Theory and Policy A Comment on Dixit and on Current Tax Theory
Author: Mr. Vito Tanzi

In a recent paper Avinash Dixit criticized the argument that when collection lags characterize tax systems, recourse to inflationary finance should be minimized. He argued that, in such cases, rather than minimizing recourse to inflationary finance, the rates of the commodity taxes should be adjusted to maintain them at an optimal level and, thus, to minimize welfare cost. This paper shows that the requirements for following Dixit’s policy prescription are almost impossible to meet. The paper argues that tax theorists should pay more attention to the constraints under which tax reforms are made.


In a recent paper Avinash Dixit criticized the argument that when collection lags characterize tax systems, recourse to inflationary finance should be minimized. He argued that, in such cases, rather than minimizing recourse to inflationary finance, the rates of the commodity taxes should be adjusted to maintain them at an optimal level and, thus, to minimize welfare cost. This paper shows that the requirements for following Dixit’s policy prescription are almost impossible to meet. The paper argues that tax theorists should pay more attention to the constraints under which tax reforms are made.

In his characteristically incisive style, Avinash Dixit has provided an interesting perspective on the optimal mix of inflationary finance and commodity taxation when collection lags are present (Dixit (1991)). The addition of collection lags differentiates his analysis from previous ones such as Phelps (1973) and Végh (1989). He concludes that the “partial equilibrium insight” (developed by Tanzi (1978)) is unlikely to give a theoretically correct policy. That “insight” suggests that, when collection lags are significant, inflationary finance should not be used or at least should be minimized, since, by raising the rate of inflation, it reduces tax revenue. Therefore, what the country could gain in revenue from inflationary finance it would lose from the fall in ordinary tax revenue while still paying a price in higher inflation. This partial analysis assumes no change in the statutory tax system. Dixit’s major preoccupation is with the welfare cost of revenue collection, whereas my own preoccupation (in the 1978 paper that stimulated Dixit’s paper) was with revenue and stabilization.

Dixit shows that, with constant average cost of collection of commodity taxes, “the optimal inflation tax is independent of the length of the lag in the collection of commodity taxes” (Dixit (1991, p. 644)). Furthermore, “if the average cost of collecting commodity taxes is a stable and rising function of the real [tax] revenue, then a longer collection lag implies a higher optimal inflation tax rate” (p. 644). In other words, “it becomes optimal to shift some of the tax burden onto the inflation tax” (p. 644). In general, the erosion in tax revenue associated with the higher rate of inflation “can be made good by changing the rates of commodity taxes” (p. 644).

Dixit recognizes that “the particular model studied [in his paper] embodies many specific assumptions” (p. 653) and is limited to commodity taxes. He believes that similar analyses can be constructed for taxes other than commodity taxes and concludes that “[w]hen we recognize the existence of a collection lag, our [optimal policy] response should be to recalculate the whole tax structure, and not merely the inflation tax” (p. 653).

Dixit’s theoretical analysis is based on a one-person economy. Although the representative consumer framework may be appropriate for addressing efficiency issues concerning optimal commodity taxation, the problem of inflationary finance has an inherent redistributive dimension (discussed below) that cannot be ignored. Thus, this class of models (including the one-person Ramsey rule) may not be particularly useful from a policy standpoint (see, for example, Ahmad and Stern (1991)). As the analysis is broadened to include different types of households, the policy recommendations implied by the theory are likely to change.

Dixit addresses the welfare costs of nonoptimal commodity taxes but not the likely more serious general costs of inflation. He thus implicitly assumes that inflation is neutral. I strongly feel that inflation has serious costs; however, I will not focus on this important aspect. Rather, I will discuss the practical limitations of his analysis. Although it may be argued that it is somewhat unfair to comment on a theoretical paper by highlighting its practical limitations, this discussion is necessary to avoid the possibility that, on some future occasion when an IMF mission recommends to a country a reduction in its reliance on inflationary finance, the minister of finance of that country, or one of his advisors, might pull out the September 1991 issue of Staff Papers to argue that this policy recommendation is not theoretically optimal. It is also necessary to highlight the great gap that yawns between normative tax theory and practical tax policy. In fact, although this comment has been stimulated by Dixit’s paper, its objective is to show more broadly the practical limitations of current normative theories of taxation.

