This chapter and the next focus on two key aspects of VAT design: the number of distinct rates charged, and the extent to which particular items or traders are exempted. These in turn relate to one of the questions that has evidently troubled many countries considering its introduction: Is VAT inherently regressive?
Background
Table 7.1 shows the number of nonzero rates adopted in all countries currently having a VAT. Over half of all countries with a VAT levy only one nonzero rate and less than a quarter charge three or more nonzero-rates.
Distribution of the Number of VAT Rates
(All countries)
Distribution of the Number of VAT Rates
(All countries)
| One rate | Two rates | Three rates | Four rates | Five rates | Six rates |
|---|---|---|---|---|---|
| 53 percent | 23 percent | 13 percent | 9 percent | 0 | 1 percent |
Distribution of the Number of VAT Rates
(All countries)
| One rate | Two rates | Three rates | Four rates | Five rates | Six rates |
|---|---|---|---|---|---|
| 53 percent | 23 percent | 13 percent | 9 percent | 0 | 1 percent |
As noted in Table 1.4, there is variation in the number of rates across regions. The number of nonzero rates tends to be greatest in North Africa and the Middle East, for instance, and least in sub-Saharan Africa. Closer examination66 also suggests that the number of rates at the time of introduction has a lasting effect on the number of rates applied. It appears too that older VATs are characterized by a larger number of rates, a point to which we return later.
It is also striking that, as Table 7.2 shows, the vast majority of countries (72 percent) that have introduced a VAT over the past ten years have opted for a single positive rate when doing so. One notable instance of this is the experience of the WAEMU countries, which is described in Box 7.1. More generally, the number of initial rates at the time of introduction has tended to fall over time. This presumably reflects, in large part, the perceived success of existing single-rate VATs.
VATs Introduced with a Single Rate
VATs Introduced with a Single Rate
| Date of Introduction | Number of Countries | Single Rate | Multiple Rates |
|---|---|---|---|
| Before 1990 | 48 | 12 | 36 |
| 1990–94 | 46 | 31 | 15 |
| 1995–April 2001 | 31 | 25 | 5 |
| Total | 125 | 68 | 56 |
VATs Introduced with a Single Rate
| Date of Introduction | Number of Countries | Single Rate | Multiple Rates |
|---|---|---|---|
| Before 1990 | 48 | 12 | 36 |
| 1990–94 | 46 | 31 | 15 |
| 1995–April 2001 | 31 | 25 | 5 |
| Total | 125 | 68 | 56 |
How Many Rates of VAT?
As discussed in Chapter 2, the archetypal VAT is a tax on consumption. At one level the question is thus why one might wish to tax different components of consumption at different rates. This question has been much discussed in the public finance literature. The question for VAT design is rather more precise: the widespread use of excises means that VAT is not the only tax on consumption, and the multistage nature of the VAT implies distinct practical consequences of setting multiple rates.
To bring out the various considerations of principle and practice that arise, this section considers in turn the potential benefits from differentiating rates of VAT, the potential costs, and finally the balance between the two.
Benefits of Rate Differentiation
There is a technical literature on the optimal degree of rate differentiation in taxing consumption; a brief guide is provided in Box 7.2. For present purposes, it is helpful to distinguish between two sets of considerations: efficiency (in the loose sense of raising revenue in a way that does not harm aggregate real income) and equity (ensuring a fair distribution of that aggregate income).
Single-Rate VATs in the WAEMU
Three members of West African Economic and Monetary Union (WAEMU) were among the first countries to introduce a VAT in sub-Saharan Africa—Côte d’Ivoire in 1960, Senegal in 1980 (beginning with a VAT at the manufacturer’s level in the 1960s), and Niger in 1986. However, these VATs had a complex rate structure (four positive rates in Côte d’Ivoire and Senegal, three in Niger) and many exemptions; their performance was not satisfactory. Taking into account these weaknesses, Benin was the first WAEMU member to adopt a VAT with a single rate and limited exemptions, in 1991. Given the revenue performance achieved in the following years (indirect tax revenue increased from 1.7 percent of GDP in 1990 to 5 percent in 1994), that country was quickly seen as a model by other WAEMU countries. Burkina Faso, Togo, Niger, and Mali adopted a single rate for their VATs in 1993, 1995, 1997, and 1999, respectively.
