Copy of the paper available here
M&S construct a novel measure of the “idiosyncratic land tax” (ILT) based on differences in how assessors and the market are valuing land (where assessors value land higher than the market, there’s an implicit land tax as you pay more taxes for an extra acre of land). This measure appears not to be spatially correlated or correlated with county characteristics so is plausibly exogeneous. Regressing it against county-level changes 2000-2020 they find that ILTs “are associated with population and earnings growth, an increase in the concentration of the population within counties, growth in diversity and an increase in business formation”. This novel ILT measure, if considered legitimate, could open up a raft of new research into the economic effects of (revenue-neutral shifts from property taxes to) LVT.
Background/Motivation: LVTs have great theoretical properties but have rarely been implemented making it difficult to evaluate them. M&S proposed a new ‘idiosyncratic land tax’ (ILT) measure which is based on idiosyncratic differences in the way assessors vs the market are valuing land.
Theoretical Model: Small open economy with endogeneous population consuming Housing services & other Consumption with Cobb-Douglas utility, housing is produced with land & structures with budget constraint: Wage Income + Land Rent Income + Tax Rebate = $I_j+\xi r \bar{L}+\kappa \bar{T} = rL_j(1+\tau_L) + p_SS_j(1+\tau_S)+C_j$ = Spending on Land + Spending on Structures + Consumption. Population in the jurisdiction depends positively on the three sources of income, divided by price levels, and a demand elasticity (wrt income). Model predictions: Proposition 1: LVTs are neutral if all land is locally-owned and tax revenues are confiscated (full capitalization so tax-inclusive cost of land is unchanged). Proposition 2: LVT is super-neutral (increases income and population density) if some land is owned by people outside the jurisdiction and the revenues are spent on public goods or rebated. Structures tax raises living costs, inducing population outflow.
LVT shifts increase population and housing per resident, and these effects are stronger for high elasticity values (eg with less strict zoning..)
Extension models a downtown with inelastic land supply and a suburb with price-elastic supply (I think because land can be converted to residential from farmland): so LVT reduces land prices by more downtown than in the suburb, leading to more housing downtown and higher population-weighted density.
Key Variable: Idiosyncratic Land Tax (ILT): derives from plausibly-exogeneous differences between how assessors and the market value an additional square foot of land (if assessors value it more than the market does, tax bills rise with land square footage faster than if these valuations were identical, so there is an implicit land tax). Measured by running hedonic models on both assessed values and actual sale prices (described in Empirics) and some decomposition which separates 1) the level of property taxes, 2) the idiosyncratic land tax ($ILT_c=\frac{\tau_{s,c}\beta_{1,c}-\tau_{e,c}\delta_{1,c}}{\delta_{1,c}}$, essentially the difference in the statutory tax rate (times assessed LV) and the effective tax rate (times the market LV)), 3) the idiosyncratic structures tax, and 4) idiosyncratic level differences across counties and neighborhoods (#2, #3 and #4 offset each other such that any positive ILT must mean undervaluing of structures or the entry fee (price levels in the county)). ILT is relatively uncorrelated over space (but highly autocorrelated within a county, as expected as they depend on local assessment practices) and with county characteristics in 2000 (land prices, white share, property values, density, population-weighted density, racial diversity, labor income, land use regulations and land supply elasticities), and a section of examples demonstrates that the ILT is independent of county-level land prices, the level of property values, and differences between tax-assessed property values and market property values.
Empirics: Data from ZTRAX and ATTOM with taxes paid, sales price, assessed value, parcel characteristics. Hedonic model with LASSO variable selection and dependent variables of both assessed value and sales price (one model uses transacted properties, another imputes market values of non-sold properties) ($V_{i,c} = \beta$) to estimate ILT (based on difference in coefficients on L in each model). Winsorize 1% and 99% giving ILTs from -3.8% to 4.4%. Regress change in outcomes on ILT with controls and state fixed-effects. Some indications that estimating the ILT may bias the main results downwards.
Results: Jurisdictions with higher ILTs have faster growth in: a) population & density, b) wages, c) diversity of income, race & age, and d) business formation. Robust to state & zip FEs, including the effective property tax, and controls for land use regulation (Wharton)/land supply elasticity (Saiz). Effects are stronger in city centres (denser places) and places with more effective governance. No effect of elected/unelected assessor or use of an algorithm on ILT or population growth.
**Discussion/Conclusions: “**Idiosyncratic land taxes are associated with population and earnings growth, an increase in the concentration of the population within counties, growth in diversity and an increase in business formation”. “A reallocation of the tax burden from structures to land has large economic benefits”. This novel ILT measure, if considered legitimate, could open up a raft of new research into the economic effects of revenue-neutral shifts from property taxes to LVT.
Questions:
Other Reads: