Dennis C. Mueller^*

University of Vienna

Considerable concern has been expressed in Europe in recent years about the potentially undesirable consequences of tax competition, and a large literature has sprung up analysing this question. As in much of the public finance literature, the analysis of this question has separated the tax and expenditures sides of the fisc. The government is assumed to have initially set the optimal quantity of public expenditures and taxes, and then the potential adverse consequences on tax revenue of tax competition are explored. There are two things wrong with this approach from the perspective of normative public choice: it ignores both the purpose for which the expenditures are intended, and any potential links between these expenditures and the tax instruments used to finance them. This paper seeks to remedy this deficiency.

More generally, we take up the question of the proper targets of taxation in a democratic state. This question has a long history in economics. One approach, traditionally referred to as the ability to pay approach, sees the issue as one of the proper sharing of the burden of taxation. The tax and expenditure sides of the state=s budget are separated, and the only question to be resolved is that of determining an equitable distribution of tax obligations. Implicitly this has been the perspective of much of the tax competition literature.

The second perspective on taxation, traditionally referred to as the benefit approach, links the expenditure and tax sides of the state=s budget. Ideally under this approach, the citizens decide both what to tax and at what level at the same time that they decide the purpose for which the tax revenues will be used. As we shall see, conceiving of taxation in this way, casts the question of tax competition in an entirely different light. Rather than concentrating on harmonizing tax rates and thereby eliminating the pressure on states to be efficient, when benefit taxes are used, the issue becomes one of the harmonized reporting of tax sources and, perhaps, of the coordinated collection of taxes.

We shall proceed as follows: Sections I and II discuss the taxation of individuals by a state that seeks to advance the interests of all citizens. Section III examines redistribution in a democratic state under the simple majority rule. Sections IV through VII examine issues that arise in the taxation of business, allowing for the possibility of factor mobility. In Section VIII the question of the taxation of individuals is taken up, under the assumption that they are mobile. The special case of sin taxes is discussed in Section IX. In Section X the predictions of this analysis are compared with the actual patterns of taxation in Europe. Some conclusions are drawn in the final section.

The normative theory of the state within the public choice tradition can be traced back to Knut Wicksell=s classic essay on Just Taxation (1896). Wicksell put forward the argument, equally as radical in our day as in his, that the state ought to provide only those goods that make all citizens better off. To determine what those goods are, each proposed governmental expenditure would be coupled to a tax which would cover it, and the expenditure/tax package would be voted upon using the unanimity rule. All citizens could determine whether their expected benefits from the expenditure would exceed their share of its costs, and thus any proposed expenditure/tax combination that secured unanimous support would achieve the desired goal of improving the welfare of all citizens.

A classic example of a government expenditure that can potentially provide net of tax benefits to all citizens is a pure public good. More than a half century elapsed between the time that Wicksell outlined his criteria for the state provision of goods, and Paul Samuelson (1954) derived the criterion for the Pareto optimal allocation of public and private goods. If each individual i=s utility is defined over a composite private good X, and a pure public good, G, U_i = U_i(X_i, G), then the Samuelsonian condition can be written as

(1) 

To satisfy this condition the government must determine both a quantity for the pure public good and a distribution of the private good. If we take the distribution of pre-tax income as given, the government must select a set of taxes that will satisfy (1). One such set was proposed by Wicksell=s countryman Eric Lindahl (1919). Each individual i=s tax price, t_i, is set equal to her marginal rate of substitution between the public and private goods,

(2) 

with the sum of the t_i constrained to equal P_G/P_X.

Lindahl taxes can be viewed as capturing elements of both the benefit principle to taxation and the ability-to-pay principle. If all individuals have the same preferences toward the composite private good, and all have the same potential command over the private good, so that MU_i/MX_i = MU_j/MX_j, for all j…i, then (2) demands higher Lindahl taxes for those who experience higher marginal utilities from the public good. Conversely, if all is have identical utility functions, U_i(X_i, G), but different incomes, then (2) requires lower Lindahl taxes for those with lower incomes.

