Editorialists throughout the United States derided

Transcription

1 Geographic Politics in the U.S. House of Representatives: Coalition Building and Distribution of Benefits Frances E. Lee Case Western Reserve University This article argues that scholars need to consider the structure of House representation to better understand distributive politics. Because House districts (unlike states) are not administrative units in the federal system, House members cannot effectively claim credit for most grant-in-aid funds. Instead, their best credit-claiming opportunities lie in earmarked projects, a small fraction of federal grant dollars. As a consequence, I expect to find: (1) political factors have a much greater effect on the distribution of earmarked projects than on federal funds generally; and (2) project grants are a better support-building tool for coalition leaders than allocations to states. I test this argument with a study of the 1998 reauthorization of surface transportation programs and find strong support for both hypotheses. Let me emphasize, when we hear people saying, well, if you eliminate the [earmarked] projects you save money, Mr. Speaker, we do not save a penny. The money, if there are no projects, simply goes back to the States or downtown [to the Department of Transportation]. Rep. Bud Shuster (R-Pa.) (Cong. Record 1998, H1887) Editorialists throughout the United States derided the 1998 reauthorization of federal transportation programs as 100 percent lard (Editorial 1998f), an all-you-can-eat pork buffet (Editorial 1998c), the bill that ate Washington (1998d), and a prime example of Capitol Hill politics as usual (Editorial 1998e). These editorials almost all took the same basic structure: after concocting a pun or two using the pork conceit, a litany of earmarks would follow, such as a new George H. W. Bush expressway (Editorial 1998b), or a billion dollar road to nowhere (Editorial 1998a). Like most journalistic coverage of the subject, these editorials perpetuate the impres- sion that high levels of federal spending are the result of wasteful parochial projects that members of Congress create for their own political benefit. Instead, as many House members pointed out during floor debates, the legislation simply set aside 5% of the total amount authorized for surface transportation for members of Congress to designate themselves. Deleting all the special projects from the bill would have had no effect on the overall amount spent. What is true of the 1998 transportation earmarks is true of earmarks generally. Although controversial, earmarking has little budgetary impact (see Schick 2000, 211 5). Citizens Against Government Waste, an organization devoted to identifying earmarked funds, estimated the cost of all such earmarked projects in 1999 at $17.7 billion (Williams 2000), approximately 0.1% of total nondefense federal outlays. By contrast, intergovernmental grants accounted for fully 22% of all federal nondefense outlays in FY Expenditures on such programs were on the rise throughout the 1990s (Conlan 1998). 1 These facts present political scientists with a puzzle: it is clear from the widespread practice of earmarking Frances E. Lee is Associate Professor of Political Science, Case Western Reserve University, Euclid Avenue, Cleveland, OH (FEL@po.cwru.edu). The author thanks Gregory Caldeira, Emery Lee, David Mayhew, Bruce Oppenheimer, Wendy Schiller, Joe White, and the anonymous reviewers for comments on this research and Karen Thornton, Brendan Elliott, and the Center for Statistics and Geospatial Data at Case Western Reserve University for their research assistance. An earlier version of this article was presented at the 2002 Meeting of the Midwest Political Science Association, and the author owes special thanks to Diana Evans for her comments as discussant. 1 Expenditures for grants-in-aid were $267 billion in fiscal year 1999, an increase in constant dollars of 61% over 1990 levels (U.S. Office of Management and Budget 2002). American Journal of Political Science, Vol. 47, No. 4, October 2003, Pp C 2003 by the Midwest Political Science Association ISSN

2 GEOGRAPHIC POLITICS IN THE HOUSE 715 that members of Congress engage in the sort of particularistic, reelection-oriented behavior described by distributive theories. However, the distribution of most federal domestic assistance cannot be explained by the usual logic of the pork barrel most federal grant money is simply not distributed in a way that maximizes creditclaiming opportunities for individual members. If distributive policy is defined as government programs that dispense funds unit by small unit, each unit more or less in isolation from other units and from any general rule (Lowi 1964, 690) it actually accounts for very little of the distribution that takes place. Legislation providing for distributive programs rarely looks like the classic logroll, in which Congress passes an omnibus bill containing long lists of specific projects. The overwhelming majority of funds for federal grant-in-aid programs are allocated according to statutory formulas, which dole out funds automatically, usually to state governments (Stein and Bickers 1995, 22). 2 Accordingly, political scientists who analyze aggregate patterns in federal spending have found that distributive theories at best explain only a small proportion of the funds distributed. One recent comprehensive study of federal spending between 1970 and 1990 reports, virtually none of the political variables were statistically significant at standard significance levels (Levitt and Poterba 1999, 207). In fact, the weak empirical evidence for distributive theories led Maass to conclude, Congress is terrible, unbelievably bad, at logrolling and porkbarreling (1983, 69). We are left then with a disjuncture between the micrologic of the electoral connection and the macropatterns in the distribution of federal funds. Overall domestic spending cannot be reduced to individual incentives. Scholars who have confronted this fact often conclude that simplifying economic assumptions about members of Congress can only take us so far. Members have other goals beyond reelection, and they pursue these goals after they have satisficed (Maass 1983, 71) the reelection imperative. No doubt this is the case. But this conclusion does not tell us when we should expect Congress to make allocational decisions on the basis of broader policy considerations, or, alternately, under what conditions Congress will distribute funds on the basis of more parochial considerations. 