Transaction costs and coalition stability under majority rule.

Johnson, Ronald N. ; Libecap, Gary D.

1. INTRODUCTION

Under a majority voting rule, representatives of specific political districts must bargain with representatives of other districts to build a coalition for enacting legislation that benefits their narrow constituencies. But the theoretical public choice literature warns of potential for unstable majorities. Minimum winning coalitions divide program benefits just among their members, creating incentives for those left out to entice defection by offering rewards to those who leave and form a different coalition. As noted in Mueller (1989) new coalitions emerge, undermining the old ones, leading to cyclical majorities, short-term programs, and highly skewed distributions of program benefits. (1) Despite these dire predictions, there is a general consensus in the literature [such as Tullock (1981)] that programs are more stable and allocations more equal than the theory suggests. (2)

In an effort to explain the discrepancy between prediction and observation, researchers have pointed to institutional rules and practices in Congress. For example, as stated in Shepsle and Weingast (1981a) and Weingast and Marshall (1988), the committee system provides a "structure-induced equilibrium" that limits the possible range of vote trading and thereby helps maintain coalitions. Further, Shepsle and Weingast (1981b), Miller and Oppenheimer (1982), and Collie (1988) argue that universalist sharing of program benefits enlarges the winning coalition and extends its political support. (3)

In this article we offer additional evidence of broad, stable sharing in many programs enacted by Congress by describing interstate distributions from the Federal Highway Trusts Fund (HTF). The allocation formula for the HTF was initiated in 1916, but despite wide divergence across the states in growth of various economic factors over the rest of the twentieth century (such as vehicle registration and population) that might have led to redirection of highway funds, there were comparatively limited HTF allocation adjustments. Analysis of state receipts from the HTF relative to tax payments into the fund reveals that some states collect much more than they contribute, whereas others pay in more than they receive. Even so, interstate ratios of HTF apportionments to payments have remained stable across the years, varying less than changes in highway use measures would suggest. Going beyond this specific program, we examine overall federal expenditure and tax shares among the states from 1975 to 1997 and show that there has been a similar continuity in the interstate distribution of federal funds and taxes. (4) As with the highway program, there is broad, stable sharing of federal expenditures across the states, with some receiving more than they contribute in taxes to the federal government.

To better understand this observed stability and use of relatively egalitarian sharing rules and to go beyond existing explanations, we emphasize the desire of politicians to minimize the high transaction costs of negotiating and enforcing political coalitions. Politicians have incentive to prevent unraveling of political agreements to avoid the costs of searching for new coalition partners, reaching agreement on the nature and distribution of program benefits and costs, and verifying compliance. These activities detract from a legislator's ability to address other voter concerns. Moreover, legislators seek to protect constituent benefits accruing from long-term programs that would be lost if coalitions unraveled. (5) Accordingly, we argue that politicians assemble greater than minimum-sized coalitions to build broad political support for their legislative programs, offering benefits to a larger constituency in exchange for additional votes. Considerable negotiation over the distribution of program benefits a nd costs may be required, so that once agreements are reached, politicians will be loath to consider a major reallocation that could undermine the coalition.

A group of politicians directly interested in specific legislation may not be large enough to achieve enactment within the legislature. Under those circumstances, coalition members must engage in additional logrolling agreements with politicians interested in different legislation to exchange votes and extend political support enough for passage. This means that political coalitions can be far ranging, involving politicians and constituencies interested in seemingly unrelated legislation. Recognizing these linkages suggests the additional hazards for legislators of attempting major ~x post adjustments in specific program allocations. Changes in the nature and distribution of benefits for one program due to a breakdown in the initiating political coalition could undermine the basis for previous logrolling agreements. Accordingly, politicians would have to not only reassemble a coalition for the narrow program but also engage in new logrolling agreements to sustain a revised program.

The article is organized as follows. The following section outlines the history of the HTF and demonstrates stability of programmatic shares. Section III examines stability within overall federal expenditures and taxes across the states, the SPEND/TAX ratios. Section IV discusses the explanations provided in the existing literature and then considers the importance of political transaction costs in coalition formation. In the concluding remarks section, we consider the implications program stability has for the performance of government and efficiency assessments.

II. THE FEDERAL HTF

The earliest comprehensive federal involvement in funding highways began with the Federal-Aid Road Act of 1916 (Pub. L. 64-355) and, five years later, the Federal Highway Act of 1921 (Pub. L. 67-212). To assemble political support for federal subsidies for road construction and maintenance throughout the country, funded by federal excise taxes on fuel, political bargains had to be struck. Early formula rules agreed to for distributing highway funds gave equal weight to state area, population, and length of postal (mainly rural) routes. The formula advantaged some states, so that other states paid more into the system than they received. Between 1916 through 1998, the year of the last major HTF legislation, there were periodic adjustments in the allocation formula, but parts of the basic framework remained. Indeed, the same formula was used for the first 40 years, between 1916 and 1956. (6) A modification for the new interstate highway system to give more weight to population and construction cost was added in 1956. (7) Other adjustments were made 26 years later in 1982 to create a separate Mass Transit Account, to ensure that states received no less than 85% of their contribution to the HTF and to include vehicle miles in the formula for distributing maintenance funds. (8) Some additional modifications in the allocation formula were made between 1983 and 1998, the year of the last major highway bill. (9) All in all, however, the basic distribution arrangement appears to have been maintained. It has provided a reliable flow of highway funds to constituents within each political jurisdiction long after the initiating coalition members passed from the scene.

To illustrate how this continuity was maintained, we examine two periods when internal pressures within the coalition threatened the basic allocation pattern: the inauguration of the interstate system and creation of a large highway trust fund in 1956 and the near completion of the interstate system in 1982 with a corresponding rise in political demands to redirect funds to mass transit. The legislative histories of these two HTF allocation adjustments reveal how modifications were made in a manner that minimized disruption to status quo distributions and thereby helped sustain the underlying political coalition for the highway program.

The new federal interstate highway system and a dramatic increase in annual federal expenditures from $25 million to $175 million under the Federal-Aid Highway Act of 1956 brought an effort in Congress by some legislators to give more weight to population and to add new variables to the distribution formula to address shifting highway "needs" across the states. This adjustment would have reduced allocations to 31 states and increased it for 17. (10) During congressional debate representatives of states such as California, New Jersey, and Indiana that contributed more than they received (see Table 1) argued that the old allocation system had to be dropped to reflect new conditions. Some, like Sen. Homer Capehart (IN), wanted emphasis on construction costs: "I can see only one proper basis upon which to arrive at a formula, and that is the number of miles and the number of lanes times the actual cost" (Congressional Record--Senate, 28 May, 1956, 9070). Sen. Thomas Kuchel (CA) added: "We need to determine the ba sis on which the Federal Government can pay the cost of the construction. I argue that it is on the basis of need alone" (Congressional Record--Senate, 28 May, 1956, 9077). But supporters of the existing formula countered, emphasizing the ambiguity of need measures and the risks of scrapping long-standing practices. Sen. Paul Douglas (IL) asked: "If the allocations for the initial 2 years are based upon the estimates of cost by the States, will not that furnish an inducement for individual States to make their estimates of cost as high as possible, because the higher the estimate, the greater the share of the initial apportionment they will get?" (Congressional Record-- Senate, 28 May, 1956, 9077). Sen. Albert Gore (TN) added, "I say that Congress cannot and must not start the distribution of the vast sums involved in the bill in such a haphazard manner.... No perfect formula can be devised. But the [existing] formula has the merit of being the legal formula for the distribution of the taxpayers' money. Moreo ver, it has the merit of having been tested and tried" (Congressional Record-- Senate, 28 May, 1956, 9076). Sen. Robert Kerr (OK, a state that also contributed more than it received) asked, "Why leave a stable foundation of operation for the precarious and insecure situation in which 31 States would find themselves if they abandoned it or permitted it to be taken away from them?" Kerr referred to the proposed change as "revolutionary" (Congressional Record-- Senate, 28 May, 1956, 9079, 9201). Finally, Sen. Dennis Chavez (NM) emphasized the progress in highway construction that had been made over the 40 years of the existing formula: "[T]he Senator from Oklahoma is certainly correct when he says it is a tried and proven formula. All the Senator from Connecticut has to do is to turn around and look at the red lines on the map [completed federal highways]. They illustrate what the formula has done" (Congressional Record-- Senate, 28 May, 1956, 9080). The Senate Committee on Public Works also emphasized the polit ical risks to the highway program of replacing the old formula:

[E]ach State's share would depend in large measure upon its ability to convince the Bureau of Public Roads of the accuracy of its estimates. Whatever may be said about the efficiency of distribution of Federal funds on the basis of need, the committee considers it contrary to the public interest to initiate a policy of distributing Federal funds on the basis of what each State claims it needs (U.S. Senate, Committee on Public Works, 1956, 3).

The ex post inclusion of "need" variables into the highway allocation formula not only could have brought political disputes over definition and measurement, as the Senate committee suggested, but it would have resulted in lower weight being assigned on the previous allocation variables and hence led to a redistribution of program funds. This reallocation was explicitly addressed in the Senate debate. It seems clear that both factors could have undermined the political coalition that created the federal highway program. The Senate defeated the proposed replacement of the old formula to one based on needs, 55 to 27, with 13 abstaining (Congressional Record--Senate, 29 May, 1956, 9203). In conference committee with the House, the old formula was kept but modified to raise the weight given to population from one-third to one-half with a provision for construction costs to be considered only for allocating funds used in completing the interstate highway system, beginning in 1960. Funds for federal primary and se condary highways continued to be allocated according to the 1916 formula (Congressional Record--Senate, 26 June, 1956, 10964).

Political pressure rose again for adjustment in the allocation formula in 1982 as the interstate highway system neared completion. This milestone was an important goal of the original political coalition, and after it was achieved, support for the highway program from representatives of states that contributed more than they received began to erode. Legislators from California, Texas, and Florida (all "donor" states, Table 1) called for an equity adjustment in the allocation formula. Rep. Bill Archer (TX) argued: "I recognize that there is a need for an Interstate System ... but certainly over a 20-year period, when the State of Texas has received less than three-quarters of the money that it has paid in, there is a gross inequity" (Congressional Record--House, 6 December, 1982, 28914). (11) But Rep. James Howard (NJ, a net donor state at least through 1973) countered:

[I]n developing this hill we did find that there is and has been throughout the history of the highway trust fund a discrepancy between the amount of money that certain States may contribute into the highway trust fund and the amount of money that they may receive. ... [T]here was a national need for a major interstate highway network across the country.... We cannot balance that out one to one. Otherwise we will have 50 separate highway situations in the country, good roads in one State and poor roads in another State.... We need a national highway system, we need good roads in every State, it we are going to have commerce and safety in our transportation. And so we have to consider having a good national highway system. And if it is going to cost some States, like my own State, to have my people in New Jersey have safe, decent, good, efficient roads when they travel to other States, that is the way I would like it (Congressional Record--House, 6 December, 1982, 28912, 28913, 28916).

Howard warned his colleagues against focusing too narrowly on how they were treated in the highway accounts when they were disproportionately benefited in other national programs. In the big picture, allocation adjustments in one dimension could require changes elsewhere that they might not want to consider:

If we are going to talk about giving back, getting back everything we give, we could talk to the State of Texas and ask them how much of the billions of dollars in the Federal Space program that goes into Texas would they like to share with New Jersey and some of the other States, or how many of our farm States would like to share money from the farm program (Congressional Record--House, 6 December, 1982, 28916, emphasis added).

The issue was resolved by increasing the gasoline tax by five cents per gallon to fund a Mass Transit Account (one cent of the increase) and to guarantee that no state would receive less than 85% of its contribution to the HTF. Because both changes were funded out of the new revenue from the tax increase, Rep. Howard could claim that "this is an amendment that helps several States and hurts no state" (Congressional Record--House, 6 December, 1982, 28913, emphasis added). No redistributions were required, and the highway coalition was maintained. Additional logrolling trades involving diverse issues were made through amendments to the 1982 highway bill that extended U.S. unemployment benefits, required states to allow much heavier trucks on federal highways, granted tax deductions for conventions on U.S. cruise ships, gave tax benefits for California utilities, and mandated funds from various other trust funds for reforestation, airport appropriations, boat safety, and fisheries development (Congressional Quar terly, Almanac, 97th Congress, 2nd Session, 1982, 315-22). These amendments illustrate the logrolling trades that were part of enacting major highway legislation.

Other transfers from the HTF were made by legislation passed in 1990 and 1993 to reduce the federal deficit. Importantly, these transfers involved across-the-board cuts and not in changes in the interstate allocation formula. In this way, status quo positions remained, as all states shared in the transfer according to the allocation formula. The Omnibus Budget Reallocation Act of 1990 (P.L. 101-508) temporarily raised the gasoline tax by five cents per gallon with half of the resulting revenues directed to the General Fund of the Treasury. Another fuel tax increase of 4.3 cents per gallon was enacted effective 1 October 1993, by the Omnibus Budget Reconciliation Act of 1993 (P.L. 103-66) with the entire amount of the increase directed to the U.S. Treasury. As political pressure to reduce the deficit waned, however, members of Congress lobbied for the return of diverted highway tax funds to the HTF. The Taxpayer Relief Act of 1997 (P.L. 105-34) redirected the 4.3-cent tax that had been going to the Treasury ba ck to the HTF effective 1 October 1997. This action underscores the importance politicians have assigned to maintaining a comparatively consistent stream of constituent benefits from the HTF.

Table 1 shows the ratio of the amount each state received from the HTF's Highway Account relative to its excise tax contributions to the fund for the periods 1956-73 and 1956-97. (12) The size of the numerator is determined by the allocation formula. Also included in the numerator are smaller amounts of discretionary funds within the HTF that congressional committees allocate to the states. (13) The denominator is based on a state's contribution from taxes on highway motor fuel, truck tires, sales of trucks and trailers, and heavy vehicle use. The table reveals the broad sharing in the benefits of the highway program that occurred over the years, including even remote states, like Alaska and Hawaii, that might not obviously be part of an interstate highway system. The table also shows that some states, such as Alaska (with a ratio of 6.69 for the entire period 1956-97), Hawaii, Montana, West Virginia, Rhode Island, and Vermont, experienced exceptionally high returns from the HTF, whereas other states, such as California, Oklahoma, North Carolina, Texas, and Michigan, contributed more than they received.

