Edward H. Clarke, USAID/Haiti, Port-au-Prince, U. S. Dept. of State, Washington, D. C. 20520

0. Introduction

This review and comment was initially stimulated by a review article, entitled "Arrow Reconsidered" by Gordon Tullock (1987, unpublished)1/ In commenting, I was motivated to consider some of the recent developments in demand revelation and their bearing on the issues raised by Tullock. For this reason, I deal with several of the controversies that were spawned by his article with Tideman (1976 and special supplement,1977), including a description of my recent efforts to deal with several of these questions. I conclude with a brief comment on the implications for several of the issues raised in Tullock's review.

In the first paragraph (and footnote) of their 1976 article, Tideman and Tullock allow that, while the demand revealing (DR) procedure does not exactly solve the Arrow problem, it avoids the conditions of the Arrow problem by using more information than the rank order of preferences. The procedure selects a unique point on "or almost" on the Pareto-optimal frontier. (When decision costs are taken into account, it provides a welfare criterion, in their opinion, superior to Pareto optimality). The authors also added that subject to any distribution of wealth, the process may be used to approximate the Lindahl equilibrium for all public goods. (See also Tullock, special supplement).

On the problem of Arrow, Tullock's recent review basically comes down to the issue of whether it is advisable to let someone (a neutral observer, called here a Ruler) make cardinal , interpersonal welfare comparisons or whether people are capable of doing so for themselves.2/ As noted by Mueller, there would appear to be two distinct justifications for wishing to exclude cardinal utility functions from a collective choice process, the reasons being that:

measurement is difficult and arbitrary

the measurement would be vulnerable to abuse by those making the cardinal measures

According to Mueller, the latter seemed to be Arrow's chief fear, mainly because he assumed that the choice process is one in which information is gathered by public officials who make the actual choices for the collective (pages 106-108). Allowing them to make cardinal utility comparisons would vest them with a good deal of discretionary power, and might be something to be avoided.

As Mueller points out, the danger of such abuse does not arise when the information provided is by the voters themselves, although most procedures permitting expression of preference intensity allow for individual strategizing. DR is immune to individual, but not coalition strategies. Yet the existance of the latter and the problem that individual measurement (of utilities) is perceived as too difficult appears to be the chief impediment to practical solutions for avoiding the Arrow difficulty.

It is agreed that if someone, a tax-setter, can come reasonably close to, in fact, determining a Lindahl equilibrium, none of these problems need arise. The ability of someone to do this, however, is questioned as betraying a naive belief in the power of an econometrician (Cox, 1981). In summarizing the status of this work in terms of the general problem raised by Samuelson, Musgrave (among others) has questioned the second of Tideman and Tullock's major claims. Musgrave (1983, p.150): "I see no basis for this".

After a decade of controversy over the basic claims, and sharp debate over several of the subsidiary issues (information and coalition incentives, the "waste" of the DR process, etc. -- see 1. below), I seek to stimulate the proponents of this and related processes to devote extra energies to problem resolution in specific practical settings. This may also require more normative agreement on the roles that various decision procedures can play in differing circumstances. Toward this end, I would like to advance some practical ideas on the role of DR in regulating a myriad of social choice proceeses. This begins with some restrictions on the DR process itself, as first suggested by Buchanan (see Clarke, 1977), and to embody these restrictions in the form of a Ruler.

The Ruler provides for greater stability in the distribution of benefits, motivated by improved incentives and provisions for ex post equity in the distribution. To also achieve greated efficiency, I have modified the objective function of the tax-setter of Tideman and Tullock (1976) and my judge (Clarke, 1971), and given the Ruler of Thirlby (1946) broad discretion in determining what decision rules should apply within an organization and in making compensating transfers to reflect external objectives imposed on the enterprise. All this is spelled out in my Demand Revealing Governance of Enterprise (Clarke, 1987). Here I will treat some aspects of the procedure that bear on more general decisional settings, suggest several important dimensions that are important in these settings (e.g. who selects the Ruler, how he is deposed) and conclude with some comment on how this bears on some of the more important controversies, including the issues recently raised by Tullock's review. My central point is, however, somewhat more modest, and deals mainly with the second of the major claims set forth in Tideman and Tullock -- the achievement of Lindahl equilibrium and the treatment of decisional costs in defining that equilibrium. (see also Tullock, 1977, p. 60).

1. How to Regulate Demand Revealing (and Democracy).

The first of two new distributional arrangements is an incentive calculus for the Ruler, suggested to me by T. N. Tideman. This procedure would serve to encourage realization of several goals of a collective decision system -- outcome and procedural efficiency, and distributional stability. 3/

The basic idea requires that the participants at the chartering stage determine the objective function of a Ruler and how he is to be compensated. They must also decide what level of social discount is to be applied to the benefits of proposals, so as to reflect the collective cost (to them) of unintended distributions.4/

The incentive arrangement would, for example, adjust the estimated net benefits of decisions for Clarke taxes and votes in favor of the status quo (and administrative costs as well) and the reward of the Ruler would be some proportion of these adjusted net benefits. This procedure would also be administered in conjunction with an insurance arrangement that would motivate the Ruler to minimize Clarke taxes and insurance premiums while maximizing net benefits (to also reflect appropriately weighted unintended redistributions). His salary would be tied to this objective function.