2 A review of the comprehensive list of grant in aid programs in 1995 reveals that more than 90% of all federal funds for grantsin-aid were allocated by formula. The recipients of formula grants, in turn, are almost always states: 86% of all formula grants-inaid went to states (Advisory Commission on Intergovernmental Relations 1995). In order to gain some new traction on the politics of geographic distribution, I propose that we look more closely at the basis of House representation. If the constituency connection is fundamental, we should consider more carefully the characteristics of those constituencies and how they shape House members incentives in distributive policymaking. After examining these below, I argue that when House members consider the geographic distribution of federal funds, they have a greater electoral interest in the distribution of earmarked projects than in the distribution of overall federal dollars. In other words, the method of distributing the funds affects the degree to which parochial considerations predominate in Congress. It is not an original idea to argue that the policy type under debate affects the politics of its consideration (Lowi 1964; Ripley and Franklin 1991). But it is not only the type of policy distributive, redistributive, or regulatory that shapes its politics. Within distributive programs themselves, the policy tools employed affect congressional incentives to credit claim and pork barrel. If this argument is correct, political factors should leave a deeper imprint on the distribution of earmarked projects than on most funds allocated in intergovernmental grants. House Districts and Federal Grants House districts are nothing but electoral entities. With the exception of the seven states with only one congressional district, House districts are not administrative or governmental units. Under the equal population requirements imposed by the Supreme Court, most House districts are not even coterminous with existing governmental units, subdividing cities and cutting across county lines. Because they are not governmental units, House districts per se are not and cannot be the direct recipients of federal intergovernmental grants of any kind; instead, most federal grants-in-aid go to state governments. There is thus a mismatch between the type of geographic unit represented in the House and the type of geographic unit that directly receives most federal grant money. This mismatch has three important effects on House members incentives when they consider programs to allocate funds geographically. First, House members often lack a bottom line. Because House districts are not administrative units in the federal system, systematic data on the amount of money they receive in federal grants is difficult to obtain. The Office of Management and Budget maintains records of the amounts states receive under all federal grant programs and has regularly published this information as the Budget Information for States. Nothing equivalent exists for

3 716 FRANCES E. LEE congressional districts. 3 Such bottom line figures are central to Senate debates over geographic allocation. Any senator proposing a change in a formula grant program is expected to provide a chart showing how much each state would receive under the proposal, and senators vote accordingly (Lee 2000). Unlike senators, House members cannot easily compare the total dollar amount their district will receive in grants with the amount that other districts will receive or with the amount their district received in the past. This lack of bottom-line data obscures House members individualparochial stakes when they consider the overall design of intergovernmental grants. By comparison, House members can easily see whether a bill contains an earmarked project located in their district. The amount of money involved in such earmarks is often trivial. But even if trivial relative to the overall program, the earmark s presence gives a House member a clearer, more visible parochial stake than an amount of dollars allocated to his state s government. Obtaining the earmark demonstrates that the member is working hard on his constituents behalf. The small size of the grant does not diminish its political value few constituents have enough knowledge to contextualize any dollar figure in light of overall program spending. 4 Second, this mismatch between House districts and the federal grant system creates more than an informational problem for House members; it also affects their ability to claim credit for funds allocated. States often enjoy wide discretion in the expenditure of federal grants. 5 In many cases, this discretion is so broad that the grant-inaid is better viewed as a funding stream for state governments than as a federal program (GAO 1998, 15). As one state official explained, These aren t federal programs, they are state programs that receive federal funds (GAO 1998, 16). As a result, even though House districts may be affected by the amount of funds the state receives 3 Scholars who have been interested in monitoring federal grants at the congressional district level have had to reconstruct the flow of federal outlays from the Federal Assistance Awards Data System (Bickers and Stein 1991). 4 Several studies (Bickers and Stein 1996; Stein and Bickers 1994, 1995; Sellers 1997) have found that measures of the number of new awards in the district perform better in House electoral connection models than increases in total district spending. We are skeptical about the appropriateness of a dollar measure of particularized benefits, Stein and Bickers (1995, 135) write; It is the occurrence of an award, not its dollar size, that is usually salient to constituents. 5 Such programs include (among many others) the federal grants for transportation, the Social Services Block Grant, Rehabilitation Services, Medicaid, aid for foster care and children s services, and Temporary Assistance for Needy Families. overall, there is no necessary connection between the two. Rep. James L. Oberstar (D-Minn.) explained the problem as follows: A statement was made earlier in this debate, decisions [about geographic allocation] should be made by the States, who know what the needs and priorities are....but let me tell my colleagues what kind of wisdom there is in state government. There was a stretch of highway in my district on which, over 15 years, 57 people have died. Where did that appear on the State priority list? (Cong. Record 1998, H2015) From the point of view of a House member, it is often not clear that working to increase the state s allocation will benefit his constituents. The benefits to the state may not trickle down to his individual district. Third, most House members face a collective action problem when considering state-level allocations. The median House member represents a district in a state with twelve other congressional districts. All else being equal, a House member s district is likely to benefit from increased funding for his state, but this is a collective good for all members of the state delegation. Any creditclaiming opportunity created by state-level benefits is available to any of the state s House members. In this sense, there are twelve potential competitors (Arnold 1990, 75) claiming credit for the achievement. Political scientists have often focused on the collective action problems that occur at the level of national policy: individual members are held electorally accountable only by local constituencies