city taxes and property tax bases***€¦ · have investigated the impacts of state and the...

CITY TAXES AND PROPERTY TAX BASES***

HELEN F. LADD* AND KATHARINE L. BRADBURY**

ABSTRACT Case, Papke, and Koenigsberg (1983);

This paper investigates the simulta-Kieschnik (1981); and Wasylenko (1981).

neous relationship between tax rates and Recent econometric investigations of the

city property tax bases using data for 86 link between state and local taxes and the

large U.S. cities in 1972, 1977, and 1982. location of economic activity fall into two

We find that a 10 percent increase in a city's categories. First are those that focus on

property tax rate decreases the property tax particular types of location decisions, typ-

base by about 1.5 percent. Local income ically the branch plant decision which, one

taxes and taxes levied by overlying juris- might argue, is the most likely to be af-

dictions also have negative impacts on the fected by taxes. (See, for example, Carlton

city's property tax base. We conclude that (1979), Bartik (1985, 1986), and Schmen-

taxes affect local property values more thanner (1982).) This approach has the advan-

is typically implied by previous studies that tage of allowing the researcher to ground

have investigated the impacts of state and the empirical analysis in the microeco-

local taxes on firms' location decisions. nomic theory of firm behavior. The sec-ond approach takes a much broader def-inition of economic activity such as the

HE property tax has traditionally been level or change in employment or capitalTand continues to be the major revenue investment either for all industries orsource of U.S. cities. Although cities have certain industrial sectors. (See for exam-increasingly turned to other local reve- ple, Papke (1986) who focuses on invest-nue sources such as sales and income ment and Bradbury, Downs, and Smalltaxes, they still rely heavily on the local (1982) who focus on changes in employ-property tax to finance their local public ment.) These more aggregate variables areexpenditures. Despite this heavy reli- harder to model precisely because theyance, surprisingly little is known about reflect a variety of economic decisions in-the impact of a city's property tax rate on cluding, for example, the decision to ex-its tax base. pand, to shut down, to set up a new branch

Many local public officials apparently plant, or to start a new firm.believe that property taxes have signifi- Even if economists better understandcant adverse effects on city economic ac- the links between local taxes and the lo-tivity. Witness, for example, their fre- cation and expansion decisions of firms,quent willingness to grant tax abatements they would still be unable to answer theto encourage economic investment in the central question of this study, namely,city. Applied economists, in contrast, are what impact do local property taxes havenearly unanimous in their skepticism both on the size of a city's property tax base?about the wisdom of such tax breaks and, The difficulty arises because the city taxmore fundamentally, about the magn'l- base includes residential as well as busi-tude of the adverse behavioral effects of ness property and because the marketstate and local taxes. Many empirical value of such property reflects not onlystudies appear to support the contention the intensity of economic development (thethat differentials in state and local tax quantity of capital) but also location rents,burdens are simply too small to offset dif- that is, the prices that firms and house-ferences in the more basic determinants holds are willing to pay to invest in theof firm location such as labor costs and central city rather than elsewhere. Highaccessibility to markets. (See surveys by property taxes may reduce the size of the

*Duke University, Durham, NC 27706 tax base either by reducing the level of'*Federal Reserve Bank of Boston, Boston, MA business or residential economic activity

02106 in the city or by being capitalized into

503

504 NATIONAL TAX JOURNAL [Vol. XLI

lower property values, or by some com- local taxes can be modeled more easilybination of both. Whatever the mecha- than could those of an aggregated set ofnism, city officials and economists ought governmental units. Hence, our goal is toto care about the responsiveness of the measure the impact of city property taxescity's property tax base to the property tax on city property tax bases, controlling forrate. To the extent that the current tax all other taxes that may affect the amountrate reduces the size of the base, the ad- of taxable property in the city.ditional tax rate needed to finance a given The basic data are for 86 U.S. centralincrease in public expenditures will be cities for three years-1972, 1977, andhigher. 1982. The 86 comprise all those American

Other state and local taxes such as in- cities with population over 300,000 in 1970come or sales taxes may also affect local or 1980 plus all the central cities of theproperty values despite their initial inci- 50 largest SMSAs in either 1970 or 1980.dence on non-property factors of produc- Thus, the analysis includes all major cen-tion or on other economic transactions. tral cities in the United States. The com-Hence, a secondary goal of this paper is bined 1980 population of the 86 citiesto provide quantitative estimates of the constitutes 21 percent of the 1980 U.S.effects of other state and local taxes on population and 94 percent of the U.S. pop-the size of a city's property tax base. This ulation in central cities containing 50,000aspect of the study can be viewed as the or more people.first step in a larger and more ambitious Property taxes are less important for citystudy that would examine the effects of governments than for other types of localall major city taxes on each of a city's tax governments such as counties and schoolbases. districts, but they still accounted for over

half of the 1982 tax revenues of the av-erage city in our sample. Dependence on

I. The Data and the Role of the the property tax varies greatly across cit-Property Tax ies and regions; cities with population un-

der 100,000 receive more than twice asOur empirical work focuses on the re- large a share of revenues from the prop-

lationship between city property taxes and erty tax as cities with population over 1the market value of potentially taxable million and cities in the Northeast de-property in U.S. central cities. City gov- pend on property tax revenues twice asemments, however, are often not alone in heavily as those in the West.having the power to levy property taxes Effective property tax rates, like depen-on the property located within city dence on the property tax, also vary sub-boundaries. In many metropolitan areas, stantially across cities. We define an ef-independent school districts, county gov- fective tax rate asernments, and special districts are alsoauthorized to tax city property if it falls t T/B,within their jurisdictional boundaries.Moreover, the economic activity gener- where T is the city's total revenues fromated by the city's property may be subject the property tax, and B is the market valueto state or local sales and income taxes. of all potentially taxable property in theRegardless of whether they apply specif- city. The tax base in this measure is in-ically to property or are levied by over- tended to be independent of how each citylying governments, all of these taxes (and defines its tax base in practice and to in-the corresponding public services they fi- clude all property other than that uni-nance) could affect the size of a city's tax versally exempt from the property tax suchbase. Hence, all must be taken into ac- as churches and government buildings.count. Data on property tax revenues are readily

Our primary perspective, however, is the available from the Census of Govern-city government itself. As an independent ments, but the market value of each city'sdecision-making entity, its decisions about potential tax base had to be estimated from

No. 41 CITY TAXES AND PROPERTY TAX REVENUES 505

data on assessed values and assessment/ alone. The inclusion of property taxessales ratios, supplemented, where possi- levied by overlying governments wouldble, with data gathered directly from cit- raise the rates substantially in some cases.)ies.' Because of missing data, information The table shows that tax rates are gen-on tax bases is available for only 68 of the erally highest for the smallest and the86 cities for all three years, but a total of largest cities in the sample. This pattern202 city-year observations are available reflects heavy dependence on the prop-for the pooled regression analysis re- erty tax relative to other revenue sourcesported below. in the small cities and heavy overall tax-

Table 1 shows the level and variation ation in the large cities. The table alsoin effective tax rates for the 68 cities with shows that among regions, rates in north-complete information. The average rate eastern cities are strikingly higher thandecreased from 0.9 percent in 1972 to 0.6 those elsewhere.percent in 1982. (These rates may appear This wide variation in effective tax rateslow, but they are averages of city tax rates provides a natural experiment for exam-

Table 1Average Effective Property Tax Rates

(percent)

NumberCities by Group Of Cities 1972 1977 1982

All 68 .86% .80% .64%

Population Size:

Less than 100,000 6 1.42 1.17 1.12100,000 - 250,000 18 .74 .69 .62250,000 - 500,000 23 .68 .58 .45500,000 - 1,000,000 16 1.00 1.01 .68Greater than 1,000,000 5 1.00 1.05 .85

Regi :

Northeast 11 1.93 2.03 1.72North Central 15 .63 .57 .47South 22 .88 .72 .62Went 20 .42 .37 .20

Notes: Averages shown for cities with data available in all threeyears.