Under this procedure, I maintain that the costly resort to interpersonal utility comparisons would likely be rare, and that when made, would probably most often result in the selection of a potentially less costly method (e. g. the one suggested by Arrow and Reynaud, Borda count, point voting, etc.) In many circumstances, the enterprise (or other collective entity) could also live with the incentives for strategic misrepresentation, although these may be substantial. See, for example, Nitzan (1985) on point voting.

In work in progress on the Demand Revealing Governance of a Community, I suggest that those who permit the use of DR in their constitution would be likely to actually use it only in the extreme event that they are led to depose the Ruler. DR, in fact becomes a welfare criterion which I have always understood to be the purpose that Tideman and Tullock envisioned for the procedure. Nevertheless, this review and the comment I solicit from Tideman, Tullock and others provides and opportunity to clarify this intent and to further explore the restrictions on the procedure as well as precisely how it might be used to regulate other processes.

I will begin with an example, drawn from a recent French public choice textbook by Hannequart and Graffe (1985) which does not mention DR but is otherwise an excellent exposition of modern public choice (and property rights) theory and a good way to improve one's French. I've borrowed their diagrams with English translations.

Hannequart and Graffe's textbook description proceeds with a demonstration first of achieving a Lindahl equibrium by a market process (Figure 1 ), then by majority voting (Figure 2) and finally, the result of assymetric preferences (Figure 3) where majority voting by a pairwise ordering of what would be preferred by each voter (A', B' or C') would lead to cycling. The idea underlying my recent suggestion is that when the conditions illustrated by Graph 3 arise, the Ruler or any individual in the collectivity be permitted to call for a demand revealing vote, in order to try to get closer to the Pareto-optimal outcome (B in Figure 3). In seeking this solution, the DR procedure also requires that the Ruler seek a Lindahl equilibrium (Figure 1) by an appropriate reallocation of the tax shares. In this way, he will maximize his objective function, which adjusts net benefits for unintended transfers, Clarke taxes and decision/administrative costs.

One might question the uncompensated transfers (the shaded areas or "pertes") in Graph 2 that are not accomodated by the procedure, which we normally live with under majority rule and simple numerical examples (which the reader can construct) under the DR procedure show that DR votes (even when there are no decision costs) leave these pertes (or Buchanan-Tullock external costs) undisturbed. However, when the pertes grow very large (Hannequart and Graffe left the large perte on individual 3 in Figure 3 to be drawn by the reader -- in its place I've substituted a diagonally shaded Clarke tax), we have need for some other procedure. The fact that DR is not even mentioned as a route out of this difficulty in a basic public choice textbook is sufficient comment on the lack of impact of that DR has in many ways had, perhaps because many continue to believe that the route to Lindahl equilibrium is too difficult or uncertain

My current efforts to reestablish faith in demand revealing is premised on the view that a procedure for moving efficiently towards a Lindahl equilibrium (Figure 1) will satisfactorily address several of the major subsidiary controversies surrounding DR -- information and decision efficiency, coalitions, budget balance (waste), as well as the problem of authority (limits on the range and applicability of the procedure) --see Clarke 1977, 1980) This also bears on the authority or discretion given the Ruler, within the enterprise or in any other collectivity.

I will turn to these subsidiary questions after elaborating on some aspests of the motivational calculus needed to obtain a Lindahl equilibrium, including the new insurance principle.

1.1 Achieving a Lindahl Equilibrium: A DR Perspective.

Nothing can be so stimulating in DR theory as thinking about running an enterprise with it -- or, for example, replacing shareholder voting in appropriate circumstances (See Clarke,1987). For hard-headed businessmen, you would have to show it contributes to "bottom-line" profits, as I will try to do in ways that suggest a "practical" solution to the problems that continue, in Musgrave's words, to also haunt "fiscal theory".

In brief, I seek "efficient unanimity" with a corporate accountant (Thirlby's Ruler)who is given the function of assigning costs to items that get on the agenda. In assigning costs, we want the Ruler to make trade-offs between the delays (and other costs) of the Lindahl tax-setting process and the pertes, or Buchanan-Tullock external costs (which are zero if the tax-price assignment process were perfect on all issues). It will, of course not be, and if the errors of the process were such that there were "systematic" correlations between who happens to lose in one application vs. another, then the law of large numbers should come close to guaranteeing that all benefit equally from all applications taken together (Polinsky,1972, Tideman, 1981).

Although one could construct a variety of "performance indicators" that would be appropriate in differing circumstances, we seek basically to motivate the Ruler to make effective trade-offs between (a.) delays and decision-costs and (b.) Buchanan-Tullock external costs, so as to maximize net benefits and the process of efficient Lindahl tax-setting.

A simple example will serve to introduce this notion, including both an "incentive calculus" for the Ruler and a new "insurance principle" which complements the Clarke tax mechanism by giving back the latter to those who suffer the most significant pertes (Buchanan-Tullock external costs) over a range of issues. Illustrating the trade-off between seeking a Lindahl equilibrium (Figure 1) and relying on majority voting (Figure 2), a Ruler faced with the appropriate decision calculus might be led to suggest voting on a given issue, even though it gives rise to the pertes in Figure 2. We could live with the pertes over a range of issues if the gains and losses offset each other, and we avoid the delays of trying to make Lindahl allocations on each issue.

