The typical general sales tax in the United States is riddled

Transcription

1 Popular Substitution Effects Popular Substitution Effects: Excess Burden Estimates For General Sales Taxes Abstract - In this paper, a partial equilibrium model is used to evaluate the efficiency of some popular sales tax structures, including structures with pyramiding from levies on production inputs. The results generally support expectations inefficiency increases when retail exemptions are large and the general rate is high but two findings are surprising. First, the input taxes considered here fall heavily on price inelastic and exempt goods and actually improve tax efficiency. Second, equal yield sales tax systems can have different median tax payments and a large majority of households may prefer exemptions and high rates. Richard R. Hawkins Department of Marketing and Economics, University of West Florida, Pensacola, Florida 3254 National Tax Journal Vol. LV, No. 4 December 22 INTRODUCTION The typical general sales tax in the United States is riddled with commodity exemptions. Food for home consumption (hereafter home food), residential utilities and services are all examples of consumption items that are spared from an otherwise uniform general sales tax rate. Since these exemptions introduce substitution effects between commodities, general sales taxes would appear to introduce inefficiencies as consumers are induced to purchase more exempt commodities and less taxed commodities. However, the size of this inefficiency is complicated by varying price elasticities. If policymakers exempt relatively price elastic commodities, they could reduce the behavioral response to a sales tax households could face relatively low prices on the goods where demand effects are large. A tax structure with this type of exemption pattern could approximate the efficiency condition discovered by Ramsey (927). Until recently, evidence on inefficiency from incomplete sales taxation was scarce (Murray, 997). General equilibrium models, including that of Ballard, Shoven, and Whalley (985) and Jorgenson and Yun (99), had been used to examine the inefficiency from sales taxes that do not reach savings or leisure, but the effects of many popular exemptions were not addressed until Baum (998) used a general equilibrium model to examine the home food exemption and found that exemption produces a small efficiency loss. Iorwerth and Whalley (22) examined another side of the home food exemption: focusing on the production shift from restaurant meals to home meals and finding inefficiency in this substitution. 755

2 NATIONAL TAX JOURNAL Home food is certainly a common target for sales tax exemptions receiving some preferential treatment in 28 of the 45 U.S. states that have a sales tax but efficiency conclusions need to be based on a more complete tax structure. This study attempts to advance knowledge about commodity taxation with a partial equilibrium model. The cost function from the Almost Ideal Demand System (AIDS) is used to examine household welfare under eight equal yield sales taxes, four of which include some taxation on business inputs. This analysis contributes to our understanding of tax efficiency in practice and to the popularity of particular tax structures. The paper is organized as follows. In the second section, the partial equilibrium approach with the AIDS is introduced. The data and demand system estimates are discussed in the third section, along with an important price adjustment technique. The simulated sales tax structures are the subject of the fourth section. The fifth section includes the household welfare estimates. Given some of the surprising results, the importance of household characteristics is addressed in the sixth section. A brief conclusion is contained in the seventh section and the demographic parameters from the AIDS appear as a table in the appendix. A PARTIAL EQUILIBRIUM APPROACH TO THE EXCESS BURDEN Excess burden estimates are typically calculated with a general equilibrium model in order to capture factor market effects that follow the introduction of a tax. An important detail for a sales tax efficiency evaluation, however, is the size of the substitution effects that should be introduced by a narrow base and a relatively high rate. For example, if a state chooses to exempt home food and services, and to tax other purchases more heavily, an efficiency evaluation should capture the increase in exempt commodity demand that follows the policy. A partial equilibrium approach is used here to measure these detailed demand effects. There are, however, three clear disadvantages to a partial equilibrium approach. First, as mentioned, this model will miss the full Harberger triangle including both consumer spending effects and supplier responses that is measured with a general equilibrium model. The latter approach can capture the effect of a sales tax on labor supply decisions, exempt commodity production decisions and capital investment decisions; potential inefficiencies that are not considered here. Second, with a partial equilibrium approach, the incidence of a sales tax has to be assumed rather than determined. In this study, full forward shifting to the final consumer has been chosen, but this assumption is not tested. 2 Third, while demand systems are generally consistent with microeconomic theory, they are not compatible with all types of empirical phenomena. The Almost Ideal Demand System (AIDS) described in this section, is a very popular demand specification, but more elaborate and more difficult to estimate specifications exist. For example, the relatively new Quadratic Almost Ideal Demand System allows for nonlinear Engel curves and this flexibility could be important. With that model, Banks, Blundell, and Lewbel (997) find that nonlinear effects are important in measuring the welfare effects for a tax on clothing. The relative performance of more elaborate systems on broader consumption taxes is unclear. See Hines (999) for a review of the general equilibrium approach to measuring excess burdens. 2 Full forward shifting is a simple assumption for what is probably a complex set of price responses. This incidence model, however, is not rejected by Poterba (996). Besley and Rosen (999) do reject the assumption for some commodities, finding over shifting of the sales tax in these cases. 756

