Political Preferencesfor the Pork Barrel:A Generalization*

Kenneth A. Shepsle, Barry R. Weingast, Washington University

This paper provides a rational explanation ror the observation or oversized coalitions,orten approaching unanimous size, in the realm or distributive policies. Distributive policiesare those which concentrate benefits in specific geographic areas (states, congressional dis-tricts) while spreading costs through general taxation. The explanation orrered here-thatlegislators may rationally prerer universalism to "hard-ball" coalition politics-generalizesearlier work or Weingast and Fiorina. In particular, it extends their results to pork-barrel poli-tics, i.e., projects that are economically inerricient, and demonstrates how packages or eco-nomically inefficient projects can nevertheless be politically popular. The main theoretical rea-ture thai induces prererences ror universalism is uncertainty over the composition or winningcoalitions and, consequently, the desire ror the political "insurance" thai universalismprovides.

The purpose of this brief paper is to establish some general conditionsunder which rational legislators institute and maintain the norm of uni-versalism in the realm of distributive politics. This practice, often noted tobe the predominant feature of congressional land and resourcepolicy, riversand harbors omnibuses, and the traditional tariff, is characterized by legis-lative support coalitions well in excess of minimal winning size and oftenapproaching unanimous size. This empirical fact, moreover, is not re-stricted to the traditional pork barrel. Congressman David Stockman (1975)finds "the distributive tendency" in many of the recent social policies issuch that, while the initial authorization and/or appropriation establishinga program may be controversial (and henceassociated with a barely winningsupport coalition and a large opposing coalition), after a time legislativesupport grows for subsequent reauthorizations as more and more districtsbecome beneficiaries of the program. As noted, this practice of uni-versalism has long been apparent in decisionson the traditional pork barrelof rivers and harbors (Maass, 1951; Ferejohn, 1974). More recently, uni-versalisticcoalitions have emergedin such diverse realms as modelcitiesandurban renewal(Plott, 1967),tax loopholes (Manley, 1970),private member

.Mr. Shepsle is Proressor or Political Science and Research Associate at the Center ror theStudy or American Business, Washington University. Mr. Wcingast is Assistant Proressor orEconomics and Research Associate at the Center ror the Study or American Rnsiness, Wash-ington University. The authors thank the editors and rererees ror their coml11ents.

Amnicln Journal or Polilical Scitnct, Vol. 2S, No. I, Ftbru8ry 19ftl

@19RIbytht Univt.sily of Trus Prn" 0092-~85]/RI/OI()(}%.lnOI."5

POLITICAL PREFERENCES FOR THE PORK BARREL 9798 Kenneth A. Shepsle, Barry R. Weingast

bills (Froman, 1967), military procurement (Rundquist, 1973), and cate-gorical grants-in-aid (Mayhew, 1974). 1n each of these areas it is not uncom-mon to observe something for (nearly) everyone in omnibus authorizationsand appropriations.

The empirical regularity of universalism in the realm of distributivepolicies stands in stark contrast to the predictions of many early (essentiallygame-theoretic) models of simple majority rule. These theories imply that aminimum winning coalition (MWC) will form in order to capture all of thebenefits of a program while paying only a fraction of the costs; yet, dis-tributive policy decisions rarely can be described in these terms.

Because most real-world legislatures are not simple majority rule insti-tutions, several modifications of the basic game-theoretical model are re-quired to make them relevant for legislative behavior. Several of these areexplored in the concluding chapter of Ferejohn's study of pork-barrel poli-tics (Ferejohn, 1974). There he notes three qualifications to majority rulemodels as they apply to real legislatures: the instability of winning coali-tions, the distribution of relative veto-power, and the agenda power of com-mittees.

Instability of Winning Coalitions

The simple game-theoretic models predict MWCs, but, as is well-known, there are many of these, no one of which is uniquely dominant.Therefore, in the view of Buchanan and Tullock (1962). some particularMWC will form, pass a distributive policy concentrating benefits on itsmembers, and, in most instances, fail to survive to the next iteration, beingreplaced instead by some different (though partially overlapping) MWCwhich passes its own program. This view accommodates a kind of over-timeuniversalism in which legislators take turns as members of the victoriousMWC, but it does not explain near-universal coalitions on any single occa-sion.