These effective property tax rates refer to property taxes of citygovernments alone; see text for definition of effective property taxrate and sources. Economic activity in these cities may also besubject to property taxes imposed by independent school districts orcounty governments.


ining the hypothesis that city tax rates sections that follow. Table 2 reports theiraffect the attractiveness of a city to mnemonic variable names, definitions,households and firms and thereby di- means, and standard deviations.rectly influence the size of the city's po- Property Tax Rate. A higher tax rate istential property tax base. Correspond- expected to reduce the value of the city'singly wide variation in city use of property tax base in part because the landalternative revenue sources and in tax component of the tax will be capitalizedburdens imposed on city residents by into lower property values. In addition, theoverlying jurisdictions also provide the improvements component of the tax is ex-information needed to determine the ef- pected to induce producers to shift awayfects of other taxes on city property tax from capital toward labor and to reducebases. the attractiveness of the city as a place

for investing capital. The reduced attrac-

11. Conceptual Framework tiveness of the city may manifest itselfeither in a change in land prices or in re-

Our model has four structural compo- duced economic activity or, most likely, innents: 1) a tax base equation, 2) a bal- some combination of both.anced budget equation, 3) a demand Estimating the relationship between theequation for public spending, and 4) a tax market value of city property tax basesmix equation. The tax base equation cap- and city property tax rates would be rel-tu'res the effects of private sector deci- atively straightforward if tax rates couldsions about locating and investing in the be viewed as exogenous. Such rates wouldcentral city. The demand and tax mix be exogenous, for example, if state lawsequations describe the behavior of local mandated binding limits on each city'spublic decisionmakers and, with the bal- property tax rate and all cities were atanced budget requirement, determine the their specified legal limits so that city taxlevel of property tax revenues in each city. rates would not be affected by the size ofOur primary focus is on the tax base the city tax bases. Perusal of the laws af-equation. For reasons discussed in the next fecting city property taxes, however, sug-section, however, careful attention to the gests that in most cases a city's propertyother three components is needed to as- tax rate, expressed as a fraction of fullsure appropriate estimation of the effect market value, is not independent of theof local property tax rates on city tax bases. city's tax base. While many cities are sub-

ject to tax limitations of some form, only

The Tax Base Equation rarely do these limitations determine theeffective tax rate. The clearest exceptions

The size of a city's property tax base re- are California cities in 1982 and Bostonflects investment decisions made over time in 1982, each of which is subject to aby households and firms in response to the binding limit ex ressed in terms of an ef-perceived costs and benefits of a city lo- fective tax rate.pcation. In modelling the outcome of these More commonly, the local property taxdecisions, we focus on the local property rate is determined in part by the size oftax rate, controlling for three sets of ex- the tax base. Consider, for example, theogenous variables in addition to year following model of city behavior. Citydummies, and the lagged dependent vari- government officials choose a level of ex-able. The first set of exogenous variables penditures in response to the demand ofare those that emerge from a simple citizen voters for public services. They thenmonocentric model of an urban economy. levy sufficient property taxes to pay forSecond are those that control for taxes whatever portion of total expenditures thatother than municipal property taxes lev- they choose not to finance from other rev-ied on economic activity generated in the enue sources. According to this model, acity. Third are measures of public ser- higher tax base would lead to a lower taxvices. The variables and reasons for in- rate for any given property tax levy.cluding them are explained in the sub- Assuming a log-linear specification of


Table 2Variable Definitions and Means

(L denotes the natural logarithm)

Mean StandardVariable Definition N=202 Deviation

Endogenous Variables

LPBASE Market value of potential property tax base 2.55 .38per capita in thousands of 1972 dollars.

LPTAX Property tax revenues per capita in 1972 4.18 .79dollars.

Other Variables in the Base Equatio

LPCY Per capita income of city residents in 1972 8.29 .13dollars.

YR82 Dumy variable that taken on the value 1 for .36 .481982 and 0 otherwise.

YR77 Dummy variable that takes the value 1 for .36 .481977 and 0 otherwise.

LPRASE-1 LPBASE lagged one period (five years). 2.46 .35

LSMPOP Population in the city's Standard 7.18 .69Metropolitan Statistical Area - in thousands(1970, 1972, 1977, 1980)

LLAND City land area in square miles. 4.37 .99

KEYCC Dummy variable that takes the value I for .74 .44dominant central cities and 0 otherwise.

TRINC Statutory tax rate for city income, earnings .0043 .0093or payroll tax.

TRSAL Statutory tax rate for city general sales .0064 .0082tax.

LOVTAX Overlying tax burden per capita in 1972 5.90 .77dollars.

LCRIME Total crimes (both property and violent -1.72 .36as reported in Uniform Crime Reports)divided by private sector employees inthe city.

LFIRESER Total per capita state and local spending 2.53 .50on fire protection in the city's state in1972 dollars deflated by FCOST.

LMISCSER Total per capita state and local spending 5.67 .24in the city's state on schools, healthand hospitals, and sewers and sanitationin 1972 dollars deflated by MCOST.


Table 2 - continued

StandardVariable Definition Mean Deviation

Potential Instruments from the Tax Revenue ficiuatio

LTSRI Total ser-vice responsibilities per capita 4.81 .60in 1972 dollars.

AVIRC Dumy variable that takes the value I if .23 .42the city uses a local income, earnings, or

payroll tax and zero otherwise.

AVGSAL Dummy variable that takes the value I if .48 .58

the city uses a general sales tax and zerootherwise.

AVSSAL DmW variable tblat takes the value I if .88 .32the city uses a selective sales tax and

zero otherwise.

ERPROP Export ratio for the property tax. .35 .11

ERINC Export ratio for the local income tax .082 .19(0 if the tax is not used).

ERSAL Export ratio for the general sales tax .077 .12(o if the tax in not used).

4 LMCOST Cost index for miscellaneous services 4.7 .21(Relative to 1972 average)

LPCOST Cost index for police services 4.7 .55(Relative to 1972 average)

LFCOST Cost index for fire services 4.6 .34

(Relative to 1972 average)

LFAID Federal aid per capita in 1972 dollars. 3.64 .99

I. LSTAID State aid per capita in 1972 dollars. 3.58 1.30

the tax base equation, we can eliminate where In denotes natural logarithm, X is

part of the simultaneity by substituting a vector of exogenous variables that in-

the definitional relationship t - PTAX/ fluence the size the tax base, and e is a

PBASE (where t is the city's property tax random error term. After substituting for

rate, PTAX is per capita property tax rev- t and solving for PBASE, the equation be-

enues, and PBASE is the market value of comes

property per capita) for the tax rate in the

equation explaining the size of the base. In PBASE - a/(l + b) + b/(l + b)

Before the substitution we have-In PTAX + c/(l + b) In X (2)

In PBASE - a + b In t + c In X + e, (1) + e/(l + b).


Thus, treating property tax revenue, the value of property per unit of land; morerather than the tax rate, as the explan- land means that the city extends furtheratory variable removes one source of the down the rent and density gradients of thesimultaneity problem, yet still makes it metropolitan area.possible to solve for b, the elasticity of the Expressing the dependent variable inbase with respect to the tax rate. per capita terms rather than per unit of

Even per capita tax revenue, however, land is more natural for the current em-may not be exogenous. A larger base pirical investigation. Hence, the depen-means that the same amount of revenue dent variable in equation 3 must be mul-can be raised with a lower tax rate' re- tiplied by the inverse of the populationducing the pain of raising taxes and density (that is, by LAND/POP, wherethereby increasing the willingness of vot- POP is the population of the city). Oneers to vote for higher taxes. (See later sec- approach at this point would be to add thetions of the paper for further discussion.) logarithm of density (with a predictedThus we treat property tax revenue as an coefficient of - 1) to the right-hand sideendogenous variable in the property tax of equation 3. The difficulty here is thatbase equation. population density is endogenous in that

Variables Derived From an Urban it, too, is determined by the exogenous de-Model. To capture the influence of the terminants of land prices in the city,private economy on a city's property tax namely SMPOP and LAND. Hence, in-base, we begin with a monocentric model cluding population density as an exoge-of an urban economy. This strategy Of nous variable in the estimating equationbuilding on the descriptive implications would not make sense. Instead, we specifyof an urban model dramatically simpli- the reduced form of the model in log-lin-fies what is in fact an enormously com- ear form asplicated and not-very-well-understoodproblem, the behavioral modeling of eco- In PBASE a + cl In SMPOPnomic activity in an urban area.' The fol-lowing log-linear specification incorpo- + C2 In LAND + e (4)rates the essential implications of thestandard urban model for the total value where PBASE is the property tax base perof property (B) per unit of land (LAND) capita and the coefficients el and C2 rep-in the portion of the metropolitan area resent the combined effects of the exoge-designated as the central city: nous variables on the base per unit of land

and on population density. Provided the

InB/LAND co'+c,'InSMPOP exogenous variables affect the populationdensity gradient less strongly than the