If pertes over time are distributed assymetrically, however, (as in Figure 3), we need to make adjustments in each individual's account (his annual income statement in ways that do not penalize him from reporting accurately (if he, in fact, is led to reveal preference information on individual issues (in which case he loses the right to have his tax-price adjusted to reflect real benefits on the particular issue).

In any case, some individuals should be compensated through a system of collective insurance, providing for expost compensation of the pertes which do not effectively balance out when Polinsky's law of large numbers is applied. One way to do this is to collect an insurance premium on taxes actually incurred and to remit the proceeds (less administrative insurance expenses) plus any Clarke taxes collected to individuals who have received the largest uncompensated pertes during the accounting period. This in no way reduces the motivation of parties with assymetric preferences to reveal their demands accurately, but compensates those (such as individual 1 in Figure 2) who lose from participation in traditional voting when the gains and losses from voting on many issues do not offset each other.

To take a simple example, if the approach suggested here were applied to all voting in the United States, we might collect an annual premium (2%) on each family's tax bill ($10,000) and remit that plus any Clarke taxes collected to those for whom the taxes departed from the Lindahl norm by the greatest amount. As will be demonstrated in Section 1.4., the approach could be made quite sophisticated and incorporate a number of efficiency features , including that of having a competitive insurance system estimate Clarke taxes for any entity using the procedure and receiving any actual Clarke taxes in return. Clarke taxes from the entity are thus zero in expectation and are relatively small in relation to the overall idemnification system.

The incentive calculus (which was suggested by Tidemann) fits naturally with the insurance principle by adjusting net benefits for some percentage of both (a.) the estimated pertes of those who lose on an issue and (b.) the Clarke taxes imposed on winners plus administrative and decision costs. A final adjustment, which is elaborated in the discussion of coalitions in Section 1.3 below, includes heavy (perhaps full) weight of negotiated tax-price assignments, which are added to net benefits. The Ruler would get some percentage of these adjusted net gains.

In a practical setting, this could be simply in the form of a salary which varies positively according to the "success indicators" (e.g. a ranking for "A" to "F" performance). If he receives an A, where there are no Clarke taxes or uncompensated pertes, he receives the equivalent of the net premiums in addition to regular salary (In a fully competitive system, he would bid for his salary). For C performance, when the premiums less expected Clarke taxes are equal to actual pertes and Clarke taxes, the Ruler receives his basic salary. When the Ruler receives a failing grade, he is deposed, and significant Clarke taxes in excess of amounts needed to compensate for pertes may have to be held over until a new accounting period and the arrival of a new Ruler.

Introducing these new feature of demand revelation,the remainer of this discussion is devoted to the key problems that have "haunted" DR over the last ten years in ways that furter elaborate on solving this basic difficulty of efficient Lindahl allocation and distribution. The DR literature of the past decade show that this problem has have some intriguing dimensions.

1.2 Rational Egoism and "Full" Information.

Margolis (1983) raised the most basic, and to me surprising, of issues regarding the DR consequences of interdependent utility functions. Tideman (1983) replied in a manner that, in my opinion, reinforced the need to ensure that the identification of benefit streams carries with it, to the extent practicable, the assignment of costs. Yet Margolis' work does motivate one to assign the problem of benefit aggregation to the Lindahl tax-setter and to rely on DR "vetoes" to efficiently control the process. When he fails, there is always the resort to DR as well as the insurance mechanism described herein (for those who have been most adversely impacted by the failure to accurately account for interdependent utilities).

The Ruler may also appeal to "sampling" procedures for eliciting individual preference information (Green and Laffont, 1977,1979). When individuals lack full information, we may also require less costly methods (Brubaker, 1986) or rely on methods of contingent valuation (Brookshire and Crocker, 1981) that may not reqiure a DR "vote", but which might reasonably approximate a Lindahl solution or which can guide cost-benefit analysts with information which faces individuals with the cost implications of their choices.

In terms of the Ruler's critical decision to make a cost assignment on an issue, we want him to undertake sufficent search (equating the marginal costs with the gains), yet not inefficiently delay decisions, by trying to eliminate the pertes (on each issue). When he is reasonably sure that the majority voting equilibrium is efficient (Figure 2) and stops the process there with no demand revealing appeals (such as might arise in the situation illustrated by Figure 3), the Ruler might simply register the estimated pertes in an accounting book. (These are the estimated losses of a minority, voting "no" as in the case of individual 1 in Figure 2). An efficient Ruler will go no further, where he is managing an agenda of issues, seeking results analogous to vote-trading and logrolling -- thus actively balancing the pertes so as to minimize them. As will be demonstrated later, the balancing process (by the law of large numbers) combined with insurance based idemnities to cope with any otherwise uncompensated pertes will, in effect, provide for a Lindahl eqiulibrium (for a range of issues), and allocative efficiency will be attained on each individual issue, taking account of the cost of information and given the opportunity for disadvantaged parties to reveal their individual demands on each individual issue, so as to arrive at the most efficient output.

