THE EXCISE TAX EFFECTS OF THE PROPERTY TAX REVISITED

Transcription

1 TE EXCIE TAX EFFECT OF TE PROPERTY TAX REVIITED Athiphat Muthitacharoen Analyst, Tax Analysis Division Congressional Budget Office George R. Zodrow Cline Professor of Economics and Rice cholar, Tax and Expenditure Policy Program Baker Institute for Public Policy Rice University International Research Fellow Centre for Business Taxation Oxford University Preliminary Draft: June 30, 2009 Comments Welcome We would like to thank Peter Mieszkowski for helpful comments. The views expressed in this paper reflect the opinions of the authors and should not be interpreted as those of the Congressional Budget Office or the Baker Institute for Public Policy.

2 ABTRACT This paper analyzes the excise tax effects of a general property tax from the perspective of a small open economy facing a perfectly elastic supply of capital, focusing on the mix of forward shifting to consumers and backward shifting to labor and landowners. The model utilized differs from most that have appeared in the property tax literature in that: (1) the property tax is applied in a four-sector model with three taxed sectors (manufacturing which produces a tradable good, and housing and services which produce nontradable goods) and a tax-exempt agricultural sector, (2) the analysis considers an intermediate run time frame in which labor is partially mobile (perfectly mobile across production sectors but fixed in total supply) while land is fixed in each production sector, and (3) all production sectors use capital, labor and land. Within this context, the imposition of the property tax is primarily borne on the sources side of incidence. This result is in marked contrast to a four-sector analog of the traditional view of the excise tax effects of the property tax with immobile labor, under which there is significant over-shifting of the tax to the consumers of nontradable goods. This result is robust to some sensitivity analysis of the parameters used in the simulations of the model. owever, when labor mobility is extended to include interurisdictional mobility such that wages are fixed, the degree of backward shifting naturally declines markedly, with a distribution of the excise tax effects of the property tax that is intermediate to the cases of immobile labor and intersectoral labor mobility. JE Codes: eywords: 21, 22, R13 Property Tax Incidence, Excise Tax Effects, Capital Tax View, mall Open Economy

3 I. INTRODUCTION AND OVERVIEW The incidence of the property tax is an issue that has been examined at length in the literature but is still far from resolved. The focus of the debate traditionally is on the fundamental question of whether the property tax is best viewed as a non-distortionary benefit tax or user charge for public services received, or a distortionary tax on the use of capital in the production of housing and other goods (Fischel, 2001; Zodrow, 2001). This paper, however, does not address this controversial issue, as it simply assumes the validity of the latter new view or capital tax view of the property tax. Instead, this analysis elaborates upon the incidence of the tax under the capital tax view, focusing on the excise tax effects of the imposition of a property tax by a single small urisdiction. These effects especially their distribution among factor owners and consumers have received less attention in the literature; for example, in a recent review of the effects of the property tax, Fisher (2009) discusses the various excise tax effects of the property tax at some length, but does not cite any evidence regarding their distribution. This does not, however, imply that such effects are quantitatively unimportant; for example, Gravelle (2007) estimates that excise tax effects account for percent of the burden of the property tax in the U.. The capital tax view was developed initially by Mieszkowski (1972) as an extension of the arberger (1962) multi-sector model of national tax incidence to a multi-urisdictional setting. It was further extended by Zodrow and Mieszkowski (1986) to include a wide variety of the characteristics of local public service provision and property taxation stressed in the literature based on the Tiebout (1956) model, including competition among local urisdictions with endogenous tax and expenditure policies, differing individual tastes for public services, community segregation by taste for public services, and a simple form of zoning. An essential 3

4 element of these derivations of the capital tax view is that they assume, quite plausibly, that the property tax applies to both residential and non-residential property and is used simultaneously by virtually all urisdictions in the country. Within this general equilibrium context and given the assumption that the national capital stock is fixed the capital tax view implies that the incidence of the property tax has two components: (1) a profits tax component that reflects the average rate of property taxation in the nation and is borne by all capital owners as a reduction in the after-tax rate of return to capital, and (2) an excise tax component that reflects the effects of local tax differentials about the average rate of taxation in the nation, with above-average tax rates reflected in some combination of higher consumer prices and lower returns to labor and land, and offsetting effects in relatively low-tax urisdictions. This paper examines more closely the excise tax effects of the property tax, in particular how they are distributed among consumers, suppliers of labor, and the owners of various types of land in the taxing urisdiction. To focus on these excise tax effects, the analysis abstracts from the profits tax component by considering the incidence of a property tax imposed by a single small taxing urisdiction that faces a fixed after-tax rate of return to capital that is, a small open economy facing a perfectly elastic supply of capital under the assumption that neighboring urisdictions hold their tax policies constant. In this context, all of the local effects of the property tax can be viewed as excise tax effects, as the effects of the tax on the return to capital owned by local residents are negligible 1 and the local effects of the tax are concentrated in 1 The analysis thus adopts a partial equilibrium approach in that it examines tax incidence solely from the perspective of a single taxing urisdiction that faces a perfectly elastic supply of capital, neglecting general equilibrium effects on other urisdictions and the profits tax effect on the return to capital. Note, however, that Bradford (1978) shows that the profits tax effect still obtains for a tax imposed in a single small urisdiction when these external effects are considered, as the burden of the tax is borne by capital as a very small reduction in the return to capital that is spread over the entire capital stock. In addition, the excise tax effects that occur in the taxing urisdiction and are the focus of this analysis are offset by similar effects of opposite sign in the non-taxing urisdiction. 4

