Fiscal Illusion from Property Reassessment? An
Empirical Test of the Residual View
Justin M. Ross
Assistant Professor
School of Public & Environmental Affairs
Affiliated Faculty
Workshop in Political Theory and Policy Analysis
Indiana University
Bloomington, IN 47405
(812) 856-7559
justross@indiana.edu
Wenli Yan
Assistant Professor
School of Public & Environmental Affairs
Indiana University
Bloomington, IN 47405
(812) 855-1978
mushbb@gmail.com
Abstract
The textbook view of the real property tax is that the tax rate is simply a residual, determined by dividing
the revenue to be raised from the property tax by the taxable base of assessed values. As such, tax rate
adjustment is automatic and changes in assessed values are neutral with respect to property tax revenue.
Critics often claim that policymakers do not fully reduce the tax rate in response to increasing assessed
values, allowing them to effectively raise revenue without raising tax rates. Using data from Virginia
cities and counties between 2000 and 2008, this paper estimates the response of levy growth rates to
property reassessment by taking advantage of a policy environment that offers a natural experiment in the
execution of mass reappraisals. A timing lag in the implementation of updated property assessments to be
used in setting the tax rate further allows the study to disentangle assessed value growth from other forms
of economic growth, including the market values of property. Our findings provide partial support for the
fiscal illusion critique, as mass reappraisals per se provide cover for politicians to raise levy growth rates;
however, overall the growth in aggregate taxable assessed values is largely irrelevant.
Keywords: Property assessment, property tax rates, fiscal illusion, median voter theory
JEL Classification: H71, H73, H83
* We appreciate the assistance of graduate assistant Larry Summers for assistance in collecting data from
Virginia county governments. We also appreciate helpful comments and feedback from John Mikesell,
Bradley Heim, Denvil Duncan, Ed Gerrish, Ryan Woolsey, George Zodrow, Bill Gentry, and two
anonymous referees. Any errors are solely those of the authors.
1.
INTRODUCTION
The administration of the real property tax makes it unlike any other ad valorem tax in the United States.
Since most individual property parcels are seldom traded, an assessment process is required to estimate
their fair market value.1 The assessment of the property can be of significant importance to individual
property owners for determining their tax liability, but in the aggregate the process is considered by
budget analysts to be irrelevant to the revenues raised (the “residual view”).2 In the United States, local
governments formally determine the property tax levy, which is the aggregate amount of revenue to be
raised from the taxation of property to meet the predetermined aggregate demands of public expenditures
(Ladd, 1991; Mikesell, 2011). The millage rate is then a matter of arithmetic as it is the result of dividing
the levy by the total assessed values in the jurisdiction. If all properties experienced a uniform 10%
increase in their assessed values, then the millage rate would be revised down so that the tax bill would be
unchanged across all taxpayers in meeting the original levy. The formal role of the property reassessment
is to maintain equity by keeping assessed values in line with their true market value. In fact, this is
exactly the way in which textbooks in public financial management and budgeting explain the setting of
the property tax rate (Lee et al., 2008; Mikesell, 2011). Still, the taxpayer’s individual property tax bill is
determined by applying the millage rate to their individual assessment, hence the widespread perception
that local governments set rates instead of levies. As explained in Fisher’s textbook on state and local
public finance (2007, p. 323):
In other words, a general rise in property values allows local governments to increase
property tax collections without increasing tax rates. Not surprisingly, some individuals
are led to conclude that the assessment increase caused the tax increase. This view is not
correct because each local government with property tax authority controls and selects,
either explicitly or implicitly, the amount of property tax revenue to levy.
“Real property” refers to land and its accompanying capital improvements. This is distinguished in tax law from
personal property, which can include vehicles or financial assets.
2
The term “residual view” is adopted here from Netzer (1964, p. 207): “…the nominal rate of the tax is essentially a
residual, derived by determining the level of expenditures, subtracting state aid and other non-property tax revenues
from budget outlays, and comparing the remainder with assessed values…[A]ssessed values are no more ‘actual’ as
determinants of property tax yield than are a number of other factors.”
1
1
The popular criticism of the residual view, frequently raised by legislators, journalists, and citizens is that
local government officials take advantage of an increase in the property assessments by only partially
offsetting the millage rate, so as to increase total real property tax revenue. Since property owners have
their individual liability partly determined by the millage rate, local government officials can increase the
levy, and raise more revenue while lowering the tax rate and thus claiming they “cut taxes.” In doing so,
these critics are arguing that the property tax reassessment process is a source of fiscal illusion that causes
the levy to inflate from what would be preferred by the voters.
Fiscal illusion has been a subject of substantial interest to public finance scholars, and it is
thought to exist in a variety of forms (Dollery and Worthington, 1996). Some forms of fiscal illusion
relate to the structure of the tax code, where indirect, hidden, or less salient taxes cause voters to believe
that the cost of public services is less than what they perceive it to be. Generally, the property tax is
thought of as a fairly salient and visible tax (Cabral and Hoxby, 2010). Renter illusion, however, is
thought to arise if renters do not expect these taxes to be shifted into higher rents from the landlord and
consequently vote for higher levels of public expenditures (Bergstrom and Goodman, 1973; BlomHansen, 2005). Another form of fiscal illusion, popularly known as the flypaper effect (Oates, 1979;
Dougan and Kenyon, 1988), occurs when categorical lump-sum grants are used to expand the public
budget by an amount greater than an equivalent increase in voter income from other sources, typically
thought to occur when politicians use the grant to feign a lower marginal cost of taxation.
The fiscal illusion most related to property reassessment comes from the revenue-elasticity
hypothesis that increases in any tax base are automatically directed into new expenditures. As Oates
(1975) observed, evidence for the revenue elasticity hypothesis is usually confounded by the fact that
revising income or sales tax rates is costly, so an absence of rate revision in response to tax base growth is
not necessarily a dismissal of voter preferences. This cannot be the case for property taxes, however,
because local governments annually determining their expenditures are, by definition, nominally setting
the tax rate, so there are no additional costs from rate adjustment.
2
This revenue-elasticity fiscal illusion critique of the residual view seems to have been partially
influential to the passage of Proposition 13 in California (Mikesell, 1980; Martin, 2008), as well as the
recent property tax revolt in Indiana when changes in assessment standards caused the taxable value of
certain properties to rise significantly.3 Contrary to the residual view, national comparisons of growth in
housing prices to property tax revenues have historically been highly correlated, as demonstrated in
Figure 1. The popular Mikesell (2011) textbook on public financial administration pays homage to this
critique in describing the administration of the property tax rate [p.494]:
A government may, of course, see the computed rate, worry about the consequences, and
revise the amount of levy it chooses to raise.
Empirical evidence to support or refute the residual view, however, has been sparse. This is in part
because policy makers may make “legitimate” changes in the levy to accommodate new growth in
demand for public services, which often occur with growth in assessed values. Areas which experience
the largest increases in property assessments might simply be high growth areas, or voters may demand
more public services due to the wealth effects accompanying appreciation in the value of their homes.
The purpose of this paper is to provide a direct test of the residual view, a view that would be
rejected if it can be shown that assessed value growth causes increases in the property tax levy. This is
done by drawing upon data from Virginia cities and counties between 2000 and 2008, where an
institutional arrangement reasonably approximates a natural experiment for testing the residual view. In
1984, the state mandated that these areas adopt a fixed reassessment frequency cycle that determines the
number of years between reassessments, which has the consequence of causing the timing of
contemporary assessments to mimic a random assignment process by making them independent of the
economic conditions of the local governments. Data from annual assessment-to-sale price ratio studies
allow us to further differentiate between changes in fair market value and assessed values, which differ
3
Mikesell (1980) describes the popular view that increases in assessments caused higher spending from activistpoliticians like Howard Jarvis which lead to their successful push for California’s proposition 13, passed in 1978.
More recently, in 2010 Indiana voters passed a constitutional amendment, which is widely seen as a consequence of
voter unrest resulting from a change in assessment on the basis of cost of replacement to current fair market values
(see Faulk, 2004).