Above, I have specified only one way (through sampling-based cost benefit analysis) of embodying the demand revealing Lindahl solution (called DRL) in a satisfactory decision process. Of course, there are two other basic ways -- representation and supplier provision (e.g. markets) through which the Ruler can promote informational efficiency (Clarke, 1980. The Ruler of a community may be able to appoint Rulers (for specific enterprises, say, for fire protection) so as to achieve reasonable production and information efficiency, but where DRL pressures for change might gradually grow unless he can arrange a satisfactory distribution for bundles of services over time. Like potential exit (and competition), the DRL exerts continual pressures to search for and maintain the Lindahl equilirium. Informationally efficient institutions will evolve, which will not require costly resort to requiring many individuals evaluate and report their preferences, even though uncompensated pertes would provide strong motivations to build coalitions that could depose the Ruler.

As will be further demonstrated in Section 1. 4.,the insurance approach serves a number of important efficiency purposes, in addition to that of providing for ex post compensation. Avoiding excessive investment in information gathering to arrive at Lindahl tax-prices is only one of these.

An informationally efficient approach also introduces the potential for individual strategy and coalitions, recognizing that the Ruler seeks to minimize all the relevant costs -- of allocative inefficiency, decision-making costs, Clarke taxes and insurance premiums. The following section treats the problem of strategy and informationally efficient solutions.

1.3. Individual Strategy and Coalition Incentives.

Despite strong-incentive compatibility (dominance), many (Riker, 1979) question the real superiority of DR because of its susceptibility to coalitions, and despite their fragility (Tideman and Tullock, 1980) Putting DR in the proper constitution, perhaps that of an enterprise, and sufficiently motivating the Ruler to seek the Lindahl solution in the manner described above, may also alleviate some of these concerns.

The previous DR literature (Clarke, 1980) has stressed that there are no coalitions with a perfect DRL solution. With less than full information, the Ruler also has strong motivations to provide transfers when potential coalitions exist, although efficiency considerations may preclude this. However, one questions whether in these circumstances, the individual would go to the trouble of forming a coalition, unless we are in a redistributive game, in which case there are powerful motivations for disadvantaged parties to directly reveal their losses.

Consider an example of incentives to coalition formation as in Figure 3, where individuals 1 and 2 have a coalition incentive under both DR and voting. The Ruler, of course, is motivated to move toward the Lindahl solution in Figure 1.

But we can't assume Nirvana, in a world of less than perfect information. In Figure 2, the DRL improves on the traditional voting process, moving the solution somewhere between A'B' and C', depending on how informationally efficient it is to approximate Lindahl. As in Clarke (1980), we can also conjecture that, if it is inefficient to make further compensating side arrangements to approximate Lindahl, coalitions would become bounded by these informational constraints.

As will be demonstrated in the next section, the specification of the Ruler's performance indicator is also critical in also determining how much search is required to propse a decision (or decision rule) and what costs to allocate to individual parties. He can propose a decision based on appropriate cost-benefit analysis as a first step towards allocative efficiency, but a party may request a delay. The next step might be a vote (majority or qualified majority) with the result that the party or other parties might then request a DR vote (further delay) in which case he is forced towards cost assignments (from Figure 3 to Figure 1). Note that in the above process, the ability of parties to recover ex post compensation would affect their desire to employ strategies for obtaining favorable cost assignments, when they can obtain the Lindahl approximation at the end of the process.

In the Figure 3 situtation, the Ruler will also further search for alternatives that may promote some allocatively efficient solution, discussed above, between A' B' and C' where a demand revealing vote could conceivably lead to an inefficient outcome from coalitions.

The liklihood that the process that I have described would ever arise at the stage where informationally efficient solutions to "positive sum" games would result in coalitions and Clarke taxes is, however, rather small But we gradually move into the political arena anytime we move from Figure 2 to Figure 3 with the creation of "potential" Clarke taxes. I suggest that my Ruler is strongly motivated to the extent it is informationaly efficient to invest in resourses to define Lindahl tax-prices so as to totally avoid coalitions, but what prevents redistributive zero-sum games (perhaps in positive sum clothes) from getting on the agenda and how do we deal with them when they are presented by members of the organization?

I would say that in the first instance, it is the Ruler's responsibility to attach apppropriate costs (individually assigned) to the benefit stream (up to the point where this is efficient), although information efficiency may also drive him to the selection of traditional voting rules, having their own coalition incentives (as in Figure 2) The Ruler is also driven by his performance criterion to exclude issues that do not contribute significantly to benefits (organizational gain) or where more search is needed to reduce possible pertes that can make net benefits barely positive or where Clarke taxes may exceed these gains. My paper, the Demand Revealing Governance of Enterprise (Clarke, 1987) illustrates this process of issue management in practical settings, including those momentus issues that involve a lot of potential redistribution and where there is a positive advantage in the potentially high Clarke taxes. These case also illustrate case where it is easy to measure the pertes and where the Ruler can more easily get to the Lindahl solution on the individual issue, and where the potential Clarke taxes provide a powerful motivational force.

Finally, coalitions are constrained in an environment of less than full information by the alternative of negotiating tax-price assignments. This also includes thae possibility of arranging insurance regarding uncertain events that can reduce or eliminate participant disagreement.5/

Thus except in cases where strong redistributional decisions are presented for resolution (and the Ruler either in turn excludes the issue or permits a DR vote to dispose of it), coalitions would seem to be bounded by the information costs constraints on the Lindahal solution. Potential costs of coalitions thus might be regarded as one of the costs of making decisions.