In December 1998, the WAEMU Council of Ministers adopted a directive to strengthen the harmonization of the indirect tax system. This directive provides that by January 2002, each WAEMU member will have a VAT with a single rate (in the 15-20 percent range) and limited exemptions (detailed in a list to be adopted by the WAEMU Commission).
The direct impact of the single-rate on Benin’s revenue performance has never been assessed. In fact, that measure was part of a significant tax administration reform package that was implemented (including the establishment of a large taxpayer unit and significant improvements in collection enforcement and audit systems). However, there is little doubt that, building on the simplicity of the VAT, the drafting of simple regulations and the design of simple forms have contributed to a reduction in taxpayers’ compliance burden and facilitated improvements in the effectiveness of the tax administration in that country.
Efficiency
Taxes on some commodities may be more distortionary than taxes on others. This notion underlies the widely known policy prescription—the “inverse elasticity rule”—of levying lower tax rates on commodities for which the demand is more elastic and higher rates on commodities for which demand is inelastic so as to minimize the impact of taxation on the pattern of consumption. A more general argument for differential commodity taxation based on efficiency considerations involves taking account of the interaction between consumption of commodities and (untaxed) “leisure,” interpreting the term to refer broadly to time spent other than in the formal labor market. Taxing most heavily those goods whose consumption is associated with the enjoyment of leisure in this broad sense will tend to mitigate the distortion of decisions away from paid work.67 Examples of such goods, however, are somewhat hard to find: golf clubs, for example, might fit the category in some countries; a more important example may be alcohol. As with alcohol, however, most of the commodities that come to mind in this context can conveniently be subjected to excise taxes. This is not a coincidence. Revenue and efficiency considerations long ago picked out a few goods as proper subjects of heavy taxation. There is, of course, scope for discussion as to what those goods might be (and the list will naturally vary across countries depending on national tastes and characteristics), and how many of them there are. The important point for present purposes, however, is that a few standard excises will be able to deal with most of the obvious cases in which there may be efficiency gains from rate differentiation.
The Optimal Tax Perspective on Rate Differentiation
The literature on optimal taxation has focused on finding restrictions on the form of consumer preferences, which imply that it is optimal to tax all commodities at the same rate. Central contributions include Besley and Jewitt (1995), who establish a necessary and sufficient condition for uniform taxation to be optimal in the single consumer case; Deaton and Stern (1986), who establish conditions for uniform taxation to be optimal in the presence of an optimal linear income tax; and Atkinson and Stiglitz (1976), who establish that indirect taxation is unnecessary in the presence of an optimal nonlinear income tax if preferences are weakly separable between consumption and leisure (that is, take the form U[F(X),L], where X denotes the vector of consumption and L leisure). (Edwards, Keen, and Tuomala, (1994) and Marchand, Nava, and Schroyen, (1996) provide a straightforward intuition for this last result.)
The general conclusion is that nonuniform taxation has a role to play whenever the pattern of consumption contains information about the consumer’s underlying—and unobservable—ability to pay taxes that is not fully exploited by the other policy instruments assumed to be available. The preference restrictions under which uniform taxation is optimal thus naturally become weaker as the range of instruments available widens. Even with fully optimal nonlinear taxation, however, the restrictions required remain implausible: the empirical evidence is that preferences are not of the Atkinson-Stiglitz form.
Moreover, this literature assumes, for the most part, that indirect taxes have no effect on factor incomes, either taking producer prices to be fixed or assuming profits to be fully taxed. In more general circumstances, nonuniform taxation might in principle have some role to play through its effect on factor incomes, although again, of course, the availability of other instruments will be critical.
One aspect of efficiency, of course, is the ability to raise revenue. As shown in Chapter 4, there are some signs that a wider range between highest and lowest VAT rates is associated with higher VAT revenues (for any given level of the standard rate). This is suggestive of some efficiency gain from rate differentiation, and to that extent runs counter to the argument just given; the evidence, however, is not strong.