I. Comments on Dixit’s Paper

This section focuses on the difficulties that would be encountered by policymakers in following the implicit policy recommendation of Dixit’s paper—that the government could increase its reliance on inflationary finance, despite significant collection lags, if it compensated for the inflationary effect on existing taxes by raising and calibrating the rates of these taxes. Let me outline some of these practical difficulties.

First, if a new rate of inflation is reached, consistent with the higher level of inflationary finance, an adjustment to the (presumably multiple) rates of commodity taxation would need to be made to keep them at their optimal level. Quite apart from the availability of the required knowledge concerning the elasticities,1 an inadequacy magnified by the fact that these elasticities would be changing during inflation, there is the question of the stability of the inflation rate. In a theoretical analysis that rate may simply be assumed to be constant, and new commodity tax rates may be calculated accordingly. In the real world the inflation rate will be highly unstable, as recent experiences of countries that have undergone high inflation rates have shown (for example, Argentina, Brazil, and Peru). This instability implies that, to remain optimal, the tax rates would have to keep changing with the rate of inflation: first, reflecting the moving inflation rate; and second, reflecting the changing elasticities. Incidentally, should these tax rates adjust ex ante, to the expected rate of inflation, or ex post, to the actual rate? Presumably, they would have to adjust ex ante in order to anticipate and counter the effects of collection lags. Ex ante adjustment would comprise an even more difficult policy to implement.2

Second, in most countries the legislature must approve the changes in the statutory tax rates.3 Given the variability in the rate of inflation and the typical sluggishness of legislators in their reactions, the difficulties can be easily imagined. By the time the changes in the rates are made, they are likely not to be optimal any more. Furthermore, the legislators wilt not be guided by the same theoretical principles that guide tax theorists and that, under the best of circumstances, might guide the policymakers in the executive branch. Thus, if the legislators are given the opportunity to change the tax rates, they are unlikely to make changes consistent with the theoretical principles.

Third, if the rate of inflation is very high, as it may easily be in situations in which the government attempts to raise its dependence on inflationary finance, the required optimal tax rates would also become very high, thus encouraging smuggling and evasion and raising the cost of collection.4 If the increase in the cost of collection is accompanied by even higher inflation tax rates, as suggested by Dixit’s model, smuggling and tax evasion would be stimulated even more, thus leading to a highly unstable situation, since increasing costs of collection would require higher inflation rates; and higher inflation rates would raise the cost of collection even more. In this situation, calibrating the tax rates to make them optimal would become a real nightmare.

Although Dixit’s paper deals with commodity taxation, it implies that the results can be generalized to other taxes. He specifically mentions taxes on labor income. One can imagine the reaction of legislators if, given significant collection lags and relatively high rates of inflation, they were told that the statutory tax rates on the incomes of individuals might have to exceed 100 percent. Such high rates could be required to neutralize the effect of the collection lag on real tax revenue. Incidentally, the ad valorem rates of commodity taxes may also have to be raised to high levels in the face of high inflation and long lags.

The preceding comments may not challenge the theoretical conclusions Dixit reaches but do call in question their practical relevance. However, a fifth comment has implications for the theory itself. In what I regard as the more realistic of the alternatives he considers, “[t]he ‘regular’ tax is levied on consumption, it is collected by the firms in advance, and held for the duration of the collection lag” before being transferred to the government (p. 653).5 This implies that the gain in paying commodity taxes with a lag accrues to the sellers (that is, the firms or the shops) and not to the buyers of the commodities who effectively bear the tax. The buyers pay the tax immediately when they buy the product, gaining nothing from collection lags. Therefore, for the consumer the increase in the tax rates necessary to compensate for the potentially high rate of inflation could be large and real. At the same time, the collection lag that pertains to sellers could well continue to grant them windfall gains.

If this alternative is, in fact, more realistic, Dixit’s implicit recommendation (that is, to rely more on inflationary finance while raising the commodity tax rates) would imply a real redistribution of income from consumers to sellers. The consumers would experience an erosion of their cash balances (because of inflationary finance) and of their disposable incomes (because of the higher commodity tax rates). This may explain why heterodox stabilization programs (as pursued by Argentina, Brazil, and Israel in the middle 1980s) were popular with consumers. By freezing prices, these programs raised the real income of consumers while they reduced the real income of sellers. Of course, one could make other assumptions (that is, that consumers themselves pay with lags for the goods they buy; or that, as Dixit also assumes, the tax reduction to the sellers is shifted partly or fully to the consumers in lower prices) that might give different results. But, in developing countries these alternative assumptions would not be more realistic.