1.4. The Problems of Budget Balance and Clarke Tax "Wastes"

These costs also constrain the "wastes" of the DR process and obviate the need for a costly budget balancing process (e.g. the Groves-Ledyard process) that would tend to give back the Clarke taxes to the median voter (Clarke 1979, 1980) and reintroduce incentives to individual strategy. In practical settings, however, the potential for significant waste to arise might argue for insuring the performance of the Ruler by having an insurance company pay-in their expected value and receive any actual Clarke taxes. What is more important in these settings, however, is the potential of the Clarke taxes to motivate additional search and information gathering designed to achieve Lindahl equilibrium.

In these settings, the real issue boils down to when an individual is motivated to impose delay costs by a DR threat and the need to have him suffer these delay costs (or some portion of them) if the outcome is not changed. If decisions on an issue move from the conditions of Figure 2 towards Figure 3, DR exerts a set of pressures on the Ruler to resolve delays by finding alternative solutions (based perhaps on more search) that will satisfy disadvantaged parties, in the preponderance of cases, avoiding the resort to "full dress" DR polling. This often means, however, also reallocating cost shares on the issue and not leaving this to the ex post reallocation process, a problem that can be dealt with through the insurance mechanism described earlier and elaborated below.

In presenting this new concept of insurance against the adverse efficiency and distributional consequences of Ruler discretion, I will start with the simpler task of showing how we cope with the problem of the perceived "waste" of the Clarke taxes. In this formulation, these are not "wastes" at all, but are subtracted from the idemnities paid by an insurance company to the enterprise or collective for the "uncompensated" pertes and other insurable events (Ruler malfeasance, excessive delays, etc.) in any accounting period. The estimated value of the Clarke taxes can be effectively "paid-in" to the enterprise when the insurance premiums are established at the beginning of the accounting period. The insurer basically says that in return for $X premium, which could be competitively determined, the company will provide idemnities to those individuals, who are most harmed as a result of uncompensted pertes.

To take a concrete example, consider the following example (Table 1) drawn from Clarke (1980). We might have an estimated 100 issues to be resolved during the decision period where the third individual in Table 1 (same as individual 1 in Figure 2) end up with uncompensated pertes amounting to $1. On each issue, the perte at the lower left hand section of Figure 2 amounts to $1/24 and we assume that the pertes balance out in three-fourths of the cases, leaving about 25 events where they do not. Thus the insurance premium (not taking into account estimated Clarke taxes) would be approximately $1.

If voters choices were confined to discrete alternatives, the $1/72 Clarke taxes in Figure 2 could not arise, as neither individual's vote for B' vs. A' (preferred by individual 1) would individually change the outcome. However, assume that on about one-twelvth of the issues, the circumstances in Figure 3 would arise, and Clarke taxes amounting to .03125 were to be generated by individual 3 in order to arrive at outcome B. For the eight events, this generates $.25 in estimated Clarke taxes. With this estimate of likely events, the actuarially correct premium, setting aside administrative expenses,would be $1 - .25 = $.75 or 1.50 percent of total project costs over the accounting period. If events worked out as estimated, $.75 would then be paid to the individuals that on balance suffered uncompensated pertes 25% of the time. These do not include, in the sense of measuring compensation, the individuals that generated the Clarke taxes, so the process of giving back the taxes does not reintroduce individual strategy.

The insurance approach serves also an important efficiency function in a fully operating DR economy (or community) where the community Ruler appoints enterprise rulers and consumer agents (Clarke 1977) gradually evolve power (through proxies (Tullock, 1977) to represent their constituents in order both to get appropriate tax-prices on individual services and to recover any uncompensated pertes. The activities or threat of large scale intervention against enterprise rulers would lead to higher experience ratings (and sanctions against the rulers) well before pressures were sufficient to get DR votes for the deposition of the Community Ruler. (The fact that an enterprise receives a failing grade and may be uninsurable might be automatic grounds for replacing the enterprise ruler).

This is an example of how we generate efficient pressures on enterprise performance from without (through insurance) but this also carries into the enterprise, Within the firm, there is a real incentive for the Ruler to perform his function based on appropriate success indicators internal to the firm, where the delays and pertes are also related to organizational success (benefits), negative votes/Clarke taxes and decision/administrative costs.

To the extent that any Clarke taxes are actually generated, they are essentially lost in the insurance premiums and idemnities, because they are so small (even in small number interactions) relative to the probable size of uncompensated pertes. In practical settings, such an insurance system would also seem quite easy to establish, with contacts written not only to compensate for the pertes but to also constrain or insure against certain abuses of Ruler discretion.

For example, an enterprise might wish to receive a bond against such events as (a.) collusion and bribery between the Ruler and individual participants who receive collusively favorable tax assignments and (b.) any coalitions that might yield a perte to other parties and which may be later detected.

There is, of course the issue of "arms-length" relationships between the "insurer" and the Ruler as well as measurement problems which we will not pursue here. For the kinds of events that have been illustrated, however, measurement would seem fairly easy, in that we only have to look at experience with voting on issues, to determine how often certain individuals were in the minority, the size of the projects involved, ruler measurements of the likely distribution of benefits, even though we do not have access to actual preference information on these particular issues.

In sum, we would appear to have a perfectly satisfactory way of solving the budget balance problem and can turn to constitutional and distributional problems.