Distribution of Relative Vetoes

Barry (1965) observes that real-world legislatures are not suitably char-acterized as symmetric games, since some (many?) legislators possess rela-tive vetoes. Ferejohn (1974, p. 238), for example. lists eight veto-individualsor veto-groups concerned with Corps of Engineers appropriations. rangingfrom the chairmen of the Public Works Appropriations Subcommittees ofthe House and Senate to the conferees from those two chambers. The morewidely distributed these relative vetoes are. the more inclusive the final win-ning coalition must be. However, unless these vetoes are so widely dis-tributed as to destroy the basic majority-rule character of a legislature, inmost instances the relative-veto argument only constrains which MWC will

form. That is, the relative-veto argument specifies who must be included inthe victorious MWC (otherwise he will exercise his veto); rarely does this re-quire dramatically oversized coalitions.

Committee Agenda Power

Yet another qualification of the simple majority rule model is the exist-ence of committees with the power to propose alternative policies to thelegislature and, within limits specified by the rules, to restrict amendmentsto their proposals. This qualification, associated with the work of Niskanen(1971), suggests not only that authorizing and'appropriating committeespossess relative vetoes; it also suggests (owing to the unrepresentative char-acter of most committees) a bias toward particular kinds of policies, e.g..big dams with lots of "bells and whistles." But it provides no basis for con-cluding that powerful committees will be disposed toward omnibus billswith projects for nearly every constituency.

Although these qualifications are important, they do not provide muchpurchase on explanations of universalism. We still do not know, that is,why universal or, at any rate, oversized coalitions are observed with con-siderable frequency in the politics of distributive policy. Two recentstudies-those of Fiorina (1978) and Weingast (I979)-as well as Ferejohn's"Conclusion" (1974) have addressed this anomaly directly. The results wereport below generalize some of their conclusions. Elsewhere (Weingast,Shepsle, and Johnsen, 1979) we examine the policy consequences of a uni-versalistic system, with special emphasis on the incentives for inefficiencycontained in its operation. Here we provide. by way of explanation, a gen-eral condition which insures that rational legislators prefer it to a pure ma-jority-rule system, on the one hand, and to the constitutional abolition ofthe pork barrel on the other. I

Theoretical Forerunners

The logic of previous analytical work on distributive politics (Wein-gast, 1979; Fiorina, 1978) establishes legislator preferences for universalismover pure majority rule. We briefly summarize Weingast's results, withoccasional reference to Fiorina. to lay the foundation for our more general

'There are really Iwo questions that need to be addressed. The first is explanatmy_ Why.or under what general conditions, does an oversized coalition form around a distrihuthepolicy? Put differently, when will a ralionallegislalor hold an ('.f alii" prdcrence fm univer-salism as opposed 10 stricl majority rule? An answer to Ihis qneslion provides an expl:uHllionfor an empirical regnlarilY. namely the rouline ohservation of large suppmting m:llitions for

dislribnlive policies. A secoud question is more descriptive or predictive. vi7... whal me thepolicy consequences of a universalistic system of decision? That is. whal kind of policie,.;, withwhat characteristics, are prodnced hy a universalistic system?

POLITICALPREFERENCESFORTHE PORKBARREL 99

conclusion. Consider a distributive policy or project Pj with the followingcharacteristics:

I. Total project benefits (present-value adjusted), b, are concentratedentirely in district}.

2. Total project costs, c, are covered through taxes with each of n dis-tricts paying (t In)c.

Thus, if a coalition of legislators S C N = 11,2,...,nl forms in order to passan omnibus of projects, one for each district represented in the coalition,the net benefits for ifS is b -(l/n)(ISIc), where ISI is the number of legisla-tors, and hence the number of proposed projects, in S. That is, the con-stituents of a coalition member receive the benefits of their project, b, andbear I/n of the costs derived from ISI projects-(Iln)(ISlc). For i~S, on theother hand, net benefits consist only of costs, namely -(l/n)(ISlc). Weingastestablishes, first, that the set of MWCs are dominant in this game, i.e., IS I =

(n + ')/2. Thu" wi,houl ru'lh" coop".liou, on, 0' Ih, (n; .) MWC, w;U

form, its members receive net benefits in the amount b _ (tIn) (n; I)c

and nonmembers receive-(~) (n; I) c.This result, as Weingast notes, "fails

to explain universalism, Le., the tendency to seek unanimous passage of dis-tributive programs through the inclusion of a project for all legislators whowant one. Indeed, this tendency constitutes evidence against the model"(Weingast, 1979,p. 249).