A vast literature beginning with Lindahl has explored the possibility of eliciting the information from citizens that is required to satisfy (1). Although no procedure has been designed that will reveal this information perfectly, in many cases taxes could be designed that might reasonably approximate the demands of eq. (1). The benefits from fire protection to an individual are proportional to the value of the property protected, the value of public highways are proportional to the amount of driving the individual does. Property taxes to finance fire protection, and tolls or gasoline taxes to finance highways are reasonable means for approximating Lindahl taxes for individuals. With roughly similar incomes, the amount of property owned or the amount driven are good proxies for the benefits a person receives from fire protection and roads. With roughly similar preferences, the rich can be expected to own more property than the poor, to make more use of public roads, and thus to pay higher fractions of their costs.

The value of a person=s property is likely to be a good proxy for the benefits derived from many locally provided public goods (e.g., police protection), and property taxes are thus an attractive choice of tax instrument for achieving Wicksell/Lindahl taxes. Other forms of user charges, like gasoline taxes and road tolls, can also be designed. For some public goods -- like the financing of the judicial system -- income may be a good proxy for the benefits an individual receives. Wealthy individuals make more use of the judicial system than the poor, and in general have more to gain or lose financially from the outcome of a given judicial proceeding. Thus, in these cases, personal income taxes could constitute reasonable approximations to Lindahl taxes.

It is not the goal of this paper, to link up each category of public service that potentially has the property of a public good to a specific tax that would reasonably approximate a Lindahl tax. These examples hopefully suffice to show, however, that such linkages are not beyond the realm of possibility. Were such linkages forged, and voters were allowed to express their preferences toward various publicly provided services, the voters themselves would reveal the appropriateness of the particular tax/expenditure link made, through their (dis)approval of the proposed tax/expenditure package.

Equation (1) in the previous section defines the condition for Pareto optimality in the allocation of a pure public good. Lindahl taxes are usually treated as possible means of attaining allocative efficiency in the provision of a public good, and Wicksell=s voluntary exchange approach to government is also assumed to apply only to the allocative side of the government=s ledger. Redistribution decisions are generally thought to be quite different from allocative efficiency decisions, and to require different taxes and different voting rules -- like the simple majority rule -- from those that might be used to make allocative efficiency decisions. Before discussing redistribution under the simple majority rule, let us briefly review the possible ways in which redistribution might come about in a community, even under the unanimity rule.

Under this theory a better-off group, R, gets utility from seeing a worse-off group, P, become better off. This assumption can be modelled by assuming that the utility of R, U_R, is a function of the private good consumption of both an R and a P, U_R(X_R, X_P). If T is the transfer from R to P, and Y_R and Y_P are R to P=s incomes, then the budget constraints of R to P, X_R = Y_R - T and X_P = Y_P + T, can be used to replace X_R and X_P in U_R yielding

(3) 

Maximizing (3) with respect to T and setting the first derivative equal to zero gives us

(4) 

An R transfers her income to a P up until the point where the marginal utility that she gets from her own personal consumption equals the marginal utility that she gets from another unit of consumption by P. The marginal utility that P gets from consumption figures only indirectly, if at all, in R=s decision to redistribute her income to P.

As such the Pareto-optimal approach to income redistribution explains why a given person R might give to a given person P, or perhaps, why an R might give to a charity that gives to a group of Ps. It does not explain, however, why government intervention is needed to bring about this transfer. To rationalize government intervention under the Pareto-optimal approach to income redistribution, one must assume that an R only gives to the Ps, if all other similarly situated Rs given, and the government intervenes to resolve the usual free-rider problem (Thurow, 1971; von Furstenberg and Mueller, 1971).

Fire insurance pools redistribute income from those who do not have fires to those who do, unemployment insurance redistributes from the employed to the unemployed. Risk averse individuals who are uncertain about whether they will be a victim of a fire, or lose their job may raise their expected utilities by joining programs that insure against these risks. As with the altruistic behavior described in the Pareto-optimal approach, the question with this explanation for voluntary redistribution is not whether it could plausibly happen, but why the government is needed to bring it about. Why cannot private insurance clubs achieve these ex ante Pareto gains?