3 Popular Substitution Effects The AIDS was introduced by Deaton and Muellbauer (98) as derivatives from the logarithm of a cost function [] log c(u,p) = a + Σ k α k log(p k ) 2 β + uβ Π k p k k + Σ k Σ j γ kj log(p k ) log(p j ) where α, β, and γ are parameters. The u variable represents a specific utility level and p k is the price of commodity k. The budget share equations are given by [2] e k (x,p) = α k + Σ j γ kj log(p j ) x + β k log ( P ) where e k represents the household s budget share for commodity k and x is the sum of total spending on all commodities (generally interpreted as household income). In the final term of equation [2], P represents a price index which is calculated with [3] logp = α + Σ k α k log(p k ) + Σ j Σ k γ kj log(p k )log(p j ) 2 For estimation purposes, three standard restrictions adding up, homogeneity, and symmetry are applied to the coefficients on prices and income to insure consistency with microeconomic theory. They are satisfied with the sum of the alphas equaling, the sum of the betas equaling, the sum of any row (or any column) of gamma coefficients also equaling zero and symmetric price responses, γ ij = γ ji. The latter restriction, Slutsky symmetry, is particularly valuable for this model; it greatly reduces the number of unique gamma coefficients that must be estimated. Once the model is estimated, the parameters are used to calculate a cost estimate (equation []) for prices that include a new, percent tax on every commodity. This estimate is the benchmark and is referred to with c(u, p ). The parameters are then used to conduct simulations, where tax rates are calculated for alternative, equal yield sales tax structures. These structures differ in the number of retail exemptions and the existence of levies on select retailer purchases (described below). The latter taxes lead to an increase in the gross price of the good, upon which the retail sales tax can be applied. Given the estimates and the calculated tax rates, p represents the final prices and c(u, p ) is the cost function for any alternative to the broad, one percent tax that produced p. The average cost differential, c(u, p ) minus c(u, p ), is divided by average revenue across households to calculate the excess burden. For example, if two structures generate $5 in new tax revenue and c(u, p ) is one dollar larger than c(u, p ), the excess burden is 2 percent ($/$5). 3 ESTIMATING RELATIVE PRICE AND INCOME EFFECTS For this study, demand equations are estimated for eight commodities: food for home consumption, utilities, gasoline, alcohol, tobacco, automobiles and furnishings, services, and generally taxable nondurables. 4 Demand data are taken from the Consumer Expenditure Survey between January, 99 and April, 993. The national Consumer Price Index (CPI) is used to measure prices. The in- 3 An important consequence of this cost function comparison is the relative weight on high spending households. For example, if one only examines the tax efficiency on the lowest and highest income households, a c(u, p ) that is 5 percent higher for the poor and percent lower for the affluent will generate a negative excess burden. 4 The generally taxable nondurables include food away from home, reading materials, apparel and apparel services, and entertainment spending. 757

4 NATIONAL TAX JOURNAL dex includes a series for each of the components within the eight composite commodities, important for constructing a price for each composite commodity, and is designed to reflect the average level of taxation in the U.S. In doing so, however, the CPI understates (overstates) the price of a commodity in a high (low) tax jurisdiction. One piece of household information, however, allows for some location specific price adjustment. The U.S. Bureau of Labor Statistics has provided state identifiers for most of the urban households in the Consumer Expenditure Survey. This limited geographic variable, and the assumption of full forward shifting, are used to adjust the CPI toward household specific prices. The price adjustment process includes two steps. First, the average sales tax rate is calculated for each commodity. Since CPI data are only gathered in some metropolitan areas, this average is only calculated for the largest central city in each of the 27 states (plus the District of Columbia) where price information is collected. 5 In the second step, household specific prices are calculated with adjustments to the average from step one. In January of 992, for example, the average home food sales tax rate across CPI areas is.8 percent. In states where food is taxed (exempt), the CPI food price is understated (overstated). The food price in jurisdiction j for that period, is adjusted according to where p food is the price reported in the Consumer Price Index. Estimates for the actual tax rate in a given jurisdiction, for a given commodity, are taken from the Advisory Commission on Intergovernmental Relations (various years). 6 There are three limitations to this price adjustment process. First, as mentioned, the full forward shifting assumption certainly does not represent every commodity price effect in every state. Second, sales tax structures are the only source of inter area price variation in this study. The approach is probably more accurate for competitive goods (e.g., computers and gasoline), and less so for services, where regional factors may be important. Finally, households in many large metropolitan areas can shop across jurisdictions with different sales tax rates, but a demand system inherently ignores the potential for border shopping as each household must face one price for each commodity. Given that the largest metropolitan areas in New York and Illinois cross several states, households in these states were deleted from the sample. In addition, no households were identified for the District of Columbia. The time period covered by the data included several sales tax rate changes. For example, the general rate in North Carolina increased in July, 99. These changes were identified with the U.S. Bureau of the Census (various years) and the price adjustments were changed accordingly. The income and price parameter estimates appear in Table and the demographic coefficients, which shift the α i paτ j [4] p food, j = p.8 food 5 The states are California, Colorado, Connecticut, Florida, Georgia, Hawaii, Illinois, Indiana, Kansas, Kentucky, Maine, Maryland, Massachusetts, Michigan, Minnesota, Missouri, New Hampshire, New Jersey, New York, Ohio, Oregon, Pennsylvania, Texas, Virginia, Washington, West Virginia, and Wisconsin. 6 The price adjustment is more complicated for alcohol, gasoline, and tobacco since states use a combination of specific taxes and ad valorem taxes. For these commodities, the total effective tax rate is calculated with wholesale prices and full forward shifting of both specific sales taxes and general sales taxes (also from Advisory Commission on Intergovernmental Relations (various years)). The price adjustment then follows the calculation above, based on our estimate of the average effective commodity tax rate in the CPI states. 758

