### A model of corporate rent-seeking through

### tax legislation

### Michael G. Williams

a### , Charles W. Swenson

b,*a_{Department of Accounting, Anderson Graduate School of Management, University of California,}
Los Angeles, CA 90095-1421, USA

b_{Leventhal School of Accounting, Marshall School of Business, University of Southern California,}
Los Angeles, CA 90089-1421, USA

Abstract

An analytic model is developed to examine the role of rent-seeking expenses on tax
legislation. Rent-seeking expenses are found to be only a fraction of the tax bene®ts at
stake. Rent-seeking expenses increase when ®rms cannot cooperate, when very general
tax legislation is proposed, and when there is legislative support for tax cuts. Ó _{1999}
Elsevier Science Inc. All rights reserved.

1. Introduction

Over the past few years, considerable attention has been given to lobbying and attempts to in¯uence members of Congress.1Particular attention has been given to campaign contributions made to members of Congress in attempts to in¯uence votes on tax legislation (see Freed and Swenson, 1995). To the extent that corporations can in¯uence such votes, there is an ethical issue for both the lawmaker and the corporation.

During the 1980s, approximately $15 million was given by corporate po-litical action committees (PACs) to tax-writing members of Congress (Freed and Swenson, 1995, p. 881). Suprisingly, the per ®rm contributions were less

*_{Corresponding author. Tel.: +1-213-740-4854; fax: +1-213-747-2815.}
E-mail address:cswenson@sba2.usc.edu (C.W. Swenson)

1_{For non-empirical works on rent-seeking and tax legislation see Surrey (1957) or Birnbaum and}
Murray (1987).

than $3 thousand,2 with some ®rms making substantial contributions, and others making virtually none (Freed and Swenson, 1995, p. 882). In empirical accounting research on rent-seeking, a signi®cant relationship was found be-tween these campaign contributions to tax-writing members of Congress and impending tax legislation during the 1980s (Freed and Swenson, 1995, p. 886). Two unexplained phenomena were also observed in that study: the seemingly insigni®cant campaign contributions (or rent-seeking expenditures) relative to the taxes at stake and the high variability of contributions across ®rms (Freed and Swenson, 1995, p. 882).

The purpose of our paper is to theoretically examine why these rent-seeking payments can be so low with such a high variability. The results have impli-cations for both empirical work and for policy. The results can help guide empirical work linking tax legislation and campaign contributions by adding covariates for probabilities of legislative success, industry settings where co-operation is likely (i.e., concentrated industries), and for shared tax bene®ts (i.e., legislation aecting a broad cross-section of ®rms). The policy implication relates to ethics: if it is desirable to discourage rent-seeking, the model may help to identify situations which are at higher risk for rent-seeking activity.

Brie¯y, the ®ndings are as follows. Under all circumstances, rent-seeking expenses are only a fraction of the tax bene®ts at stake. One reason is that incremental expenditures are progressively less eective, so that inframarginal expenditures are highly cost-eective relative to marginal expenditures. A second reason for relatively small rent-seeking expenses is cooperation among ®rms. When industries are concentrated, they are better able to coordinate eorts and reduce expenses.

Rent-seeking expenses are also reduced when there is free riding on tax bene®ts. Thus, where very general tax laws are proposed which apply to all industries (e.g., MACRS depreciation), per ®rm rent-seeking expenditures are lower than where industry or ®rm-speci®c tax provisions are at stake, e.g., the Tax Reform Act of 1986 (TRA86). Finally, overall increases in legislative support for tax cuts result in increased rent-seeking expenditures if tax bene®ts are not shared. Conversely, legislative antipathy toward tax cuts results in decreased rent-seeking expenditures under those conditions.

2. Background

2.1. Taxes and politics

Although no formal theoretical (analytic) research exists on rent-seeking and taxation, theoretical work does exist linking interest group theory and

2_{Similar small per ®rm amounts have been widely reported with respect to members of Congress}
not involved in tax legislation issues (cf. Godwin, 1990, p. 292).

taxes (cf. Hettich and Winer, 1988). They (1988, pp. 702±711) found that small, easily organized groups are most successful in obtaining rents.

Various non-empirical (e.g., Birnbaum and Murray, 1987) studies have considered seeking and business tax breaks. Because other types of rent-seeking expenses are not publicly available (e.g., employment after leaving oce) or cannot be linked speci®cally to tax legislation (lobbying expenses and speaking honoraria), contributions by PACs have been the focus of these studies. See Freed and Swenson (1995, p. 877) for a discussion of this literature. In the only published regression analysis of rent-seeking and taxes, Freed and Swenson (1995) examined individual ®rms' contributions and their tax bene®ts at stake prior to passage of Economic Recovery Tax Act of 1981 (ERTA) and TRA86. Campaign contributions included both PAC contribu-tions and bundled individual campaign contribucontribu-tions by the ®rms' employees, ocers, and directors. They found that such contributions were related to all tax law changes of ERTA, but only to special industry-speci®c preferences on the table prior to the passage of TRA86 (Freed and Swenson, 1995, pp. 885± 887).