The answer typically given invokes asymmetric information and moral hazard behavior. Individuals know whether they are industrious or not, and thus whether they are likely to lose their jobs or not. Employers and insurance companies cannot evaluate these probabilities. Thus, highly industrious people would find it optimal to self-insure against unemployment, less industrious people would join insurance programs, and the familiar Alemons problem@ leads to the insurance market=s disappearance. By forcing everyone to Abuy@ unemployment insurance, the government solves the lemons problem and recreates Aa market@ for insurance.

Both of the preceding positive theories explain why a community might unanimously agree to implement a redistribution program. Each theory might also be given a normative interpretation. A rich person ought to obtain utility from altruistic behavior toward the poor; a rich person ought to assume that he might someday be poor. The latter prescription can lead to the conclusion that the income distribution ought to be egalitarian. To see this assume again the existence of two individuals, R, and P. The utility of each is solely a function of his own private goods consumption, U_R = U_R(X_R), U_P = U_P(X_P). T is again the transfer from R to P, and Y_R and Y_P are R to P=s incomes with Y_R > Y_P. Each individual assumes that with probability p_R he will be R and with probability (1-p_R) he will be P. After substituting the budget constraints, X_R  = Y_R - T and X_P = Y_P + T, into U_R and U_P, each individual chooses T to maximize the same expected utility function

(5) 

The first order condition from this maximization yields

(6) 

If one makes the further assumption that an ethical individual assumes that he has an equal probability of being R or P, and that they both have identical utility functions, then the righthandside of (6) becomes one, and the optimal T equates the incomes of the two individuals. Uncertainty over future position produces an egalitarian distribution of income.

Under the simple majority rule redistribution can take place without any uncertainty on the part of those who are taxed, or without any altruism on their part. If the Ps are in the majority, they can take advantage of this fact under the simple majority rule. The simplest of all explanations for redistribution in a majority rule democracy is that a winning majority uses its political authority to transfer income to itself from the losing minority.

Businesses benefit directly from a variety of goods and services provided by the state. An obvious example of such a public good is highways. The annual costs of a system of highways can be divided into the amortization of their initial cost, A, and maintenance of the highways, M, which we can think of as including the cost of highway police, etc. Maintenance costs can be expected to be proportional to traffic on the highways. Thus, the maintenance costs that must be incurred because of the activities of a given firm are proportional to the amount of goods it ships. To begin, let us assume that the initial costs of constructing the highways have been entirely amortized, A = 0. Let m be the highway maintenance cost per unit of output shipped in industry X, c_i production costs per unit of output for a firm I in this industry. If the state levels a tax on firm I to cover the highway maintenance costs, which the firm forces the state to incur, then the unit costs of the firm are c_i + m. Assume that X is a differentiated product, and each firm i faces a demand schedule, p_i = a_i - bx_i. A firm with a higher a_i can be thought of as having a higher quality product, as consumers are willing to pay a higher price for this firm=s product at any level of output. Under Cournot behavioral assumptions, a profit maximizing company in this industry chooses an output x_i = (a_i - c_i - m)/2b, and earns a profit p_i = (a_i - c_i - m)²/4b. Firms with higher quality products as measured by a_i, or lower costs of production, c_i, have larger outputs and profits in the industry.

This method of financing the taxes constitutes a form of benefit taxation. Each firm covers that proportion of highway maintenance costs that arise because of its use of the highways. Each firm chooses an output at which its marginal revenue just equals the cost of producing the last unit of output plus the social cost of shipping it to a customer. If all markets were perfectly competitive, this method of financing highways would result in a Pareto optimal allocation of resources. We have, however, assumed that each firm faces a negative sloped demand schedule, and thus possesses some market power. This market power allows each firm to earn a profit, and these profits become an alternative source of tax revenue for the state.

If one wished to apply the ability to pay principle to firms, the obvious target of taxation would be profits. Suppose, therefore, that the state chooses to cover the maintenance costs of the highway by levelling a proportionate tax t on each firm=s profits, instead of charging it a user fee proportional to its shipments over the highways. With m = 0, each firm I=s output and profits become

(7) x_i = (a_i - c_i )/2b, p_i = (a_i - c_i )²/4b

The state must choose a profit tax rate, t, that will cover the total maintenance costs incurred with m = 0.

