EVIDENCE ON THE DISTRIBUTIONAL EFFECTS OF A LAND VALUE TAX ON RESIDENTIAL HOUSEHOLDS. Elizabeth Plummer

Transcription

1 Natioal Tax Joural, March 2010, 63 (1), EVIDENCE ON THE DISTRIBUTIONAL EFFECTS OF A LAND VALUE TAX ON RESIDENTIAL HOUSEHOLDS Elizabeth Plummer This study examies how replacig a uiform property tax with a lad value tax (LVT) would shift the tax burde for sigle-family residetial properties i Tarrat Couty, Texas, over the period Results suggest that a LVT would shift the tax burde away from sigle-family properties ad o to other property classes. For the more recet years i the sample, the average tax liability for sigle-family properties would decrease about 30 percet, regardless of household icome, ad Suits Idices suggest that, withi residetial properties, a LVT would be slightly more progressive tha a property tax. Horizotal equity problems would be greatest for the lowest-valued properties relative to other properties. This study also examies how a LVT would chage property values due to the effects of tax capitalizatio. Keywords: tax burde, horizotal equity, tax capitalizatio JEL Codes: H21, H22, H71 I. INTRODUCTION The LVT is a variatio of the property tax. A traditioal property tax taxes lad ad buildig values at the same rate, while a pure LVT taxes lad value but exempts buildigs ad improvemets. 1 Propoets of lad value taxatio argue that a traditioal property tax distorts ecoomic decisios by taxig, ad thus pealizig, improvemets. A LVT is cojectured to be more efficiet because the supply of lad is fixed ad will ot be affected by a tax o lad value, assumig that lad is valued for tax purposes at its curret highest ad best use (Bell, Bowma, ad Germa, 2009, p. 171). I additio to efficiecy gais, the purported beefits of a LVT are that it ecourages more itesive lad use, reduces urba sprawl, ad reduces lad speculatio (Oates ad Schwab, 1997). 1 As discussed i Bourassa (2009), a less extreme versio of a LVT is a split-rate tax, which taxes lad ad improvemets, but at differet rates. The lad tax rate is usually the greater of the two. This is also referred to as a two-tier or graded property tax. Elizabeth Plummer: Departmet of Accoutig, Neeley School of Busiess, Texas Christia Uiversity, Fort Worth, TX (c.e.plummer@tcu.edu)

2 64 Natioal Tax Joural Regardless of a LVT s efficiecy ad ecoomic developmet advatages, public ad legislative support for a LVT depeds a great deal o how the tax would affect the distributio of the tax burde. Would a LVT be more or less progressive tha a traditioal property tax? Who are the wiers ad losers? There are few studies that examie the distributioal effects of a LVT, although icreased iterest i lad value taxatio has motivated two recet studies. Eglad ad Zhao (2005) examie the distributioal cosequeces of chagig from a uiform property tax to a LVT for Dover, NH. They fid that a LVT would icrease the tax o sigle-family residetial property ad that chages withi this property class would be regressive. I cotrast, Bowma ad Bell (2008) provide a similar aalysis for Roaoke, VA, ad fid the opposite result. They fid that chagig to a pure LVT would reduce the aggregate tax burde of residetial properties ad would be more progressive tha the traditioal property tax. This study examies how replacig a uiform property tax with a LVT would affect the tax burde of ower-occupied, sigle-family residetial properties i Tarrat Couty, TX, over the 10-year period Ower-occupied, sigle-family residetial properties make up the largest property class i Tarrat Couty, represetig a little over 50 percet of the total market value of all real properties i the couty. 2 Durig this time period, the umber of ower-occupied, sigle-family properties icreased from 249,731 to 330,835, ad the total market value grew from $21.4 to $49.1 billio. I focus o ower-occupied, sigle-family residetial properties for two reasos. First, I assume i the aalysis that ecoomic ad statutory icidece are idetical. Of course, i may cases, a tax burde ca be shifted from oe taxpayer to aother so that the tax s statutory icidece differs from its ecoomic icidece. For example, owers of retal property could shift the tax to reters (higher retal paymets), while owers of busiess property could shift the tax to cosumers (higher retail prices). However, focusig o ower-occupied, sigle-family residetial properties icreases the probability that the property ower bears the tax s ecoomic icidece, sice opportuities for tax shiftig are relatively limited. Secod, i much of my aalysis, I assume that property value proxies for household lifetime icome. While this may be a reasoable assumptio for ower-occupied residetial properties, it is less likely to be a reasoable assumptio for retal properties. 3 This study s results are summarized as follows. O a aggregate basis, a reveue-eutral LVT would shift the tax burde away from sigle-family residetial properties ad o to other property classes. Over the 10-year period, a LVT would decrease taxes for 85.5 percet of residetial homeowers, ad the media average chage i tax liability 2 If retal houses ad secod homes are icluded, sigle-family residetial property represets about 65 percet of the total market value of all real properties i Tarrat Couty. 3 Lifetime icome measures are preferable to aual icome measures because lifetime measures are less subject to temporary fluctuatios (Metcalf, 1994), ad because idividuals are likely to make cosumptio decisios based o lifetime icome ad ot aual icome (Fullerto ad Metcalf, 2002). Cosistet with prior research (Poterba, 1989, 1991; CBO, 1990; Metcalf, 1994), I use cosumptio as a proxy for lifetime icome, ad use residetial property value to measure cosumptio.