5 wages, commodity prices (unless the price of a good is determined on a national or international market) and land rents. By comparison, in the derivation of the capital tax view, all urisdictions effectively raise their taxes simultaneously, so that the excise tax effects are calculated as differentials about the average rate of tax, as the excise tax effects of the average rate of tax imposed by each urisdiction cancel across urisdictions (and the tax has no effect on the aggregate supply of capital by assumption). The small taxing urisdiction analyzed in this paper might be interpreted as a single large city, or perhaps a large suburb with tax and expenditure authority, that raises its property tax to finance public services under the assumption that other communities that do not change their tax policies, 2 or as a group of communities in a metropolitan area that have relatively high property tax rates. Although the excise tax effects of the property tax have been analyzed in the literature, much of which is rather dated, our paper extends this literature in several ways. 3 Most importantly, most such analyses focus exclusively on the effects of the property tax within a single sector, typically residential housing, 4 or in some cases an industrial sector. 5 But, as stressed in the derivation of the capital tax view, a central feature of the property tax is that it applies to both residential and nonresidential or business property, typically at the same rate, so that its excise tax effects should be examined in a model that has at least two sectors. This is especially important since the market characteristics of the two sectors are typically so different, 2 The analysis thus overstates the responsiveness of capital to property tax increases to the extent that such increases are matched in neighboring communities, as is consistent with the strategic property tax competition literature (Brueckner, 2003). 3 For a review of this literature, see Muthitacharoen and Zodrow (2009). 4 For example, see obson (1986), in (1986), Brueckner (1981), eroy (1976) and Arnott and Mackinnon (1977); all of these studies take a long run perspective in which land is flexible across uses in different sectors and labor (if considered at all) is perfectly mobile across urisdictions. 5 For example, ullivan (1984) considers an industrial property tax that exempts housing. 5

6 with residential housing a nontradable good with a price that is locally determined, while the nonresidential sector is likely to produce tradable goods that face much more price competition from national or international competitors and may even effectively face a fixed price. Moreover, even a two-sector approach will miss two essential features of the property tax as it is applied in the U. First, as shown in the two-sector model analyzed by Muthitacharoen and Zodrow (2009), the excise tax effects of the property tax differ significantly depending on whether the nontradable goods sector is a relatively capital intensive sector, such as residential housing in the typical analysis, or a relatively labor intensive sector, such as services. In particular, somewhat counter-intuitive results such as lower prices in the nontradable goods sector may result in the latter case. econd, agricultural property, especially land, is typically taxed very lightly, implying the need to include a relatively low-tax or tax-exempt sector in the model. Another issue is that most existing studies assume two-factor production functions, ignoring the need to model separately the effects of the property tax on capital, land and labor, especially when focusing on the nature of the excise tax effects of the property tax. Finally, most studies tend to take rather long run views of property tax incidence, often with individuals fully mobile across urisdictions and land mobile across production sectors; by comparison, we consider an intermediate run case in which labor is partially mobile and land is immobile across sectors. More specifically, in this paper the excise tax effects of the property tax are analyzed in the context of a model of a small open economy with four production sectors and three factors of production (capital, labor and land) in each sector. The property tax is assumed to apply to both capital and land used in the production of housing and two of the non-housing goods, broadly defined as manufacturing and services, while the fourth production sector, agriculture, is 6

7 assumed to be exempt from property taxation. 6 In two of the production sectors, corresponding to housing and services, goods are assumed to be nontradable with prices that are determined locally. In the other two sectors, manufacturing and agriculture, the goods are assumed to be tradable, with prices that are determined in national or international markets and are thus fixed from the perspective of the taxing urisdiction. This allows us to examine the effects of the imposition of a property tax in three of the four sectors, which results in an admittedly still highly stylized, but nevertheless relatively more realistic, representation of property taxation in the U.. In addition, and in contrast to many models with two-factor production functions, all four production sectors are assumed to use all of the three standard factors of production of capital, labor and land. We follow most of the literature in assuming constant elasticity of substitution (CE) production functions, as well as a CE utility function defined over consumption of the four private sector goods, with public goods separable in the utility function. This assumption, however, is restrictive in that requires identical elasticities of substitution between each pair of factors in a sector, and between each pair of goods in the utility function. Note in particular that our model of local property tax incidence draws on some recent studies of the incidence of the corporate income tax in the international taxation literature, which has shown that a many-sector, many-factor approach provides important insights into tax incidence analysis that are not captured in the standard two-sector, two-factor model, especially when some goods are tradable while others are not (Gravelle and metters, 2006; arberger, 2008). 6 Although this assumption is only an approximation, it is nevertheless fairly realistic, as agricultural property typically benefits from very generous treatment under the property tax in the U.. These benefits include the valuation of agricultural property according to its current actual use which is typically much lower than its market value, assigning a lower assessment ratio for agricultural property, and providing various tax credits or exemptions to qualified farmers (National Conference of tate egislatures, 2002). 7