Weingast establishes, secondly, that an (expected) cost is borne by alllegislators not taken into account in this simple legislative model and that,by taking this cost into account, the tendency toward universalism can beexplained. In the simple legislativegame, a MWC forms, but which one?There are many and, while there are some models that attempt to predictwhich MWC will form (Ferejohn, 1974, pp. 233-252; Buchanan and Tul-lock, 1962, pp. 131-171; Fiorina, 1978), it is reasonable to suppose that

(

n-I

)each legislator, ex ante, is uncertain whether one of the n; I MWCsof

( n-I,\which he is a member, or one of the \n; 1-) MWCs of which he is no/

a member, will form. This uncertainty may be contrasted with the certaintyof universalism. Under the latter arrangement, every legislator ob-

100 Kenneth A. Shepsle, Barry R. Weingast

tains a project for his district so that his net benefits are b-(~)(nc) =b-c. Weingast contrasts the expected net benefits of simple majority rule(making reasonable assumptions about the probability of inclusion in thewinning MWC) with the certain net benefits of universalism, obtaining hisUniversalismTheorem:

If legislators maximize Ihe nel benefils which accrue to Iheir dislricts, if all MWCs areequally likely. and if b>c (with band c the same for each district). then the net bcnefilsunder universalism exceed the expected net benefits of majority rule for every legislator.

Weingast's Universalism Theorem exploits legislators' uncertaintyabout the composition of winning coalitions in order to demonstrate the exante preference for the outcome deriving from a "norm" of universalismover the expected outcome of "hardball" MWC politics. It is, of course,plausible that in some circumstances the uncertainty is insufficient to gen-erate this result, a point to which we return in the conclusion. However,there are some additional features of this theorem which are restrictive and,

therefore, will occupy our attention in the remainder of this paper. The re-sult of our analysis is a more general universalism theorem. Specifically, wefirst discard the assumption that project benefits and project costs are thesame for all projects (a point noted by Fiorina, 1978). Second, we distin-guish two different kinds of project costs, differentially evaluated by legis-lators in their net benefit computations-expenditure costs and externali-ties. Third, and possibly most significantly, we do not require (in contrast toboth Fiorina, 1978, and Weingast, 1979) that project benefits exceed projectcosts; thus, unlike previous work, we are able to extend the UniversalismTheorem to pork-barrel (read: inefficient) activities. Facilitating this exten-sion, fourth, is a reconceptualization of political benefit-cost analysis whichallows legislators to seek expenditures in their districts even when these arenot linked to strictly economic benefits.

Legislator Objectives

Consider a generic distributive policy-project P(Xf), where Xj is ameasure of project scale, which has the property of concentrating projectbenefits in the}th district while spreading costs (through taxation) across alldistricts. Analytically, we let h(xf) represent the present value of projectbenefits, all of which are "consumed" by the }th district. Total resourcecosts, c(xf), are represented as the sum of three, politically relevant,categories:

C(Xf)= c,(xj) + C2(Xj)+ c.l(xj)

POLITICAL PREFERENCES FOR THE PORK BARREL 101

Total resource costs, that is, are the sum of

(i) expenditures for project inputs spent in the district in which theproject is located (c,(xj»);

(ii) expenditures for project inputs spent outside the district in whichthe project is located (cz(Xj));and

(iii) nonexpenditure costs, e.g., externalities, imposed on the districtin which the project is located (cJ(xj»),z

Finally, we assume that taxes are raised to cover expenditures so that

T(xj) = c.(xj) + cz(Xj).

In particular, there exists a vector of tax shares for each district, t =n

(t.. . . .,tn), with Ii ~ 0 and 1: t, = I. If P(xJ requires expenditures c,(Xj) +;=1

Cz(Xj), and if taxes are required to cover expenditures-T(xj) = c.(Xj) +cz(xj)-then district i's tax bill for P(Xj) is its share of the expenditures-t,T(xj) = li(C.(Xj)+ cz(Xj»).

A legislator, in order to provide an answer to the quintessential retro-spective question on election day, viz., "What have you done for melately?", seeks to maximize his district's net benefits. Regarding own-dis-trict projects, benefits are derived from two distinct sources. First, there arethe economic benefits produced by the project, b(xj). Second there are ex-penditures made in the district, c.(xj), which have the salutary effect of in-creasing the local demand for project inputs, thereby conferring windfall-like gains on current input owners in the district, as well as producing localeconomy-wide effects in the form of increased employment, profits, andlocal tax revenues.) On the other side of the ledger are the district-ear-marked costs of a local project, namely nonexpenditure costs in the form ofexternalities, cJ(Xj),and the district's tax share, tj(c,(Xj) + cz(Xj)).Regardingprojects in other districts, th.ejth district derives "benefits" from expendi-tures made within its boundaries, say, cZj(x,) from project i( * J)', whileshouldering its share of taxes, tAc.(x,) + cJ(x,)),

'11 is. of COurse, possible 10 imagine spillover-like nonexpenditure costs for other districts,

complementing (',(x,). Owing, however, to the concentrated nature of benefits, we choose toignore these.