+ C2' In LAND + e' (3) rent and business density gradients, westill predict a positive sign for SMPOP and

where e' is a random error. Controlling a negative sign for LAND.'for the amount of land in the city, more The sample cities, and the metropolitanactivity in the metropolitan area as mea- areas in which they are located, vary insured by metropolitan population (SMPOP) how well they fit the simple monocentricleads to a higher value of property in the model. Of most concern is that some of thecity per unit of city land. According to the central cities in the sample are not themonocentric model, this occurs both be- primary centers of economic activity incause larger metropolitan areas have their respective SMSAS. Thus, for exam-higher land prices at the center and also ple, Everett WA, with its 1982 populationbecause the higher price of land induces of 57,000, has far less claim to being themore intensive economic development in center of the Seattle-Everett metropoli-the city in the form of business and res- tan area than does Seattle with 490,000idential structures. residents. To help control for such varia-

More city land, controlling for metro- tions, an additional variable (KEYCC)politan population, is predicted to reduce takes on the value one for those cities that

m - -


dominate their SMSAs and zero other- as a proxy for the burden of overlyingwise. Dominant cities are defined as the taxes. The variable is calculated as thecentral city in SMSAs with one central city average state and local per capita taxand, somewhat arbitrarily, as those that burden in the city's state multiplied by thehave 60 percent or more of either popu- "non-city" share of taxes in the state, andlation or employment in SMSAs with two is based on the following logic. If the citycentral cities or 50 percent or more in itself imposed no taxes, all residents (in-SMSAs with three central cities. Donii- cluding residents of the city) could benant central cities are predicted to have viewed as being subject to the averagelarger tax bases per capita than do non- state and local tax burden in the state. Butdominant central cities because of their since each city's major taxes (property, in-positions as centers of urban economic ac- come, and sales) are modeled separately,tivity. city taxes must be removed from the total

Per capita income of city residents (PCY) state and local burden. This adjustment iscompletes the specification of this part of accomplished by using statewide data tothe equation. This variable is expected to subtract the average per capita taxes col-have a positive sign because higher in- lected by all those jurisdictions in the statecome increases the demand for housing.' that perform the same functions as the city

Nonproperty Tax Variables. Three ad- in question. Thus, for example, in con-ditional tax variables control for the other structing the overlying tax variable for ataxes imposed on economic activity in city such as San Francisco which has re-central cities. Alternative taxes levied by sponsibility for county functions as wellthe city itself are represented in the tax as municipal functions and hence has nobase equation by statutory city tax rates overlying county, care is taken to makefor income taxes (TRINC) and for general sure that the calculated overlying burdensales taxes (TRSAL), with a rate of zero does not include county taxes. 7 The ex-if the city does not use the tax. Close to pected sign of this variable is negative;

L half the sample cities make no use of lo- higher overlying tax burdens are likely tocal general sales taxes and only 18 of the depress economic activity in the city.86 sample cities had some form of a city Measures of Public Services. Public

6income or payroll tax in 1982. Negative services are an important component ofsigns are expected on both tax rate vari- any study of the effects of state or localables. A higher income tax rate lowers the taxes. Since public services are valued andnet-of-tax income of city residents, and taxes are used to pay for them, failure tomay induce firms to pay higher wages control for services could lead to incorrectthan they otherwise would, thereby re- estimates of the effects of taxes alone. Inducing net profits and discouraging in- principle, we need to control for all ser-vestment in the city. These effects will re- vices available to city firms and house-duce the demand for property in the city holds regardless of whether the servicesexcept in the unlikely event that the in- are provided by the city government itselfcentive for firms to substitute away from or by some overlying government such aslabor in favor of land and capital is large a state or county government or an in-enough to offset the output effect. A higher dependent school district.sales tax rate may reduce the demand for Public expenditures typically serve astaxed goods, and, similarly, lead to less the standard measure of public servicesinvestment in the city. in studies of tax capitalization (e.g. Oates,

The tremendous complexity and varia- 1969), but, as shown by Rosen and Ful-tion in the division of taxing responsibil- lerton (1977), output measures such asities among city and non-city govern- educational test scores are far superior.ments across states makes it impossible Hence, we have constructed three publicto control separately for each of the non- service measures that are intended tocity taxes imposed on city economic activ- measure public sector outputs rather thanity. Instead, we constructed a single vari- simply public inputs. The first is theable (OVTAX), based on statewide data, number of crimes in the city expressed as


a fraction of the city's total private sector where k is the proportional adjustmentemployment (CRIME). Vocal public con- occurring in one (five-year) period. Tak-cern about crime makes it plausible that ing natural logarithms yields:the decision to invest in a particular citymight be influenced by the perceived risk In PBASET \In PBASET*of being assaulted or robbed. Higher crime + (1 )L)In PBASE, (6)rates per worker in the city are

predicted

to lead to lower property values. The sec- Hence, the model now includes all the de-ond and third variables (FIRESER and terminants of PBASE* described aboveMISCER) are proxies for fire protection plus the lagged base, PBASE, 1. Recall-and miscellaneous services (consisting of ing the formulation of equation (2), thislocal schools, sewers, sanitation, and yields the complete modelhealth and hospitals) constructed from percapita state and local expenditure data by In PBASE k[a/(l + b) + b/(l + b)state deflated by the estimated costs ofproviding the respective service or sets of -In PTAX + c,/(l + b) In SMPOPservices in each city. The cost indexes are

+ C2/(l + b) In LAND + C3/(l + b)designed to measure the effects of city-specific environmental and demographic -KEYCC + C4/(l + b) In PCYfactors on the costs of providing a givenpackage of public services and are de- + C5/(l + b) TRINC + C6/(l + b)

scribed in more detail below. To the ex- TRSAL + C7/(l + b) In OVTAXtent that the indexes correctly measurethe costs per unit of final output, deflat- + C8/(l + b) In CRIME

ing expenditures by them should lead to + cg/(l + b) In FIRESER + clo/(l + b)reasonable proxies for the level of ser-vices actually available to city residents. -In MISCSER + cll/(l + b) YR82

We expect both variables to enter the base + C12/(l + b) YR771equation with positive signs since higherservice levels should increase the size of + (1 - X) In PBASET 1 + e. (7)a city's property tax base, ceteris paribus.

Lagged Base and Year Dummies. The Determinants of Property Tax Revenuesfinal three variables are two-year um-mies (YR82, YR77) and the dependent The key to identifying the crucial coef-variable lagged one (five-year) period ficient of the endogenous variable PTAX(PBASE-1). The year dummies are needed in equation 7 is that there be good iden-in the pooled regression to control for tifying variables, that is, variables thatcyclical trends in the aggregate economy would clearly belong in a revenue equa-and for secular trends such as the fall in tion but not in the base equation. Onetransportation costs that tend to decen- distinguishing characteristic of this studytralize business activity. The lagged de- is the availability of such variables. ThesePendent variable accounts for the fact that variables are derived from the followingland rents and the intensity of economic identity, which reflects the balanced bud-activity that determine city property val- get requirement that local revenues equalues may not adjust immediately to chang- local expenditures:ing conditions over time.