Freed and Swenson (1995, p. 881) found that observed PAC contributions varied widely by ®rm, and were only a small percent of tax bene®ts at stake. This ®nding was similar to all published non-tax empirical studies [see cites in Freed and Swenson (1995, pp. 893±894)]. As an indication of the magnitude of PAC contributions, Table 1 in Freed and Swenson (1995, p. 881) reported aggregate PAC contributions during 1980s, when a number of tax bills were enacted. The absolute and relative contributions were substantial; $5.3 million was given to tax writing members of Congress by corporate PACs prior to the enactment of TRA86, which was about 12% of PAC contributions to all po-litical candidates during this period (Freed and Swenson, 1995, p. 881). Freed and Swenson's (1995) analysis also found that prior to ERTA, which had very general provisions aecting virtually all ®rms, contributions were relatively small (Freed and Swenson, 1995, p. 881). In contrast, Freed and Swenson (1995, p. 881) found that equivalent per ®rm contributions just prior to the TRA86 were substantially higher. These observations agree with analytic propositions later in the paper.

2.2. Rent-seeking models

Independent of the empirical studies above, a theoretical literature (e.g., Tullock, 1967, 1980; Berry, 1993) has developed that models ®rms' decisions to make rent-seeking expenditures. The models are general in nature, and do not refer to rents from tax legislation.

With the exception of Berry (1993), there is a single winner of the prize. The prize is awarded to the highest bidder, or bidders in the case of Berry (1993). The probability of winning depends on the size of the agent's expenditures relative to expenditures across all agents. A number of features of the tax legislative process require that these more general models of rent-seeking be expanded. The additional features of our paper's model are: (1) the prize may or may not be won by rent-seekers, depending on a stochastic process; (2) ®rms can cooperate or compete; (3) the prize can be won by any number of rent-seekers, each with a probability dependent on its expenditures; and (4) there may be asymmetric legislative support across agents.

3. The basic model

Assume that there are Nrisk neutral ®rms who seek favorable tax legisla-tion, andmof these ®rms will actually receive or win such bene®ts. The ®rms cannot cooperate (e.g., form an industry PAC) to pool in¯uence expenditures, and instead compete for such legislation. Aggregate tax bene®ts are denoted by

R, whereRW>0 is the payo to a single winner andRLP0 is the payo to

each of the losers if there is at least one winner. Aggregate tax bene®ts can then be written as RRW Nÿ1RL. If there is more than one winner, the sharing rule is:

RW mÿ1RL

m : 1

This ®xed-pie representation of the payo is common to other rent-seeking models and is representative of most tax legislation. In most tax legislation, ®rms seek a speci®c law (or set of laws) that have a revenue cost to the gov-ernment that can be estimated.

The setting where ®rms do not share tax bene®ts is characterized as a winner-take-all environment whereRL 0 and RWRin the case of a single winner. Similarly, sharing of tax bene®ts implies thatRL >0 andRW<Rfor a single winner. Unlike the more general regulatory setting where the govern-ment can, for example, award a franchise to a single ®rm, tax legislation typ-ically involves laws which can be used simultaneously by a number of ®rms and/or industries (although ®rm-speci®c or ri¯e shot laws occasionally are written). For example, tax bene®ts resulting from rapid depreciation under the ERTA accrued to almost all ®rms, even though lobbied for by only certain capital intensive ®rms (see Birnbaum and Murray, 1987).

Successful legislative outcome for any one issue is governed by a
probabi-listic mechanism H
ai, where H is a probability function, and ai are
rent-seeking expenditures (constrained to be non-negative). An example ofHis the
binomial distribution, in which H1ÿ
1ÿpal_{. In this speci®cation, each}

rent-seeking ®rm can purchase a number of independent attempts to in¯uence tax legislation, each with a probabilitypof success. Across time, any one at-tempt can result in a legislative success. The cumulative probability of success increases with each attempt, but the incremental probability falls with each attempt. This setting is representative of real world tax legislation if each at-tempt is with respect to a separate lawmaker or committee, or lawmakers' preferences can stochastically ¯uctuate from attempt to attempt (say, because of ¯uctuating pressures from other interest groups).