5 Popular Substitution Effects TABLE SELECTED PARAMETER ESTIMATES FROM THE ALMOST IDEAL DEMAND SYSTEM (Standard Errors in Parenthesis) Demand Determinant Demand Equation γ i γ i2 γ i3 γ i4 γ i5 Generally Taxable Nondurables.75 (.35) Services.7 (.25).357 (.33) (.).8 (.9).5 (.4) (.7).73 (.3). (.2).64 (.3).4.63 (.3).74 (.).8 (.24).4 (.2).38 (.28) Home Food.36 (.36).464 (.3).44 (.5).38 (.36).37 (.46) Utilities.324 (.26).562 (.2).32 (.4).37 (.36).7 (.38) Automobiles and Furnishings Demand Determinant Demand Equation γ i6 γ i7 γ i8 α i β i Generally Taxable Nondurables (.93).27 (.9) Services (.4).27 (.2) (.88).9 (.2) (.25).33 (.3) (.97).98 (.2) Home Food.384 (.4) (.262).843 (.5) Utilities.335 (.89).67 (.94).8.35 (.87).55 (.5) Automobiles and Furnishings Notes: Estimates are for the constrained model and standard errors are not listed for calculated parameters. Generally taxable nondurables include food away from home, reading materials, apparel and apparel services, and entertainment spending. Fourteen additional household factors were used to allow shifts in the α parameters and these coefficients appear in the Appendix. rameters, appear in the Appendix. The more familiar own price and income elasticity estimates appear in Table 2. For the eight commodities, six of the own price elasticity estimates are negative and significant. 7 The income elasticity estimates range from.3 (tobacco) to 2.62 (automobiles and furnishings) and all are signifi- 7 For alcohol and utilities, the elasticity estimates are not significantly different from zero. 759

6 Demand Equation Generally Taxable Nondurables NATIONAL TAX JOURNAL TABLE 2 INCOME AND OWN PRICE ELASTICITY ESTIMATES FROM THE ALMOST IDEAL DEMAND SYSTEM (Standard Errors in Parenthesis) Average Budget Share.7 Income.7 (.55) Elasticity Uncompensated, Own Price.35* (.263) Compensated, Own Price.6* (.26) Services.44.6 (.27).72* (.8).24* (.8)..8 (.88).55 (.3692).54 (.3692).4.7 (.7).83* (.733).8* (.733)..3 (.7).7* (.988).7* (.988) Home Food.3.47 (.32).58* (.4).5* (.6) Utilities.8.44 (.47).6 (.6). (.58) Automobiles and Furnishings (.6) Significant from. at the 95 percent confidence level. * Significant at the 95 percent confidence level..33* (.294).6* (.294) Notes: Average budget share is calculated on aggregate spending. All estimates are evaluated at the sample mean. cantly different from one. 8 Income elastic commodities also include generally taxable nondurables, services, and alcohol. Inelastic demand is important when a state enacts a new exemption or increases a tax rate. If home food is the first exemption, final food spending will fall and taxable spending on everything else will increase. 9 With additional exemptions, one cannot be sure if the tax base will increase as the household might choose more of the other exempt commodities. Finally, if policymakers increase the tax rate, final spending on taxable goods should also increase, but some of the original purchase quantity should be lost to the numerous cross price effects that follow the changing prices. The compensated price elasticities also appear in Table 2. Under the Ramsey rule, these estimates indicate that gasoline and tobacco should receive the lowest tax rates. However, the range of compensated elasticities is small and a full exemption could be too large. Finally, while this paper focuses on the general sales tax, the Ramsey rule applies to the set of all commodity taxes. SIMULATED TAX STRUCTURES AND CONSUMER SPENDING As mentioned, the first tax structure is the benchmark, a destination based tax on spending with no consumption exemptions. The second through fourth tax struc- 8 Deaton and Muellbauer (98) also found highly income elastic responses for transport and communication and for other goods. 9 In other words, spending on everything else is a complement to a price inelastic commodity. 76