Dixit also discusses the possibility that the erosion connected with collection lags might be neutralized by charging interest on the tax liability. This alternative has been recommended before (see, for example, Tanzi (1977, p. 165)) and has been tried by a few countries, especially by Brazil. Unfortunately, it has been difficult to apply this measure equitably and efficiently (see Gianbiagi (1987)). It becomes particularly difficult to implement fairly when the rate of inflation is high. A better policy option is to reduce the rate of inflation. If that fails, the proper practical policy is to reduce the length of collection lags while trying to index the liabilities.

In conclusion, for the reasons that I have outlined, and also because inflation has so many costs, which may easily exceed those connected with the use of nonoptimal taxes, I feel that the message obtained from the “partial equilibrium insight” is still the best practical message to give to policymakers: especially when collection lags are significant, inflationary finance is a highly unsatisfactory way of raising money. By casting doubts on this message, economic theory, even if correct as theory, may encourage the pursuit of inferior and costly policies.

II. General Observations on Current Tax Theory

I would like to take advantage of this comment to make a few general and highly personal observations on the type of analysis typified by Dixit’s paper. The normative analysis, based largely on theoretical developments that took place in the early 1970s, has monopolized advanced courses in public finance.6 As I have indicated, apart from a few details, I have no basic disagreement with Dixit’s theory as theory. If the informational, administrative, and political requirements necessary to implement it could be met, I would surely like to base the advice that we give to countries on conclusions reached by papers such as Dixit’s.7 However, normative theoretical developments have largely overlooked those requirements, so it is an open question whether the conclusions reached by that work could be implemented. If applied, they might easily lead to results that are less optimal than the theory would lead one to believe.

Much of the current theoretical work in taxation has been concerned with minimizing welfare costs in highly constrained and abstract situations, rather than with attempting to make these theories implementable in a highly imperfect world. Theorists blame practitioners for ignoring theory; practitioners blame theorists for ignoring reality. In either case there is little dialogue between the two groups. As a consequence, much of the recent theoretical writing on taxation has provided little guidance to tax reformers despite the considerable talent of those who formulate the theories. (See also Aaron (1989), Slemrod (1990), and Harberger (1990) on this point.) Many recent tax reforms have been inspired by principles other than those espoused by taxation theory. In fact, it may not be much of an exaggeration to say that concerns about underground economic activities and tax evasion may have had more influence on tax policy than recent taxation theory. This might not have happened if theorists had made a greater effort to popularize the main results of tax theory and, especially, to deal with the practical limitations of that theory.


Let me consider informational requirements first. As Arnold Harberger (1990) forcefully argued in a recent paper: “While the mathematics underlying the Ramsey Rule solution cannot be denied … [that] solution requires far more knowledge of supply and demand relations than we actually possess” (p. 4). He added that this “… represents the area of the economy where our data are poorest” (p. 6). A similar conclusion was reached by Angus Deaton (1987), who, after surveying these statistical requirements concluded that “[t]he global knowledge of demands and of preferences required for optimal taxation is simply not obtainable in developing countries nor probably in developed countries” (p. 112).8

Without this information, tax practitioners who attempt to use optimal taxation theory are forced to rely on highly inadequate data (say, old and imperfect data or data borrowed from other countries) or on arbitrary and debatable guesses. Of course, if reliable data are not available in normal conditions, they are going to be even less available under conditions of high inflation. When information about elasticities is available (as with some products, or, for example, with respect to the supply of labor of different groups, such as head of household, spouses, and other dependents), this information should be, and occasionally has been, used. The point is that the theory should not continue to assume that elasticity information is normally available.9


As far as the administrative requirements are concerned, the implicit assumption of much recent tax theory seems to be that administrative problems are trivial or, alternatively, that tax administration is always capable of enforcing any theoretical solution. However, especially in developing countries, the cost of administering alternative tax systems can vary a lot. More important, some systems may not be administrable at almost any cost. When countries try to enforce these systems, they end up with “effective” systems that are very different from the legislated ones. Even if the legislated systems appear optimal, the effective ones are unlikely to be so. Therefore, the efficiency gains of alternative options cannot be compared with the additional administrative costs imposed by these options to determine which alternative is more efficient.10 It is puzzling how little attention tax administration problems have received in the public finance literature.11 A better knowledge of tax administration on the part of those who contribute to the theoretical literature would probably change some of the theoretical conclusions and would make that literature more influential among practitioners.

