Split-rate taxation increases urban construction

At a glance — Taxing land more heavily than buildings tends to increase construction relative to comparable jurisdictions — consistent across Pennsylvania studies, the Pittsburgh natural experiment, and a Finnish natural experiment — though a New Zealand null and a timing critique keep it short of decisive. Evidence: Moderate–strong (three peer-reviewed Pennsylvania studies, the Pittsburgh natural experiment, and a peer-reviewed Finnish natural experiment agree on direction; a New Zealand null, self-selected Australian data, and a timing-neutrality caveat keep it short of Strong) · 17 supporting sources · 2 challenging Strongest support: Oates & Schwab (1997) — Pittsburgh's building-permit value rose roughly 70% in the 1980s after it taxed land about 5× buildings, near the top of a 15-city comparison despite a shrinking economy. Strongest counter: Gemmell, Grimes & Skidmore (2019) — an Auckland rating-base shift found no significant effect on new building development.

The Claim

When a jurisdiction taxes land at a higher rate than buildings (a "split-rate" or "two-rate" property tax), construction tends to increase relative to comparable jurisdictions. Taxing improvements less removes a penalty on building; taxing land more makes holding a site idle costly.

The 30-second version. Three of the strongest studies point the same way. In a tract-level panel of Pennsylvania's split-rate adopters, Banzhaf & Lavery (2010) find adoption raises the capital-to-land ratio, and that the effect shows up as more housing units (about 4–5 percentage points per decade), not merely bigger ones — the density margin. Across 15 Pennsylvania municipalities, Plassmann & Tideman (2000) find a higher land-to-building tax ratio produces a statistically significant increase in the number and value of building permits. And in the flagship natural experiment, Oates & Schwab (1997) find Pittsburgh's building-permit value rose roughly 70% in the 1980s after it taxed land ~5× buildings — near the top of a 15-city comparison group — even as the city was losing population and steel jobs. Honest limit: the rigorous evidence is almost all from Pennsylvania (the main US jurisdiction permitting split rates), one short foreign panel found no effect, and a theoretical "timing" critique warns some of the extra building may be over-hastened rather than pure efficiency gain (see Counter-Evidence).

A historical prologue sits behind the modern studies: New York City exempted new residential buildings (but not land) from local taxation from 1921 to 1931 — a split-rate-like episode at metropolitan scale. The wiki's entry on the episode carries the contemporaneous Bureau of Labor Statistics assessment (construction volume visibly boosted; rents and the shortage not obviously solved) and the modern skeptical reading (the 1920s boom was national; zoning capacity and the absence of rent control did much of the work) — suggestive rather than causal evidence, but the largest untax-buildings experiment on record.

That prologue is wider than New York alone. Gaffney's "New Life in Old Cities" (2006) surveys eight further US and Canadian cities — Cleveland, Toledo, Detroit, Milwaukee, Chicago, San Francisco, Houston, and Vancouver, 1890–1930 — where Georgist-aligned leaders shifted assessment toward land and away from buildings, using population growth (not building permits, which he notes are unavailable in uniform form for the period) as his outcome measure. The pattern is consistent with, but not a substitute for, the modern construction-permit evidence above: it is historical correlation across nine cities and three countries' worth of comparison groups, argued by a committed Georgist advocate rather than isolated with a matched research design. Gaffney's Georgist City-Growth Case Studies supplies the fuller, unabridged workpapers behind the Cleveland and Chicago capsules in that same 2006 survey — including Gaffney's own explicit downgrade of the Chicago case's evidentiary strength — but, like New Life in Old Cities, is not itself added as supporting evidence here (see that page's Standing and Limits).

A separate, earlier Gaffney data point on the same city: Gaffney's own 1960s land-value study of Milwaukee County — an "isovalic" contour map built from several thousand actual land sales, comparing bare ("challenger") land values to combined land-plus-old-building ("defender") values — found that in roughly a quarter of the city the bare renewal value already nearly equaled the defender value, and estimated that a small further rise in challenger over defender values would trigger renewal of 20% of the central-city area, with full exemption of buildings from the property tax triggering renewal of about 50%. The study's primary publication is his 1969 AIP Journal article "Land Planning and the Property Tax", written for an audience of professional planners; his 1998 book chapter "The Philosophy of Public Finance" restates the same study and figures twenty-nine years later. This is a different exercise from the 1916–40 political history above: a single-researcher microdata study, not independently replicated or published with the underlying dataset, so it is Gaffney's own modeled projection rather than a controlled test. The 1969 article also supplies a supplementary theoretical account of why the building tax delays renewal: new construction is typically debt-financed, so a tax proportional to total property value (rather than just the debt portion) doubly leverages a builder's cash squeeze relative to an equivalent rise in mortgage rates.

