Skip Navigation Planning & Markets
Subscribe Submission Requirements Editorial Board Archive Links Search Home


Nobel Prize pundits (Science, October 18, 1996) credit successful applications of William Vickrey's idea to areas such as the spectrum and Treasury debt auctions, as a contributing factor in the award of Vickrey's prize. Why not now, in Vickrey's honor, extend the idea to expenditure allocation affecting distributive Federal programs? Following Vickrey's idea, an incentive-compatible approach to grant-in-aid design for distributive programs was initially set forth in an unpublished paper (Clarke, 1981), entitled "Reforming the Grant-in-Aid Pork Barrel," and was briefly described with respect to "A Limited Fund Mechanism" for allocating Superfund by Tozzi and Clarke (1983). More recently, Brough, Clarke and Tideman (BCT, 1995) described a compensated incentive-compatible approach first elaborated by Tideman (1979) to airport slot management and noise regulation. This method, described in BCT (1995) effectively separates the allocation of a resource (i. e. airport slots) from distributional questions, i. e., whether airlines should be entitled to landing rights as opposed to communities or the public at large. The method is an adaptation of the second price auction (Vickrey, 1961) and Clarke's demand revealing method for public goods allocations (Clarke, 1971).

With respect to distributive grants-in-aid (see Stein and Bickers, 1995), the following section of this paper illustrates the use of the procedure for allocating annual obligations to the state revolving funds (SRFs) under the existing EPA Construction Grant Program. The method could also be adapted to the implementation of the new Safe Drinking Water Act and Hazardous Waste grant management (Superfund). See, for example, Brough (1995).

By comparing a more practical, administrative approach with an admittedly more visionary one, it is possible to better gauge why the latter may appear less acceptable to real political actors and the general political science in the foreseeable future (see Introduction). Winning the hearts and minds of the general political science may simply require proving that the ideas work in simple administrative settings before trying them in somewhat more controversial settings where the new voting rules might supplant existing voting rules. The Nobel Committee noted that the Vickrey auction anticipated (by a decade) the theoretical development of the pivotal mechanism as a means of ensuring truth telling in public project tenders. I am hopeful that, in actual experience and within a decade, the use of the principles of the former (in spectrum and Treasury note auctions) could lead to use of the latter (also in voting on public projects) in appropriately designed institutional settings.

Revisiting the Limited Fund Mechanism: A Practical Approach

There is currently a wellspring of activity in converting from traditional grants-in-aid to revolving funds. Holcombe (1992) examined the general topic of revolving fund finance, focusing on the organization and administration of State Revolving Funds (SRFs) for wastewater treatment. A compensated incentive compatible (CIC) approach towards administration of revolving funds in general can be illustrated as follows.

Take, for example, the existing construction grant program funded in FY 1996 at approximately $2 billion in annual capitalization grants. Under the CIC method, an administrator would establish an initial interest rate, which might be 7% (say 2% below the market rate, so as to reflect the "disutility" associated with administrative strings and Federal requirements). This rate, would be based on evaluations by a national cost-benefit advisory committee and, in future years, also reflect for any given jurisdiction the revealed willingness to pays (WTPs) of other jurisdictions.

Suppose that the average state (J1) had an initial entitlement of S = $100 million. As shown in the following figure, however, J1 would take more and others would take less than their initial shares, and the willingness and ability to flex funds makes the derived supply of funds to each individual jurisdiction rather inelastic. Here, the supply of funds to J1 would be determined by subtracting the amounts that all others would take at 7% from the fixed $2 billion supply of funds. Assume, given a high elasticity, that this would equate with J1's demand for funds at a rate of approximately 7.25%. The amount that J1 would pay is determined in two parts. For an extra $10 million (A - S) over the entitlement level, it pays 7% and other jurisdictions are compensated by a similar amount. For the remaining or next $10 million (B - A), it pays an average of 7 and 7.25%, which is shown by the shaded area in the figure. This is essentially a VCG tax, also here called a general progress payment or GPP, which motivates jurisdictions to provide an accurate expression of preferences, albeit creating a small surplus of about $17,000 in extra annual interest payments for J1 over what is needed to compensate other jurisdictions at the rate of 7%.