The basic distribution of highway funds was determined when the initial highway coalition was assembled. As the legislative histories presented indicate, once the coalition was established to create a national highway system politicians were reluctant to fundamentally change the interstate distribution of funds. As a result, the ratios of apportionments to taxes presented in Table 1 should be stable across time. An immediate problem that has to be resolved is the selection of the statistical technique to examine stability. Although the term stability implies stability over time, there are two dimensions to this problem. First, there is the issue of whether the means and variances of the ratios have changed over time. If the underlying distributions have changed, that would indicate that the allocation scheme has changed. Failure to find a significant change, however, does not imply stability of individual state shares, because winners can replace losers while the underlying distribution remains unaltered. Thu s the second issue revolves around the stability of states' shares over time.

Table 2 provides descriptive statistics for the HTF for each year in the sample, from 1974 to 1997. (14) The mean apportionment/tax ratio for each year is substantially above unity, reflecting the fact that many states enjoy an advantage in the allocation of funds. It is also readily apparent that the distribution of the ratios is highly skewed to the right in each year of the sample for the same reason. Tests for the equality of the means of the ratios in each year yields an F-statistic of 0.189, degrees of freedom (23, 1,176). The critical value at the 5% level is 1.76, thus we cannot reject equality for the ratio means. A Brown and Forsythe test for the equality of the variances between the series yields a statistic of 0.231, and this measure fails to reject equality of the variances. Other tests were performed, such as excluding Alaska and Hawaii from the sample and testing for a break in the sample with the introduction of the Mass Transit Account in 1982. (15) Taken collectively, however, these tests co nsistently indicate that the underlying HTF distributions reflected in the apportionment/tax ratios have remained stable over time. (16)

A problem with only examining the stability of the underlying distributions is that they may remain stable over time even if state shares are fluctuating widely from year to year. A simple yet direct test of stability is to examine whether the past is a good predictor of the future. We conducted a variety of tests employing that concept.

Consider first the premise that changes to the allocation formulas, although numerous, have amounted to largely fine-tuning. If so, the earlier formulas should be good predictors of more recent allocations. The original 1916 highway formula as codified in 1921 apportioned funds on the following basis:

One-third in the ratio which the area of each State bears to the total area of all the States; one-third in the ratio which the population of each State bears to the total population of all the States, as shown by the latest available Federal census; one-third in the ratio which the mileage of rural delivery routes and star routes in each State to the total mileage of rural delivery and star routes in all the states at the close of the next preceding fiscal year, as shown by certificate of the Post-master General. (Pub. L. 67-217)

To update the 1921 formula we used data on population and land area from the Statistical Abstract of the United States. Treating star routes as essentially the equivalent of today's rural interstate highways and adding that mileage to a state's own system of rural highways provides a comparable mileage measure. (17) Consider, for example, how well the 1921 allocation formula predicts the share of each state's 1997 allocations from the HTF. The regression results are

Apportionment Share, 1997

= 0.002 + 0.90 * Allocation Formula (1921), (0.70) (7.58)

with t-statistics in parentheses; number of observations = 50, R-squared = 0.54. Clearly, the 1921 allocation formula doesn't provide a perfect fit, but the coefficient on the allocation formula is highly significant and close to unity. Moreover, the regression includes Alaska and Hawaii and neither were states in 1921. Dropping those two states increases R-squared term to 0.72.

Alternately, if past shares predict future shares, that is prima facie evidence of stability. Let the dependent variable be the state's average apportionment/tax ratio over the period 1974-97, and the explanatory variable equal the state's average ratio over the period 1956-73, as shown in Table 1. The regression results are

Apportionment/Tax [Ratio.sub.i] (1974-97)

= 0.34 + 0.70 * Apportionment/Tax [Ratio.sub.i] (1956-73), (4.76) (18.11)

with t-statistics in parentheses; number of observations = 50, R-squared = 0.87.

The high R-squared term and high t-statistic for the coefficient on the explanatory variable indicate that the past is a good predictor, and stability is implied. The coefficient on the apportionment/tax ratio (1956-73) is less than unity, suggesting that some changes in the allocation scheme have occurred. The legislative histories of congressional debate over the distribution formula show that there was political pressure for gradual modification to direct funds to more populous states with growing highway demands. As a consequence, some states, such as Alaska, experienced a decline in their ratios. Additional testing revealed, however, that for the majority of states there was no statistically significant trend in their ratios.

Moreover, consistent with the results, the apportionments received by each of the states have not been very sensitive to changes in key economic variables, such as motor vehicle registrations, that reflect highway use or need. To see this, we regressed the log of annual apportionments by state from 1974 to 1997 against the log of motor vehicle registration, controlling for individual state effects: (18)

[LogApportionment.sub.ij]

= 0.02 [LogMotorVehicleRegis.sub.ij] (1.57)

with t-statistics in parentheses; number of observations = 1,200, R-squared = 0.90. Although total apportionments expanded almost 40% in real terms over the 24-year period, a state's apportionments from the HTF were not elastic with respect to motor vehicle registration. (19) The estimated elasticity for apportionments with respect to motor vehicle registrations is close to zero and is not statistically significant. Of course it would be difficult for large states to experience substantial shifts because they are constrained by the aggregate size of the fund, but that, too, speaks to the stability of allocations from federal programs. Once the highway program was agreed to, it is evident that only minor manipulation of HTF distributions took place.

III. STABILITY IN OVERALL FEDERAL EXPENDITURES AND TAXES

We have just examined the federal highway program, which has provided broad, stable shares of highway funds across the states since it was inaugurated in 1916. The funds have been disbursed through formulas that have not been drastically changed, even though the states have had divergent growth patterns in population and vehicle miles. Examination of aggregate federal expenditures and taxes across the states also reveals similar stability.

Consider Table 3, which shows the ratio of the amount of federal spending on all programs in each state divided by the amount of federal taxes collected in the state for the years 1975 and 1997. At first glance, there appears to be considerable variation in the SPEND/TAX ratios, and some states have experienced substantial shifts in their rankings over time. Nevertheless, the simple correlation coefficient between the ratios for 1975 and 1997 is 0.74. The correlation coefficient using state rankings is 0.79. These results imply considerable continuity in overall federal tax and expenditure patterns over time. (20)

Other evidence besides the SPEND/TAX ratios reveals legislative stability. The analysis of congressional roll call voting by Poole and Rosenthal (1997) reveals that despite the wide array of issues faced by legislators over the past 200 years, voting records are remarkably stable and predictable along one or two dimensions. Long-term variables, such as party and urban/rural constituencies, appear to limit a politician's ability to switch coalitions. (21) Similarly, Peacock and Wiseman (1961) and Higgs (1987) have noted that once government programs were enacted in response to crises, they endured long after the crises have passed.

IV. UNIVERSAL SHARING, TRANSACTION COSTS, AND PROGRAM STABILITY

There is convincing evidence of universalist sharing and stability in the ratios of receipts to tax payments across the states for the highway program and more broadly in aggregate federal allocations and taxes. These conditions are inconsistent with predictions of cycling majorities in legislative coalitions and corresponding fluctuation in the distribution of program benefits and costs. Awareness of the potential for intransitivity of social choice under majority rule dates, at least, to the work of Marquis de Condorcet in the late eighteenth century. He showed that an equilibrium may be nonexistent if a committee uses majority rules to choose from among a set of alternatives. The search for conditions that would yield a stable outcome has been one of the most intensely explored areas in public choice.

