant-land-sales index valuing Manhattan's developable land at ~$1.47T — published in Regional Science and Urban Economics 2018, not the JUE), the AEI Land Price Indicators (the 70.9% SF land-share figure), and Saez & Zucman (2016) (the wealth-concentration primary). The three remaining minor Part-1 data points — all reaching Doucet second-hand through Counting Bounty — are now de-referenced (registry rows pending ingest, annotated on land-rent-could-fund-government): Steven Cord's estimate is Cord (1985), AJES 44(3):279–293, whose primary headline is 28% of 1981 national income (Smith/Doucet quote it as ~24%); Richard Ebeling's federal-lands figure is the ~$5.5T federal land+mineral value in his There Is No Social Security Santa Claus (Future of Freedom Foundation, 21 Dec 2015), which Doucet himself flags as method-free; and the PLACES Lab machine-learning estimate is Christoph Nolte's Boston-University land-value maps (Nolte 2020, PNAS 117(47):29577–29583; placeslab.org). The ranked ingest shortlist lives in sources/inbox/T1-TODO-doucet-parts-1-3-dereferencing.md.

Part 3 — Can Unimproved Land Value Be Assessed?

Part 3 is the series' assessment installment and the least-ingested of the three. Doucet prepared for it by attending Ted Gwartney's five-week Henry George School course Assessing Land Values — Principles and Methods, surveying ~a dozen mass-appraisal papers from Google Scholar, and reading the International Association of Assessing Officers (IAAO) standards. His structure:

• The basics. Georgist or not, assessors accept the identity Total Value = Land Value + Improvements Value and solve for the third term as a residual using three standard approaches — market, cost, and income. Gwartney's practitioner claim is that land is easier to value than buildings because location has fewer "fiddly variables," so he estimates land directly and takes the building as the residual.
• Checking assessments. Ratio studies against market transactions, plus owner complaints as a feedback signal — Doucet relays the practitioner benchmark (via Hefferan & Boyd) that under-~2% objection rates signal good assessments. On required precision, Gwartney (quoted): "You have a lot of wiggle room. It doesn't have to be perfectly precise. The idea is to improve on what's already being done" (Part 3). Doucet adds the safety-margin argument that because real-world LVT targets ~85% not 100%, moderate over-assessment still leaves rent uncollected rather than over-taxing.
• Staffing. Gwartney ran British Columbia's province-wide assessment with a staff of ~690 for ~5 million people (≈1 per 7,250) — Doucet contrasts this with the IRS at ≈1 per 4,425 Americans to rebut the "army of assessors" objection.
• Modern methods. The empirical core, from recent mass-appraisal papers: multiple regression analysis, nonparametric kernel regression, nonparametric adaptive-weights smoothing, semiparametric regression, and expert-weighted models (AHP / iLVM). His headline evidence is the Berlin study by Kolbe, Schulz, Wersing & Werwatz: the semiparametric method reached a 0.845 correlation with the traditional expert-assessed land-value map (BRW) — roughly ±15% — while the plain kernel method managed only 0.704 and was "not able to produce estimates for the city center." He also cites Almy (2014) for a ~€20-per-property assessment cost (≈1% of receipts) and the developing-country iLVM study (Bencure et al., BayBay City, Philippines) explaining ~67% of variance.

Doucet's conclusion is deliberately hedged and should be quoted as such: "By George, Unimproved Land Value can (probably) be accurately assessed" (Part 3). His stronger claim is only the negative one — that the objection "valuing land separately from improvements is fundamentally impossible" is "plainly ruled out." He closes by calling for a single-location cross-comparison of every mass-appraisal method (he suggests Pennsylvania) — a research gap the wiki records rather than resolves.

Part 3 citation ledger (de-referenced)

• Already on the wiki: Gwartney's assessment essay/seminar; the mass-appraisal-methods concept page (which currently summarizes Doucet's account of the Berlin/Kolbe methods rather than citing Kolbe directly); the land-cannot-be-assessed objection page; IAAO standards.
• Now de-referenced (updated 2026-07-11) — the load-bearing Part-3 sources have their own pages: Kolbe, Schulz, Wersing & Werwatz (the 0.845 semiparametric figure), Hefferan & Boyd (2010) (the <1%-objection / <2%-complaint benchmarks), and Almy (2014) (the €20/property cost figure).
• Now de-referenced (updated 2026-07-11): the Kolbe et al. adaptive-weights-smoothing companion paper (2015) — folded into the Kolbe Berlin page as a marked companion section — and Bencure et al. (2019), iLVM (BayBay City — the sole developing-country case in the set).
• Now de-referenced (updated 2026-07-18): Gloudemans & Almy (2011), Fundamentals of Mass Appraisal (IAAO textbook — the methods backbone) has its own page, built from IAAO's course materials and standards rather than the paywalled full text, which remains inaccessible.
• Still second-hand only (ranked ingest shortlist in sources/inbox/T1-TODO-doucet-parts-1-3-dereferencing.md): Yalpir & Unel (2017) (Konya MRA); Fuest et al. (2018) (ifo Grundsteuer cost comparison — distinct from the fuest-peichl-siegloch incidence paper already on the wiki); plus minor cites Kilić et al. (2019), Raslanas et al. (2014), Aragonés-Beltrán et al. (2008), Xue et al. (2008), and Kettani & Khelifi (2001).