and thus focus on particularized benefits, neglecting the wider national interest (Fiorina 1989, 107 8; Jacobson 2001, 241). Collective action problems can also occur when the House considers distributive policies if House members focus on the small projects earmarked for their own constituents rather than the much larger allocations that go to state governments. In this sense, even pork barrel politics is vulnerable to the problem of collective accountability. In sum, most federal grant-in-aid dollars are distributed at the state level, but the best credit-claiming opportunities for House members come in the form of special projects and earmarks, which are rather inconsequential to overall federal spending. This mismatch between the structure of House representation and the structure of the federal grant-in-aid system is likely to affect distributive politics in the House of Representatives. Even though project grants account for only a small percentage of federal domestic assistance funds, House members value them because of the credit-claiming opportunities they afford. By contrast, House members may care deeply

4 GEOGRAPHIC POLITICS IN THE HOUSE 717 about the larger allocational decisions from a policy perspective, but funds that go to state governments give them little tangible evidence to demonstrate that they are working hard on behalf of their constituents. Given these incentives, we should expect to find two patterns in House distributive politics. First, the distribution of earmarked projects in the House will be more influenced by political factors than the overall distribution of federal grant-in-aid funds. Second, the coalitionbuilding process in the House should focus more on the allocation of earmarks than on the overall distribution of funds. If earmarks have greater electoral value to House members, they will serve as a better tool for coalition leaders to build support for the bill than manipulating the funding distribution to states. I test this argument with a study of the politics and distributive policy outcomes of the 1998 House reauthorization of federal surface transportation programs. In reauthorizing these programs, the House simultaneously considered the formula allocating transportation funds to states, a host of earmarks requested by individual members, and the discretionary programs administered by the federal Department of Transportation. 6 This reauthorization provides a good test of these expectations about House incentives for three reasons. First, it permits a direct comparison of the politics of earmarks with the politics of overall state allocation while holding the program type constant. The earmarks and the state-level grants did not serve different policy goals. In fact, the earmarked projects would have been eligible for funding out of the state s overall allocation (although those projects might not otherwise have been prioritized by state officials). If distribution patterns dramatically differ between these two methods of distributing funds, it strongly suggests that House members regard them differently for political, as opposed to policy, reasons. Second, it provides an opportunity to study coalition building in distributive policy. Because both the state-based formula programs and the earmarks were authorized in the same legislation, it is possible to directly compare how House members receipt of projects and increases in overall state funding relate to their support for the legislation. Third, transportation policy is important in its own right as one of the federal government s largest domestic assistance commitments. Surface transportation programs alone consistently account for at least 20% of all the federal government s grants-in-aid to state and local governments. 7 6 The funding for each type breaks down as follows: 88% of the funds were allocated directly to states, 5% were distributed in earmarked projects, and the federal discretionary programs accounted for the remaining 7%. 7 Figure is calculated with data from the 1998 Statistical Abstract of the United States. Data and Variables Data on the allocation of funds to states is easily available. Charts showing the amounts states would receive under the House bill were circulated in Congress and printed in the Congressional Quarterly Weekly Report (Gettinger 1998, 1068; Hosansky 1997b, 1067). Locating the congressional district for each of the 1,506 earmarks in the House bill was more difficult. Some of these were identified by finding the member who had requested the earmark in hearings before the committee (House 1997). When the city or county designated to receive the earmark was represented by a single member of Congress, the congressional district could be determined using the Congressional District Atlas (Department of Commerce 1993). The remaining projects were identified on maps that showed both highways and congressional district lines, which were generated with geographic information software. 8 In cases where the project was located in more than one congressional district, each district was counted as receiving a grant, and the total dollar amount for the project was divided between them. Determining whether Congress allocates funds on the basis of substantive policy or political considerations (or some combination of the two) must begin by identifying the major policy goals relevant to geographic distribution. 9 According to Rep. Bud Shuster, chair of the House Committee on Transportation and Infrastructure, a top priority of the legislation was to increase investment in the infrastructure that tie[s] our country together, particularly the federal-aid mileage that bears most of the interstate traffic (Cong. Record 1998, H1886). A second goal was to guarantee states a better return on their federal gas tax and highway trust fund contributions. A GAO study (1995) has recognized highway trust fund contributions as a valid measure of state need for transportation spending because they closely reflect use of the transportation network in the state. 10 Two variables Number 8 Two data layers were needed to create these maps: highways and roads, and congressional district lines. Both were available in a CD- ROM set entitled StreetCD99 available from GeoLytics. The GIS software used was ESRI s ArcView. 9 Some goals of the legislation such as increasing transportation spending overall or stiffening the penalties for drunk driving are separate from matters of geographic distribution. 10 The House proposal to increase the minimum allocation for states is also motivated in part by parochial considerations. In 1998 most (30) states fell into the donee category, but most House members represented constituencies (219) located in donor states. But this kind of parochialism is different from that posited by distributive theories, in that members who gain benefits through a broad policy change like this are distributed across the entire House; possessing useful committee assignments or majority party status reaps no extra benefits.

