Choi & Sjoquist (2015): Economic and Spatial Effects of Land Value Taxation in an Urban Area

Overview

"Economic and Spatial Effects of Land Value Taxation in an Urban Area: An Urban Computable General Equilibrium Approach" by Ki-Whan Choi (Massachusetts Department of Revenue) and David L. Sjoquist (Dan E. Sweat Chair, Andrew Young School of Policy Studies, Georgia State University) appeared in Land Economics 91(3), August 2015, pp. 536–555 (JEL H21, H71).[1] It is a member of the wiki's computable general-equilibrium site-value-tax cluster — this page is the dedicated deep-dive for that cluster's most recent and most elaborated member, obtained and read in full (2026-07-18), not summarized from the abstract alone.

The paper builds an urban CGE model of a monocentric city (a central business district producing a non-housing good, surrounded by a residential ring, surrounded by an unmodeled agricultural sector) and calibrates it to the Atlanta, Georgia metropolitan area, then simulates replacing the local capital property tax with a revenue-neutral land value tax.[1] It explicitly extends DiMasi (1987) — the anchor study of the wiki's CGE cluster — and Sullivan (1984, 1985), differing from both in three ways the authors flag as significant: (1) residents, not absentee landlords, own land, so a land-price change also moves resident income; (2) three income groups rather than one, enabling a distributional read; and (3) housing capital can be modeled as completely mobile, partially mobile, or completely immobile, capturing short- versus long-run adjustment, plus a labor-leisure choice Sullivan's models omit.[1]

Core Findings

• Full capitalization onto the landowner when boundaries are fixed. In the benchmark case, the tax-inclusive capital property tax rate is 18.9% (a 22.9% nominal rate).[1] Under fixed CBD and urban-area boundaries, "landowners bear the entire burden of the tax... it is not possible for landowners to avoid the tax when the boundaries of the CBD and urban area are fixed."[1] Ring-1 land rent falls from $9,940 to $1,346/acre/year (−86.5%), and the tax-exclusive housing-service price falls from $16,669 to $13,973 (−16.2%); the wage rate relative to the user price of capital rises 24.6%.[1]
• A revenue-neutral pure LVT is fiscally feasible for Atlanta — unlike DiMasi's (1987) Boston finding, but at a very high rate. DiMasi (1987) found that switching Boston fully onto land value did not raise enough revenue and had to fall back to a graded (split-rate) system; Choi & Sjoquist attribute their opposite result to Boston's substantially higher property tax rate.[1] For Atlanta, the required revenue-neutral land tax rate is 66.8% with endogenous CBD/urban boundaries, lower than the fixed-boundary case because the urban area spatially contracts.[1] [VERIFY: internal inconsistency in the source] — the paper states the fixed-boundary revenue-neutral rate as 82.9% in one passage ("a tax on just land rent must be 82.9%... in the case of fixed boundaries (Table 6)," p. 544) and as 89.2% in another, three pages later, when the endogenous-boundary rate is compared back to it ("66.8%, which is lower than the 89.2% in the case of fixed CBD and urban boundaries," p. 546) — both verified verbatim against the primary PDF text at their respective locations; this page reports the discrepancy rather than silently picking one.
• Large welfare gain — roughly triple DiMasi's estimate. The switch to a revenue-neutral LVT raises money-metric welfare (equivalent variation) by 19.2% of tax revenue with fixed boundaries and 18.2% with endogenous boundaries — "somewhat larger than" DiMasi's (1987) 6.6% figure for a graded property tax on Boston.[1] A companion split-rate variant (a 12% capital tax paired with a revenue-neutral land tax, which comes out to 50.6%) produces a smaller welfare gain of 8.8%, closer in scale to DiMasi's result — the authors read the size gap between the pure-LVT and split-rate welfare gains as evidence the efficiency gain scales with how completely the capital tax is removed.[1]
• The paper's most distinctive result: switching to a revenue-neutral LVT is progressive. Extending the model to three income groups calibrated to Atlanta's actual income distribution (Group 1, high-income, 22% of households; Group 2, middle-income, 31%; Group 3, low-income, 47%; Group 1 endowed with 1.5× the land and capital of Group 2), the wage rate relative to the cost of capital falls 10.7% for the high-income group, rises 13.7% for the middle-income group, and rises 22.2% for the low-income group — "in other words, switching from a capital property tax to a revenue-neutral LVT is progressive."[1] This is a genuinely new finding for the wiki's evidence base: the existing capitalization studies (Borge–Rattsø, Buettner, DØRS) measure whether the tax capitalizes onto owners, not how the resulting welfare change is distributed by income.
• The urban area expands, not contracts, under the LVT reform when boundaries are endogenous — contrary to Song & Zenou (2006), consistent with Banzhaf & Lavery (2010). The paper explicitly weighs in on a live theoretical dispute: Song & Zenou (2006) predict a property tax contracts the urban area; Choi & Sjoquist "obtain the opposite result," which they read as consistent with Banzhaf & Lavery's (2010) empirical finding (from Pennsylvania split-rate data) that land taxation curbs sprawl rather than causing it.[1] Under the endogenous-boundary simulation, the urban radius shrinks 16.9% (from 30.57 to 25.41 miles) and CBD area shrinks 15.9% when the capital property tax is replaced by land taxation — i.e., removing the capital tax is what increases density and shrinks the urban footprint.[1]
• The welfare gain shrinks sharply — and incidence changes — when housing capital is not freely mobile. With completely immobile housing capital (a short-run proxy), ring-1 land rent falls only 72.9% (vs. 86.4% under perfectly mobile capital) and the welfare gain drops to 5.9% of tax revenue, versus 19.2% in the base case.[1] The authors note the limiting case explicitly: if all capital (housing and non-housing) were perfectly immobile, the capital property tax would fall on a perfectly inelastically supplied factor, so that "a reduction in the tax on capital and an increase in the tax on land would be fully capitalized into prices, and thus there would be no change in welfare" — a clean theoretical benchmark for why the size of the efficiency gain, not its direction, depends on capital mobility.[1]
• Resident vs. absentee ownership changes the magnitude, not the direction. Re-running the model with absentee (nonresident) landowners, the ring-1 land-rent decline is smaller (39.5%) than under resident ownership (43.8%), because resident landowners' own incomes fall with land value, which dampens the demand response that drives the price decline further under the resident-ownership assumption.[1]

