incidence of the property tax on commercial real estate - CiteSeerX

National Tax Journal Vol. 46, no. 2, (June, 1993), pp. 109-20

INCIDENCE OF THE PROPERTY TAX ON COMMERCIAL REAL ESTATE: THE CASE OF DOWNTOWN CHICAGO JOHN F. MCDONALD* Abstract - This paper presents an empirical study of office rents and property taxes for individual buildings in downtown Chicago in 1991. The results show that 45 percent of property tax differentials are shifted forward to tenants as higher gross rents. Application of the Hausman (1978) test for specification error reveals that assessed value per square foot (the base for the property tax) is exogenous with respect to gross rent per square foot.

INTRODUCTION The incidence of the local property tax on nonresidential real estate is an important empirical issue in local public finance. The conventional view, which stems from Mieszkowski (1972), is that interjurisdictional differences in nonresidential property taxes create excise tax effects, because capital and labor are mobile in the long run. Land is immobile, but its share of costs is too small to absorb the entire differential tax burden. The excise tax effect means that some of the burden of the tax is shifted forward to consumers, whose de*University

of lllmols at Chlcago, Chicago, IL 60680

109

mands must be less than perfectly elastic at the jurisdiction level. The only direct empirical tests of the conventional view were conducted by Wheaton (1984). Wheaton found that interjurisdictional differences in actual property taxes paid per square foot (as opposed to statutory rates) had no effect on gross rents in office buildings in metropolitan Boston. This finding means that, in contrast to the conventional view, the burden of the tax was not being shifted from real estate capital and land onto labor or consumers. Wheaton’s results are consistent with the view that land and real estate capital are inelastically supplied to a jurisdiction for the time period under consideration. However, if the demand for offices is perfectly price elastic within a metropolitan area, interjurisdictional differences in property taxes create differences in aftertax returns to capital and therefore can lead to location shifts for capital in the long run to eliminate those after-tax return differentials. Another study of rent per square foot in office buildings by Clapp (1980) included a property tax variable. However, the variable included is the total property tax bill for the building, which is highly correlated with the size of the build-

National Tax Journal Vol. 46, no. 2, (June, 1993), pp. 109-20 ing. The variable has a positive and statistically significant coefficient, but the fact that Clapp (1980) did not use property tax per square foot makes the interpretation of his results difficult. In this study, the focus is on office buildings located in downtown Chicago; the study is of intrajurisdictional differences in property tax rates. It is found that property tax rates per square foot vary enormously within this relatively small area. Indeed, so little of the variation in property taxes per square foot cart be explained by conventional measures’ of the economic success of thle office building (e.g., rent, vacancy rate, age of building, etc.) that the tax rate can be taken as an exogenous variable. The paper thus provides a strong test of whether variations in property taxes cannot be shifted forward. The finding in this paper is that 45%1 of an increase in the property tax per square foot applied to an office building was shifted forward to the tenants. The empirical results thus provide some support for the above-mentioned conventional view of property tax rncidence, albeit in the context of a single taxing jurisdiction. The paper does not address the issue of whether workers or consumers bear the burden of that portion of the property tax that is shifted forward. Other empirical studies of the office market in downtown Chicago include Brennan et al. (1984), Hough and Kratz (1983), Mills (1992), and Posner (forthcomrng). These studies examined the effects on rent per square foot of various building characteristics and location within the downtown area. (Mills (1992) also included data on buildings in the rest of the metropolitan area.) Rents were found to vary a great deal within the downtown area, so location must be taken into account. Brennan et a/. (1984) used individual leases as the unit of observation and were primarily interested in discovering how rent varies with other terms in the lease. Hough and Kratz (1983) tested the most extensive set

of building characteristics and found that age of the building, number of floors in the building, availability of conference rooms, and recognized architectural distinction were statistically significant determinants of average rent per square foot for the building. Mills (1992) found that asking rents were related to the age and size of the building and to the availability of retail services in the building. Mills employed a set of locatron dummies for the various parts of downtown Chicago that is vrrtually identical to the one used in this study. Posner (forthcoming) used a similar set of variables and found that years since last major renovation and presence of security personnel also were statistically significant determinants of average rent per square foot. However, none of these four studies included the property tax in its set of determinants of rent. THEORETICAL