5 718 FRANCES E. LEE of Federal-aid Highway Miles and Return on Trust Fund Contributions 11 are included in the analysis to account for the effect of these goals on geographic distribution. Committee power has long been recognized as a source of political influence in distributive politics (Arnold 1979; Ferejohn 1974). Three variables account for committee members influence on geographic distribution: Committee Leader, Republican Committee Member, and Democratic Committee Member. In the districtlevel models, Committee Leader indicates that the district is represented by a chair or ranking member on the House Transportation and Infrastructure committee or the Subcommittee on Surface Transportation. In the state-level models, Committee Leader indicates that a House member from the state holds one of the committee leadership posts. The committee membership variables are measured in the same way, indicating in the district-level models that the district is represented on the House Transportation and Infrastructure Committee and in the state-level models that at least one House member from the state is a committee member. More recently, scholars have focused on the parties role in distributive politics (Carsey and Rundquist 1999; Cox and McCubbins 1993; Levitt and Snyder 1995). If members of Congress believe that credit claiming for distributive benefits helps them get reelected, then the majority party should not regard minority party members as full partners in distributive coalitions (Balla et al. 2002). Accordingly, majority party leaders should be especially concerned that their most vulnerable party members receive distributive benefits (Denemark 2000). Two variables measure the importance of party: Party Affiliation and Percent of the Vote in 1996 Party. In the district-level models Party Affiliation is coded 1 when the district is represented by a Republican member, otherwise 0; in the state-level model, the variable refers to the percentage of the state delegation that is Republican. PercentoftheVote in 1996 Party is an interaction term to determine if the majority party targets benefits to its marginal members. In the state-level models, this variable reflects the percentage of each state s Republican delegation that won reelection by less than 55% of the vote in Data on number of federal-aid lane miles and return on trust fund contributions were drawn from the Federal Highway Administration s 1997 Highway Statistics. Including both these policy variables in the regression analyses does not present multicollinearity problems (r =.29). 12 Other measures of this state-level variable were tested such as whether even one majority party member from the state delegation won in a close race and the average percent of the vote obtained by delegation Republicans without changing the results in any significant way. Geographic distribution may also be affected by differences in members incentives and abilities to obtain funds. Senior members of Congress may enjoy advantages in the distributive politics game, simply because they are more effective in legislative politics generally (Hibbing 1993, 80). Marginal members may have stronger than average incentives to seek local benefits (Bickers and Stein 1996; Levitt and Snyder 1997; Stein and Bickers 1995). Two variables are included to account for these sources of political influence on distribution, a member s number of YearsofServiceand the member s Percent of the Vote in House apportionment may also influence the distribution of funds to states. Populous states have more representatives in the House, and in theory this could affect policy outcomes. If state delegations cooperate on behalf of the whole state, then populous states may be advantaged. As discussed above, however, a collective action problem may inhibit effective cooperation among larger delegations. If so, populous states will receive no advantage and may even suffer a disadvantage relative to smaller states. To test for this effect of House apportionment, a variable Number of Districts in the State is included in the analysis of state-level allocations. House apportionment may also shape the incentives of members representing single-district states. Such members can claim credit for the much larger sums obtained by increases in state-level funding, and it may even be politically beneficial for them to forego projects. By not participating in the pork barrel game that so many editorialists deplore, these members can assume the mantle of fiscal responsibility. To test for this effect on the models of project distribution, a dummy variable One District State wasincluded(coded1formemberselectedatlarge and 0 for all others). The Politics of Geographic Distribution The 1998 transportation reauthorization provided funds to every state; in addition, an overwhelming majority of congressional districts 86% of House districts (374) received at least one earmark. 14 In this sense, coalition leaders appear to have followed a policy of universalism, 13 Data on members seniority and the percent of the vote received in 1996 were drawn from Vital Statistics on Congress, (Ornstein, Mann, and Malbin 1998). 14 The mean House district received 3.6 earmarked projects and the median received 3.0. Other descriptive statistics for these variables are as follows: minimum = 0, maximum = 24, variance = 11.9, and standard deviation = 3.44.