We assume d=daiH ai>0;d

2

=da2_{i}H
ai<0. That is, the function is
in-creasing and concave in ai. Incremental expenditures become marginally less
eective over time for the simple reason that previous expenditures have
al-ready provided a signi®cant probability of success and there is not as much left
to gain. The probability can increase signi®cantly from 1%, but not from 99%.
Assume that rent-seeking expenditures,ai, can be made in unitsa(assumed
to be continuous),3which have a constant per unit cost ofc. For example, if
the ®rm hires a lobbyist,amight represent the hours of the lobbyist's time, and

c the lobbyist's hourly billing rate. An analogy can be made for the use of employees in lobbying, or campaign contributions, which are often made in multiples of $100. Firms simultaneously choose their levels ofa. The expected pro®t for a ®rmias a function of the units of rent-seeking expenditures is:

E Y

where each of theNÿ1 other ®rms chooseajinvestments.

Eq. (2) has three terms: the ®rst is expected pro®t, givenihas one or more legislative successes. It is the product of the probability of ®rm i having a legislative success and the expected value of such a success. The value of a success is the sum of expected successful legislation for 1 toNpossible winners. Each of these possible situations (shown in braces) is the value for ®rmi for

that m number of winners times the number of combinations of getting m

winners out ofn®rms, the joint probability of success by themÿ1 other ®rms, and the joint probability of no success by the remaining Nÿm ®rms. The second term is expected pro®t foriwith no successes. It is the product of the probability that ®rmihas no success, the value of being a legislative loser, and

3_{The binomial distribution discussed above would suggest that}

the probability that there is at least one legislative success among the Nÿ1 other ®rms. The third term is the certain constant cost of rent-seeking units of expenditures. We can dierentiate (2) byai, set the result to zero to obtain the pro®t maximizing level of in¯uence expenditures as follows:

0

SinceHis concave inai, this condition is sucient for a maximum, i.e., the second-order condition holds. A symmetric non-cooperative Nash equilibrium in rent-seeking expenditures for a single period is de®ned by (2) withaiaj, i.e., each ®rm makes identical rent-seeking expenditures.

The term in braces is the bene®t of winning, the dierence between the ex-pected payo for the company as a winner and as a loser. Multiplying that by the marginal probability of winning due to an increase in ai, gives you the marginal bene®t of increased expenditures. That must equal the marginal cost ofc.

The concavity ofHensures that inframarginally, each company has positive marginal bene®ts from lobbying. Also, each company gains at least some ex-pected tax bene®ts even if no lobbying is done (the losers' share). Therefore, each ®rm invests less in lobbying than its expected tax bene®ts.

Proposition 1.Rent-seeking expenditures are only a fraction of the rent at stake.

What is surprising about this finding is how relatively small such rent-seeking

expenditures can be. The exact ratio of expenditures to rents is of course a

function of the other variables in the model:H,N,m, and the payos,RWand,

RL. Consider one plausible example: H1ÿ0:9a_{;} _{N} _{}_{2;}_{RW}_{}_{$}_{100;}_{000;}

RL0 (i.e., there are no tax bene®ts to legislative ``losers''), and C $100. Here, the predicted per ®rm expenditure is $3,780. Total expenditures are 7.56% of the rents at stake. This may explain why observed PAC contributions appear to be low relative to tax bene®ts at stake, for example (Freed and Swenson, 1995, p. 888).

The low total cost of lobbying relative to the total tax bene®ts received suggests that ®rms will strongly desire to play this game and prevent others from joining (and compete away the rents) if possible. Thus, an ethical cor-ollary is:

Corollary 1.Firms will value a lobbying presence and will attempt to make the

system as inaccessible to others as possible.

4. Expenditures in a cooperative setting

Suppose that instead of the non-cooperative model presented above, ®rms costlessly form industry coalitions such as an industry PAC [the reader is re-ferred to Linster (1994) for the case of cooperation for adeterministic prize]. Expected group pro®t is then:

E Y

This represents Rtimes the probability of at least one success for any ®rm less the costs ofN®rms. Dierentiating (4) with respect to group rent-seeking expenses and setting the result equal to zero gives maximized expected group pro®ts:

To see how this cooperative solution is dierent from the non-cooperative one, we compare it to the setting of where the rewards to successful legislation are fully appropriable so thatRL0:Then (3) can be rewritten as:

0R 1ÿH aj

Note that the second term in (6) represents the marginal bene®t to ®rmias a result of legislative success by rival ®rms, or ties. The possibility of being tied represents a private return for non-cooperating ®rms, since gains from legis-lative success are shared by all ®rms in the tie. This term is absent in (5), thus the incentive for over-investment. Said another way, in the non-cooperative setting a ®rm is better o tying because it can share in tax bene®ts. On the other hand, there is no bene®t to more than one legislative (across ®rms) success in the cooperative setting. Accordingly:

Proposition 2. When firms cooperate to seek tax legislation, per firm

rent-seeking expenditures are lower than if they do not cooperate in a non-sharing

contest.