(8) 

which yields a value for t of

(9) 

A given firm i=s taxes under the profit tax exceed its taxes when charged a fee proportional to use, depending on the direction of the following inequality

(10) 

This inequality is equivalent to

(11) 

A firm whose share of industry profits exceeds its share of industry output pays a greater share of the costs of highway maintenance. Under the assumptions made in this example, firms with higher quality products and lower costs have both higher outputs and profits. Since output increases proportionally to (a_i - c_i ), while profits increase in proportion to (a_i - c_i )² , substituting a proportional tax on profits for a highway user fee shifts the burden of paying for the highways onto the more efficient, higher quality firms.

Applying the ability-to-pay approach to the taxation of firms can reduce the efficiency of the economy in two ways. First, it reduces the costs of shipping goods, and thus leads to greater outputs for firms that ship their goods over the highways. In a perfectly competitive industry, this would imply that the marginal social benefits from the last units produced were less than the marginal social costs of producing and delivering them. In industries where competition is imperfect, and some firms earn economic profits, taxing profits rather than applying user fees subsidizes low quality/inefficient firms at the expense of high quality/efficient companies. Should there be additional fixed costs that need to be covered, the use of a profit tax rather than a user fee could result in an inefficient firm=s survival under the profits tax, where it would have disappeared if it had to pay for the costs it imposes on society.

The task of charging a user fee to cover the costs of public highways is relatively easy when all of these costs are maintenance related and proportional to use of the highways. Consider now the problem when A > 0, and these fixed amortization costs must be covered. Here we have a situation in which the optimal tax package is a two-part tariff with each highway user charged for the maintenance costs he induces, and the overhead amortization costs are covered through lump sum taxes. If we rule out lump sum taxes on feasibility grounds, the most obvious way to cover highway amortization costs that is consistent with the benefit approach would be to prorate them to use. Thus, each firm would pay a user charge, u, for the highways that was proportional to its output. This user charge would equal the highway maintenance costs plus a share of A, u = m + s, where s is defined by

(12) 

With A > 0, the user fee a firm pays to ship goods on the highways exceeds the marginal social costs of using the highways, u > m. If all industries were perfectly competitive, this property would result in a welfare loss relative to the first-best method of covering A through lump sum taxes. But when competition is imperfect, as we have assumed, such a change in output might increase or decrease social welfare. Covering A through a tax on economic profits would, on the other hand, leave output unchanged from the level that would be optimal under the perfect competition assumption. Such a method of covering social overhead costs might seem attractive, therefore. Use of the profits tax would continue to shift the burden of taxation from the relatively less efficient/low quality producers to the relatively more efficient/high quality producers, however.

Consider now the case of two countries, Ö and S, which are situated next to one another. Both have highway systems with identical maintenance cost structures and identical initial costs that must be amortized. Producers to the north of the two countries ship goods across them to customers to the south. Transit routes through the two countries are identical in all respects. Producers choose shipping routes entirely on the basis of costs. We continue to assume that there are no crowding or other external costs to shipping.

Suppose that each country decides to charge a user fee to cover both the maintenance and amortization costs of their highways. Each uses eq. (12) to fix s, setting n equal to the number of domestic users, d, plus one half of the total number of potential transit users. Call this s, s_n. Under the assumption that d is the same in both countries, both initially set the same u and divide the transit traffic equally. By cutting its user fee slightly, however, one of the two countries can capture all of the transit traffic, and cover a greater fraction of it=s A out of the user fees levied. A competition between the two countries for transit traffic ensues with the much lamented Arace to the bottom@ in user fees.

There are two things to note about this outcome. First, bottom in the case where the countries attempt to cover the costs of the highway system through user fees, consists of u = m, not u = 0. Neither country has an incentive to subsidize transit by charging transit vehicles a price that does not cover the maintenance costs the transitor causes. Assuming that the Bertrand equilibrium is reached with each country setting u = m, each country must cover the highway amortization costs out of other sources of tax revenue. If we assume that the transitors are incorporated outside of Ö and S, then each of these two countries=s own citizens must bear the full amortization costs of their highway system. If we ignore the differences in tax burdens within each country, a country=s citizens are no worse off financially setting u = m + s_d, where s_d is determined using eq. (12) under the assumption that the number of highway users is just the number of domestic users, d, than they are setting u = m. With u = m at the Bertrand equilibrium, a country can expect to have half of the total potential transitors crossing its borders, with u = m + s, and s based on n = d, a country can expect no transitors. If there were any costs at all beyond highway maintenance fees to having foreign firms ship their goods across a country, the second solution would dominate the first, and each country would set its user fee at the higher level.