A fourth Gaffney data point sits outside the urban-construction literature entirely but tests the same mechanism in a different sector. Rising Inequality and Falling Property Tax Rates (1992) is a 50-state, 1900–1987 cross-sectional and time-series study of US farmland: after average farm property tax rates fell 40% between 1930 and 1987, Gaffney finds states with higher property tax rates consistently show more building investment per dollar of land value, more intensive land use, and smaller, more equally distributed farms — the Wisconsin-vs-Florida comparison shows the high-tax state leading 5-to-1 in farm buildings per dollar of farmland value. This is observational cross-sectional evidence, not a natural experiment, and Gaffney himself leaves the causal direction formally open, but it extends the land-tax-raises-the-capital-to-land-ratio finding beyond urban permits into agriculture and over a much longer time horizon than any other source on this page.

The Evidence in Numbers

The findings are mutually reinforcing: the Pittsburgh case shows a large effect in one aggressive adopter; the Pennsylvania panels show the effect holds across many municipalities, appears specifically as more dwelling units, and shows up in the tax base itself.

International corroboration — Victoria, Australia (Harrison 1983): Harrison (1983) presents Australian state-level data providing independent corroboration from a different country and institutional setting. In Victoria (1966–78), cities taxing only site value (SVR) saw dwelling growth of +12.9%, while cities taxing land plus buildings (NAV) saw only +2.8% (Harrison 1983, Ch. 15). Building permits in 1975–78 were 39.9% of the 1966–69 level in SVR cities versus only 9.5% in NAV cities — a more than fourfold difference (Harrison 1983, Ch. 15). When Caulfield switched from SVR to composite rating in 1969–70, building permits dropped 66%, compared to a 16% drop in SVR cities (Harrison 1983, Ch. 15). The Australian evidence is observational rather than experimental, but it predates and independently corroborates the Pennsylvania findings from a different continent and tax system. A methodologically independent modern study strengthens the Victorian case: Murray & Hermans (2019), using historical Victorian council data in a difference-in-differences design, find that municipalities switching from capital-improved-value rating (land plus buildings) to site-value rating saw a 20% increase in the value of new residential construction — the same direction as Harrison's raw comparison, from a non-movement source with a causal design. The full case, with its self-selection caveats, is at Victoria's SVR vs NAV rating comparison.

Peer-reviewed corroboration from Finland (Lyytikäinen 2009): the strongest non-US, non-Australian test comes from a Finnish natural experiment. After 2001, municipalities could tax undeveloped residential land at a higher rate than developed land; using a fixed-effects Poisson panel of 1998–2006 data, Lyytikäinen (2009), in the Journal of Urban Economics, finds adopting this "three-rate" system raised single-family housing starts by roughly 10% a year — the land-favoring-tax prediction confirmed on a fresh continent and in the same top field journal as the Pennsylvania panels. (It carries its own timing nuance: the data hint that development density may have fallen somewhat, attenuating part of the gain — the Finnish echo of the Bentick–Mills caveat below.)

The Evidence — Paper by Paper

Each supporting page carries its own external citations and honest caveats:

• Oates & Schwab (1997) — the most-cited empirical LVT study in a US city and the flagship natural experiment: Pittsburgh's 1979–80 shift to heavy land taxation was followed by a ~70% building-permit boom that comparable rust-belt cities did not see. Its own caution — a concurrent downtown redevelopment push, so the tax was a "meaningful, plausibly significant" contributor rather than the sole cause — is carried in Counter-Evidence.
• Plassmann & Tideman (2000) — a Markov Chain Monte Carlo analysis of 15 Pennsylvania municipalities that turns the single Pittsburgh case into a multi-jurisdiction panel, finding a significant positive effect of the land-to-building tax ratio on both the number and value of building permits.
• Banzhaf & Lavery (2010) — a tract-level difference-in-difference-in-differences panel that swaps building permits for direct Census measures of housing units and dwelling size, and decomposes the effect: the extra construction is more units, not bigger units (concentrated in structures of five-plus dwellings), the strongest density-specific evidence in the literature.
• Yang & Hawley (2022) — approaches the question from the tax base: aggregate market values rise after adoption, corroborating the construction story without relying on permit data.
• Brueckner (1986) — the foundational modern theoretical model of graded (split-rate) taxation: shifting a property tax off improvements and onto land raises the capital-to-land ratio, with an ambiguous-to-positive effect on land value itself. This is the mechanism the empirical studies test.
• Harrison, The Power in the Land (1983) — the Australian SVR-vs-NAV dwelling-growth and building-permit data above (a C-claim: empirical data reported in an advocacy book, not peer-reviewed; corroborating, not load-bearing).
• Bourassa (1987) — the earliest econometric Pittsburgh study (monthly data 1978–84), finding a significant new-housing response; its own caveat is that the effect is carried by the improvement-tax cut rather than the land-tax rise.
• Edwards (1984) — across six Australian states (1951/52–1974/75), jurisdictions taxing land more and improvements less show significantly higher new-house values and a larger housing stock.
• Tomson (2016) — Tallinn, under Estonia's land-only tax since 1993, shows higher inner-city density and more construction than comparable Riga — Baltic external validity.
• Victoria SVR-vs-NAV comparison — the municipal-level Victorian rating-system comparison behind Harrison's Australian numbers, held to its own honest limits.
• Cho et al. (2013) — a revenue-neutral two-rate simulation for Nashville predicting residential density rising 18–83% (model evidence).
• CGE site-value simulations — the general-equilibrium family (DiMasi, Haughwout) predicting higher capital-to-land ratios and more construction under land-shifted taxation (model evidence, weaker than the quasi-experimental tier).
• Murphy & Seegert — implicit land taxes created by assessment practice have measurable real effects on construction, corroborating the mechanism from an oblique angle.
• Gaffney (1992): Rising Inequality and Falling Property Tax Rates — a 50-state, 1900–1987 cross-sectional study of US farmland finding higher property tax rates consistently associated with more building investment per dollar of land value and more equal farm-size distribution (observational; direction of causation not statistically resolved by the paper itself).

A Related but Distinct Result — the Conventional Property Tax and Sprawl

A neighbouring peer-reviewed finding is easy to conflate with this outcome but must be kept separate. Song & Zenou (2006), in the Journal of Urban Economics, show — theoretically and in a year-2000 IV analysis of US urbanized areas — that cities with higher conventional property tax rates have smaller urbanized footprints. That is anti-sprawl evidence, but for the ordinary property tax and through a partly opposite mechanism: the conventional tax restrains sprawl in part by discouraging housing capital generally, whereas split-rate/LVT design encourages structures on valuable land while penalizing idle holding. Banzhaf & Lavery explicitly contrast their density result with Song & Zenou's compactness result and offer a partial reconciliation (differing tax rates across a metro can reshape the density gradient without changing measured urbanized-area compactness) but do not resolve it empirically. The two results point in a broadly similar direction — more tax on land-related holdings, less sprawl — but the wiki cites them side by side, never interchangeably, and Song & Zenou is therefore not listed as direct support for the construction claim.

Counter-Evidence and Limits

The direction of the effect is consistent, but the honest case has real qualifications.

1. The Bentick–Mills timing critique (theory). Bentick (1979) and Mills (1981) show that a land tax assessed on current market value (which capitalizes future development options) is not neutral: it can bias owners toward quick-yield projects and hasten development timing at real resource cost. This cuts two ways for this page. It does not undercut the empirical finding — a timing bias toward earlier/denser building actually reinforces the observed construction and density increase. But it complicates the welfare reading: part of the extra construction may be over-hastened rather than a pure efficiency gain. Tideman's 1982 rebuttal locates the flaw in the assessment basis, not the land tax as such — a tax on pre-development site value restores neutrality — so the practical exposure depends on how a jurisdiction assesses land. (Full treatment: The Bentick–Mills timing critique.)

2. Pittsburgh-specific confounds. The flagship case has two. First, Oates & Schwab themselves note Pittsburgh pursued an aggressive downtown public–private redevelopment push in the 1980s, so they claim only a "meaningful, plausibly significant role" for the tax, not sole causation. Second — and cutting the other way, in the tax's favour — Pittsburgh's 1980s were the years of the regional steel collapse: the city was losing population and industrial jobs. A general demand surge therefore cannot explain the building boom, which strengthens rather than weakens the causal story. The multi-municipality panels (Plassmann & Tideman, Banzhaf & Lavery) exist precisely to move beyond this single, confounded case. Gaffney's New Life in Old Cities adds a further, distinct caveat: Pittsburgh's building-permit success under the graded tax did not translate into population growth — its population fell sharply after 1980 — which he attributes to the plan applying only to the city's own levy (not the overlapping county or school-district tax), a 1980s wage tax, and county-level assessment practice that left land at only a reported 10% of the property tax base by 2000. This is a caveat about a different outcome (population, not construction) and does not contradict the construction findings above, but it is a reminder that a partial or poorly-assessed graded tax can decouple the two.