Another and increasingly important aspect of efficiency deserves some emphasis. As noted in Chapter 2 (and discussed further in Chapter 17), the destination principle by which VAT is normally levied is intended to ensure that the tax provides a means of taxing the consumption of domestic residents. There are circumstances, however, in which commodities can be purchased tax-paid by residents of other countries. The simplest example is that of cross-border shopping by final consumers, either legally (within duty paid allowances) or illegally. Another example arises from the taxation of some services in the EU by reference to the location of the supplier. The importance of all such cases is that domestic tax rates may affect the volume of sales to nonresidents, and the appropriate rate structure will then reflect the opportunities that these consumers have to buy elsewhere. In particular, countries may find themselves at a disadvantage if they tax more heavily goods that foreigners can find less heavily taxed elsewhere.
The significance of these concerns is likely to vary greatly across countries. Cross-border indirect tax differences are often greater for excises than for VAT, so the issue will in most cases be of less importance for VAT (though there are exceptions). It may also be that services will prove particularly vulnerable: as discussed in Chapter 17, there have already been suggestions in the EU to zero-rate items delivered over the Internet in order to protect domestic producers. One area in which the treatment of purchases by nonresidents is both pressing and quantitatively significant for many developing counties is that of tourism. This is discussed in Box 7.3. In this, as in other instances of this general problem, the possibility that countries will do themselves mutual harm by competing for mobile tax base may ultimately point to a potential gain from coordination of tax rates, perhaps on a regional basis.
The VAT Treatment of Tourism
It is often advised that services provided to tourists should be fully taxed. It is increasingly argued, however—and in some cases is the practice—that they should be zero-rated or exempted.
The issue is sometimes couched in terms of the question of whether these services are properly regarded as consumption (and so subject to normal tax), because they are consumed domestically, or exports (and so zero-rated), because they are consumed, typically, by foreigners. This, however, is not the issue; nor is the treatment of tourism in the national accounts of any relevance.
The key economic point is that the ability of tourists to substitute between services offered by different countries can quite plausibly imply highly elastic demand for services that are broadly the same in different countries. This high elasticity points to a low tax rate. The argument applies most obviously to facilities—palm-fringed beaches, for instance—that are much the same in different countries. Where facilities are genuinely unique, however, the same considerations will point to a high rate, naturally implemented as an excise or entrance fee. Even manifestly matchless resources may, however, in a wider sense, be substitutes: a visit to the Pyramids may be a substitute for one to the Forbidden City. Even in such cases, the ability to extract rents may therefore be limited.
A strong case can thus be made for setting a low tax on generic tourist services, perhaps even a zero rate. Or, to discriminate between less elastic domestic demand and more elastic foreign demand, tax might be rebated to nonresidents (as with the hotel tax in Canada); this, of course, is potentially troublesome to administer. Genuinely distinctive tourist resources might be made the subject of user fees, which could in any event be better tailored to congestion and other external effects of their usage.
Three arguments can, however, be made against taxing generic tourist services at a low rate. The first is the administrative: that multiple rates complicate administration and compliance, and create opportunities for abuse. The second is that there is a coordination problem: the elasticity of demand for tourist services offered by any particular country may be much greater than that for the services offered collectively. While the service offered by any particular game park is not unique, that offered by game parks in general is much more so. Thus, all countries could gain by collectively agreeing to raise the tax rate applied. This argument points to regional coordination. Third, to the extent that such services are purchased by foreigners, who presumably appear in policymakers’ objective function with a lower weight than do their own nationals, the social loss from raising taxes on these goods will be reduced: taxing foreigners is always attractive.
Finally, while environmental considerations have come to receive considerable attention in the formulation of tax policy, they do not provide a rationale for differential VAT rates. The VAT is intended as a tax on consumption, while corrective taxes should apply to all uses of externality-generating commodities; intermediate use, in particular, should be taxed just as heavily as final use. Externalities call for excises, not consumption taxes.