'11 is not unusual for these expenditures to be mistaken for and counted as a measure of

project benefits. Arnold (1979), for example, equates project benefits with local expenditures.11 is important to note that (',(x;) is not an economic benefit, but is transformed into and

treated as a "benefit" by political actors.'The amount (',(Xi) is total cost of project i spent in districts other than the ith. Specifi.

cally, the amount (',,(x,) is spent in districtj so that (',(x,) = E c,,(x;).Ic_i

102 Kenneth A. Shepsle, Barry R. Weingast

Consider, then, a vector or an omnibus of projects, x =(X,,xh . . . ,xlt . . . ,xn), where xj is the scale of the project in district j.Some of these, of course, may be identically equal to zero if a district'slegislator is not part of the prevailing coalition. The generic legislator's ob-jective function, which he seeks to maximize,consists of

(i) own project benefits: b(xj)(ii) own project in-district expenditures: c,(xj)

(iii) own project externalities: -cJ(Xj)(iv) own project tax share: -tAc.(xj) + cz(xj»)(v) other projects' districtj-targeted expenditures: 1: cZj(x,)

(vi) tax share of other projects' expenditures: ,,;.j

-tj I: (c,(x,) + cz(x,)).i*i

Items (i)-(iv) obtain only if j is part of the prevailing coalition. The politicalobjective function of each legislator, then. is a net benefit functionBj(x" . . . ,xn):

Bj(x" .' . . .xn) = Ib(xj) + c,(xj) - cJ(xj) - tAc,(xj) + cz(xj)]l

+ II: cZj(x,)- tj I: (c.(x,) + cz(x,)]I (I)i*j i_iThe first collection of terms in equation (I) are "own project" net benefits,while the second collection comprises "other projects" net benefits. LettingN(xj) represent the former, equation (I) becomes

Bj(x" . . . ,xn) = N(xj) + I: c:Jj(x,)- tj 1: (c,(x,) + c,(x,)) (2)/,;.j ,,;.j

lJ

:\

Project Evaluations

What omnibus of projects would the jth legislator enact if he had his"druthers"? Since net benefits from "own project" and net benefits from"other projects" are additively separable in equation (2)-total net benefitsdoes not depend on interactions among projects-the jth legislator's opti-mal omnibus is one in which each project scale is one that equates his mar-ginal benefits and his marginal costs:

own project:

N' (xj) = b' (Xj) + c;(xj) - c;(xj) - tj(c;(Xj) + c;(xj») = 0 (3)

other project: C;j(x,) - tAc;(x,) + c;(x,)) = 0 (4)(alli*1)

Let X; bej's "own project" scale that maximizesequation (2), i.e., the scalethat satisfies equation (3). We have shown elsewhere that the own-projectscale that maximizesdistrict net benefits, x;. generallyexceedsthe efficient-

POLITICAL PREFERENCES FOR THE PORK BARREl 103 104 Kenneth A. Shepsle. Barry R. Weingast

project scale.' This is a consequence of two political mechanismsthat trans-form the economic calculus into a political one:

(i) political cost accounting in whichgeographically-earmarked expen-ditures are transformed into political benefits; and

(ii) taxation mechanism in which beneficiaries bear only a fraction(their tax share) of the total expenditure costs.

As well, it follows that x7. the scale of the ith project regarded as optimal bythe ith legislator, exceeds what would be regarded as optimal by the jthlegislator (that is, the left-hand side of equation (4) is negative for x, = x,").

The consequence of this argument is the following. If, as a general rule,a legislator supports a project or a proposal if and only if it yieldshis districtnonnegative net benefits, then any singlepork-barrel project, set at the leveloptimal for its recipient district, can only be assured of the support of therecipient district's legislator. Indeed, not only will equation (4) be violatedfor nonrecipient districts; total net benefits for nonrecipient districts, aswell, will typically be negative-cIAx,') - tAc.(x,') + Cl(x,')]< O. Thus, inmost instances, the degenerate omnibus (0, . . . O.x.-:O, . . . ,0) will lose bythe vote of n-I: I. In order to bring home some pork, consequently, legisla-tors must either turn to cooperative logrolling arrangements or (possibly)readjust their respective project scales downward. In the remainder of thispaper we explore the former alternative.