3 A Lad Value Tax o Residetial Households 65 would be 28.6 percet. To provide evidece o the distributioal effects of a LVT across households with differet icome levels, I rak all sigle-family properties from high value to low value ad divide them ito three groups top 30 percet, middle 40 percet, ad bottom 30 percet. Evidece suggests that a LVT would decrease the average tax liability for all three groups. Although the chage i tax liability would be largest i dollar terms for the highest-value properties, the percetage chage i tax liability is comparable across groups at least i the years from I also compute a Suits idex for each year to measure the LVT system s overall progressivity. Results suggest that that a reveue-eutral LVT would be slightly more progressive tha the property tax for the years subsequet to I also examie the effects of a LVT o horizotal equity, defied i this case as the variatio i effective tax rates across property owers with similar property values. The evidece suggests that the variatio i effective tax rates would be greatest for the lowest-valued properties (properties valued at less tha $40,000), implyig that horizotal equity would be lower for this group compared with other properties. I also examie how chages i tax liability resultig from a LVT are related to aual icome measures from the 2000 U.S. Cesus. Results are cosistet with those which use property value to measure household icome. The average decrease i tax liability (i dollar terms) will be greatest for taxpayers i high-icome cesus tracts, but the percetage chages i tax liability are comparable across icome levels. Last, I examie how chagig from a property tax to a LVT will affect property values. If a LVT chages a property s expected future tax liabilities, the the property s market value will chage a effect referred to as tax capitalizatio. Capitalizatio effects will also deped o how soo lad values are reassessed after a LVT system is implemeted. I therefore examie two scearios oe with o reassessmet of lad values ad oe with immediate reassessmet of lad values. These two cases represet the boudaries o the rage of outcomes. Results suggest that, if lad values are ot reassessed whe a LVT is implemeted, sigle-family residetial property values would icrease a media average of 5 12 percet. If lad values are reassessed as part of the LVT system s implemetatio, the the media icrease i property value would be about percet. If lad values are reassessed but with some delay after a LVT is implemeted, the media chage i property values would fall somewhere withi the rage of these two cases. This study makes several cotributios to the study of lad value taxatio. First, it examies how chagig to a LVT system would affect property values through the effects of tax capitalizatio. Chages i property values will shift wealth amog curret property owers, resultig i distributioal effects that have bee igored i prior studies, icludig the two studies cited above. Cosiderig these capitalizatio effects provides a much more complete aalysis of how a LVT system would affect the tax burde tha does a aalysis of sigle-year tax burde chages. Secod, this study provides detailed empirical evidece o the vertical ad horizotal equity effects of a LVT. Most empirical studies focus o the LVT s ecoomic efficiecies (e.g., reductio i deadweight losses) or its effects o ecoomic developmet (e.g.,

4 66 Natioal Tax Joural reductio i urba sprawl). There are few studies that provide detailed evidece o how a LVT would affect vertical equity, ad o studies of which I am aware that provide evidece o how a LVT would affect horizotal equity. 4 The lack of evidece seems especially problematic, give that the public cares deeply about equity ad less about efficiecy (Schwartz, 1998, p. 260). Uderstadig the LVT s equity effects is essetial if the LVT is ever to be cosidered a serious alterative, or supplemet, to the traditioal property tax. Academicias, policymakers, ad taxpayers eed empirical evidece o the LVT s distributioal effects to help facilitate iformed policy decisios. 5 Third, this study cotributes to the evidece o the effects of a LVT by examiig a large umber of residetial properties i orth cetral Texas over a exteded period of time (i.e., a average of 288,000 orth cetral Texas properties over a 10-year period). This icreases our uderstadig of the geeralizability of results o a LVT s distributioal effects over time, ad across differet geographic areas ad jurisdictio types. Fially, although this study focuses o a lifetime measure of icome, it also tests the sesitivity of the results to this approach by followig Bowma ad Bell (2008) ad ivestigatig the effects of usig a aual measure of curret household icome. The remaider of this paper is orgaized as follows. The ext sectio reviews prior research that examies the distributioal effects of a LVT. Sectio III describes the data ad sample, icludig descriptive statistics, ad Sectio IV presets results examiig how a LVT would affect vertical ad horizotal equity for ower-occupied, sigle-family residetial properties i Tarrat Couty. Sectio V presets results that use cesus tract icome data to examie how a LVT would affect vertical equity. Sectio VI provides a aalysis of tax capitalizatio effects, ad the fial sectio presets coclusios. II. PRIOR RESEARCH Icreased iterest i lad value taxatio has motivated researchers to examie the differetial tax icidece of a LVT compared with a traditioal property tax. I geeral, these studies focus o a specific geographic area (e.g., a sigle city) ad estimate the tax burde of a hypothetical LVT that would geerate the same amout of tax reveue as the curret property tax. The hypothetical distributio is compared with the actual property tax distributio to determie how a LVT would redistribute the tax burde across property owers, idetifyig chages i tax liability across taxpayer groups. 4 Bibliographies provided by Grote (2007) ad Dye ad Eglad (2009) show that there are a relatively small umber of studies that focus o the LVT s equity implicatios. These studies are discussed i Plummer (2009). The oly study to address horizotal equity is Sjoquist (2007). However, he does ot compare the horizotal equity of a LVT with that of a uiform property tax. He examies Jamaica s LVT system, ad shows how its curret structure (with tax caps) decreases horizotal equity whe compared with a LVT without tax caps. 5 For example, i 2001 ad agai i 2003, tax bills were itroduced i both houses of the Miesota state legislature. Both bills icluded provisios for a trasitio to a state-wide LVT, ad i both cases, the LVT provisios were evetually dropped. Major reasos icluded ucertaities about the ultimate burde of a LVT ad a lack of familiarity amog legislators regardig lad value taxatio (Havema, 2004).

