Comparing Tax Expenditures and Direct Subsidies: The Role of Legislative Committee Structure*

Transcription

1 Comparng Tax Expendtures and Drect Subsdes: The Role of Legslatve Commttee Structure* Dhammka Dharmapala Economcs Program Research School of Socal Scences Australan Natonal Unversty Canberra ACT 0200 Australa Fnal verson: September 1998 Forthcomng n Journal of Publc Economcs Abstract Ths paper analyzes the mpact of legslatve commttee structure on polcy outcomes, comparng a tax commttee (enactng tax expendtures) and a decentralzed system of specalzed commttees (undertakng drect spendng). An endogenous commodty taxaton framework s combned wth models of legslatve barganng and commttee votng. The man concluson s that the tax commttee structure gves rse to lower subsdy levels under a wde range of crcumstances. However, t s only under a more restrctve set of assumptons that socal welfare s hgher n the tax commttee regme. Ths theoretcal analyss s llustrated by two examples of nsttutonal change n Congress. Keywords: Tax Expendtures; Subsdes; Commttees; Interest Groups; Legslatve Barganng JEL Classfcaton: D72; H20 * Ths paper s based on materal n my Ph.D dssertaton at the Unversty of Calforna at Berkeley. I would lke to thank Alan Auerbach for nvaluable suggestons and comments, and Aaron Edln, Robert Innes, Krsten Landeryou, John Qugley, Todd Sandler, Eleen Segarra, semnar partcpants at varous nsttutons, two anonymous referees and the Edtor for helpful comments. Avnash Dxt kndly provded access to unpublshed materal. All remanng errors are, of course, my own. Fnancal support from the Alfred P. Sloan Foundaton (through a Doctoral Dssertaton Fellowshp) and the Robert D. Burch Center for Tax Polcy and Publc Fnance (at the Unversty of Calforna at Berkeley) s gratefully acknowledged.

2 1) Introducton Economsts have tradtonally approached the analyss of tax-related ssues from a normatve perspectve, such as that employed n the optmal taxaton lterature. However, n recent years, a growng body of research (exemplfed by Grossman and Helpman (1994), Dxt (1995; 1996a, b) and Dxt, Grossman and Helpman (1997)) has sought to endogenze the determnaton of economc polcy by explctly modelng the relevant poltcal processes. Ths paper apples ths postve approach to the comparatve analyss of tax expendtures 1 and drect government subsdes. The am s to examne the mpact of dfferent types of legslatve commttee structures (n partcular, a tax commttee enactng tax expendtures, as opposed to specalzed commttees undertakng drect spendng) 2 on polcy outcomes. Ths queston s nspred by a recent debate among legal scholars (e.g. Zelnsky, 1993) concernng the relatve merts of tax expendtures and drect spendng as means of provdng government assstance for favored actvtes. The central am of ths paper s to characterze the polcy outcomes whch arse under the tax commttee and specalzed commttee structures, and to evaluate the welfare consequences of each regme. In so dong, the model of endogenous commodty taxaton developed by Dxt (1995) (and nspred by Grossman and Helpman (1994)) s used as the underlyng framework. Ths s combned wth a legslatve barganng approach (n the tradton of Baron and Ferejohn (1989)) to the modelng of each commttee s nternal decsonmakng processes. Ths explctly takes nto account the strategc nteracton among legslators wthn commttees, and enables the polcy outcomes to be derved as equlbra of clearly specfed commttee votng games. Three alternatve characterzatons the agenda setter domnated, majortaran, and unversalstc models of nternal commttee decsonmakng are used, to ensure the robustness of the results. In addton, dfferent patterns of lobbyng wthn the specalzed commttee structure, arsng from dfferent assumptons about the transacton costs facng lobbysts, are examned. The man result of ths analyss s that the tax expendtures determned by a tax commttee wll, under a wde varety of crcumstances, be smaller than the correspondng drect subsdes enacted by a system of specalzed commttees. However, t s then shown 1 Pechman (1987, p. 355) defnes tax expendtures as... specal provsons of the ncome tax laws that reduce the tax lablty of those who make payments or receve ncomes n certan desgnated forms. The noton was developed by Surrey (1970) as a conceptual tool for understandng such tax provsons. See Pechman s (1987) Appendx C or Surrey and McDanel (1985) for detaled dscussons. 2 Examples of tax commttees n the US Congress are the House Ways and Means Commttee and the Senate Fnance Commttee, whle the Agrculture commttees of both the House and the Senate are examples of specalzed commttees. 1

3 that ths does not necessarly mply that socal welfare s hgher wth a tax commttee structure; the theoretcal characterzaton of the (more restrctve) set of condtons requred for a hgher welfare level s renforced wth a numercal llustraton. The major contrbuton of ths paper s to establsh a framework for comparng tax expendtures and drect subsdes on procedural grounds. In so dong, t reformulates common agency models of poltcs to address a queston whch has not htherto been nvestgated formally. Moreover, t s one of relatvely few studes to explctly combne specal nterest lobbyng wth models of legslatve barganng. 3 Fnally, the paper s theoretcal fndngs are used to llumnate two hstorcal epsodes nvolvng nsttutonal changes n the US Congress. The frst of these relates to aspects of the hstorcal development of tax expendtures (wth partcular reference to the elmnaton of many such provsons by Congress as part of the Tax Reform Act of 1986). The second concerns reforms to the House Ways and Means Commttee whch were nsttuted n In each case, t s argued that the theoretcal nsghts provded by ths paper explan observatons whch prevously seemed puzzlng. The paper begns wth a revew of the relevant lterature n Secton 2. Secton 3 brefly presents the basc model of endogenous commodty taxaton, whle Secton 4 ntroduces and dscusses the two alternatve commttee structures. The polcy outcomes under varous assumptons about commttee decsonmakng and lobbyng patterns are derved n Secton 5. The welfare comparson s undertaken n Secton 6, and the hstorcal applcatons are addressed n Secton 7. Secton 8 concludes the paper. 2) Revew of the Lterature Although there are few studes of tax expendtures by economsts 4, legal scholars have wrtten extensvely on the topc. 5 The school of tax expendture analyss whch has been developed by Surrey (1970) and hs assocates s characterzed, at least mplctly, by a publc nterest model of poltcs, n whch legslators pursue the general welfare to the best of ther abltes. Scholars n ths tradton (e. g. Surrey and McDanel, 1985) argue that drect spendng programs are superor to tax expendtures for delverng government assstance to 3 Persson (1998) also does so, though n a somewhat dfferent context. 4 A major excepton s Feldsten (1980), whch consders the crcumstances under whch a government subsdy (n the form of a tax preference) to prvate chartable donatons s more effcent than drect government provson. 5 Tax expendtures have also receved some attenton from poltcal scentsts and publc admnstraton scholars. See Howard (1995; 1997) for a dscusson of the ssue from ths perspectve. 2