Five Years Later (2026)

In a September 2026 guest post on Astral Codex Ten, Doucet revisited the three-part series and gave a part-by-part self-assessment. He reports that Part 1, on land's magnitude, "has aged the best of the three," and reaffirms its conclusion rather than revising it. His main update there is South Korea: drawing on his own reading of official Ministry of Land, Infrastructure and Transport valuation data and his interviews with Korean researchers, Doucet gives South Korea's land-value-to-GDP ratio as the highest in the OECD — over 500% currently, having run roughly 400–600% over the past 50-plus years — and contrasts this with his own earlier estimate that put total US land value at around 200% of GDP. These figures are Doucet's own calculation, not an independently sourced or verified estimate. He also adds two smaller updates to the magnitude argument: that upzoning is necessary but not sufficient for housing affordability, and that ground-rent leasing deserves more attention as a value-capture route alongside land value tax itself.

On Part 2, tax incidence, Doucet still holds that the weight of the evidence he reviewed shows land value tax is generally not passed on to tenants, but now reports one narrow exception. He argues that under pervasive rent control — where a landlord is specifically permitted to raise a below-market rent by the amount of a tax increase — the tax can be mechanically passed through, and he states that this describes the situation in Denmark, citing the 2024 working paper by Nielsson, Wroblewski & Yding, whose quasi-experiment found a precise zero price effect from Danish land taxes and inferred that the burden is shared with tenants and future purchasers. The pervasive-rent-control explanation for that zero-price-effect result is Doucet's own account of the mechanism at work in Denmark; the authors' own summary reports the zero-price and burden-sharing findings without attributing them to a rent-control channel, so the mechanism should be read as Doucet's gloss rather than their stated explanation. Doucet frames the exception as narrow — "the Danish case is only one data point among more than a dozen others" he originally reviewed — and lists several empirical tests (why landlords don't cut rents when their costs fall; why property developers still discount land-tax holding costs into offering prices) that he says a general capitalization-denying account would still need to answer.

Part 3, on assessment, is the installment Doucet says has aged least well — not because his hedged 2021 conclusion changed, but because years spent working full-time on property valuation gave him a much more detailed picture of the field. His single largest update again comes from South Korea: he now treats the country's national land-valuation system, which values land itself (not merely total property value) down to the individual parcel every year in about five months, as an existence proof that granular annual land assessment is administratively achievable at national scale. He also reverses part of his earlier skepticism of the "cost approach" to assessment, now calling it "basically fine" when properly implemented, and argues that the single most important discipline is simply keeping valuations current, citing Pittsburgh's split-rate repeal as a case study in what stale assessments do to public trust and political viability. He sets these revisions against a broader claim of growing legislative momentum, pointing to the 2026 Virginia and Kentucky land value tax enablement laws as evidence that implementation work, more than persuasion, is now the active front for the movement.

The four questions drew a published rebuttal from a pseudonymous Substack writer and a reply from Doucet in its comments; the exchange, and what the incidence literature makes of it, is set out on Landlords Cannot Pass LVT to Tenants. In October 2026 Doucet began a series on whether land supply is fixed: Part 1, on the Netherlands and land reclamation, syndicated on Progress.org, and Part 2, a parable of pencils and parcels contrasting sales and holding taxes, carried on Deadweight Loss.[4]

Bears On

• Outcome: Landlords cannot pass LVT on to tenants
• Objection: Land value can't be assessed accurately
• Concept: Mass appraisal methods · ATCOR · Henry George Theorem
• Book: Land Is a Big Deal — the book-length expansion of this series

Sources

1. Lars Doucet (2021–22), Does Georgism Work? — used as a navigation / advocacy-adjacent source (§3 source hierarchy: an advocate representing his own position) that works through the three empirical cruxes (is land a big share of value; can LVT be passed to tenants; can land be assessed); the underlying magnitudes and assessment results are attributed to, and de-referenced to, the primary studies on the pages this one links. Part 1: Is Land Really a Big Deal? · Part 2: Can Landlords Pass LVT to Tenants? · Part 3: Can Land Be Assessed?, Astral Codex Ten. Verbatim quotes above (Krugman, Tideman, Gwartney, and Doucet's own hedged conclusion) transcribed from Parts 1 and 3, each ≤50 words with part-level locators. Primary source for the Krugman quotation (which Doucet carries second-hand): Michael Scott Moore, "This Land Is Your Land," Pacific Standard, from Moore's 2008 Berlin interview with Krugman. psmag.com
2. Lars Doucet (2022), Land Is a Big Deal (book). landisabigdeal.com — the book-length treatment; see book page.
3. Lars Doucet (2026), "Does Georgism Work? Five Years Later," guest post on Astral Codex Ten (Scott Alexander), 24 September 2026. astralcodexten.com — a teaser cross-post also runs on the Progress and Poverty Substack at blog.landeconomics.org. Read in full — used as a Tier 2 practitioner-author retrospective (D-claim throughout; his own first-person account of what aged well or poorly in the original series) for the Part 1–3 self-assessment, the South Korea land-value-to-GDP figures (Doucet's own calculation from MOLIT data, not independently verified), the pervasive-rent-control account of the Danish incidence exception (his own gloss, not the Nielsson et al. paper's stated mechanism), the South Korea national land-valuation system as an assessment existence proof, and the framing of 2026 state enablement wins.
4. Lars Doucet, "Is land supply fixed? Part 1: The Netherlands and land reclamation," Progress.org, 11 October 2026, progress.org; "Part 2: A parable of pencils and parcels," Progress and Poverty, 13 October 2026, blog.landeconomics.org — used for the existence and subjects of the series (D-claims; Tier 2).