There are a number of factors that could encourage compliance with political agreements and thereby limit coalition defection and vote cycling. One is the value of a legislator's reputation for adherence. Presumably, if a legislator violated coalition agreements, he or she would not be viewed as trustworthy in negotiations and hence not included in political bargains. Exclusion could be costly to the politician's constituents because they would not share in program benefits if the legislation were written around them. A politician considering defection, then, would have to weigh the costs to his constituents of a damaged reputation with the benefits of joining a new coalition. If the new coalition can be formed at low cost and the redistribution is significant, then it is conceivable that the constituent gains from defection could outweigh the losses of membership in fewer future coalitions. Hence, the role of a politician's reputation in maintaining coalition stability is not obvious. A test of the empirical importance of reputation as a constraint on behavior is provided by examination of the votes of lame-duck politicians relative to those facing reelection. It would seem that maintaining a reputation would be less important for lame ducks and that they might behave opportunistically, voting for one-time gains for their constituents through defection or for programs that satisfied their individual ideological tastes. However, research on the behavior of last-period legislators does not reveal significant shifts in their voting patterns (shirking). (22) Accordingly, although reputation may play a role in maintaining coalitions, its theoretical and empirical contribution is unclear.

Legislative rules that protect committee proposals from alternation also could inhibit cycling. As described in Shepsle and Weingast (1981a), Gilligan and Krehbiel (1987), and Weingast and Marshall (1988), these rules grant committee chairs and members agenda-controlling powers that favor their preferences over those of the entire legislature. Because committee members tend to be more homogenous than the legislature as a whole, signaled-peaked preferences, coupled with the gate-keeping actions of committee chairs, could provide stability within the committee. There remains, however, the problem of enforcing agreements once legislation reaches the broader legislature. To secure passage in the entire chamber, committee members must engage in logrolling trades with other politicians. But as we have noted, logrolling coalitions are themselves subject to competitive unraveling. What, then, enforces those agreements?

Political reputations and committee structures appear to be only part of the solution to the potential problem of cycling. We emphasize legislative design for universalist sharing of program benefits. Broad programmatic sharing expands the size of the constituent group that has a stake in the legislation and reduces the incentive of politicians to defect from the coalition. If the transaction costs of reassembling coalitions were very low, then universalist sharing would not be important, at least for maintaining agreements. If cycling occurred, coalitions could be restructured quickly, dropping some politicians and their constituents while adding others. (23) Under these circumstances, cycling would be associated with narrowly shared but repeatedly shifting benefit allocations. By contrast, if the transactions costs of searching for new coalition partners, negotiating the distribution of program benefits and costs, and monitoring compliance are high and if constituents demand long-term stable benefits, then universalist sharing becomes important for stability. Furthermore, once broad distributions are agreed to, politicians will be reluctant to adjust them in an important way because of the threat of undermining the coalition agreement.

Avoiding cycling is not the motivation for universalism described in some of the key papers on the issue. Weingast (1979) and Shepsle and Weingast (1981b) argue that universal coalitions are preferred by legislators because of uncertainty regarding the makeup of the winning coalition. Because no politician can be sure of membership, a more inclusive group is desired. Miller and Oppenheimer (1982) assert that fairness norms lead to more general sharing than would result from formation of a minimum winning coalition. Glazer and McMillan (1992) make a more direct linkage between stability and universal sharing. They argue that narrow proposals appealing to a bare majority are especially vulnerable to amendments and hence are subject to costly renegotiation. Proposals that offer some benefits to a larger majority are less likely to be amended and are more stable and productive for constituents. Congleton and Tollison (1999) also note that uniform benefits to all voters increase stability. Broader coalitions raise the transaction costs of redistribution and reduce the incentive to change the status quo. They argue that inclusion of a transaction costs parameter in redistribution games can prevent cycling. In their framework, when the transaction costs parameter is sufficiently low, cycling may continue, resulting in a downward spiral and increased rent dissipation. (24)

These articles are instructive, but they do not develop the importance of transaction costs as an incentive for stability to the degree described here, nor do they place the discussion into the context of an actual legislative setting, such as the HTF. Transaction cost analysis entails more than simply introducing a per-unit charge for each coalition formation. As the HTF case reveals, transaction costs are complex and far reaching in their influence on politicians and political behavior. Indeed, as North (1990) emphasized, transaction costs and efforts to reduce them shape the institutions (rules of the game) we observe in actual practice.

The notion of endless cycling ignores the cost to politicians of repeatedly forming and reforming coalitions. It also neglects the opportunity costs of failed coalitions and the loss of related government programs that bring valuable benefits to constituents. If transaction costs are positive, and they most assuredly are, endless cycling would at some point have to bankrupt legislators and government programs. Politicians would have to spend their time contacting, negotiating, and monitoring agreements, with little or no time to serve constituents. Furthermore, the associated instability inherent in cycling would suggest that legislation involving long-term programs and allocations could never be enacted. With negative-sum redistribution games, welfare would be reduced as resources were continually used both by politicians in negotiation and renegotiation and by constituents in repeatedly adjusting to program instability. VanDoren's (1991) analysis of the complex political process underlying the development a nd implementation of U.S. energy policy between 1945 and 1976 illustrates the immense amount of work that legislators must do to forge and maintain programs desired by multiple constituents. Benefits and costs must be balanced in determining agendas, investing political support, and trading votes. They would not want to create extra conflict for themselves by routinely defecting from beneficial coalitions that had been costly to assemble.

Accordingly, our explanation for programmatic stability focuses on the high political and resource costs of allowing legislative deals to collapse. Coalitions do not fail in isolation. Because agreement among politicians directly concerned with a narrow program may not carry enough votes for passage, logrolling exchanges with politicians interested in other legislation are required. The basis for these trades, however, are placed at risk with any significant adjustment in program distributions. Tullock's (1981, 195) observation that "congressional boodle is passed around more or less equally," across political jurisdictions within the United States is germane. The comparatively egalitarian outcome reflected in the SPEND/TAX ratios presented in Table 3 is the result of many logrolling exchanges where stable coalition formation requires assembling political support among many politicians and their constituents. (25) The individual programs that make up the aggregate SPEND/TAX ratios need not (and likely do not) benefit all states in the same portion as the aggregate ratios. Constituencies and demands vary across political jurisdictions. But individual programs are enacted as part of broader vote trades and logrolling so that no states are entirely left out. In this manner, all politicians have an interest in the maintenance of the various programs because their success affects the endurance of coalitions for other legislation. Accordingly, efforts to change, say, agricultural programs must not only confront the congressional committee structure that Shepsle and Weingast (1981a) point to as a stabilizing influence but also all other politicians who participated in logrolling exchanges in the general legislature that involved agricultural legislation. Because of these far-reaching effects of coalition breakdown and the associated costs of reassembling new agreements across many legislative interests, politicians will resist major changes in programmatic sharing.