The potential scale of this eect is illustrated by the example in the previous
section withH1ÿ0:9a_{,}_{N} _{} _{2,}_{R} _{} _{$100,000, and}_{C} _{} _{$100. Recall that}
absent collusion, each ®rm spends $3780. With collusion, each ®rm only spends
$2210. Collusion leads to even greater cuts in spending when N >2. An
in-teresting feature of the binomial distribution (which is not true for any other
distribution) is that total spending across ®rms is the same in the collusion case,
independent ofN. 1000 ®rms would spend the same amount as one ®rm.

Some interesting observations can be made from Proposition 2. An impli-cation for empirical work is that observed PAC contributions should be lower for organized groups of ®rms than for others. Thus, ®rms in highly concen-trated industries, especially if they have existing trade associations, should have small PAC/tax bene®t ratios. An ethical implication is that, because uncon-centrated industries make relatively larger contributions, law makers can re-ceive more rents by proposing more widespread legislation, or:

Corollary 2. Ceteris paribus, politicians find it more profitable to propose tax

legislation to relatively unorganized industry groups.

Note that this does not mean we should observe more rents actually going to such unorganized groups. Instead, it means that politicians can receive more money from entertaining tax legislation relevant to such groups.

Note that Proposition 2 assumes that ®rms costlessly form coalitions. This might occur where ®rms already belong to an industry trade association, perhaps with a PAC, such that the marginal cost of rallying member ®rms behind any one tax legislation eort approaches zero. On the other hand, it might be that ®rms are not already organized so that the cost is high, or that they are organized but there is still some marginal cost for any particular rent-seeking eort. Assume that the cost of organizing rises strictly withN. Rewrite (4) and (5) as:

E Y

N

a

" #

R1

ÿ 1ÿH aN

ÿNa co; 40

0N R 1

ÿH aNÿ1dH

da ÿco

; 50

whereois the per ®rm organizing cost. RewritingNR 1ÿH aNÿ1

asb_{N}0 and

rearranging (50) we get:

dH da

co

b_{N}0
a

: 7

Recall that themarginalprobability of legislative success is strictly decreasing in units of rent-seeking expenditures (concavity); i.e., the left-hand side (LHS) of

(7) is decreasing ina. Since the right-hand side (RHS) of the ®rst order condition (FOC) in (7) becomes larger aso increases, for the equality in (7) to be main-tained, the LHS must increase as well. As noted above, when dH=daincreases, this implies a decrease in the units of investment, because of the inverse relation between marginal probability of success and marginal units of expenditure. Accordingly, the model predicts that the higher the per ®rm organizing costs, the lower the rent-seeking investments in this cooperative model.

Proposition 3. In a setting where firms can cooperate to obtain favorable tax

legislation, rent-seeking expenditures are lower when organizing costs are

greater.

The above analysis assumes that the ®rms play a cooperative game. To the extent that coalitions are formed without enforceability of agreements, then the public goods (free-riding) phenomenon will occur and perhaps some or no ®rms will provide rent-seeking expenditures. This is essentially the non-collu-sive case with complete sharing of tax bene®ts in which minimal expenditures are optimal. According to Peltzman (1976, p. 239) we would expect cooperative behavior in more concentrated industry groups, because organizing costs and free riding are lower.

5. Expenditures assuming sharing of tax bene®ts

To examine the impact of sharing of tax bene®ts, ®rst rewrite the model in (2) assuming no sharing. To simplify notation, de®ne b^ aj as the expected payo for ®rmifrom winning givenaj:

^ b aj

XN

m1 RW

m

Nÿ1

mÿ1

H ajmÿ1 1ÿH ajNÿm

which is decreasing in aj, and is substituted for most of the ®rst term in (2). Since tax bene®ts are not shared,RL0, and the second term in (2) drops out. Substituting inb^ ajwe write (3) as:

^ b aj

dH dai

c;

or marginal bene®t equals marginal cost, which can be rewritten as:

dH dai

_{^}c

b aj

: 8

ÿc aj b aj dH dai

c; 9

where c aj RL1ÿ 1ÿH ajNÿ1 which is the expected payo for i from losing,4and

b aj

XN

m1

RW mÿ1RL m

Nÿ1

mÿ1

H aj mÿ1

1ÿH aj Nÿm

;

both of which are decreasing inaj. Eq. (9) can be rewritten as

dH dai

c

ÿc aj b aj

: 10

Recall that dH=dai is decreasing and convex in ai when ai is viewed as continuous. Note that b^>b> ÿcb. This in turn implies that the de-nominator in (8) is larger than the dede-nominator in (10), which means that:

Proposition 4.When firms do not share tax benefits with other firms,per firm

rent-seeking expenditures are larger than when firms must share tax benefits.