Relation to the CGE Literature

Choi & Sjoquist situate the paper as the direct successor to DiMasi (1987) — the anchor of the wiki's CGE cluster — and to Sullivan (1984, 1985), whose closed-region models found landowners bear the entire burden of a property or land tax under full labor mobility, with welfare losses of 6.5% (residential) and 4.6% (industrial) of tax revenue for the reverse direction (LVT → property tax).[1] It also references Follain & Miyake (1986) (the Jamaica CGE, also in the wiki's cluster) and Arnott & MacKinnon (1977) (the methodological precursor), and positions itself as the first of these to calibrate to actual US metropolitan data (Atlanta) rather than hypothetical parameters, and the first to combine three income classes with a labor-leisure margin and variable housing-capital mobility in a single model.[1]

Bears On

• Benefit page (model-evidence tier): Landlords cannot pass a land value tax on to tenants — the paper's fixed-boundary result ("landowners bear the entire burden of the tax... it is not possible for landowners to avoid the tax") is a general-equilibrium confirmation of the same asset-price incidence mechanism the page's capitalization evidence establishes econometrically; carried there as model evidence per the cluster's own tiering, not as a primary anchor.
• Research (cluster anchor): Site-value CGE simulations — this page is the detailed member entry; DiMasi (1987) remains the cluster's headline reference point for the welfare-gain comparison.
• Research: Borge & Rattsø (2014) · Buettner (2003) — the econometric capitalization studies this paper's simulated full-capitalization result is the CGE analogue of.

Limits and Honest Caveats

• Model evidence, not measurement. As with the rest of the CGE cluster, these are disciplined simulations — internally consistent given the assumed elasticities and functional forms, but not observations of an actual Atlanta tax reform (no US jurisdiction has run this experiment at this scale). The wiki's evidence ladder ranks this tier below the quasi-experimental capitalization studies (DØRS, Borge & Rattsø).
• The required LVT rates (66.8–89.2%) are Atlanta-specific and reflect a low starting property-tax burden, not a universal statement about how high an LVT "has to be." England's (2007) five-city calculation, which the paper itself cites, found a revenue-neutral switch infeasible for Chicago and Milwaukee and feasible but at "very high" rates for Philadelphia and Washington, D.C. — the required rate is a function of the tax base being replaced, not a fixed number.[1]
• The two stated values for the fixed-boundary revenue-neutral rate (82.9% vs. 89.2%) could not be reconciled from the text alone — see the finding above; a future editor with access to the paper's replication files could resolve which is correct (or whether they answer subtly different questions the prose does not fully distinguish).
• Progressivity is a within-model result, not a real-world distributional measurement. It follows from the assumed endowment structure (Group 1 has 1.5× Group 2's land and capital) and the labor-leisure margin, not from observed Atlanta household data on actual land ownership by income.
• Doucet's Land Is a Big Deal (Ch. 21) cites this paper for its full-capitalization conclusion but not for its progressivity or welfare-magnitude findings, which were not previously on the wiki and are new content from this direct read.

See Also

• Site-value CGE simulations (cluster page) — DiMasi (1987), Haughwout (2001), Follain & Miyake (1986), and the other members of this model family
• Landlords cannot pass a land value tax on to tenants
• Borge & Rattsø (2014), Norway capitalization · Buettner (2003), German land-tax capitalization · Capozza, Green & Hendershott (1996) — the incidence cluster this paper's own headline finding corroborates
• Tax Capitalization · Split-Rate Taxation
• Banzhaf & Lavery, Pennsylvania sprawl — the empirical evidence this paper's urban-area-size result agrees with

Sources

1. Ki-Whan Choi & David L. Sjoquist (2015), "Economic and Spatial Effects of Land Value Taxation in an Urban Area: An Urban Computable General Equilibrium Approach," Land Economics 91(3): 536–555 (August 2015). DOI 10.3368/le.91.3.536 (JSTOR/Project MUSE, paywalled). Free full-text PDF (published-article scan, gwern archive): gwern.net/doc/economics/georgism/2015-choi.pdf — fetched and read in full (2026-07-18) — used for every finding on this page: the benchmark and reform tables (Tables 5–15), the revenue-neutral rate figures and their internal inconsistency, the welfare-gain percentages, the three-income-group progressivity result, the immobile/mobile-capital and absentee-landlord sensitivity results, and the comparison to DiMasi, Sullivan, Follain & Miyake, Song & Zenou, and Banzhaf & Lavery.