11 The Threshold

Liam Ebrill, Michael Keen, and Victoria Perry
Published Date:
November 2001
  • ShareShare
Show Summary Details

Experience has taught, sometimes harshly, that a critical decision in designing a VAT is the threshold level of firm size above which registration for the tax is compulsory. This chapter reviews the issues at stake.


This is an area in which FAD advice has clearly been at odds with practice: the results from the survey reported in Chapter 6 indicate that thresholds actually adopted have in many cases been substantially lower than those recommended, and have rarely been higher. Table 11.1 compares thresholds recommended by the IMF with those applied (at the time of the survey101) in those survey countries for which the information was obtained. On average, those adopted are less than 80 percent of those recommended. This, however, understates the issue: while some follow FAD advice on this matter fairly closely, others set vastly lower thresholds. This dissonance suggests that this is an issue that needs closer thought than it has previously received, and a central task of this chapter is to begin that task.

Table 11.1.VAT Thresholds: Actual and Recommended(U.S. dollars)
Actual ThresholdFund Recommendation
Burkina Faso80,00080,000
$300–$1,200(varies locally)$10,500
El Salvador6,00012,000
Sri Lanka33,00030,000
Vietnam(a) Zero for state and foreign enterprises and others under invoice method;

(b) 1.5 times minimum civil servant salary for individuals and households
Source: IMF staff compilation.

$20,000 at introduction; $37,000 in January 2000.

Firms above the upper threshold are fully subject to VAT. Those between lower and upper thresholds charge a turnover tax at 6 percent (compared to 15 percent standard rate for VAT), which is then creditable against output tax for firms above the upper threshold. In this intermediate range the system can thus be thought of as a simplified VAT. The IMF recommendation relates to this lower threshold.

Adjusted for inflation since time of recommendation (1986).

$20,000 at introduction.

Source: IMF staff compilation.

$20,000 at introduction; $37,000 in January 2000.

Firms above the upper threshold are fully subject to VAT. Those between lower and upper thresholds charge a turnover tax at 6 percent (compared to 15 percent standard rate for VAT), which is then creditable against output tax for firms above the upper threshold. In this intermediate range the system can thus be thought of as a simplified VAT. The IMF recommendation relates to this lower threshold.

Adjusted for inflation since time of recommendation (1986).

$20,000 at introduction.

There can be no doubt as to the practical importance of the threshold issue. Several survey responses—those for Albania, Croatia, Georgia, Ghana, and Uganda—explicitly mention the low level of the threshold as having been one of the principal weaknesses of the VAT adopted. In Ghana, the low level of the threshold is cited as one reason for the failure of the 1995 VAT.102 One of the key differences between the VAT reintroduced in Ghana in January 1999 and its predecessor is a much higher threshold: about $75,000 in the new version, compared to $20,000 in 1995. In Uganda, the near failure of the VAT in 1996 is in large part attributed to a low threshold, which in the event was quickly raised from a level of $20,000 at the time of introduction to $50,000 only five months later. Within the EU, the considerable structural harmonization accomplished under the Sixth Directive does not extend to the level of the threshold;103 variation across member states continues to cause some friction.

To elaborate on current practice, two features are striking. First, there is considerable variation in the level at which VAT thresholds are actually set. The average for the survey countries in Table 11.1104 is about $34,500 with variation from zero in Vietnam and $6,000 in El Salvador to around $80,000 in the WAEMU countries and over $200,000 (for distribution) in China.

The variation is far wider, and the mean higher, across the full set of countries with a VAT: for those countries for which we have been able to obtain information, the mean VAT threshold105 is $90,000; the lowest is zero, and the highest (found in Singapore) about $700,000. The variation is also considerable when thresholds are expressed relative to GDP per capita—greater, indeed, than the variation in GDP per capita. Even within the EU, there is significant variation in the level of the threshold: from zero in several member states (albeit with simplified schemes) to about $75,000 in the United Kingdom and $100,000 in France (the last of these reflecting a substantial increase, from about $17,000, at the start of 1999).