This is hardly surprising since the larger the sharing of bene®ts, the smaller the incremental payo for winning and the smaller the incentive to win. The implication for empirical work is that observed per ®rm PAC contributions should be larger when ri¯e-shot (®rm speci®c) tax legislation is proposed, than when widespread (e.g., MACRS depreciation, which applies to all ®rms) leg-islation is proposed. An ethical corollary is that politicians can receive more money in the non-shared tax bene®t setting, or:

Corollary 4. Ceteris paribus, politicians prefer tax legislation which provides

benefits to a clearly defined set of firms.

One conjecture resulting from this is that much of the complexity in the Internal Revenue Code might be due to numerous provisions which carefully de®ne conditions for receiving tax bene®ts, i.e., to avoid bene®t-sharing (free riding).

Note that shared tax bene®ts are essentially a public good. For the setting of

adeterministicprize which is a public good, the interested reader is referred to

Katz et al. (1990) and Linster (1993).

4_{Note that this term is absent from (8) because legislative losers do not share in tax bene®ts in}
that winner take all setting.

6. Parametric shifts in legislative probabilities

Suppose that there occurs a parametric shift inH, or the probability that tax legislation will be enacted at a certain level of rent-seeking expenditures. Such a shift can be either positive or negative. An example of the former was the legislative mood prior to the passage of ERTA, where a pro-business, tax cutting mentality had broad popular and electorate support (see, for example, Freed and Swenson, 1995, p. 877). One example of a negative shift is the legislative mood prior to TRA 86, where tax loopholes of various corporations were targeted by the US Congress (see Freed and Swenson, 1995, p. 877).

To examine the eect of changes in legislative support for the provision of tax bene®ts, we will de®ne legislative support as a parameter g, where 0<g<1. Now H ai gH ai where H ranges from 0 to 1. This can be thought of as a two-stage lottery. In the ®rst stage, it is determined whether legislators are or are not susceptible to lobbying. This stage is in¯uenced solely by the degree of legislative support. In stage two, it is determined whether the ®rm's lobbying is successful. This stage is in¯uenced by the amount of rent-seeking expenditure. The ®rm must win both lottery stages to gain the desired tax bene®ts. Ifg is high, then legislators are very amenable to providing tax bene®ts, while if g is low, then the political environment is hostile and no amount of lobbying is likely to be successful.

The direction of the eect of a change in legislative support on rent seeking expenditures depends on how much sharing of tax bene®ts occurs and how many ®rms are competing. Consider the two extreme cases of no sharing and full sharing.

In the no sharing case, the ®rst order condition is

cb^ adH

To determine the eect of g on a, this equation must be dierentiated with respect tog:

Sincecis a constant, this can be rearranged as

da

^

The derivative ofb^can now be determined:

db^

Now it can be shown that

^

Therefore, given (11), da=dg must be positive. This proves the following proposition:

Proposition 5. When there is a shift in legislative support for favorable tax

legislation,the change in rent-seeking expenditures are in the same direction as

the change in legislative support assuming that there is no sharing of tax benefits.

An increase in g increases the total probability of success, but it also in-creases the marginal probability of success since dH=dagdH=da. Therefore, each unit ofais more eective, encouraging more expenditure. An increase ing also increases the probability that other ®rms will win, reducing the expected value of the prize. This eect discourages expenditures but is smaller than the eect on the marginal probability of winning.

The result is contingent to some extent on the speci®c way in which we model legislative support. In particular, Proposition 5 can only be proven by assuming that the legislative support eect on H is proportional. The most basic mathematical concept of legislative support is that it increases the impact that lobbying has on the probability of success. In other words, the cross-partial derivative ofHwith respect toaandg is positive. Unfortunately, this cross-partial derivative condition is not sucient for lobbying to increase with g. This is due to the eect that g can have on the probability of other ®rms winning, which can oset the cross-partial derivative eect. Counterexamples (with a positive cross-partial derivative) can be constructed in which lobbying falls as the legislative support increases. However, such counterexamples are very unusual. We feel that the additional structure used in our de®nition of legislative support (that it has a proportional eect) is a relatively harmless

necessity to allow a de®nitive proof and is representative of the more general case of a positive cross-partial derivative.

When tax bene®ts are shared, the situation is more complicated. To see this, now consider the extreme case of complete sharing in whichRWRL. In this scenario, the only incentive to lobby is to improve the probability of at least one winner. As long as someone else wins, you do not care whether you do as well. In this case, (10) can be simpli®ed as

dH

Rearranging and dierentiating as in (11) yields:

da

The term in brackets is 1 ± expected number of winners. If the number of expected winners exceeds 1 (which it will if N is suciently large), then an increase ingleads to a decrease ina. This is due to the fact that an increase in legislative support signi®cantly increases the probability of another ®rm win-ning, in which case there is less need for you to win as well. On the other hand, if the number of expected winners is less than 1 (which it will be ifNis su-ciently small), then an increase ingleads to an increase ina. In this case, other ®rms' experiences are less relevant to your decision, which is dominated by the fact that your lobbying is now more eective.