Of course, if both countries set the same user fees under the assumption that with these fees they would have no transitors, they would again share the perhaps now lower total number of transitors. The transitors would now be covering more than their proportional share of the highway amortization costs, and implicitly subsidizing other activities in each country. One might expect in this case that user fees would began to fall, and a downward spiral would again ensue, but one might not. Once each country=s politicians understand the nature of the competitive game, they realize that they are unlikely to be able to choose a positive s that is just less than that of the other country, and capture all of the transit traffic. The three outcomes that might be expected to be sustained are s = 0, s = s_d, and s = s_n. Since the first is the least attractive of the three for each country, there is little reason to expect it to be the likely outcome.

Paying for the highways by taxing profits, or by some other method that corresponds to the ability to pay approach, obviously has no attraction for either country. Such taxes could be levied only on each country=s own citizens. Transitors would cover none of the costs of the highway system. The advantages of benefit taxes to finance highway systems are enhanced, once the highways are opened to transit traffic, even in the presence of potential intercountry tax

competition.

City fire departments protect business property from fire; city police protect business property from theft and vandalism; city sanitation departments dispose of the trash and sewage of businesses. Each of these services, and many others that governments provide, benefit businesses in obvious ways. Each of these services could easily be financed by some form of benefit taxation. The benefits from fire protection are proportionate to the value of the plant and equipment of a business. So too, to a large extent, are the benefits from police protection. Property taxes are the obvious source of revenue for these public services, when they are financed under the benefit principle.

In other cases the benefits from government services, as with highway use, may be more closely related to output or value added. In all cases, however, business= benefits from government services are far more likely to be proportional to some measure of the scale of the business -- replacement costs of capital stock, output, value added -- than to profits or other measures of company financial success. Thus, taxes on property, sales and value added become logical choices for financing government activities, when business is taxed under the benefit principle. As in the case of taxes to pay for highways, if a government service like fire protection is financed out of a proportional tax on profits, rather than say by a property tax, the share of the costs of the services borne by the relatively more efficient/high quality producers rises. Applying the ability-to-pay approach to business taxation, rather than the benefit approach, effectively redistributes tax burdens to subsidize inefficient producers at the cost of more efficient ones.

As with highways, the costs of most public services can be separated into a fixed component that does not vary with output, and a variable component proportional to output. Considerable empirical research estimating a so-called Adegree of publicness@ parameter for urban public services implies, however, that in the long run the fixed costs for these services are zero and variable costs are constant. A city with a population of one million spends roughly twice as much on police, fire protection, etc. as a city of 500,000. Thus, for the kinds of public services that local communities supply, the task of choosing a benefit tax is fairly simple, as no formula for sharing the fixed costs of the service needs to be determined. Each business can be charged a share of the average costs of providing a service that is proportional to its benefits from the service.

When the long run marginal costs of providing a public service are constant, and communities rely on benefit taxation to finance these services, no community needs to be concerned about tax competition and races to the bottom, provided that it is as efficient as other communities at providing public services. No business will move into a community that does not provide fire protection and the other public services, and it will willingly pay for these services if that they are provided as efficiently as in other communities, and thus charged for at the same levels. No community will charge businesses a price for these public services that is less than its costs of providing them, because to do so would raise their average tax burden, and thus lower their welfare from living in this community. Any community that followed a strategy of taxing its citizens to subsidize the costs of providing public services to business would find that in the long run its citizens would move to other communities, where businesses are not subsidized and thus the net of tax incomes of citizens are higher. In the long run communities must compete for both businesses and citizens, and if they discriminate against either one, they can expect to see them leave.