Second, there is also significant variation in the form that thresholds take and in the extent and nature of related measures. In most cases there is a single threshold, specified as a monetary amount of turnover. However:106

  • A number of countries specify different thresholds for different kinds of activity. The most common form of variation involves a lower threshold for services than for other activities; this is the case, for example, in Burkina Faso, Indonesia, Ireland, and Togo.107 Moreover, those instances in which the VAT only extends to some stage of production also imply, in effect, threshold differentiation: the threshold is infinite for excluded activities.

  • A few countries have also adopted measures to smooth the discontinuity that otherwise arises when all activities are brought into tax once the threshold is reached: in Japan, for example, entities with turnover above ¥30m (below which no VAT is payable) but below ¥50m (above which VAT is fully payable) have their liability reduced proportionately. The Netherlands also provides a tax reduction to those whose tax liability is between NLGs 2,173 and 4,150, ensuring a gradual rather than discrete entry into a taxpaying position.

  • More generally, many countries, including those with a threshold of zero (such as Italy and Spain), apply simplified schemes, notably a presumptive tax based on firm characteristics or with reduced reporting requirements.108 Some countries (including Bangladesh, the Philippines, China, and Tanzania) levy a simple turnover or gross receipts tax on such firms,109 typically at a low rate (in the order of 2 to 4 percent). Somewhat more common in developing countries is for those firms below the threshold to be subject to some form of presumptive tax. Diversity in practice was also evident in the survey response: of those cases with usable replies, in about 13 of the countries surveyed no sales tax was levied on those below the threshold; about five imposed some turnover tax and eight were subject to some form of presumptive taxation.

Most countries, but not all,110 allow those below the VAT threshold to register voluntarily. Compliance costs aside, this will be in the commercial interests of those selling to registered traders (since registration will enable them to recover tax on their inputs, and the tax that must then be charged on output will in any event be recovered by the purchaser) and of those selling to final consumers at a sufficiently low tax rate (so long as the tax recovered on inputs exceeds that which has to be charged to those consumers). The zero-rating of exports, in particular, means that all exporters will find registration beneficial. Voluntary registration is typically seen as a means of limiting competitive distortions and avoiding inequities. Restrictions are needed, however, to prevent companies registering to take input tax credits and then disappearing before paying any positive net tax.

Considerations in Setting the VAT Threshold

A Crucial Empirical Regularity

The rationale usually given for a high threshold rests on the empirical observation that the size distribution of enterprises is typically such that a relatively small proportion of firms account for a large proportion of potential VAT revenue. Deploying scarce administrative resources so as to raise revenue most effectively thus calls, it is argued, for a concentration of those resources on the largest taxpayers; the revenue to be raised from the smaller firms is seen as insufficient to warrant the resources required for its collection.

It does indeed appear to be an empirical regularity that value added is very strongly concentrated among a relatively few firms. Table 11.1 shows the distribution of turnover by size of firm for selected countries. Despite significant variation, a useful rule of thumb is that the largest 10 percent of all firms commonly account for 90 percent or more of all turnover.111

This seemingly universal feature has important implications for the relationship between the threshold and the tax base: starting from a low level, a $1 increase in the threshold is initially very cheap in terms of revenue foregone, but becomes much more expensive at higher levels of turnover. That is clearly the case for the distributions shown in Table 11.2. Nevertheless, experience shows that many countries have not found entirely compelling the case for a high threshold that this would seem to imply. Quite why is not always clear. In part, there seems to have been a belief that a lower threshold than that advised would prove more productive of revenue. There may also have been concern over the potential inefficiencies and inequities arising from the differential treatment of those above and below the threshold. To assess the force of both the standard argument and possible objections to it, it is helpful to consider more systematically the underlying issues of principle that arise in considering the appropriate VAT threshold.

Table 11.2.Distribution of Turnover in Selected Countries1

EgyptGeorgiaPakistanSri LankaUganda
Source: IMF compilation

Entry is percentage of turnover (or, in the case of Egypt, GST revenue) accounted for by largest firms.