8 The resulting model is sufficiently complicated that our analysis is limited to numerical simulation of the model. owever, some intuition regarding into the operation of the model is provided in our companion analytical incidence study (Muthitacharoen and Zodrow, 2009), which examines the excise tax effects of the property tax in a two-sector, three-factor model that considers both the case in which the nontradable good is capital-intensive housing and the case in which it is labor-intensive services. As noted above, a separate issue is the time frame of the incidence analysis, which is reflected in the degree of mobility of the various factors of production. In all cases, we adopt the relatively standard small open economy assumption that the supply of capital to the taxing urisdiction is perfectly elastic, effectively assuming that changes in new and replacement investment occur rapidly enough to reach the desired capital stocks in each sector quickly. 7 owever, rather than assuming a full long run equilibrium with perfectly mobile households and free reallocation of land across production sectors, we focus on an intermediate run case characterized by fixed supplies of land in each sector and partial mobility of labor in the sense that labor does not change its urisdiction of residence (so the total supply of labor within the taxing urisdiction is fixed), but can move freely among the four sectors within the taxing urisdiction. (In the sensitivity results, we also consider a case in which labor is free to seek employment in other neighboring urisdictions while continuing to reside in the taxing urisdiction.) Thus we effectively assume that over the relevant time period the benefits of the additional public services received by living within the taxing urisdiction (which are assumed to be separable in the individual utility function) as well as any attachment to community and other 7 ee Fullerton (1983) for a ustification of this assumption in the context of a closed economy. In addition, most recent empirical work suggests that the international capital mobility is increasing over time (Zodrow, 2008). 8

9 transactions costs associated with moving are sufficiently large to preclude a change of residence by households. We believe that this intermediate run case is of considerable interest. By comparison, most other studies of the excise tax effects of the property tax have taken very long run view of property tax incidence; these studies typically assume that labor is perfectly mobile across all urisdictions, which generally implies that it bears little if any of the tax, and that land is perfectly mobile among all uses, implying that all landowners within a urisdiction bear equally any part of the tax that is capitalized in land values. Although these long run incidence results are certainly of interest, from a policy perspective, incidence results over the alternative intermediate run time period analyzed in this paper are ust as critical (if not more so) than very long run results. At a minimum, these results provide insight into the nature of the effects that would occur during the (lengthy) transition to a new long run equilibrium after an increase in a general property tax by a single taxing urisdiction. In addition, this partial equilibrium perspective, with capital perfectly mobile, is highly relevant from the perspective of a single urisdiction considering an increase in the tax. 8 everal earlier papers have examined the excise tax effects of the property tax. These studies tend to fall into two camps. One suggests that the excise tax effects of the tax will primarily be reflected in higher consumer prices. For example, in his seminal contribution, Mieszkowski (1972) suggests that in a single sector model, changes in wage rates will be relatively small as labor is partially mobile and can also be substituted for capital. In addition, he argues that changes in land values are likely to be substantial from the perspective of land 8 This contrast with other studies that follow in the new view tradition and assume a fixed supply of capital, thus effectively analyzing incidence from the perspective of the nation (or perhaps a very large region), under the assumption that changes in after-tax returns have no effect on the aggregate capital stock. 9

10 owners but will not be large in the aggregate due to the low share of land rents in total costs. Instead, Mieszkowski predicts that commodity prices will rise so that at least 75% of the excise tax effects of the property tax will fall on consumers. This is roughly consistent with a third view of the incidence of the tax the traditional view which argues from a partial equilibrium perspective that the capital component of the property tax is fully shifted forward into housing prices and the prices of non-housing goods (Zodrow, 2006; Wildasin, 1986). For example, Wildasin (1986) shows that this result obtains exactly within the context of a two-factor mobile in which capital is perfectly mobile and labor and land are both immobile when the demand elasticity for the taxed good equals the elasticity of substitution in production. 9 Wassmer s (1993) empirical analysis suggests that 87 percent of the excise tax effects of a general property tax are shifted forward as higher prices for housing and other goods. Finally, Youngman (2002) and Fisher (2007) note that much of the concern about the regressivity of the property tax is based on the assumption that the residential portion of the tax, and perhaps much of the nonresidential portion as well, is shifted forward in the form of higher consumer prices, implying regressivity (at least when measured with respect to annual income). In contrast, a second strand suggests that forward shifting of the property tax will be much less important. For example, in the same single-sector, two-factor (mobile capital and immobile land/labor) model, Wildasin shows that the distribution of the excise tax effects in general depends on the relative magnitudes of the demand and production substitution elasticities, and in the special case in which demand is perfectly elastic all of the excise tax effects of the property tax are naturally borne by the immobile factors. imilarly, in a study that 9 In this case, the reduction in demand for the immobile factor due to tax-induced commodity price increases and the associated reductions in output demands are ust offset by increases in demand for the immobile factor due to substitution away from the taxed mobile factor, leaving the price of the immobile factor unchanged, that is, resulting in full forward shifting of the property tax. 10