4 favored areas because of the greater expertse (relatve to the tax commttees) of the specalzed Congressonal commttees whch are responsble for spendng programs. In contrast, Zelnsky (1993) employs a publc choce (or nterest-group) approach. He contends that the specalzed commttees are more vulnerable to capture by specal nterests than s the tax commttee, because the latter faces a greater varety and number of competng consttuences. The essence of hs thess s that... because tax nsttutons are subject to more numerous and dverse consttuences than the specalzed, lmted-clentele organzatons that desgn and mplement drect government spendng, they... are less susceptble to nterest group capture... than ther drect expendture equvalents (p. 1166). The theoretcal approach whch seems most approprate n formalzng Zelnsky s hypothess s a common agency framework wth multple prncpals (e.g. Bernhem and Whnston, 1986a), n whch lobbysts are conceptualzed as prncpals and legslators as agents. Common agency models have ncreasngly been employed n recent years by economsts seekng to explan economc polcy outcomes generated by poltcal processes nvolvng specal nterest lobbyng. For example, Grossman and Helpman (1994) characterze the poltcal equlbrum whch results from specal nterest lobbyng for protecton. In ther model, nterest groups (defned by ther sector-specfc factor endowments) offer contrbutons, condtoned on polcy outcomes, to the government (whch seeks to maxmze a weghted sum of total campagn contrbutons and socal welfare). The equlbrum polces nvolve the organzed groups drvng a wedge between the world prces for the goods whch they produce and the producer prces whch they receve, at the expense of unorganzed sectors. Dxt (1995, 1996b) generalzes the Grossman-Helpman model n order to analyze the endogenous determnaton of commodty tax rates. 6 3) The Basc Model The basc model on whch the analyss n ths paper s based s a smple lnearquadratc formulaton of the Grossman-Helpman and Dxt framework. Ths s characterzed brefly here, as more detaled formulatons are to be found n Grossman and Helpman (1994), Dxt (1995, 1996b) and Dxt et al. (1997). A small open economy s assumed, wth n sectors, each requrng a specfc factor of producton. There are N ndvduals (each endowed wth the th specfc factor) n the th group and N ndvduals n the economy as a whole. Each specfc factor defnes a latent nterest group; an exogenous subset (denoted L) 6 Dxt, Grossman and Helpman (1997) develop an extremely general formulaton of ths approach. 3

5 consstng of k of these are organzed nto lobby groups. The th good has a world prce p *; the domestc producer prce (receved by producers) s p and the domestc consumer prce (pad by consumers) s q. The vector of producer prces s denoted by p, and that of consumer prces by q. A quadratc proft functon π( p ) s postulated, yeldng a lnear supply functon for each good: y( p) = gp f (1) Grossman and Helpman (1994) assume unform consumer tastes, whereas Dxt (1995) ntroduces dfferent preferences for members of dfferent groups. For the purposes of ths paper, the former postulate s adopted. 7 Thus, each ndvdual s assumed to have a (common) quadratc consumer surplus functon s(q ) and a lnear demand functon: d ( q) = a b q (2) for the th good (note that cross-prce effects are assumed to be zero n both demand and supply). The market demand for ths good s: where A D = A Bq (3) = Na and B = Nb. The ntegral of ths expresson gves the total consumer surplus S(q ) generated by the consumpton of good. Summng these over all yelds the total consumer surplus S(q) n the economy. The government determnes the vectors of domestc producer and consumer prces p and q. For nstance, f p - p * < 0, t s mposng a producton tax on the th good. Budget balance s acheved through per-head lump-sum transfers, determned by the net revenue of the government: 1 r ( p, q) = [( q p*) D ( q ) ( p p*) y ( p )] G (4) N where G s some exogenous revenue requrement. In makng ts decson, the government s nfluenced by the lobbyng actvtes of the organzed groups. The lobbyng process s modeled as follows: each group offers a schedule of campagn contrbutons, contngent on polcy outcomes. Thus, the th group offers a schedule C (p, q). The welfare of ndvduals n the th group (gnorng ther campagn contrbutons, f any) s: W ( p, q) = N s( q) + π ( p ) + N r( p, q) (5) 7 The assumpton of unform consumpton tastes mples that no sector wll lobby to change any element of q from ts world prce. Thus, for the entre analyss below, the decsonmakng process focuses solely on p. However, the more general formulaton s presented n Eqs. (4)-(7) for the sake of completeness. 4

6 where s(q) s each ndvdual s total consumer surplus from all goods. Socal welfare s gven by the expresson: The government s objectve functon s assumed to be: W ( p, q) = S( q) + π ( p ) Nr( p, q) (6) + θw( pq, ) + ( pq, ) (7) where θ s the weght placed on socal welfare (relatve to campagn contrbutons). An objectve functon of ths form can be regarded as resultng from a desre to maxmze contrbutons, subject to a reelecton constrant. 8 It should be emphaszed that ths framework, despte the appearance of beng desgned to address nternatonal trade ssues, serves as a general model of endogenous commodty taxaton. It s also mportant to note that, n ths model, taxes and subsdes are purely redstrbutve n nature, rather than servng to correct market falures. It s possble to construct a model n whch the socally optmal taxes or subsdes are nonzero for some sectors (due, for nstance, to externaltes). Ths, however, would complcate the analyss wthout alterng the basc thrust of the conclusons. C 4) The Tax Commttee and Specalzed Commttee Structures In the model outlned n Secton 3, the government s treated as a sngle unfed decson-makng entty. Here, ths framework s adapted to analyze decsonmakng by legslatve commttees. The fundamental am s to compare two possble nsttutonal structures through whch a legslature may choose taxes and/or subsdes for each of the economy s n sectors. The frst of these nvolves empowerng a commttee to decde smultaneously on the polces relatng to all n sectors. As ths unfed approach resembles the process by whch the tax-wrtng commttees of Congress (such as House Ways and Means) determne the tax laws, t proves convenent to characterze t as the tax commttee (hereafter, TC) approach, and to regard the subsdes granted to favored sectors by the unfed commttee as tax expendtures. The alternatve structure nvolves establshng n separate specalzed commttees, each wth jursdcton over one sector of the economy; n ths specalzed commttee (hereafter, SC) structure, each commttee chooses a tax or subsdy for ts sector n solaton from the other commttees. 8 See Grossman and Helpman (1996) for an account of the mcrofoundatons of ths objectve functon. 5