Figure 1

Figure 1: Interest Rate Determination for State Revolving Fund (SRF)

The procedure can be viewed as a potentially more effective way of reconciling needs as provided by a formula or determined by a central administrator and the needs as perceived by subnational entities. As Holcombe notes (1992, page 55), the states receive very disparate portions of their EPA determined needs through the formula. "Two states -- Wyoming and North Dakota -- will receive more in SRF capitalization grants from the federal government than their EPA-estimated needs through the year 2008, while the bottom four states -- Arizona, Florida, Massachusetts, and Washington -- will receive only 5% of their estimated needs through 2008 in capitalization grants." For comparison purposes, Holcombe (footnote 11) notes that the median state receives about 14% of estimated needs through the year 2008; the large percentages given to some states is a result of an agreement that each state will receive a minimum percentage of the total of the money regardless of other factors.

With respect to a detailed elaboration of this approach in a separate paper, it is fairly easy to see how the administered system might be superior to simple trading or flexing of funds by simple agreement between the jurisdictions so as to, for example, achieve more efficiency in equating the marginal product of investment in a particular activity (i. e. toxic waste cleanup). Except for removing incentives to strategic manipulation to alter the equilibrium price of funds or the interest rate in small number interactions, the device may not be all that superior relative to trading, but it does have growing advantages when we adjust the aggregate level of expenditure. See BCT (1995), particularly in comparing CIC with zero revenue auctions which have been advanced in the environmental permit trading area as a means of achieving more allocative efficiency while preserving distributional stability.

The Limited Fund Mechanism (LFM) was originally advanced in the context of programs where the interjurisdictional spillovers (clean water, hazardous waste sites) were arguably small. However, this ignores the problem of substate allocation, where districts will vie for funds while letting the rest of the State pay for them. Also distributional considerations will intervene in that other jurisdictions may prefer, for example, that there be more clean-up of sites where, say, children are residing nearby as opposed to brownfields development where state WTPs might be higher.

Consider some additional examples of potential application to now evolving new revolving fund programs. EPA is now in the process of writing rules for the formula allocation of a new Safe Water Drinking Act program enacted this year. One could envision a set of simple per capita entitlements, followed by needs or cost-benefit assessments in which jurisdictions with higher "needs" would borrow amounts above their per capita entitlements and others would be compensated for receiving less (at assigned rates of compensation). In this and similar programs, the usual discretionary accounts (5 to 10% of the amounts allocated subnationally) would be allocated to states that exceed their entitlement levels by indications of their WTPs -- thus in effect paying variable matching rates depending on their demand for discretionary funds. In this way, a state with high needs makes a move from a simple per capita distribution (S in Figure 1) to point A (determined by the benefits assessment, which can also take account of projects with spillovers) and then to B (reflecting actual WTPs), creating a minimum surplus (as measured by the shaded area in Figure 1). This approach also minimizes distributional struggles in the initial allocation, where small states usually end up with some "minimum" allocation unrelated to "needs" (see Holcombe, 1992), and we avoid asking them to explicitly loan back funds when program managers may come to believe that these are their owned funds (a sort of flypaper effect).

Federalist Resource Allocation: Agenda Setting and the Pivotal Mechanism

I now turn, in more detail, to a second example involving the establishment of State Infrastructure Banks (SIBs) for transportation infrastructure. Currently, States are allowed to use up to 10% of their formula funds to capitalize the SIBs. I will not dwell here on programmatic features of the SIB that could lend itself to more administrative flexibility in allocation via the LFM. Suffice to say, a few large states would have likely used it intensively. About 10 large states with slightly less than one-half of the population had in fact about two-thirds the originally proposed allocations. In these contexts, the policy analyst would also have concern about strategic influences on the interest rate. Also the supply of funds may be relatively inelastic. In this program, for example, it was anticipated that in the startup the large States might be allowed to exceed outlay restrictions (15% in first year, then 52%, up to 100% by fourth year) as long as the total outlays for all states did not exceed an aggregate outlay limit characteristic of Federal highway aid programs.