3
because of assessment errors and time lags between mass reappraisals. The identification approach
allows us to separate the accumulated demand effects related to economic growth from the potential role
of reassessment as a source of fiscal illusion. Our findings are able to detect some amount of fiscal
illusion, where increases in assessed values and mass reappraisals cause higher levies.
This paper is structured as follows: in the next section, we provide a summary of the previous
literature, followed by a section providing the necessary background to understand Virginia as a natural
experiment for testing the residual view. The fourth section discusses the model and research
methodology, and results are presented in the fifth section of the paper. The paper concludes with a
summary of the findings and a discussion of policy implications.
2.
PREVIOUS LITERATURE
To the best of our knowledge, there are just four papers – Mikesell (1978, 1980), Bloom and Ladd (1982),
and Ladd (1991) – whose purpose is to test the residual view. The general difficulty with the existing
literature on the residual view is that assessment growth could be correlated with broader economic
growth, particularly if there is a marginal propensity to consume public services out of property
appreciation that is distinct from other forms of economic growth.
Bloom and Ladd (1982) studied changes in property tax levy growth in 241 Massachusetts towns
and cities between 1960 and 1978. After controlling for area and year fixed effects, the results suggested
that the enactments of mass reappraisals were accompanied by statistically significant 3.5 percent increase
in the levy growth rate for cities, but no effect was found in towns. However, the regressions had no
other control variables, raising the possibility of omitted variable bias, and the mass appraisals were
performed at the discretion of the municipalities, perhaps resulting in an endogeneity bias. Ladd (1991)
replicated the specification from Bloom and Ladd (1982), this time using data from North Carolina over
1960-1984, where she argued the state mandates of reappraisal frequency set in earlier years allowed for a
more exogenous reappraisal process than was observed in the earlier study of Massachusetts. The
4
findings from North Carolina in Ladd (1991) indicated that mass reappraisals were correlated with
increased property tax levy growth of 3 to 8 percent. However, while there were no other annual control
variables, just year and area fixed effects, the results were reestimated after splitting the sample by high
versus low income and population growth. The evidence from these specifications becomes somewhat
more mixed, suggesting these could be potentially important omitted variables. In fact, both Bloom and
Ladd (1982) and Ladd (1991) framed their model as a median voter public service demand function, in
which case population and income would play a strong theoretical role.
The earliest test of the residual view came from Mikesell (1978), who estimated a time-series
regression of total property tax levies in the state of Indiana between 1950 and 1972 on state personal
income, household personal property, and a dummy variable indicating the presence of a mass
reappraisal. The mass reappraisal dummy was associated with a 0.01 percent increase in property tax
levies, a magnitude which is neither substantively nor statistically significant. While this finding is
consistent with rate adjustment, it must be qualified by the observation that it was local governments
rather than the state who determined levies and assessments, and so the state-level time-series approach
suffered from aggregation bias.
In a 1980 paper, Mikesell again studied the residual view, this time with counties and
independent cities in Virginia from 1973 to 1975. Mikesell (1980) finds reassessments to have had an
effect on the nominal rate consistent with rate revisions that offset changes in assessed values, but not on
the effective tax rate. 4 This provided some evidence that Virginia counties were not completely offsetting
the change in property assessed values with tax rate adjustments, but measurement of base growth using
the assessment-to-sales price ratio did not allow for a clear inference on how large or small the effect
could have been.
4
This is a true limitation in drawing inferences on the residual view for any paper with property tax rate as the
dependent variable. For example, Cornia and Walters (2006) present a regression explaining effective property tax
rates in five Utah MSAs over 20 years with changes in assessed values, though their interest is in the impact of full
disclosure policies rather than the residual view. While they find a negative relationship, consistent with rate
adjustments, it is not possible to determine if the rate revisions were sufficient to remain levy-neutral.
5
Beyond papers directly interested in the residual view, related research on the relationship
between housing prices and property tax revenues can be informative in terms of testing the hypothesis.
One of the more recent examples is presented in Lutz (2008), the main result of which estimates property
tax revenue growth among local governments located in Metropolitan Statistical Areas between 1985 and
2005 as a function of multiple lags in the Federal Housing Finance Agency’s (FHFA) housing price
index.5 The regression results reveal that the property tax revenue elasticity with respect housing prices is
greater than zero, but less than one, causing the author to conclude that property tax rates must partially
decline in response to changing housing prices. Interestingly, Lutz (2008, p. 566) also suggests that this
result is partially explained as an institutional consequence of unchanging property assessments:
Institutional features of the property tax, such as delays in bringing assessed values into
line with market values and caps and limitations on the tax, likely explain the lag between
house price and tax revenues (and may also influence the magnitude of the relationship).
In essence, Lutz’s proposed interpretation seems to be one where increases in assessed values do cause
property levy increases and rate changes are caused by voter demand, which would be somewhere
between the residual view taught in textbooks and the opposing view of its critics. The paper provides a
useful point estimate on the general elasticity of housing prices to property tax increases, but it also
highlights why research explaining property tax levies with only fair market value growth do not provide
direct evidence on the residual view.6 Other papers (e.g. Alm et al., 2011; Doerner & Ihlanfeldt, 2011;
Lutz et al. 2011) which investigate property tax revenues and house prices have this same qualification
when it comes to implications for the residual view: without assessed values being part of the regression,
it is not possible to differentiate between voter demand shocks and policy makers taking advantage of
fiscal illusion in revising the rate.
5
At the time of the Lutz (2008) paper, the housing price index was published by the Office of Federal Housing
Enterprise Oversight.
6
Furthermore, since the housing price index is quality-adjusted it need not be the case that the assessed values
constituting the property base changed in the same manner. Increases in the taxable property base for any reason
should permit the form of fiscal illusion required to reject the residual view.
6
In summary, the previous literature directly interested in the residual view has produced mixed
results, and may suffer from omitted variable bias. Specifically, what remains to be contributed is an
approach where change in assessed values is not contaminated by other growth variables. The next
section provides the institutional background on Virginia’s administration of property taxes and real
property assessments as the basis for this study.
3.
BACKGROUND ON VIRGINIA AS A NATURAL EXPERIMENT
Detractors of the residual view are functionally claiming that changes in the assessed values cause
changes in the property tax levy through a form of revenue-elasticity fiscal illusion among the voters.
Since the most significant changes in assessed values result from mass reappraisals, the ideal natural
experimental setting for testing the effect of property reassessments would be to observe a set of local
governments whose true property value growth could be observed by the researcher, but have property
reappraisals determined by random assignment. As a result, the property reassessment process would be
exogenous to the determination of the levy. Furthermore, to be a viable application of the median voter
demand model, the policy makers of these governments would have no external constraints imposed by
the state in setting their budget with respect to their ability to levy from the property tax.7
As previously described, changes in the property assessments might reflect underlying economic
growth effects, which could confound an empirical investigation in the determination of the levy. In most
instances, however, assessment values are usually the only widely available data on underlying property
growth. It is also conceivable in some instances the property assessments are conducted in some
endogenous manner, particularly if there is no cyclical basis for reappraisals. This would be particularly
problematic if updates in the property tax rolls coincided with planned changes in property tax levies,
perhaps to assure local voters of a more equitable expansion of the tax base or take advantage of fiscal
illusion. Furthermore, some states have statutory limitations on property taxes which create a different set
7
Some states set limits on growth rates on property levies, while others limit the growth in taxable property values.
7
of incentives. Local governments may be constrained, for instance, in the amount of revenue growth
from one year to the next (e.g. California) or with caps on the property tax liability as a share of assessed
values (e.g. Indiana). These types of constraints might induce governments to behave strategically to
accommodate median voter demand, both in terms of setting the levy and/or in terms of setting the
assessment level.
Virginia is home to 134 counties and independent cities, generically referred to here as
“counties,” where the institutional conditions reasonably approximate the natural experimental setting.8
First, the state imposes no serious limitation on policy makers in taxing property. The only requirement
they face is that notice must be given in a local newspaper if the property tax levy is to grow more than
one percent from the previous year. The Virginia Department of Taxation also conducts an annual “sale
price ratio study,” in which fair market transactions are randomly sampled across the state, and
consequently each property’s selling price is compared to the assessed value at the time of the sale. The
median ratio of assessed value-to-sale price (sales ratio) and other common measures are computed and
used to evaluate the accuracy of property assessments in each county. These sales ratio studies are
published annually, but with an approximate lag of 2-3 years after the sales actually take place, and
therefore 1-2 years after the budgeting process. These annual estimates of fair market value growth can
be used to capture the amount of growth that might induce demand for more public services.