Equity
These efficiency considerations abstract from the possibility that alternative commodity tax structures will alter the distribution of real income. In terms of equity considerations, it will be desirable, all else equal, to tax most heavily those goods that account for a greater share of the expenditure of the better off. If, on the contrary, all households allocated their budgets across commodities in the same way—that is, if budget shares for all goods were independent of income—then there would be no equity gain in taxing some goods more heavily than others. However, since there is no reason to presume that this condition holds in general, in the absence of other instruments to achieve distributional goals it may be desirable to set more than one rate of indirect taxation.
The presence of other instruments, however, makes it less likely that social gains will be had from setting more than one rate of VAT. Most obviously, the presence of an income tax provides a more effective means of pursuing distributional objectives, and differentiation is consequently less likely to be needed. For example, if the way in which consumers allocate their spending across commodities is the same at all income levels—more precisely, if budget shares do not depend on the pretax wage rates at which individuals can sell their labor—then indirect taxation is entirely unnecessary. (This is the Atkinson-Stiglitz (1976) result, referred to in Box 7.2.) The reason is straightforward: there is no information on individuals’ underlying (and unobserved) ability to pay taxes contained in the pattern of their expenditures that is not already contained in their (observable) earnings; thus one can do no better than simply tax the latter. This, however, is a restrictive condition, unlikely to hold in practice. More generally, in the presence of an optimal nonlinear income tax, it will be appropriate to have as many distinct rates of indirect taxation as there are distinct groups of commodities that are perfectly substitutable among themselves.68
It has also been increasingly argued that expenditure policies, in areas such as education and health, may be more effective tools for pursuing equity objectives than the use of differential VAT rates. This has received less formal attention, however, and does not seem to have been investigated empirically. Insofar as such expenditure measures are akin to negative income tax payments, however, the same arguments as above would suggest a reduced need for rate differentiation.
The availability of other instruments thus weakens the case for rate differentiation. It does not, however, eliminate it. That said, the availability of other instruments is likely to be crucial, and on this there may be an important distinction between developed and developing countries. For a developed country with a well-functioning income tax, the case for rate differentiation is likely to be relatively weak. Davis and Kay (1985), for example, show how the elimination of the considerable zero-rating of VAT in the United Kingdom could be accompanied by an increase in the threshold for the income tax and increases in benefit levels in such a way as to maintain revenue without jeopardizing progressivity. Using a computable general equilibrium (CGE) model calibrated to the United States, Ballard and Shoven (1987) conclude that differentiating the rate of a VAT used to partially replace the income tax would have a large efficiency cost but do relatively little to mitigate the adverse distributional impact of the change. Many developing countries, however, do not have well-functioning income taxes and the case for differentiation may thus be stronger. Using a CGE model for Bangladesh, Hossain (1995) concludes that some zero-rating and selective use of excises would significantly diminish the regressive impact of using the VAT to replace the preexisting indirect tax structure.69
How large could be the gains that can be achieved by redistribution though indirect taxation? The scope for redistribution depends on the extent of dissimilarity in spending patterns of rich and poor. For quite plausible degrees of dissimilarity, however, the amount of redistribution that can be achieved through any indirect tax alone is quite limited.70 Suppose, for example, that there is some particular good—food, say—that accounts for a much larger share of the expenditure of the poor than it does of the expenditure of the rich. Then to achieve a more equitable distribution of the tax burden one might think of taxing food more lightly than other commodities, and even subsidizing it if the funds are available. The extent to which money can be channeled to the poor by this means is limited, however, by the fact that while the proportion of their income that the rich spend on food may be relatively low, the amount of food they consume may be very large. Thus much of the money spent on subsidizing food will actually benefit the rich: in one country, FAD staff estimated that for every $100 (say) in VAT revenue forgone as a result of zero-rating food, less than $15 accrued to the poorest 30 percent of the population, while about $45 benefited the richest 30 percent.