Transaction costs and universalist distributions provide a different explanation for the observed crisis/ratchet phenomena in the growth of government programs. Higgs (1987) argues that these ratchets reflect shifts in ideology among voters toward greater acceptance of government intervention during a crisis and subsequent interest group dependence on transfers. These factors may play a role, but they do not consider the political bargaining underlying the adoption and possible elimination of government programs. Responding to new legislative demands during a crisis involves political agreements and logrolling trades among many politicians. Universalist sharing to build broad coalitions suggests that a crisis will lead to an increase in expenditures, possibly accompanied by new programs, so that all political jurisdictions benefit. The response to the crisis will not be allowed to upset the status quo SPEND/TAX rankings by having only a few political jurisdictions receive the bulk of new spending. The expansi on of New Deal government programs in the 1930s appears to be consistent with this view. (26) Once coalitions are assembled, there will be resistance to program cuts that threaten benefit streams. Any adopted reductions will be spread across politicians and constituents, rather than rely upon elimination of a few programs. In this way, there will be universalist sharing in the reductions and a maintenance of the general distribution of government benefits. (27) The high transaction costs of changing major policies when there is broad sharing explains North's (1990) observation that institutional change involving legislation takes place only incrementally without the wide swings suggested by cyclical legislative majorities.

V. CONCLUDING REMARKS

Despite the near completion of the interstate highway network by the late 1970s and changing demands that emphasized urban transit systems, allocations from the HTF have been remarkably stable. A standard of stability is whether the past is a good predictor of the present. In the case of HTF allocations, aggregate state HTF apportionment/tax ratios for 1956-73 are significant predictors of 1974-97 state ratios. Furthermore, comparing the stability of state apportionments from the period 1974-97 to a relevant measure of highway use, motor vehicle registration, reveals that state apportionments have not been elastic to changes in vehicle registration. Although HTF apportionments expanded by almost 40% in real terms between 1974 and 1997, the estimated elasticity for apportionments with respect to motor vehicle registration is close to zero and not statistically significant. Hence, the apportionment series appears quite stable and is not responsive to shifts in a major highway use variable. Political factors det ermined the allocation scheme, and politicians have been reluctant to importantly change interstate distributions.

We argue that the high transaction costs of coalition formation and maintenance contribute to program stability and that the opportunity costs of allowing logrolling exchanges to collapse induces members of Congress to avoid major programmatic changes associated with cycling majorities. Though we do not provide a direct test of this hypothesis, HTF changes have been minor and incremental. (28) The general continuity of all federal distributions over time is revealed in the aggregate SPEND/TAX ratios from the period 1975-97. There is no evidence of cycling. If cycling were to occur, it is more likely to take place in very small programs where the overall egalitarian sharing constraint does not matter much. The readjustment in other programs if the coalition surrounding a small program were to collapse could be minimal, and the logrolls that brought it about would be of little consequence to most legislators. In contrast, if the coalition surrounding a major program were to fail, the egalitarian constraint impl ies that either a new program must be assembled quickly to take its place or that a significant reallocation of all distributions must occur. Given the high transaction costs involved, the latter seems unlikely.

Consideration of transaction costs in coalition negotiation and enforcement makes it more difficult to draw clear efficiency conclusions regarding government programs. Although the allocation rules adopted by Congress to preserve long-term political coalitions serve well the objectives of the elected members, they can generate outcomes that have little semblance to what most economists would consider efficient. But high costs for reaching an agreement, high policing costs, and the interrelated nature of logrolling exchanges constrain program changes, even as underlying demands change. For these reasons, the rules Congress adopts for allocating funds will generate outcomes that reflect heterogenous demands and side payments, rather than the expenditure of funds in a benefit- and cost-effective manner. (29) The alternative to such programmatic sharing rules could be an inability to reach any stable agreement on long-term constituent programs.

To illustrate the nature of the problem, those who have examined universalistic distributions, such as Weingast et al. (1981) and Inman and Rubinfeld (1997), have commented on the apparent inefficiency of such program allocations. But if minimizing the transaction costs of coalition formation and maintenance in an effort to ensure long-term constituent benefits is a motivation for broad sharing rules, then these inefficiency conclusions may be inappropriate. (30) Clearly, transaction costs in politics are positive, and we should not expect outcomes in either the market place or the political arena to lie along the same frontier as they would in the absence of these costs. Consideration of transaction costs, however, does not imply that outcomes examined on the basis of their apparent costs and benefits are of no value. Rather, it suggests that institutional change that would remedy the situation may be too costly to achieve.

In the empirical case at hand, a study by U.S. General Accounting Office (1995, 21-24) reported that certain factors, such as rural population and highway mileage factors, used to allocate highway funds across the states were irrelevant or outdated and called for apportionments to be more reflective of a state's contribution to the HTF. As we have argued, although allocating funds to the states based on their contributions would seemingly yield the highest economic return, such criteria would not generate the political consensus necessary to achieve a national highway program.

The allocation rules for the HTF also conflict with those who have

argued that user taxes and formula earmarking, as in the case of the federal excise tax on gasoline, are a means of generating more efficient outcomes. Teja (1991, 13-26) suggests that earmarked allocations are adopted to avoid perverse outcomes possible under general fund financing. (31) Our analysis suggests that the real motivation is the need to build political coalitions and maintain them. As such, any direct cost-benefit gains, in the traditional sense, may be little more than coincidental.

TABLE 1

Ratios of Approtionments to Tax Payments, HTF

 Cumulative Cumulative
State 1956-73 1956-97

Alabama 1.15 1.11
Alaska 8.86 6.69
Arizona 1.48 1.17
Arkansas 0.89 0.99
California 0.84 0.95
Colorado 1.16 1.31
Connecticut 1.22 1.79
Delaware 1.31 1.50
Florida 0.66 0.89
Georgia 0.82 0.92
Hawaii 3.06 3.78
Idaho 1.59 1.72
Illinois 1.10 1.11
Indiana 0.80 0.87
Iowa 0.93 1.14
Kansas 0.91 1.10
Kentucky 1.16 1.06
Louisiana 1.42 1.23
Maine 1.12 1.15
Maryland 1.14 1.45
Massachusetts 1.06 1.72
Michigan 0.85 0.90
Minnesota 1.18 1.26
Mississippi 1.07 0.99
Missouri 0.96 0.95
Montana 2.69 2.37
Nebraska 0.97 1.14
Nevada 2.08 1.51
New Hampshire 1.40 1.40
New Jersey 0.82 1.02
New Mexico 1.64 1.34
New York 0.98 1.22
North Carolina 0.60 0.86
North Dakota 1.95 1.96
Ohio 1.01 0.92
Oklahoma 0.80 0.86
Oregon 1.43 1.21
Pennsylvania 0.97 1.15
Rhode Island 1.55 2.22
South Carolina 0.77 0.88
South Dakota 1.95 1.97
Tennessee 1.11 0.98
Texas 0.77 0.85
Utah 2.13 1.58
Vermont 2.70 2.13
Virginia 1.21 1.12
Washington 1.32 1.55
West Virginia 2.30 1.99
Wisconsin 0.70 0.90
Wyoming 2.97 1.91
Total 1.04 1.12

Source: Highway Statistics, Table FE-221.