Property tax levies in Virginia are determined by a county council of elected representatives,
usually going by the title “Board of Supervisors.” This council varies in number according to the county,
and council members are typically staggered so that different members are up for election in different
years. The county is the primary level of general purpose local government, and it takes on most local
public services such as health, welfare, and education.9 There does exist varying numbers of special
governments within the counties which may be delegated differing responsibilities (library, fire, etc.) and
8
In Virginia, there are several independent cities which have the same status as counties in terms of administrative
powers and responsibilities. This means they have their own independent general purpose budget.
9
Public school districts are operated at the county level in Virginia.
8
coordinate with the county, usually by proposing their own budget to the council for approval. The
property assessment process is also undertaken at the county level, but under the auspices of the county
Commissioner of Revenue, another elected official. The local Commissioner of Revenue will either
maintain the property tax rolls directly, or appoint a distinct property assessor office whose department
maintains the responsibility. Thus, in Virginia, the elected body responsible for determining the tax
burden is distinct from the one which determines the property assessments.
Virginia is a strong Dillon Rule state, so by and large the public service responsibilities are
relatively homogeneously delegated across county governments.10 In addition to making this a
reasonably comparable group of governments, this also establishes a useful scenario for the remaining
critical aspect of a natural experiment in terms of the timing of mass reappraisals across counties. In
1984, the Virginia Tax Code required that all county governments adopt a frequency cycle for property
reassessment.11 Historically, many Virginia counties would go long stretches of time between
reappraisals, over 10 years in some cases (Mikesell, 1980). The law stipulates that, for all properties, the
frequency of reassessment be greater for the higher population counties than for those with lower
populations. Independent cities with populations above 30,000 are required to reassess biannually, while
those below the cut-off may choose to reassess either annually or every four years. Counties must
reassess every four years, but they can elect a fixed 1, 2, 5, or 6-year cycle if their population is below
50,000. Counties do have the ability to change their assessment cycle frequency, but this is rarely the
case and is largely driven by compliance with the state laws as their population grows or shrinks.12 The
difference in frequency of reassessment cycles results in the contemporary situation where every year has
a different set of counties conducting mass reappraisals. Between 2000 and 2008, there were no two
years in Virginia with exactly the same set of counties conducting mass property reassessments. As a
“Dillon’s Rule” is the legal precedent that local governments only have the powers and authorities that have been
explicitly granted to them by the state.
11
Discussion of the virtues and trade-offs of options in setting reassessment cycles can be found in Mikesell (1980)
and Stine (2010).
12
The only example we have found of the changing reassessment frequency in the time frame of our data occurred
from 2000 to 2001 as the independent city of Clifton Forge lost its status as county equivalent, and was merged with
the encompassing Alleghany County and effectively increasing its population.
10
9
result, the specific year in which a county reappraises property is exogenous to the determination of the
property levy, and therefore is probably as close to random assignment as can be hoped for in the natural
world. It would seem unlikely that counties determining their assessment frequency cycle in 1984 would
do so for the explicit purpose of having a property reassessment take place in a particular year during the
millennium decade to hide a tax levy increase.
Virginia’s reassessment cycles and sales ratio studies also play an important role in distinguishing
between growth in fair market values and assessed values, since in principle the two should be relatively
collinear. However, multiyear lags between mass reappraisals create a situation where the year-over-year
change in assessed value growth will reflect cumulative changes in fair market values rather than just the
single year growth. Furthermore, the accuracy of the assessment process, as measured by Virginia’s
previously described sales ratio studies, is not universal across counties. Typically, a county which
conducts a mass reappraisal will have a median sales ratio indicating that parcels have an assessment
which is 90 to 110 percent of estimated fair market value.
In order to identify the possible role of property reassessment in concealing levy growth, there is
another important institutional detail in the timing between when the change in assessed value is known
to the owners and when it is used in setting the rate. Most property reassessments are conducted within
the first three months of the calendar year, with the latest being conducted in June, and the balance of the
year is employed for giving notice to the residents and the opportunity to appeal. The budget is set prior
to the end of the fiscal year, and as a consequence, the property reassessments are not reflected in the new
millage rate until the next calendar year. As an example, even if the reappraisals were conducted in
calendar year 2005, it would not be reflected in the millage rate confronting the taxpayer until the next
year when the budget was being set for Fiscal Year 2006-2007. This institutional setting allows us to
model demand for public services as being a function of contemporaneous market value appreciation, and
the lagged period where the intermittently-updated assessed values are finally adopted into the property
tax rate as representing the opportunity for policy makers to take advantage of any fiscal illusion.
10
A preliminary look at data from Virginia also reveals why critics may feel the principle of the
residual view is not being met. Figure 2 compares the average real property tax levy against the average
property tax millage rate between 1999 and 2009. As Figure 2 reveals, on average property tax rates have
remained relatively constant, even fallen slightly, while the average revenue raised from real property
taxation has more than doubled over the same period. For the pattern in Figure 2 to hold, it had to be the
case that growth in assessed values kept pace with the levy so as to remain rate neutral, and Figure 3
confirms this to be the case. These averages, of course, mask a considerable amount of underlying
heterogeneity in the conditions experienced across counties, but one can observe why critics of the
residual view might be concerned that rates do not properly adjust. Furthermore, levy growth is higher in
years when the budget is based on mass reappraisals at 5.1 percent, as compared to just 2 percent in nonmass reappraisal years. However, Figure 3 also demonstrates that assessed values have largely been in
accordance with estimates of fair market value growth in the areas. This suggests that demands for public
services that accompany economic growth may be the main predictor of levy growth in a more
sophisticated multivariate regression, the derivation of which is the subject of the next section.
4.
MODEL DETERMINATION AND VARIABLE SELECTION
Empirical Model
The essence of the popular concern is that local officials can deviate from the median voter’s preferences
in setting expenditure levels under the cover of rising property assessments. To test this rejection of the
residual view, the model in this paper starts with the median voter demand model in the tradition of
Borcherding and Deacon (1972), a model frequently employed in empirical work to test the degree of
non-rivalry in local public expenditures.13 This model is motivated with the assumption of a decisive
voter in area i receiving utility from the quantities of private (Xi) and public goods (gi). Using Gi to
represent total public good provision, the degree of population (Ni) rivalry of the good or service
13
The model was similarly derived by Bergstrom and Goodman (1973). Other papers employing this model for
public expenditures include McMillan et al. (1981), Edwards (1986), and Holcombe and Sobel (1995).
11
determines the individual level of consumption, resulting in gi=Ni-γGi, where γ is a congestion parameter.
The decisive voter’s income (Ii) can be used to allocate expenditures across the two goods, with the
resulting budget constraint of Ii=PxXi+TiPgGi, where Ti is the median voter’s share of the tax burden.
Substituting giNγ in lieu of Gi into the budget constraint results in a demand equation of the form
gi=gi(Ii,Px,Ti,Pg,N). Assuming that input prices to public service production are constant across the state,
while allowing for individual voter taste characteristics to be represented as vector Vi, a constant elasticity
𝛾
𝜇
demand function can be defined as 𝑔𝑖 = 𝑉𝑖 (𝑇𝑖 𝑁𝑖 )𝛼 𝐼𝑖 . Substituting gi=Ni-γGi into the left-hand side and
rearranging results in
𝜇
(1) 𝐺𝑖 = 𝑉𝑖 𝑇𝑖𝛼 𝑁𝑖𝜆 𝐼𝑖 ,
where λ=(1+α)γ. Since the quantity of government services is typically unknown, the literature proxies
for these services using expenditures, implicitly assuming that there is a constant dollar per unit of public
goods and services. The log-log form of equation (1) is then estimable using linear regression methods,
and is adopted in this paper for the specific case of identifying local government expenditures. From an
accounting perspective, balanced budgets require that government spending (Gi) in a given year equals
the property tax levy (Li) plus revenue from other sources (Ri). In Virginia counties, these other nonproperty revenue sources are predominantly composed of intergovernmental transfers, various user fees,
and cigarette taxes. Once the local government has identified revenue from all other sources it determines
total spending by issuing the property tax levy. Median voter income is composed of annual income flow
(Yi) and the income from the fair market value of their real property (Hi), although data limitations require
us to proxy the latter with just the current market value.14 Income in this form is assumed to be the
multiplicative function Iiμ=YiηHπ, with the intuition that the marginal propensity to spend on services may
differ between income and their property values. Substituting this function into (1) and adding a temporal
dimension results in the following logged empirical demand function for the property tax levy:
14
Ultimately, the implicit assumption is that the growth in the fair market value of the property will capture the
influence and correlation of the income flow from properties in the behavior of the median voter.