From an arbitrary majority omnibus consisting of w projects set at thescale Xi = x,"and n- w projects set at the scalex, = 0, it is clear that the netbenefits to districtj, and consequently the support of thejth legislator, de-pend on an uncertain event, namely whether jfW or j4 W. An extension ofWeingast's (1979)argument-also see Fiorina (l978)-shows that the solu-tion to this majority-rule legislativegame involves minimum winning coali-tions (MWCs)only, so that we need only compute expectationsover the set ofx:. vectors.6From the point of viewof each legislatorj, the principal ques-tion is whether the minimum winningcoalition that prevails contains him asa member thereby assuring his district a project. If he assumes that each ma-jority coalition (and hence each x:. vector) is as likelyas any other, 7 then theprobability' that his project is included (jf W) is

(n-I

)_ w-I_~

p - (:) -. 2n

(5)

n2" + I

n even

and the probability that he is excluded from the winning coalition (j4 W)isn -I

l-p=2il' (6)

Thus, he may compute the expected valueof majority rule, Ej(MR):

Ej(MR) = p(net benefits from inclusion in MWC)

+ (I - p) (net benefits from exclusion from MWC)

=pR + (I-p)S (7)

To evaluate Rand S in equation (7) it is convenient, and in no way re-stricts our analysis, to incorporate the following expectations about j's ex-penditure benefits deriving from other projects (Cli(x7')-typebenefits) andj's tax share of other projects (tj[c.(x,")+ Cl(x,")]-typecosts):

Majority Omnibuses and UnIversalism

Consider, then, a majority omnibus, x:" consisting of w projects,where

w=n + I

2 nodd

'See Weingast et al. (forthcoming) for a derivation of this result.

'Weingast. like us, assumes that every member of a winning coalilion oblains a project.Fiorina shows that, in the absence of this condition, "bribe" coalitions may form in whichinitial losers bid away members of a previous MWC and promise to support on/." Ihe projeclsof these "bribed" individuals. This latter situation is unstable. However, Fiorina establishes

that the imposition of institutional rules of procedure restricting amendmenls reasserts theimportance of (specific) MWCs.

'We should nOle Ihat each member of a winning coalition not only obtains his projecl, but

obtains it at the scale optimal for him (according to the maximization of equation (I n. Fiorinahas argued that, when the cost of projects is known, there will be strong pressure for a specificMWC to form-namely that consisting of the w cheapest projects. This, in turn, could set abidding war into motion. We do not examine this prospect, so there is room for furlher gen-eralization (or qualification) of our results.

'These computalions are based on n odd; comparable results are obtained when n is even.

and the projects are distributed, one per district, to a winning coalition oflegislators, W. Thus, ~ is a vector with the jth component equal to x: ifjf Wand equal to zero if j4 W. If jf W, then district j receivesthe b- and C.-type benefits from its own project, the c1rtype benefits from projects of itscoalition partners, and bears tax and externality costs. These facts may beconfirmed from an inspection of equation (2).

POLITICAL PREFERENCES FOR THE PORK BARREl 105

(i) C,j is the expected (or average) expenditure in districtj from the w-Iother projects represented in the MWC. Since there are a =

(n-I

)w-I MWCs which contain j, then, letting B be the set of theseMWCs and C a generic coalition,

- 1CZj =- E E C'j (xf)a C.B I*JI.C

(ii) T is the expected (or average) tax bill for the other w-I projectsrepresented in an MWC containingj:- 1

T = - E E [c.(x,") + cz(x,")]a C.B I*J;fC

Thus, if jf W, he obtains "own project" net benefits, N(x'!), and "otherproject" net benefits, cZj - t/r:

R = N(x'/) + czr t/r (8)If j~ W, on the other hand, then N(xj) = N(O) = 0, i.e., no "own project"benefits, and "other project" net benefits are now based on w otherprojects. Thus, if cz/(w-I) and T/(w-l) are per project average expendi-tures in districtj and per project average tax burden, respectively,we have

S = w CIJ _ w tJT (9)w-I w-I

Substituting (5), (6), (8), and (9) into (7) yieldsn + 1 - n-I[

w -]Ej(MR) = 2n [N(x'/)+ CIJ- tJT] + 2n w-I (czj- tJT) (10)

In equation (10), since w = (n + 1)/2 (n odd-we develop only thiscase), w/(w - I) = (n+ I)/(n-I) and [(n-I)/2nJ[w/(w-I)] = (n+ 1)/2n.Thus equation (10)simplifies to

n+1 --EAMR) = 2n [N(x'/)+ 2(cIJ - tJT»)

which, since (n- 1)/(w- I) = 2, is rewritten (for reasons that are clear below)

I

In + 1 n-I - -EAMR)=- [N(x'/)+_ I (C1J - tjT»)2n w-

Having determined in equation (I I) j's expected benefits under pure ma-jority rule in which some MWC is expected to form, we may now computehis expectations under a system of universalismin order to compare the twoinstitutions.