6 GEOGRAPHIC POLITICS IN THE HOUSE 719 in which every member, regardless of party or seniority has a share in the benefits (Mayhew 1974, 88). But a universalistic coalition need not be egalitarian. Some members of the coalition may receive a larger share than others, and the geographic distribution of earmarks clearly illustrates this. 15 Tables 1 and 2 show OLS regression analyses of the number of earmarked projects congressional districts received and the dollar value of those projects. 16 The results are calculated both with and without the policy variables included. Given that only a small percentage of transportation funds were allocated in earmarks, it is perhaps not surprising that members did not use these to carry out the broader policy goals of the legislation. Neither of the policy variables had the expected effect on the geographic distribution of projects. Both these tables show, however, that political factors clearly influenced the distribution of earmarks. First, there is a substantial premium for committee members and leaders. The multivariate regression results reveal that, all else being equal, committee members of both parties receive about $20 million more in projects than the average noncommittee member and about three more projects each. Committee members received around twice as many projects and twice as much allocated for them as noncommittee members. 17 The bonuses for committee leaders are even larger. 18 Second, there is a solid advantage for the majority party in the distribution of project dollars. Controlling 15 The mean House district received $20.2 million in earmarked projects and the median received $16.8. Other descriptive statistics for these variables are as follows: minimum = 0, maximum = 97, variance = 279.8, and standard deviation = Because the number of projects received is a count variable, OLS results may be biased and inefficient (King 1988; Long 1997). Furthermore, Poisson regression is inappropriate because of overdispersion in the dependent variable (variance = 11.9; mean = 3.6). To address these concerns, negative binomial regression models (NBRMs) were also estimated using STATA. The results do not differ substantially from the OLS models: all the coefficients take the same sign and with one exception each takes the same level of statistical significance. The one discrepancy is that the coefficients for Percent of the Vote in 1996 Party are not statistically significant in the NBRMs: p =.06 for the model without the need variables and p =.12 for the model with the need variables. Because the results from OLS and NBRM are so similar overall, I present the simpler OLS results in Table The mean committee member received 6.7 projects while the mean noncommittee member received 3.0 projects (p <.001). The mean committee member received $39.8 million in projects, while the mean noncommittee member received $16.5 million (p <.001). 18 The mean committee leader received 14.7 projects while the average nonleader received three projects (p <.001). The mean committee leader received $66.7 million in projects, while the mean nonleader received $19.7 million (p <.001). TABLE 1 Political Effects on the Number of Earmarked Projects, 1998 Transportation Reauthorization (N = 435) Model I B (t-value) Model II B (t-value) Federal-aid Highway Miles ( 1.11) Return on Trust Fund.94 + Contributions (2.00) Leader of Transportation and Infrastructure (5.45) (5.57) Committee Republican Member of Transportation and (6.13) (6.28) Infrastructure Democratic Member of Transportation and (4.73) (4.69) Infrastructure Seniority (1.46) (1.63) Party (1.27) (1.46) Percent of the Vote in General Election ( 2.17) ( 2.02) Percent of the Vote in General Election Party ( 1.68) ( 1.95) One District State ( 2.00) (.97) Constant (3.96) (4.81) F Mean of Dependent Variable Adj. R p <.05; p <.01; p <.001 (one-tailed test) + p <.05; ++ p <.01; +++ p <.001 (two-tailed test) for other factors in the model, Republicans received an estimated $16 $17 million more than the average Democrat (p <.05). The partisan effect is not quite as apparent on the number of projects; in these models the coefficient for party affiliation is positive but not statistically significant. 19 Interestingly, these results are consistent with 19 Bickers and Stein (2000) show that Republican members have greater demand for certain types of spending (e.g., contingent liability programs) than Democrats, and thus changes in aggregate federal spending since the Republican take-over reflect the parties

7 720 FRANCES E. LEE TABLE 2 Political Effects on the Dollar Value of Earmarked Projects, 1998 Transportation Reauthorization (N = 435) Model I B (t-value) Model II B (t-value) Federal Aid Highway Miles ( 1.45) Return on Trust Fund 4.87 Contributions (2.23) Leader of Transportation and Infrastructure (3.79) (3.92) Committee Republican Member of Transportation and (8.21) (8.34) Infrastructure Democratic Member of Transportation and (7.36) (7.29) Infrastructure Seniority (1.59) (1.76) Party (2.20) (2.43) Percent of the Vote in General Election (.82) (.64) Percent of the Vote in General Election Party ( 2.07) (2.38) One District State ( 1.24) (.06) Constant (2.57) (3.52) F Mean of Dependent Variable Adj. R p <.05; p <.01; p <.001 (one-tailed test) + p <.05; ++ p <.01; +++ p <.001 (two-tailed test) the conclusion of Balla et al. s partisan blame avoidance account of pork barrel politics: by giving the minority some pork, the majority inoculates itself against charges of wasteful spending, but by granting the minority a smaller share of the federal pork pie, the majority party boosts the electoral fortunes of its own members (2002, 524). The findings also suggest that marginal Republican districts different preferences. There is, however, no reason to suppose based either on the findings of Bickers and Stein or on the parties voting behavior that there is a party dimension to demand for transportation spending. did come in for special treatment. In Table 2 the interaction term Party 1996 General Election takes a negative, statistically significant (p <.05) coefficient meaning that Republicans who won their last election by narrower margins received more funds for their districts projects than other members. 