To move now from the administrative realm to a more "speculative" realm, let us add in a feature of prominent "Truth in Budgeting" legislation (in the transportation area). This is the proposed use of the 4.3 cents (about $5.2 billion in gas tax money) used for deficit reduction. Suppose states could use a prorata share of this amount, while also assuming the debt repayment obligation, as long as they attempt to meet a performance goal of "maximizing the likelihood of minimizing harm" through beneficiary-related taxes and charges and the use of the funds is approved by the Congress. This idea, and Congressional use of a preference-intensity revealing voting rule to make decisions on budget size and restrictions, will be elaborated in Section C.

In terms of the following example, I will deal only briefly with LFM issues. However, suppose in the following table, option A represented the net present value (NPV) of projects financed by $100 million (10% of New York's unified grant) and option B represented the NPV of projects displaced in New York under current programs. Under this scenario, the LFM would allocate an additional $15 million in NPV costs over and above the interest rate applied to interstate allocations for comparable programs (Construction Grants, Drinking Water) described earlier. This might drive the rate to well over 8% as opposed to 7% in the previous illustrative construction grant project discussion (see also shaded area in Figure 1), and the rate may also be higher if the SIB programs are not subject to project restrictions (e. g. grant assurances) characteristic of Federal grant and loan programs.

Note here that the interest rate escalation may be such that we consider allowing states to turn to the $5 billion deficit reduction account, through procedures elaborated in the following section. Here we assume that the supply of funds is sufficient to finance most worthwhile projects with positive NPVs at reasonable social discount rates.

The examples presented below describe how we make decisions on improved budgetary flexibility while also describing the pivotal mechanism in a couple of discrete case contexts. Suppose our voters were Metropolitan Planning Organizations (MPO's) or perhaps New York's 31 Congressional districts. The two alternatives below continue to be competitive projects. Option B is the project initially favored by the National committee, which uses pivotal voting among the national sample of citizens in project selection (see Bailey, 1996a). However, a competitive project is advanced by the Executive officers in the form of option B and adjustments are made in the internal cost allocations within the state by the Electoral Commission to ensure that there is no redistributive harm (i .e there are no negative redistributive impacts by jurisdiction when the projects are considered singly). The most impacted jurisdictions are then polled, again using pivotal voting to determine which project to select. If the benefit impacts were fairly diffuse, none of the Congressional districts might have (singly) a pivotal impact. However, suppose that New York (Urban) in the example below changed the outcome with the result that it paid a $5 million "Clarke tax" or GPP penalty (see Table 1). That is when this jurisdiction (New York City or the New York MPO) voting as a whole has its preferences taken into account, its vote is pivotal (i. e. when its preferences are excluded, option B would have been accepted by a vote of $15 million to $10 miliion with a difference of $5 million). Note also that the executive officers representing Other Regional and Other Regions could also be pivotal, except in this particular case when their preferences are excluded, the result would have been a tie, with no GPP penalty.

These other actors, of course, may have preferences for a particular configuration of services. Drawing from an example for agenda setting in a related paper, neighboring Vermont proposes a modification to City Hall's plan (a further development of an intermodal project involving New York's airport system which includes improved surface transportation access). In this context, a private party in neighboring Vermont (see Table 2) proposes a modification (option C) to City Hall's plan (option A) which has net benefits (also including option B).

Table 1: Agenda Setting (2 options)

Options (Millions $)
New York (Urban) 10. 0.
New York (nonurban) 0. 15.
Other Regional 5. 0.
Other Regions 5. 0.

20. 15.

Table 2: Agenda Setting (3 options)

Options (Millions $)
New York (Urban) 10.0 0. 10.1
Other Regional 9.8 15.0 9.6
Vermont 2.0 0. 1.0

20.0 15.0 20.7

In this case, Vermont is able to organize a complementary set of expenditures (using say a portion of 10% of its unified grant or $6.7 of $67 million), including those of New York's $100 million, to effect a $.7 million addition to total net benefits. With the selection of option C rather than the next highest option A, there is a slight improvement to New York, a $.2 million "leakage" (disbenefit) to other regions, and a $.8 million improvement for Vermont.

To effect this selection, Vermont would make a GPP payment of $.1 million because in the absence of its expression of net benefit, option A would have been chosen by a vote of $19.8 million to $19.7 million, or $.1million.