5 A Lad Value Tax o Residetial Households 67 The studies most relevat to this aalysis are Eglad ad Zhao (2005) ad Bowma ad Bell (2008). Eglad ad Zhao (2005) use tax parcel data for Dover, NH, to provide evidece o how a split-rate tax would redistribute the tax burde across property owers. Dover is a small but growig city orth of metropolita Bosto. Eglad ad Zhao s sample cosists of 5,250 sigle-family residetial properties for They fid that, as a whole, sigle-family residetial properties i Dover would bear a larger percetage of the total property tax uder a split-rate tax system tha uder the existig uiform tax rate system. The authors use assessed property value to measure household icome ad fid that the tax chages across residetial homeowers would be regressive uless some type of tax relief were also implemeted. Bowma ad Bell (2008) use residetial property data for the city of Roaoke, Virgiia, to examie the distributive effects of chagig from a traditioal property tax to a pure LVT. Their sample cosists of 28,680 sigle-family residetial properties for They perform a similar aalysis as Eglad ad Zhao (2005), but with very differet results. Bowma ad Bell fid that chagig to a pure LVT would reduce the aggregate tax burde of residetial properties. The authors the examie the distributioal effects across households usig differet measures of household icome. The first measure of household icome is assessed property value, which is the same icome measure used by Eglad ad Zhao. Bowma ad Bell also lik the Roaoke property tax data to icome data from the U.S. Cesus Bureau, ad use several cesus tract icome measures to measure household icome (e.g., media family icome). Regardless of the icome measure used, the authors fid that a LVT would beefit most those areas with the lowest icomes ad the highest poverty rates. I cotrast to Eglad ad Zhao, Bowma ad Bell s results suggest that a LVT would be more progressive tha the traditioal property tax. 6 I recocilig their results with those of Eglad ad Zhao, Bowma ad Bell (2008) suggest that the differet results reflect betwee-city differeces i lad value ratios (i.e., the ratio of lad value to total property value). The authors suggest that these differeces arise because Roaoke has characteristics of older cetral cities cities that have experieced some populatio loss but remai a employmet ceter whereas Dover is more like a growig bedroom commuity orth of Bosto. This results i residetial properties i Roaoke havig a much lower lad value ratio tha residetial properties i Dover. O average, lad value costitutes 14.3 percet of a sigle-family s total property value i Roaoke ad 34.7 percet i Dover. 7 I additio, there is more 6 Bowma ad Bell (2004) explore the distributioal implicatios of chagig from a traditioal property tax to a pure LVT for two additioal Virgiia local govermets: the couties of Chesterfield (metropolita, with over 250,000 residets) ad Highlad (rural, with fewer tha 2,500 residets). For both couties, they fid that movig from a traditioal property tax to a LVT would decrease the share of taxes paid by residetial property owers ad icrease the share paid by busiess property owers. 7 Bowma ad Bell (2008), Table 3, reports the average ratio of buildig value divided by assessed lad value for sigle-family homes. The average ratio for Dover is 1.88, ad the average ratio for Roaoke is These values imply average lad value ratios (lad value divided by total property value) of ad 0.143, respectively.

6 68 Natioal Tax Joural variatio i the lad value ratios for sigle-family residetial properties i Roaoke tha i Dover. 8 9 III. DATA AND SAMPLE A. Tarrat Appraisal District Data Tarrat Couty is a urba couty located i orth cetral Texas that covers approximately 898 square miles ad curretly has about 1.7 millio residets. Most of those residets live i Fort Worth, the couty s largest city. Tarrat Couty s populatio grew by 24 percet over the period ; it was raked as the fifth fastest growig couty i the Uited States for the 12-moth period begiig July Data for this study come from the Tarrat Appraisal District (TAD) database for the years , ad from the 2000 U.S. Cesus. The TAD data cotais parcel-level iformatio for all real properties i TAD, icludig the values of residetial, busiess, agricultural, ad vacat lad. TAD is resposible for appraisig all properties i Tarrat Couty, ad TAD s boudaries are approximately the same as those of Tarrat Couty. Real property market values are geerally uobservable, uless a property is sold durig the year. This study therefore uses appraisal value to proxy for uobservable market value. Although ot a perfect proxy, there are several reasos why TAD s lad values are likely to be reasoable estimates of market values especially for residetial properties. 10 First, Texas law requires that real property be appraised at 100 percet of market value ad that appraisal records report separate market values for lad ad improvemets. I additio, TAD has adopted a aual reappraisal cycle. This meas 8 Plummer (2009) discusses other studies that examie the distributioal effects of a LVT. These studies iclude Schwab ad Harris (1997) for Washigto, DC; Havema (2004) for commercial ad idustrial property i Miesota; Gloudemas (2000) for Caada; De Cesare, et al. (2003) for Brazil; ad Sjoquist (2007) for Jamaica. 9 All studies that examie the distributioal effects of a LVT geerally make two assumptios. First, calculatios of differetial tax icidece are geerally based o the assumptio that there is o chage i lad values after a LVT is implemeted. However, if a LVT chages a property s expected future tax liabilities, this may cause the property s market value to chage a effect referred to as tax capitalizatio. I examie the effects of tax capitalizatio o property values i Sectio VI. Secod, studies geerally focus o how a LVT would affect statutory (legal) tax icidece. They geerally do ot estimate how taxes might be shifted to other groups i the log-ru (e.g., a busiess tax beig shifted to cosumers i the form of higher retail prices). Because of these assumptios, these distributios studies do ot provide log-term aalysis, which would icorporate possible market resposes such as chages i lad values ad tax shiftig. Although a log-term aalysis could provide valuable iformatio about a LVT s ultimate ecoomic icidece, there are several advatages to examiig a LVT s short-term effects o legal tax icidece. First, short-term aalysis requires far fewer assumptios ad estimatios, makig the results more iterpretable ad reliable tha results from a log-term aalysis. Secod, a short-term aalysis is the first step toward a more complete, log-term aalysis of a LVT s distributive effects. Third, a short-term aalysis provides iformatio o how a LVT would iitially shift the tax distributio across property owers. These iitial chages i actual tax paymets are likely to be a domiat factor i ifluecig voter attitudes about adoptig a LVT. 10 The followig iformatio is take from the 2006 TAD Reappraisal Pla, ad was cofirmed i discussios with TAD officials.