7 Thus, the two nsttutonal regmes dffer n terms of the breadth of the jursdcton exercsed by commttees. The TC has broad jursdcton (encompassng the entre economy), n that t can decde the entre vector p of producer prces. Ths entals that t can mpose (dstortonary) taxes on the unorganzed sectors of the economy, as well as offerng (dstortonary) subsdes to the organzed sectors. In contrast, each SC can only determne the polcy pertanng to one sector of the economy, so that SC chooses only p. It follows that each specalzed commttee s precluded from mposng dstortonary producer taxes on sectors other than that over whch t exercses jursdcton. Ths nsttutonal feature proves to be of consderable sgnfcance n the welfare comparson between the two regmes, undertaken n Secton 6; the ratonale for the assumpton s dscussed n more detal there. The TC s choce of taxes and subsdes satsfes a balanced budget condton; f revenue from the (dstortonary) taxes does not equal the amount spent on subsdes, the lump-sum per-head taxes or transfers gven by Eq. (4) result n budget balance. However, as the SC regme nvolves decentralzed decson-makng, each SC s choce wll not, n general, satsfy budget balance. Moreover, n Congress, specalzed commttees only has the power to authorze expendtures, and these authorzed expendtures must be approprated before they can be undertaken. Thus, t s necessary to ntroduce an Appropratons Commttee wthn the SC structure. The assumpton adopted here s that the authorzatons determned by the specalzed commttees are smply ratfed by the Appropratons Commttee, whch acheves budget balance through the per-head lump-sum taxes (gven by Eq. (4)) whch fnance these appropratons. Whle ths assumpton s undoubtedly rather extreme, more complex models of barganng between the Appropratons Commttee and each of the specalzed commttees lead to qualtatvely smlar results (albet wth the dfferences n outcomes between the two structures beng somewhat attenuated). The two commttee structures also ental substantally dfferent lobbyng opportuntes for the organzed nterest groups. Groups lobbyng the TC can condton ther campagn contrbutons on the entre vector p, rewardng the legslators for choosng a hgher subsdy for tself, and also nducng legslators to support hgher taxes on other sectors. Each organzed group pursues the second objectve because the TC can fnance the subsdes t enacts ether by unform lump-sum taxes or by dstortonary taxes on unorganzed groups; the organzed groups, of course, favor the latter, as t results n more redstrbuton. On the other hand, a group lobbyng a partcular SC can only condton ts contrbutons on the (scalar) choce varable of that commttee, and any subsdy granted by a SC can only be fnanced by unform lump-sum taxes. 6

8 Gven ths formulaton of the SC structure, the queston then arses of whch groups lobby whch of the specalzed commttees. Consder group, where L; clearly, group has a drect and obvous nterest n the decson made by SC. However, group s also harmed (through hgher taxes) by subsdes gven to sector j by SC j. In these crcumstances, f ts lobbyng opportuntes were unrestrcted, would group lobby SC j? The earler lterature (e.g. Zelnsky, 1993) has tended to assume that the pattern of lobbyng n the SC regme would nvolve group only lobbyng SC. 9 A detaled analyss of ths ssue s undertaken after the nternal decsonmakng processes of the commttees are presented. Secton (5.1.2) below consders the ssue of whether there s any dfference n polcy outcomes between TC and SC structures when all organzed groups lobby each of the specalzed commttees. Proposton 2 suggests that, under the assumptons of ths model, there s no dfference. Ths mples that, whle the earler lterature has taken such a dfference as beng self-evdent, ther dervaton requres that further restrctons be placed on the nsttutonal structure. However, t s not beng clamed that the equvalence result n Proposton 2 s realstc ; rather, t s an extreme benchmark whch llustrates the strngent condtons under whch polcy outcomes are nvarant to commttee structure. As dscussed more fully n Secton (5.1.2), the relatonshp between lobbysts and legslators, n practce, s one requrng mutual trust and credblty. Lobbysts must therefore nvest substantally n establshng ther credblty wth legslators. The costs of dong so, along wth other transacton costs, are lkely to be ncreasng n the number of commttees lobbed. The equvalence result apples only to the specal case n whch these transacton costs are zero or constant. 10 Thus, Secton (5.1.3) also consders the opposte case, n whch each nterest group only lobbes ts commttee. The two extreme cases (n Sectons (5.1.2) and (5.1.3)) defne the bounds wthn whch the polcy outcome under the SC regme les; n general, t s shown that ths outcome wll dffer from that to whch the TC structure gves rse. Fnally, t s mplctly assumed that the legslatve commttees examned here are ndeed the relevant decson-makng enttes wthn Congress. In partcular, the decsons made n commttee are not overturned on the floor of Congress. There are at least two possble justfcatons n the lterature for such an assumpton. The frst s that a unversalstc norm of deference prevals among dfferent Congressonal commttees, for 9 Persson (1998, p. 319) makes a smlar assumpton, argung that the bulk of campagn contrbutons n the Unted States goes to representatves from the same dstrct as the donor, or to a member of the commttee holdng jursdcton of regulaton or grants that apply to the donor group. 10 In other crcumstances, a free rder problem among groups may emerge, as a group cannot capture the full benefts of lobbyng a commttee other than that exercsng jursdcton over ts sector see Secton (5.1.2). 7