20 Individual member abilities and incentives also have an effect on the distribution of projects. Marginal members generally received a larger number of individual projects (p <.05), but there is no statistically significant effect on the amount of money allocated to their projects. As expected, the seniority variable is positive in each model; however, it only has a statistically significant effect (p <.05) on the dollar value of projects. Speaking on the House floor, Rep. Steve Largent (R-Okla.) indicated that the benefits to marginal members were part of a conscious political strategy: I want to say that I am opposed to the bill... I would ask the question, is it right that the campaign committee chairmen are consulted before the special road projects are given to Members who live in or represent politically sensitive districts where they have tough races coming up in November? (Cong. Record 1998, H2013). House apportionment appears to have an impact on the incentives of at-large members to seek special projects. The variable generally takes a negative coefficient, revealing that the seven members who represent one-district states were less likely to receive projects. The coefficient reaches statistical significance (p =.02) in one of the models of the number of earmarked projects. Only two states, Montana and Delaware, received no earmarked projects in the transportation bill and both have only one House district. In fact, Delaware Rep. Mike Castle held a press conference denouncing the earmarks in the House bill, and as a consequence came in for special criticism on the House floor when Shuster singled him out during a tirade against the hypocrisy of members who feel that nobody s cause but their own is worthwhile (Cong. Record 1998, H1887). Overall, these findings reveal a strong imprint of political factors on the distribution of earmarks. Committee power, partisan motivations, and individual members abilities and incentives all have a measurable impact. However, only a small amount of federal money was distributed with these projects relative to the overall 20 The size of the coefficient indicates that, all else being equal, a Republican who won reelection in 1996 with only 55% of the vote received an additional $10 million in projects compared to a Republican member who was unopposed in the last election and an additional $4.6 million compared to a Republican member who received 75% of the vote.

8 GEOGRAPHIC POLITICS IN THE HOUSE 721 TABLE 3 Political Effects on Total Funds Allocated, Measured as the Percent Increase or Decrease for Each State in the 1998 Transportation Reauthorization (N = 50) Model I B (t-value) Model II B (t-value) Federal-Aid Highway Miles.0002 (2.20) Return on Trust Fund Contributions ( 5.45) Number of Congressional Districts in the State ( 1.88) (.83) Leader of Transportation and Infrastructure (.11) (.74) Committee Republican Member of Transportation and (1.58) (.96) Infrastructure Democratic Member of Transportation and (.10) (1.30) Infrastructure Seniority (Average) (3.40) (.02) Party (Percent Republican) (1.80) (1.44) Percent of House Members who Are Marginal (<55%) ( 1.16) (.79) Percent of Republican Members who Are Marginal (<55%) (.39) (.02) Constant (1.55) (.74) F Adj. R Mean of Dependent Variable p <.05; p <.01; p <.001 (one-tailed test) + p <.05; ++ p <.01; +++ p <.001 (two-tailed test) authorization. Eighty-eight percent of the authorized funds were directed to state governments. Table 3 examines the effect of policy goals and political factors on the geographic allocation of these grants to states. The dependent variable in the model is the percent increase (or decrease) in the state s allocation under the House bill compared to existing law at the time. 21 Unlike the earmarked projects, these funds were not distributed 21 This measure has the advantage of distinguishing between the base and the increment in spending decisions (Ray 1980, 6). at the district level, and so the model must be estimated with state-level data. 22 Each of the variables thus reflects the characteristics of the state s House delegation. As shown in Table 3, political factors left virtually no imprint on the distribution of funds to states in the House bill. The only political variable with a statistically significant effect in the expected direction (p <.01) was the average seniority of the House delegation. Although the coefficients on the committee representation variables are generally positive, none is statistically significant. It is thus not possible to conclude that a state s representation on the House committee on Transportation and Infrastructure yielded any consistent benefits in overall state funding. Similarly, being represented by a more Republican delegation produced no statistically significant effect on overall distribution. Finally, there is no evidence that larger state delegations successfully cooperate to produce extra gains for their states. Instead, the policy factors account for most of the variation in the allocation of funds to states. States with more federal-aid highway miles and lower returns on their trust fund contributions gained larger funding increases. In sum, the distribution of state funds does not reflect the political influence or incentives of the state s House delegation. The dramatic difference between these findings and those from the analysis of earmark distribution may partly stem from the different levels of analysis. House members may have a tacit understanding of how their districts will be treated by their state departments of transportation, so that even if their state overall does not receive large funding increases their districts interests will still be protected. However, there is strong reason from the Congressional Record to think otherwise. Many What the state received in the past is due to the work of previous members; it is the increase (or decrease) that reflects the influence of House members in However, the findings for the political variables do not substantively change if the dependent variable used in the model is the actual dollar amount each state would receive. Descriptive statistics for the dependent variable analyzed in the model are as follows: minimum = 28.9, maximum = 89.1, mean = 48.9, median = 56.4, variance = 553.1, and standard deviation = Unfortunately, no official data sources exist that allow one to analyze the amount of these state funds received by each congressional district. The Federal Assistance Awards Data System maintained by the Bureau of the Census does not track the suballocation of state funds once they are paid to the state government. The Consolidated Federal Funds Report (U.S. Bureau of the Census 2002) tracks the allocation of funds down to the county level, but counties map poorly onto congressional districts. Many counties especially those in large metropolitan areas contain multiple congressional districts, and federal data sources provide no way of determining how much each district received. It is important to note, however, that House members themselves faced this information problem when considering the distribution of funds to states.

