I am pleased that Vito Tanzi takes my article (Dixit (1991)) sufficiently seriously to comment in such detail (Tanzi (1992)). Perhaps he takes it too seriously. I do not believe that in any foreseeable future a finance minister or ministry official will justify a high inflation rate by appealing to my results: if any made the attempt to read my article, they would surely quit at the first equation. And if one did persevere and try the argument, the IMF need only point out that the policy that emerges as optimal in my model is a package—a combination of inflation and commodity taxes—and one cannot pick the former aspect and ignore the latter.

As one might expect, I agree with some of Tanzi’s comments and have varying degrees of disagreement with others. Let me briefly state the main points of agreement and disagreement and then elaborate.

I agree with Tanzi when he places my treatment squarely in the Ramsey-Diamond-Mirrlees approach to public finance, which combines the Arrow-Debreu general equilibrium analysis with the Bergson-Samuelson social welfare maximization. To the extent that he criticizes the underlying assumptions of rational expectations, symmetric information, and complete markets as inadequate for modeling inflation and capturing all its costs, I agree. When he says that administrative and political constraints on taxation are important, I also agree but do not think he recognizes their full implications. But when he says that the inadequacy of general equilibrium theory justifies partial equilibrium insights, I disagree. To me, his arguments call for going beyond general equilibrium by recognizing the importance of asymmetric information and incomplete markets, not retreating to partial equilibrium, which ignores the most important and robust feature of general equilibrium—namely, the interconnections of different parts of the economy and of a policy package.

In the Arrow-Debreu framework, or its rational expectations extensions to deal with uncertainty, all private contracts can be, and will be, fully indexed. The only distortion induced by inflation will be the one associated with the need to hold non-interest-bearing money because of the transactions technology usually captured in cash-in-advance constraints. This is what the Friedman-Phelps model began, and what the models of Végh (1989) and others, as well as my own model (Dixit (1991)), have adopted.

I fully agree that the Friedman-Phelps formulation can handle only a relatively minor aspect of the economic cost of inflation. In a model with rational expectations and full indexation of contracts, there is no “arbitrary rearrangement of riches [that] strikes not only at security, but at confidence”; that framework does not allow “all permanent relations between debtors and creditors [to] become so utterly disordered as to be almost meaningless”; Keynes (1920, pp. 220) rightly emphasized these aspects.

But when one adopts a theoretical framework, one must follow its logic. Therefore I cannot agree with Tanzi’s “fifth comment” (p. 960 and footnote 5) that pertains to my theory itself. I do not assume that in equilibrium competitive firms would pass on to the consumer their windfall gain from holding the tax revenues for the period of the collection lag; rather, it is a logical implication of the free-entry condition of competitive equilibrium, going back to Marshall and beyond. Tanzi’s “real redistribution of income from consumers to sellers” can occur only in disequilibrium, and even then one will have to ask what happens to income accruing to impersonal entities called firms.

Tanzi’s criticism of my one-person framework is also weak unless it is construed as a criticism of the whole Arrow-Debreu framework. Heterogeneity becomes important only in conjunction with disequilibrium. In a world where all contracts are appropriately indexed, the distributional effects of inflation arise only because different people have to hold different amounts of fiat money for transaction purposes. It is not clear to me which way the redistributive impact will work. The poor will generally have a lower shadow value of time and so can make more trips to the bank and hold lower average cash balances, so they should be affected less in this respect than the rich. But in any case, I would expect the effect to be quantitatively small. In a world with incomplete markets and costly transactions, the poor may have less access to indexed contracts, and this may be a bigger and normatively perverse source of redistributive cost.

Tanzi’s criticism regarding practicality is twofold—administrative and political. He raises some pertinent administrative issues for commodity tax theory as a whole: gathering reliable information about elasticities, the weak links in the chain from an economist’s prescription to the actual tax code, and so on. I agree with many of these points, but do not see that collection lags present many additional difficulties. In my base case, charging interest on the accrued tax revenues held by firms is entirely equivalent to changing the rates to new optimal levels. This takes care of the problems Tanzi points to in his first four comments in Section I. Ex ante legislation can lay down a rule that such interest will be compounded daily at the then prevailing rate of interest. That removes the need for high statutory rates, or for precise forecasting of future inflation. Tanzi asserts later that many high-inflation countries cannot administer such interest charges. But he does not give any arguments, and merely cites a paper in Portuguese not accessible to me, so I am unable to comment.

Incidentally, the administrative cost argument may end up implying a greater reliance on the inflation tax. Inflation is generally regarded as an easy tax to collect, bypassing the normal channels of legislation and administration. As Friedman (1962, p. 39) put it, “control of money has, from time immemorial, enabled sovereigns to exact heavy taxes from the populace at large, very often without the explicit agreement of the legislature.” Therefore if collection of commodity taxes is more costly, it becomes optimal to rely more on inflation.

Tanzi correctly points out that inflation and taxation policies emerge from political processes and approvingly cites the “public choice” work on tax theory in contrast to the Ramsey-Diamond-Mirrlees approach. I share his admiration for the many insights the political economy approach to taxation has provided. But I do not think Tanzi realizes its full import. If carried to its logical conclusion, that view makes any normative or prescriptive work in public finance pointless. The tax system, like all government policy, is the endogenous outcome of the political process. As economists, all we can do is to study and describe this politicoeconomic equilibrium. There are no degrees of freedom to manipulate or affect the outcome.

Finally, let me repeat what I regard as the central and robust message of my paper. While the general equilibrium approach has flaws, one should distinguish between the “general” and the “equilibrium” aspects. The latter, particularly in the Arrow-Debreu framework, leaves much to be desired. The former, emphasizing that different sectors of an economy and different components of a policy package are interconnected, is important in any framework. Thus, I believe that the starting point for my article retains its validity independent of its model-specific development there. If collection lags are important, we should recognize their implication for the whole tax system and not just for one component of the tax package. When this is done in a model that satisfies Tanzi’s concerns for disequilibrium and administrative constraints, inflation may turn out to play a small role in the optimal package, but that remains to be seen.

I hope this exchange will spur further research in this area, recognizing many practical aspects that Tanzi emphasizes, but respecting, as I stress, the need to examine the implications of these issues for the whole tax system.