These pivotal payments are, of course, small change and could be considered as small adjustments in the interest rate on funds charged to the relevant beneficiary jurisdictions and their private partners. For example, if a base interest rate of 7% yielded some $75 million in present value over 20 years, over and above the $107 million in New York and Vermont capital investment in Option C, pivotal payments of a few million would be a small element in their decision calculations.

Also of importance is the incentive structure driving the agenda setters -- the National committee and the potentially cooperative (or competitive) executive officers as well as the Commission. The above example provides a convenient vehicle for introducing a specific incentive structure to properly motivate these actors. Following Bailey's constitution (1996a), a possible set of incentives would include payments that are a positive function of budget size (.1%), social gain over the status quo (3%) and a negative function (0.5%) of harms. For the advisory committee, this would be about $95,000 when we compare Option A over the status quo (B). Now if a private party is able to convince the committee to add in something like Option C, the incentives might be some 10 times larger. In addition to the small positive reward for budget size (0.1%), the incentive would be 30% of the net social gains less 5% of the harms. This would be shared between the private party and the committee.

The private party could also shop in regional and local legislative fora, putting together options that would compete formally with those advanced by the national committee. There would also be strong incentives to use funds earmarked for particular projects in areas where they may not be used most efficiently. For example, we have a lot of smaller states (like Vermont) as well as in the West and Hawaii that get a lot more in transportation funds than they pay in Federal gas taxes. The GPP gives them the choice of (1.) saving the funds if the opportunity costs are high enough (and reducing their own gas or other taxes with the interest proceeds), (2.) spending the funds for improvements that will improve the use of our cultural heritage while advancing conservation goals or (3.) more traditional projects (airport improvements and highways). To the extent that projects are financed through beneficiary taxes (including land rents), then the local taxpayers are able to invest the funds in projects elsewhere at rates approximating the social cost of capital.

National Resource Allocation: Combining the Pivotal Mechanism and Thompson Insurance

We now present a situation where the efficient rent seeking on the part of potential competitors (the national committee, the EOs and private parties) is very aggressive. Large projects with high rates of return are identified, along with reasonable cost sharing and beneficiary-tax plans that can be approved under the aegis of the Electoral Commission (and appropriate state authorities), so that the incentive-based rewards from selection of the more aggressive projects promise to be substantial.

Let me address first the issue of when we go from agenda setting with the pivotal mechanism to combined pivotal/Thompson representative referenda. The national committee, of course, uses the pivotal mechanism in its deliberations and to the extent this leads to the wisest decisions, no referenda will be needed. This will also be true the extent to which the potentially competitive EO's tailor regional/local solutions to achieving the highest net social benefit.

In addition, we have utilized, in the Federalist setting above, representative referenda guided by the agenda setting activities of the committee and the EO's. However, they operated under budgetary constraints and inflexibilties that might not be always appropriate. In effect, we might carry the agenda setting process to the point that most of the parties are in reasonably unanimous agreement about most resource allocational decisions. However, a remnant of controversy remains to be addressed in the national budget game. In addition, this forum is the appropriate place to make determinations on the question of overall budget size, as instanced by decisions to permit subnational governments to dip into the deficit reduction set aside through a particular decisionmaking procedure designed to implement "truth-telling" in budgeting.

To also address questions when a project requires referendum consideration, we might apply certain tests. In the above examples, we could have required that no project where the ratio of material gains less redistributive harms is less than 1.2 would be considered. This is the point where the incentive payments (i.e. 3% of net gains less 0.5% of harms turns approximately negative, at a more precise ratio of 1.16). This might be combined with a presumption against a pure pivotal payment of any significant size -- if pivotal (GPP) payments are generated, a Congressional referendum might be called instead. In the examples previously presented, we would guess the referendum test would be passed as the payments would be regarded as insignificant. In the examples presented in this section, the GPPs are quite large, even though the preferred alternative satisfactorily, but barely, passes the 1.2 test.