Sources typically do not break down turnover shares by precise percentiles of population; figures are for percentiles close to that shown.

Source: IMF compilation

Entry is percentage of turnover (or, in the case of Egypt, GST revenue) accounted for by largest firms.

Sources typically do not break down turnover shares by precise percentiles of population; figures are for percentiles close to that shown.

The Trade-Off Between Revenue and Collection Costs

If it were not for the costs of administering a VAT (incurred by the authorities) and of complying with it (incurred by taxpayers), the best threshold would be zero: this would maximize revenue (at any given tax rate) while also minimizing distortions of competition between firms of different size. Thus the need for some threshold arises from the willingness to forsake some revenue in order to save on collection costs.

To see what this trade-off might imply for the appropriate level of the threshold, suppose that the government values an additional $1 of revenue at $δ. Clearly one expects δ > 1, since the only rationale for raising revenue is the belief that resources are more valuable to society in the hands of the government than in those of taxpayers. Put differently, since taxation involves costs to the private sector additional to those of the resource transfer itself—because it distorts economic activity—an additional $1 of revenue should only be raised if the uses to which it is put are valued by society at more than $1. Indeed δ-1 can be thought of as corresponding precisely to the deadweight loss associated with the distortion of economic behavior.

Suppose then that such a government considers raising the threshold level of turnover, denoted z, by $1 (taking as given the rate τ at which VAT is levied). For each firm consequently taken out of the tax net, the government loses revenue of τvz (where v denotes value added per unit output, so that tax paid at the threshold level of turnover is τvz) but saves administration costs of, say, A; each firm taken out of tax, on the other hand, gains after-tax income of τvz and saves compliance costs of C. Weighting the net loss to the government by δ and equating it to the gain to the private sector gives an optimal threshold of:

As would be expected, the optimal threshold is higher the more costly is administration or compliance and the less urgent is the need for funds (the lower, that is, is δ). Clearly too it is higher the lower is the ratio of value added to sales. All else equal, there is thus indeed a case for setting a lower threshold for more profitable and/or labor intensive activities.

More important than these qualitative insights, however, are the illustrative calculations that the simple rule in (11.1) allows. For OECD countries, as noted in Chapter 5, Cnossen (1994) estimates that a well-functioning VAT involves administration costs in the order of $100 per registrant and compliance costs of around $500.112 Studies for the United States suggest a value for the marginal value of dollar of tax revenue on the order of $1.20 to $1.50; for illustration, take δ = 1.2. Suppose too that the tax rate is 15 percent and the ratio of value added to sales is 40 percent, the simple rule thus suggests a threshold of about $52,000.

One of the striking features of the rule in (11.1) is that it defines the optimal threshold without reference to the underlying size distribution of firms. This is in apparent contrast to the standard argument above. This difference reflects the importance the standard argument has attached to the existing capacity of the tax administration. That is, the threshold is calculated not by reference to an explicit calculation along the lines above but rather as whatever is needed to restrict the number of taxpayers to fit some given (usually very limited) administrative capacity. This, however, is simply the same rule in another guise; the key feature is simply that when capacity is limited the administrative cost A is implicitly very large.

More generally of course, it is important to recognize that administrative costs are not exogenous: the costs of coping with each taxpayer depend on such design choices as the frequency of audit, the nature of audit, the complexity of the tax structure, and so on. Thus one interpretation of the regularity with which FAD has advised thresholds higher than those subsequently adopted is that FAD has thought the proper costs of audit to be greater than the authorities have. There are clear links here with audit policy, which is discussed at length in Chapter 14.

Distortionary Effects of Differential Treatment Above and Below the Threshold

Another and potentially important set of issues that arise in considering the VAT threshold are the potential distortions of competition, and inequities, associated with the differential treatment of those above and below the threshold. This raises several issues.