7 Popular Substitution Effects tures are also only applied to consumer purchases, but with sequential exemptions for home food and many services (tax base 2), home food, services and residential utilities (base 3), and the above mentioned commodities plus gasoline (base 4). While tax base reaches all current consumption spending, this total is not as large as the annual expenditures reported by the U.S. Bureau of Labor Statistics. An important reason for the difference is the exclusion of personal insurance and pension and cash contributions. It should be noted, however, that the exclusion means potentially inefficiencies from exempt savings, insurance payments, and charitable contributions could be missed with the new, hypothetical taxes. Input purchases are another source of sales tax revenue in the United States and the second set of tax structures is designed to follow the estimates of Ring (999) and the tax system descriptions of Due and Mikesell (994). These structures feature the same consumption exemption pattern, from base to base 4, but the tax is also applied to select input purchases by sellers of goods and services. For this paper, input purchase levies are assumed to be collected on retailer purchases of new capital and on three operating expense categories for these retailers: communication equipment, utilities, and office supplies. Each category was chosen by the process of elimination as input sales taxes are generally not collected on raw materials, labor, and firm to firm services. In response to the input purchase levies, retailers are assumed to pass these taxes onto consumers, according to the ratio of taxable input spending to revenue. For example, if 9 percent of home food retailer revenues are subject to the input sales taxes, the gross price of food increases by 9 percent of the stated tax rate and the effective rate to the consumer is greater than the stated rate. When home food is exempt, the effective rate is 9 percent of the stated rate on other goods. When food is taxed, the rate is slightly higher than 9 percent of the stated rate. The degree of pyramiding for each consumption commodity appears in Table 3 and the effect of this pyramiding on the commodity tax rates appears in Table 4. The reader should note that for calculation purchases, a retailer SIC has been assigned for the individual goods and services. The ratio of taxable spending (by the retailer) to revenue determines the size of the pass through. These input burdens vary from.9 percent for automobiles to over 45 percent for shelter. Industry revenue totals, as well as the capital and operating expenditure totals (in Table 3), are generally taken from the U.S. Bureau of the Census (996a and 996b). Final tax rates for the equal yield tax structures (Table 4) are based on the expected household response to the changing prices, but the exempt commodities in the top half of the table actually carry a zero tax rate. The bottom half includes lower explicit tax rates as collections are also derived from the retailer purchases. 2 Note that each commodity in the structures with pyramiding carries a non zero tax rate, but the lowest effective rate is.7 percent. Due and Mikesell (994). If the unadjusted price is $, the tax rate is 5 percent and 9 percent of input purchases are also taxed, the final price is roughly $5.47 and the effective rate is 5.47 percent or the product of the explicit rate and one plus the taxable input share. 2 The treatment of business purchases, however, does not change across the structures in the bottom half of Table 2. This is consistent with sales tax policy in many states, where states with a residential utility exemption continue to tax non residential purchases (Federation of Tax Administrators, 997). 76

8 NATIONAL TAX JOURNAL Commodity Class Generally Taxable Nondurables TABLE 3 DEGREE OF PYRAMIDING FROM SALES TAX ON SELECT INPUT PURCHASES Taxable Inputs as a Share of Total Revenue 6.6 Component Commodities Food Away From Home Housekeeping Supplies Apparel and Apparel Services Fees and Admissions Television, Radios, and Sound Equipment Other Goods Taxable Inputs as a Share of Total Revenue Retailer SIC other 59 Services 25.5 Shelter Other Lodging Household Operations Other Vehicle Expenses Health Care Personal Care (excluding 54) Home Food 3.7 Home Food Utilities 4.7 Electricity Automobiles and Furnishings 2.6 Furnishings and Equipment Automobiles Notes: For most retailers of goods and services, the input sales tax is applied to spending on capital equipment, utilities, office supplies, and telephone and other communications. See article text for specifics on housing and utilities. The Other 59 SIC includes miscellaneous retail stores excluding drug and liquor stores. Since other lodging is not taxed by the local jurisdiction, the commodity is included in the services class. SOURCE: Author s calculations on U.S. Bureau of the Census (996a and 996b) and Federal Energy Regulatory Commission (997) Given the retailer purchase orientation of the pyramiding constructed here, it is not surprising that the input taxes only reach a maximum of 32.5 percent of all sales tax collections (bottom row of Table 4), a range generally below the estimates in Ring (999). For these hypothetical sales taxes, the structures are not applied to purchases by the producer or distributor of goods since these firms are likely to be located elsewhere. Therefore, when the household buys a soft drink, local tax is collected on equipment and select other purchases by the retailer, but not on purchases in the course of production and distribution of the product. 3 One important limitation with this approach is the absence of any local sales and use taxation on the export economy. The size and composition of this economy varies across jurisdictions and collections on these commodities is speculative, but the relationship between the export economy and tax efficiency in a jurisdiction is an important topic for future research. 4 3 This does not mean that wholesale prices are free of any sales taxes, rather they are free of the local sales tax. 4 For example, a jurisdiction with a bedroom suburban orientation and a jurisdiction with relatively immobile exporters could desire different tax structure choices. 762