7. Unequal costs and bene®ts across ®rms

7.1. Dierential legislative support across ®rms

The foregoing analysis assumed identical legislative support levels across ®rms. Under some circumstances this might not be the case: some ®rms might be under particular congressional scrutiny for whom tax bene®ts would be unpopular; ®rms may have already garnered too many (too few) tax or non-tax bene®ts and are less (more) likely to obtain rents through favorable non-tax rules. Alternatively, some ®rms are more adept at working the legislative process and have better connections (due to past expenditures, for example). One example illustrates these ideas. TRA86 targeted repeal of special tax rules for companies whose ®nancial statements revealed that they had large incomes but paid no federal corporate income taxes (see Birnbaum and Murray, 1987).

Consider the case in which one ®rm has a dierent legislative support level, gi. The FOC for the no sharing and sharing cases are respectively,

g_{i}dH

dai

_{^}c

b aj

and

g_{i}dH

dai

c

ÿc aj b aj :

Changes in gi have two eects on these equations. First, the LHS is
pro-portional tog_{i}. Second, the RHS is negatively related tog_{i}, due to the strategic
reaction of other ®rms. Other ®rms will reduce their expenditures ifg_{i}increases
because the expected incremental value to them of winning will be lower (since
®rmiis more likely to win). The reaction of the other companies will increase
the incremental value of winning to ®rm i, which is the denominator of the
RHS. Both eects of a change ing_{i} have the same eect onai, since dH=dai
must change in the opposite direction ofg_{i} to maintain equality in the above
equations. Hence the following proposition:

Proposition 6. When a firm's legislative support changes, its expenditures

change in the same direction,while competing firms change their expenditures in

the opposite direction.

It follows from this proposition that ®rms with greater legislative support will spend more on lobbying than ®rms with less legislative support, one possible explanation for heterogeneity across ®rms.

Unlike Proposition 5, Proposition 6 does not require that legislative support be speci®cally modeled as having a proportionate eect on H. Proposition 6 requires that the cross-partial derivative ofHwith respect toa andg be pos-itive. That is because there is now a negative eect of a change in one ®rm's

legislative support on other ®rms' success probabilities, resulting in a second-ary eect that enhances rather than osets the primsecond-ary eect.

7.2. Size dierences and rent-seeking expenditures

One of the most important dierences between ®rms is relative size. In this context, size refers to sensitivity to changes in the tax code. This will roughly correspond to traditional measures of size as big ®rms have more to gain or lose from changes in tax rates and depreciation schedules, for example. In addition, some ®rms, particularly in industries with either signi®cant excise taxes (e.g. tobacco) or tax subsidies (e.g. ethanol) will have greater size in this context.

The eect of size in the model is to change the magnitude of the prizes for a
particular ®rm. The prizes arew_{i}RWandw_{i}RL wherew_{i} is the ®rm's size. The
FOC in the no-sharing and sharing cases become, respectively,

dH dai

c

wib^ aj

and dH dai

c

w_{i}ÿc
aj b
aj
:

These equations are similar to the dierential legislative support case withw_{i}
replacing g_{i}. As in that case, ai is an increasing function of wi. Hence, the
following proposition:

Proposition 7.Larger firms that are more sensitive to the tax code spend more

resources on rent-seeking than smaller,less sensitive firms.

7.3. Dierential rent-seeking costs across ®rms

The foregoing analysis assumed that the marginal cost of rent-seeking ex-penditures, c, was identical across ®rms. There are a number of situations where this would not be true. If ®rms already have well-organized lobbying eorts, the marginal cost of rent-seeking might be relatively low. For example, well-established ®rms may already have lobbied before and politicians may be familiar with them and their lobbyists. Or, ®rms may belong to well-organized industry trade associations who have regular lobbyists with whom tax-writing lawmakers are familiar, and who have already set up a PAC.

will reduce their spending in the face of stier competition from ®rmi. Both eects driveaiup. Hence, the following proposition:

Proposition 8. Firms with lower per-unit costs buy more units of rent-seeking

than higher cost firms.

Since the above is based on quantity of in¯uence units purchased, total expenditures can be obtained only by multiplying by the cost per unit. The sign of the change in dollars spent depends on the cost elasticity ofai. That in turn depends on the exact speci®cation ofH. Since expenditures are observable and units purchased are not, the unfortunate implication for empirical work is that, unless the researcher can otherwise distinguish between ®rms,total rent-seeking

efforts might appear similar or even inverted across high and low cost ®rms

based on observed PAC contributions,ceteris paribus. In contrast, dierences in behavior due to variation in marginal bene®ts have an unambiguous eect on expenditures.