These inherent limitations to the extent of redistribution that can be achieved by indirect taxation are developed in Box 7.4. They mean that even the best-informed government will be severely constrained in the redistribution it can achieve by rate differentiation.71
The discussion has presumed so far that the main distributional effect of VAT is felt through the prices that people pay for the goods they consume. That may not be the case. The real burden of the VAT tax may not fall entirely on consumers but may in part be passed back to suppliers of factors through lower prices received by producers.72 The distributional consequences of a VAT may thus be more complex than at first appears, and in some cases a VAT might be deployed in a deliberate attempt to tax producers rather than consumers. For example, the availability of imported foods may mean that small farmers—that is, those below the VAT threshold—are unable to recapture in higher consumer prices the VAT that they are charged on their fertilizer usage. Despite the production inefficiency it induces, such a tax may thus be desirable as a means of taxing this otherwise hard-to-tax group.73 On somewhat similar grounds, Ab Iorwerth and Whalley (1998) argue that the commonplace exemption of food exacerbates the distortion toward excessive home production of meals implied by the non-taxation of time inputs into home cooking.
Limits to Redistribution Through Indirect Taxation
Assume that there are only two income groups—“rich” and “poor”—and only two goods. Assume also that the poor spend all of their income on good 2, while the rich buy both goods. The sole object of government, for argument’s sake, is to redistribute income to the poor. Clearly the best way to do this is to tax good 1 (at tax-inclusive rate t) in order to finance a subsidy s on good 2.
How much redistribution can be achieved? Since the poor spend all of their income on the subsidized good, the amount they gain is simply the subsidy received. Expressed as a proportion of their income, this is simply the subsidy rate s. For the government’s budget to balance, it must be that sw2 = tw1 where wi denotes the proportion of total income (of rich and poor combined) that is spent on good 1. Thus, s = t(w1/w2); and so, since t cannot exceed unity, the gain to each poor person can be no more than w1/w2.
Assume, for example, that 80 percent of all income accrues to the rich, who spend 90 percent of their income on good 1. Since the rich are the only group who buy good 1, it follows that w1 = 0.72. Then the poor can gain no more than 250 percent of their income.
This is a surprisingly small amount, given the propitious circumstances of the example. Even though the rich spend relatively little of their income on the subsidized good, the volume of that spending on that good soaks up much of the subsidy that the taxes that they pay finance.
In less extreme cases, the potential redistributive gain is even less. Change the numbers above, for instance, to have the rich receive 70 percent of all incomes and spend 50 percent of their income on good 2, and the potential gain to the poor falls to only 54 percent of their incomes.
A striking instance of the use of rate differentiation to pursue distributional objectives through effects on factor incomes is the recent EU decision to allow member states to set (for a period of no more than three years) a reduced rate of VAT on a limited number of items from a specified list of labor-intensive services, such as construction, renovation, child care and hairdressing. The avowed purpose of this initiative is to stimulate employment in these sectors. Such measures might indeed be rationalized as counteracting the possibility that taxation of such items at the full rate would encourage consumers to adopt inefficient methods of self-supply (an argument developed by Piggott and Whalley (1998)): cutting their own hair, for example, rather than purchasing in the market. They may also serve to reduce the extent of tax evasion, potentially a worth-while objective in itself. Against this, however, the distributional consequence of setting a reduced rate in this way may even be adverse, as small tax exempt firms cease to benefit from the ability to sell at the gross prices charged by their taxable competitors. Moreover, it is not clear why one might wish to encourage employment in these sectors rather than in general; for which better tools—reducing payroll taxes, for instance—would seem to be available.
Costs of Rate Differentiation
Although there may thus be advantages to some rate differentiation, there are evident costs as well. These can be divided into two main categories, both of which also apply to the introduction of exemptions, which is discussed in the next chapter.
Increased administration and compliance costs
The reasons that applying a single rate of VAT to all items except exports generally74 simplifies both compliance and administration include:
The use of a single positive rate simplifies taxpayer books and invoices, and eliminates the need to open separate records for purchases and sales depending on the rate to be applied. It also facilitates simple and easy-to-understand tax forms—a single page for a single-rate VAT in New Zealand, instead of two to three pages in the case of VATs with multiple rates (including reduced positive rates and zero-rates) in EU countries. Simple records, invoices, and tax forms in turn support the effective operation of a self-assessment system based on voluntary compliance. They also support more effective taxpayer education and staff training programs.