TABLE 2

Descriptive Statistics for the Apportionment/Tax Ratios

 1974 1975 1976 1977 1978 1979 1980 1981

Mean 1.44 1.40 1.41 1.29 1.33 1.40 1.38 1.37
Median 1.13 1.12 1.18 1.06 1.15 1.08 1.15 1.16
Maximum 11.08 9.62 9.91 6.13 7.23 8.41 7.76 7.29
Minimum 0.60 0.60 0.69 0.52 0.58 0.49 0.56 0.58
SD 1.48 1.28 1.30 0.87 0.98 1.22 1.07 1.01
Skewness 5.69 5.49 5.69 3.83 4.53 4.18 4.47 4.28

 1982 1983 1984 1985 1986 1987 1988 1989

Mean 1.39 1.31 1.29 1.27 1.26 1.29 1.31 1.32
Median 1.10 1.06 1.10 1.07 1.12 1.06 1.07 0.99
Maximum 6.41 5.63 4.46 4.22 4.13 5.02 6.66 6.80
Minimum 0.57 0.66 0.73 0.69 0.69 0.72 0.65 0.62
SD 0.97 0.82 0.68 0.70 0.65 0.78 0.97 1.09
Skewness 3.38 3.38 2.76 2.65 2.60 3.08 3.83 3.67

 1990 1991 1992 1993 1994 1995 1996 1997

Mean 1.32 1.27 1.32 1.29 1.23 1.24 1.27 1.23
Median 0.94 0.94 0.99 1.02 1.00 0.98 1.03 0.98
Maximum 6.55 4.16 6.92 5.61 5.31 5.93 5.20 4.62
Minimum 0.53 0.69 0.71 0.66 0.66 0.52 0.74 0.81
SD 1.09 0.83 1.04 0.91 0.79 0.80 0.73 0.65
Skewness 3.35 2.29 4.06 3.43 3.43 4.21 3.40 3.26

TABLE 3

Aggregate SPEND/TAX Ratios for 1975 and 1997

 1975 1997
State Ratio Rank Ratio Rank

Alaska 2.44 1 1.44 8
New Mexico 1.93 2 1.94 1
Mississippi 1.76 3 1.61 3
Hawaii 1.58 4 1.50 6
Washington 1.40 5 0.98 31
North Dakota 1.35 6 1.66 2
Utah 1.35 7 1.02 27
Virginia 1.34 8 1.50 5
Alabama 1.34 9 1.38 12
Arizona 1.31 10 1.09 23
South Dakota 1.29 11 1.35 14
Montana 1.28 12 1.49 7
Idaho 1.25 13 1.16 19
Arkansas 1.24 14 1.32 15
Oklahoma 1.22 15 1.40 9
West Virginia 1.21 16 1.60 4
Kentucky 1.21 17 1.37 13
Wyoming 1.21 18 1.13 22
Maryland 1.20 19 1.31 16
Colorado 1.20 20 0.91 37
South Carolina 1.19 21 1.23 17
Vermont 1.17 22 0.98 32
Louisiana 1.16 23 1.38 11
Georgia 1.16 24 0.99 30
Tennessee 1.13 25 1.16 20
Maine 1.12 26 1.38 10
California 1.11 27 0.93 35
Missouri 1.10 28 1.23 18
Texas 1.03 29 0.99 29
Florida 1.00 30 1.06 24
New Hampshire 1.00 31 0.72 47
North Carolina 0.98 32 1.03 26
Kansas 0.98 33 0.96 34
Nevada 0.96 34 0.74 46
Massachusetts 0.95 35 0.91 39
Oregon 0.94 36 0.92 36
Rhode Island 0.92 37 1.14 21
Connecticut 0.92 38 0.67 50
New York 0.89 39 0.84 41
Pennsylvania 0.87 40 1.01 28
Nebraska 0.84 41 0.97 33
Minnesota 0.83 42 0.76 44
Indiana 0.73 43 0.89 40
Wisconsin 0.73 44 0.80 43
Illinois 0.72 45 0.70 48
Ohio 0.70 46 0.91 38
Iowa 0.69 47 1.03 25
Delaware 0.66 48 0.80 42
New Jersey 0.66 49 0.67 49
Michigan 0.65 50 0.75 45

Source: Statistical Abstract of the United States 1978, Table 433, 267,
and Statistical Abstract of the United States 1999, Table 555, 356.

Notes: SPEND is the amount of Federal spending within the borders of a
state, including expenditures on defense. TAX represents the amount of
federal tax revenue collected from or apportion to each state.

(1.) For summaries, see Enelow (1997) and Rae and Schickler (1997) on cycling and majority rule.

(2.) Only limited empirical analyses on the extent of cycling have been undertaken. Stratmann (1996) offers the most detailed examination. He provides evidence of stable collective choices in Congress regarding pork barrel expenditures at the congressional district level from 1985 through 1990. Our analysis covers a much longer time period.

(3.) Collie defines universalistic voting as those votes where at least 90% of the voting members vote in the same direction. Our emphasis in this article, however, is on the share of distributive benefits.

(4.) In his study of pork barrel expenditures by the U.S. Army Corps of Engineers, Ferejohn (1974) finds distributional patterns similar to those described for the HTF in the text. He also (193-94) notes strong, stable congressional support for the HTF maintained through the distribution of rewards. With regard to Corps of Engineers projects (237-40, 246-47) he describes stable winning coalitions that were larger than a bare majority. These coalitions were assembled via a complex series of bargains and logrolling trades among members of many congressional committees. Within the coalition there was broad sharing of Corps of Engineers expenditures, so that all parties developed a stake in the associated legislation.

(5.) See Hall (1987) for a discussion of how members of the House of Representatives have an almost unlimited range of demands on their time and are constantly seeking ways to better allocate their time.

(6.) A provision for ensuring that no state received less than 0.5% of annual apportionments was added early in 1921 and maintained through 1982.

(7.) Although highway funds had been collected and distributed by the federal government since 1916, the separate HTF was created in 1956 by the Federal-Aid Highway Act of 1956 (Pub. L. 70-627). Federal-aid primary and secondary roads received apportionments according to the past formula. The interstate system included cost of completion variables and a provision that no state receive less than 0.75% of annual apportionments.

(8.) The 85% figure was raised to 90% for 1990 and 90.5% in 1998.

(9.) For an outline of the apportionment formulas across time, see U.S. Department of Transportation, Highway Statistics, various years. A history of the HTF and description of the various user taxes can be found in U.S. Department of Transportation (1998, 1999, 31-35).

(10.) See statement by Sen. Robert Kerr (OK) in Congressional Record--Senate, 28 May, 1956, 9079.

(11.) For statements from legislators from other donor states, see also pages 28884 and 28915.

(12.) These ratios are generally referred to as donor/donee ratios, although they actually reflect the ratio of apportionments to excise tax payments. States that contribute more than they receive back are referred to as donor states. The last row in Table 1, labeled Total, shows the ratio for the fund as a whole. Because there are years when the HTF pays out less than it takes in and other years when the opposite is the case, the state observations for each year are normalized by dividing each state's ratio by the total ratio for that year.

(13.) For a discussion of how these discretionary funds are used to influence votes on highway bills, see Evans (1994).

(14.) Annual state-level ratio data are apparently available only from 1974 on. We searched data bases for the Federal Highway Administration and contacted the National Transportation Library but located no additional data. As indicated in the table the distribution of the apportionment/tax ratios is skewed to the right because many states, such as Alaska, Hawaii, Montana, West Virginia, Vermont, and Wyoming, have received much more than they have contributed to the highway trust fund for the political reasons we describe in the text. For these states the ratios are above one. There is no reason to believe that the ratios would be distributed normally. The table also reveals that the skewness has gradually declined over time. As we point out in the text, states that have benefited and received more than they have paid into the HTF have had to give up some trust fund allocations to more urban states that historically had ratios below unity. Even so, the overall distributions have not changed substantially, and we cannot reject equality of the means.