1.5 Constitutional and Distributional Problems.

This moves us into the question, not only of arms length relations with the insurer but of basically how the Ruler is appointed, regulated and deposed. Basically, too, what precisely is he? (As an almost certified public accountant - I received only a 74 on the Virginia CPA exam in 1964, but I still prefer that the Ruler, appropriately trained in public choice legal procedure come from the accounting ranks. However, the Ruler might be someone else, as for example a conseillier financier. After writing a recent article on the Ruler (Clarke 1987), I actually found him in a french management book in a Moroccan bookstore. The book deals with the organization of a federated group of enterprises and with the "choix d'une forme juridique" when one has a problem of collective decisions which require a "coordinateur". Several federated groups, in fact, empower such a person to determine when a decision that would normally be made unanimously, would be made by a majority, and vice versa. (CNCF, 1976, p.52) 5/

If there is disagreement, and with respect also to the sanctions that can be exercised against the Ruler (before going to the insurance company), the matter could also be referred to a Comitê des Sages (made up under CNCF procedure, of several heads of enterprise in the federation).

Beyond the issue of appointing, rather than electing my Ruler, there are a range of constitutional issues that might be delegated to him regarding (a.) choice of decision rules (e.g.majority voting vs. qualified majority, and the problem of when society moves from ordinal toward cardinal-utility based decision procedures (Borda, point voting, and DR). It is argued here that if

we are properly insuring against certain (hopefully improbable events), we can safely delegate decisions on what rules to use in which circumstances to the Ruler (see also Clarke, 1987) DR itself provides safeguards against the use of alternative decision procedures for rent-seeking as Tullock (1986) and Clarke have also stressed and the insurance approach herein can appropriately insure against DR abuses.

A final issue is ruler deposition and the problem of new appointments. I would currently suggest that we use the DR process to depose a Ruler and to appoint a new one by a rule that comes close to unanimity. As stated earlier (introduction), it is also quite likely that in a well functioning system with a large number of voters, DR would rarely be seen except in the case of a deposition issue. This would occur probably after sanctions and low grades on performance , and after the pertes of a rather substantial number of voters have led them to organize representation so that he is overthrown, well before sufficient numbers would otherwise be led to exit because of his ineffficient behavior.

As in Clarke (1980, page 100), I would tend to again set aside distributional (or equality) issues and the question of a social welfare function (or a bliss point on it). Nevertheless, Tullock's review article stimulated me not only to try to prove that we can get to a reasonable Lindahl solution but also to adequately comment on his review of Arrow and Reynaud. This has led me to deal with a narrower distributional problem of how we get from here to there vis a vis democracy, and to advance some "practical suggestions" for the effective blending of DRL, not only from the standpoint of efficiency but also with respect to distribution.

2. Tullock's "Arrow Reconsidered"

Before dealing with transitional problems, Tullock's review should encourage those dissatisfied with unregulated democracy to reconsider the Lindahl solution, endowing a neutral observer with a good deal of discretion and motivating him to move there while properly regulating democratic social choice processes.

In practical settings, I suggest starting with those which Arrow and Reynaud addressed in their recent book (a large enterprise or government bureau) where the issue of the feasibility or admissibility of cardinal utility comparisons should be addressed from a DR or DRL perspective. On this question, practical considerations (and the DR literature over the past decade) indicate to me that the full information assumption, and there resort to individual demand revealing, is not very helpful, except in special cases where parties are significantly disadvantaged by some rent-seeking redistribution (see Clarke, 1987,for examples; also Tullock, 1986). In these cases , the Clarke tax rears its ugly head to prevent rent seeking and should be viewed as a very useful safeguard in the corporate or community charter, assuming that the priviledge of invoking the process is not abused (because disadvantaged parties would pay the process costs when they do not change the outcome).

For non-redistributive problems, I have tried to show how the DRL process will, in fact, drive the overall process of selecting decision rules in a world of less than full information. In these settings, efficiency requires that we make compromises with democracy in ways that better us all. Tullock's enthusiasm for the DR process leads him to suggest extending it more broadly (at least as a thought experiment ) comparing the small waste of DR against the wasted sales effort in Presidential campaigns. I resist such an application, though it is interesting, but in so doing I risk being put in the embarrassing position of suggesting what I would do in the alternative.

In my opinion, Tullock has knowingly raised the opera singer problem, leading one to suggest not that the second singer be substituted for the existing process of electoral competition, but rather as a social welfare function guiding community and enterprise governance. The former entities, as do the latter, would have as an important feature the exit option and would compete for citizens, who have more equal entitlements to land and natural resourses (Tideman, 1984)

In turn, a nation would be a federation of communities with an appointed ruler who can constrain the behavior of exectutives and legislatures (perhaps lodged somewhere in the judiciary and dealing also with important social accounting questions that arise among the communities).

Until this day, I would prefer to consider blending DR into the constitutions of individual enterprises and communities. In these settings the sense in which DR might solve or "mostly solve" the Arrow problem is more than theoretically intriguing. To the extent that enthusiasts for DR can build convincing practical applications that get people to adopt institutions resembling DRL, those which use it, in my opinion, can evade the Arrow problem. This does not mean that societies throughly infected with redistributive problems of democracy, however, can have an easy time making such a transition (see Clarke, 1980).