CONSIDERATIONS

Mieszkowski (1972) argued that a uniform real estate tax across all jurisdictions will have no effect on the allocation of resources and will simply reduce the Ireturns to capital (and land). Rather, it is real estate tax differentials that alter real estate prices and resource allocation. Mieszkowski (1972, pp. 90-93), as part of his more general analysis, discusses the incidence of a relatively high real estate tax in a central business district. In this model, the market for downtown real estate is assumed to be competitive with demand and supply elasticities greater than zero and less than infinity. The model leads to the standard result that an Increase in the real estate tax (compared to that prevailing elsewhere) will change the rent on real estate according to dR/dt = 1 /(l

-- Ed/E,)

where R is rent, t is the tax per unit of real estate, Ed is the market elasticity of demand for real estate (defined as a positive

I


PROPERTYTAX INCIDENCE IN CHICAGO

number), and Es is the market elasticity of supply. The tax t is assumed to be a specific tax on floor space (dollars per square foot), because this is how it is regarded by the participants in the market for office space in downtown Chicago. Obviously, if Ed is infinite, dR/dt = 0; and if Es is zero (infinite), dR/dt = 0 (dR/dt = 1). This paper is concerned with individual office buildings within a downtown area. What happens if the real estate tax imposed upon an individual building is higher than that prevailing elsewhere in that same downtown area? One’s first reaction might be simply to say that the individual building faces a demand of infinite elasticity and therefore there will be no impact on the rent charged tenants. But there are reasons to doubt that demand is infinitely elastic. Moving is costly. Suitable quarters must be found, and moving expenses must be paid. If suitable quarters cannot be found in the immediate vicinity, then the tenant’s customers, suppliers, and employees must all make adjustments. Some customers may be lost, relationships with suppliers may be changed, and some valuable employees may be lost as well. In short, the owner of the office building very likely has some market power over the tenants. However, that market power is limited by the fact that landlords do not wish to lose good tenants; landlords and tenants normally form a long-term relationship. Consider a simple model of pricing for a firm with market power. Rent is determined according to

0 R = (1 + m)(C + t) where m is the markup over marginal cost, C is marginal cost excluding the real estate tax, and t is (as earlier) the tax per unit of real estate. The effect changing the tax is

q dR/dt = (1 + m) + t(dm/dt) 111

In the standard C ham berlinian case the firm maximizes profits, or

(1 + m) = ed/(ed - 1) where ed is the elasticity of demand faced by the firm. Because a profit-maximizing firm with local monopoly power will have marginal revenue greater than zero, e,, must be greater than one. If demand elasticity is a constant, dm/dt = 0 and dR/dt must exceed one; the firm with local monopoly power will increase rent by an amount greater than the increase in the tax. Such a result seems implausible and is not consistent with the empirical findings presented subsequently. However, if demand becomes more elastic as price increases, then dm/dt < 0 and it is possible that dR/dt < 1. For example, in the textbook case of a linear demand curve, exactly one-half of a tax increase is passed forward to the tenants. To see this, assume a linear demand curve R = a - bq, where q is quantity. Setting marginal revenue equal to marginal cost (C + t) leads to the profit-maximizing price of R = (a + c + t)/2. Consider another possible model. Following Gordon (1967), Cauley and Sandler (1974), and Sebold (1970), suppose that the firm does not strictly maximize profits. Office building owners normally engage building managers who are responsible for setting rents, marketing the building, etc. There may be some separation of ownership and control which permits building managers to engage in activity that does not lead to profit maximrzation. Building managers may adjust the markup downward (dm/dt < 0) in response to an increase in the tax rate, and rent may adjust by an amount that is less than the increase in the tax. Alternatively, consider a modified version of the model of taxation under oligopoly developed by Stern (1987) and Delipalla and Keen (1992), for example. Firms maxi-