9 722 FRANCES E. LEE House members including some of the most influential members on transportation issues complained that they did not know where transportation funds would be spent in their states and feared that their districts needs would be neglected. 23 Members defended the use of earmarked projects for precisely this reason. If House members were able to provide for their districts and claim credit for projects without the use of earmarks, it is difficult to see why they would have incurred editorialists wrath by including more than 1,500 such projects in the bill. Taken together, these findings are compatible with the idea that the basis of representation in the House affects members incentives in geographic distribution. Funds earmarked for projects in congressional districts provide House members with credit-claiming opportunities; members thus perceive them as a political commodity. As such, they are distributed on a political basis in the House, and members advantaged by committee and majority party status secure more. Funds allocated to states, however, do not generate the same kind of creditclaiming opportunities. While House members care about the distribution of these funds from a policy perspective, they do not appear to distribute them as a political commodity. Coalition Building and Geographic Distribution The 1998 transportation bill also provides a good opportunity to examine which sort of geographic benefit provided a stronger incentive for members to support the committee s bill and to resist amendments that undercut it. Transportation bills almost always pass by a wide margin in Congress, but this does not mean that coalition leaders face no challenges in winning approval of their proposals. In 1997 and 1998 the top priority of House legislators involved in drafting the bill was to dramatically increase spending on transportation, and the 40% boost they sought exceeded the budget resolution s limits by $16 billion (Weisman 1997). This goal put committee leaders at odds with the Republican leadership of both chambers as well as President Clinton (Hosansky 1997a). The process of building a successful coalition took 16 months, during which time the committee suffered one 23 This complaint was registered by Shuster (Cong. Record 1998 H1887, H2014), Rep. James Oberstar (Cong. Record 1998, H2015), Rep. Tom Davis (Cong. Record 1998, H1898), and Rep. Steve LaTourette (Cong. Record 1998, H2016). major defeat on the floor 24 and had to recommend a sixmonth extension of program funds to avert construction stoppages. 25 House committee leaders had two kinds of distributive resources to expand support for their bill: they could allow individual House members to designate highpriority transportation projects in their districts, and they could manipulate the formula that allocated most of the funds so that House members would see increases for their states. 26 They took both approaches from the start of the process. When the House committee began to markup a bill for the first time, the formula distributions to states had already been revised. In addition, in January 1997 long before they began drafting a bill the leadership of the committee invited all House members to submit written requests for individual projects for transportation needs in their districts (Cong. Record 1997, E90). Which approach had a bigger impact on House support for the bill? Evans (1994) has shown that earmarked projects are effective as side payments to sway members to support the broader legislative goals of committee leaders. The question posed here is whether such projects had a larger effect on House members support for the committee s bill than the decisions about state-level funding, which are far more significant for the geographic distribution of transportation funds overall. If the basis of representation in the House affects members incentives in the hypothesized way, members are likely to view changes in state-level spending as not closely connected to their parochial or electoral interests. State-level increases will constitute one policy consideration among others that members take into account as they weigh their support for the proposal, but they provide little credit-claiming opportunities that members can use to shore up their electoral support. As such, they do not constitute side payments for members who need extra incentive to go along with the committee. By contrast, projects earmarked for particular congressional 24 This occurred on May 21, 1997 when House Transportation and Infrastructure Chair Bud Shuster failed to amend the 1997 budget resolution to increase transportation spending beyond the caps imposed by the Republican party leadership. 25 The 1991 transportation authorization expired on September 30, 1997, and the House did not pass its regular reauthorization bill until April 1, It is important to note that the formula distribution is highly manipulable, and it is easy to target funds to individual states. This is not just because of the wide availability of potential formula factors and personal computers to calculate alternatives (Nathan 1993). It is also because regardless of the formula factors Congress adopts the actual funding outcome for every state is determined by the minimum allocation factor, which is a figure that is separately negotiated and specified for each state (GAO 1986).