9 Popular Substitution Effects TABLE 4 EXPLICIT AND EFFECTIVE TAX RATES FOR TAX STRUCTURES WITHOUT AND WITH PYRAMIDING Explicit Rate on Consumer Purchases (No Exemptions) % Tax Base 2 (Most Services and Food) Without Pyramiding 2.38% 3 (Most Services, Food and Utilities) 2.9% 4 (Most Services, Food, Utilities and ) 3.3% Generally Taxable Nondurables Services Home Food Utilities Automobiles and Furnishings % 2.38% % % With Pyramiding Explicit Rate on Consumer Purchases.87%.76% 2.3% 2.2% Generally Taxable Nondurables Services Home Food Utilities Automobiles and Furnishings.93% % % % Exhibit: Pyramided Revenue as a Share of Total Revenue 2.8% 25.9% 29.9% 32.6% In constructing the input sales taxes, shelter and utilities were problematic. For utilities, financial data for the industry is limited. The industry was not subject to a complete 992 Census and expenditures for construction & acquisition of plant are the only inputs taxed. This data, also for 992, is taken from the cash flow summary for major investor owned electric utilities in the United States (Federal Energy Regulatory Commission, 997). 5 The price of shelter is assumed to reflect sales taxes on construction inputs, but there is reason to believe the input share (within the services commodity in Table 3) is overstated. Land rent is not separately reported in the Consumer Expenditure Survey and the shelter ratio includes an input tax on the land portion of the payment. The overstatement could be greatest when the consumer has purchased an older house with few capital improvements. 6 5 The available cash flow data do not separate land acquisitions and the input tax share unintentionally includes these purchases. But in total, the utility industry could be under taxed in this model since spending data on other factors (e.g., communications) is not included. 6 Another problem area is the circular nature of contractor to contractor payments measured in U.S. Bureau of the Census (999). Since the interest here is in the share of a construction project that is subject to the sales tax, sub contractors are assumed to have the same ratio for material spending to revenue. 763

10 NATIONAL TAX JOURNAL EXCESS BURDEN ESTIMATES AND EXEMPTION POPULARITY Excess burden estimates for these new, relatively small sales taxes can be found in the third column of Table 5. In the top half of the table, for example, the second structure includes exemptions for highly inelastic services and moderately inelastic food. The second structure produces the same revenue and a cost estimate that is roughly $5 higher. This differential, divided by per household revenue of $63.5 produces an excess burden of 23.3 percent. The third and fourth tax structures include additional exemptions for price inelastic commodities and the excess burden increases to 3.6 percent and 38.5 percent. The excess burden estimates indicate that inter commodity substitution effects are an important source of efficiency loss in the typical sales tax and the burden increases when exemptions and rate increases are added. In the bottom half of the table, however, the pyramiding of taxes on retailer purchases brings a surprising effect. These input levies fall relatively heavily on inelastic and frequently exempt commodities (e.g., services and utilities), and the addition of input taxation reduces the rate differential between taxed and exempt commodities, reducing the excess burden. 7 Under tax base 4, for example, the cost estimate with pyramiding is $6,43.38 and this is lower TABLE 5 EXCESS BURDEN ESTIMATES Tax Base (Exemptions) Explicit Sales Tax Rate Cost Estimate Excess Burden Estimate Median Household Liability Households With A Lower (Higher) Tax Burden Sales Tax Structures Without Pyramiding (None).% $ $ (Most Services and Home Food) 3 (Most Services, Home Food, and Utilities) 4 (Most Services, Home Food, Utilities and ) (None) 2 (Most Services and Home Food) 3 (Most Services, Home Food, and Utilities) 4 (Most Services, Home Food, Utilities, and ) Sales Tax Structures With Pyramiding.87% $ % % $ ,525 (8399) 8,895 (729) 9,58 (6776),2 (4,73) 7,78 (844) 8,987 (6937) 9,82 (6742) Notes: Each excess burden estimate is calculated with the difference in the cost estimate (relative to tax structure without pyramiding) divided by the average tax liability. Cost estimates and median liabilities are based on quarterly reporting. 7 This type of efficiency was introduced in Stiglitz and Dasgupta (97). 764