Consider the universalistic omnibus, xZ, in which every district obtainsits optimum scale project, x:. In exchange for the certainty of its own

(II)

J06 Kenneth A. Shepsle, Barry R. Weingast

project, however, is the more unpleasant certainty for j that it must pay itstax share of n projects:

n -I - -EJ(U) = N(x'/) + _ I (czr tjT) (12)w-

Here, in addition to the net benefits from j's project, N(x'/), we must addthe n - I terms representing cz-type expenditures and tax shares from the

n - 1 other projects, i.:.e.,n - 1 times the per-project expenditure and taxterms, czj/(w-I) and tJT/(w-I), respectively.

We are now in a position to pose these questions: (I) under what condi-tions will rational legislators choose to employ a universalistic omnibusrather than a majority omnibus in the formulation of distributive policy?and (2) is there any basis for abolishing the pork barrel?

The first question directs us to the quantity Ej( U) - Ej(MR). In particu-lar, subtracting equation (I I) from equation (12), and dividing through by(n - 1)/2n, yields

PROPOSITIONI: EJ(U) -EAMR) > 0 if and only ifn-I - -N(x'/)+_ I (cZ)- tJT) > 0 (13)w-

To give the reader some intuition on condition (13), consider the sym-metric special case in which

(i) tJ = I/n-equal tax shares;(ii) c.(xf) = c.(x'/)-c.-type costs are the same for each project, k =

1,2,3; and(iii) e,j = (w-I)[cz(xf»)/(n-I) = (w-I)(cz)/(n-l)-cz-type expendi-

tures from every project are uniformly distributed across all n- Iother districts.

Noting that T = (w - I)(c. + cz) in this circumstance, equation (13) re-duces to

1(n -I

)[w-I 1

][b + c,-cJ-,,(cl + Cz»)+ w-I n-I cz-,,(w- I)(c. + c:z) > 0

which, with some manipulation, becomes

b-cJ>O (14)

This condition asserts that, in the symmetric special case, if legislator j be-lieves the benefit-cost ratio of his project (and, by symmetry, other projects)exceeds

cJ(x'/)c.(Xj"') + cz(Xj"') + cJ(Xj"')

then universalism is preferred to majority rule. Since this latter ratio is nor-mally much less than unity, it is not an especially demanding or unrealistic

POLITICAL PREFERENCES FOR THE PORK BARREL 107108 Kenneth A. Shepsle. Barry R. Weingast

Second, we regard Ihe geographic incidenceof costs as wellas benefitsas profoundly important in Lowian distributive politics. The hallmarks ofthese policies are (1) economic benefits concentrated geographically, (2)financing burdens dispersed geographically through the taxation mecha-nism, and (3) expenditures for project inputs with their associated geo-graphic incidence. Discussions of distributive politics tend to focus exclu-sivelyon the first two points to the exclusion of the third. This, we believe,is a significant omission, especially in light of the recent emphasis in con-gressional scholarship on the role of legislators in chasing down grants, con-tracts, and other program expenditures for their districts.

As a consequence of point (3), our third variation on the analyticalthemes of Weingast (1979) and Fiorina (1978) is to include c,-type and C'Ftype expenditures in legislator evaluations. Indeed, in some pork-barrelcontexts, e.g., the siting of military bases, expenditures constitute the prin-cipal "benefit" legislators secure for their constituents since the allegedpublic good (in this case, defense) is often a fiction.'

Finally, as noted above, our conclusion about the universalism normdoes not require the restrictive condition that projects be efficient. Univer-salism, in our model, applies without serious qualification in the realm ofpork-barrel omnibuses because political net benefits systematically differfrom their economic counterparts; pork-barrel collections of economicallyinefficient projects, that is, are not political liabilities.

In concluding, we offer three observations which may, though we donot pursue them here, provide some empirically testable consequences. Theprimary culprit in Weingast's UniversalismTheorem and our generalization(Proposition I) is uncertainty. In a pure majority-rule system, MWCs pre-vail (Weingast's Proposition I), but there may be considerable uncertaintyabout which of the many MWCs actually forms. The norm of universalismis a hedge against this type of uncertainty. Uncertainty, however, maymanifest itself in various ways. We have chosen a particularly simple (andperhaps extreme) representation, namely that of maximal uncertainty inwhich no MWC is more likely to form than any other. Like Rawls(1971)weemploya "veil of ignorance" behind which legislators rationally opt for thesecurity of universalism. And, as in Rawls' analysis, this hypothesis maywell be implausible in those circumstances in which prior knowledge aboutgroup structure and coalitional possibilitiesis present.

constraint. Moreover, it renders universalismpreferable to majority rule de-spite the fact that there may be an ex ante anticipation of pork-barrel (read:inefficient) projects. Put differently, inasmuch as c,(x) and c,(x) are oftenthe major cost components of a project, the criterion ensuring a preferencefor universalism over majority rule-b - CJ > O-is astoundingly weak,constituting a considerably weaker requirement than that of efficiency, viz.,b - c, - c, - c) > O.