8. Numerical analysis

To see the comparative static results in action, a numerical analysis is
per-formed. De®ne the probability function asH
1ÿ0:9a_{}_{;} _{c}_{}_{$}_{1;} _{RW}_{}_{$}_{50;}

RL0; andN 4 ®rms for the winner take all setting. In the sharing of tax bene®ts setting,RW$29 and RL $7 (per ®rm), which is equal in expected value to the winner take all setting. The symmetric Nash and cooperative so-lutions are shown in Table 1 for the case of the above parameters. The solu-tions were obtained using a computer spreadsheet. For comparison, the cases of Hs0:5 1ÿ0:9a (low legislative support), and the non-symmetric case

where two of the ®rms faceH and two faceHsare shown.

Table 1

Numerical example: rent-seeking expenditures per ®rm, Nash solutions (cooperative solutions in parentheses)a

Setting Legislative probability function,H Tax legislation outcome,a

No sharing of tax bene®ts

Sharing of tax bene®ts

1
1ÿ0:9a_{ 8}_{®rms} _{7.77 (3.94)} _{4.15 (3.94)}

2 0:5
1ÿ0:9a_{ 8}_{®rms} _{5.87 (3.92)} _{2.18 (3.92)}

3
1ÿ0:9a_{}_{for 2 ®rms} _{11.10 (7.89)} _{6.92 (7.89)}

0:5
1ÿ0:9a_{}_{for 2 ®rms} _{1.73 (0)} _{0 (0)}

a_{a}

amount of rent-seeking expenditures made by each ®rm. Value of rent is $50; cost per unit of rent-seeking is $1; Number of ®rms4. Cooperation assumes that ®rms can make side payments to each other.

As can be seen, all results are consistent with the comparative statics. The Nash solutions exceed the cooperative solutions in all winner take all cases, and expenditures in the winner take all settings exceed expenditures in the sharing of tax bene®ts setting. When legislative support decreases, expenditures de-crease in both the sharing and no-sharing settings. In the asymmetric legislative support setting, expenditures are much higher for the legislative-favored ®rms than for non-favored ®rms. Indeed, in the sharing case, the non-favored ®rms do not bother to spend anything. Another notable result is the cooperative case with asymmetric support. The optimal strategy is to do all coalition lobbying through the favored ®rms and have the non-favored ®rms subsidize them through side payments. Also note that expenditures are higher in the cooper-ative case than in the sharing case in setting 2. This is evidence of free-riding in which ®rms fail to account for the positive externality they provide to others through their expenditures which increase the probability of at least one win-ner, enabling everyone to gain at least the loser's prize.

9. Conclusions

Our paper has examined rent-seeking in the case of tax legislation. The model makes no assumptions about the time limit or direction of rents sought. For example, in any year lawmakers might propose taking away tax bene®ts, and ®rms would make expenditures with the intent of not losing the rent; this is what was observed as a large infusion of corporate PAC contributions prior to the passage of TRA86 (Godwin, 1990, p. 293, no. 3). If tax bene®ts are viewed as being on the table for each year's legislative budget, then ®rms must con-stantly make rent-seeking expenditures, i.e., not just before attempting to ob-tain new legislation.

The large increase in rent-seeking expenditures prior to TRA86 is attribut-able to two separate eects: an increase in the number of ®rms aected by tax legislation (all ®rms were aected), and an increase in the amount of tax bene®ts on the table vis-a-vis other legislation. It is obvious that an increase in available rents increases per ®rm rent-seeking expenditures. However, holding available rent constant, an increase in ®rms decreases per ®rm expenditures through an erosion of the expected incremental value of winning (since there are more expected winners). In the non-sharing case in which losers gain nothing, the combined eect of a doubling of both is a net increase in per ®rm expenditures,5a result consistent with observed campaign contributions prior to TRA86 (e.g., Freed and Swenson, 1995, p. 881).

5_{This is evident from the fact that}_{b}^_{increases as both}_{R}

One of the most important results which falls out of the model is that in a setting of probabilistic legislation (or other types of regulation), observed rent-seeking expenditures are only a fraction of possible tax bene®ts. In the fairly typical case discussed in Section 3, total rent-seeking expenditures are 7.56% of the tax bene®ts at stake. This general result may partly explain why observed PAC contributions in empirical studies are very small relative to the rents at stake; see discussion in Godwin (1990, pp. 291, 292), and empirical evidence in Freed and Swenson (1995, p. 888).

According to Halperin (1991, p. 7) an analytic model's results should make sense notwithstanding the assumptions.6Here the model assumes a speci®c competitive environment and a balanced government budget. For the former, the non-cooperative model assumesN®rms are competing for tax legislation. Instead, we might view this as N industries who are competing; if ®rms co-operate within industries and organizing costs are negligible, the model's predictions would not change. For the balanced budget assumption,Ris as-sumed to be ®xed. However, we might viewRas being a function ofa, i.e., the politician may make the tax bene®t larger thanR(i.e., go beyond the budget) if rent-seekers value it enough. If we can make an assumption as to the recursive (implicit function) relationship between R and a then the model should be solvable.