The issue of administrative constraint has been faced most dramatically in the case of whether a value-added tax should have one or several rates. An increase in the number of rates might make the value-added tax theoretically more efficient (and more equitable) but it entails large administrative complications and thus raises collection costs and tax evasion. This is the main reason why value-added taxes with uniform rates are supported by many tax experts, although they conflict with the conclusions of optimal taxation theory.12 As Harberger has argued, the case for uniformity of rates in indirect taxation is based, and should be based, on practical rather than on theoretical considerations. Similar considerations have led other economists to similar conclusions (see Deaton, (1987, p. 112)). If the alternative is to use questionable data to bring into existence a (presumably optimal) tax system that is not administrable, it is not clear that the final product is better than one reached by following more traditional but less theoretically sound tax principles.

Political Constraints

Let me now add a few remarks about political requirements. Tax reform is an eminently political activity. It is an activity carried out by policymakers who may have all kinds of professional backgrounds and may be guided by all kinds of objectives rather than by the maximization of a “social welfare function.” These policymakers may be lawyers, accountants, administrators, farmers, architects, medical doctors, professional politicians, and, occasionally, even economists. These individuals, including those in the legislature who vote on the tax laws, often feel that they are fully competent to judge the merit of tax reform proposals.13

They are unlikely to delegate to professional economists the major role in making the choices, and they are unlikely to give the same weight to the objective of allocative efficiency as professional economists do.

All this means that, in order to have influence, economists must try to convince these individuals of the importance of efficiency considerations. To do this, they must learn to speak the language of average policymakers and must promote their points of view while addressing the concerns of these policymakers. They cannot leave this function to others. Arguments that require advanced economic theory to be understood are unlikely to impress those who make policies. The latter often dismiss these arguments as “academic,” by which they normally mean that they think that these arguments are irrelevant. Citing Harberger again: “if the economists are to carry out their responsibility to society, they must as a group maintain communication with the public at large… [t]he average economist should be able to communicate effectively and simply with the average citizen” (Harberger (1989, p. 3)). This is especially true for taxation, although, of course, it does not mean that each article written must satisfy this requirement. Articles written largely for graduate students must be rigorous and theoretical. But their authors should not come to believe that these articles are setting the guidelines for tax reform in the real world. Before they can make this claim they must pay full attention to the informational, administrative, and political constraints mentioned above and must incorporate the implications of those constraints in their writing. It would also be beneficial if the students themselves became aware of these constraints. On the other side, those who deal with administering tax systems must make a greater effort to draw from their work and experience basic lessons that can be used by theorists.

III. Concluding Remarks

Recent theoretical work on taxation has played an important role in calling attention to the efficiency costs of taxation, which can be significant. These costs had been largely ignored or minimized by public finance experts until the early 1970s and continue to be ignored today by many tax practitioners. Recent academic work has been conducted at too high a level of abstraction and has ignored the important constraints faced by tax reformers. It may thus be time to redirect attention from what is optimal to what is possible while trying to retain as much as is feasible of the main message of tax theory. This more down-to-earth approach to tax theory would surely increase the practical value of taxation theory.


  • Aaron, Henry J., “Politics and the Professors Revisited,” Richard T. Ely Lecture, American Economic Association, Papers and Proceedings, Vol. 79, (May 1989), pp. 115.

    • Search Google Scholar
    • Export Citation
  • Ahmad, Ehtisham, and Nicholas Stern, The Theory and Practice of Tax Reform in Developing Countries (Cambridge: Cambridge University Press, 1991).

    • Search Google Scholar
    • Export Citation
  • Brennan, Geoffrey, and James M. Buchanan, The Power to Tax: Analytical Foundation of a Fiscal Constitution (Cambridge: Cambridge University Press, 1980).

    • Search Google Scholar
    • Export Citation
  • Deaton, Angus, “Econometric Issues for Tax Design in Developing Countries,” in The Theory of Taxation for Developing Countries, ed. by David Newbery and Nicholas Stern (New York: Oxford University Press, 1987).

    • Search Google Scholar
    • Export Citation
  • Dixit, Avinash, “The Optimal Mix of Inflationary Finance and Commodity Taxation with Collection Lags,” Staff Papers, International Monetary Fund, Vol. 38 (September 1991), pp. 64354.