9 the reasons outlned n Wengast (1979), and Wengast, Shepsle and Johnsen (1981). 11 The other nvolves the theory of commttee specalzaton developed by Gllgan and Krehbel (1987); n ther model, the floor of Congress precommts to the passage of blls reported out by commttees n order to provde commttee members wth an ncentve to specalze n gatherng nformaton about ther area of jursdcton. 5) Legslatve Barganng and Polcy Outcomes Ths paper focuses prmarly on comparng polcy outcomes n the TC and SC regmes. In addressng ths ssue, the nternal processes by whch these commttees reach decsons are, of course, crucal. A number of dfferent models of votng wthn legslatures and legslatve commttees exst n the lterature. The most common approach to the analyss of decsonmakng wthn Congress draws on the legslatve barganng framework of Baron and Ferejohn (1989). Typcally, such models of legslatve barganng take legslators preferences as exogenously gven. A recent excepton, however, s Persson s (1998) model, whch explctly ncorporates specal nterest lobbyng, so that legslators engage n the barganng process to pursue objectves determned, n part, by lobbyng actvtes. The lterature on legslatve barganng has tended to fnd that outcomes are hghly senstve to the precse nsttutonal assumptons adopted (see, for nstance, the dscusson n Baron and Ferejohn (1989, p. 1182)). Thus, ths paper uses three dfferent stylzed characterzatons of the commttee decsonmakng process to ensure the robustness of the man results concernng polcy outcomes. These wll be referred to as the agenda setter domnated (A), majortaran (M) and unversalstc (U) models. There are two man dmensons along whch these models of a legslatve commttee s nsttutonal structure can be characterzed - the recognton rule (whch specfes whch legslators have the rght to make proposals) and the amendment rule (whch determnes whether amendments can be offered once a proposal s made). In the A model, one member of the commttee s exogenously chosen to be the agenda setter. The recognton rule s that only the agenda setter has the power to make a proposal, and the amendment rule s closed (so that no other member can offer an amendment). The M model s characterzed by both an open recognton rule and an open amendment rule. Votng takes place sequentally on each decson before the commttee, and the outcomes are determned by majorty rule. The U model s characterzed by a norm of deference (Wengast, 1979; Wengast et al., 1981). The 11 Ths noton should not be confused wth the dea of unversalsm wthn a commttee, as dscussed below. 8

10 recognton rule s that each legslator enjoys exclusve proposal power over a subset of the commttee s sphere of jursdcton; once these proposals are made, votng takes place under a closed rule. The am of ths paper s to focus on the effects of lobbyng on legslators behavor (gven partcular nsttutonal structures). Thus, the models presented below abstract from dfferences n legslatve dstrcts; all legslators are assumed to be ex ante dentcal (except nsofar as one s exogenously chosen to be the agenda setter), and face the same reelecton consttuency. It would be relatvely straghtforward to ntroduce dfferences among dstrcts, but ths would dstract attenton from the man ssue. Moreover, assumng unform dstrcts bases the results aganst fndng any dfference n polcy outcomes (such as those whch emerge from the analyss below), n the sense that there s no purely electoral ncentve for legslators to favor sectors whch are dsproportonately represented n ther dstrcts. 5.1) Polcy Outcomes wth Agenda Setter-domnated Commttees As noted above, the man features of the A model are a recognton rule whch vests exclusve proposal power n an agenda setter (say, the charman of the commttee), and a closed rule for votng. The commttee conssts of l members (where t s assumed throughout that l 5), each of whom seeks to maxmze a weghted sum of campagn recepts and socal welfare. It s mportant to note that the argument of socal welfare s the p that s actually enacted, not each legslator s preferred p ) The Tax Commttee Consder a tax commttee, as descrbed n Secton 4. Suppose that ths commttee operates n accordance wth the A model, so that one member of the commttee s exogenously chosen to be the agenda setter (hereafter, A; the tax commttee wll also be referred to as agenda setter domnated (A)). The vector of producer prces enacted by the A tax commttee s denoted by p AT, and the proposal made by the agenda setter by p ATP. The polcy outcome s determned by a game nvolvng both lobbyng and legslatve barganng. The tmng of the game, as depcted n Fg. 1 below, s as follows. In the frst stage, the organzed nterest groups lobby A by offerng contrbuton schedules condtoned on p AT. As A s the only legslator wth proposal power, t s never worthwhle for any nterest group to drectly lobby any of the other commttee members. In the second stage, A chooses a vector of producton prces p (.e. p ATP ) to propose to the commttee; ths vector s chosen to maxmze the objectve functon: AT AT θ W ( p ) + ( p ) (8) C L 9

11 In the thrd stage, A engages n barganng wth the other (l - 1) members of the commttee (who wll be referred to as the junor members). Suppose that ths barganng takes the form of a take-t-or-leave-t offer by A to make a transfer to each junor legslator s campagn fund, condtonal on the legslator s vote n the fnal stage of the game. 12 Each junor legslator j has an objectve functon of the form T j + θw(p AT ) (9) where T j s the transfer promsed by A. Suppose that the transfers T j are nfntesmally small (.e. ε), 13 and assume further that A s constraned to treat the junor members symmetrcally. 14 Note that the assumpton that ths barganng s between A and each junor member s made for convenence; the results are unchanged f t s assumed that t s the nterest groups whch offer transfers to the junor members (as utlty s transferable between A and the groups). In the fnal stage, the commttee votes on p ATP under a closed rule, wth the vector of (undstorted) world prces p* beng the default alternatve. To solve for the outcome of ths game, consder the fnal stage. All legslators other than A are dentcal (n that each faces the same objectve functon, by assumpton), so any votng equlbrum among these (l 1) legslators must be symmetrc. It can be shown that: Proposton 1: The unque (symmetrc) equlbrum of the commttee votng game s for all l commttee members to vote for p ATP. It follows that p AT = p ATP Proof: By constructon, p ATP s A s preferred outcome. Consder the (l 1) junor commttee members. There are two possble symmetrc votng patterns for these legslators: All vote aganst p ATP. Consder the j th legslator. If all (l 2) other junor members vote aganst p ATP, then (for l 5), the outcome wll be p*. Thus, j s payoff from votng for p AT s ε +θw(p*), whch s greater than the payoff θw(p*) from votng aganst p ATP. Thus, ths s not an equlbrum. All vote for p ATP. Consder the j th legslator f all (l 2) other junor members vote for p ATP, then (for l 5), the outcome wll be p ATP. Thus, j s payoff from votng for p ATP s ε +θw(p ATP ), whch s greater than the payoff θw(p ATP ) from votng aganst p AT. Hence, ths s an equlbrum. Thus, the outcome whch emerges from the commttee s vote s that p AT s enacted. Gven ths result, A s problem n the second stage s: 12 More generally, ths transfer could represent any valuable poltcal servce, such as permttng the junor member to ntroduce a bll favorng her dstrct. 13 Thus, the transfers do not sgnfcantly affect the jont welfare of A and the nterest groups. 14 Ths s purely for ease of exposton, and s not crucal to the results; Dharmapala (1998) obtans smlar outcomes wth A engagng n votebuyng n order to buld a majorty coalton. 10