12
(2) 𝑙𝑛𝐿𝑖𝑡 = 𝛽𝑙𝑛𝑉𝑖𝑡 + 𝛼𝑙𝑛𝑇𝑖𝑡 + 𝜆𝑙𝑛𝑁𝑖𝑡 + 𝜂𝑙𝑛𝑌𝑖𝑡 + 𝜋𝑙𝑛𝐻𝑖𝑡 + 𝛿𝑙𝑛𝑅𝑖𝑡 + 𝜀𝑖𝑡 .
The expected signs are negative for tax price (α) and positive for housing values (π), population (λ), and
income (η). The theoretical literature on the flypaper effect of other revenue sources suggests that the
expected sign of Rit is ambiguous (Fisher, 1982; Hamilton, 1983; Hines and Thaler, 1995). Empirically, it
is likely that there exist unobserved time invariant fixed effects, and as a consequence equation (2) will be
first differenced to eliminate them:
̃𝑖𝑡 + 𝜂𝑙𝑛𝑌̃𝑖𝑡 + 𝜋𝑙𝑛𝐻
̃𝑖𝑡 + 𝛿𝑙𝑛𝑅̃𝑖𝑡 + 𝜀̃𝑖𝑡 .
(3) 𝑙𝑛𝐿̃𝑖𝑡 = 𝛽𝑙𝑛𝑉̃𝑖𝑡 + 𝛼𝑙𝑛𝑇̃𝑖𝑡 + 𝜆𝑙𝑛𝑁
Using tilde to represent the first differences, the model in equation (3) also squares with the general view
in public finance that local public officials inherit the previous year’s budget, and then update it for the
upcoming year to reflect new circumstances and information. This is done during the budgeting stage of
the process, where policymakers have the opportunity to revise the levy, while anticipating the effect on
the millage rate. As already described, critics of this process have claimed that property reassessments
provide cover for public officials to increase the levy beyond that preferred by the constituency. If the
residual view is correct, then reassessing property values will have no impact on the levy itself. To
consider this possibility, we augment the model in equation (3) with the first differenced log of assessed
values (AV) and a dummy variable indicator for the reassessment (A).
̃𝑖𝑡 + 𝜑(𝐴𝑖𝑡 ×
̃𝑖𝑡 + 𝜂𝑙𝑛𝑌̃𝑖𝑡 + 𝜋𝑙𝑛𝐻
̃𝑖𝑡 + 𝛿𝑙𝑛𝑅̃𝑖𝑡 + 𝜌𝐴𝑖𝑡 + 𝜃𝑙𝑛𝐴𝑉
(4) 𝑙𝑛𝐿̃𝑖𝑡 = 𝛽𝑙𝑛𝑉̃𝑖𝑡 + 𝛼𝑙𝑛𝑇̃𝑖𝑡 + 𝜆𝑙𝑛𝑁
̃𝑖𝑡 ) + 𝜀̃𝑖𝑡
𝑙𝑛𝐴𝑉
The indicator Ait takes the value of one if the levy was set with assessed values following a mass
reappraisal, whereas the continuous variable AVit is used to calculate the magnitude of change in assessed
values used in setting the rate.15 The interaction between the two is employed to allow for the model to
15
Note that, while a mass reappraisal takes place between January and June of the calendar year, the millage rate
derived from the property tax levy is not based on these updated values for the budgeting process until the next
calendar year. So AVit is based on changes in assessed values relevant to the budget process, even though the
assessor’s office may have already updated their parcel records.
13
distinguish the behavioral response between years with and without mass appraisals. The critical
assumption implied here is that current period public service demands are a function of current fair
market value appreciation, and that the median voter does not update their demands on a cycle that
mimics the cycle between reassessments adopted for the budget process. Since some areas go as much as
six years between mass reappraisals, the percentage increases can be orders of magnitude larger than the
years between assessments, and so the adjustment of the rate may be substantially different in these two
sets of years. The marginal effects of interest to testing the residual view are:
(5)
𝜕𝑙𝑛𝐿̃𝑖𝑡
̃ 𝑖𝑡
𝜕𝑙𝑛𝐴𝑉
(6)
𝜕𝑙𝑛𝐿̃𝑖𝑡
𝜕𝐴𝑖𝑡
= 𝜃 + 𝜑𝐴𝑖𝑡
̃𝑖𝑡
= 𝜌 + 𝜑𝑙𝑛𝐴𝑉
Consider first the marginal effect of changes in assessed values, equation (5), during years where budgets
are not based on mass reappraisals (Ait=0).16 The claim of the residual view is that rates fully adjust and
there is no effect on the property tax levy, so that θ=0. In the strongest version of revenue-elasticity fiscal
illusion, rates do not adjust at all so that increases in the base are automatically translated into new
spending (θ=1). In years with mass reappraisals (Ait=1) this intuition on θ is unchanged, but it becomes a
test of 𝜃̂ + 𝜑̂ = 0 for the residual view and 𝜃̂ + 𝜑̂ = 1 for critics. In principle, however, any substantive
magnitude between zero and one demonstrates some amount of support for fiscal illusion in the
reassessment process. There is no obvious expected sign for φ, but it is likely negative because of the
constant elasticity implied by the log-log specification. Since the percentage change in assessed values are
often orders of magnitude larger than those in the intervening years, the levy-growth responsiveness
might be smaller in percentage terms than in non-mass reappraisal years.
In equation (6), the marginal effect on the levy of a mass reappraisal is provided. Just to illustrate
the intuition of this marginal effect, suppose taxable assessed value growth was zero following mass
̃𝑖𝑡 = 0; this would be the case if the appraisal only served to redistribute assessed
reappraisal, so that 𝑙𝑛𝐴𝑉
16
Assessed values can change between reappraisals due to appeals, property improvements, and land use changes.
14
values across parcels. If the mass reappraisal allows for a form of fiscal illusion where taxpayers attribute
their new tax bill solely to the reassessment, then local politicians could increase the levy and ρ would
become positive.17 This was the specific test conducted by Bloom and Ladd (1982) and Ladd (1991),
with the exception that the model employed here controls for factors other than the mass reappraisal. The
full marginal effect in equation (6) also allows for the interaction with total assessed value growth so that
it can possibly vary by magnitude of the assessment growth.
Variable Definition and Data
While the range of available data slightly differs across the variables, all series hold in common the 2000
to 2008 period, and therefore all summary statistics and regression estimates are based on this time
period. Virginia has 134 cities and counties, but three of them (Alleghany County, Newport News City,
and York County) are excluded because they exchanged territory in 2001, causing large reported changes
in several key variables. All dollarized variables are inflation-adjusted to 2005 equivalent prior to any
transformations. Real property tax levies and other sources of revenue were obtained from the Virginia
Auditor of Public Accounts.
The school-age share of the population, share of the population over age 65, and racial
homogeneity are included as regressors in equation (4) to capture the effect of changing voter
characteristics (𝑉𝑖𝑡 ). These are all common demographic variables of interest to the literature on demand
for public services (e.g. Alesina et al, 1999; Lind, 2007; Fletcher and Kenny, 2008). Also included is the
share of housing units which are owner occupied, as there is some support for “renter illusion” that may
cause renters to not recognize the full cost of services financed by property taxes (Bergstrom and
Goodman, 1973; Blom-Hansen, 2005).18 Annual population and housing unit estimates are provided by
the U.S. Census, as are the estimates by age and race. Information on assessed values and the true fair
The interpretation of ρ is that the implementation of a mass reappraisal increased the levy growth rate by (ρ x 100)
percent.