This last fact, however, underscores the importance of the second ques-tion above. If the condition that renders universalism superior to majorityrule permits inefficient projects as part of the universalistic omnibus, mightthere not be some basis for constitutionally abolishing the pork barrel alto-gether? To investigate this possibility, wecompare two omnibuses:

(i) universalistic omnibusxZ = (x,'~, . . . .xZ);and(ii) the zero omnibus z = (0,0,. . . ,0)

At the individual level, j prefers xZ to z if E,(U) > 0 as given in equation(12). An examination of equations (12)and (13)establishes the following:

PROPOSITION2: E,(U) > 0 if and only if

E,( U) - EAMR) > 0

Therefore, given the peculiaritiesof political cost accounting, a univer-salistic system is preferred by political actors to both majority rule omni-buses and a constitutionally restricted zero-omnibus under the same condi-tion-in the symmetric case, this reduced to b(x7) > CJ(x7).SO long aspork-barrel projects are not "too inefficient," this inefficient arrangementis politicallypopular.

Discussion

One of the limitations of the analyses of Weingast (1979)and Fiorina(1978) is their respective discoveries that there is an ex ante preference foruniversalismOverpure majority rule if project benefits exceedproject costs.One might reasonably ask, however, "Where's the pork barrel?" The porkbarrel consists of inefficient projects-those for which benefits do not ex-ceed costs. Thus, their conclusions do not cover the observation of the uni-versalism norm in such traditional pork-barrel policies as the rivers andharbors omnibus, which often consists of a set of inefficient projects. Ourresults given in Propositions 1 and 2 provide a more general rationale forthe universalism practice in all realms of distributive policy including thepork barrel in several respects.

First, along with Fiorina (1978),we do not require project benefits andcosts to be identical across districts. Rather, associated with each projectP(xj) are its own distinct benefits and costs.

'One referee noted that our result derived rrom a fiscal illusion, namely, Ihe one in which

in-district expenditures are counted as benefits. He noted rurther that similar resulls might wellbe derived rrom related illusions, for example, from underestimating tax share, or rrom nol

laking externalilies inlo account. We agree with his latter point, though we are more uncertainaboUI tagging expenditure-related benefits an "illusion." In fact, expendilures are beneficial ina very real way to their recipients (current input owners).

POLITICALPREFERENCESFORTHE PORKBARREL 109

1 a tightly organized partisan legislature, for example, in which inter-partMWCs are proscribed, a legislator knows (with certainty) that he willbe member of the victorious MWC, and his project authorized, only ifheis a lrtisan of the majority party. Equations (5) and (6)-and hence (7)-are )w inappropriate: if m is the number of members of the majority

part then there are (~) feasible MWCs, (~~:) of which the jth ma-jorillegislator is a member; thus, for j, the probability he is in the vic-tori ISMCW, p~, is zero if he is not a majority partisan and, if he is, Pm =

~:~~ = (n + I)/2m > p-'" equat;on ('). Fo, th, majo,ity P"'y m,,"-berswho organize and control the legislature, majority rule, that is,Ej(N), is much more attractive than in the higher-uncertainty, veil-of-ignance, nonstratified legislature. However, in accord with precisely thesamlogic by which Proposition I was derived, one should expect to ob-serva "distributive tendency" within the majority party. Majority-partyunirsalism, then, is the ex ante expectation. This suggests that highlystra'ied legislatures, e.g., the British House of Commons or the nine-teen-century U.S. House of Representatives, should exhibit a restrictedforrof universalism (namely, majority partism) in comparison to lessstraled legislatures, ceteris paribus. Additionally, the degree of ineffi-den in pork-barrel projects will vary with the degree of stratification (thatis, t' size of the dominant group within which universalismoperates).