12 ATP ATP Max θ W ( p ) + C ( p ) (8) ATP p as A antcpates that the proposal wll be enacted. The soluton s closely analogous to the government s polcy choce n Dxt (1995); however, t s derved here as the equlbrum of a commttee votng game, rather than as the choce of a unfed governmental entty. A chooses p ATP to maxmze the expresson n (8) n the second stage, gven the contrbuton schedules C offered by the organzed nterest groups. In the frst stage, the groups offer contrbuton schedules to A, playng a Nash equlbrum among themselves. In analyzng ths game, Bernhem and Whnston (1986b), Grossman and Helpman (1994) and Dxt (1995) restrct attenton to truthful strateges, n whch each group s contrbuton functon C s equal to ts beneft functon (up to an addtve constant). Equlbra n such truthful strateges have the attractve propertes of coalton-proofness and Pareto-effcency (n the sense of maxmzng the jont surplus between A and the organzed groups). When the groups use truthful strateges, the soluton to the game s equvalent to that whch arses when each group seeks to maxmze the blateral surplus between t and A. Followng the dervaton n Dxt (1995), (8) s equvalent to the problem: 15 L Max θ W ( p) + ( p) (10) p W L Substtutng n the lnear-quadratc functonal forms n Eqs. (1)-(4), the FOCs are: θ[π (p ) y (p ) (p p *)g ] + π (p ) α L [(p p *)g I + y (p )] = 0 L where α L θ[π (p ) y (p ) (p p *)g ] α L [(p p *)g I + y (p )] = 0 L (11a) (11b) N / N s the proporton of the populaton whch s organzed n all lobby L groups. Rearrangng Eqs. (11a) and (11b), the producton tax or subsdy for group can be expressed: p AT p I α g p f L * = (12) θ + α L g where I s an ndcator varable equal to one f the th group s organzed (.e. f zero otherwse. Ths s essentally dentcal to Dxt s (1995) Eq. (19). L) and 5.1.2) The Specalzed Commttee Structure wth Unrestrcted Lobbyng Now consder the specalzed commttee structure, n whch the legslature s dvded nto n separate commttees. As dscussed above n Secton 4, ths rases the ssue of whch 15 An alternatve approach (adopted by Goldberg and Magg (1997)) whch yelds the same soluton s to vew the nteracton between the groups and the government as a Nash barganng game. As any soluton must maxmze the jont surplus, ths leads drectly to the program n Eq. (10). 11

13 groups lobby whch commttees. Suppose, ntally, that all organzed groups smultaneously lobby all the specalzed commttees (and that dong so does not nvolve any costs for the groups n terms, for example, of the credblty of ther promses). The tmng of ths game wthn each SC, as shown n Fg. 2, s as follows. In the frst stage, all organzed groups lobby commttee s agenda setter, A. Then, A chooses a proposal p ASP. In the thrd stage, A offers transfers to the junor members Fnally, the commttee votes on the proposal under a closed rule. Each commttee takes the decsons of all the others as gven. Let p ASL be the th producer prce enacted when all organzed groups lobby SC. Then, the outcome n the fnal stage s: Proposton 1 : The unque (symmetrc) equlbrum of the commttee votng game s for all l commttee members to vote for p ASP. It follows that p ASL = p ASP Proof: Analogous to that of Proposton 1. Solvng through the earler stages of the game, the followng equvalence result can be establshed: Proposton 2: Gven the assumptons of the model, p AT = p ASL. ASLP Proof: Consder A s problem (.e. choosng p subject to lobbyng by all organzed groups). Ths choce must maxmze the jont surplus: Max θw ( p ) + W ( p ) p The FOCs are: θ[π (p ) y (p ) (p p *)g ] + π (p ) α L [(p p *)g + y (p )] = 0 L L θ[π (p ) y (p ) (p p *)g ] α L [(p p *)g I + y (p )] = 0 But, ths s dentcal to Eqs. (11a) and (11b) above. Thus, p AT = p ASL Moreover, by Proposton 1, p ASL = p ASP L Thus, a SC regme n whch lobbyng s unrestrcted leads to dentcal polcy outcomes to those arsng from the TC structure. One mplcaton of ths equvalence result s that the assumpton n the earler lterature that the two structures would automatcally lead to dfferent outcomes must be qualfed; t would seem, nstead, that some nsttutonal constrants on the process of contractng between lobby groups and legslators are requred n order to generate dfferng polcy outcomes. As noted n Secton 4, however, t s not beng argued here that Proposton 2 s lkely to descrbe the operaton of real-world legslatures, nor that t represents a robust result. Rather, t consttutes a benchmark, llustratng the restrctve condtons under whch polcy outcomes are nvarant to commttee structure. 12