mize profits but set marginal cost equal to perceived marginal revenue which depends partly upon thle conjectured reactions of other firms’ output to changes in its own output. Assume that the price the firm charges is a function of its own output and of the output of the industry, or R = R(Q, o), where R is rent, Q is industry outpl-lt (floor space), and g is the firm’s output. The total (differential of total revenue is

In summary, this section has discussed three basic models of the firm to examine the effect of a specific tax on office floor space on office rent. The standard model of a profit-maximizing local monopolist, a model of a nonprofit rnaximizing firm, and an oligopoly model with conjectural variation all lead to the conclusion that the extent to whch an increase in the tax will be shifted forward is indeterminate. THE DATA

dl:Rq) = Rd9 +- ( f3), and it is also known that adjustments of some sort are made for the age of the building and its location. It is also obvious that the formula for AV includes a sizable error term, and it is reasonable to assume that this error is larger for larger buildings. This equation for AV implies that assessed value per square foot of rentable space, the vanable used in this study, is

7).

av = AV/9 The property tax per square foot increases with gross rent according to

= 8 + (/3 - @(S/q)

+ f(age, location) + u. As such, this variable is exogenous rent equation.

(dt/dR)R,

= [(l -

e)T/(r

+ T)IR,

where R, is the marginal effect of attribute 1 On gross rent. For example, if 7 = 0.04 (a typical value for 7 in Chicago), e = 0.4, and r = 0.06, then dt/dR = 0.24. In short, variables that increase gross rent are also supposed to increase property taxes. This means that the disturbance term in the gross rent equation in column 1 of Table 3, which depends upon a collection of omitted building and/or location attributes, may be (positively) correlated with the property tax variable. The actual method used to determine

as117

in the

The Hausman (1978) specification test is used to check for possible simultaneity bias. Under the null hypothesis in large samples, the assessed value and disturbance term are uncorrelated. The alternative hypothesis is that these two variables are correlated and that OLS does not yield consistent estimates of the coefficients. The Hausman test involves estimation of a reduced-form equation for assessed value. The predicted values from this estimated equation are then included in an expanded OLS regression for gross rents. The null hypothesis is not rejected if the hypothesis that the coefficient of the predicted as-


sessed value is zero cannot be rejected. The coefficient of the predicted assessed value is the difference between the OLS and the two-stage least-squares coefflclents of assessed value. Tvvo alternativle versions of the Hausman test were run. In the first version, stage one of the Hausman test is an OLS regression of assessed value per square foot on all of the independent variables listed in column 1 of Table 3 plus the total square feet of rentable space and the proportion of rentable space devoted to offices (as opposed to retailing and other uses). Preliminary regression results (available upon request) indicated that these two variables are not determinants of office rents, and equation 11 indicates that the proportion of rentable space devoted to offices is an irnportant determinant of the assessment. As shown in column 1 of Table 4, the R2 for this regression is 0.22. The predicted values from this regression are then entered in the gross rent regression, and the results are shown in column 2 of Table 1. The coefficient of predicted assessed value is small (0.025) and statistically insignificant (1 = 0.115). This version of the Hausman test thus leads to the conclusion that assessed value per square foot can be considered an exogenous variable. SlDrne readers of an earlier version of this paper expressed concern that the results of the Hausman test can be sensitive to the choice of exogenous variables included in the first-stage regression. A second version of the Hausman test has been run to exatlline this issue. In this case, it is hypothesized that, in addition to the variables included in column 1 of Table 4, assessed value per square foot is determined by an “unofficial” program of granting property tax breaks to individual buildings Cook County has an official program that lowers assessed value to 42% of its normal level, bdt this program does not apply to the dl2wnto’wn area. Nevertheless, individual