7 A Lad Value Tax o Residetial Households 69 that a property s appraised market value, icludig its specific lad ad improvemet compoets, are reviewed ad adjusted each year. If TAD cotiued to reappraise lad values aually after a LVT system was implemeted, it is less likely that ay capitalizatio effects would be delayed log. TAD s Residetial Appraisal Departmet is resposible for appraisig sigle-family residetial property. TAD s primary approach to valuig residetial properties is the market or sales compariso approach. TAD uses recet sales withi a eighborhood to determie the appropriate market adjustmet for the eighborhood. This geerally ivolves comparig the recet sales prices of eighborhood properties to their appraised values to determie the appropriateess of the sold properties appraised values (i.e., sales ratio aalysis). 11 Evidece idicates that TAD s assessed market values for sigle-family residetial properties are of good quality by idustry stadards. The Texas Property Tax Code requires the State Comptroller to coduct a aual Property Value Study, which measures appraisal performace for all appraisal districts i Texas. For TAD s sigle-family residetial properties for the years , the average media appraisal ratio is 0.98 (i.e., 98 percet of market value), the average coefficiet of dispersio (COD) is 4.79, ad the average price-related differetial (PRD) is COD ad PRD are measures of appraisal uiformity. COD is a measure of horizotal equity ad measures whether appraisal districts are appraisig properties at a equal percetage of market value. A smaller COD value suggests greater uiformity, with COD=0 showig perfect uiformity. PRD is a measure of vertical equity, ad is used to idicate whether assessmet ratios of high-value ad low-value properties are systematically differet. The appraisal ratio ad the COD ad PRD values suggest that the TAD results are well withi the Iteratioal Associatio of Assessig Officers (IAAO, 2007) guidelies. 12 Evidece also suggests that the TAD appraisers are careful i estimatig the separate market values for the lad ad buildig compoets of real property. Each year, TAD s Residetial Appraisal Departmet coducts residetial lad aalysis to develop lad-specific values for sigle-family residetial properties. The appraisers use sales data where available, or abstractio ad allocatio methods, to help esure that the lad values best reflect the lad s cotributio to overall property value. 13 TAD makes extesive use of geographic iformatio system (GIS) techol- 11 TAD maitais a file that cotais sales data for vacat ad improved residetial real properties. This iformatio is collected from a variety of comprehesive sources, icludig data from the Board of Realtors ad other sales vedors, builders, brokers, district survey letters set to buyers ad sellers, field discovery, ad protest hearigs. I 2005 ad 2006, the umbers of sales received ad processed by the residetial research staff were 24,756 ad 25,856, respectively. 12 I its Stadard o Ratio Studies (2007), the Iteratioal Associatio of Assessig Officers (IAAO) suggests that acceptable stadards are a media appraisal ratio betwee , a COD value of betwee for sigle-family residetial properties, ad a PRD of betwee As discussed i Bell, Bowma, ad Germa (2009), the abstractio method starts with the property s total market value ad subtracts the depreciated cost of replacig improvemets. The residual value is attributed to lad. The allocatio method allocates a percetage of the total property value to lad, where the percetage is determied usig market evidece.

8 70 Natioal Tax Joural ogy ad thus has access to iformatio about umerous lad characteristics, icludig locatio, shape, size, view, ad topography. I sum, the evidece suggests that TAD s curret appraised lad values are reasoable estimates of the appraised lad values that would occur uder a LVT system. This is likely to be especially true i the early years of a LVT, as appraisal districts would be trasitioig from their curret property tax systems with their existig lad values. However, measuremet error of lad value remais a cocer uder a LVT system (Bell, Bowma, ad Germa, 2009). Measuremet error ot oly affects estimates of tax shiftig, but would also affect actual tax shiftig uder a LVT system if lad values were ot reassessed appropriately whe a LVT system is implemeted. I the aalysis below, I discuss the effects that measuremet error would have o shifts i the tax burde. B. Descriptive Statistics Table 1 presets descriptive iformatio o ower-occupied, sigle-family residetial properties i Tarrat Couty over the period The umber of properties grew by 32.5 percet over the 10-year period, while total market value grew by Year Table 1 Descriptive Iformatio for Ower-Occupied, Sigle-Family Residetial Properties i Tarrat Couty ($ Millios) Number of Properties Total Market Value ($) Total Lad Value ($) Total Improvemet Lad Value Ratio 1 Value ($) Mea Media ,731 21,358 4,701 16, ,210 22,560 4,811 17, ,690 24,568 5,077 19, ,377 28,075 5,485 22, ,180 31,715 5,829 25, ,394 37,276 6,294 30, ,087 40,018 6,581 33, ,062 42,712 6,997 35, ,401 46,579 7,755 38, ,835 49,082 8,254 40, year growth 32.5% 129.8% 75.6% 145.1% (26.2%) (26.0%) Notes: The 10-year growth i the Cosumer Price Idex (CPI) for the Dallas-Fort Worth area is 25.6 percet. (1) Lad value ratio is defied as the proportio of a property s total market value attributable to lad value ad is equal to the market value of lad, divided by the market value of lad plus improvemets. The mea ad media values are computed usig parcel-level property data.