11 Popular Substitution Effects than the estimate for tax base 4 without pyramiding. In fact, the excess burden falls from 38.5 percent to 26.7 percent. Since the indirect taxes introduced above only reach retailer purchases, three additional simulations were conducted under the assumption that the system understates the level of pyramiding. The simulations included () increasing the estimates in Table 3 by. for each commodity, (2).2 for each commodity and (3).2 for automobiles and furnishings and generally taxable nondurables. In each case, the explicit tax rate fell and the excess burden estimates (not shown) were close to the estimates for systems with pyramiding in Table 5. 8 In other words, the result that pyramiding improves sales tax efficiency does not appear to be limited to the pyramiding structure specified in Table 3. The estimates in Table 5 are smaller than many of the excess burden estimates for additional income taxation that appear in Ballard, Shoven, and Whalley (985) and Jorgenson and Yun (99). Three points, however, need to be clear in order place this study in the literature. First, many estimates for the average excess burden of income taxation (summarized in Jorgenson and Yun, 99) are smaller than the estimates here, meaning that a change in the sales tax structure could be just as important as replacing an income tax with a non distorting tax. Second, it appears that the composition of the tax base can generate roughly the same excess burden as that of additional broad based sales taxation in a general equilibrium setting, the above mentioned studies finding marginal excess burdens of.388 and Finally, this study only captures gains in efficiency for changing the sales tax structure across relatively low rates. The substitution of tax base 4 for tax base at a higher rate should produce a larger excess burden as the new tax rates move further up the compensated demand curves for taxed goods. In general, the excess burden estimates in Table 5 indicates that tax exemption and rate driven substitution effects are an important source of welfare loss, even for small sales tax rates. There are, however, three reasonable arguments for concluding any discussion of sales tax efficiency at this point. First, the excess burden for the sales tax may be acceptable vertical equity concerns generally dominate sales tax policy debates while efficiency debates are rare (Murray, 997). Second, tax structures also apply to visitor spending and to some portion of inputs for export commodities. Finally, Fox (992) identifies a considerable debate in expanding a sales tax base to services, where administration and compliance costs are uncertain. The above arguments, however, neglect two empirical details in the right side of Table 5. First, thousands of households fare better under the structures with exemptions. With tax structure 2, for example, more than 7, households face a lower tax payment while 8, have a higher payment. Similarly, the tax payment for the median household is lower when the structure includes exemptions and a higher tax rate. Both of these features are independent of whether the structure reaches input purchases. It is possible that households actually prefer structures with slightly higher personal tax payments. In fact, the average excess burden for the households with a lower tax payment (not shown) under structure 2 is very small, but positive. 8 For the simulation where the taxable input share values for each commodity were increased by.2, pyramided revenue reached between 45 percent (tax base 2) and 53 percent (tax base 4) of total revenue estimates that are consistent with those in Ring (999). 9 The inter commodity substitution effects captured with this model are larger than the findings of Baum (998), but that paper only examined the effect of a food exemption. 765

12 NATIONAL TAX JOURNAL Under tax structure, however, a policymaker can tell a majority of taxpayers an alternative exists that will lower their taxes, without affecting government spending. This policy alternative could increase a microeconomic cost function, but a large coalition of voters, including the household with the median liability, has an opportunity to decrease the group share of total tax revenue. Public choice theory indicates that a majority of voters will support one of the tax bases with exemptions and a higher rate. SOME HOUSEHOLD TYPES Clearly, estimates based on a demand model for an average household fail to capture the inter household differences in budget shares and tax payments. For example, a large household spends a larger share on food at home and a smaller share on alcohol and the treatment of these commodities under the tax structure will be important to that household. In general, one might suspect that households who fare better under exemption laden tax structures are somehow different from households who pay less under a broad base (and a low rate). Table 6 has been constructed to examine evidence on the suspicion. The table includes tax payment estimates for three hypothetical households that only vary in household characteristics, with the exception of the single adult who also differs in choosing to rent. In other words, the households live in the same region, have the same income and face the same prices. If the sales tax applies a percent rate to all consumer spending, each household pays a tax of $63.5. In the table, household type is important to the tax payment tax structure relationship. For example, the one adult household does not disproportionately benefit from the exemptions here and faces the lowest tax burden under a broad base and a low rate. Conversely, a family of four can reduce their tax bill by about 5 percent if the jurisdiction chooses the narrowest base/highest rate. 2 Given the Two Adults, Two Children TABLE 6 QUARTERLY TAX LIABILITY PROJECTIONS FOR SELECT HOUSEHOLD TYPES (No Exemptions) $ (Most Services and Home Food) Tax Base 3 (Most Services, Home Food, and Utilities) Sales Tax Structures Without Pyramiding $6.69 $ (Most Services, Home Food, Utilities, and ) $54.8 One Adult Two Senior Citizens Sales Tax Structures With Pyramiding 54.7 Two Adults, Two Children $63.34 $6.82 $57.98 $56.7 One Adult Two Senior Citizens Notes: The household with two adults and two children has a reference person between the ages of 35 and 44 and does not rent. The one adult, middle household, is between the ages of 25 and 34 and rents. The two seniors household does not rent. All households live in the Midwest, face an unemployment rate of 6 percent and spend $6,43 in the quarter. 2 This comparison is for the structures without pyramiding. With pyramiding, the savings is roughly percent. 766