building owners have the opportunity to contact the assessor’s office in advance of a reassessment in order to point out financial problems. Also, the appeal process IS available once the reassessment is issued. It IS obvious that some building owners have used these opportunities to their advantage, but which ones? The available data can be used to determine which buildings lhave been granted a tax break that is equal to (or greater than) the official program that exists in other parts of the ctty. The “tax break” variable IS defined (3s follows. According to Table 1, commercial buildings in the assessment area that includes downtown had a median assessment that is 20.4 percent of market value. A tax break that reduces the assessment to 42 percent of its normal level would mean an assessment of 8.6 percent of market value. Data on the market values of buildings are not available, but recall that market value V = (1 - e)R/ (r -t +I. Suppose that the ratio of assessed to market value of 0.086 indicates the awardtng of a tax break. From the equation for V, the ratio of assessed value to rent per square foot for such buildings is

w/R

== O.OSC,[(l - e)/(r + T)]

If e = 0.4, r = 0.06, and 7 = 0.04, then w/R = 0.52. A dummy variable for the awarding of a tax break has been defined for bulldings with av/R less than 0.50. Thirty-four buildings (13 percent) fall into thts category. A probit analysis (not reported) using the tax break dummy as the dependent variable reveals that the tax break IS not related to some obvious criteria, such as age of building, vacancy rate, or building size. The tax break variable is providing new information. q

The first stage regressi’on for the Hausman test has been run with the tax break vari-


PROPERTYTAX INCIDENCE IN CHICAGO

TABLE 4 REGRESSION ANALYSIS OF ASSESSED VALUE PER SQUARE FOOT: OFFICE BUILDINGS IN DOWNTOWN CHICAGO IN 1991 (T-RATIOS IN PARENTHESES)

(2)

(1) Constant Age of building Number

of floors

Percentage Building

(years)

vacant

built before

Percentage

1950, not rehabbed

of rentable

Total rentable

space devoted

since 1979

to offices

space (1000 square feet)

0.324 (0.15) 8.206 (2.20) -2.195 (0.85) -7.935 (2.76) -12.443 (2.46) 0.225 259

-0.568 (0.28) 6.309 (1.83) -2.094 (0.89) -7.678 (2.91) -8.896 (1.91) 0.348 259

of building

North of Chicago River Wacker

24.755 (4.56) -0.136 (4.43) -0.095 (1.13) -0.049 (1.13) - 1.304 (0.74) 0.090 (1.95) 0.002 (0.93) - 15.242 (6.83)

-

Tax break granted Location

18.124 (3.12) -0.125 (3.73) 0.001 (0.00) -0.045 (0.94) -1.611 (0.84) 0.121 (2.41) 0.000 (0.12)

Drive (West Loop)

East of Loop West of Chicago River South of Loop R2 Sample size

able included, and the results are shown in column 2 of Table 4. As one would expect, the tax break variable is highly statistically significant and the ti is 0.35 (compared to 0.22 without the tax break variable). The predicted assessed value from this new regression equation is entered into the regression equation for gross rent, and the results are shown in column 3 of Table 3. Once again, the predicted value of the assessment is not statistically significant, although the t ratio is 1.19 (compared to 0.15 for the earlier Hausman test in column 2 of Table 3). The entire Hausman test was then repeated using a critical value of av/R of 0.65 rather than 0.50, and the predicted value of the assessment once again was not statistically significant. In short, these additional Haus-

119

man tests support the conclusion that the assessed value per square foot is an exogenous variable in the equation for gross rent.