Another aspect of the balanced budget setting is what group receives a tax increase equivalent to R (the tax rent). The model follows all prior analytic research by ignoring this aspect, or assuming it falls on individuals. Indeed, the latter is a reasonable assumption (see Peltzman, 1976, pp. 212±220) since in-dividuals rarely exert enough political in¯uence to de¯ect tax increases. However, the possibility exists that ®rms who are not legislative winners could be hit with the corresponding tax increase. This is modeled in Appendix A; as can be seen, the eect is to increase per ®rm rent-seeking expenditures.

Our paper makes no policy prescriptions with respect to the eciency gains/ losses in this setting. Undoubtedly, rent-seeking expenses are used to eect both deserved and undeserved tax bene®ts. An obvious ethical consideration is if the expenditures enable ®rms to buy votes as opposed to supporting their positions on tax legislation by funding campaigns. If the anecdotal evidence is correct and such vote buying occurs, thenanyrent-seeking expenditurea>0 is unethical. Looking back at the paper's ®ndings, we see that such ethical vio-lations can be reduced by:

· _{A cooperative setting. If ®rms are likely to cooperate on policy matters,}

then, if given the opportunity to lobby for tax legislation, they are likely to spend less;

6_{One assumption, consistent with the literature, is the simultaneous moves of the ®rms. For an}
example of sequential moves in a tax setting, see Beck et al. (1996).

· _{Unorganized industries. If ®rms do not already have trade associations or}

industry PACs, and there are high transactions costs in doing so (many and/or widely dispersed ®rms), the ®rms will spend less;

· _{Tax laws that are written to apply generally to many ®rms, instead of }

indus-try-or ®rm-speci®c laws;

· _{Potential tax legislation is considered only when there is hostility among }

leg-islators toward the provision of tax bene®ts; and

· _{Increasing the cost of expenditures. One way to do this would be to have a}

tax on PAC contributions. On the other hand, tax deductibility of rent-seek-ing costs (currently this is not allowed) would increase unethical activity.

Appendix A. Eects of having unsuccessful rent-seekers receiving tax increases

In the case of a tax increase on legislative losers to balance the budget, the pro®t-maximizing condition is:

Eq. (A.1) can be rewritten as:

dH dai

c

d aj b aj

: A:2

Comparing, we see that the denominator in (A.2) is smaller than the de-nominators in (8) and (10). The result is that per ®rm rent-seeking expenditures are higher when legislative losers are taxed. Of course, if ®rms can collude, then no rent seeking will take place at all in this environment, since rent seeking imposes externalities on other ®rms equal to the expected tax bene®ts.

References

Berry, S.K., 1993. Rent-seeking with multiple winners. Public Choice 77 (3), 437±443.

Birnbaum, J.H., Murray, A.S., 1987. Showdown at Gucci Gulch: Lawyers, Lobbyists, and the Unlikely Triumph of Tax Reform. Vantage, New York.

Freed, G., Swenson, C., 1995. Rent-seeking and US corporate income tax laws. Contemporary Accounting Research 11 (2), 873±894.

Godwin, R.K., 1990. Investments in rent-seeking. Public Choice 64 (2), 291±297.

Halperin, R., 1991. Analytical methodology in tax research. In: Enis, C. (Ed.), A Guide to Tax Research Methodologies, American Taxation Association, Sarasota, Fla, pp. 10±25. Hettich, W., Winer, S., 1988. Economic and political foundations of tax structure. The American

Economic Review 78 (3), 701±712.

Katz, E., Nitzan, S., Rosenberg, J., 1990. Rent-seeking for pure public goods. Public Choice 65 (1), 49±60.

Linster, B., 1994. Cooperation rent-seeking. Public Choice 81 (1), 23±34.

Linster, B., 1993. A generalized model of rent-seeking behavior. Public choice 77 (4), 621±635. Peltzman, S., 1976. Toward a more general theory of regulation. Journal of Law and Economics 19

(2), 211±240.

Surrey, S., 1957. The congress and the tax lobbyist: how special tax provisions get enacted. Harvard Law Review 32 (3), 1145±1182.

Tullock, G., 1967. The welfare cost of taris, monopolies, and theft. Western Economic Journal 5 (2), 224±232.

Tullock, G., 1980. Ecient rent-seeking. In: Buchanan, J.M., Tollison, R.D., Tullock, G. (Eds.), Toward a Theory of the Rent-Seeking Society, Texas A&M University Press, College Station, pp. 211±272.