9 A Lad Value Tax o Residetial Households percet from $21.4 billio to $49.1 billio. 14 The growth i residetial property values greatly exceeded the 10-year growth i the Cosumer Price Idex (CPI) for the Dallas-Fort Worth area, which was oly 25.6 percet. Table 1 also shows that lad values grew more slowly tha improvemet values, with 10-year growth rates of 75.6 percet ad 145 percet, respectively. Accordigly, lad represets a lower percetage of total market value i 2006 tha i earlier years. The differece i growth rates betwee lad ad improvemet values is due i part to ew residetial properties some quite large beig built o less expesive lad located i suburba ad more rural areas of the couty. The last two colums of Table 1 are computed usig parcel-level property data ad preset mea ad media for the lad value ratios. Lad value ratio is defied as the proportio of a property s total market value attributable to lad value ad is equal to the market value of lad divided by the market value of lad plus improvemets. The average lad value ratio for ower-occupied, sigle-family residetial properties has decreased over the past 10 years. Betwee , the mea lad value ratio declied from 22.5 percet to 16.6 percet, while the media ratio declied from 20.5 percet to 15.2 percet. The more recet lad value ratios are similar to the 14.3 percet average for Roaoke, Virgiia, ad smaller tha the 34.7 percet average for Dover (Bowma ad Bell, 2008; Eglad ad Zhao, 2005). IV. RESULTS A. Effect of a LVT o the Tax Burde of Ower-Occupied Sigle-Family Residetial Properties This study assumes that the curret uiform property tax would be replaced by a reveue-eutral LVT, which requires calculatio of the tax rate o lad values that would geerate reveues equal to a jurisdictio s property tax reveues. Over the period , the actual property tax rate averaged 2.42 percet ragig from a low of 2.36 percet i 1998 to a high of 2.50 percet i For this same 10-year period, the reveue-eutral LVT rate would average percet ragig from a low of 8.94 percet i 1997 to a high of percet i I first ivestigate how chagig from a property tax to a reveue-eutral LVT would affect the average tax burde of all ower-occupied, sigle-family residetial proper- 14 I use TAD assessed market values to proxy for market values, but for ease of expositio, I refer to all values simply as market values. 15 Cosistet with Eglad ad Zhao (2005) ad Bowma ad Bell (2008), this study uses existig lad values to calculate the reveue-eutral tax rate (t L ) ad also uses a static defiitio of reveue eutrality. Over the log-ru, shiftig to a LVT would have capitalizatio, reassessmet, ad costructio effects. As these effects occur, the two reveue trajectories (property tax versus LVT) would diverge. Furthermore, if shiftig to a LVT system chages lad values, ad if lad values are reassessed to reflect these chages, the the reveue-eutral tax rate based o existig lad values may differ from the rate based o post-lvt lad values. Estimatig a tax rate based o post-lvt lad values requires oe to estimate lad values for all properties i the taxig jurisdictio (residetial ad busiess), which are i tur a fuctio of t L. Such a simultaeous aalysis is left for future research.

10 72 Natioal Tax Joural ties (hereafter, residetial properties ). For each residetial property, I calculate the tax liability uder the curret property tax ad the reveue-eutral LVT. The chage i a property s tax liability is equal to the property s LVT liability mius its property tax liability. Table 2 presets the results for the years The first four colums show the average (mea ad media) dollar amouts paid each year uder the curret property tax system ad a reveue-eutral LVT system. The ext two colums show average values for the chage i tax liability, while the ext two colums show average values for the percetage chage i tax liability (i.e., chage i tax liability divided by property tax liability). Table 2 shows that chagig to a LVT would decrease the average tax liability for residetial properties for each year I 2006, for example, the mea property tax liability is $3,536, while the mea LVT liability is $2,708 a $828 lower tax liability. The percetage chage i tax liability values are useful because they help cotrol for chages i the dollar s value over time. Table 2 shows that the average percetage chage i tax liability has become more egative over the 10-year period. This icreasigly egative percetage chage i tax liability is due to the decreasig lad value ratios for residetial properties as show i Table Across all years, Table 2 shows that a LVT would decrease a residetial property ower s tax liability by a mea of 21.8 percet ad a media of 28.6 percet. These values are similar i magitude to those reported i Bowma ad Bell (2008). For their sample of sigle-family resideces i Roaoke, VA, the mea ad media percetage chage i tax liability for 2003 would be 20.7 percet ad 24.2 percet (see their Table 4). Of course, ot all sigle-family residetial property owers would experiece a tax decrease if a LVT replaced the property tax. The last colum of Table 2 shows the percetage of property owers that would experiece a tax icrease if a LVT were implemeted. This percetage is geerally decreasig over time, ad oly 12.7 percet of property owers would face a higher tax burde i 2006 if a LVT were implemeted. Over the 10-year period, a LVT would decrease taxes for 85.5 percet of residetial homeowers. These results are quite similar to Bowma ad Bell (2008), who report that a LVT would decrease the 2003 tax liability for 84.8 percet of their sigle-family properties (see their Table 4) The mea ad media lad value ratios for all other property classes, except the multi-family residetial property class, remai stable across time. The ratios for multi-family properties decrease over time. 17 The chage i tax burde for sigle-family residetial properties also depeds o the quality of lad values for other property classes (e.g., busiess properties). If the ratio of assessed-to-market lad value for other property classes is lower (higher) tha the ratio of assessed-to-market lad value for sigle-family residetial properties, the the estimated LVT burde for sigle-family residetial properties as a whole will be too high (too low). Evidece suggests that the lad values for other property classes are reasoable estimates of their lad values. TAD s Commercial Appraisal Departmet reviews values aually, compares assessed lad values with recet commercial lad sales i each desigated lad market area, ad extesively codes a property s lad characteristics. Objective measures for the quality of TAD s assessed market values for other property classes are well withi the IAAO guidelies (i.e., average values of 0.99 for media appraisal ratio, 6.67 for COD, ad 1.02 for PRD).