More formally, if PBASE,* is the long-run equilibrium tax base per capita in pe- PTA), - E, ORriod t and PBASE, and PBASET I are the (8)actual bases per capita in periods t and t (PTAX/TTAX)- 1, then in multiplicative form

where PTAX is property tax revenues,PBASE,/PBASEt 1 (5) Y,,' , E, is all city operating expenditures

(PBASEt*/PBASEt 1)', summed over I individual spending cate-


gories, OR is nontax exogenous revenues tax revenue equation and can serve tosuch as lump sum intergovernmental aid, identify the coefficient of the property taxand TTAX is total revenues from all local variable in the base equation, In the es-tax sources. To model this relationship, timated equations, per capita incomewe specify first the demand for public ser- (PCY) serves as a reasonable proxy for thevices and hence total revenue require- relevant income measure and is expectedments and second how cities choose among to have a positive effect on taxes. We con-tax sources. trol for the decisive voter's tax share with

Demand for Local Public Expendi- a measure of property tax exporting (thetures. The literature on local public ex- greater the exporting, the lower the bur-penditures is well developed and needs den on city residents), which is also ex-only brief review here.' The typical start- pected to influence the choice among rev-ing point is that the quantity of public enue sources, described in the nextservices demanded is a function of resi- subsection of the paper.11dent income, per unit costs, and tax prices. As specified so far, desired expendi-Desired expenditure on the ith expenditures are not explicitly dependent on theture category (E,) is the product of costs size of the property tax base. This pri-and quantity demanded. Hence, marily reflects the simplicity of the de-

cisive voter model, particularly in its me-E, C,Q, f(Y, C,, TS), (9) dian voter form, rather than economic

reality. In more realistic models, the taxwhere Y is the income of the decisive voter, base or its components might well influ-C, is per unit cost of the ith expenditure ence demand. The value of residentialcategory, and TS is the decisive voter's property, for example, might better re-share of the local tax burden. flect residents' permanent income and,

Appropriately measured cost variables consequently, their willingness to pay fornot only would reflect the costs of inputs, public services than does current income.but, following the logic of Bradford, Malt Or higher-valued property (whether res-and Oates (1969), would also incorporate idential or business) may require greater

4 the amount of intermediate goods such as services in the form of protection frompolice patrols that are needed to produce crime. Hence, using the simple modela given level of output such as protection presented in equation 9 to justify exclud-from crime. The idea here is that envi- ing the size of the property tax base as anronmental conditions in the city that are explanatory variable would be a mistake.outside of the control of city officials, such Moreover, as discussed below, the reve-as city density and the incidence of pov- nue mix component of the model also imerty, may affect the costs to the city of plies that the property tax base is a de-providing a given level of the output, such terminant of property tax revenue.as police protection, ultimately valued by The analysis to this point argues for in-city residents. To reflect such cost vari- cluding in the revenue equation standardations, we have included three cost in- determinants of the demand for spendingdices for miscellaneous public services on individual public services: resident in-(MCOST), police protection services come, a tax exporting measure, service-(PCOST), and fire protection services specific measures of costs, and the prop-(FCOST). The cost indices are derived from erty tax base. One additional variable re-a regression model that estimates the av- lated to expenditures dominates all theerage impact on city expenditures of each others in terms of its relevance to thiscost factor, controlling for other determi- study, namely a measure of the servicesnants of city spending such as resident for which each city government in theincome and preferences." sample is responsible. Given the cross

Given the purpose of this study, we need sectional nature of the data set, variationnot be precise about the identity of the in service responsibilities is likely to be adecisive voter. The goal is simply to use primary determinant of the variation inthe essentials of a basic expenditure model total expenditure. Stated differently, theto determine which variables belong in the number of expenditure categories over


which the summation applies in equation variables indicating the legal availability8 varies substantially among cities. of alternative local taxes, and ER is a

The variable measuring the total ser- vector of export ratios for alternativevice responsibilities assigned to each city taxes."(TSRI) indicates the per capita spending The larger is the property tax base pernet of user charges that would be re- person, the easier it should be for the cityquired in each city to achieve national av- to raise revenue through the property tax.erage per capita state-local spending on A larger tax base allows the city to raiseeach of 17 designated services, given the a given amount of revenue with a lowerparticular allocation of spending respon- tax rate and thereby to avoid the poten-sibilities in each city's state. Constructed tially distorting effects and heavy taxfrom statewide rather than city-specific burdens of high tax rates.expenditure data, the measure avoids the In most cases, authorization of alter-potential problem of attributing high ser- native revenue options is a state, rathervice responsibilities to cities which choose than a city, decision. Yet, authorization toto provide high service levels. The mea- use nonproperty taxes plays a key role insure varies substantially across cities. The the city tax decision given the stringentmost obvious variation relates to services restrictions often placed on city revenue-such as elementary and secondary edu- raising authority. As already noted, onlycation and municipal hospitals for which 18 of the 86 sample cities are currentlya city has either complete spending re- allowed to use some form of local incomesponsibility or none. Some cities are also or payroll tax, and 47 cities a general salesresponsible for services provided by coun- tax. A higher proportion of cities are per-ties in other states. Variation across states mitted to impose some form of selectivein the role of state government also ac- sales tax. Availability of alternative taxcounts for a substantial portion of the sources is likely to decrease a city's reli-variation in service responsibilities across ance on the local property tax, Availabil-states; the greater the state role in wel- ity is indicated with dummy variables:fare, health, and corrections, the lower, in AVINC refers to local income, earnings,

12general, is the city role. or payroll taxes; AVGSAL refers to theChoice among Revenue Sources. Sub- general sales tax; AVSSAL refers to se-

tracting nontax exogenous revenues (OR lective sales taxes.in equation 8) from total desired expen- Not all taxing instruments impose equalditures yields the amount of revenue that burdens on resident voters. A portion ofmust be raised from local tax sources. the burden of a local sales tax, for ex-Nontax revenue is primarily intergovern- ample, might be shifted onto nonresidentmental aid, but in principle, only the aid tourists and commuters in the form ofthat is truly exogenous. As measured, higher prices. Or a payroll tax may fallhowever, some of the per capita federal partially on nonresident commuters. Sim-aid (FAID) and per capita state aid ilarly, part of the property tax burden may(STAID) may be matching aid. ultimately be bome by nonresidents in the

How the remaining revenue require- form of lower profits, higher prices, orments are allocated among local tax lower wages. The export ratios (ER), de-sources has received much less attention fined as the proportion of the tax burdenfrom economists than has the expendi- associated with each of the major localture decision. 13 Consistent with the deci- taxes that can be shifted to nonresidents,sive voter approach, one might hypoth- are included in the equation to account foresize the following model of the city this burden shifting. The hypothesis is thatdecision regarding how much to rely on local voters choose taxes in such a way aslocal property taxes: to n-dnimize burdens on themselves. Hence,

a higher export ratio for the property taxPTAX/TTAX = F(PBASE, AV, ER) (10) (ERPROP) is expected to lead to greater

reliance on property taxes, but higher ex-where PBASE, as before is the per capita port ratios for sales (ERSAL) or incomeproperty tax base, AV is a vector of dummy taxes (ERINC) are expected to reduce city


reliance on property taxes. coefficient of the endogenous revenueThe calculation of each export ratio re- variable in the base equation. One must

lies first on assumptions about which be careful, however, to assure that eachgroups -consumers, workers, or owners of the proposed instruments is truly ex-of property-bear the burden of each tax ogenous in the two-fold sense that 1) it doesand second on estimates of the proportion not belong in the equation as an explan-of each group that lives outside the city. atary variable and 2) conditional on 1) itThe incidence assumptions are straight- is uncorrelated with the error term.forward for the income and sales taxes; Following Lugar and Stahl's (1986) ex-local income taxes are assumed to be borne position of Hausman (1978), we performfully by workers in the form of lower wages a set of simple t-tests to determine whetherand local sales taxes by consumers in the each of the variables excluded from theform of higher prices. The incidence as- base equation is truly exogenous in thesumptions used for the property tax are sense just stated. For example, consider amuch more complex and are spelled out variable such as the index of service re-elsewhere.'5 Calculated export ratios for sponsibilities. Based on our conceptualall three taxes vary substantially across framework, this variable belongs in thecities because of differences in how cities revenue equation but not in the tax basedefine their tax bases, the diversity of city equation. The division of service respon-roles in metropolitan areas, and varia- sibilities among levels of governmenttions in the mix of shoppers, job-holders, should have no effect on city property val-or property types across cities. ues, controlling for public service levels,

In summary, the property tax revenue aside from its effect through property taxequation incorporates the determinants of liabilities. By adding it to the base equa-total revenue requirements, including tion and testing the hypothesis that itsresident income, service-specific cost in- coefficient is zero, we are testing the jointdices, the property tax base, and a mea- null hypothesis required for exogeneity.sure of city government service respon- Through a process of trial and error, wesibilities, as well as the determinants of obtain a set of instrumental variablesrevenue mix, such as exogenous intergov- whose members all pass the exogeneityernmental aid and the availability and test.exportability of nonproperty tax sources: These exogenity tests lead us to reject

as instruments two of the excluded vari-PTAX F(PCY, MCOST, PCOST, ables: the export ratio for property tax

burdens (ERPROP) and the cost index forFCOST, PBASE, TSRI, FAID, STAID, police services (PCOST). Failure to passAVINC, AVGSAL, AVSSAL, ERINC, the exogeneity test, however, need not

man that the variables belong in the base

eERSAL,ERPROP,YR82,YR77). (11) equation as explanatory variables; con-ditional on a true model that excludes

As in the base equation, the variables are them, the variables may simply be cor-expressed in logarithmic form for esti- related with the error term. This lattermation.'6 interpretation applies to both variables.