11 is in some ways similar to our own, Wilson (1984) considers a general property tax imposed on perfectly mobile capital and land in both tradable and nontradable sectors, with both sectors employing capital, labor and land. is analysis, however, considers only two production sectors and takes a long run, general equilibrium view of property tax incidence, as households are perfectly mobile between regions, and land is perfectly mobile between production sectors within a region. Wilson also allows individuals to differ in their demands for the nontradable good. e focuses on the excise tax effects of the property tax on the prices of the nontradable good in the presence of perfectly mobile households that are heterogeneous in non-wage income. e shows that within this context the elasticity of demand for the nontradable good is infinite, as households with different incomes and different tastes costlessly migrate between regions in response to any difference in non-traded good prices, changing the mix of households in each urisdiction until this difference is eliminated; that is, perfect mobility of households across urisdictions implies the absence of excise tax effects in the form of higher prices for nontradable goods. Although Wilson does not address incidence issues directly, presumably land bears all or most of the burden of the property tax in his model. Thus, while Wilson focuses on the long run implications of perfect household mobility, this study examines property tax incidence from the perspective of a single small taxing urisdiction in an intermediate run model with partial labor mobility and land supplies that are fixed across production sectors. The remainder of the paper is organized as follows. The following section contains a description of the model and the parameter values used in the simulations. ection III presents the simulation results, and ection IV concludes. 11

12 II. MODE AND CAIBRATION II.1 Model Description The model used in this paper to calculate the intermediate run incidence of an increase in a local property tax has four production sectors: agriculture (A), manufacturing (M), housing services (), and non-housing services (). The goods produced by the manufacturing and agricultural sectors are tradable, while housing and non-housing services are nontradable. All production sectors use capital, labor and land as inputs, with land in each production sector fixed. The taxing urisdiction is assumed to be a small open economy that faces a fixed net rate of return r on capital (the supply of capital is perfectly elastic), a fixed price for the tradable agricultural good ( p A), and a fixed price for the tradable manufactured good, which is the numeraire ( p M = 1). Prices for the two nontradable goods ( p, p ) are determined endogenously. All markets are assumed to be perfectly competitive. Each resident of the urisdiction owns one unit of labor, an equal share of the fixed supply of local housing land, and an equal fixed share of a national portfolio that includes all of the fixed national supply of capital and the fixed national supply of land used for production of all goods other than housing. The small open economy assumption implies that the actions of the single taxing urisdiction do not affect the aggregate value of the national portfolio. The property tax rate is imposed on all uses of capital and land in all sectors except the agricultural sector (A), and is stated on a tax-exclusive basis. For each unit of capital, the capital owner receives the after-tax return r while capital costs producers r(1 + T ), where the property tax rate is T = T, = M,, and T = 0, = A. ince land supplies are fixed in each sector ( V, = AM,,, ), it is convenient to use a restricted profit function approach, with gross 12

13 returns to land, the fixed factor in sector, expressed as Π [ p, wr, (1 + T); V] and landowners receiving a net return of Π / (1 + T ). Producer Optimization technology where All production sectors are characterized by constant elasticity of substitution (CE) ρ ( ) 1/ 1 ρ ρ ρ α α α, σ Q =Ψ + + VV =, 1 ρ Q is the amount of good produced within the urisdiction, used in sector, production in sector, and the is the amount of capital is the amount of labor used in sector, σ is the elasticity of substitution in αi are the weights for factor i in sector, and for sector, with units in each sector chosen so that Ψ = 1. Ψ is a scale factor As described above, the model is characterized by partial mobility of labor, as labor is immobile across urisdictions, but perfectly mobile across productions sectors within the taxing urisdiction, earning wage w. Thus, the total supply of labor ( ) within the taxing urisdiction is fixed, so that A + M + + =. Using the restricted profit function approach for the CE production functions, gross returns to land (residual profits) are outputs σ ( ) ( (1 )) Π = Ψ + σ /( 1) 1 1 σ σ σ σ 1 σ αv V P α r T α w 1/(1 σ ) Differentiation of the restricted profit expressions with respect to output prices yields Π 1 σ Q P ( P ) ( r(1 T )) w V P 1 /(1 ) /( 1) 1 σ σ σ σ σ σ σ σ σ 1 σ = = α Ψ Ψ α + α V. 13