13 Popular Substitution Effects large coefficient on the seniors as a share of the household parameter, δ, with respect to services and the negative coefficient on δ with respect to gasoline (both in the Appendix), it is not surprising that a household with two elderly members receives the best treatment under tax base 3. Finally, the estimates change if the jurisdiction moves to structures with pyramiding, but the rank order from tax structure to tax structure 4 does not. The tax payment estimates in Table 6 also suggest that tax structure debates are affected by the characteristics of the population. For example, proposals for generous exemptions and a higher rate will appeal to the family of four and not to the single adult. It is not entirely clear, however, whether state tax bases will gravitate toward structure 4. The family of four could be a majority in the jurisdiction, a vocal minority or a group more interested in some other tax criteria. 2 The relationship between household types and sales tax policy is another important area for future research. CONCLUSION General sales tax structures with high rates and generous exemptions have been thought to create inefficiencies. This hypothesis is based on the assumption of sizable inter commodity substitution effects, but evidence in the literature has been scarce. The extension of sales taxes to producer inputs and the popularity of tax structures that should introduce distortions have called into question the accuracy and importance of the excess burden hypothesis. The estimates in this paper, from a partial equilibrium model applied to several general sales tax structures, indicate that exemptions and high rates do introduce inefficiencies. The excess burden the reduction in economic welfare beyond the revenue raised ranges between 7 and 39 percent. The value increases when the tax structure includes more generous exemptions and a higher tax rate. These estimates fall below many of the welfare costs for additional income taxation, but the hypothetical taxes examined here are roughly as damaging as the average excess burden of income taxation and the additional excess burden of higher sales taxation with a general equilibrium approach. When the tax structures are expanded to include select purchases by retailers, the excess burden estimate for each set of exemptions declines for two reasons. First, the input levies are disproportionately large for inelastic commodities that are commonly exempt, e.g., services and utilities. Second, each structure considered here raises an equal yield and the input levies are offset by a reduction in the explicit sales tax rate. Calculations on the empirical sample reveal an important justification for sales tax structures that introduce inefficiencies. The combination of exemptions and a higher rate generally allow a majority of households to pay less taxes. The application of the tax structures to some hypothetical household types illustrates the importance of consumption differences across the population of a jurisdiction. For example, a family of four should have a lower tax liability under generous exemptions and the highest tax rate. Ultimately, variations in tax liability across household types, for equal yield tax structures, introduce an important public choice component to a jurisdiction s decision among sales tax structures. 2 Further complicating this hypothesis is the fact that other tax bases exist (not considered here) that would create an even lower liability for the family of four and that across observations, considerable spending variation exists even when household characteristics are held constant, e.g., some families of four spend relatively little on services and home food. 767

14 NATIONAL TAX JOURNAL Acknowledgments The author wishes to thank Nestor Arguea, Matt Murray, Mary Beth Walker, and two anonymous referees. REFERENCES Advisory Commission on Intergovernmental Relations. Significant Features of Fiscal Federalism Volume : Budget Processes and Tax Systems. Washington D.C.: Advisory Commission on Intergovernmental Relations, Various years. Ballard, Charles L., John B. Shoven, and John Whalley. General Equilibrium Computations of the Marginal Welfare Costs of Taxes in the United States. American Economic Review 75 No. (March, 985): Banks, James, Richard Blundell, and Arthur Lewbel. Quadratic Engel Curves and Consumer Demand. Review of Economics and Statistics 79 No. 438 (November, 997): Baum, Donald N. Economic Effects of Eliminating the Sales Tax Exemption for Food: an Applied General Equilibrium Analysis. Journal of Economics 24 No. (Number, 998): Besley, Timothy J., and Harvey S. Rosen. Sales Taxes and Prices: an Empirical Analysis. National Tax Journal 52 No. 2 (June, 999): Deaton, Angus, and John Muellbauer. An Almost Ideal Demand System. American Economic Review 7 No. 3 (June, 98): Due, John F., and John L. Mikesell. Sales Taxation: State and Local Structure and Administration. 2d ed. Washington, D.C.: The Urban Institute Press, 994. Federal Energy Regulatory Commission. Financial Statistics of Major U.S. Investor Owned Electric Utilities. Washington D.C.: Federal Energy Regulatory Commission, 997. Federation of Tax Administrators. Sales Taxation of Services: 996 Update. Research Report No. 47. Washington, D.C.: Federation of Tax Administrators, 997. Fox, William F. Sales Taxation of Services: Has Its Time Come? Chapter 4 in Sales Taxation: Critical Issues in Policy and Administration, edited by William F. Fox, 5 6. Westport, Connecticut: Praeger Publishers, 992. Hines, James R. Three Sides of Harberger Triangles. Journal of Economic Perspectives 3 No. 2 (Spring, 999): Iorwerth, Aled Ab, and John Whalley. Efficiency Considerations and the Exemption of Food from Sales and Value Added Taxes. Canadian Journal of Economics 35 No. (February, 22): Jorgenson, Dale W., and Kun Young Yun. Tax Reform and U.S. Economic Growth. Journal of Political Economy 98 No. 5, Part 2 (October, 99): Jorgenson, Dale W., and Kun Young Yun. The Excess Burden of Taxation in the United States. Journal of Accounting, Auditing & Finance 6 No. 4 (Fall, 99): Murray, Matthew N. Moving the Retail Sales Tax to a Retail Tax: Optimal Tax Considerations. Chapter 6 in The Sales Tax in the 2 st Century, edited by Matthew N. Murray and William F. Fox, Westport, CT: Praeger Publishers, 997. Poterba, James M. Retail Price Reactions to Changes in State and Local Sales Taxes. National Tax Journal 49 No. 2 (June, 996): Ramsey, Frank P. A Contribution to the Theory of Taxation. Economic Journal 37 No. 45 (March, 927): Ring, Raymond J. Consumers Share and Producers Share of the General Sales Tax. National Tax Journal 52 No. (March, 999):