14 For nstance, Proposton 2 requres the absence of cross-prce effects. 16 The most sgnfcant qualfcaton to Proposton 2, however, may relate to the transacton costs of lobbyng, and, n partcular, to the ssue of credblty. Recall that the frst stage of the commttee votng game nvolves the nterest groups makng promses to donate campagn contrbutons, contngent on the enacted polcy outcomes. Clearly, these promses must be credble for legslators to respond to the ncentves that they embody. 17 Thus, the SC regme wth unrestrcted lobbyng nvolves the mplct assumpton that an nterest group can develop a relatonshp of complete credblty wth all n commttees, whle lobbyng each of them smultaneously. One mght magne a credblty functon c(n ) for each nterest group, expressng legslators belefs concernng whether the group wll delver on ts promsed contrbutons, where n s the number of commttees lobbed by the group. Then, the polcy nvarance result n Proposton 2 holds only f c(n ) = 1 when n = n. Introducng a small probablty that a group s credblty s adversely affected by lobbyng all the commttees s suffcent to overturn the result n Proposton 2. In practce, as argued n Secton 4 above, lobbysts must nvest substantally n establshng ther credblty wth legslators. These transacton costs, whch are lkely to be ncreasng n the number of commttees lobbed, wll be balanced aganst the expected benefts from lobbyng, and, n general, groups can be expected to reach an nteror optmum n whch they choose to lobby some, but not all, commttees. Moreover, ntroducng transacton costs whch rse wth the number of commttees lobbed gves rse to a possble free rder problem among nterest groups. 18 A subsdy granted to sector j by SC j harms all sectors j through ncreased taxes. If group lobbes SC j aganst the subsdy to the j th sector, group wll only enjoy a small fracton of the socal beneft resultng from a reduced subsdy to sector j. However, group wll ncur the addtonal costs (possbly n the form of reduced credblty wth ts own commttee, SC ) of lobbyng SC j. Clearly, groups wll generally fnd t worthwhle to lobby the SC wth jursdcton over ther own sector, as the benefts of recevng subsdes are large. However, the free rder problem dscussed here wll tend to dscourage groups from lobbyng 16 Thus, a general nonequvalence result s obtaned n Dharmapala (1998), where nteracton effects among the polcy varables, operatng through an electoral support functon, are assumed. 17 In practce, reputaton effects may partally mtgate these problems. However, t may be argued that reputatonal consderatons may actually be more salent n a one-on-one lobbyng framework. 18 Ths should not be confused wth the free rder problems nvolved n ndvduals wthn a sector organzng nto a lobby group; n ths framework, t s assumed that certan groups have exogenously formed. What s meant here s that, gven that they have formed, the groups themselves wll nonetheless experence a free rder problem n lobbyng multple commttees. 13

15 commttees other than ther own. For example, agrcultural subsdes clearly harm all nonagrcultural sectors of the economy. Nonetheless, t may be that a partcular nonagrcultural lobby group (such as a bankng lobby), taken n solaton, has no ncentve to lobby the Agrculture commttee to reduce these subsdes, f dong so entals that t wll be less effectve n lobbyng the Bankng commttee. Thus, the followng secton consders the polar opposte case to that of unrestrcted lobbyng namely, a pattern of lobbyng n whch each organzed group lobbes only ts own commttee. However, t should be borne n mnd that ntermedate cases are more lkely to be realstc ) The Specalzed Commttee Structure wth Restrcted Lobbyng The restrcted pattern of lobbyng consdered n ths secton can be conceptualzed as resultng from a contractng restrcton mposed on the lobbyng actvtes of nterest groups. Specfcally, ths nsttutonal structure resembles the TC structure wth the restrcton that each organzed group can only condton ts contrbuton functons on the tax or subsdy t receves. 19 There may exst a ratonale for mposng such a restrcton; t may be that each group s assumed to have greater expertse regardng ts own sector than wth regard to other sectors, so that more heed s pad to s lobbyng regardng sector than regardng other sectors. Alternatvely, such a pattern may arse endogenously f there are costs of lobbyng commttees other than one s own (as n the credblty story above). In ths case, ths pattern represents the polar opposte to that n Secton (5.1.2). The smplest vsualzaton of ths lobbyng pattern may be to magne the commttees as meetng smultaneously n separate commttee rooms; each lobbyst can thus only be n one commttee room. The tmng of the game s depcted n Fg. 3. The th SC s lobbed by th group f that group s organzed; otherwse, t s not lobbed at all. In other respects, the game s smlar to that represented n Fg. 2. The notaton s also smlar: p AS s the th producer prce authorzed by the commttee, and p ASP s the agenda setter s proposal. Solvng for the votng equlbrum, the result s analogous to that n Proposton 1. Proposton 1 : The unque (symmetrc) equlbrum of the commttee votng game s for all l commttee members to vote for p ASP. It follows that p AS = p ASP Proof: Analogous to that of Proposton 1. Thus, the agenda setter s proposal s enacted. In the second stage, two cases can be dstngushed. If L, the agenda setter s problem s equvalent to choosng p ASP drectly to maxmze the blateral surplus between the lobbyst and the agenda setter: 19 There s an analogy, though n reverse, to Dxt s (1997, p. 381) recommendaton that a smlar contractng restrcton be mposed n order to provde stronger ncentves wthn bureaucraces. 14

16 ASP ASP Max θ W ( p ) + W ( p ) (13) pasp Through a dervaton whch parallels that of Eq. (12), ths yelds the outcome: p AS 1 ( N / N ) g p f p* = θ + ( N / N ) g (14a) If L, then A does not face any lobbyng, and smply maxmzes socal welfare, yeldng the proposal p *, whch s dentcal to the default alternatve. 20 Clearly, the votng outcome must be p AS = p * (14b) 5.1.4) Comparng the Outcomes Ths secton focuses on comparng the outcomes p AT and p AS. Note that the fcus here s on the SC outcome wth restrcted lobbyng. As suggested earler, actual SC outcomes are bounded by p ASL (whch s dentcal to p AT by Proposton 2) and p AS AS. Showng that p dffers systematcally from p AT, then, establshes that polcy outcomes under the TC and SC regmes wll dffer, except n the lmtng case of completely unrestrcted lobbyng. In undertakng the comparson, t proves helpful to make two mld assumptons, and derve Lemma 1, whch proves useful, both here and later on n the analyss. The frst assumpton s: A1: card L 2 Ths smply ensures that at least 2 groups are organzed. The second assumpton s: A2: y ( p *) > 0 Ths requres that some postve quantty of each good would be forthcomng at the world prce, and rules out the possblty that some sectors are brought nto exstence solely by the prospect of recevng subsdes. The expressons for the th tax or subsdy (Eqs. (12) and (14a)) can be expressed n general form as p p I γ g where γ s an arbtrary parameter. Then, the followng result holds: Lemma 1: Gven A2, p f j * = (15) θ + γ g Proof: Rearrangng Eq. (15): dp dγ < 0 20 As there are no cross-prce effects, A cannot rase socal welfare by dstortng p n such a way as to mtgate the dstortons created by those commttees whch are subject to lobbyng. 15