11 A Lad Value Tax o Residetial Households 73 Table 2 Differetial Tax Icidece: Effect o Tax Liability of Chagig from a Property Tax to LVT System, Ower-Occupied Sigle-Family Residetial Properties Property Tax Liability ($) LVT Liability ($) Chage i Tax Liability ($) Percetage Chage i Tax Liability Percet of Properties with Icrease i Tax Liability Year Mea Media Mea Media Mea Media Mea Media ,033 1,676 1,683 1,341 (350) (348) (15.4) (23.0) 20.5 (=249,731) ,123 1,757 1,738 1,401 (385) (381) (16.5) (23.9) 19.0 (=250,210) ,264 1,859 1,831 1,487 (432) (424) (17.1) (25.0) 18.5 (=257,690) ,509 2,067 2,001 1,596 (508) (499) (18.7) (26.3) 16.4 (=273,377) ,705 2,214 2,095 1,622 (610) (600) (22.2) (29.1) 13.9 (=282,180) ,080 2,537 2,307 1,792 (773) (758) (25.5) (31.8) 11.2 (=296,394) ,272 2,730 2,451 1,906 (821) (815) (25.9) (31.8) 10.8 (=301,087) ,441 2,877 2,560 1,940 (880) (876) (26.6) (32.3) 10.6 (=310,062) ,563 2,961 2,709 2,052 (854) (875) (25.6) (31.7) 11.3 (=326,401) ,536 2,920 2,708 1,954 (828) (841) (24.4) (31.0) 12.7 (=330,835) Average 2,853 2,360 2,208 1,709 (644) (642) (21.8) (28.6) 14.5

12 74 Natioal Tax Joural B. Effects of a LVT o the Distributio of the Tax Burde To provide evidece o the distributioal effects of chagig from a property tax to a LVT, Eglad ad Zhao (2005) ad Bowma ad Bell (2008) rak all sigle-family properties from high value to low value, ad the divide them ito three groups top 30 percet, middle 40 percet, ad bottom 30 percet. For each group, they report statistics o the average chage i tax liability ad o the percetage of property owers that would face a tax icrease if a LVT were implemeted. I perform a similar aalysis for my sample for each of the years ad preset results i Table 3. For each of the three property wealth groups, Table 3 provides the mea ad media chage i tax liability, mea ad media percetage chage i tax liability, ad the percetage of properties that would experiece a tax icrease uder a LVT system. For ease of expositio, I graphically depict some of the iformatio i Figures 1, 2, ad 3 below. I focus my discussio o the figures ad refer to additioal details provided i the table. Figure 1 shows the media chage i tax liability, i dollars, for each property wealth group. The average tax liability for each group would decrease if a LVT were implemeted, ad this decrease becomes more egative over time. The average decrease would be largest for the high-value properties ad smallest for the low-value properties. The media reductio i tax liability rages from $ for the lowest-valued properties, $ for the middle group, ad $883 1,623 for the highest-value properties. The relative patter is similar if mea averages are compared across groups ad over time, although the decreases are slightly smaller i absolute value terms (Table 3). Figure 2 below shows the media percetage chage i tax liability for each property wealth group. For the years , the highest-valued 30 percet of properties would experiece the largest media percetage decrease (ragig from percet), while the lowest-valued 30 percet of properties would experiece the smallest percetage decrease (ragig from percet). The media percetage decrease for the middle group rages from percet. I the years from , however, the media percetage reductio i tax liability is comparable across groups ragig from percet i all cases. The post-2002 results of this study suggest similar percetage chages across all three wealth groups. This differs from Bowma ad Bell (2008), who fid that the media percetage chage i tax liability becomes more egative as property value decreases. For their 2003 sample, the media percetage chage i tax liability would be 21.9 percet for the highest-valued properties, 23.1 percet for the middle group, ad 29.4 percet for the lowest-valued properties (see their Table 4). Bowma ad Bell s results suggest that a LVT would icrease progressivity more for Roaoke tha it would for Tarrat Couty. Figure 3 below shows the percetage of property owers i each group that would experiece a tax icrease if a LVT were implemeted. For all years, owers of the lowest-valued properties would be the most likely to face a tax icrease. However, the percetage of this group facig a icrease drops sigificatly over this period. I 1997, almost 36 percet of property owers i the lowest-value group would experiece a

13 A Lad Value Tax o Residetial Households 75 Table 3 Differetial Tax Icidece: Chage i Tax Liability by Property Wealth Group Top 30 percet Media ΔTL ($) (883) (929) (1,035) (1,147) (1,297) (1,505) (1,551) (1,623) (1,612) (1,597) Media %ΔTL Mea ΔTL ($) (704) (752) (846) (949) (1,098) (1,304) (1,334) (1,400) (1,304) (1,304) Mea %ΔTL % with tax icrease Middle 40 percet Media ΔTL ($) (370) (409) (460) (543) (645) (807) (834) (929) (933) (897) Media %ΔTL Mea ΔTL ($) (297) (336) (381) ($463) (551) (715) (770) (839) (827) (794) Mea %ΔTL % with tax icrease Bottom 30 percet Media ΔTL ($) (101) (119) (124) (169) (242) (362) (420) (457) (488) (457) Media %ΔTL Mea ΔTL ($) (66) (83) (86) (127) (203) (320) (376) (415) (442) (397) Mea %ΔTL % with tax icrease Total sample size 249, , , , , , , , , ,835 Notes: ΔTL = Chage i tax liability i dollars, %ΔTL = Percetage chage i tax liability, ad % with tax icrease = percetage of properties with a tax icrease uder a LVT.