3. "Density in the wrong place" is not necessarily less sprawl. Banzhaf & Lavery are careful: they "do not claim that our empirical work shows that the split-rate tax … has resulted in less sprawl." Their result is that split-rate taxation raises density where it is applied; if adopted by a fringe or exurban jurisdiction, that added density could count as more metro-area sprawl. The construction claim is well-supported; the leap to "reduces sprawl" is not automatic (see LVT reduces sprawl).

4. A foreign null result. Gemmell, Grimes & Skidmore (2019) studied Auckland's forced move away from land-value rating after the 2010 council amalgamation and found little effect on new building consents (only on alterations) over a short two-year window. It is a confounded reform and a short panel, so it does not overturn the Pennsylvania results — but it is a reminder that the construction effect is not automatic in every institutional setting.

5. External validity. The rigorous US evidence is overwhelmingly from Pennsylvania, the principal US jurisdiction permitting split rates. But the base is broader than it once was: the Australian evidence is now backed by the Murray & Hermans (2019) difference-in-differences preprint (though Victorian municipalities self-selected their rating base), and — most important — Lyytikäinen (2009) supplies a peer-reviewed, quasi-experimental Finnish result (~10% more single-family starts) in the same journal as the Pennsylvania panels. The New Zealand null still points the other way over a short window, so generalization to every property-tax institution remains only partly established — but "true in one US state" now understates the evidence.

Strength of Evidence

Moderate to strong. The direction is consistent across three peer-reviewed Pennsylvania studies, the most-cited US natural experiment (Pittsburgh), independent Australian data, and a peer-reviewed Finnish natural experiment (Lyytikäinen 2009), and it matches theory (zero deadweight loss on land plus a removed penalty on building). What holds it short of "Strong": the US base is still concentrated in Pennsylvania's ~18 adopters, one foreign panel (New Zealand) found no effect, the flagship case carries an acknowledged redevelopment confound, and the Bentick–Mills timing critique — echoed in Lyytikäinen's own reduced-density finding — means some of the measured extra building may reflect over-hastened rather than efficient development. As a land-core benefit claim, the construction effect can be stated with confidence; the further step to "and therefore reduces sprawl / raises welfare" should stay attributed and qualified.

See Also

• Victoria's Site-Value vs Net-Annual-Value Rating — the 1966–78 Australian natural experiment
• Bourassa (1987): Land Value Taxation and New Housing Development in Pittsburgh — the earliest econometric study finding Pittsburgh's split-rate tax significantly increased new-housing construction
• The Houston Plan of Taxation (1912–1915) — an early, legally-terminated American split-rate experiment predating Pittsburgh, showing a construction surge under land-favoring assessment
• Land Value Tax · Harrisburg, Pennsylvania · Split-rate taxation
• The Bentick–Mills timing critique — the neutrality caveat on the welfare reading
• Song & Zenou (2006) — the conventional-property-tax-and-sprawl result, kept distinct
• Lyytikäinen (2009) — the peer-reviewed Finnish three-rate natural experiment (~10% more housing starts)
• Gemmell, Grimes & Skidmore (2019) — the New Zealand null result
• Gaffney (2006): New Life in Old Cities — the wider nine-city historical prologue, and the Pittsburgh population-vs-construction caveat
• Gaffney (1998): The Philosophy of Public Finance — Gaffney's own 1960s Milwaukee County land-value microdata study
• Gaffney (1969): Land Planning and the Property Tax — the primary source of the Milwaukee study above, plus the building-tax credit-leverage financing mechanism
• Gaffney (1992): Rising Inequality and Falling Property Tax Rates — the agricultural cross-sectional data point, outside the urban-construction literature
• Harrison, The Power in the Land — the Australian SVR vs NAV evidence
• Victoria's SVR vs NAV rating comparison (1966–78) — the Australian natural-experiment case page, including the independent Murray & Hermans (2019) corroboration
• LVT Dampens Land Speculation — the companion outcome page using the same Australian data
• LVT reduces urban sprawl — the adjacent claim the density evidence bears on