17 j θ + γ 1 γ f p = p * θ + 2γ 1 θ + 2γ 1 g Dfferentatng wth respect to γ, A2 => p * > (f /g ) j dp 1+ θ = dγ ( θ + 2γ 1) 2 p f * g dp < 0 dγ γ That s, p s monotoncally decreasng n γ. Ths result makes the comparson of the TC and SC outcomes (wth restrcted lobbyng n the latter case) extremely straghtforward: Proposton 3: Gven A1 and A2, AT p < AS p Proof: Consder the case where L: A1 N / N < α L AT AS Lemma 1 p < p L Consder the case n whch L: AT AS Eq. (12) p < p *, whle p = p * from Eq. (14b) AT AS p < p L AT AS p < p Thus, the TC structure leads to lower subsdes than does the SC regme (bearng n mnd the caveat above concernng the restrcton on lobbyng). 5.2) Polcy Outcomes wth Majortaran Commttees As noted earler, there are a wde varety of models of legslatve decsonmakng n the exstng lterature. Thus, ths secton, and Secton (5.3) below, examne the robustness of the result n Proposton 3 by analyzng alternatve specfcatons of the recognton and amendment rules. In ths secton, an nsttutonal structure whch may be termed majortaran (M) s assumed. Unlke the A structure, ths s hghly decentralzed, wth open recognton and amendment rules. Votng s assumed to take place sequentally on each tem on the commttee s agenda, and the outcome s determned by a majorty vote, undertaken separately on each dmenson of choce. 21 A full characterzaton of the commttee s delberatons under these open rules would nvolve each of the l commttee members (where l s assumed to be odd) makng a proposal, 21 Ths assumpton of sequental votng (n cases where multple decsons must be made) ensures the exstence of a votng equlbrum under relatvely mld condtons. As wll be seen below, t does not mply that legslators gnore the mpact of ther vote on one ssue on ther optmal choce of how to vote on other ssues. 16

18 and each offerng amendments to each of these proposals. Gven the restrcton that votng takes place separately along each dmenson, and that legslator preferences are sngle-peaked along each of these dmensons (whch emerges from the lobbyng pattern dscussed below), the Medan Voter Theorem mples that a votng equlbrum exsts along each dmenson, and that the equlbrum s the preferred poston of the medan commttee member along that dmenson. The smplest way to characterze the process of commttee votng s to magne that each legslator smply announces her preferred value of p (when the vote on p s held); then, the medan announcement becomes the enacted polcy. In vew of the hghly decentralzed nature of commttee decsonmakng wthn the M structure, t s assumed that lobby groups contrbuton functons are condtoned on a legslator s announcements, rather than on the polcy outcomes whch are enacted. Furthermore, the socal welfare term n the legslators objectve functon has as ts argument the prce vector whch the legslator prefers (and announces), rather than the enacted vector. Ths latter assumpton reflects a stuaton n whch the socal welfare component of the legslators objectve functon arses from a reelecton constrant; as commttee members operate n an autonomous manner, voters evaluate them on the bass of ther votes, rather than the outcome. Note that these assumptons ensure that votng wll be sncere ; legslators have no ncentve to vote strategcally ) The Tax Commttee Outcome The tmng of the M commttee votng game n a TC settng s shown n Fg. 4. In the frst stage, each organzed group chooses (l + 1)/2 members of the TC to lobby; as the members are all dentcal ex ante, ths choce s made at random. 22 Then, each group offers a contrbuton schedule to those legslators t has chosen, condtonal on the legslators votes n the fnal stage. In the next stage, each legslator chooses a vector of producer taxes and subsdes. Member j s choce wll be denoted by p j ; the vector s chosen to maxmze the objectve functon: j j θ W ( p ) + C ( p ) (16) where M j L denotes the subset of nterest groups whch lobby legslator j. Note that ths objectve functon depends only on p j, and not on the vector whch s actually enacted. Fnally, the n votes are taken, wth each member announcng her preferred value for each element of p. M j 22 Note that the groups have no ncentve to lobby more than ths mnmal majorty of members. 17

19 The expected number of groups lobbyng legslator j s k(l+1)/2l, so that E[card M j ] = k(l+1)/2l, though the actual number of lobbysts s random. Member j s choce of p j to maxmze Eq. (16), subject to ths lobbyng, must maxmze the aggregate jont surplus between herself and the lobby groups n M j :.e. Ths yelds the result: p j p j θw ( p ) + W M j I α M j g p f * = (18) θ + α M g j M where I s an ndcator varable (=1 f M j ), and α N / N s the proporton of the populaton belongng to lobby groups n M j. The votng outcome for sector can be characterzed as follows. If L, then there wll be (l/2) members not lobbed by the th group; they wll announce values below the world prce p * (.e. taxes). The (l + 1)/2 members lobbed by wll announce values above p * (.e. subsdes). The outcome wll be the smallest announced subsdy: p MT = mn(p j M j ) If L, every legslator wll announce a tax (as each legslator seeks to redstrbute from the unorganzed groups to those who are lobbyng her), and the outcome wll be the medan announcement: p MT j = medan p 5.2.2) The Specalzed Commttee Outcome The SC structure wthn a M framework s closely analogous to that descrbed above, except that SC s only choosng p. The same restrcton on lobbyng actvty as n Secton (5.1.3) s also mposed here, to ensure comparablty of the results. The tmng s depcted n Fg. 5. In the frst stage, f L, the group randomly chooses (l + 1)/2 commttee members to lobby, and then offer them contrbuton schedules condtoned on ther vote for p. In the next stage, each member chooses her vote for p, denoted p j, to maxmze an objectve functon smlar to Eq. (16). Fnally, the commttee votes: each member announces her p j, and the medan announcement s enacted. j Consder the case where L: a legslator who s lobbed by the group wll choose p to maxmze the blateral surplus: M j M j ( p j ) M j (17) 18