Variation in neither variable can be saidto cause variation in the tax base. In-III. Estimation and Results stead, the causation probably goes in the

Careful specification of the revenue other direction; cities with larger per cap-equation provides a rich set of potential ita tax bases simply tend to have greaterinstruments for our 2SLS estimation of potential to export property tax burdensthe tax base equation. In particular, the and higher costs of providing police ser-12 variables that appear as control vari vices than those with smaller tax bases."ables in the tax revenue, but not in the An additional econometric problem couldbase equation, are logical candidates for arise from the presence in the base equa-instruments than can serve to identify the tion of the lagged dependent variable. As

No. 41 CITY TAXES AND PROPERTY TAX REVENUES515

is well known, the coefficient of the lagged ita propertyvaillo8 o,,Id be 16 Percent

variable will be inconsistently estimated lower in the hi,"Iier tax city- long-runif the equation error in time t is corre- A more direct estimate of the,Uation with-lated with the error in time t - 1. For- response emerges@from the eq le. The@pendent variabtunately, however, inconsistency is not a out the lagged de ne per-problem here; using a Hausman test of the coefficient of -.. 1,4 implies that a 0,y tax rateform elaborated by Lugar and StaM (1986), cent difference in a city's PrOPe". the size

,ent difference inwe reject the null hypothesis that the leads to a. 12 pere Ong-run equilibrium as-lagged base is correlated with the error of its base. The makesterm.18 @;umption iinplicit it., this equation less ap-

Table 3 presents the coefficients esti- this somewhat-srrialler estimate the par-mated using two-stage least squares with pealing than the estimate fromdata for 1972, 1977, and 1982 for two tial adjustment Diodel, ariables Pro-equations: the partial adjustment equa- The three additional tax.v .ting -new insights abouttion that corresponds to equation 7 and, vide some intere8 rnative revenue sourcesfor comparison, an equation that excludes the effects Of a)te tax base.the lagged dependent variable. Columns on the size of a (,ityl, property axes bothCity income taxes and non-city t2 and 4 represent the structural coeffi- @lcity's Property .

tax base,cients of the model derived from the es- reduce the size of ' y salestimated parameters of each equation, but, somewhat surprisingly, elt estaxesThese can be interpreted as the long-run taxes do not. This finding about Sal riianyisiness owners,responses of the logarithm of the tax base may not surprise bi ges overprefer sales tato each variable. The following discussion of whom seerl to.al taxes on the groundsfocuses primarily on the partial adjust- other state and loc tax burden to cOn-ment model, which we prefer because it that they can shift the prices. Toallows for the dynamics of property mar. sumers in the fol,nl of higher.

ket adjustments over time. economists, however, the finding is stillrising; by discouraging

Overall, the equations perform well. The somewhat surp higherconsumption of taxed items, thecoefficient of primary interest, that of the . I level ofproperty tax variable, is -0.080 and sig- prices should reduce the city'nificantly different from zero. Ignoring for economic activity. by our es-a moment the lagged adjustment term and Once it has been TnultiPliedthe coefficient of therecalling from equation 2 that this esti- timate of ((I + b)/\), LOVTAX)mated coefficient equals (b/(l + b)), we overlying tax burden variable (

ed directly as a long-runestimate the short-run elasticity of a city's can be interpret fficient ofper capita property tax base with respect elasticity. In contrast, the coeto its property tax rate to be approxi- the income tax rate variable (TRINC),

not in logqrithrn form, niust bemately -0.075, where the short run is a which is s mean value in additionperiod of 5 years. multiplied by it nvert it to an elasticity.

The long-run response can be calcu- to ((I + b)/X) to cOl followinglated by combining this estimate with our These calculations produce the taxes andestimate of \, the partial adjustment pa- elasticities: -0-10

for overlyingmagni-

rameter. The estimated coefficient on the -.065 for income taxes, Theseble to, but, not surprlr,-

lagged base variable, 0.54, implies that k tudes are cOrnPara the property tax rateis 0.46 and that slightly less than half of ingly, smaller than e Policy clgnificancethe adjustment to long-run equilibrium elasticity of -.15, Th agnitudes@sof these Ynoccurs during a five-year period. This ad- of the relative size tion.justment speed means that our best esti- is discussed in the coricludilig sec es, thee tax variab)tnate of the long-run elasticity of the tax in contrast to th iables isbase to the property tax rate is -0.15 perfomance of the service var ures of(equals -0.075/0.46). Thus, if the prop- somewhat disappointing. The. meas the,Ieous services entererty tax rate in one city were twice that crime and iniseellai n with the predictedof another city and this difference per- preferred equatiosisted over time, we predict that per cap- signs, but with relatively large ,taiaard


Table 3TAX BASE EQUATION

Dependent Variable: LPBASE

Partial Adjustment Full Adjustment(Equation 7) (Do la2zed dependent variable)

Independent Estimated Structural Estimated StructuralVariables Coefficients Parameters Coefficients Parameters

(1) (2) (3) (4)

LPTAX -.081* -.15 -.14* -.12(.032) (.040)

LPCY .95* 1.8 1.5* 1.3(.16) (.18)

LSMPOP .061* .11 .17* .15(.033) (.040)

LLAND -.026 -.048 -.053* -.046(.022) (.026)

KEYCC .015 .028 .12* .10(.047) (.057)

TRINC -8.3* -15 -16* -14(2.7) (3.3)

TPSAL *61 1*1 1,1 1*1(2.4) (2.8)

LDVTAX -.054 -.10 -.098* -.086(.034) (.043)

LCRIME -.041 -.076 -.10 -.088(.046) (.070)

LFIRESER -.0024 -.0045 .061 .054(.046) (.058)

LMISCSER .097 .18 -.11 -.10(.097) (.12)

YR82 .064 .12 .21* .18(.043) (.051)

YR77 .025 .046 .11* .099(.041) (.049)

LPBASE-1 .54* - -(.056)

Constant -7.0* -9.5*(1.4) (1.6)


Table 3 - continued

Partial Adjustment Full Adjustment(Eauation 7) (No latted dependent YLtLLbLe)

independent Estimated Structural Estimated StructuralVariables Coefficients Parameters Coefficients Parameters

(2) (3) (4)

Adjusted R2 .67 .48

Standard Error .22 .29

Sample Size 202 225

Coefficient significantly different from zero at 5% level, one-tailed test.