The examples illustrate a participatory approach to budgeting that grows out of a 1979 debate (Ferejohn, Forsythe, and Noll vs. Clarke) about the use of demand revealing in making budget allocations, e.g. for the Public Broadcasting Corporation's Station Program Cooperative (SPC). When I advanced the view that demand revealing was also indeed a potentially superior method of public budgetary decisionmaking (Tideman and Tullock, 1976), critics Ferejohn, et. al. (1979) Riker (1979) and others questioned its superiority by illustrating the susceptibility of the method to defeat by coalitions via zero sum games of traditional politics, including the traditional budgetary process.

Bailey (1996) shows that the pivotal mechanism working together with Thompson insurance is relatively immune from coalitions. In addition, the pivotal mechanism preserves the truth telling integrity of Thompson insurance.

Incentive-based Procedures

This section elaborates on the characteristics of the operation of the incentive-based budget procedure when there are major conflicts or controversies between 2 or more separately constituted bodies (the committee, the EOs, and private parties). The competitive process drives each body towards the most socially efficient solution, so there is likely to arise, at the margin, some small gain along with GPP payments, which is always the case if we were evaluating a continuum of budgetary choices. The conflicts also arise when the 2 or more committees are drawing up plans from two separate population samples (i. e. a regional advisory committee vs. a national one), or particularly when some specially chartered committee (essentially a private government) pursues plans at odds with the general population.

The following examples pertain to a wide variety of transportation issues (controversies over major mass transit or intercity rail investments in the New York region or the Northeast Corridor, building a new Interstate (I-31) or a NAFTA corridor through the heartland linking Canada and Mexico, or building a replacement bridge or tunnel for the Woodrow Wilson bridge in the Washington region). They are also being used, in a slightly different format, to illustrate national budgeting of air traffic control and airport investments.

In the current contexts, let me now introduce some ways of resolving the previously identified problems in relation to the actual budgetary parameters used in the earlier debate. I also posit fundamental differences between two competitive committees (national vs. regional) which allows me to explore further the properties of Bailey's approach to incentive compatible governance. In the following example, we have two alternative budgets (A and B) made up of different program packages -- see also tables 11-13 drawn from FFN and Clarke (1979) and reproduced here as Tables 3 and 4 -- that generate different patterns of net benefits for the users. The combinations based on gross costs and benefits presented in Table 4 boil down two most preferred packages, CDE and CDA respectively. Also, our nine regions have been cumulated into five where we have the result for the Middle Atlantic (Region 4), with 15% of the cost shares, presented in Table 3.

Table 3: Comparative Net Benefits for One Region

Options (Billions $)
Total budget level 3.000 2.850
Cost to region 0.450 0.413
Gross benefit 0.510 0.232
Less GPP -0.128
Net 0.382

Note: With respect to illustrative budget allocations, the $3 billion total budget level in the example outlined here conforms closely to the FFN and Clarke model of demand revealing and an adaptation (by FFN) of Groves-type procedures in the allocation of an approximately fixed collective common user budget. We use the same gross payoffs, cost shares and net payoffs (after costs) as FFN and Clarke, the only difference being that all payoffs are multiplied by 10 to the 8th x a $30 budget level = $3 billion.

Table 4: Costs and Values of Programs to Regions

Billions $
Program element
Cost of
Region 1
Region 2
Region 3
Region 4
Region 5
A 0.60 0.30 0.18 0.24 0.12 0.48
B 0.90 0.12 0.36 0.60 0.18 0.30
C 1.20 0.48 0.84 0.66 0.24 0.54
D 1.05 0.36 0.60 0.36 0.30 0.72
E 0.75 0.12 0.54 0.36 0.42 0.18
Share of costs (%)
10% 25% 30% 15% 20%


    CDA (Option B) = $2.85 billion
    CDE (Option A) = $3.00 billion

The model created here now suggests that in lieu of the GPP payments to achieve the preferred outcomes (those shown in Table 5 being a theoretical upper limit that is reduced towards zero in large number interactions involving as many as 435 representatives), we substitute Thompson insurance. In the aggregate, region 4 buys .5 times (.510 less .232b) = .278b in insurance, receiving back a refund measured by an almost certain positive Thompson insurance surplus less incentive pay (see later discussion).

This net benefits and incentive tax calculations (GPP payments) among the five regions are as follows. The figures in parentheses reflect the net differences between options A and B (see Table 5).