20 j j (19) θw ( p ) + W ( p ) The outcome s dentcal to the expresson n Eq. (14a), so that the preferred outcome s the same for all lobbed members. Clearly, ths wll also be the medan legslator s announcement; thus, the outcome s: p MS M j 1 ( N / N ) g p f p* = θ + ( N / N ) g (20a) If L, each legslator smply maxmzes socal welfare, so that the preferred p j = p * j. The votng outcome s smply: p MS = p * (20b) 5.2.3) Comparng the Outcomes To compare the TC and SC outcomes, t s helpful to make two assumptons, each of whch relates to the noton that the pattern of lobbyng of the TC whch results from the random choces of the lobby groups s suffcently dspersed. Note that, once A2 s mposed (so that Lemma 1 holds), the enacted p under the TC s determned by the largest value of α M j whch s found among the legslators lobbed by group. Usng Lemma 1, whch outcome s larger depends crucally on the values of max ( α M j ) and (N /N). As M j, the proporton α M must, by defnton, be at least equal to (N j /N). Moreover, the extremely mld assumpton that one of the legslators lobbed by s also lobbed by at least one other group s suffcent to ensure that max ( α A3: For all L, some j such that M j and card M j 2 M j M j M j ) s strctly greater than (N /N): That s, each organzed group lobbes a legslator who s also lobbed by at least one other organzed group. The other assumpton s that a majorty of commttee members are lobbed by at least one group: A4: For each SC, at least (l + 1)/2 members for whom card M j 1 Then, the followng result can be derved: Proposton 4: Gven A2, A3 and A4, p MT < p MS Proof: Recall that A2 Lemma 1 If L, the result s a straghtforward applcaton of Lemma 1 j If L, Lemma 1 and A4 medan α M > 0 medan p j < p *.e. p MT MS < p * = p 19

21 Thus, the TC structure leads to lower subsdy levels, as n Secton (5.1.3). 5.3) Polcy Outcomes wth Unversalstc Commttees The fnal nsttutonal framework for nternal commttee decsonmakng examned n ths paper s known as the Unversalstc (U) model, developed by Wengast (1979) and Johnsen et al. (1981). As formulated here, t shares wth the agenda setter framework the features of exclusve proposal power and a closed amendment rule; however, the power assocated wth these nsttutons s dspersed among dfferent commttee members, rather than beng concentrated n a sngle agenda setter. Ths s acheved by parttonng the set of decsons before the commttee nto subsets, over each of whch a dfferent legslator exercses proposal power. The proposals made by the dfferent members are then combned nto an omnbus bll, on whch the commttee votes under a closed rule. 23 The type of arrangement descrbed here can be supported by what s termed a norm of deference among legslators, where each defers to the others n matters fallng wthn ther spheres of authorty ) The Tax Commttee The TC under a U system s assumed to have l members, where member j enjoys deference (.e. exclusve proposal power) wth regard to a subset S j of the sectors of the economy. The subsets are exhaustve and mutually exclusve. The tmng s as shown n Fg. 6. In the frst stage, legslator j s lobbed by all the organzed groups over whose sectors she enjoys deference (.e. all groups n S j L). These groups offer contrbuton schedules condtoned on the elements of the enacted vector of producer prces p UT whch are n S j. Then, member j chooses a (card S j ) 1 vector p jp to propose. In the next stage, j proposes p jp ; ths proposal becomes part of an omnbus bll, denoted p UTP, whch smply combnes p jp j. Fnally, the commttee votes on p UTP under a closed rule. Consder the fnal stage, where the followng result holds: Proposton 5: An equlbrum of the votng game s for all commttee members to vote for the proposal p UTP Proof: Consder commttee member j. If all (l 1) other members vote for p UTP, then p UT = p UTP. Member j s payoff (a weghted sum of campagn contrbutons and socal welfare) depends, by assumpton, only on the enacted p UT. Thus, j s payoff from votng for p UTP s dentcal to that from votng aganst p UTP. Votng for p UTP s a best response for j. 23 An example of ths modus operand s provded by the account n Cohn (1988, p. 6) of the Senate Agrculture commttee n the 1970 s; the Charman s reported to have smply asked each Agrculture member, What do you need, Senator? and compled the answers nto a farm bll. 20

22 Ths equlbrum s not necessarly unque. However, t s possble to argue that t may be focal; n a repeated-game context, t would be possble for commttee members to punsh those among them who dd not observe the norm of deference by votng aganst the defector s proposals, thus denyng them campagn funds. 24 Antcpatng an equlbrum n whch p UTP passes, legslator j wll choose p jp to maxmze the jont surplus: (21) The outcome s: where α N / N. S j S j L p UT jp jp θw ( p ) + W ( p ) S j L p I α g p f S j * = (22) θ + α S g 5.3.2) The Specalzed Commttee Outcome, and the Comparson of Outcomes The SC regme wth each commttee operatng accordng to a U modus operand s closely analogous to the agenda setter model. Indeed, the tmng s dentcal to that shown n Fg. 3, except that an organzed group lobbes a randomly chosen commttee member, rather than an exogenously determned agenda setter. The votng equlbrum gven by Proposton 1 holds, and the polcy outcome (the vector p US ) can be characterzed by Eqs. (14a) and (14b). In comparng the TC and SC outcomes, t proves helpful to assume that: A5: l < k.e. the number of legslators on the TC s less than the number of organzed groups, so that at least one legslator s confronted by multple lobbysts. Accordng to the Congressonal Budget Offce (1981, Table 1, p. 35), 104 tax expendture provsons were n operaton n It seems reasonable to assume that the number of nterest groups nvolved n the taxwrtng process s at least as large as the number of provsons. In contrast, the Senate Fnance commttee has 20 members, and the House Ways and Means Commttee has 39 members (e.g. Duncan, 1997). Gven A5, t follows that: j 24 It mght be argued that repeated nteractons among legslators on the same commttee wll lead them to moderate proposals whch (by requrng hgher taxes) mpose costs on the sectors lobbyng other members, and thereby reduce the jont welfare of these sectors and ther colleagues. Ths factor may mtgate the U tendency towards overspendng to some degree. However, t should be borne n mnd that legslators are prmarly nterested, not n the welfare of the groups whch lobby them, but n the share of the surplus whch they receve from beng lobbed. Thus, t s by no means clear that a low spendng norm wll emerge. 21