First, as noted, there are important circumstances in which it is commercially advantageous to be fully liable for VAT. This includes firms selling zero-rated items and, potentially even more important, those selling to other firms that would wish to register for VAT in order to effectively reclaim tax paid on their own inputs. For these reasons, it is normal practice to allow firms to register for VAT voluntarily. This right is subject, typically, to provisions guarding against temporary or fraudulent registration simply to obtain refunds: deregistration is commonly restricted, for instance (as noted above). Indeed there is a sense in which a reduction in the threshold is self-enforcing: the more firms that are subject to VAT the greater the likelihood that a trader will find themselves selling to registered traders and so will find it advantageous to register too.

Even when it is commercially advantageous to be below the threshold, however, the extent of that advantage should not be overstated. Small traders will be unable to recover VAT on their inputs: it is only their own value added, not the full value of their sales, which escapes taxation. Nevertheless, there clearly is potentially some cause for concern. In particular, firms characterized by a high ratio of value added to sales and selling to unregistered purchasers—small traders providing services directly to final consumers being the key group here—are likely to find it worthwhile to be exempt from VAT. Distortions of this kind have both equity and efficiency aspects.

In equity terms this is, in a sense, an odd worry, since although the size of an enterprise is not necessarily a reliable indicator of equity concerns, the presumption would be that smaller traders are generally poorer—and so presumably more deserving of support—than large. Thus equity considerations would tend to point toward higher thresholds than would otherwise be the case.

The implications of the existence of a threshold for economic efficiency, and their consequences for the appropriate level of that threshold, are complex. One distortion arising from the exemption of small traders is the potential for cascading that exemption always introduces. The provisions for voluntary registration noted above, however, provides a safeguard against such cascading on sales by small traders to registered ones. It is only in connection with transactions between small traders that unrecovered input taxes are likely to cascade into final prices. Of greater concern is the potential distortion of competition, in favor of smaller enterprises, implied by the additional costs—in terms of both the tax liability itself and compliance costs—imposed on those above the threshold. Since smaller firms are likely to be characterized by higher costs than larger enterprises, this effect tends to reduce the efficiency of the market outcome. Indeed, the presumption in contexts of imperfect competition is that it is the small firms, if any, who should be disfavored by tax policy.113 Moreover, the threshold may itself distort output decisions as those firms that would otherwise find it attractive to produce some amount in excess of the threshold choose instead to remain below the threshold, the saving in tax and compliance costs more than offsetting the reduction in sales volume.

Quite what these effects imply for the appropriate threshold remains largely unstudied (a recent exception being Keen and Mintz (2000), on which the analytical parts of this chapter draw). It seems likely, however, that they on balance point to a lower threshold than would otherwise be the case: although one could conceive of leveling the playing field between most companies by setting a high threshold (so that most compete on a tax-exempt basis), this strategy would call, if revenue is to be maintained, for a higher tax rate on those above the threshold, in turn intensifying distortions between the two groups.

Moreover, quite apart from the potential distortion of real economic activity is the potential scope that the threshold creates for the avoidance of VAT by organizing production in a series of sufficiently small enterprises. Legislation typically provides for related firms to be aggregated for the purposes of applying the threshold, though the detection and identification of common control needed for this purpose can be problematic. Artificial splitting of enterprises has certainly been a concern in some countries.114 Again, while the most effective response to this issue is likely to be in terms of anti-avoidance rules, it may also on this account be appropriate to set a somewhat lower threshold than would otherwise be the case, as a means of raising the cost of avoidance by this route.

Relevance of Special Taxes on Those Below the Threshold

The extent and nature of the distortion between those above and below the threshold will depend, of course, on how those below the threshold are taxed. As noted, a number of countries (such as China and Tanzania) apply a gross receipts tax, at a low rate (typically 2 to 4 percent) to such firms; sometimes, indeed (as in China) this tax has its own threshold. Rather more common, however, is some form of presumptive tax.

