Computable General-Equilibrium Simulations of Site-Value Taxation

What This Page Collects

Most of the wiki's empirical evidence on land taxation is reduced-form: it observes what happened when a real jurisdiction shifted its tax mix (Pittsburgh, Tallinn, Danish municipalities). That evidence is strong on whether an effect occurred but weak on magnitude and general-equilibrium consequences — how a large shift onto land ripples through wages, output, city size, and welfare all at once. A separate strand of the literature fills that gap by building computable general-equilibrium (CGE) models of an urban economy, calibrating them to a real place, and simulating a move toward site-value taxation.

This page is the cluster entry for that strand. Its unifying result is that, across independent models with different structures and calibrations, shifting the tax burden onto land is predicted to lower land rents and house prices, raise wages, shrink the developed footprint of the city, and yield a net welfare gain. Its unifying caveat is equally important: these are simulations, not experiments. Their numbers are only as good as the elasticities and preference parameters fed in, several of which the authors themselves flag as poorly identified. This is model evidence — a distinct and weaker tier than the quasi-experimental and natural-experiment studies elsewhere in the wiki (see the tiering note below).

The Studies

A theoretical companion sits just outside the computational family: Brueckner (1986) derives, analytically rather than by simulation, that a graded (split-rate) tax raises capital intensity and — in the city-wide case — lowers house prices, the same qualitative pattern the CGE models produce numerically. DiMasi (1987) is in effect the computational general-equilibrium extension of Brueckner's analytic result, and the two are routinely cited together.

Reading the Results

The consistent direction. DiMasi's headline — lower land rents and house prices, higher wages, a smaller urban area, and a welfare gain up to 6.6% of revenue — is the reference result the whole cluster orbits. Haughwout's NYC model produces very large gains when the City's high-distortion tax mix (sales, income, property, corporate) is replaced by land taxation: private capital stock rising over 200%, output and land values over 100%, in the most responsive scenario. Choi & Sjoquist's Atlanta model reaches broadly similar qualitative conclusions with a more realistic ownership and labor structure.

The honest complications the models themselves surface. These are not one-way advertisements for LVT:

• Magnitudes hinge on the land share and the elasticity of substitution. Haughwout notes his large numbers sit "near the tops of the ranges reported by Nechyba (1998)," whose comparable US state-local model found capital stock rising between 14% and 122% and output rising 10%–90% depending on the land-capital substitution elasticity. Lower assumed land shares mechanically inflate the estimated benefits. The authors treat these parameters as under-identified and call for more empirical estimation.
• A revenue-neutral pure land tax need not raise land values. Haughwout's Table 4 finds that when the public sector is held to its baseline size, financing it with a 21.7% land tax can depress land values — because a large public sector funded that way "is not worth its cost." Landowners benefit unambiguously only when both the structure of taxation and the level of revenue can change. This mirrors Brueckner's finding that the land-value effect flips sign between the small-zone and city-wide cases, and it cautions against reading "LVT always raises land values" out of these models.
• LVT raises less revenue than a broader capital tax. Follain & Miyake's central negative result — an LVT has "less revenue potential" than a capital value tax and its welfare edge is economy-dependent — is a direct simulation-based statement of the revenue-adequacy objection, and a useful corrective to over-claiming.

On Plassmann & Tideman. The task of assembling this cluster asked whether a general-equilibrium variant of Plassmann & Tideman exists. It does not: their two-rate construction study is a Markov-chain Monte Carlo econometric analysis of Pennsylvania building permits, and their land-assembly and losses-of-nations work uses different methods again. None is a spatial CGE simulation of a site-value shift, so no Plassmann–Tideman row appears above.

Evidence Tier: Model Evidence, Handle With Care

These studies are internally rigorous — they impose full market clearing, budget balance, and mobility, so their predictions are mutually consistent in a way back-of-envelope arguments are not. But a calibrated simulation is only a disciplined statement of "if the economy has this structure and these elasticities, then a land-tax shift does this." It is not observation of a real reform. On the wiki's evidence ladder this ranks below natural experiments and quasi-experimental designs (e.g. the Danish land-tax capitalization study, or Löffler & Siegloch's 5,500-municipality pass-through design) and below strong panel evidence. Where the wiki's benefit and narrative pages cite these simulations, they should be presented as the general-equilibrium complement to the reduced-form record — predicting the direction and rough scale of a large land-tax shift — not as direct measurement of one.