When all communities employ benefit taxes and individuals are mobile, their movement from one community to another can increase the welfare of all individuals. This is the famous Tiebout (1956) proposition. Individuals who seek higher quality schools for their children move to those communities that provide such schooling, and are willing to pay the higher taxes that are required. Some communities offer many high quality public services and have high tax rates, others attract citizens who prefer minimal public services and low taxes. No race to the bottom results, unless we assume that all communities start the race with bundles of public goods and services that are of a higher quality than their citizens want. Even under this assumption, Abottom@ will consist of public goods bundles that match the preferences of each community=s citizens.

Tiebout=s demonstration that the free mobility of individuals improves the allocation of publicly provided goods and services rests on a variety of assumptions that may not be reasonable in a real world context (Mueller, 1989, Ch. 9; Inman and Rubinfeld, 1997). For example, crowding may result in some communities that attract citizens, while a community that loses population may find itself left with a stock of public goods that exceeds the desires of its remaining residents. But even in this situation the problem is not one of an excessive tax competition between the two communities, nor would tax harmonization improve matters. Indeed, one way to deal with this problem would be for the community that is gaining population to raise its taxes to keep new residents away.

Any community that is inefficient in its provision of public services and sets higher taxes than other communities, which provide comparable levels of services, is likely to lose population, of course. Its government will feel a pressure to improve the efficiency with which it supplies its public services, and then to lower taxes (or raise quality). Requiring that the more efficient communities raise their taxes to harmonize them with those of the less efficient ones would only worsen the welfare of citizens in both communities. We conclude that allowing for citizen mobility in a world in which all communities employ benefit taxation to finance their bundles of goods and services will not result in a downward spiral in taxes that harms the citizens in each community, and may actually benefit citizens in some, by making their governments more responsive and efficient.

All of the forms of redistribution that could conceivably come about even if a community used the unanimity rule provide each citizen benefits, and could therefore be financed by a benefit tax of some sort. With Pareto-optimal redistribution, the rich benefit from seeing the poor better off. A tax on the incomes of the rich with the funds transferred to the poor is perhaps the most obvious way to bring about this redistribution, but other tax/transfer combinations are also possible. For example, if many local services like fire protection are financed through proportional property taxes, an additional tax on high property values could be levied, converting the property tax into a progressive tax, with the additional funds raised transferred to the poor.

In any case, there would be no reason for the rich to try to avoid paying the tax by exiting the community, if they obtain utility from seeing the poor members of their community better off. If the rich exited they would not pay this tax, but they would also not receive the benefits from seeing their poorer fellow citizens better off. It would be no more rational for a rich person to exit from a community to avoid a tax that was used to finance the kind of redistribution she favors, then it would be to exit to avoid a tax that was used to provide the level of education for her children, which she wanted.

A similar argument holds with respect to government run insurance schemes. If unemployment benefits are financed out of a pool that is created through a tax on the incomes of potential beneficiaries, no one has an incentive to exit the community to avoid the tax, so long as they remain uncertain in their new community about the chances of being unemployed. If the argument that the market cannot supply unemployment insurance efficiently is valid, then no one would have an incentive to leave a community that provides unemployment insurance at its actuarial value and enter a community that does not provide unemployment insurance.

The same is true for anyone who supports redistribution, because she accepts one of the normative arguments for it. If someone votes to tax themselves, because they believe that one ought to behave altruistically toward the poor, or assumes that she has an equal chance of being herself or a poor person, she is unlikely to decide after the tax that she voted for is implemented to move to another community to avoid paying it.

Someone might move to avoid paying a tax, which they did not vote for, of course. Redistribution from a losing minority to a winning majority, as occurs under the simple majority rule, does provide the losing minority with an incentive to move to a community which does not engage in this form of redistribution. Thus, citizen mobility might constrain a government=s ability to redistribute the income of the community toward its supporters, and would then encourage it to try and prevent the targets of its involuntary redistributive programs from moving to other communities in which they were not so heavily taxed. Tax harmonization agreements, which equate tax rates across communities and thereby discourage migration are likely to be attractive to any government that seeks to engage in involuntary redistribution.

Social concern about drinking arose most prominently during the 19th and early 20th centuries in Scandinavia, Northern Europe, and North America. ATemperance movements@ were led by women (wives) who sought to prevent their husbands from stopping at the local pub on the way home from work and squandering the family=s resources. Thus, the original opposition to alcohol consumption was based on its financial impact on working class families, and the proposed solution to the misery caused by alcohol consumption was to prohibit or significantly restrict its consumption.