18
We are grateful to an anonymous referee for suggesting the owner-occupied and school age children share as
additional control variables.
17
15
market value of property is provided as part of the annual assessment sales ratio studies conducted by the
Virginia Department of Taxation, and typically represent about three to four percent of the total parcels
located in the state.
Annual estimates of median adjusted gross household income are provided by the Virginia
Department of Housing and Community Development. This measure of median household income had a
change in definition for 2008, causing a large statewide increase in local estimates of income, and as a
result time fixed effects are included in every model specification. The preferred measure of the median
voter tax share typically used in empirical studies is the median assessed value as a share of total assessed
values, with the assumption being that the median value approximates the median, or decisive, voter.
Unfortunately, median assessed value is not available for any year except 2000, so instead we use the
share of median income divided by total personal income in the county as a proxy for the tax share.19
Both the ideal and the employed measures of tax price implicitly adopt the view that it is only the
residential tax price that is considered by local public officials, and that consequently non-residential tax
prices are not influential.
5.
ESTIMATION RESULTS
Main Results
Estimates of equation (4) are provided in Table 2 with alternative restrictions on the main variables of
interest, and all specifications include year fixed effects. Breusch-Pagan diagnostics for
heteroskedasticity supported the use of robust standard errors, which have been clustered by county.
The control variables generally take the expected signs. The point estimates across the
specifications in Table 2 indicate that a one percent increase in income growth was correlated with about
a 0.30 percent increase in the property tax levy change. A one percent increase in population growth was
19
A pairwise correlation between the preferred median tax share and the proxy variable employed in the paper was
0.95 in 2000, which was the only year such a correlation could be computed.
16
correlated with roughly a 0.10 percent increase in the levy, while a percentage increase in the median tax
share was correlated with about a 0.30 percent decline. The negative sign on median tax share, which
proxies for the price to the median voter for public goods, is consistent with the existence of a downward
sloping voter demand curve. Together, the population and tax price point estimates indicate that the
Samuelson “publicness” index of the spending is between 0.51 and 0.40, roughly midway between a pure
public and pure private good.20 Revenue from other non-property sources was positive and statistically
significant, consistent with a complementary correlation with property tax revenue. The demographic
variables were not statistically significant.
The estimates of Table 2 also indicate that a one percent increase in the growth of the fair market
value of property increases the property tax levy by a statistically significant 0.04 to 0.07 percent. This is
consistent with the notion of wealth effects, where property owners support more public spending as the
value of their assets increase, but is also a qualitatively small effect.21 In specifications B-E, variables
capturing the effect of changing taxable assessed values are included, with the estimates of the full model
in equation (4) appearing in specification E. As described earlier in this paper, changes in taxable values
of property differ from the fair market value in important ways. First, the biggest changes in assessed
values appear in years with mass reappraisals, which differ in frequency across areas. This causes the
changes in assessed values to be more volatile since the reappraisals must reflect property value growth in
all the intervening years. Secondly, taxable assessed values differ from fair market value by various
property exemptions. In addition, assessor errors can result in assessment-to-sale price ratios that differ
both over time and across areas. Finally, assessed value changes are not reflected in millage rates until
the budget of the next fiscal year. There is some remaining multicollinearity between fair market and
20
As derived in Borcherding and Deacon (1972) and discussed in the model derivation section, the Samuelson index
of publicness is given as the coefficient on population divided by one plus the coefficient on tax price. From column
E of Table 2, this calculation is 0.280/(1-0.304)=0.401. The highest Samuelson index in Table 2 is 0.51 in column
A.
21
There is nothing about the institutional design in Virginia that allows us to infer that the coefficient on the fair
market value growth of property can be interpreted as a causal estimate of property wealth effects. It serves only to
capture the accumulated wealth effects from economic growth that are not otherwise captured by the other variables.
17
assessed values, but it is relatively weak with variance inflation factors of 10. Multicollinearity is not
particularly troubling here since both variables have high levels of statistical significance.
The specifications in B-E of Table 2 are all supportive of some amount of fiscal illusion that
would violate the residual view. Regardless of the specification, growth in assessed values and mass
reappraisals are statistically significant determinants of property tax levy growth. From specification E it
can be seen that in those years without mass reappraisals (A=0), a one percent increases in assessed values
result in a statistically significant 0.068 percent increase in the property tax levy. For a 3.2 percent
increase in assessed value growth, the observed sample mean, there is a 0.21 percent increase in the
property tax levy, a magnitude that is 7 percent of the mean observed levy growth in the sample.22 The
marginal effect in equation (5) of assessed value growth during mass reappraisals (A=1) diminishes to
0.051 and loses statistically significance.23 At best, it seems there is a very small amount of fiscal illusion
from assessed value growth that is statistically significant, but it is substantively small and thus consistent
with the revenue neutral adjustments in the property tax rate predicted by the residual view.
In evaluating the marginal effect of a mass reappraisal, provided in equation (6), we first consider
the case with zero assessment growth (Δln(Assessed Values)=0) where the potential for fiscal illusion
would come from a zero-sum redistribution of the property assessments across individual owners. The
point estimate in column E of Table 2 suggests that mass reappraisals increase the levy growth rate by 1.5
percent.24 This effect shrinks by 0.017 percent for every one percent above zero assessment growth,
though this is not statistically significant.
In Table 3, specifications are alternated with the inclusion of area and year fixed effects.
Although the influence of time invariant omitted variables relevant to explaining the levels of the
variables would have been eliminated with first differencing, the specifications with the inclusion of area
fixed effects discard the influence of any remaining omitted variables which might affect the rate of
22
Calculations: 3.2x0.068=0.218; 0.218/3.0 = 0.07 (7%).
Calculation: 0.068-0.017=0.051.
24
Calculation: 100 x 0.015=1.5%.
23
18
change in the levy. Both appraisals and assessed value growth retain at least 5 percent level of statistical
significance, with magnitudes remaining relatively unchanged from those presented in Table 2. The
primary consequence of including fixed effects is the sensitivity of statistical significance of the other
variables.
The main findings presented in Tables 2 and 3 demonstrate there are some detectable amounts of
fiscal illusion that is statistically significant, but substantively small.25 The primary interpretation is that
mass reappraisals provide some cover for politicians to raise levies, but the aggregate growth in taxable
assessed values is largely irrelevant. In essence, the politicians can raise a small amount of additional
revenue through the levy under the guise that most owners will simply attribute it to their new property
reassessment, and without the cover of the mass appraisal that revalues the properties across the board
this is much harder to do. Our findings are also substantially smaller than those found in the most
comparable previous literature on fiscal illusion with respect to property reassessments. Bloom and Ladd
(1982) found that mass reappraisals were associated with a 5.1 percent levy increase in Massachusetts,
but these findings had no other controls for economic growth and mass appraisals were likely determined
endogenously to annual levy growth. Ladd (1991) found the impact of assessments to increase property
tax levies by an additional 3 to 8 percent in North Carolina counties where reassessments were likely
exogenous in their timing, but once again there were no controls for economic growth. Our approach
finds a substantively smaller amount of levy growth at approximately 1.5 percent.
Specification Checks
In the results presented in Tables 2 and 3, the assumption was made that wealth effects arising from the
median voter came from the true fair market value of their property, which is derived from the state’s
annual sales ratio study of assessment accuracy. An alternative possibility is that the owner’s primary
25
Interestingly, although the detectable amount of fiscal illusion is small, the magnitude is comparable to the
coefficient size on fair market value appreciation, although we cannot make comparable claims of causality.