Bears On

• Outcome: LVT improves housing affordability — the CGE models predict lower house prices from a large land-tax shift, the general-equilibrium counterpart to that page's reduced-form price evidence (added there at the model-evidence tier).
• Outcome: Split-rate taxation increases construction — the models' rising capital intensity is the GE version of the split-rate construction result.
• Outcome: Taxing land and rents increases productivity — the simulated welfare gains and higher wages.
• Outcome: LVT can replace capital taxes without efficiency loss — the deadweight-loss and welfare comparisons (with Follain & Miyake as the honest limiting case).
• Narrative: The Housing Crisis Is a Land Crisis — where DiMasi and Haughwout are deployed as the simulated evidence for lower house prices.

See Also

• A Modern Analysis of the Effects of Site Value Taxation (Brueckner 1986) — the analytic companion to these computational models
• Split-rate taxation increases construction
• LVT improves housing affordability
• Objection: LVT can't raise enough revenue — the concern Follain & Miyake give simulation support to
• Split-Rate Taxation · Deadweight Loss · Site Value

Sources

1. Joseph A. DiMasi (1987), "The Effects of Site Value Taxation in an Urban Area: A General Equilibrium Computational Approach," National Tax Journal 40(4), December 1987. DOI (paywalled) — abstract verified verbatim this session: "A long-run spatial general equilibrium model of an urban area is used to examine the effects of a change from a tax system in which land rents and capital values are not differentiated for tax purposes to one in which land rents are much more heavily taxed than are capital values… the welfare gain in making such a policy change can be as high as 6.6 percent of tax revenues raised… Structural changes included lower land rents and housing prices, higher wages, and a redu[ced size of the urban area]." — used as the anchor of this cluster. Existing sources/registry.csv row points at the housing-crisis narrative and should be repointed here.
2. James R. Follain & Tamar Emi Miyake (1986), "Land Versus Capital Value Taxation: A General Equilibrium Analysis," National Tax Journal 39 (1986). DOI — abstract verified verbatim this session: a CGE model of "a reduction of the Jamaican income tax in favor of either a land value tax or a capital value tax," concluding "(1) There is less revenue potential from an LVT than from a CVT; (2) the welfare gains from the LVT are sensitive to the type of economy; and, (3) the declines in deadweight loss that result from a switch to a CVT are modest and it is possible that deadweight loss increases." — used for the LVT-vs-CVT revenue and welfare comparison. A new source for this wiki, not yet in sources/registry.csv.
3. Andrew F. Haughwout (2001), "Land Taxation in New York City: A General Equilibrium Analysis," paper prepared for the conference in honor of Dick Netzer, Federal Reserve Bank of New York. PDF — full text fetched and read this session — used for the calibrated NYC CGE; the fixed-rate replacement results (private capital stock >200%, output and land values >100%, population >80% in the full-congestion scenario); the revenue-neutral required land tax rate of 21.7% and the finding that land values can fall under a revenue-constrained pure land tax; and the comparison to Nechyba (1998) (capital stock +14% to +122%, output +10% to +90%). Existing sources/registry.csv row points at the housing-crisis narrative and should be repointed here.
4. Richard J. Arnott & James G. MacKinnon (1977), "The Effects of the Property Tax: A General Equilibrium Simulation," Journal of Urban Economics 4(4): 389–407 (also Queen's University Working Paper No. 205). RePEc record — abstract verified this session ("a general equilibrium model of residential land use to study the effects of the property tax in a closed city… solved… numerically"). — used as the methodological precursor to the site-value simulations. A new source for this wiki, not yet in sources/registry.csv.
5. 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. DOI — abstract verified this session: an Atlanta-calibrated urban CGE of switching from a capital property tax to an LVT, distinguished from "prior simulation studies" by resident (not absentee) landowners, three income groups, imperfectly mobile housing capital, and a labor-leisure choice — used for the modern Atlanta extension row above. Doucet's Land Is a Big Deal (Ch. 21) lists this paper among those supporting full capitalization of LVT into land prices — the landlords cannot pass LVT to tenants claim carries it via this cluster as model evidence.
6. Companion analytic model: Brueckner (1986) — wiki summary — the theoretical (non-computational) counterpart, already in the corpus.