The use of a single rate facilitates VAT audits. In a single-rate VAT, basic desk verifications, including cross-checking between VAT, income tax, and customs data, help assess the accuracy of VAT liabilities reported by the taxpayer. These verifications are complicated in a multiple-rate VAT, where the breakdown of purchases and sales by VAT rate requires checking taxpayers’ records. Moreover, in the case of a multiple-rate VAT, significant audit time (30 to 40 percent) is typically needed to verify the breakdown of inventories, purchases, and sales between the respective rates. Conversely in a single-rate VAT, audit work can focus essentially on two main issues: underreporting of output VAT; and over-reporting of input VAT.
The use of a single rate also helps limit the number of refund claims to, essentially, exporters and enterprises purchasing a large amount of capital goods. Conversely, the adoption of a reduced positive rate and the use of a zero rate for domestic sales can result in sustained credit positions for those taxpayers using these rates. This results in additional refund claims, which increase the burden of the tax administration and impact the effectiveness of its operations.
Differential treatment of different goods creates scope for argument as to the proper treatment of items lying on the borderline between distinct categories. In the United Kingdom, for example, disputes have arisen as to whether deepening a house’s foundations should be treated as repairs (taxable) or an extension, and whether a Jaffa Cake is a cake or a biscuit. Cnossen (1987) lists many murky borderlines in the Dutch VAT: smoked eel taxable at the reduced rate, for example, while smoked salmon is subject to the standard rate.
Differentiation also creates incentives for the deliberate misclassification of items. One important instance of this is in the zero-rating of exports, which creates the potential to gain by diverting to the domestic market goods claimed to have been exported. Border formalities can discourage such fraud by providing a check on the fact of export. In the EU, however, the removal of such frontiers has given rise to increasing concern that the VAT chain is weakened by the zero-rating of exports to registered traders. Moreover, in many developing countries the need to prevent abuse of these provisions whilst providing timely refunds to honest traders is a major and continuing concern.
Consistent with these difficulties, some tentative evidence that multiple rates do indeed jeopardize revenue by weakening enforcement is provided by Agha and Haughton (1996), who find for a sample of OECD countries that the ratio of actual VAT revenues to the estimated yield with perfect enforcement is significantly lower the greater is the number of VAT rates.
Political vulnerability
One common concern in considering measures of rate differentiation or exemption is that, however deserving the measures in their own right, they make it harder to resist pressures for other and perhaps less deserving measures of differentiation. By permitting some differentiation, governments may reveal themselves more willing to live with the complications that differentiation brings and so increase their vulnerability to further erosion of the tax base. Thus it might be, for example, that the theoretical considerations discussed above point to a system with two distinct rates of VAT. But political dynamics may mean that the real choice is between one rate and many more than two.
The legitimacy of this concern is to some degree empirically testable: does the number of distinct VAT rates tend to proliferate over time, and does the number of rates at the time of introduction have any lasting effect? The regression reported (in equation 7.1) above tends to answer both questions in the affirmative. There is in this sense some empirical support for the “slippery slope” argument against multiple rates at the time of introduction: through this route, starting off with a relatively small number of rates may go some way toward limiting the number of future rates.75
Comparing Benefits and Costs
Whether or not some differentiation in VAT rates is appropriate will depend on whether the benefits outweigh the costs. But while the nature of those benefits and costs is fairly clear, their magnitude is not. In one sense the costs are evident enough—or at least will be made evident to the policymaker by the complaints of taxpayers and tax administrators. There appears, however, to be no hard evidence on how large those costs are. The benefits are in a sense less apparent: overstretched tax administrations will complain to the minister, but the excess burden triangles of optimal tax theory knock on no one’s door.
Moreover, the benefits are likely to be quite country-specific, depending as we have seen on the nature of spending patterns (and the availability of other instruments).
One approach to this issue is to extend the approach in Box 7.4 to get some sense of how large the possible gain from rate differentiation might be. This is done in Box 7.5. In the example there, by incurring additional administrative costs equal to 3 percent of total incomes it is possible to increase the net incomes of the poor by about 38 percent. Whether this is worthwhile (bearing in mind that the rich lose out by the amount of the gain to the poor plus the deadweight loss of the extra costs) will depend on the importance one attaches to bringing about a more equal distribution of income.