14 and differentiation with respect to factor prices yields the factor demands σ Π 1 σ /(1 σ ) σ /( σ 1) α σ σ σ 1 σ = = α Ψ α + α V 1 σ ( ) ( (1 )) P r T w V w w σ Π 1 σ /(1 ) /( 1) 1 σ σ σ α σ σ σ σ 1 σ = = α ( ) ( (1 ) V Ψ α + ) α [ r(1 + T)] r(1 + T) P r T w V Consumer Optimization The analysis assumes that the utility function of the representative resident of the taxing urisdiction is also characterized by a CE function, defined over consumption of the four goods ρ ρ ρ ρ ( δ δ δ δ ) 1/ U( AM,,, ) = AX A + MXM + X + X, where σd = 1/(1 ρ) is the elasticity of substitution between each pair of goods, and δ is the share parameter for good. The associated indirect utility function is ρ. 1/( σ D 1) σd 1 σ D V( P, Y) = Y δ P, = M, A,, where Y is the income of the representative resident of the taxing urisdiction and reflects returns on individual holdings of capital, labor and land, or Π Y = r + w +Π A +Π M +Π +, 1 + T where w is the wage rate, is the total amount of capital owned by residents in the initial equilibrium, and is the total amount of labor supply provided by residents of the taxing urisdiction. Note that capital services (and labor services in the interurisdictional mobility case analyzed below) can be exported but the income generated from the exported capital or labor must be spent within the taxing urisdiction, on goods that are either produced within the taxing urisdiction or imported from the other urisdictions. Total property tax revenue (TR) is 14

15 T TR = rt ( M + + ) + ( Π M +Π +Π). 1+ T Revenues are assumed to be spent on a public good that is separable from the individual utility function, with public sector spending on the four goods in the model assumed to be proportional to private sector spending. olving the utility maximization problem yields consumer demands: σ 1 δ σ 1 σ X = Y δ P P. = AM,,, Incidence of the Property Tax To focus on the excise tax effects of the property tax on different types of factor owners, we assume that individuals own only one type of factor (with consumption of each good proportional to income, given the CE utility function). In this case, the change in utility for the suppliers of labor measured in dollar term can be approximated as U λ = w ω ( X P + X P ), where denotes a finite change in variable, λ is marginal utility of income and ω ( =,,,, V, ) are the income shares of each type of factor income in the initial equilibrium ( ω + ω + ω + ω + ω + ω = 1). 10 V Following conventional terminology, this approximation for the utility change associated with the property tax increase can be divided into tax incidence on the sources side, w, and tax incidence on the uses side, ω ( X P + X P ), that is, suppliers of labor could be harmed by any tax-induced reduction in wages and any increases in the prices of the nontradable goods. 10 This approximation is derived by differentiating the utility function for a household whose only source of income is wages, and substituting from the first order conditions and the result of differentiating the budget constraint, taking into account the small open economy assumptions. 15

16 The analogous expressions for the other factor owners are U λ U λ V = ω ( X P + X P ), N = Π ω ( X P + X P ), V N where net land returns are Π =Π / (1 + T ). The total dollar value of the change in utility of the residents of the taxing urisdiction can be approximated as U = Y X P + X P, λ which can also be divided into incidence on the sources side Y and incidence on the uses side ( X P + X P ). The excess burden experienced by local residents due to local use of the property tax is EB = ( U / λ TR)/ TR. II.2 Model Calibration The parameters used in the model simulations are as follows. Production cost shares in each sector are calculated using the U.. Benchmark Input-Output Accounts for Denoting θ i as the factor cost share of factor i in the production of good, these data indicate that the agricultural sector is heavily land intensive ( θ A = 0.17, θ = 0.32, θ = 0.51), the manufacturing sector is heavily labor intensive ( θ M = 0.35, θ X = 0.61, θ VX = 0.04 ), the housing sector is relatively capital intensive ( θ = 0.45, θ = 0.26, θ V = 0.29 ), and the services sector A 11 ee Bureau of Economic Analysis, U.. Benchmark Input-Output Accounts, which are available at 16

17 is heavily labor intensive ( θ = 0.30, θ = 0.67, θ = 0.03 ). The substitution elasticities in both production are based on the values used in Morgan, Mutti and Partridge (1989), and are σ = 1, σ = 0.8, σ = 0.1, and σ = A M Consumption expenditure shares, which are calculated from data from the Bureau of abor tatistics consumer expenditure survey (2006), are β = 0.14, β = 0.16, β = 0.20, and A M β = 0.5, where β is the consumption share of good, and the elasticity of substitution in consumption, also drawn form Morgan, Mutti and Partridge (1989) is σ D = III. IMUATION REUT Overview This section provides the results of several simulations of the model. As explained in more detail in Muthitacharoen and Zodrow (2009) in an analysis of a related but much simpler two-sector, three-factor model, these results are driven by several key factors. Most importantly, the tax drives up production costs in all sectors in which the property tax is imposed (that is, all sectors other than agriculture), especially in the relatively capital intensive sectors, manufacturing and housing. This tends to reduce capital demands and outputs in those sectors, especially in the tradable taxed sector (manufacturing) where forward shifting is impossible; by comparison, these effects are mitigated by forward shifting of the tax burden in the two nontradable goods sectors, housing and services. 12 Although dated, this is the most recent complete set of elasticities of substitution for all four sectors that we could find. There is, however, considerable variation in the empirical literature on these elasticities. For example, Chirinko, Fazzari and Meyer (2004) argue that the aggregate elasticity of substitution of production in the U.. is in this neighborhood of 0.4, and Epple, Gordon and ieg (forthcoming), argue that the elasticity of substitution between land and non-land inputs in the production of housing is in the neighborhood of one. Accordingly, we perform some sensitivity analysis for different values of the substitution elasticities in production below. 17