(15.) We also examined whether using the tax/apportionment ratio instead of the apportionment/tax ratio affected our results. It did not. The Brown and Forsythe test for the equality of the variances of the tax/apportionment ratios yielded a statistic of 0.798.

(16.) Stratmann (1996) proposed a unique test for cycling. Essentially, if there is substantial variance in allocations/outcomes on a periodic basis relative to the variance of the allocations, summed up over the entire period, that could be indicative of cyclical majorities. The underlying notion is that cycling generates stark differences between winners and losers in the short run, but over the long run the returns average out. A variant of his test as applied to the apportionment/tax ratios is to first compute the coefficients of variations for each year, sum these up, and then divide by 24. This provides an average measure of the amount of variation occurring annually. This measure can then be compared to the coefficient of variation derived by first computing the average ratio for each state over the 24-year period, and then obtaining the mean and standard deviation of these average ratios. Using the apportionment/tax ratios of the HTF, the coefficient of variation for the former measure is 0.713, and t he latter, 0.655. Although these results can not reject cycling, there is little difference between the two measures.

(17.) Data on milage by state are from U.S. Department of Transportation, Highway Statistics (1997, Table HM-20).

(18.) In addition to state fixed effects, state individual rhos were estimated and used to correct for auto correlation. The apportionment figures were deflated using the Consumer Price Index.

(19.) Inclusion of a trend variable does not affect the results. Moreover, if the individual state effects are dropped from the regression the R-squared term drops to 0.20, suggesting that motor vehicle registrations alone are not a good predictor of apportionments.

(20.) Although there is evidence of redistribution, from higher to lower income states as noted by Peltzman (1985), the allocation of federal funds is far more broad based than the simple game theory models suggest.

(21.) Poole and Rosenthal's contention that members of Congress vote along one or two dimensions that reflect mainly ideology has not gone unchallenged. Heckman and Snyder (1997) argue that the number of the dimensions is somewhat larger and reflects issue-specific attributes of Congress. Koford (1990, 1994) analyzes multiple roll-call votes and raises similar objections but also argues that the consistency found by Poole and Rosenthal and others can be explained by a desire to minimize transaction costs in logrolling. He does not detail the nature of the transaction costs involved, but suggests that political parties have a huge stake in satisfying constituencies via successful coalitions.

(22.) Coleman (1983) in his article on trustworthiness suggested that there would be less stability in a lame duck legislature. Lott (1987, 1990) examined whether or not last-period politicians indulged their ideological preferences and voted for programs that were not in their constituents' interests when no longer constrained by reelection. Other than some reduction in attendance rates, he found no significant change in voting patterns for politicians about to retire. In addition to these issues, the incentive to defect to form a new political coalition must also be tempered by expected retaliation from members of the old coalition who can make counterbargains to those considering a new coalition and punish the defector and his or her constituents.

(23.) If transaction costs were zero, implying that information costs were zero, there would be no reason to expect cycling. Costless contracting and enforcement imply a stable equilibrium. Hence, cycling is most likely to occur where transaction costs are relatively low but not equal to zero.

(24.) This outcome suggests implausible myopia among legislators. However, in examining the proposal and gate-keeping powers of committees Baron and Ferejohn (1989) argue that under an open amendment rule, the larger the discount factor facing legislators, the more egalitarian the distribution of benefits. A summary of some of the literature on majority rule, cycling, and broad sharing is provided by Rae and Schickler (1997) and Stratmann (1997). Stratmann (1997, 334) adds discussion of transaction costs as contributing to stability.

(25.) Accordingly, universalism and stability apply to both individual pieces of legislation and to larger legislative bargains. It might be argued that if cross-legislative deals support universality and stability there is no need for balance within a particular piece of legislation. But this notion is incorrect. As we argue, a lack of balance in a particular program would encourage coalition defection, undermining the basis for broader logrolling exchanges. Moreover, a small coalition for narrow legislation would rely even more on broad logrolling exchanges for enactment. The costs of devising logrolling exchanges across heterogeneous legislators likely are higher than forming agreements among the more homogeneous legislators interested in the narrow program. This suggests that politicians would form as large a coalition as possible in support of a specific bill before turning to broader logrolling exchanges to ensure passage. An example of broad sharing within a narrow program is the distribution of ethano l plants planned under the early ethanol subsidy. Forty-six of 50 states were to receive at least one ethanol production plant, even though ethanol was primarily made from corn grown in the Midwest. See U.S. National Alcohol Fuels Commission (1981, 57).

(26.) For analysis of the distribution of federal programs and expenditures during the New Deal, see Wallis (1998).

(27.) Similarly, the dramatic reforms undertaken by New Zealand in the 1980s in response to a financial crisis led to broad-based cuts, rather than elimination of a few programs. There appears to have been a universalist sharing of the expenditure reduction. For discussion regarding the sharing of budget cuts and policy reforms see Scrimgeour and Pasour (1996).

(28.) A direct test would require a measurable change in the transaction costs of negotiating and enforcing logrolling agreements along with observations of changes in sharing rules and/or stability. Direct measures of transactions costs arc, however, unlikely ever to be complete. Alternatively, consider the situation where political parties were adept at enforcing deals but subsequently lost that power, then the transaction costs of coalition maintenance should rise. Under these circumstances, elected representatives would respond by increasing the degree to which program benefits are shared. There is empirical support for this prediction. Collie's (1988) results indicate that as party power lessened in the postwar period, universalism increased.

(29.) Discussions of the consequences of universalism have also pointed to its potential for inefficient outcomes. See Shepsle and Weingast (1981a, b).

(30.) As argued in Johnson and Libecap (2001), once transaction costs are introduced, the standard concept of Pareto efficiency loses meaning. For discussion of some of the key issues regarding transaction costs and redistribution in politics, see Williamson (1996, 195-213). Williamson (1998, 11-16) attempts to resolve some of the conflicting issues regarding the efficiency of government policies by introducing a remediableness criterion for evaluating programs.

(31.) See, for example, the papers in Wagner (1991).

REFERENCES

Baron, D. P., and J. A. Ferejohn. "The Power to Propose," in Models of Strategic Choice in Politics, edited by P. C. Ordeshook. Ann Arbor: University of Michigan Press, 1989, 343-66.

Coleman, J. S. "Recontracting, Trustworthiness, and the Stability of Vote Exchanges." Public Choice, 40(1), 1983, 89-94.

Collie, M. P. "Universalism and the Parties in the U.S. House of Representatives, 1921-80." American Journal of Political Science, 32(4), 1988, 865-83.

Congleton, R. D., and R. D. Tollison. "The Stability Inducing Propensities of Very Unstable Coalitions: Avoiding the Downward Spiral of Majoritarian Rent-Seeking." European Journal of Political Economy, 15(2), 1999, 193-205.

Enelow, J. M. "Cycling and Majority Rule," in Perspectives on Public Choice: A Handbook, edited by D. C. Mueller. New York: Cambridge University Press, 1997, 149-62.