We might begin by opening our eyes and providing convincing comparisions of demand revealing and traditional institutions, and also demonstrating how new institutions and procedures can be effectively blended with the traditional ones, making necessary compromises along the way.

Needless to say, improvements in welfare can sometimes be elusive, paricularly when exit is diificult and some memebers of the existing collective can continue to coerce others. To return to our illustrations, the move towards Lindahl equilibrium (Figure 1) will likely disadvantage individual 3 (as in Figure 2) or individuals 1 and 2 (as in Figure 3) relative to what that can obtain under existing procedures or in a majority rule coalition.

In turn, one should not discount the range of practical circumstances in which such a coalition believes it has a valuable "property right" for which the difficulty of explicitly arranging a compensatory "buy-out" might be quite elusive. In the real world, we are faced then not only with the inescapable difficulties of moving towards Lindahl equilibrium but in arranging compensation for the move from existing institutions where the system of democratic entitlements may be quite large. The difficulty of doing this, however, should not serve as a criticism of demand revealing, when the compromises permit, for example, greater opportunity for inefficient coalition behavior than would be feasible under an informationally efficient Lindahl equilibrium. (See section 1.3.)

More fundamentally, the difficulty of convincing people that they should permit circumstances in which those endowed with unequal wealth to express preference intensities embodying this wealth is perhaps one of the most important, and related impediments to applying demand revealing institutions in real life. It is also the chief difficulty that inhibits the resort to cardinal utility comparisons when we talk about designing institutions that otherwise deal efficiently with the measurement difficulties (see introduction).

While the procedures introduced here illustate how we might solve the problem of making efficient cardinal utility comparisons, the ethical appeal of these procedures may be limited to a narrower range of circumstances, absent efforts to cope more squarely with problems of equality. Thus there may be normative as well as practical objections to seriously considering the replacement of real world electoral processes with demand revealing institutions.

While I fully share Tullock's enthusiasm for demand revealing, and concur with the usefulness of comparing the decision costs of existing processes (including sales effort) relative to the decision and information costs embodied in procedures that would incorporate wealth and preference intensity to a larger extent, I am skeptical that people would readily give up their preference for symbolic equality in the larger electoral setting. They might be induced to do so, however, in settings of individual enterprises and communities where the fruits of blending the non-traditional procedures and the traditional ones could be quite large, and where symbolic equality would be largely preserved.

The central assertion of this review is that, once agreement on the validity of Tideman and Tullock's second major assertion (Lindahl equilibrium) is established, we can make further headway in understanding the extent to which demand revealing "mostly solves" the Arrow problem. But once we have established the validity of a satisfactory means of obtaining Lindahl equilibrium, having carefully opened to door to cardinal utility, we still have the mammoth task of designing real world institutions that will permit such comparisons to be made by a neutral observer or by individuals themselves in an ethically satisfactory manner.

The demand revealing Lindahl solution presented here, combined with (a) a properly specified objective function and set of incentives guiding the Ruler and (b.) an appropriately designed insurance arrangement to guard against abuse by the neutral observer could be an attractive option, at least in a reasonably wide variety of circumstances. I suggest further research on both normative and practical aspects of this question.

Discussion: If in Figure 2, the average efficient output were 1/2 over 100 trials , the average perte or Buchanan/Tullock external cost (lower left hand shaded area) would be 1/24 or approximately $1 over 100 trials. If one tvelfth of these trials resulted in Clarke taxes under the conditions of Figure 3 (see diagonally shaded area, the average Clarke tax would be $.03125. With this expected distribution of events, the insurance premium would be $1. - .25 =.75. Idemnities would, in turn, reflect administrative costs and actual Clarke taxes.

Table 1

1/ In this review, Tullock basically reasserts the premise that the existance of individual strategy-free decision processes (which mostly solve the Arrow problem) should enable us to more squarely confront the implications of the Arrow problem vis a vis democracy

2 / This is my interpretation, of course, of the basic thrust of Tullock's review. Demand revealing oprens up the possibility of making cardinal utility comparisons in a individually strategy-free environment. Nevertheless, the problems of less than full information and possible coalition strategies, as well as discretionary authority accorded a neutral observer must be confronted, before ethically satisfying demand revealing systems will be accepted. See also Mueller (1979) on cardinal utility and the Arrow problem.

3/ These are the criteria, including that of equlity, against which demand revealing systems mightr be compared with other voting systems (see Tideman, 1977).

4/ "Suppose we decide that a dollar unintentionally redistributed from one person to another imposes a cost of 20 cents . Then the true net benefit of implementing a new proposal is its net benefit as revealed by DR minus 20% of the Clarke taxes minus 20% of the vote of the status quo minus something for the administrative cost of the proposal. This is the measure of the performance of the Ruler: it would be reasonable to take some fraction of this as his compensation.

"When a proposal is not approved, the cost of having considered it is 20% of the Clarke taxes plus the administrative cost. This assmemtry results form postulating an entitlement to the status quo." T. N. Tideman, private communication, April 29, 1987.

5/ This idea is pursued at some length in a companion paper. It further illustrates my repeated assertion that if the participants in a coalition have information not possessed by the Ruler, they can be induced to reveal it through a set of negotiated tax-price assignments. In Figure 3, for example, individuals 1 and 2 "could" organize a coalition, but before mutuaaly revealing an excessive demand for an outcome near A'B', why not negotiate through the Ruler for adjustments? My incentive calculus for the Ruler also heavily weights these transfers so as to encourage the Ruler to seek out such arrangements.