18 The fact that forward shifting is limited to the nontradable sectors naturally has important implications for the mix of excise tax effects between forward and backward shifting. Because forward shifting in the tradable manufacturing sector is impossible, wages must fall, and by a relatively large amount since land rents in that sector are relatively small. The resulting outflow of labor into the nontradable sectors puts downward pressure on wages in those sectors and thus limits the extent of forward shifting that occurs before a new equilibrium is attained; this is especially true in labor intensive services but also, although to a lesser extent, in the capital intensive nontradable sector, housing. Thus, the traditional view of the property tax, under which consumer prices go up by the full amount of the property tax, will not obtain in the foursector model. abor can also move to the untaxed sector, but that reallocation is limited, given the relatively small amount of labor used in agriculture. Note that another way of thinking about these changes is that, with respect to the two nontradable goods sectors, the potential for backward shifting to labor is more limited in capital intensive housing than in labor intensive services, so that more forward shifting is expected in housing. The imposition of the property tax naturally tends to drive mobile capital out of the taxing urisdiction, although this effect is mitigated somewhat by a reallocation of capital (and labor) to the untaxed sector (agriculture). Finally, net land rents tend to decline in all the taxed sectors, especially in the tradable goods sector where forward shifting is impossible, as land bears the land component of the tax and some of the capital component as well due to backward shifting. This effect is mitigated to the extent that the tax is shifted to labor, or shifted forward in the nontradable sectors, which is more likely to be important in the capital intensive nontradable sector (housing) as described above. Net land rents increase in the untaxed sector as capital and land are reallocated into agriculture. 18

19 The Base Case Turning to the simulations, the first set of results is shown in Tables 1 and 2. The simulations assume a marginal increase in the property tax rate from zero to 5 percent, that is, to r(1+t)=1.05r, or T=0.05, with T/(1+T)= The first case considered (the second column of Table 1) assumes that labor, like land, is completely immobile within each production sector. This benchmark case is the four-sector, three-factor analog to the single sector, two-factor ( capital perfectly mobile and labor/land immobile ) model used to generate the traditional view of full forward shifting of the excise tax effects of the tax (under the appropriate circumstances), and in fact this simulation generates a four-sector version of the traditional view. Although the prices of the manufacturing and agricultural goods are fixed, the prices of housing and services increase significantly, with the slightly larger increase in the price of services, relative to housing, reflecting the relatively greater ease of substituting labor for capital in that sector, which is important enough to offset the higher level of taxes imposed on this capital intensive sector. Wages fall significantly in the manufacturing sector, reflecting the absence of forward shifting possibilities, but wages increase in the two nontradable sectors (housing and services), capturing the benefits of forward shifting. This is especially true for services, where the elasticity of substitution is relatively large, resulting in a relatively large consumer price increase and thus increased relative demand for the fixed factor, and because services is relatively labor intensive, so that its property tax burden is relatively low. A similar pattern is observed for land rents. Although land rents in all taxed sectors decline, partly due to the burden of the land component of the tax, the decline is largest in the tradable goods sector (manufacturing), and smallest in labor-intensive, high substitution elasticity services. In the agriculture, the assumptions of fixed land and labor, coupled with a 19

20 fixed price of capital and property tax exemption, imply that this sector is not affected by the imposition of the tax. As expected, the imposition of the property tax drives capital out of the single taxing urisdiction, with the largest effect in the manufacturing sector where forward shifting is impossible, and a greater outflow from services than from housing, due to the larger production elasticity of substitution in the former sector. The changes in output are of course smaller, but of generally the same relative magnitudes. The incidence results for the simulation in which labor is immobile are presented in the second column of Table 2. All individuals lose from reform but, consistent with the traditional view, this loss is concentrated on the uses side of incidence. In the aggregate, labor income increases, as wage losses in the manufacturing sector are offset by gains in the housing and services sectors. For the reasons discussed above, the income (sources) losses experienced by landowners in the manufacturing and housing sectors are relatively large, and these individuals also experience the largest overall percentage losses, with somewhat smaller reductions in land rents and welfare for owners of land used for services. With labor immobile, the results listed in Table 2 under Tax Burdens on Factor Owners demonstrate the four-sector version of the traditional view of the excise tax effects of the property tax. Indeed, in this benchmark simulation, there is approximately 25 percent overshifting of the tax, as the aggregate change on the sources side is positive, with roughly 125 percent of the tax borne on the uses side of incidence. The burden of the tax on factor owners is only 1.4 percent greater than the amount of revenue raised, indicating that the efficiency costs of a small increase in the property tax from an initial zero tax equilibrium are fairly small. Moreover, the results listed under Tax Burdens on 20