In the presence of a simple turnover tax on those below the threshold, the calculation underlying (11.1) must be recast in terms of the differential revenues and collection costs associated with the two taxes around the threshold.115 Assume, for instance, that one retains the same parameter values as in the illustrative calculation above, but now supposes a turnover tax at 3 percent to be charged below the threshold. Assuming too that the collection costs associated with the sales tax are one-quarter those of the VAT, the optimal threshold rises from $51,000 to $77,500. More generally, the cheaper the alternative tax is to collect, and the higher the rate at which it is levied, the higher is the optimal threshold for the VAT. As discussed, this simple rule neglects the social costs of the production inefficiency associated with the differential treatment of those above and below the VAT threshold. These costs will be mitigated by the existence of some alternative tax, such as a turnover tax, on those below the threshold. Thus, one would expect the existence of such a tax to imply, all else equal, an optimally higher threshold for the VAT.

Because of the empirical regularity emphasized above, such taxes commonly raise little revenue. Moreover, the administrative resources required to implement even a gross receipts tax may be better allocated to the deployment of the VAT, especially if no such tax originally exists. A turnover tax on small traders is also liable to intensify the cascading effect of their unrecovered VAT. It may be, however, that by assuaging the inequity felt by traders caught in the VAT, a simple tax on those below the threshold will in some contexts increase the political acceptability of a VAT levied with an appropriately high threshold.

What Form of Threshold?

The standard FAD recommendation is for a single threshold, uniform across activities, specified in terms of turnover. One implication of the simple rule (11.1), however, is that, all else equal, the optimal threshold is lower the greater is the ratio of value added to sales and hence the larger is the potential VAT revenue at any given level of sales. This provides some rationale for the practice, which as noted above is observed in a number of countries, of applying a lower threshold to services, since these are typically relatively labor-intensive activities. Against this, however, are to be weighed the practical difficulties of distinguishing between service activities and in dealing with traders conducting multiple activities. These difficulties, which traders have an incentive to exploit, can be expected to intensify over time as economic activity becomes more complex.

A more fundamental question is whether there are better ways of specifying the threshold than in terms of turnover. As noted above, there are two quite different issues. The first is the discontinuity associated with current practice. The second is the case for differentiating thresholds by activity. It might seem that both could be accommodated by specifying a threshold in terms of value added and applying tax, as under the income tax, only to the excess of actual value added over that threshold. This is easily achieved under the subtraction form of VAT. One can simply tax value added above some threshold, exactly the same structure as is normal under the income tax, and in principle apply any rate structure one likes.116 The dominant invoice method, however, does not lend itself to such an outcome. This is partly because calculation of value added requires knowledge of input and output costs net of tax.117


The choice of threshold has proved a crucial element of VAT design.

  • Experience indicates that setting too low a threshold can significantly compromise the political and administrative feasibility of a VAT. This, together with the remarkable degree to which the VAT base is concentrated among a relatively small number of taxpayers, and the limited administrative capacity in many countries, lends support to setting a relatively high threshold (especially at the time of introduction). However, authorities often appear not to have been persuaded of the wisdom of this approach. The reasons for this are not entirely clear: a belief that high thresholds may forego significant revenues, and perceptions of unfair competition, appear to be among the most prominent reasons. It may be that the experiences of Ghana and elsewhere will make the case for a sufficiently high threshold more persuasive in the future.

  • Countries appear more inclined to differential thresholds by sector than FAD has advised; there is, though, some theoretical support for a more differentiated threshold.

  • The tax treatment of those below the VAT threshold has received scant attention, and practice varies, though countries quite frequently adopt simplified methods for small traders. This is evidently a second-order issue in terms of direct revenue impact, and innovation in this area at the time of introduction of a VAT may divert resources. Given, however, the potential importance of treatment below the threshold for the extent of registration under the VAT, for the development of compliance capacity, and—not least—the importance that governments appear to attach to the competitive positions of those above and below the threshold, a simple tax below the VAT threshold may have a beneficial effect disproportionate to the revenue it yields.

    Other Resources Citing This Publication