19
information source for property value is their assessed value.26 To explore this possibility we take
advantage of a difference in timing between when the assessed values are known to the property owner
(Assessed Values†) and when they are employed for the purpose of tax rate setting (Assessed Values‡), a
lag that exists to allow for an appeal period to occur. As previously described, the latter assessed values
is the same measure previously employed in the measure of Assessed Value growth in Tables 2 and 3
since that was the relevant opportunity for taking advantage of voter fiscal illusion. Table 4 and Table 3
differ only in that the growth in fair market property values has been replaced with the growth in assessed
values as they are known to the property owners. We can see that the assessed value growth known to the
owners (†) is positive and statistically significant, albeit with smaller magnitudes than when the
counterpart measure of fair market value growth was used in Table 3. More importantly for our purposes,
however, the magnitudes and statistical significance for the variables of interest of interest, Mass
Reappraisal and Assessed Values‡, were virtually unchanged.
The results in comparing Table 4 to Table 3 also present some evidence on another possible
limitation of the test. If there is error in the measure of fair market value, the positive correlation between
assessed value and market value appreciation could introduce some positive bias to the estimated
coefficient on the assessment variables. This correlation is weakened by the budgetary lag in adopting the
assessed values used for rate-setting, but Table 4 also reveals that when fair market value is replaced by
contemporaneous assessed value growth the coefficients are virtually unchanged. Since the
contemporaneous assessed value carries an even larger difference from market value, it seems likely that
this potential bias is small.
Though not presented, some additional sensitivity checks are discussed in the remainder of this
section and are available upon request. For brevity, all of the sensitivity assumptions will be compared to
specification D of Table 3, the full model with both year and county fixed effects. The final year of the
dataset, 2008, would have reflected adjustments to the budgets with the housing market crash towards the
26
We appreciate an anonymous referee who suggested this alternative possibility.
20
end of the 2007 calendar year. Estimating the model without 2008 data has virtually no effect on the
results. The point estimate on mass reappraisals is 1.5 percent and is statistically significant at the 5
percent level, down from 1.7 percent with 1 percent significance. The coefficient on assessed value
growth remains at 0.059 with a p-value below 0.05.
One of the attributes of Virginia is that state law creates heterogeneity in the frequency of mass
reappraisals. A potential limitation of this law, however, is that areas with more frequent mass
reappraisals will carry more weight in the mean-centered parameter estimate on the Mass Reappraisal
indicator. To examine the sensitivity of the results to this phenomenon, the full model with both area and
year fixed effects was estimated with a weighted regression, where the weight was based on the inverse
frequency of their number of mass reappraisals in the sample. Comparing to the unweighted regression in
specification D of Table 3, the point estimate on Mass Reappraisal increased to a 2.3 percent increase in
property tax levies that was statistically significant at the five percent level. The point estimate on
assessed value growth was 0.055, only slightly lower than in the unweighted regression, though its
statistical significance fell below the 5% level. Finally, there is a considerable literature in local public
finance on tax competition among governments (for e.g., see Wilson, 1999; Brueckner and Saavedra,
2001). Using spatial econometrics to allow for neighbor’s property tax levy growth as another control
variable demonstrated no substantive difference in the main variables of interest, and the statistical
significance of the spatial lag carried a p-value of just 0.10.27
Alternative specifications of non-property revenue were included or replaced for the one
presented, including separate controls for state and federal transfers, but this had no substantive effect on
our variables of interest and the r-squares were virtually unchanged. Finally, as previously described,
three areas were excluded from the regression because they changed territorial boundaries with each other
in 2000 to 2001, causing relatively large swings in several of their variables. Including them in the
27
The spatial weight matrix most favorable to a finding of spatial dependence was the first-order contiguity matrix,
for which the tax mimicking parameter was about 0.07. The result was estimated using the maximum likelihood
approach described in LeSage and Pace (2009). We are grateful to an anonymous referee for suggesting this
robustness check.
21
regression, however, did not affect the main variables of interest within the rounding tolerance of the
reported result in specification D of Table 3.
6.
CONCLUDING REMARKS
The residual view of the property tax is that changes in assessed values have no direct influence on the
amount of revenue raised through real property taxation because rates adjust as a matter of arithmetic
(Netzer, 1964). Critics of property taxes suggest that the reassessment process creates a form of fiscal
illusion, analogous to the revenue elasticity hypothesis (Oates, 1975; Dollery and Worthington, 1996).
The few papers which have previously tested these competing views directly, however, have considerable
shortcomings with respect to a causal identification approach. Particularly problematic for the previous
literature has been the inability to control for economic growth in an area that may be associated with
changes in assessed values, as well as endogeneity in the decision to conduct reassessments. Using 2000
to 2008 data from Virginia cities and counties, this paper provides the clearest evidence to date on the
validity of the residual view. We argue that the institutional structure results in a reassessment cycle that
is exogenous to a given year’s levy growth, and we are able to control for confounding economic growth
factors. Our findings suggest that there is a statistically significant but analytically trivial amount of fiscal
illusion that allows for levy growth in response to increases in taxable assessed values. Mass reappraisals,
on the other hand, do provide some cover for politicians to hide levy increases on the order of about 1.5
percent, but the aggregate growth in taxable assessed values that occurs due to the reassessment is largely
irrelevant. We speculate that the mass reappraisals allow politicians to raise a small amount of additional
revenue because most owners will likely ascribe changes in their tax bill primarily to their new
assessments. Our finding on mass reappraisals is substantially smaller, however, than the most
comparable literature found in Bloom and Ladd (1982) and Ladd (1991), which produced estimates of 5.1
and 3 to 8 percent, respectively.
22
Despite the generally high visibility of the property tax to which many scholars attribute to its
unpopularity, this paper joins Cabral and Hoxby (2010) in finding evidence that this instrument is not
completely exempt from fiscal illusion. To the extent the mass reappraisals allow for politicians to push
levy growth above trend, this fiscal illusion detracts from the median voter theory as a descriptive theory
of local government. The evidence presented here, of course, is qualified by the external validity of
applying Virginia results to other states. States like California, Florida, and Michigan place limits on the
ability of local governments to raise revenue through the property tax, which may induce local
governments to strategically raise levies beyond current median voter demands in anticipation of being
unable to raise levies again in the future.
It is possible that additional transparency laws may help circumvent this phenomenon. Some
states have also adopted “Truth in Taxation” laws that apply to the property tax, which can include the
publication of important ballot information or direct mailings of budget information (Youngman and
Malme, 2005). In Virginia, cities and counties must only publish an announcement in the local paper if
levies are to increase more than one percent in a particular year. It is possible that strengthening such laws
further, perhaps by informing property taxpayers what their individual liability would have been under the
previous year’s levy, might assist in overcoming the fiscal illusion that occurs in mass reappraisal years.
Limiting the discretionary power local politicians have over general purpose budgets, such as by requiring
voter approval for significant tax levy increases, might also be a step towards constraining such behavior.
REFERENCES
Alesina, Alberto, R. Baqir, and William Easterly. 1999. Public Goods and Ethnic Division. Quarterly
Journal of Economics 114: 1243-84.
Alm, James, Robert D. Buschman, and David L. Sjoquist. 2011. Rethinking Local Government Reliance
on the Property Tax. Regional Science and Urban Economics 41:320-331.
Bergstrom, T. C, and R. P. Goodman. 1973. Private demand for public goods. American Economic
Review 63: 280-296.
Blom-Hansen, Jens. 2005. Renter Illusion: Fact or Fiction? Urban Studies 42(1): 127-40.
23
Bloom, Howard S., and Helen F. Ladd. 1982. Property Tax Revaluation and Tax Levy Growth. Journal of
Urban Economics 11 (1):73-84.
Borcherding, Thomas E., and Robert T. Deacon. 1972. The Demand for Services of Nonfederal
Governments. American Economic Review 62 (Dec.):891-901.
Brueckner, Jan K. and Luz A. Saavedra. 2001. Do Local Governments Engage in Strategic Property-Tax
Competition? National Tax Journal 54(2): 203-229.
Cabral, M. and C. Hoxby. 2010. The Hated Property Tax: Salience, Tax Rates, and Tax Revolts. Stanford
University Working Paper. Retrieved from http://www.stanford.edu/~mcabral/Tax_Salience.pdf.
Cornia, Gary C. and Lawrence C. Walters. 2006. “Full Disclosure: Controlling Property Tax Increases
During Periods of Increasing House Values.” National Tax Journal 59 (September): 735-49.