21 Residents illustrate that from the perspective of residents of the taxing urisdiction, the use of the property tax is desirable; that is, the tax burden is about 3 percent less than tax revenues raised due to tax exporting, as virtually all of the burden of the tax borne by land rents in the manufacturing and services sectors is borne by nonresidents. Consider next the incidence of the property tax in the model of intersectoral labor mobility analyzed in this paper, shown in the third column of Tables 1 and 2. In this case, partial labor mobility dramatically changes the distribution of the excise tax effects of the property tax. Wages fall significantly and the increase in the prices of nontradable goods (housing and services) is much more modest, especially for services. The primary driver of these effects is the large outmigration of labor from the taxed tradable goods sector (manufacturing), where forward shifting is impossible and capital outmigration is large. With a fixed overall stock of labor, the resulting inmigration into the other three sectors puts downward pressure on wages and land rents. Because services are labor intensive, the fall in wages implies especially moderate forward shifting of the tax, while this effect is less important but still significant in capital intensive housing. and rents fall in all taxed sectors, but especially in manufacturing, and increase in untaxed agriculture due to inflows of both capital and labor. The property tax drives capital out of all taxed sectors, with the largest effect in manufacturing where there is no forward shifting, and the smallest in housing where substitution possibilities are small and forward shifting is greatest. Because the services sector is labor intensive, however, its output increases slightly, while outputs in the other taxed sectors decline. Both capital and labor are diverted to agriculture, where output increases. These changes imply a very different pattern of incidence than in the case of immobile labor. On net, all factor owners lose from the reform except for the owners of untaxed 21

22 agricultural land, with the largest percentage losses again concentrated among owners of land used in manufacturing and housing, and to a slightly smaller extent, services. Downward pressure on wages coupled with declines in net land rents in all the taxed sectors, coupled with modest increases in the prices of nontradable goods, especially services, imply that virtually all (77.4 percent) of the burden of the excise tax effects of the property tax is borne on the sources side of incidence, and roughly 72 percent of this is borne by labor. Thus, intersectoral labor mobility in the four-sector model is enough to eliminate the over-shifting of the tax that occurred with labor immobile, and indeed results in over three-quarters of the burden of the tax being borne on the sources side of incidence, primarily by labor in marked contrast to the results typical of the traditional view obtained in the model with labor immobile. 13 ensitivity Results There are of course many directions in which a sensitivity analysis of the results presented above might take. We present the results of three sensitivity analyses in this section. First, as noted above, there is considerable uncertainty about many of the parameters in the model, especially the various elasticities of substitution in production. For example, Chirinko, Fazzari and Meyer (2004) note that there is a wide range of estimates of the aggregate elasticity of substitution between capital and labor in production in the literature, and argue that the best estimate is closer to 0.4 than the values of 0.8 for manufacturing and 1.0 for services used in our analysis. Accordingly, we run a simulation of the model in which all of the production elasticities of substitution are set equal to Note that the excise tax effects of the property tax are less regressive in this case than under the traditional view, as both land ownership and wage income are more highly concentrated among the wealthy than consumption. 22

23 The results of this simulation, which are presented in Tables 3 and 4, suggest that lowering the affected elasticities of substitution changes the relative magnitudes of the results, without altering the basic conclusion regarding the distribution of the excise tax effects of the property tax. The primary effects are that, with lower elasticities of substitution, capital outflows are smaller, as are capital reallocations especially from the manufacturing and services sector where the declines in the substitution elasticity are greatest. As a result, less labor leaves manufacturing to migrate to the manufacturing and services sector. With equal elasticities of substitution in manufacturing and services, the differences in the percentage changes in capital and labor employed, and in the declines in net land rents, are largely eliminated in these sectors. owever, as above, services output still increases slightly while housing output declines, and the percentage increase in housing prices is still considerably greater than the small increase in the price of services. The utility changes experienced by the owners of the various factors are very little changed across the two simulations. Most importantly for our purposes, the distribution of the excise tax effects of the property tax is virtually unchanged. When labor is immobile, the foursector version of the traditional view again obtains, with overshifting of the property tax into higher prices of housing and services by about 26%. With intersectoral labor mobility, however, 74% of the burden of the tax is borne on the sources side of incidence, with labor bearing roughly 76% of this burden. ocal residents again benefit to roughly the same extent from the imposition of the property tax due to the exporting of the burden on land used in manufacturing and services. Alternatively, one might argue that the elasticity of substitution in housing used in the base case simulation (0.4) is too low. Albouy (2006) notes that estimates in the empirical 23