In a world of uncertainty about future states of the world, the Ruler may also arrange insurance (in lieu of transfers) that will reduce disagreement over decisions and increase collective expected values over uncertain events. This is explained in more depth in relation to the insurance system described herein and a "pareto optimal process" first advanced by Thompson (1966).

Arrow, Kenneth (1951) Social Choice and Individual Values. New York: John Wiley

Arrow, Kenneth and Reynaud, Herve (1986), Social Choice and Multidimensional

Decision Making: Cambridge: the MIT Press

Brubaker, Earl R. (1986) Efficient Allocation and Unanimous Consent with

Incomplete Demand Disclosures? Public Choice 48 (3): 217-227

Brookshire, David S. and Crocker, Thomas D. (1981), The Advantages of

Contingent Valuation Methods for Benefit-Cost Analysis,Public Choice: 36:235-252

Buchanan, James M. and Tullock, Gordon, The Calculus of Consent: Ann Arbor:

Michigan University Press

Chambre Nationale des Conseilliers Financiers (1976) Qu'est-ce qu'un

groupement federatif d'enterprises, Paris, Dunod

Clarke, E. (1971) Multipart Pricing of Public Goods Public Choice 11; 17-33

Clarke, E, (1980) Demand Revelation and the Provision of Public Goods

Cambridge, Ballinger Pub. Co.

Clarke, E, 1979) The Search for Budget Balance in a Theory of Demand

Revelation, in Public Choice in New Orleans, Center for Study of Public Choice, Virginia Polytechnic and State University, Blacksburg, Va.

Clarke, E. Some Aspects of the Demand Revealing Process, Public Choice

(special spring supplement 29(2): 37-49

Cox, Charles,(1981) review of Demand Revelation and the Provision of Public

Goods, JEL

Green, J. and Laffont, J. J. (1977) Imperfect Personal Information and the

Demand Revealing Process: A Sampling Approach Public Choice (special spring supplement 29:2 79-94

Green, J. and Laffont, J. J. Incentives in Public Decision Making: Amsterdam

North Holland

Groves, T. and Ledyard, J. (1977) Optimal Allocation of Public Goods; A

Solution to the "Free Rider" Problem, Econometrica 45: 783-809

Hannequart, A. and Greefe, X. (1985) Economie des Interventions Sociales,

Paris: Economica

Margolis, Howard (1982) Selfishness, Altruism and Rationality. Cambridge U. Press.

Mueller, Dennis M.,(1979) Public Choice: Cambridge: Basic Books

Musgrave, Richard A. "Public Ggoods" in Paul Samuelson and Modern Economic

Theory, E Cary Brown and Robert M. Solow, ed., New York: McGraw Hill 139-156

Nitzan, S. (1985) the Vulnerability of Point-Voting Schemes to Preference

Variation and Strategic Manipulation Public Choice 47(2): 349-370

Polinsky, A. M. (1970) Probabilistic Compensation Criteria QJE 86: 407-425

Riker, W. (1979) Is the New and Superior Process Really Superior, JPE 87

(4): 875-890

Thompson, E. A. (1966) A Pareto Optimal Group Decision Process in Tullock, G. Papers in Nonmarket Decision Making.

Tideman, T. N. (1977), introduction Public Choice (special spring supplement)

Tideman, T. N. (1983) Some Economic Conceptions of Value

Tideman, T. N. 1984) Toward restructuring of normative economics, American

Behavioral Scientist 28(2): 233-247

Tideman, T. N. and Tullock, Gordon (1977) A New and Superior Process for

Mmaking Social Choices JPE: 84: 1145-1160

Tideman, T. N. and Tullock, Gordon (1981) Coalitions under Demand Revealing

Public Choice: 36: 323-328

Tullock, Gordon (1977) Practical Problems and Practical Solutions; the Demand

Revealing Process as a Welfare Indicator, Public Choice (special spring supplement) 27-36, 51-64

Tullock, Gordon (1986) Demand revealing, Rent-Seeking and Redistribution

Tullock, Gordon (1987) Arrow Reconsidered: A Review Article (unpublished)

After sufficient search, the Ruler could endorse a decision to change the status quo. An internal Ruler-administered insurance arrangement would provide ex post compensation to those whose net benefits are negative in the long run. Thus, if those who prefer the status quo are proved right in looking to an unfavorable state-of-the-world for investment, the actual losses would be taken into account with possible gains on other issues to determine compensation. In principle, the Clarke taxes plus insurance premiums (adjusted to reflect expected Clarke taxes) would be used to finance this ex post compensation, although practical arrangements might be somewhat simpler.

The arrangement to provide ex post compensation may, of course, sufficiently change the pattern of expected adjusted net benefits, so that the project can be implemented. When the Ruler has insuffient information to change the cost-assignments so as to better achieve a Lindahl equilibrium, he might also assist the parties (who may have better information in the negotiation of modified cost assignments (see Clarke, 1980, pages 81-89). In concept, the opportunity for such transfers will safeguard against coalitions and thier value should also be included in the objective function and incentive calculus facing the Ruler.5/