Evans, D. "Policy and Pork: The Use of Pork Barrel Projects to Build Policy Coalitions in the House of Representatives." American Journal of Political Science, 38(4), 1994, 894-917.

Ferejohn, J. A. Pork Barrel Politics; Rivers and Harbors Legislation, 1947-1968. Stanford, CA: Stanford University Press, 1974.

Gilligan, T. W, and K Krehbiel. "Collective Decisionmaking and Standing Committees: An Informational Rationale for Restrictive Amendment Procedures." Journal of Law, Economics, and Organization, 3(2), 1987, 287-335.

Glazer, A., and H. McMillan. "Amend the Old or Address the New: Broad-Based Legislation When Proposing Policies Is Costly." Public Choice, 74(1), 1992, 43-58.

Hall, R. L. "Participation and Purpose in Committee Decision Making." American Political Science Review, 81(1), 1987, 105-28.

Heckman, J. J., and J. M. Snyder Jr. "Linear Probability Models of the Demand for Attributes with an Empirical Application to Estimating the Preferences of Legislators." RAND Journal of Economics, 28, 1997, S142-S189.

Higgs, R. Crisis and Leviathan: Critical Episodes in the Growth of American Government. New York: Oxford University Press, 1987.

Inman, R. P., and D. L. Rubinfeld. "The Political Economy of Federalism," in Perspectives on Public Choice: A Handbook, edited by D. C. Mueller. New York: Cambridge University Press, 1997, 73-110.

Johnson, R. N., and G. D. Libecap. "Information Distortion and Competitive Remedies in Government Transfer: The Case of Ethanol." Economics of Governance, 2, 2001, 101-34.

Koford, K. "Dimensions, Transactions Costs and Coalitions in Legislative Voting." Economics and Politics, 2(1), 1990, 59-82.

-----. "What Can We Learn about Congressional Politics from Dimensional Studies of Roll-Call Voting?" Economics and Politics, 6(2), 1994, 173-86.

Lott, J. R. Jr. "Political Cheating." Public Choice, 52(2), 1987, 169-87.

-----. "Attendance Rates, Political Shirking, and the Effect of Post-Elective Office Employment." Economic Inquiry, 28(1), 1990, 133-50.

Miller, G. J., and J. A. Oppenheimer. "Universalism in Experimental Committees." American Political Science Review, 76(3), 1982, 561-74.

Mueller, D. C. Public Choice II. Cambridge: Cambridge University Press, 1989.

North, D. C. Institutions, Institutional Change, and Economic Performance. New York: Cambridge University Press, 1990.

Peacock, A. T., and J. Wiseman. The Growth of Public Expenditures in the United Kingdom. Princeton: Princeton University Press, 1961.

Peltzman, S. "An Economic Interpretation of the History of Congressional Voting in the Twentieth Century." American Economic Review, 75(4), 1985, 656-75.

Poole, K. T., and H. Rosenthal. Congress: A Political-Economic History of Roll Call Voting. New York: Oxford University Press, 1997.

Rae, D. W, and E. Schickler. "Majority Rule," in Perspectives on Public Choice: A Handbook, edited by D. C. Mueller. New York: Cambridge University Press, 1997, 163-80.

Scrimgeour, F. G., and E. C. Pasour Jr. "A Public Choice Perspective on Agricultural Policy Reform: Implications of the New Zealand Experience." American Journal of Agricultural Economics, 78(2), 1996, 257-67.

Shepsle, K. A., and B. R. Weingast. "Structured Induced Equilibrium and Legislative Choice." Public Choice, 37(3), 1981a, 503-19.

-----. "Political Preferences for the Pork Barrel: A Generalization." American Journal of Political Science, 25(1), 1981b, 96-111.

Stratmann, T. "Instability of Collective Decisions? Testing for Cyclical Majorities." Public Choice, 88(1-2). 1996, 15-28.

"Logrolling," in Perspectives on Public Choice: A Handbook, edited by D. C. Mueller. New York: Cambridge University Press, 1997, 322-41.

Teja, R. S. The Case for Earmarked Taxes: Government Spending and Public Choice. London: Institute of Economic Affairs, 1991.

Tullock, G. "Why So Much Stability?" Public Choice, 37(2), 1981, 189-205.

U.S.Department of Transportation. Federal Highway Administration. Highway Statistics, various years.

-----. Highway Trust Fund Primer. November 1998.

-----. Financing Federal-Aid Highways. Publication No. FHWA-PL-99-015, August 1999.

U.S.General Accounting Office. Highway Funding: Alternatives for Distributing Federal Funds. GAO/RCED-96-6, November 1995.

U.S.House of Representatives. "Federal Highway and Highway Revenue Acts of 1956." 84th Cong., 2nd Sess. H. Rep. 2022. Washington, DC: U.S. Government Printing Office, 1956.

U.S. National Alcohol Fuels Commission. "Fuel Ethanol: An Energy Alternative for the 1980s," Final Report. Washington, DC: U.S. Government Printing Office, 1981.

U.S. Senate, Committee on Public Works. "Federal-Aid Highway Act of 1956." 84th Cong., 2nd Sess. Report No. 1965. Washington, DC: U.S. Government Printing Office, 1956.

VanDoren, P. M. Politics, Markets, and Congressional Policy Choices. Ann Arbor: University of Michigan Press, 1991.

Wagner, R. E., ed. Charging for Government: User Charges and Earmarked Taxes in Principle and Practice. London: Routledge, 1991.

Wallis, J. J. "The Political Economy of New Deal Spending Revisited, Again: With and without Nevada." Explorations in Economic History, 35(2), 1998, 140-70.

Weingast, B. R. "A Rational Choice Perspective on Congressional Norms." American Journal of Political Science, 23(2), 1979, 245-62.

Weingast, B. R., and W J. Marshall. "The Industrial Organization of Congress; or, Why Legislatures, Like Firms, Are Organized as Markets." Journal of Political Economy, 96(1), 1988, 132-63.

Weingast, B. R., K. A. Shepsle, and C. Johnsen. "The Political Economy of Benefits and Costs: A Neoclassical Approach to Distributive Politics." Journal of Political Economy, 89(4), 1981, 642-64.

Williamson, O. E. The Mechanisms of Governance. New York: Oxford University Press, 1996.

-----. "Public and Private Bureaucracies: A Transactions Cost Economics Perspective." Working paper, Haas School of Business, University of California, Berkeley, May 1998.

RONALD N. JOHNSON and GARY D. LIBECAP *

* Earlier versions of this article were presented at the 2000 Meetings of the Western Economics Association, Vancouver, BC, Canada, 29 June-3 July; 2000 Meetings of the International Society for the New Institutional Economics, Tuebingen, Germany, 22-24 September; and the Public Choice Center Workshop, George Mason University, 9 November 2000. We benefited from comments at those meetings as well as from comments from the editors, referees, and Rob Fleck.

Johnson: 2206 River Run Drive, Unit 38, San Diego, CA 92108. Phone 1-619-284-6629, E-mail Ronnilsjohnson@aol.com

Libecap: Professor of Economies and Law, Karl Eller Center, University of Arizona, 202D McClelland Hall, Tucson, AZ 85721, and Research Associate, National Bureau of Economic Research, Phone 1-520-621-4821, Fax 1-520-626-5269, E-mail glibecap@eller.arizona.edu