Such calculations can yield no general conclusions, other than to emphasize that the issue of rate differentiation is in large part ultimately an empirical one, and that some attempt to quantify the factors involved may be fruitful. Some of the considerations raised above—including the political economy considerations—are much harder to quantify. What is widely agreed, however, is that while the benefits of introducing an additional VAT rate are likely to decrease quite quickly with the number of rates so long as even a very rudimentary set of alternative instruments is available, the costs also increase quite rapidly. While there may be some gain from introducing a second VAT rate, there is unlikely to be any great gain from introducing a fourth or fifth. As discussed in Chapter 15, it is also now all too apparent that the zero-rating of exports has given rise to severe administrative problems in many developing countries. There would need to be very clear evidence of substantial offsetting benefit to warrant extending these to domestic transactions by zero-rating items consumed by the less well off.
Comparing the Costs and Benefits of Rate Differentiation
Pursuing further the two-good, two-types example in Box 7.4, assume that the government needs to raise a proportion g of aggregate income in order to meet its expenditure needs.
If it sets only a single rate of tax, that rate will then have to be precisely g. The gain to the poor from moving to a two-rate structure is thus g - s, where, as before, s, is the subsidy on good 1. Assume that implementing this second tax rate involves administrative costs equal to a proportion γ of total income. The government must now set t and s so that tw1 + sw2 = g + γ. Since t can be no greater than unity, the gain to the poor can now be no more than (1 - g)(w1/w2) - (γ/w2).
Take the same illustrative numbers as at the end of Box 7.4: the rich receive 70 percent of all income and spend 50 percent of their income on good 1. Assume in addition that the government must raise 20 percent of all income for general spending, and that the second rate increases administrative cost by 3 percent of income. The largest possible gain to the poor from introducing a second rate is now about 38 percent of their income.
Notice, moreover, that (through the term γ/w2) the increased administration costs associated with the second rate lead to an even greater proportionate reduction in the gain to the poor: in the example above, the 3 percent administration costs actually reduce the largest gain to the poor by over 4.5 percent. In this sense, those costs are borne disproportionately by the poor (rather than by, say, a 3 percent reduction in the net incomes of rich and poor alike). The reason, intuitively, is that a government that seeks only to raise the incomes of the poor will always extract as much as possible from the rich in order to benefit the poor; if the amount that the government has to pay in administrative expenses increases, this can consequently be found only by reducing the subsidy—which bears more harshly on the poor than on the rich, since it is they whose expenditures are most concentrated on the subsidized good. Thus, there appears to be a multiplier-like effect at work, with the administrative cost of multiple rate structures having a more than proportionate effect in reducing the redistributive potential of indirect taxation.
Conclusions
The principal conclusions suggested by this discussion are:
Support for setting only a single positive rate is based both on experience with the administrative and compliance difficulties associated with multiple rates and on the realization that the amount of redistribution that can be achieved through indirect taxation is inherently limited. This position is matched by an increasing tendency for VATs to be introduced with only a single rate.
The extent to which equity gains can be achieved by differential rates of VAT depends on the range of other instruments available. A few excises on goods in inelastic demand may be able to reap the main efficiency gains from differentiation, so that the case for multiple rates of VAT is primarily an equity one.
The equity case for differential VAT rates will be stronger the more restricted is the set of other tax-spending instruments available to government. It will be weaker, for instance, when an effective income tax is in place. But while the benefits from differential VAT rates may consequently be greater in developing countries, these are also the countries in which the problems of compliance and administration associated with a multiple rate structure will be greatest.
The standard advice to fully tax tourism appears to be widely ignored. It may have underestimated the optimal tax argument for a lower tax on services of a fairly non-unique kind. Effort might be better focused on securing some degree of coordination between competing countries.
The importance of the rate differentiation issue may in any event have been overstated. Even leaving aside non-VAT instruments, there are design features within the VAT itself other than the statutory rate structure that can be—and are—rightly or wrongly used to mitigate the distributional effects of the tax. The widespread use of exemptions, for instance, discussed in the next chapter, means that even a VAT that has only a single statutory rate will have many effective tax rates, some of which (such as those associated with the exemption of basic foodstuffs, health, and education services) may have significant equity effects. A high threshold can have similar effects.