24 literature range from , and Epple, Gordon and ieg (forthcoming) argue that the best estimate is in the neighborhood of one. Accordingly, we run a simulation for the case in which the elasticity of substitution in housing production is equal to 0.8, with all of the other elasticities the same as in the base case. The results of this simulation are presented in Tables 5 and 6. In this case, with a higher elasticity of substitution in the housing sector, capital outflows from the taxing urisdiction are slightly greater, as are the capital and labor outflows from the manufacturing sector. With nearly equal elasticities of substitution in the two tax nontradable goods sectors, the percentage reductions in capital and increases in labor (and the decline in net land rents) in these sectors are very similar. With slightly larger factor flows, the reductions in manufacturing and housing output and the increase in services output are slightly larger. With respect to the incidence of the property tax increase, the utility changes experienced by the various factor owners are again roughly similar. Moreover, the distribution of the excise tax effects of the property tax is again virtually unchanged. When labor is immobile, the foursector version of the traditional view again obtains, with overshifting of the property tax into higher prices of housing and services by about 31%. In marked contrast, with intersectoral labor mobility, 72% of the burden of the tax is borne on the sources side of incidence, with labor bearing roughly 74% of this burden. The gain to local residents from exporting some of the property tax burden is little changed, relative to either of the previous simulations. Finally, we consider the sensitivity of the results to a different variant of partial labor mobility. The analysis thus far has indicated that in the case of residence immobility coupled with intersectoral labor mobility, labor bears a significant portion of the sources side incidence of the property tax, which is in turn a significant fraction of the total burden. uppose instead that labor is mobile not only across production sectors, but is also sufficiently mobile across 24

25 urisdictions so that the net wage is unchanged (while maintaining its residence in the taxing urisdiction). This scenario might reflect a moderately large suburb increasing its property tax rate while its residents can relatively costlessly change employment, but not residence, by taking obs in neighboring suburbs. In this case, the wage rate is fixed, while the amount of labor employed in the taxing urisdiction (but not the number of residents) can vary in response to the tax. 14 The results of this simulation are presented in Tables 7 and 8. Not surprisingly, since backward shifting to labor is eliminated by assumption, the excise tax effects of the property tax are significantly altered. With increased mobility of labor, labor outmigration from the taxed tradable goods sector (manufacturing) more than triples, with some of this labor going to the taxed nontradable goods sectors (housing and services) but less than in the case of only intersectoral labor mobility since employment outside the urisdiction is now allowed, and a significant amount of labor is no longer employed in the taxing urisdiction. Capital outflows from manufacturing are also far more pronounced, and the total capital stock declines somewhat more. The extent of capital outmigration from the housing and services sectors declines, however, reflecting the larger increases in forward shifting of the tax in these two sectors coupled with constant wage costs. For the same reason, the relative degree of forward shifting is greater in the labor intensive services sector, which no longer benefits from a decline in wages. Output declines much more markedly in manufacturing, remains unchanged in agriculture (since land, the prices of capital and labor, and the output price are all fixed), and again declines slightly in housing and increases slightly in services. 14 abor is of course still only partially mobile, as full mobility would require a change of residence if household utility fell below an exogenously determined level. 25

26 Turning to the incidence of the tax, recall that when labor was immobile, the property tax was overshifted by 25 percent, while with intersectoral labor mobility, only 23 percent of the tax was shifted forward. With interurisdictional as well as intersectoral labor mobility implying no backward shifting to labor, the extent of forward shifting is increases dramatically, to 70 percent, although it is still significantly less than in the case of immobile labor. The sources side of incidence accounts for the remaining 30 percent of the tax, which is borne entirely by the owners of land in the taxed sectors, especially the owners of land in the capital intensive manufacturing and housing sectors. The increase in labor mobility in these simulations is thus sufficient to restore much of the traditional view of the excise tax effects of the property tax, although the degree of forward shifting is still considerably less than in the four-sector model with labor immobile. Finally, with labor mobile, the excess burden of the tax on factor owners increases to 3.43 percent (relative to 2.25 percent with intersectoral labor mobility), and the gain to residents from the use of the property tax increases by a factor of five, reflecting a near doubling of the exportable tax burden on the owners of land used in manufacturing and services. IV. CONCUION This paper analyzes the excise tax effects of a general property tax from the perspective of a small open economy facing a perfectly elastic supply of capital, focusing on the mix of forward shifting to consumers and backward shifting to labor and landowners. The model utilized differs from most of the other models that have appeared in the property tax literature in that: (1) the property tax is applied in a four-sector model with three taxed sectors, corresponding to manufacturing which produces a tradable good and housing and services which produce nontradable goods, and a tax-exempt agricultural sector, (2) labor is partially mobile, as it is perfectly mobile across production sectors but fixed in total supply, while land is fixed in each 26