14 76 Natioal Tax Joural Figure 1 Media Chage i Tax Liability, by Group: Ower-Occupied Sigle-Family Residetial Properties Figure 2 Media Percetage Chage i Tax Liability, by Group: Ower-Occupied Sigle-Family Residetial Properties tax icrease uder a LVT, but the percetage drops to 17 percet i 2002 ad remais there (or slightly lower) i all subsequet years. For the highest-valued properties, there is little chage across time i the percetage of owers who would experiece a tax icrease. For all years, about percet would pay higher taxes uder a LVT. Owers of the middle 40 percet of properties are the least likely to face a tax icrease

15 A Lad Value Tax o Residetial Households 77 Figure 3 Percetage of Properties with Tax Icrease, by Group: Ower-Occupied Sigle-Family Residetial Properties if a LVT were implemeted. I all years except 1997, the percetage of property owers who would pay higher taxes uder a LVT is the smallest for this group, ad i each year subsequet to 1999, fewer tha 10 percet of these property owers would face a tax icrease. This study s results for the low ad middle groups are similar to those of Bowma ad Bell (2008), who fid that a LVT would icrease the 2003 property taxes for 17.8 percet of the lowest-value properties ad 8.2 percet of the middle properties. Bowma ad Bell s high-valued properties are slightly more likely to experiece a tax icrease uder a LVT tha this study s high-valued properties (21.9 percet versus percet). I cotrast, this study s results differ sigificatly from Eglad ad Zhao (2005), who fid that a pure LVT would icrease property taxes for 99.2 percet of their sample s lowest-valued properties, 80.3 percet of the middle properties, ad oly 26.1 percet of the highest-valued properties (see their Table 7). 18 The differet patters across studies reflect differeces i lad value ratio patters. For Dover, lad value ratios appear to icrease as property values decrease, makig a tax icrease more likely for lower-valued properties. For Roaoke ad Tarrat Couty, average lad value 18 Measuremet error i lad values withi the residetial property class should ot affect the relative distributioal aalysis uless the ratio of assessed-to-market lad value varies systematically with property value. If the ratio of assessed-to-market lad value is lower for high-value properties tha for low-value properties, the the estimated LVT burde for high-value properties is too low. The aalysis will uderstate the LVT s progressivity, compared with the case where lad values were assessed correctly. Coversely, if the ratio of assessed-to-market lad value is higher for high-value properties, the the estimated LVT burde for high-value properties is too high ad the aalysis will overstate the LVT s progressivity. Measuremet error i lad values across differet property classes is discussed i the footote above.

16 78 Natioal Tax Joural ratios are smallest for the middle-valued property group, makig these properties the least likely to experiece a tax icrease uder a LVT. C. Suits Idices for a LVT System The data preseted above provide evidece o how a LVT would chage the tax burde across icome groups as well as direct comparisos with the fidigs of Eglad ad Zhao (2005) ad Bowma ad Bell (2008). However, the aalysis does ot provide a summary measure of the LVT system s overall vertical equity. Researchers examiig property tax icidece ofte use the Suits idex to measure a tax system s overall progressivity (Suits, 1977). 19 The Suits idex ca vary from 1 to +1, with 1 idicatig maximum regressivity ad +1 idicatig maximum progressivity. A idex value of 0 idicates a proportioal tax. I compute a Suits idex for the LVT system separately for each year ad report results i Table 4. The idex is slightly egative (S = ) i 1997, ad the steadily icreases to i The tred of values suggests that the LVT becomes slightly more progressive over the 10-year period. However, all idex values are very close to zero, with values ragig from to Because I use property value to measure household icome, the Suits idex for the property tax Table 4 Suits Idices for a LVT System: Ower-Occupied, Sigle-Family Residetial Properties Year Number of Properties Suits Idex , , , , , , , , , , The Suits idex = 1 [Σ 10 k=1 (1/2)(PT k + PT k 1 )(Y k Y k 1 )]/5,000 where PT k =cumulative percet of property tax burde for populatio deciles 1 through k, ad Y k =cumulative percet of total icome for populatio deciles 1 through k.

17 A Lad Value Tax o Residetial Households 79 system without exemptios is zero (i.e., a perfectly proportioal tax). 20 Therefore, the Suits idex values i Table 4 suggest that the LVT would be slightly more progressive tha the property tax i years subsequet to D. Effects of a LVT o Horizotal Equity Horizotal equity requires that taxpayers with equal before-tax icomes should have the same effective tax rates (Gravelle ad Gravelle, 2006). However, uder a LVT, property owers with idetical total property values will owe differet tax liabilities if their lad values differ. To examie the effects of a LVT system o horizotal equity, I compare the variatio i the LVT effective tax rates across property owers of equal icomes. For each of the 10 years separately, I divide taxpayers ito 51 groups based o their similarity i property value (a proxy for icome), ad all members of a group are cosidered to have equivalet icome. 21 I compute the effective tax rate for each property, defied as the property s tax liability uder the LVT system divided by the property s total market value. I the compute the coefficiet of variatio (CV) for the effective tax rates separately for each of the 51 groups of icome-equivalet property owers. 22 Accordigly, there are 51 coefficiets of variatio computed for each of the 10 years. I aalyzed horizotal equity separately for each year, but for parsimoy, oly preset results for Figure 4 plots the CV value for each group for Higher CV values idicate greater variatio i effective tax rates for the group, ad thus less horizotal equity. Figure 4 shows that the lowest-valued properties (i.e., those groups with properties valued at less tha $40,000) have the greatest variatio i effective tax rates. The CV values are 91, 88, ad 74. The CV values for all other groups are 58 or less. Groups with properties valued betwee $80, ,000 have the lowest CV val- 20 Perfect proportioality results because taxpayers are beig taxed at a uiform rate o the full market value of their property, which is also the measure of their icome. The property tax would ot be perfectly proportioal if exemptios were allowed to reduce a property s taxable value below its market value (Plummer, 2003). 21 Properties are divided ito groups as follows. For properties with values betwee $10, ,000, groups are formed by $10,000 icremets (i.e., all properties with values betwee $10,000 20,000 are Group 1, all properties with values betwee $20,000 30,000 are Group 2, etc.). Properties with values betwee $200, ,000 are formed ito groups by $25,000 icremets; properties with values betwee $500,000 to $1 millio are formed ito groups by $50,000 icremets; ad properties with values betwee $1 millio to $2 millio are formed ito groups by $100,000 icremets. I delete properties with values below $10,000 ad above $2 millio. This deletes fewer tha 2,000 properties i ay year. For 2006, this deletes 350 properties. 22 The CV is a widely-used measure of horizotal equity ad measures the relative dispersio i effective tax rates amog members withi a sigle group. It ca be expressed as CV N = SD N /Avg(T N ) 100, where CV N = the coefficiet of variatio of effective tax rates for group N, SD N = the stadard deviatio of effective tax rates for group N, ad T N = mea effective tax rate for group N. 23 Results for other years are available from the author upo request.