Such policies could be rationalized using standard externality arguments. The consumption of alcohol by men imposed significant costs on the women and children who depended on their incomes. A similar rationale for constraining the consumption of alcohol cannot be sustained today in the developed countries of the world, however, first because families are less dependent today on one member=s income, and second because alcohol consumption today, even in extreme cases, would constitute a far smaller proportion of most families= income than it did a century ago.

The rationale often stated today for restricting the consumption of alcohol through taxation, and analogously tobacco, rests on their likely adverse effects on health. These effects are borne by the consumer herself, however, and thus cannot justify state intervention to curb an externality. Nor can taxes on alcohol or tobacco be reconciled with either the ability-to-pay or the benefit approaches to taxation. If one justifies taxing alcohol and tobacco on the basis of their adverse effects on health, then the beneficiaries from these taxes are the persons who do not drink or smoke as a result of these taxes. Thus, the beneficiaries from taxes on alcohol and tobacco do not pay the taxes, and the drinker or smoker who does receives no services from the state for the extra taxes she pays that are not equally available to a person who does not drink or smoke.

The most perverse of all sin taxes are those levied on gambling, however. Gambling, like alcohol consumption, was originally thought to be a social evil because of its addictive nature, and because of the financial hardship it brought to those families that possessed an addictive gambler. The social response to the intrafamily externalities caused by gambling was originally, as with alcohol consumption, outright bans of it. Until fairly recently, virtually all forms of gambling were illegal in the United States outside of Nevada. One state after another has introduced gambling, however, until it is now quite wide spread in the United States, and an important source of revenue in many states other than Nevada. Yet, gambling has no adverse effects on the health of the gamblers, and no negative externalities, except for the intrafamily financial externalities, which were the cause of its originally being banned. By taxing gambling, however, the state reduces the income of gamblers, and thus worsens the only external effect of gambling that could possibly justify its taxation in the first place. Thus, as with taxes on alcohol and tobacco, gambling taxes cannot be reconciled with either the ability-to-pay or the benefit approaches to taxation. Indeed, in the case of gambling taxes, they fall heavily on those with a great inability to pay.

The only normative justification for taxing alcohol, tobacco and gambling that one can give is that they impose less excess burden than other sorts of commodity taxes owing to the highly inelastic nature of their respective demand schedules. But these high inelasticities exist because of the addictive nature of the activities taxed. Thus, the very high inelasticities that justify taxing these activities to minimize excess burden, imply that the taxes do not significantly deter consumption, and thus fail to accomplish their alleged goal of deterrence. The main effect of these sin taxes is merely to transfer consumers= surplus to the state, whereby they admirably accomplish the state=s goal of raising revenue -- which of course explains why the taxes exist.

Because sin taxes involve no quid pro quo of public service benefits for citizen tax payments, and because they typically are very high percentage multiples of the values of the goods taxed, the incentives to avoid them are quite large. Cross-border shopping in response to tax differentials has been common in both North America and Europe. Since the taxpayer gets nothing in return for the taxes she pays, she has an incentive to avoid all of these taxes. AA race to the bottom@ in sin taxes is possible, if communities are situated close enough to make cross-border shopping practicable. In a highly mobile world, tax harmonization may be the only way to preserve the revenue, which the state gets from these sorts of taxes.

If governments provided only those goods and services, which citizens and businesses demanded, and taxed the beneficiaries of its policies in proportion to the benefits that they receive, one might expect that tax and expenditure patterns would be similar for countries at similar stages of economic development. Such is not the case, however. Government expenditures as a percentage of GDP in the major industrialized countries range from around 30 percent in Australia, Switzerland and the United States to some 50 percent in the Netherlands and Denmark (see Table 1). An explanation for these great disparities, which is consistent with the assumption that governments do provide the goods and services that citizens and businesses demand would be that Atastes@ for government outputs differ greatly even across countries at similar levels of economic development.

• This paper has received financial support from the Hochschuljubiläumsstiftung of the City of Vienna.