    • Search Google Scholar
    • Export Citation
  • Gianbiagi, Fabio, “O ‘Efeito Tanzi’ e o Imposto de Renda da Pessoa Física: Um Caso de Indexaçao Imperfeita,” Revista de Finanças Públicas, Vol. 47, (July/August/September 1987).

    • Search Google Scholar
    • Export Citation
  • Harberger, Arnold C., “The Economist and the Real World,” Occasional Papers, No. 13 (San Francisco: International Center for Economic Growth, 1989).

    • Search Google Scholar
    • Export Citation
  • Harberger, Arnold C., “The Uniform Tax Controversy,” in Public Finance, Trade, and Development, Proceedings of the 44th Congress of the International Institute of Public Finance, Istanbul, 1988, ed. by Vito Tanzi (Detroit: Wayne State University Press, 1990).

    • Search Google Scholar
    • Export Citation
  • Newbery, David, and Nicholas Stern, eds., The Theory of Taxation for Developing Countries (New York: Oxford University Press, 1987).

  • Phelps, Edmund S., “Inflation in the Theory of Public Finance,” Swedish Journal of Economics, Vol. 75 (March 1973), pp. 6782.

  • Slemrod, Joel, “Optimal Tax Systems,” in Journal of Economic Perspectives, Vol. 4 (Winter 1990), pp. 15778.

  • Stern, Nicholas, “Aspects of the Genera! Theory of Tax Reform,” in The Theory of Taxation for Developing Countries, ed. by David Newbery and Nicholas Stern (New York: Oxford University Press, 1987).

    • Search Google Scholar
    • Export Citation
  • Tanzi, Vito, “Inflation, Lags in Collection, and the Real Value of Tax Revenue,” Staff Papers, International Monetary Fund, Vol. 24 (March 1977), pp. 15467.

    • Search Google Scholar
    • Export Citation
  • Tanzi, Vito, “Inflation, Real Tax Revenue, and the Case for Inflationary Finance: Theory with an Application to Argentina,” Staff Papers, International Monetary Fund, Vol. 25 (September 1978), pp. 41751.

    • Search Google Scholar
    • Export Citation
  • Végh, Carlos A., “Government Spending and Inflationary Finance,” Staff Papers, International Monetary Fund, Vol. 36 (September 1989), pp. 65777.

    • Search Google Scholar
    • Export Citation

Vito Tanzi is the Director of the Fiscal Affairs Department.

He thanks Ehtisham Ahmad, Dieter Bös, Parthasarathi Shome, and Howell Zee for comments on an earlier draft.


This issue is discussed later.


Governments typically do not have direct control of the inflation rate as they do (in theory at least) of tax rates. This is because the translation from monetary growth rates (which they can control) to inflation rates does not have a one-to-one correspondence except in steady states. Steady states, of course, are useful for classroom discussions but not for guiding real world policies. Thus, it may be questionable to treat the inflation tax (which is an implicit tax) as something commensurate with other regular, explicit taxes.


It is highly unlikely that the legislature would ever give the executive branch the power to make these changes.


These difficulties can be assumed away in a theoretical model but not in real life situations.


Dixit also assumes that, in equilibrium, competitive firms would pass on this gain to the consumers in the form of lower prices. This may not be a realistic alternative in the situation he is dealing with.


So far, the main challenge to this analysis, at least in public finance, has come from James Buchanan and the public choice literature. (See Brennan and Buchanan (1980).)


would still be allergic, however, to recommending policies that raise the rate of inflation.


Nicholas Stern (1987) was more optimistic about the extent of our knowledge in this area.


Ramsey rules have found more ready applications in public sector pricing than in taxation, in part because of the greater availability of information concerning demand elasticities in public enterprises.


If the question were mainly one of cost of administration, and if these costs could somehow be quantified, the issue would be one of simply comparing at the margin the quantified efficiency gains of a tax reform with its administrative costs. The optimal alternative would be the one that would maximize the difference between gains in efficiency and additional administrative costs including compliance costs.


This comment is not directed just to the “theorists.”


Of course, no value-added tax is ever truly uniform, since some activities are always exempt or, as in the case of leisure, cannot be taxed. A combination of value-added taxes with single rates with a well-selected use of excise taxes might bring the tax system in the direction of the theoretical conclusions. But even in this case, the informational requirements are substantial.


And surely the lobbies that represent various vested interests know exactly what they want from a tax reform.