18 80 Natioal Tax Joural Figure 4 Coefficiet of Variatio for LVT Effective Tax Rate, by Property Value Group for 2006 ues (ragig from 34 39), cosistet with greater horizotal equity. The CV values for properties valued at greater tha $160,000 ted to icrease as property values icrease, but all CV values fall betwee Results for other years ( ) exhibit a similar patter, with the very lowest-valued properties exhibitig the greatest variatio i effective tax rates. This suggests that, uder a LVT system, horizotal equity would be lower for the lowest-valued properties. V. EVIDENCE ON VERTICAL EQUITY EFFECTS USING CENSUS TRACT INCOME DATA All of the aalysis above uses property value as a measure of household icome. Property value as a measure of icome has several advatages: (1) the amout is taxpayer-specific, (2) it is available for each year, ad (3) it may better reflect lifetime icome tha a aual icome measure. However, it is also useful to examie how chages i tax liability resultig from a LVT are related to aual icome measures (Bowma ad Bell, 2008). I this sectio, I examie how chages i tax liability are related to aual icome measures, usig data from the 2000 U.S. Cesus Bureau. To use the cesus data, I must first match each idividual parcel of real property with its correspodig cesus tract. This results i properties beig grouped ito 313 differet

19 A Lad Value Tax o Residetial Households 81 cesus tracts. 24 Cesus tract data are aual measures for a group of households withi a cesus tract. I use a cesus tract s icome measure as a proxy for the icome level of all residetial properties i that cesus tract. I use three icome measures from the cesus tract data: per capita icome, media family icome, ad percetage of idividuals below the poverty level. For each cesus tract, I also compute the media property value (usig the 2000 TAD property data) ad use this as a measure of household icome for the cesus tract. This allows me to compare results usig aual icome measures with results usig lifetime icome measures. The icome measures are highly correlated with oe aother. Per capita icome ad media family icome are highly correlated (a correlatio coefficiet of 0.900), ad media property value is highly correlated with per capita icome ad media family icome (correlatios of ad 0.826, respectively). The poverty rate is egatively correlated with the other icome measures (correlatios ragig from to 0.905). Cesus tracts with greater aual icome levels have higher property values ad lower poverty rates. I first provide a aalysis similar to that i Table 3 above. For each icome measure, I rak cesus tracts from high icome to low icome, ad divide them ito three groups top 30 percet, middle 40 percet, ad bottom 30 percet. I the use all cesus tracts i a group to compute the media chage i tax liability (MediaΔTL) ad the media percetage chage i tax liability (Media%ΔTL) if a LVT were implemeted. 25 Table 5 reports the results. Regardless of the icome measure used to rak cesus tracts, Table 5 shows that residetial properties i the highest-icome cesus tracts would experiece the largest average decrease i tax liability (i dollar terms), while the lowest-icome cesus tracts would experiece the smallest average decrease i tax liability. However, the media percetage chages i tax liability (Media%ΔTL) are more comparable across icome groups, regardless of the icome measure used to rak cesus tracts. Media%ΔTL values rage from 25.2 percet to 28.2 percet for the high-icome cesus tracts, 22.5 percet to 27.7 percet for the middle group, ad 20.3 percet to 23.3 percet for the low-icome cesus tracts. Overall, the Table 5 results are cosistet with the Table 3 results above: the average decrease i tax liability (i dollar terms) is greatest for high-icome taxpayers, but the percetage chage i tax liability is comparable across icome levels. I ext replicate Bowma ad Bell s (2008) aalysis usig my sample of properties. Usig a sample of 23 cesus tracts i Roaoke, Virgiia, Bowma ad Bell, compute the correlatio betwee Media%ΔTL ad several cesus tract icome ad housig measures. For my sample of 313 cesus tracts, Table 6 presets the Spearma correlatios betwee MediaΔTL ad several cesus tract icome ad housig measures 24 A appedix, which describes the matchig process i detail ad provides a table with summary iformatio for the 313 cesus tracts, is available from the author upo request. 25 I compute the media chage i tax liability ad the media percetage chage i tax liability separately for each cesus tract, usig all properties i the cesus tract. I compute the media chage i tax liability ad the media percetage chage i tax liability separately for each group (high, medium, ad low), usig all cesus tracts i the group. These are the amouts preseted i Table 5.