Notes: Pooled time series and cross-section using data for 1972, 1977, and

1982.Estimated with two-stage least squares; LPTAX treated as endogenous.LPBASE-1 is LPBASE lagged one period (five years); missing data for LP&UE in1967 account for the difference in sample size between the two equations.

lnztruments: LTSRI, LMCOST, LYCOST, AVINC, AVGSAL, AVSSAL, ERINC, ERSAL,LFKID, LSTAID, and exotenous explanatory variables in the equation.Asymptotic standard errors in parentheses below coefficients.

errors. This provides at best weak support lagged base term increases the indepen-

for the conclusion that higher crime rates dent variation in these two variables and

reduce property values and that higher yields more precise estimates Of their ef-

miscellaneous services, such as education fects.

and sanitation, increase property values We interpret the positive sign Of met-

in the city. Our proxy for fire protection ropolitan population and the negative s-ignservices enters with an unexpected neg- on city land area to mean that the im-

ative sign but its coefficient is virtually pacts of these variables on rent and busi-

zero. ness density gradients are greater than

The variables derived from the urban on population gradients. For example, the

monocentric model all have the expected positive coefficient of the metropolitan

signs in both equations. As predicted, population variable suggests not only that

higher metropolitan population, higher the market value of property in the cen-income of city residents, and status as a tral city per unit of land is higher in largerdominant central city all produce higher metropolitan areas, but also that the

per capita property values in the central higher value per unit of land is not offset

city while larger land area decreases per by a sufficiently higher population den-capita values. However, the presence in sity in the central city to produce a lower

the preferred equation of the city's tax base value of property per city resident."

in an earlier year makes it difficult to ob-

tain precise estimates for KEYCC andIV. Interpretation: The Effects of

LLAND. This follows because a city's sta-Taxes on City Property Tax Bases

tus as a dominant central city within amulti-city SMSA typically does not change Our best estimate of the long-run cla-over time. In addition, its land area does ticity of a city's property tax base with re-not change much except in cities engaged spect to its tax rate is 0.15. This esti-in extensive annexation. Omitting the mated response appears to be well


identified in a statistical sense thanks to cities, with more going to large centralthe nature of the data base and the at- cities than to their wealthier suburbs.tention devoted to the specification of the With the recent dramatic decline in directrevenue equation as well as to the base federal aid to cities at the same time thatequation. costly-to-serve households such as the poor

This -0.15 would be easy to interpret tend to be increasingly concentrated inif the implicit experiment underlying our central cities, such cities may need to in-approach were unambiguous. The esti- crease local tax rates more relative to theirmating equation, however, focuses on suburbs than they have in the past (Laddcomparisons among large central cities, and Yinger (1989)). If this outcome oc-setting aside intrametropolitan differ- curs, city tax bases may decline more thanences in tax rates and service levels .20 If predicted in this study because investorsone were willing to assume that city and would find cities even less attractive thansuburban tax rates and service levels move in the past relative to their surroundingtogether, the coefficient estimate would areas.imply that a 10 percent increase in a cen- How reasonable is the preferred esti-tral city's property tax rate, accompanied mate? In the absence of comparable stud-by a proportionate change in suburban tax ies, one approach is to evaluate it withrates, would decrease the central city's tax reference to the theory of capitalization.base by 1.5 percent. Although the level of Consider, for example, what would hap-tax rates and services in central cities pen if property tax rates were fully cap-typically differ significantly from those of italized into property values. By way ofother cities within the metropolitan area, illustration, consider a parcel of propertythe assumption that central city and sub- worth $100,000 and subject to a 0.8 per-urban tax rates and service levels move cent tax rate, the average city's rate in thetogether over time is plausible because sample. Then consider the change in valuecities within a metropolitan area are sub- predicted to occur with a 10 percent in-ject to similar state fiscal institutions and crease in the tax rate, to 0.88 percent. As-economic pressures. For example, a re- suming the resulting $80 tax increase wereduction in state aid for local governments expected to continue indefinitely, fullor a turndown in the state economy are capitalization would lead to a 1.5 percentlikely to put upward pressure on local tax reduction in the value of the property (asrates both in central cities and in their our equation predicts) provided the rele-suburbs. vant discount rate were 5.3 percent (since

In reality, however, central city and $80/0.053 - $1500 - 1.5 percent ofsuburban tax rates exhibit some indepen- $100,000). Thus, at a discount rate of 5.3dent variation. This observation implies percent, the estimated long-run elasticitythat the - 0. 15 elasticity indicates the av- of -0.15 would be fully consistent with 100erage long-run effect of a change in the percent capitalization. For higher dis-central city tax rate plus the effects of count rates, the estimated elasticity in-corresponding changes in suburban tax dicates a larger impact than would be im-rates and service levels that occurred plied by capitalization alone and for lowerduring the sample period. Provided sub- rates, it indicates a smaller impact.urban tax and service levels continue to Even if one accepted 5.3 percent as themove in the same way as they have done correct discount rate, one should not in-in the past relative to changes in the cen- terpret our results as evidence that prop-tral city tax rate, the reported elasticity erty taxes are fully capitalized into val-accurately predicts the long-run effects of ues. In contrast to previous studies thata change in central city tax rates. explicitly measure capitalization (e.g.

The reported elasticity could, however, Oates, 1969), we purposely do not controlunderstate the effects of future increases for the amount of capital in the city. Hencein central city tax rates. The elasticity was the estimated response of property valuesestimated for a period of time character- represents the combined effects of someized by rapid increases in federal aid to capitalization (price change) and some


physical disinvestment in the city (change cal sales taxes which appear to have vir-in the quantity of capital). That is, part tually no effect on the property tax base.of the reduction in the base probably rep- Table 4 compares the effects of the var-resents the decision of some potential res- ious taxes, using the coefficients from theidents (households and firms) not to move partial adjustment equation. The entriesinto the city, the decision of some firms in the first column show the impact on theto let their city property depreciate as they logarithm of the property tax base of a 10invest elsewhere, and the decision of some percent increase in a particular tax ratecity residents to disinvest in their resi- or, equivalently for nonproperty taxes, indential structures by reducing mainte- tax revenues. As changes in logarithms,nance. This decrease in investment oc- the entries (multiplied by 100) can be in-curs simultaneously with a fall in the price terpreted as percentage changes in theof city land and stops once the after-tax base. Thus, a 10 percent increase in therate of return to investment is again property tax rate is predicted to reduce theequalized across jurisdictions. base by 1.5 percent, while a 10 percent in-

As noted earlier, this study provides new crease in the local income tax rate is pre-evidence that taxes other than city prop- dicted to reduce the base by only 0.7 per-erty taxes also affect the size of a city's cent. The estimated impact of a 10 percentproperty tax base. The one exception is lo- increase in overlying taxes is a 1.0 per-

Table 4Impacts on the Property Tax Base

(By Type of Tax)

10 PercentIncrease in Tax $10 Per Capitaa

Rate or Tax Revenues Increase in Taxes

Property Tax -.015 -.014 or -.020b

City Income Tax -.0065 -.025

City Sales Tax 0 0

Taxes Imposedby OverlyingGovernments -.010 -.0026

Notes: Entries are predicted impacts on the natural logarithm of thePer capita property tax base. When multiplied by 100 they can beinterpreted as percentage changes in the tax base. Calculated fromestimated coefficients from the partial adjustment equation shown inTable 3.

al972 dollars.bthe first entry was evaluated at the average tax base and tax ratein the sample. The second entry was evaluated for average taxrevenues.

520 NATIONAL TAX JOURNAL (Vol. XLI

cent decline in the base, halfway between finance them by higher noncity taxes, ourthe impacts of local income and property estimates suggest that the city's propertytaxes. These findings have two clear im- tax base will increase. Reducing propertyplications. First, a 10 percent increase in taxes by $10 per capita would increase thea city's property tax rate will produce only tax base by 1.4 to 2.0 percent, an increasean 8.5 percent increase in property tax that more than offsets the much smallerrevenues on average in the long run. Sec- decrease of 0.3 percent associated with theond, increases in city income tax rates or $10 increase in the overlying tax burden

21in taxes levied by overlying jurisdictions on city residents and firms.such as state and county governments willreduce the property tax revenues col- V. Conclusionlected with a given property tax rate.