Sources

1. Oates & Schwab (1997), "The Impact of Urban Land Taxation: The Pittsburgh Experience," National Tax Journal 50(1) — wiki summary · PDF. Used for the Pittsburgh ~70% building-permit rise, the 15-city comparison, the shrinking-economy point, and the authors' own redevelopment-confound caution.
2. Plassmann & Tideman (2000), "A Markov Chain Monte Carlo Analysis of the Effect of Two-Rate Property Taxes on Construction," Journal of Urban Economics 47(2) — wiki summary · publisher. Used for the statistically significant permit effect across 15 Pennsylvania municipalities.
3. Banzhaf & Lavery (2010), "Can the Land Tax Help Curb Urban Sprawl? Evidence from Growth Patterns in Pennsylvania," Journal of Urban Economics 67(2) — wiki summary · DOI. Used for the capital/land-ratio and housing-unit density findings, the multi-unit-structure concentration, and the authors' own "do not claim … less sprawl" and density-in-the-wrong-place caveats.
4. Yang & Hawley (2022), "Effects of Split-Rate Taxation on Tax Base," — wiki summary · publisher. Used for the tax-base corroboration (aggregate market values rise after adoption).
5. Brueckner (1986), "A Modern Analysis of the Effects of Site Value Taxation," — wiki summary. Used for the theoretical capital/land-ratio mechanism the empirical studies test.
6. Song & Zenou (2006), "Property Tax and Urban Sprawl: Theory and Implications for US Cities," Journal of Urban Economics 60(3) — wiki summary. Used for the distinct conventional-property-tax compactness result (cited as a contrast, not as direct support).
7. Bentick (1979) / Mills (1981) / Tideman (1982) — the development-timing neutrality debate — wiki summary. Used for the timing non-neutrality caveat on the welfare interpretation.
8. Fred Harrison (1983), The Power in the Land, Shepheard-Walwyn. Publisher · wiki summary — used for the Victoria SVR vs NAV dwelling growth and building-permit data (C-claim: empirical data reported in an advocacy book, not peer-reviewed; corroborating, not load-bearing).
9. Gemmell, Grimes & Skidmore (2019) — Auckland amalgamation and building consents — wiki summary. Used for the short-window foreign null result.
10. Cameron K. Murray & Jesse Benjamin Hermans (2019), "Land value is a progressive and efficient property tax base: Evidence from Victoria," OSF Preprints — free PDF · IDEAS/RePEc record. Used for the independent difference-in-differences finding that switching from capital-improved-value to site-value rating is associated with a 20% increase in the value of new residential construction (working paper, not yet peer-reviewed).
11. Teemu Lyytikäinen (2009), "Three-rate property taxation and housing construction," Journal of Urban Economics 65(3): 305–313 — wiki summary · free VATT DP 419. Used for the peer-reviewed Finnish natural-experiment finding that adopting a three-rate (higher tax on undeveloped land) system raised single-family housing starts ~10% a year, with a reduced-density attenuation caveat.
12. Mason Gaffney (2006), "New Life in Old Cities," masongaffney.org — wiki summary · PDF. Used for the wider nine-city historical prologue and the Pittsburgh population-vs-construction anomaly (advocacy-authored historical synthesis, not peer-reviewed; corroborating, not load-bearing).
13. Mason Gaffney (1998), "The Philosophy of Public Finance," Ch. 7 in Fred Harrison (ed.), The Losses of Nations (Othila Press, 1998) — wiki summary. Used for the 1960s Milwaukee County "isovalic" land-value study (advocacy-authored, single-researcher microdata exercise, not independently replicated; corroborating, not load-bearing; this chapter restates the primary 1969 source below).
14. Mason Gaffney (1969), "Land Planning and the Property Tax," AIP Journal, May 1969, pp. 178–183 — wiki summary · PDF. The primary source of the Milwaukee isovalic study cited above (item 13 restates it); also used for the building-tax credit-leverage financing mechanism (advocacy-authored planning-journal article, not peer-reviewed; corroborating, not load-bearing).
15. Mason Gaffney (1992), "Rising Inequality and Falling Property Tax Rates," Chapter 10 in Gene Wunderlich (ed.), Ownership, Tenure, and Taxation of Agricultural Land (Westview Press) — wiki summary · PDF. Used for the US farmland cross-sectional and time-series data (observational, book chapter in an edited academic volume; corroborating, not load-bearing).