Table 5: Net Reported Benefits and Calculation of Incentive Tax (GPP)

Millions $
Option A
Option B
Delta = A - B


Note: *The GPP is calculated by the difference is net benefit to all others when the other option (B) is chosen in the absence of the region's expression of net benefit (i. e. region 4's GPP = 3338 - 3210 = 128 million). See also Clarke (1979, Table 13).

To illustrate further, we take Clarke's (1979) analysis of the critical choice between CDA and CDE in the FFN model (see Table 5). With our new numbers, this critical choice implies a difference of $3.0 vs. 2.85 billion in the annual budget, for a difference of $150 million. There is also another proposed reallocation of $600 million from expenditures on A to be spent on E. However, if we apply the incentive tax procedure to the gross aggregations shown above in making the efficient choice of option A over B, given the cost shares, the result is $282 million in penalty taxes or GP payments.

The Electoral Commission, of course, has the incentive to try to get as close as possible to the ideal efficient distribution of cost. With a good knowledge of benefits and costs, the Electoral Commission could allocate $282 million more to parties 2 and 4 (54.2% vs. 45% of the cost of having E vs. A) and 9.4% less to parties 3, 5 and 1. We get $150 million in social gains at no cost. The extent to which this result can be achieved will depend, of course, on available information (including how well the sample of the citizenry predicts benefits for the population as a whole), on the costs of further search (including the tailoring of programs by the executive officers to particular regional/local circumstances), as well as such factors/limitations as available tax and budget flexibility (i .e. reasonable uniformity in tax assessments at the subnational level).

The above example reflects a situation where there would be strong pressures for a Thompson insurance referendum, absent successful negotiations between the opposing parties (2,3, and 4 vs. 1 and 5). Absent such agreement, the parties would now ask the Electoral Commission for odds on either side winning. Let us take these as equiprobable. Also the preferences are expressed solely in terms of material gain with the result that A is selected with total compensation of $525 million paid to regions 1 and 5, while reducing their harm to .5 x $525 million = $263 million. There is also a Thompson insurance surplus of $75 million which would be used to make incentive payments and cover administrative costs with the remainder returned to the participants through the refund procedure described earlier.

Returning now to the incentive payments described briefly above in the context of the New York-Vermont examples, we have the national committee with the winning proposal (A) receiving approximately $29 million in incentive payments. Of this there is $2 million relating to budget size, part of which has been shared with the other committee for common elements (CD) in the proposed budgets. There is another $27 million (or $45 - 18 million measured by 30% of social gain less 5% of uncompensated harm). According to Bailey, there would also be smaller payments to the members of the Electoral Commission unrelated to budget size, thus inducing that body to make appropriate Lindhal tax-transfers, to the extent practicable, to maximize social gain while minimizing redistributive harm.

The idea that Congress or any legislative body would ever use preference intensity voting rules may seem farfetched (see also Mueller, 1996, Ch. 11, also with reference to the use of demand revealing). However, as shown in a related paper on aviation governance, it is easier to see a government-controlled corporation administering airports and air traffic control governed by procedures analogous to those contained herein. The corporation would be subject to strong agenda control and the use of preference intensity voting would carry right up to a representative referendum described in this section where some form of determining budget size and restrictions is obviously called for. At that point, we might have to turn to more conventional tools such as qualified majority rules in the traditional legislature to resolve contentious issues. Nevertheless, the relative properties, and advantages/disadvantages of the preference-intensity revealing rules should be a lively arena for research, also applied to particular institutional settings explored in this and related papers.

In this paper, I have applied the same principles of agenda control to the allocation of distributive programs by the Congress. Professor Rubin and other skeptics may be right about the potential reluctance of "real political actors" and the "general political science" to adopt such procedures in real political settings. However, this should not be viewed as resulting from technical or operational limitations as much as the obvious difficulties inherent in changing our approach to the exercise of power and influence in matters affecting public decisionmaking.

page 3


USC Seal

Main Page | Subscribe | Submission Requirements | Editorial Board | Archive | Links

ISSN 1548-6036

Copyright 1999-2000
University of Southern California
Los Angeles, California 90089-0626