)ur idea, then, is that if a legislator knows he is a member of the ma-jori party, and if it is supposed that only majority partisans may be in-cIud in winning legislativecoalitions, then uncertainty over which of hismajity partisans will be included in a winning coalition induces each ma-jori legislator to support majority party universalism. The logic may beextded further. If there is uncertainty over the identity of the majoritypar in the future, then under fairly broad circumstances, career-orientedlegators wiII prefer ful1universalism to universalism restricted to the ma-jori party. Consequently, though the results in the body of this paperigne political parties, the presenceof parties does not undermine the logicof tiversalism.

ndeed, the logic of universalism may undermine parties. Our secondobsvation, then, is that the insurance-likecharacter of the norm of univer-salh may serve to inhibit or delay party discipline or other bases of strati-ficam, on the one hand, and may contribute to their erosion, on the other.Thas, because universalismoffers a different basis for distributing bene-fitshan does partisanship, and because legislators may wish to insure

110 Kenneth A. Shepsle. Barry R. Weingast

against adverse events (e.g., not being in the majority party), their choice ofthe former comes at the expense of the latter. This, of course, will dependon a number of factors, perhaps most prominent of which is the time hori-zon and career perspective of legislators. Thus, the late nineteenth-centurydevelopment of careerism and professionalism in the U.S. House may, incomplex ways, have provided the fertile soil for universalismto flourish andparty government to deteriorate.

Finally, let us underscore a theme elegantly developed by Fiorina(1978): reciprocity across committees and policy jurisdictions is univer-salism as manifested in a structural1ydifferentiated legislature. The basisfor this observation is, once again, the implausibility of our veil-of-ignor-ance argument. While it might appear reasonable to treat as a "stylizedfact" the general and pervasivepopularity of rivers and harbors projects inthe nineteenth century-every district had rivers to be dredged, canals andleveesto be built, or harbors to be developed-it is certainly less true in thetwentieth century. This fact surely has contributed, as Ferejohn (1974)notes, to the efforts by the Corps of Engineers to redefine its mission. Forour purposes, the implication of this development is that districts, their eco-nomic and demographic compositions, are sufficiently diverse and dissimi-lar so as to induce their representatives to seek membership on differentlegislative committees and influence in different policy domains. Reci-procity replaces universalismas legislative"insurance" in this context.

A theory of legislator uncertainty has been employed here to explainsome aspects of legislator preferences for the pork barrel. Because of thepolitical basis for calculating benefits and costs, "pork," in various forms,will always serve as part of the legislator's response to his voters' retrospec-tive question, "What have you done for me lately?" And, consequently,economic inefficiency will likely be a permanent characteristic of the dis-tributive policiesof legislativeinstitutions.

Manuscript submitted 31December1979Final manuscript received8May 1980

REFERENCES

Arnold, R. Douglas. 1979. Congress and the bureaucracy: A theory of influence. New Haven.Conn.: Yale UniversilY Press.

Barry, Brian, 1965, Political argument. London: Routledge and Kegan Paul.Buchanan, James M., and Gordon Tullock. 1962. The calculus of consent. Ann Arbor: Uni-

versity of Michigan Press.

Ferejohn, John A. 1974. Pork barrel politics. Stanford: Stanford University Press.Fiorina, Morris P. 1978. Legislalive facilitation of government growth: Universalism and reci.

procity practices in majority rule institutions. Paper presented al the Conference on theCauses and Consequences of Public Sector Growth, Dorado Beach. Puerlo Rico.

POLITICAL PREFERENCES FOR THE PORK BARREL III

Froman, Lewis. 1967.The congressionalprocess. Boston: lillIe, Brown.Lowi, Theodor~ J. 1964. American business, public policy, case.studies, and political theory.

World Politics, 16: 677-715.-.1979. The end of liberalism , 2nded. New York: Norton.

Maass, Arthur. 1951. Muddy waters. Cambridge, Mass.: Harvard University Press.Manley, John. 1970. The politics of finance. Boston: Little, Brown.Mayhew, David. 1974. Congress: The electoral-connection. New Haven, Conn.: Yale Univer-

sity Press.

Niskanen, William A. 1971. Bureaucracy and representative government. Chicago: Aldine-Atherton.

Plolt, Charles. 1968. Some organizational innuences on urban renewal decisions. AmericanEconomic Review, 58: 306-311.

Rawls, John. 1971. The theory of justice. Cambridge, Mass.: Harvard University Press.Rundquist, Barry S. 1973. Congressional innuences on the distribution of prime military con.

tracts. Unpublished Ph.D. dissertation, Stanford University.Schallschneider, E. E. 1935. Politics, pressures and the tariff. Englewood Cliffs, N.J.:

Prentice-Hall.

Stockman, David. 1975. The social pork barrel. Public Interest, 39: 3-30.Weingast, Barry R. 1979. A rational choice perspective on congressional norms. American

Journal of Political Science, 23: 245-263.

-, Kenneth A. Shepsle, and Christopher Johnsen. Forthcoming. The political economyof benefits and costs: A neoclassical approach to distributive politics. Journal of PoliticalEconomy.