Conclusions This study of 259 private office buildings in downtown Chicago has shown that assessed value per square foot strongly influences a building’s average gross rent per square foot. In 1991, 45 percent of property tax differences across buildings were shifted forward to tenants. This result means, of course, that 55 percent of property tax differentials were absorbed by the owners of the building and land. The effect of the property tax on gross rent can be estimated with considerable


precision, because assessed value per square foot turns out to be an exogenous variable with respect to gross rent. The property tax is supposed to be determined by the market value of the building, which clepends upon gross rent and expenses. However, the apparent difficulty with establishing accurate estimates of market values tums the property tax into an exogenous-some rnight say arbitrary---vanable. This study pertains only to the office market in downtown ChIcago in 1991. Further cross-sectional and time-series studies of other commercial real estate mark.ets are needed to determine the generality of the results presented here.

tlvltles ” loun7,1/ of RegIonal Soence 20 (Augu,l,

Delipalla,

Sophia

Cornpetition

” Journal of Public Economw

Gordon,

Robert.

Be-

49 (1992): 351-

“The lncldence of the Corporate

Tax in U S M jnufactunng, Review 57 (Seiltember, Hausman,

Jerry.

Econometnca Hough,

1925 -62 ” Amencxn

1967)

Douglas

Income Econom/c

73 l-58

“Speclficatlon

~16 (December,

Tests in Econometrics 1978)

and Charles

”

1251-72

Kratz.

“Can ‘(3ood’ Archltec-

ture Meet the Market Test?” /ourna/ of Urban riconom/cs

14

(July, 1983). 43 -54 Illinois

Deparlment

ment/Sales Revenue

of Revenue.

“FIndIngs of 1990 Assess-

R~IIO Study ” Spnngfleld: November,

Mieszkowskc

llllnols Department

of

1992

Peter. “The Property Tax An Excise Tax or a

Prottts Tax”’

Journal of Pub//c iCconom/cs I (January. 1972)

Mills, Edwin.

asststarlce and the refereer for hcllpful comments REFERENCES “l-4

1991 Occupancy

Survey ” Chicago

Building Owners’ and Managt?rs’ Assoclatlon

of Chlcago,

1991 Roger Cannaday

and Peter Colwell.

Rent in the Chicago CBD ” Amerkn

Lvban Econom/cs A:;soaat/on

Real Estate and

Journal 12 (Fall, 1’384) 243-

60

1992)

Posner,

Analysis of Offlce Rents In

Kenneth.

“An Amenity

Sandler.

“lhe

Short-Rbn Shtftlng of

Income Tax A Theorrtlcal

/K F/nance 29 (1974)

Investqatlon

” Pub-

19-35

Clapp, John. “The Intrametropolltan

Associdtm, forthcom-

Iv F.D. ’ Short-Run Response In a Utility-Maxlmlzatton

Framework ” A’aOonal Tax lournal

23 (December,

1970)

365-72 Stern,

Nicholas.

“The Effects of Taxation, Pnce Control

C Intracts

petltion ” /our/~a/ Wheaton,

William.

37 (December,

In Oligopoly

and Monopolistic

of Public Econonxcs 32 (1987)

and

Com-

133-58

“The Incidence of Inter-JurIsdIctIonal

ferences In Commercial Locatlon of Offlce Ac-

273-87

Business District ” Journal of the Amen-

can Real Esta’t and Urban Et-onom/cs

Government Jon and Todd

In the Chicago

Red/Estate and Urban [co.

nonvcs ,&socr,t~o,? 20 (Summer,

Sebold, Thomas,

the Corporate

“Offtce Rent Deiermlnants

the Chlcago CIW,al

BOMA/Chicago.

Cauley,

Keen. “The Comparison

67

Area ” Journal of the Amencan

*The author thanks Mike Thorn of BOMA/CtlIcago

“OffIce

and Michael

tween Ad Va c rem and Specific Taxation Under Imperfect

73 96

ENDNOTE

Brennan,

1980)

387. 399

Dif-

Property Taxes ” Nat,on,3/ Tax lournal

1984): 515-27