15 Popular Substitution Effects Stiglitz, Joseph E., and P. Dasgupta. Differential Taxation, Public Goods and Economic Efficiency. Review of Economic Studies 39 No. 2 (April, 97): U.S. Bureau of the Census. Quarterly Summary of Federal, State, and Local Tax Revenue. Washington, D.C.: U.S. Bureau of the Census, Various years. U.S. Bureau of the Census. 992 Census of Retail Trade, Subject Series: Measures of Value Produced, Capital Expenditures, Depreciable Assets, and Operating Expenses RC92 S 2. Washington, D.C.: U.S. Bureau of the Census, 996a. U.S. Bureau of the Census. 992 Census of Services, Subject Series: Measures of Value Produced, Capital Expenditures, Depreciable Assets, and Operating Expenses SC92 S 2. Washington, D.C.: U.S. Bureau of the Census, 996b. U.S. Bureau of the Census. 992 Census of Construction Industries: SIC Summary of Findings web site ( sicsof.html). Accessed on November 3,

16 NATIONAL TAX JOURNAL Demand Equation Generally Taxable Nondurables APPENDIX DEMOGRAPHIC COEFFICIENTS FROM THE ALMOST IDEAL DEMAND SYSTEM Demand Determinant δ δ 2 δ 3 δ 4 δ 5 δ 6 δ 7.37 (.5).3 (.42).69 (.5). (.2).4 (.6).36 (.6).78 (.9) Services.84 (.62).325 (.64).326 (.22).5 (.2).5 (.8).293 (.23).3 (.24).67 (.4).5 (.).26 (.4).2 (.).2 (.).9 (.4).29 (.5).24 (.9).49 (.7).42 (.6).9 (.).6 (.2).84 (.6).4 (.7).99 (.6).7 (.3).5 (.4).4 (.). (.2).4 (.4).33 (.5) Food (at Home).77 (.4).66 (.36).42 (.2).23 (.).5 (.5).27 (.3).43 (.5) Utilities.63 (.29).7 (.26).25 (.8).6 (.).7 (.3).334 (.9).4 (.) Automobiles and Furnishings Demand Equation Generally Taxable Nondurables Demand Determinant δ 8 δ 9 δ δ δ 2 δ 3 δ 4.6 (.27).62 (.2).74 (.53).238 (.48).99 (.48).2 (.48).6 (.48) Services.7 (.33).2 (.6).4 (.27).9 (.5).3 (.66).75 (.4).54 (.6).65 (.3).545 (.6).4 (.3).437 (.6). (.3).39 (.62).3 (.3).34 (.9).22 (.7).3 (.2). (.8).34 (.8).45 (.8).68 (.8). (.7).6 (.6).8 (.7).27 (.5). (.5).2 (.5).2 (.5) Food (at Home).87 (.2). (.7).52 (.42).359 (.39).248 (.38).55 (.38).48 (.38) Utilities.33 (.4).6 (.).475 (.3).27 (.27).235 (.27).56 (.27).5 (.27) Automobiles and Furnishings Notes: The demographic variables are ) seniors as a share of the household, 2) children as a share of the household, 3) number of household members, 4) number of household members squared, 5) the national unemployment rate, 6) a renter dummy, 7) a south region dummy, 8) a northeast region dummy, 9) a west region dummy, and dummy variables for the reference person age ) below 25, ) between 25 and 34, 2) between 35 and 44, 3) between 45 and 54 and 4) between 55 and