The second column of Table 4 focuses Economic theory predicts unambigu-attention on equal-yield revenue changes ously that, controlling for service levels,by showing the impact on the (logarithm an increase in local property tax burdensof the) property tax base of a $10 increase reduces the size of the local property taxin per capita tax revenues from each base. This outcome occurs through eithersource. The two entries for the property or both of two mechanisms: 1) the tax re-tax reflect evaluation at different aver- duces the level of economic activity or 2)ages. The 1.4 percent decline was evalu- it is capitalized into lower land prices.ated at the average property tax rate and Despite this clear prediction, economiststax base in the sample, while the 2.0 per- know little about the magnitude of thecent decline was evaluated at the average impact. This situation reflects the fact thatper capita tax revenues in the sample. previous researchers have tended to focusSurprisingly, the calculations indicate that either on the location and investment de-a $10 increase in local income taxes re- cisions of firms and households or alter-duces the size of the property tax base on natively, controlling for the amount ofaverage by more than does a comparable capital in the city, on the extent to whichper capita increase in property taxes. This taxes are capitalized into property or landfinding is easily reconcilable with the ap- values, but never on both mechanisms si-parently contradictory results in column multaneously.1: a $10 per capita increase in revenue Theoretical predictions about the ef-from income taxes requires a much larger fects of local income taxes on propertyabsolute and percentage increase in in- values are somewhat more ambiguous be-come tax rates than a $10 increase in cause a higher tax on labor income couldproperty tax revenues does in property tax induce firms to substitute away from la-rates. bor in favor of land and capital. In prin

These results can be used to determine ciple, this substitution effect could offsetthe net effect on the property tax base of the output effect, but the standard pre-substituting revenues from one tax for sumption is that, like higher propertyanother. Assuming constant total tax rev- taxes, higher local income taxes reduce theenue and after allowing time for all ad- size of the local property tax base. Thisjustments, the larger response for income study contributes to our understanding oftaxes means that shifting away from these behavioral responses by specifyingproperty taxes toward income taxes re- and estimating a simultaneous model thatduces the size of the property tax base. permits us to determine the total impactShifting away from property taxes in fa- on the size of the local tax base of variousvor of local sales taxes, in contrast, in- forms of local taxes, where the total im-creases the size of the local property tax pact includes both the investment andbase. price effects of tax changes.

Alternatively if local officials choose to We conclude that taxes affect localreduce property taxes by inducing higher property values more than is typicallylevels of government to take over some of found in previous studies. In particular,the city's 3ervice responsibilities and to we find that a 10 percent increase in a


city's property tax rate decreases the cit- industrial real property values in those cities fullyy's tax base by about 0.75 percent during taxing personal property.a five-year period and by 1.5 percent in 2This means that the equation modeling tax reve

nues (discussed below) is less appropriate for Bostonthe long run. In addition, local income and the California cities in 1982 than for other cities.taxes and taxes levied by overlying juris- But deleting these observations does not affect thedictions (such as county and state govem- coefficient estimates.ments) also have negative impacts on the 'See Bradbury, Downs, and Small (1982) for an at-

tempt to model the simultaneous determination of citycity's property tax base. In contrast, local population, employment, and per capita income.sales taxes appear to have virtually no 'More precisely, we haveimpact. One striking finding is that for agiven amount of revenue, local income In (B/LAND) el' + cl'ln SMPOPtaxes have more deleterious effects than + C2' In LAND+ e' anddo local property taxes on the size of the In (POP/LAND) - 4 + d, In SMPOPlocal property tax base. Thus, not only + d2 InLAND+ u,property taxes but also other forms of lo-cal taxes have adverse effects on the local so thatproperty tax base. These effects are im-portant because lower property tax bases ln(B/POP) (c4' do)+(cl'-d,)InSMPOPrequire higher local property tax rates to + (C2' d2) In LAND+ (e' - u)finance a given package of public ser- Hence, cl should be interpreted as el' - d, andC2 asvices. e2' - d2-

'This prediction should not be confused vnth themore ambiguous effects of income that typically emergefrom an urban model. According to that model, house-

ENDNOTES hold income of the residents of a metropolitan areaexerts two opposing forces on the slope of the pricegradient On the one hand, higher income increases

***The authors thank Ilhong Cho for research as the demand for space, which tends to flatten the gra-sistance and John Yinger, Lynn Browne, and anon- dient. On the other, higher income increases the valueymous referees for advice and conunents. They are also of travel time and thereby raises the cost of trans-grateful for helpful suggestions from the participants portation which leads to steeper gradients. With theat the NBER Conference on State and Local Govern relevant variable specified as the per capita incomement Finance (Cambridge, MA, December 12-13, of city residents alone, the effects of income on city1986) at which an earlier version of this paper was vs. suburban location are already accounted for andpresented The research reported here is part of the the issue becomes solely one of the demand for hous-NBEWs research project in State and Local Finance ing services.Any opinions expressed are those of the authors and 6 Statutory rates for cities imposing a sales tax werenot those of the National Bureau of Economic Re- pieced together from a variety of sources. The foursearch. main sources are the Commerce Clearing House, State

'The Census of Governments reports assessed val- Tax Guide; Due and Mikesell, (1983); Due (1971) andues of locally-assesscd taxable and exempt real and Advisory Commission on Intergovernmental Rela-personal property and state-assessed property, as well tions, Significant Features of Fiscal Federalism, var-as the percentage distribution of real property values ious years.among residential (total and single-family housing), 'The estimated overlying tax burden for all otheracreage, vacant platted, conunercial and industrial and California cities in the sample exceeds that for Sanother uses These value data were combined with as- Francisco, because they all have overlying countiessessment/sales ratios (reported by the Census of Gov- See Ladd and Yinger (1989), ch. 7, for a complete dis-ernmentr for all real property, single-family residen- cussion of the complexities of calculating measures oftial property, and in some cases, vacant property), overlying tax burdens in a slightly different context.Information on assessment practices for personal and 'Previous studies have shown that higher crime ratesstate-assessed property from the Census of Govern- reduce housing values. See, for example, Rizzo (1979)nients, and economic data from other sources to es- and Gray and Joelson (1979). We have normalized thetimate the market value of property in eight classes number of crimes (both property and violent crimesfor each city. These estimates are described in more as reported in The Uniform Crime Reports) by emdetail in the appendix to Bradbury and Ladd (1985). ployees rather than by city population because crimesBecause personal property is partially or completely ratei3 per resident appear to be strongly correlated withexempt from property taxation in many states, the re the amount of a city's nonresidential activity, eape-Ported "exempt" values for personalty were generally cially commercial activity. Hence, normalizing byincomplete Personal property values were imputed resident population rather than by employees wouldfor most cities based on their commercial and indus- lead to a severe reverse causation problem; higher taxtrial real property values and the relationship be bases per capita would be associated with higher crimetween personal property values and commercial and rates per capita.


'Fhe literature on local public goods contains two scribed later and thus are appropriately excluded fromapproaches to the modeling of the local decision to tax the revenue equation.and spend. First and most common is the positive, 15See appendix to Bradbury and Ladd (1985) for apredictive theory of the decision-rfiaking process that full description of how the export ratios were calcu-starts with the decisive voter's demand for local pub lated. As noted in the text, the analysis is most comlic goods. According to this approach, local public goods plicated for the property tax. In brief, the share of theare viewed as providing consumption benefits to con property tax falling on land is assumed borne bysumer-voters in the same manner as private goods landowners. Capital is assumed to be mobile acrossSee, for example, Bergstrom and Goodman (1973), and cities, so the portion of the tax with an initial ineiBorcherding and Deacon (1972). The alternative ap- dence on capital is assumed to be borne partially byproach emphasizes the investment aspect of public consumers in those markets where taxed producersgoo& provided by local governments According to this dominate, and partially by labor and land, approxiview, voters choose an expenditure and tax package mately in line with factor shares. The share of the taxto maximize the value of their property independent that is exported then further depends on the fractionsof their demand for public goods. See, for example, of consumers, landowners, and workers who live in.Sonstelie and Portney (1978), Brueckner (1979), and side vs. outside the city.Brueckner (1982). Implicit here is the view that ra- '6Estimates of the revenue equation are not re-tional asset-owning voters should behave this way ported in this paper. Coefficient estimates are avail-since consumption interests "can be satisfied through able from the authors upon request.migration, if necessary" (Sonstelie and Portney, p. 271). 17Consistent with our model, the Hausman testsTwo considerations lead us to use the standard de show that the state and federal aid variables are ap-mand approach. First, we question the applicability propriately excluded from the base equation. An al-of the assumptions of the investment approach to het temative view that intergovernmental aid belongs inerogeneous voters in the large central cities of our the base equation makes sense only if tax burdenssample; to satisfy their preferences for public ser- are not included in the equation. The logic would bevices, many residents of central cities would face high that, after controlling for service levels and their costscosts of moving to other jurisdictions. Furthermore, of provision, higher intergovernmental transfers leadmany of the voters in central cities are not property to lower local tax burdens and that these lower taxowners Second, the demand approach forms the basis burdens are capitalized into a higher valued tax base.for many previous empirical studies of local public In our model, the tax variables fully capture the in-spending that yield reasonable results. We know of direct effects on the tax base of federal and state aidfew studies that use the investment approach as the '8'Fhe test works as follows: first the residuals arebasis for predicting the behavior of local voters (an calculated from an auxiliary regression of the laggedexception is Brueckner (1982)). We emphasize, how base on the set of exogenous variables in the baseever, that this paper does not explicitly test one ap- equation, not including the lagged base, plus otherproach over the other. variables hypothesized to be exogenous. Second, one