10 GEOGRAPHIC POLITICS IN THE HOUSE 723 districts clearly impact House members parochial interests and provide easy credit-claiming opportunities. They thus offer a more direct appeal to members electoral selfinterest. If this argument is correct, House members receipt of earmarked projects will be more closely connected to their support of the legislation than state-level funding increases. To assess the effect of geographic benefits on House members support for the bill, I analyze members voting behavior on the legislation. From the point of view of committee leaders, five significant roll-call votes took place on the 1998 transportation reauthorization. 27 The votes are briefly described below, and a + or symbol next to each indicates whether the committee s proposed funding increases for a member s state or district should increase or decrease a member s willingness to vote yes. + Shuster Amendment to the Budget Resolution: Would increase outlays for surface transportation programs to $137 billion, one of the committee leadership s highest priorities. Defeated Spratt (D-S.C.) Amendment to Extend ISTEA: Would reject the committee s bill and extend the existing highway programs for two months. President Clinton favored this amendment. Defeated Graham (R-S.C.) Amendment to Strike High Priority Projects: Would remove all provisions that provide funds for earmarked projects. Defeated Kasich (R-Ohio) Amendment to Turn Back Federal Funds to States: Would return federal transportation dollars directly to the states and reduce the federal gas tax over a four-year period. Defeated Surface Tranportation Reauthorization (BESTEA): Passage of the bill authorizing $219 for surface transportation over six years, a 40% increase. Adopted Logistic regression analyses of these votes were conducted in order to measure the influence of geographic benefits on members voting behavior. The dependent variable is coded 1 for members voting (or pairing) in support of the committee leadership s position and 0 for members voting (or pairing) in opposition. Three variables in these models account for the effect of earmarked 27 Two additional roll-call votes were taken on matters that divided the committee: an amendment offered by Rep. Marge Roukema (R-N.J.) to end the Transportation Department s minority set-aside programs in contracting, and an amendment by Rep. Danny Davis (D-Ill.) to increase the authorization for a welfare-to-work transportation program. Given the subject matter (welfare and affirmative action) these were highly ideological votes that divided both the chamber and the committee along party lines. As expected, analysis of these roll-call votes shows a different pattern than those for which the committee leadership was unified. projects and overall funding increases: Number of Projects, Dollar Value of Projects, and Percent Funding Change for State. 28 Other influences on members votes were also included as control variables. Member of the Committee (coded 1 for members, otherwise 0) takes into account the likelihood that members of the Transportation and Infrastructure committee will be more willing to support the leadership of the committee. Party(coded 1 for Republicans, otherwise 0) captures the partisan influence on the vote. 29 Percent of the Vote in 1996 is included to account for the possibility that marginal members face different pressures than other members. Table 4 reports the maximum-likelihood estimates of the regression coefficients for each of the roll-call votes. Because there is no consensus on the best measure, goodness-of-fit in these models is reported in three ways: percent reduction in error, percent of cases correctly predicted, and the Nagelkerke R 2 (Pampel 2000, 48 53). As expected, the control variables in the models have some predictive power. Members of the Transportation and Infrastructure committee tended to support their committee s leadership more than other members. The coefficients on Party reveal that the committee generally drew more support from Democrats than Republicans. 30 Although the effect is not great, marginal members were somewhat more likely to oppose the committee and to favor the anti-pork position taken by editorialists (even though they received more earmarked projects than average). 31 On every roll-call vote, members who received earmarked projects were more likely to support the committee s position, and in each model the coefficient for Dollars in Projects was statistically significant in the 28 The first two variables are the total number and dollar value of the projects each member s district received. The percent-change variable reflects the percentage increase or decrease that the member s state would receive under the new formula allocation above the amount his state had received under existing law. 29 Ideology variables were also included in the model in lieu of party without changing the results significantly. It was not possible to include variables for both party and ideology in the model, given the very high correlation Pearson s r of at least.93 (p <.001) between the two. 30 The exception was on the Spratt amendment to extend ISTEA, the one vote for which Party was not statistically significant. Clinton s support of this amendment appears to have swayed some Democrats to support the President rather than the committee on this vote. 31 The coefficient on percent of the vote shows that marginal members opposed the committee more often than others on every vote, but it was only statistically significant (p <.05) for two roll-call votes.