Doerner, William M., and Keith R. Ihlanfeldt. 2011. House Prices and City Revenues. Regional Science
and Urban Economics 41:332-342.
Dollery, Brian E. and Andrew C. Worthington. 1996. The Empirical Analysis of Fiscal Illusion. Journal
of Economic Surveys 10(3): 261-297.
Dougan, W.R. and Kenyon, Daphne A. 1988. Pressure Groups and Public expenditures: The Flypaper
Effect Reconsidered. Economic Inquiry 26: 159-70.
Edwards, J. H. Y. 1986. A note on the publicness of local goods: Evidence from New York State
municipalities. Canadian Journal of Economics 19: 568-573.
Faulk, Dagney. 2004. How We Got Here from There: A Chronology of Indiana Property Tax Laws.
Indiana Business Review 79(2): 1-3.
Fletcher, Deborah and Lawrence W. Kenny. 2008. The Influence of the Elderly on School Spending in a
Median Voter Framework. Education Finance and Policy 3(3): 283-315.
Fisher, Ronald C. 1982. Income and grant effects on local expenditure: The flypaper effect and other
difficulties. Journal of Urban Economics 12 (3):324-45.
———. 2007. State & Local Public Finance, 3rd Edition: Thomson South-Western.
Hamilton, Bruce W. 1983. The Flypaper Effect and Other Anomalies. Journal of Public Economics 22
(3):347-361.
Hines, James R., and Richard H. Thaler. 1995. Anomalies: The Flypaper Effect. The Journal of Economic
Perspectives 9 (4):217-226.
Holcombe, R. G., and R. S. Sobel. 1995. Empirical evidence on the publicness of state legislative
activities. Public Choice 83: 47-58.
Ladd, Helen F. 1991. Property Tax Revaluation and Tax Levy Growth Revisited. Journal of Urban
Economics 30 (1):83-99.
LeSage, James P. and R. Kelley Pace. 2009. Introduction to Spatial Econometrics. CRC Press: Taylor &
Francis Group.
Lind, Jo Thori. 2007. Fractionalization and the Size of Government. Journal of Public Economics 91: 5176.
Lutz, Byron F. 2008. The Connection Between House Price Appreciation and Property Tax Revenues.
National Tax Journal 41 (3):555-572.
Lutz, Byron, Molloy Raven, and Hui Shan. 2011. The Housing Crisis and State and Local Government
Tax Revenue: Five Channels. Regional Science and Urban Economics 41:306-319.
Martin, Isaac William. 2008. The Permanent Tax Revolt: How the Property Tax Transformed American
Politics. Stanford: Stanford University Press.
McMillan, M. L., R. W. Wilson and L. M. Arthur. 1981. The publicness of local public goods: Evidence
from Ontario municipalities. Canadian Journal of Economics 14: 596-608.
Mikesell, John L. 1978. Property Tax Assessment Practice and Income Elasticities. Public Finance
Quarterly 6 (1):53-63.
———. 1980. Property Tax Reassessment Cycles: Significance for Uniformity and Effective Rates.
Public Finance Review 8 (1):23-37.
———. 2011. Fiscal Administration: Analysis and Applications for the Public Sector. 8 ed. Boston:
Wadsworth Cengage Learning.
24
Netzer, Dick. 1964. Income Elasticity of the Property Tax: A Post-Mortem Note. National Tax Journal 17
(2):205-207.
Oates, Wallace E. 1975. Automatic Increases in Tax Revenues – The Effect on the Size of the Public
Budget. In W.E. Oates (ed) Financing the New Federalism: Revenue Sharing Conditional Grants
and Taxation. Baltimore: John Hopkins University Press.
Oates, Wallace E. 1979. Lump-sum Intergovernmental Grants have Price Effects. In P. Mieszkowski and
W. Oakland (ed) Fiscal Federalism and Grants-in-Aid. Washington D.C.: The Urban Institute.
Robert D. Lee, Jr., Ronald W. Johnson, and Phillip G. Joyce. 2008. Public Budgeting Systems. Sudbury,
Massachusetts: Jones and Bartlett Publishers, Inc.
Smith, Adam. 1776. An Inquiry into the Nature and Causes of the Wealth of Nations. New York: The
Modern Library.
Stine, William F. 2010. Estimating the Determinants of Property Reassessment Duration: An Empirical
Study of Pennsylvania Counties. Journal of Regional Analysis and Policy 40 (2):143-159.
Wilson, John Douglas. 1999. Theories of Tax Competition. National Tax Journal 52(2): 269-304.
Youngman, Joan, and Jane Malme. 2005. Stabilizing Property Taxes in Volatile Real Estate Markets.
Lincoln Instiute of Land Policy: Land Lines 17 (3).
25
Figure 1: Growth of U.S. Housing Prices and Local Property Tax Revenue, 1997-2009
2.50
1997=1.0
2.00
1.50
1.00
0.50
0.00
1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009
Housing Price Index
Local Property Tax Revenue
Data: Housing Price Index (Federal Housing Finance Agency); Local Property Tax Revenue (U.S. Census
of Government Finance).
26
70
20
30
40
50
60
Property Tax Levy (Millions, 2005$)
1
.8
.6
.4
.2
0
Property Tax Millage Rate
Figure 2: Average Local Property Tax Rates Versus Real Property Tax Levies in Virginia, 1997-2009
1995
2000
2005
2010
Year
Property Tax Millage Rate
Property Tax Levy (Millions, 2005$)
Note: Levy data from Virginia Auditor of Public Accounts; millage rate from Virginia Department of
Taxation (available only starting in 1998).
27
Figure 3: Average Local Assessed Values Versus Estimated Fair Market Values for Real Property in
Virginia, 1998-2009
Notes: Data from Virginia Department of Taxation
28
Table 1: Summary Statistics and Variable Description
Variable
Δln(Property Tax Levy)
Δln(Non Property Revenue)
Δln(Fair Market Property Value)
Δln(Median Tax Share)
Δln(Population)
Δln(Median Household Income)
Δln(Share School Age)
Δln(Share Age 65+)
Δln(Share Owner Occ.)
Δln(Racial Homogeneity)
Mass Reappraisal
Δln(Assessed Values)
Mean
Std.Dev. Min
Max
0.030
0.067
-0.356
0.510
0.005
0.069
-0.478
0.396
0.052
0.133
-0.234
0.814
0.022
0.153
-0.502
1.434
0.010
0.025
-0.090
0.193
0.034
0.143
-0.528
1.417
-0.007
0.027
-0.232
0.272
0.012
0.101
-0.688
1.269
0.005
0.006
-0.034
0.074
-0.001
0.071
-0.650
0.620
0.321
0.467
0.000
1.000
0.032
0.234
-0.529
1.629
Sample Size: 1,172
Notes & Sources: All currency data deflated to real 2005 dollars prior to transformations; Δln(variable)
indicates the year-over-year change in the natural log of the variable. Property Tax Levy: Total revenue
to be raised from taxation of real property (Virginia Auditor of Public Accounts); Non Property Revenue:
local revenue raised from all sources other than real property taxation (Virginia Auditor of Public
Accounts); Fair Market Property Value: total true fair market value of all taxable properties as estimated
in the state assessment-sale price ratio study (Virginia Department of Taxation); Median Household
Income: median adjusted gross income from tax returns (Virginia Department of Housing and
Community Development); Median Tax Share: median household income divided by total personal
income in thousands of dollars (Bureau of Economic Analysis); Share School Age: Share of population
that is age 5-19 (U.S. Census); Share Age 65+: population over age 65 divided by total population (U.S.
Census); Share Owner Occ.: share of the housing units that are owner occupied (occupancy status and
building permit data on residential construction from the U.S. Census); Racial Homogeneity: a
Herfandahl-Hirschman Index of racial population composition between whites, blacks, Asians, American
Indians, and miscellaneous divided by 10,000 (authors’ calculation from U.S. Census Data); Mass
Reappraisal: dummy variable indicating that the year’s budget is based upon assessed values following a
mass reappraisal (Virginia Department of Taxation); Assessed Values: total assessed values of taxable
properties used in that year’s public budget, as estimated in the state assessment-sale price ratio study
(Virginia Department of Taxation). Δln(M.H.I.)xY2008: interaction of median household income with
year 2008 dummy to account for change in measurement.
29
Table 2: OLS Estimates of Property Tax Levy Growth for Virginia cities and counties with Year
Fixed Effects, 2000-2008.
Dep. Var.: Δln(Property Tax Levy)
Δln(Non Property Revenue)
Δln(Fair Market Property Value)
Δln(Median Tax Share)
Δln(Population)
Δln(Median Household Income)
Δln(M.H.I.)xY2008
Δln(Share School Age)
Δln(Share Age 65+)
Δln(Share Owner Occ.)
Δln(Racial Homogeneity)
A
B
0.079**
0.075**
C
0.082**
D
0.079**
E
0.079**
(0.036)
(0.035)
(0.036)
(0.036)
(0.036)
0.041***
0.053***
0.072***
0.072***
0.073***
(0.015)
(0.015)
(0.015)
(0.015)
(0.015)
-0.342*** -0.292*** -0.330*** -0.306*** -0.304***
(0.091)
(0.088)
(0.085)
(0.084)
(0.084)
0.341***
0.280***
0.307***
0.282***
0.280***
(0.095)
(0.094)
(0.091)
(0.091)
(0.091)
0.124
0.101
0.110
0.101
0.100
(0.116)
(0.111)
(0.109)
(0.108)
(0.107)
0.252***
0.218***
0.254***
0.236***
0.235***
(0.087)
(0.084)
(0.082)
(0.082)
(0.082)
0.087
0.037
0.070
0.047
0.045
(0.062)
(0.061)
(0.059)
(0.060)
(0.060)
0.021
0.014
0.010
0.008
0.009
(0.023)
(0.023)
(0.021)
(0.022)
(0.022)
-0.074
0.121
0.049
0.128
0.173
(0.274)
(0.285)
(0.245)
(0.256)
(0.281)
0.000
0.005
0.005
0.006
0.006
(0.023)
(0.022)
(0.022)
(0.021)
(0.021)
0.014***
0.015***
Mass Reappraisal
0.027***
(0.005)
Δln(Assessed Values)
(0.005)
(0.005)
0.071***
0.057***
0.068***
(0.011)
(0.011)
(0.021)
Mass Reappraisal x Δln(A.V.)
-0.017
(0.024)
Intercept
0.021**
0.012
0.020**
0.016*
0.016*
(0.008)
(0.008)
(0.008)
(0.008)
(0.008)
2
0.107
0.140
0.159
0.166
0.167
R
Notes: Sample size is 1,172, with 131 cross-sectional units. Statistical significance level indicated
with *** (p<0.01), ** (p<0.05), * (p<0.1). Robust standard errors clustered by area reported in
parentheses. All specifications include year fixed effects.
30
Table 3: Property Tax Levy Growth Determinants for Virginia cities and counties with Area and
Year Fixed Effects, 2000-2008
Dep. Var.: Δln(Property Tax Levy)
A
Δln(Non Property Revenue)
0.064*
Δln(Fair Market Property Value)
Δln(Median Tax Share)
Δln(Population)
B
C
0.079**
0.056
D
0.070*
(0.035)
(0.036)
(0.037)
(0.038)
0.071***
0.073***
0.056***
0.058***
(0.015)
(0.015)
(0.015)
(0.015)
-0.198*** -0.304***
-0.044
-0.105
(0.073)
(0.084)
(0.083)
(0.102)
0.401***
0.280***
0.254***
0.105
(0.088)
(0.091)
(0.094)
(0.095)
0.022
0.100
-0.176*
-0.165
Δln(Median Household Income)
(0.092)
(0.107)
(0.104)
(0.132)
0.163***
0.235***
0.197***
0.302***
(0.055)
(0.082)
(0.055)
(0.087)
0.043
0.045
0.034
0.036
(0.065)
(0.060)
(0.072)
(0.067)
-0.001
0.009
-0.013
-0.003
(0.021)
(0.022)
(0.023)
(0.024)
0.097
0.173
0.062
0.209
(0.277)
(0.281)
(0.304)
(0.328)
0.008
0.006
0.000
-0.001
(0.023)
(0.021)
(0.026)
(0.023)
Mass Reappraisal
0.014***
0.015***
0.016***
0.017***
(0.005)
(0.005)
(0.006)
(0.006)
Δln(Assessed Values)
0.070***
0.068***
0.060**
0.060**
(0.021)
(0.021)
(0.025)
(0.025)
-0.019
-0.017
-0.020
-0.020
Δln(M.H.I.)xY2008
Δln(Share School Age)
Δln(Share Age 65+)
Δln(Share Owner Occ.)
Δln(Racial Homogeneity)
Mass Reappraisal x Δln(A.V.)
(0.025)
(0.024)
(0.031)
(0.031)
0.013***
0.016*
0.017***
0.026***
(0.003)
(0.008)
(0.003)
(0.009)
Area FE?
No
No
Yes
Yes
Year FE?
No
Yes
No
Yes
Intercept
2
0.135
0.167
0.088
0.127
R
Notes: Sample size is 1,172, with 131 cross-sectional units. Statistical significance
level indicated with *** (p<0.01), ** (p<0.05), * (p<0.1). Robust standard errors
reported in parentheses; specifications without area fixed effects are clustered by
area. For models with area fixed effects R2 , reported is the within-R2 .
31
Table 4: Property Tax Levy Growth Determinants for Virginia Cities and Counties with Area and
Year Fixed Effects, 2000-2008
Dep. Var.: Δln(Property Tax Levy)
A
Δln(Non Property Revenue)
0.067*
Δln(Assessed Values)†
Δln(Median Tax Share)
Δln(Population)
B
0.080**
C
0.058
D
0.070*
(0.035)
(0.036)
(0.037)
(0.038)
0.028***
0.032***
0.022***
0.026***
(0.007)
(0.008)
(0.008)
(0.008)
-0.206*** -0.320***
-0.041
-0.111
(0.074)
(0.085)
(0.084)
(0.104)
0.425***
0.301***
0.264***
0.115
(0.089)
(0.092)
(0.094)
(0.096)
0.041
0.125
-0.171
-0.156
Δln(Median Household Income)
(0.093)
(0.108)
(0.106)
(0.134)
0.144***
0.226***
0.184***
0.298***
(0.055)
(0.083)
(0.056)
(0.087)
0.034
0.037
0.026
0.029
(0.066)
(0.060)
(0.073)
(0.066)
-0.005
0.007
-0.018
-0.005
(0.021)
(0.023)
(0.024)
(0.025)
0.140
0.216
0.079
0.229
(0.292)
(0.293)
(0.312)
(0.336)
0.008
0.006
0.000
-0.002
(0.024)
(0.022)
(0.026)
(0.024)
Mass Reappraisal
0.015***
0.015***
0.016***
0.017***
(0.005)
(0.005)
(0.006)
(0.006)
Δln(Assessed Values)‡
0.066***
0.066***
0.057**
0.058**
(0.022)
(0.021)
(0.025)
(0.025)
-0.019
-0.018
-0.019
-0.021
Δln(M.H.I.)xY2008
Δln(Share School Age)
Δln(Share Age 65+)
Δln(Share Owner Occ.)
Δln(Racial Homogeneity)
Mass Reappraisal x Δln(A.V.)‡
(0.025)
(0.024)
(0.031)
(0.031)
0.015***
0.017**
0.019***
0.027***
(0.003)
(0.008)
(0.003)
(0.009)
Area FE?
No
No
Yes
Yes
Year FE?
No
Yes
No
Yes
Intercept
2
0.129
0.162
0.084
0.124
R
Notes: Sample size is 1,172, with 131 cross-sectional units. Statistical significance
level indicated with *** (p<0.01), ** (p<0.05), * (p<0.1). Robust standard errors
reported in parentheses; specifications without area fixed effects are clustered by
area. For models with area fixed effects R2 , reported is the within-R2 . †- indicates
assessed value growth at the time in which the households are informed of them; ‡ indicates assessed values when used for determining millage rate.
32