"For a complete description of the basic method tests the null hypothesis that the coefficient of theology, see Ladd and Yinger (1989) A similar meth- residuals is zero when they are included as an ex-odology based on Massachusetts communities is de- planatory variable in the base equation.scribed in Bradbury et al (1984). '9That is, (cl' dl) and (C2' d2) from footnote 4

"Use of the property tax export rate for the deci- are positive.sive voter's tax share implicitly assumes that the 'Given the number of communities in the typicalproperty tax is the marginal tax. If income or sales metropolitan area, determining the relevant indicataxes were instead the marginal tax, we would expect tors of suburban tax rates and service levels is a com-their export ratios to be positively associated with to- plicated task. Furthermore, the data to implement anytal spending and hence with property tax revenues. such measures are not readily available.But in the choice among revenue sources, higher ex- 21 Furthermore, if it typically costs the city more onport ratios for nonproperty taxes should have a neg- a per capita basis to provide the shifted service thanative effect on property tax revenues, as discussed be- it costs other municipalities within the jurisdictionlow. Export ratios for income and general sales taxes assuming the responsibility, overlying taxes may riseconsistently obtain negative signs in the estimates of by less than $10 per capita; hence, the net increasethe property tax revenue equation (not reported) and in the property tax base may be slightly greaterthose for the income tax differ significantly from zero.This implies either that the property tax is the marginal tax or that the negative effect of higher non-property tax exporting on revenue choice more than REFERENCESoffsets its positive effect on total revenue raising wherenonproperty taxes are used at the margin.

12 For a complete denvation and discussion of this Advisory Commission on Intergovernmental Rela-measure, see chapter 8 in Ladd and Yinger (fortheom tions (1981). "Regional Growth: Interstate Taxing) Competition," Report A-76. March 1981 (Washing-

"One notable exception is Inman (1982) ton, D.C.)"To the extent that city officials are concerned about Bartik, Timothy J. (1985). "Business Location Deci-

deadweight loss, tax rates on alternative tax instru- sions in the United States: Estimates of the Effectsments also belong in the equation. The data allow us of Unionization, Taxes, and Other Characteristicsto reject this hypothesis: The tax base variables of States," Journal of Business and Economw Sta(TRINC and TRSAL) pass the exogeneity test de- tistics, 3.


Bartik, Timothy J. (1986). "Tax Effects on the Loca Crime, Sage Criminal Justice System Annuals; Voltion of Now Branch Plants in the United States," 12, pp. 47 61.Working Paper 86-Wl9, Department of Economics Hausman, J. A (1978). "Specification Tests in Econo-and Business Administration, Vanderbilt Univer- metrics," Econometrtca, 46 (6), 1251 1271.sity. Inman, Robert P. (1982). "The Local Decision to Tax,"

Bergstrom, T., and R. Goodman (1973). "Private De unpublished paper presented at the 1982 meetingsmands for Public Goods," American Economic Re- of Taxation, Resources, and Economic Developmentview, vol. 63, pp. 280-296. (TRED), Lincoln Institute of Land Policy Research,

Bloom, H S, and H. F. Ladd (1982). "Property Tax Cambridge, MA.Revaluation and Tax Levy Growth," Journal of Ur Kieschnick, Michael (1981). Taxes and Growth: Bustban Economics, Winter. ness Incentives and Economic Development. Council

Borcherding, T., and R Deacon (1972). "The De of State Planning Agencies, Washington, D.C.mands for Services of Non-Federal Governments," Ladd, Helen F and John Yinger (1989). America'sAmerican Economic Review, vol. 62, pp. 891-901 Ailing Cities: Fiscal Health and The Design of Ur

Bradbury, Katharine L., Anthony Downs, and Ken ban Policy. Baltimore: Johns Hopkins Press.neth A. Small (1982). Urban Decline and the Future Lugar, Michael I., and Dale 0. Stahl, 11, (1986) "Spec-of American Cities. Washington, D.C.. The Brook ification Errors in Models of Aggregate Labor Sup.ings Institution. ply," The Review of Economics and Statistics, LXVIR

Bradbury, Katharine L., and Helen F. Ladd (1985). (2), pp 274 283."Changes in the Revenue-Raising Capacity of U.S. Oakland, William H. (1978). "Local Taxes and Intra-Cities, 1970-1982," New England Economic Re Urban Industrial l@ocation: A Survey." In Georgeview, March/April. Break (ed.), Metropolitan Financing and Growth

Bradbury, Katharine L., Helen F Ladd, Mark Per- Management Policies Madison: University of Wis-rault, Andrew Reschovsky, and John Yinger (1984) consin Press."State Aid to Offset Fiscal Disparities Across Com- Oates, Wallace E. (1969). "The Effects of Propertymunities," National Tax Journal, vol. 37, no. 2 Taxes and Local Public Spending on Property Val-(June). ues: An Empirical Study of Tax Capitalization and

Bradford, David F, R A. Malt, and Wallace E Oates The Tiebout Hypothesis," Journal of Political(1969). "The Rising Cost of Local Public Services: Economy, 77 no 6 fNovember/December); pp. 1004Some Evidence and Reflections," National Tax 08.Journal, vol 22 no. 2 (June) Papke, Leslic (1986) 'Subnational Taxation and Cap-

Brueckner, Jan K (1982) A Test for Allocative Ef ital Mobility Estimates of Tax-Price Elasticities"ficiency in the Local Public Seewr,"Journal ofpublic MIT mimeo.Economics, 19, 311-331 Rizzo, Mario J - (1979). "The Loss of Time to VictimF,:

Brueckner, Jan K. (1979). "Property Values, Local An Empirical Analysis," in the Journal of LegalPublic Expenditures, and Economic Efficiency," Studies, Vol VIII (1), pp 477 506.Journal of Public Economics, 11, 223 245. Rosen, Harvey S. and David J. Fullerton (1977) "A

Carlton, Dennis (1979). "Why New Firms Locate Note on Local Tax Rates, Public Benefit Levels, andWhere They Do: An Econometric Model," in In- Property Values," Journal of Political Economy, 85,terregional Movements and Regional Growth, edited no 21, pp. 433 440.by William Wheaton. Washington, D C.: The Ur Schmenner, Roger W (1982). Making Business Loban Institute. cation Decisions Englewood Cliffs, New Jersey:

Case, Karl, Leslie Papke, and Susan Koenigsberg Prentice Hall, Inc.(1983). Taxes and Business Location, final report Sonstelie, Jon C and Paul R. Portney (1978). "Profitsubmitted to New York State Legislature Maximizing Communities and the Theory of Local

Due, John F. (1961). "Studies of State-Local Tax Dif- Public Expenditure " Journal of Urban Economics,ferences on Location of Industry," National Tax 5,263 277Journal, 14. Wasylenko, Michael (1981) "The Location of Firms:

Due, John F. (1971). State and Local Sales Taxation. The Role of Taxes and Fiscal Incentives." In R. WChicago: Public Administration Service. Bahl (ed.), Urban Government Finance: Emerging

Due, John F., and John Mikesell (1983) Sales Tax- Trends, Vol. 20, Urban Affairs Annual Reviews,ation: State and Local Structure and Administra Sage.tion. Baltimore: Johns Hopkins University Press. White, Michelle J. (1986) "Property Taxes and Firm

Gray, Charles M. and Mitchell R Joelson (1979) Location" in Harvey S Rosen (ed.) Studies in State"Neighborhood Crime and The Demand for Central and Local Public Finance. Chicago. University ofCity Housing," in Charles M. Gray, The Costs of Chicago Press.

city taxes and property tax bases***€¦ · have investigated the impacts of state and the...

Documents