Foreign Aid and Domestic Politics: Voting in Congress and the Allocation of USAID Contracts across Congressional Districts.

Robert K. Fleck [*]

Christopher Kilby [+]

This paper investigates the relationship between congressional support for foreign aid and the distribution of United States Agency for International Development (USAID) contract spending across congressional districts within the United States. The extent to which such a relationship matters has become increasingly important in recent years, as the end of the Cold War and the advent of the Republican-controlled Congress have eroded the traditional base of support for foreign aid. We develop a model to illustrate how the distribution of contract spending could be used to increase support for foreign aid, but at the expense of development impact, in effect trading quality for quantity. Data on domestic foreign aid contract spending and votes in the 104th Congress House of Representatives allow us to test whether the geographic distribution of USAID contract spending within the United States is consistent with a systematic attempt to build support for foreign aid in Congress. Econometric results provide little e vidence of such attempts, apparently because voting on this issue is insensitive to the distribution of contract spending.

1. Introduction

Although U.S. foreign aid historically has catered to national security, commercial, and humanitarian interests, in recent years aid agencies have increasingly focused on domestic commercial benefits when presenting their case to lawmakers. The United States Agency for International Development (USAID) describes itself as a provider of "direct economic benefits" to "almost every state in the Union," a claim that it supports with state-by-state data on its Web site (USAID 1998b, c). This direct appeal to commercial interests is not surprising given the demise of the Cold War--era foreign aid coalition and the advent of the Republican-controlled Congress. After the fall of the Berlin Wall, a Democratic Congress presided over a substantial decline in foreign aid spending. A Republican Congress in 1995 shifted the debate from simply cutting spending to dismantling agencies. Although USAID survived, it has been forced to reorganize, drastically reduce its size and scope, and campaign actively for funding.

This paper examines the domestic political economy of foreign aid, investigating the relationship between support for foreign aid in the House of Representatives and the geographic distribution of USAID contract spending across congressional districts within the United States. We develop a theoretical model to illustrate how manipulating the distribution of contract spending could increase political support for foreign aid, though at the expense of development impact, in effect trading quality for quantity. Data on domestic foreign aid contract spending and votes in the 104th Congress House of Representatives allow us to test whether the distribution of USAID contract spending within the United States is consistent with a systematic attempt to build support for foreign aid in Congress. The foreign aid literature posits a quality-quantity trade-off; this paper provides a theoretical model of a mechanism through which such a trade-off might operate and econometric evidence of its magnitude.

Anecdotal evidence and univariate statistics do appear to link spending and support for aid. In addition to USAID's sales pitch, other anecdotal evidence includes American firms lobbying Congress for profitable foreign aid policy (Morgan 1995), members of Congress pressuring USAID to obtain contracts for their constituents (Fiorina 1989, pp. 64-5), and, despite its bidding process, USAID influencing contract awards (Kamen 1996). In the votes we analyze, House Democrats are five times more likely to vote in favor of aid, and they enjoy two times as much USAID spending in their home districts as do House Republicans. Republican members of the Foreign Operations Subcommittee of the House Appropriations Committee are three times more likely to vote in favor of aid and get three times as much USAID spending as do other Republicans.

However, multivariate analysis of USAID spending data covering over 1000 contractors and 3000 contracts points to a different conclusion. We find only weak links between congressional voting and the geographic distribution of contract spending. After accounting for differing levels of contractor qualifications across congressional districts, the level of contract spending does not depend substantially on the representative's position on foreign aid (strong supporter, swing voter, or opponent) or on other political variables, such as committee membership, tenure, or party. The apparent explanation for this is that aid contract spending has little influence on congressional voting. Controlling for party, general liberal-conservative divisions in Congress, and constituency characteristics, the level of contract spending in representatives' home districts does not substantially influence their voting positions. Washington-area "Beltway" Republicans are the exception, as their districts receive high levels of con tract spending, and they vote in favor of aid much more often than do typical Republicans.

These results shed light on the emerging post--Cold War aid regime. Although USAID activities do provide "direct economic benefits" to "almost every state in the Union," there is little indication that USAID systematically manipulates the allocation of contracts in an attempt to garner political support. This is the rational outcome. Given that the geographic distribution of USAID contracts appears to be an ineffective tool for strengthening political support for foreign aid, efforts to forge a new coalition around commercial interests are unlikely to succeed. The stylized fact of a "quality-quantity" trade-off does not hold in this case; manipulating the geographic distribution of USAID contracts--thereby including less qualified contractors--is unlikely to increase aid funding.

The paper proceeds as follows. Section 2 provides historical and institutional background. Section 3 develops the theoretical model of spending. Section 4 presents the empirical analysis. Section 5 concludes.

2. Political Coalitions and Support for Foreign Aid

The amount of foreign aid, as well as how and where that money is spent, depends critically on the degree to which aid spending serves a variety of interests and thereby helps to maintain a coalition. From the end of World War II to the end of the Cold War, support for foreign aid relied heavily on national security and domestic economic interests as well as humanitarian concerns. National security was an explicit objective of foreign aid and an essential factor in maintaining foreign aid budgets (Krueger, Michalopoulos, and Ruttan 1989; Griffin 1991; Zimmerman and Hook 1996). [1] Similarly, the desire to serve domestic economic interests was explicitly acknowledged. Presidential administrations routinely highlighted domestic economic benefits of foreign aid. [2] Even at such watershed points as Truman's Point Four program and Kennedy's establishment of USAID, presentations to Congress emphasized economic advantages to the United States (Daniels 1951; Reeves 1988).

USAID has explored ways of strengthening this coalition. In a 1984 volume, Rice and Donald (coeditors of USAID's Development Digest) discuss "constituency-building" through development education and strengthening ties with private voluntary organizations, universities, the banking community, and other commercial organizations (Rice and Donald 1984). Another angle considered by USAID is the potential importance of the geographic distribution of contract spending. As far back as the 1970s, USAID tracked contract spending by congressional district (Rice and Donald 1984, p. 352). More recently, it has established a database to substantiate its claims of providing widespread benefits to American firms and workers. [3]

The coalition of groups with humanitarian, commercial, and security concerns has created an uncomfortable partnership with often conflicting objectives (Wood 1996; Zimmerman and Hook 1996). Among advocates of humanitarian aid, the ability to build political support for foreign assistance based on security and domestic economic interests is viewed as a double-edged sword. Practices that promote donor interests--such as tying aid to purchases of the donor country's goods and services and the use of mixed credits--clearly reduce the real value of aid to the recipient relative to dollar amounts (Jepma l991). [4] Other donor-driven biases, such as intensive use of capital and imports, further reduce the development effectiveness of aid (e.g., Tendler 1975; James 1995). There is a quality-quantity trade-off when linking aid to the donor country's economic interests increases the total aid budget but reduces the value to the recipient of a dollar's worth of aid (Jay and Michalopoulos 1989).

The fall of the Berlin Wall generated considerable uncertainty within the development establishment about the future of foreign aid. Some predicted a boon from the peace dividend and new opportunities in Eastern Europe and the former Soviet Union. [5] Others foresaw a withering of foreign aid as the national security motive for aid diminished. [6] This uncertainty ended with the success of conservative Republicans in the 1994 congressional elections: The number of USAID staff fell by over 3,000, the number of countries with USAID programs was projected to fall by 45, and the number of foreign missions was projected to drop by 40 (Lippman 1996). [7] According to USAID, "U.S. foreign assistance programs are at the lowest levels, in real dollar terms, that they have been in over 50 years" (USAID 1998b).

The real crisis for American foreign aid began in 1995. Republican control of Congress brought outspoken opponents of foreign aid into key committee positions. In the House, Sonny Callahan (R-Ala.)--an admitted aid neophyte who campaigned on promises to limit foreign aid--headed the Foreign Operations Subcommittee of the Appropriations Committee. The chairman of the Senate Foreign Relations Committee, Jesse Helms (R-N.C.), armed with leaks from USAID, launched a public campaign against foreign aid in general and USAID in particular, circulating press releases and newsletters with such titles as "Captured Enemy Documents" and "Jesse Helms' Believe It or Not!" Capitalizing on weakening support for foreign aid, this campaign changed the terms of the debate from trimming the president's aid requests to abolishing the entire foreign aid apparatus. The success of Helms's efforts was driven by one central factor; in the words of USAID then-director J. Brian Atwood, "The Cold War consensus in support of adequate spe nding for international programs no longer exists" (Lippman 1996).

A pivotal issue in the FY96-97 Foreign Aid and State Department Authorization Bill was the future of USAID. The House version, passed on June 8, 1995, after heated debate, called for the abolishment of USAID as an independent agency (folding it into the State Department) and a reduction in its staff and funding. Helms's Senate bill, which also provided a detailed State Department restructuring plan and would have undermined the position of foreign aid, reached the floor later in the summer but was halted by threat of a Democratic filibuster. Helms countered by holding up 18 ambassadorial appointments until John Kerry (D-Mass.) brokered a deal calling for large administrative cuts but allowing an independent USAID. The Senate finally passed the bill on December 14. In March 1996, both the House and the Senate adopted a conference report eliminating USAID, the United States Information Agency (USIA), and the Arms Control and Disarmament Agency (ACDA) but permitting the president to waive elimination for any tw o agencies. The president vetoed the bill on April 12, 1996, alleging that elements of the State Department reorganization interfered with presidential decision making. On April 30, a House override attempt failed. No foreign aid authorization bill has been enacted since 1985.

Appropriations bills have more often become law, but passage has been complicated by controversy over spending levels and seemingly intractable abortion issues. In June 1995, when an already lean foreign aid appropriations bill reached the House floor, the Republican majority pushed through a rule allowing unlimited amendments, a move widely seen as an attempt to allow further paring down of aid spending. After a number of amendments, including Christopher Smith's (R-N.J.) reintroduction of Reagan-era "Mexico City Restrictions" related to abortion, the House passed the bill on July 11. [8] On September 21, the Senate passed an amended version of the House bill, removing the abortion restrictions and increasing funding by about $500 million. The conference committee quickly reached a compromise on the level of funding but was unable to resolve the conflict over abortion restrictions. [9] After another round of voting, a compromise was reached late in January 1996. This compromise attached the spending measure to the Interim Spending Bill, dropped the abortion restrictions, and effectively reduced spending on USAID population programs by 35%. This bill was passed by the House on January 25, then passed by the Senate and signed by the president on January 26. [10]

The appropriations process began again in May 1996, and the House passed the Fiscal 1997 Foreign Operations Appropriations Bill on June 11. This bill called

for further reductions in foreign aid and again included Smith's abortion restriction. The Senate passed the bill on July 26 but included amendments to eliminate Smith's restriction and to increase population program funding. In a complex deal to resolve these differences, the conference report replaced abortion restrictions with a delay in the release of family planning funds. To complete the appropriations process before the start of the new fiscal year, the House passed foreign aid funding as part of the Omnibus Spending Bill on September 28, 1996. The Senate passed the bill on September 30, and the president signed the bill on the same day.

The congressional attack on USAID's budget and bureaucratic base left the agency, in the words of the Washington Post, "shriveling like a sun-dried tomato" (Lippman 1996). In an effort to defend itself and build support for foreign aid, USAID has increasingly advertised its role as a provider of "direct economic benefits" and emphasized the geographic distribution of its contract spending within the United States. These "direct economic benefits" are sizable; of the $12.9 billion in contracts active during 1995, $11 billion were awarded to firms based in the United States. [11] That the geographic distribution of contracts matters in at least some cases seems clear from anecdotal evidence, but whether it plays an important role more generally is an empirical question.

3. A Model of the Distribution of Spending

This section presents a model to explain the distribution of contract spending across congressional districts. A government agency (USAID in our case) with discretion to allocate spending across congressional districts may have an incentive to use the allocation to strengthen congressional support. Our model examines how the effects of those incentives depend on (i) cross-district differences in contractor qualifications, (ii) differences in legislator preferences, and (iii) the extent to which "direct economic benefits" for legislators' home districts increase their support for the agency.

We assume that the agency attempts to maximize its budget subject to a simple majority legislative approval constraint. [12] Following the logic of the "agenda setter" model (e.g., Romer and Rosenthal 1978; Rosenthal 1990), we take the agency to be rational and forward looking when proposing policies that the legislature can either accept or reject in favor of the status quo. [13] Stated formally, the model is as follows.

Parameters and Variables

[alpha] Actual budget share for each of n districts, where [alpha] = ([[alpha].sub.1],...,[[alpha].sub.n]).

[[alpha].sup.*] Technically efficient budget shares, where [[alpha].sup.*] = ([[[alpha].sup.*].sub.1],...,[[[alpha].sup.*].sub.n]). [[alpha].sup.*] reflects the level of contractor qualifications in each district. Deviating from [[alpha].sup.*] shifts contracts from qualified contractors in one district to less qualified contractors in another district, reducing the impact of spending.

B Actual budget.

[B.sup.SQ] Status quo budget.

[B.sub.i] Measures legislator i's attitude toward spending.

[omega] Weight given to own-district spending in legislator's utility function.

Assumptions

A.1 The agency is the agenda setter. The agency is rational and forward looking and seeks to maximize its budget. The agency proposes the share of contract spending for each district ([alpha]) and the overall budget for contract spending (B). If the legislature accepts the proposal, the proposal becomes policy. If the legislature rejects the proposal, policy reverts to the status quo.

A.2 The status quo agency budget arises from legislators' preferences. If policy were to revert to the status quo, the agency would allocate shares of funds across districts according to [[alpha].sup.*]. [14] The status quo agency budget ([B.sup.SQ]) is determined as the level of spending that would be the Condorcet winner if [alpha] = [[alpha].sup.*].

A.3 Legislators' votes depend on the total level of spending and the allocation of spending across districts. Each legislator votes to accept the agency's proposal if and only if his or her utility from the agency's proposal is at least as great as the utility from the status quo. The utility of policy is

[U.sub.i](B, [alpha]) = [omega][[alpha].sub.i]B - [(B - [B.sub.i]).sup.2] - [[[sigma].sup.n].sub.k=1] [([[alpha].sub.k] - [[[alpha].sup.*].sub.k]).sup.2].

The first term, [omega][[alpha].sub.i]B, indicates the value to the legislator of direct benefits to his or her home district. The second term, -[(B - [B.sub.i]).sup.2], reflects the legislator's preferences toward aid spending in general. The final term, -[[[sigma].sup.n].sub.k=1] [([[alpha].sub.k] - [[[alpha].sup.*].sub.k]).sup.2], indicates that, all else constant, legislators prefer the efficient use of contractors ([alpha] = [[alpha].sup.*]). [15]

Policy Outcomes

The agency solves

[max.sub.B,[alpha]] [U.sub.A](B) s.t. [U.sub.i](B, [alpha]) [greater than or equal to] [U.sub.i]([B.sup.SQ], [[alpha].sup.*]) for a majority of the legislators,

where

[U'.sub.A] [greater than] 0; [omega] [greater than or equal to] 0; [[[sigma].sup.n].sub.k=1] [[alpha].sub.k] = 1; [[[sigma].sup.n].sub.k=1] [[[alpha].sup.*].sub.k] = 1; [[alpha].sub.k] [greater than or equal to] 0, [[[alpha].sup.*].sub.k] [greater than or equal to] 0 [forall] k;

[B.sub.i] [greater than or equal to] 0 [forall] i.

Figure 1 illustrates the resulting distribution of contract spending. [16] Legislators are ordered along the horizontal axis in terms of decreasing support for aid under the status quo. The vertical axis measures the deviation of the actual district share of the budget from the technically efficient or status quo share. If no weight is given to own-district contract spending ([omega] = 0), there is no deviation ([alpha] - [[alpha].sup.*] = 0), and the distribution of spending depends only on contractor qualifications. However, if contract spending does matter ([omega] [greater than] 0), legislators divide into three groups from left to right. Loyalists vote for the budget despite getting a reduced share of spending. Swing, or Pivotal, Voters vote for the budget because of more favorable treatment. Opponents vote against the budget and also get a reduced share of spending. Among the middle group, the agency must give the most marginal legislators (those who would favor aid spending the least under the status quo) the greatest extra share of the budget in order to win their votes. With [omega] [greater than] 0, the distribution of contract spending depends on both the level of contractor qualifications in the district and the legislator's attitude toward aid spending.

The model illustrates a quality-quantity trade-off when [omega] [greater than] 0. The agency can garner legislative support for a larger budget by distorting the distribution of contract spending in favor of legislators who would otherwise be marginally opposed to the larger budget. However, distorting the distribution of contract spending away from the technically efficient shares reduces the per dollar impact of spending.

4. Empirical Analysis

This section tests the spending model's implications. We estimate a spending equation that can identify favoritism toward swing legislators' districts and a voting equation that measures the impact of spending on voting. We start by describing the variables used in the analysis.

The Data: SPENDING, AIDSCORE, and Qualifications

The unit of analysis throughout is the representative/congressional district. Our measure of contract spending is derived from USAID Yellowbook data (USAID 1996, 1997). SPENDING covers all USAID contracts active in either fiscal 1995 or 1996. We use contractors' nine-digit ZIP codes to aggregate data on individual contracts to totals by congressional districts. SPENDING is an annualized dollar amount expressed in per capita terms.

Table 1 provides statistics on SPENDING. Average district contract spending is roughly $9 per capita per year with an average of $5.90 in Republican districts and an average of $12.50 in Democratic districts. Within the Beltway, spending averaged $132 in Republican districts and $154 in Democratic districts. [17]

We use roll-call voting records from the 104th Congress to identify legislators' positions on foreign aid, selecting the 10 House votes with clear pro- and anti-aid positions. Figure 2 presents the chronology of the votes in relation to the debates discussed in section 2 (see Appendix B for details). AIDSCORE, the pro-aid percentage of a representative's votes, summarizes representatives' positions on foreign aid. As Table 2 shows, the average AIDSCORE of 49 masks substantial differences between parties and across the Beltway. AIDSCORE averages 18 for Republicans and 85 for Democrats overall, 70 and 98 within the Beltway. The difference between Republicans and Democrats inside and outside the Beltway is also apparent in correlations between SPENDING and AIDSCORE. For Republicans, the correlation coefficient is 0.245 overall but -0.011 outside the Beltway. For Democrats, the correlation coefficient is 0.128 overall and 0.119 outside the Beltway.

Contractor qualifications also influence the distribution of USAID spending. Several district-level measures reflect the size of the pool of qualified potential USAID contractors and hence act as proxies for contractor qualifications. The most obvious is the level of education because advanced degrees in economics, development studies, accounting, engineering, law, public health, and other fields are often prerequisites to winning a bid. The level of employment in the public and nonprofit sectors may also indicate the size of the labor pool qualified for USAID contract work. Some nonprofit organizations (or nongovernmental organizations [NGOs]) specialize in Third World development, some even in USAID contracting. Appendix B defines these variables and gives summary statistics.

The Spending Equation

To test the implications of the spending model in section 3, we estimate a multivariate tobit with SPENDING as the dependent variable. [18] We use two different measures of legislators' positions on foreign aid--AIDSCORE and predicted AIDSCORE--to verify the robustness of our findings. Using predicted values of AIDSCORE avoids a potential endogeneity problem, as the estimated effect of AIDSCORE on SPENDING could be conflated with the effect of SPENDING on AIDSCORE. [19] Predicted AIDSCORE is the fitted values from an ordinary-least-squares regression using Poole and Rosenthal's NOMINATE scores as explanatory variables. [20] Because the two variables yield similar results, in the interest of expedience we simply refer to AIDSCORE in the following discussion.

The spending model predicts that swing legislators' districts may receive a disproportionate share of spending. Controlling for qualifications and proximity, this would give rise to an inverted-U, or "shark fin," shape (as in Figure 1) relation between SPENDING and AIDSCORE with a peak near the median AIDSCORE value. We test for this in the SPENDING equation in two ways. First, with a quadratic AIDSCORE specification, the inverted U translates into a positive linear term and a negative squared term that peaks near the median AIDSCORE. Second, with a piecewise linear function broken at the median AIDSCORE, favoritism of swing legislators should be directly apparent.

Table 3 presents estimation results. Each specification includes proxies for contractor qualifications as well as BELTWAY, the location variable discussed previously. There is no evidence of favoritism of swing legislators in any of the four specifications. Rather than an inverted U, the quadratic specification results in a U shape (negative linear term and positive quadratic term), but the coefficients are statistically insignificant. [21] Likewise, the piecewise linear specification results in a V shape, again statistically insignificant. [22] BELTWAY and proxies for contractor qualifications have the expected positive signs in all specifications.

For completeness, we briefly consider some alternative political models of spending. The agency could reward loyal supporters (an optimal strategy in some repeated settings), implying a positive relation between AIDSCORE and SPENDING after controlling for contractor qualifications and location. However, neither the results in Table 3 nor simple linear specifications provide much evidence of such a strategy. [23] Agency objectives may be furthered by favoring influential representatives (e.g., those from the pro-aid party, those with the longest tenure, or key committee members); again, this pattern is not evident in the data. [24] Finally, the agency might spread contract dollars more evenly to develop a broad base of support and avoid creating strong opponents, a strategy implicit in publicizing the "direct economic benefits" for every state and in agency attempts to diversify its contractor base away from the Beltway. Again, we test for this possibility but find little support. [25]

The Voting Equation

Apparently, the distribution of contract spending does not depend systematically on political variables. The theoretical model provides a potential explanation: strong favoritism of swing legislators is absent if [omega], the marginal effect of contract dollars on legislative support, is small. We now turn to this possibility, estimating a voting equation with AIDSCORE as the dependent variable.

Our goal here is to see whether variations in own district aid contract spending cause representatives to vote differently than they would otherwise have voted on foreign aid-related measures. A representative's overall voting record is perhaps the best predictor of the representative's votes in the absence of home district benefits. Poole and Rosenthal's NOMINATE scores are very useful for this purpose, as they describe legislators' locations in a two-dimensional, vote-predicting space. Scores range from --1 to 1, with Republicans tending to have high scores on the X dimension (Poole and Rosenthal 1997). [26] We include BELTWAY to capture proximity effects and various district demographic, educational and occupational variables to reflect the influence of constituent preferences not already summarized by the NOMINATE scores. [27]

Because the distribution of the dependent variable AIDSCORE is censored, we estimate a tobit. A heteroskedasticity correction allows for cases where a representative did not vote on all measures. In view of the sharp contrast between Republican and Democrat voting patterns, we estimate separate equations for each party. [28]

Table 4 reports the results. In contrast to the bivariate correlation in Table 2, SPENDING does not have a significant effect on voting for either Republicans or Democrats. Among Republicans, the bivariate correlation between SPENDING and AIDSCORE is largely explained by differences across the beltway; SPENDING is insignificant in any specification, including BELTWAY. [29] Among Democrats, the key is a split between liberals and conservatives within the party; the estimated coefficient on SPENDING is small and insignificant in any specification including XCORD, the primary NOMINATE dimension. [30]

A clear picture emerges from these estimates. Once we consider representatives' overall voting records and constituent characteristics, the geographic distribution of USAID contract spending across congressional districts does not appear to influence congressional support for aid in general. The possible exceptions are Beltway Republicans whose districts benefit from high levels of contract spending and who are much more supportive of aid than are typical Republican representatives. We find no evidence that Democrats from high-contract-spending districts are substantially more supportive of foreign aid than Democrats from low-contract-spending districts after controlling for overall voting record. Relating these findings to the theoretical model, the real-world counterpart to is small, and, with [omega] small, efforts to increase support for foreign aid by manipulating the allocation of contract spending would likely be ineffective.

5. Conclusion

A variety of factors--USAID's promotional literature, anecdotes about political influence, and the simple correlation between contract spending and congressional votes--suggest links between the geographic distribution of USAID contract spending within the United States and support for foreign aid in Congress. Yet econometric analysis, based on data for all USAID contracts active during the 104th Congress, reveals only weak links. Once we control for differences in contractor qualifications across districts, the level of contract spending does not depend substantially on the representative's support for foreign aid or other political variables. Although USAID activities do provide "direct economic benefits" to "almost every state in the Union," there is little indication that USAID systematically manipulates the allocation of contracts in an attempt to garner political support. Furthermore, the level of contract spending in a representative's home district has at most a small effect on his or her support for aid (except in the case of Beltway Republicans).

The larger question raised by this research is whether domestic economic benefits significantly increase support for foreign aid programs. Although we explored only one dimension of the issue, we find little evidence that the economic benefits of aid translate into support for foreign aid in Congress. Traditional pork-barrel politics, which link votes to the distribution of benefits across districts, are not apparent in the data. If the commercial benefits of foreign aid programs have little effect on support for aid, a coalition that substitutes commercial interests for waning national security concerns is unlikely to win increased funding. Yet the costs of such a coalition may be high. Catering to commercial interests is likely to reduce development effectiveness and, especially in the long run, undermine public support (Jay and Michalopoulos 1989; Zimmerman and Hook 1996). In sum, trading away quality is unlikely to obtain a substantially higher quantity of foreign aid.

(*.) Department of Agricultural Economics and Economics, Montana State University, Bozeman, MT 59717, USA; E-mail [e-mail]; corresponding author.

(+.) Department of Economics, Vassar College, Poughkccpsie, NY 12604, USA; E-mail [e-mail].

For helpful comments, we wish to thank an anonymous referee, Weiner Baer, Beth Davenport, and Peter Kilby. We extend special thanks to Sally Scholz.

Received January 1999; accepted March 2000.

(1.) The literature on donor motives confirms the importance of security concerns in the allocation of U.S. foreign aid during the Cold War (e.g., Wittkopf 1972; Maizels and Nissanke 1984). More recent evidence suggests a diminished influence of the security motive in U.S. aid allocations since the fall of the Berlin Wall (Meernik, Krueger, and Poe 1998).

(2.) Because of foreign aid's apparent utility as a foreign policy tool, the executive branch has generally taken a more favorable view than has Congress. Democratic administrations have generally shown more support for foreign aid than have Republican administrations (Eggleston 1987).

(3.) In support of the claim that USAID contracts provide "direct economic benefits" to "almost every state in the Union," USAID lists state-by-state data on its Web site. For example, a section titled "FOREIGN AID FOR ALABAMA" explains, "The principal beneficiary of America's foreign assistance programs has always been the United States. Close to 80 percent of the U.S. Agency for International Development's (USAID's) contracts and grants go directly to American firms." Under "CREATING JOBS IN ALABAMA," the document lists all USAID contracts active in Alabama over a one-year period, including contractor, location, and contract amount (USAID 1998a).

(4.) "Mixed credits"--mixing subsidized aid loans with commercial loans to finance commercial projects--allow a domestic firm to win contracts it might otherwise not. This leverages the aid budget but blurs the line between assistance and trade, usually promoting the latter at the expense of the former.

(5.) For example, Zimmerman and Hook (1996, p. 57) point Out USAID's optimism in a 1994 report stating: "With the end of the Cold War, the international community can now view the challenge of development directly, free from the demands of superpower competition."

(6.) In 1991, Griffin stated flatly that "foreign aid is a product of the Cold War" and predicted that, without "its raison d'etre," aid from OECD countries would decline rapidly (Griffin 1991, pp. 647, 670).

(7.) The foreign mission figure was based on a projection of closures by the year 2000.

(8.)The Mexico City restrictions, imposed by Reagan in 1984, barred "U.S. aid to international organizations that performed or 'actively promoted' abortions" (Congressional Quarterly 1997, pp. 10-9). These restrictions expanded on the already existing 1973 law prohibiting the direct funding of abortions with foreign assistance funds.

(9.)The House bill called for $11.9 billion in total foreign operations spending, the Senate bill called for $12.4 billion, and a conference report called for $12.1 billion. The $5.74 billion subtotal for USAID in the conference report was actually lower than in either the House version ($5.76 billion) or the Senate version ($5.91 billion) (Congressional Quarterly 1996, pp. 11-41).

(10.)In the end, the appropriation was $12.4 billion (Congressional Quarterly 1997, pp. 10-49).

(11.) Derived from USAID (1996).

(12.) This characterization is reasonable even if the agency has other objectives. If, for example, the agency cared solely for the impact of its programs, it would still be concerned with the size of its budget. What matters for the model is that the policy outcome may depend on trade-offs between the size of the budget and the per dollar effectiveness of spending. For simplicity, we introduce this trade-off through the legislators' objective functions.

(13.) Although our model introduces a new quality-quantity trade-off, it draws on ideas in the literature on incentives to maintain the support of pivotal or loyal supporters (e.g., Wright 1974; Kiewiet and McCubbins 1985; Cox and McCubbins 1986; Stratmann 1992; Grier, McDonald, and Tollison 1995; Fleck 1999).

(14.) This is equivalent to a lexicographic agency objective function with budget size first and the per dollar impact of spending second.

(15.) There is a substantial literature that addresses the issue of why legislators vote the way they do (see, e.g., Goff and Grier 1993; Coates and Munger 1995; Bender and Lott 1996). In our model, legislators' preferences may reflect constituency preferences and/or constituent-independent preferences (e.g., legislators' personal ideological beliefs). Similarly, legislators may act as voters desire (in the manner of "delegates") and/or based on their own judgment (in the manner of "trustees").

(16.) See Appendix A for solution method.

(17.) Proximity to USAID headquarters may be important for a number of reasons. The terms of USAID contract bidding may work to the advantage of better connected Beltway firms that can use their contacts to find out what sort of proposals the agency favors and what projects are in the pipeline. Location decisions of frequent contractors reinforce this link.

(18.) We use a tobit estimation because SPENDING is zero in 123 districts.

(19.) In terms of the spending model, predicted AIDSCORE reflects a representative's position on foreign aid under the status quo (i.e., [[B.sup.0].sub.i]), a function of [B.sub.i] (the legislator's attitude toward aid spending in general) and [[[alpha].sup.*].sub.i] (district qualifications). AIDSCORE itself reflects the representative's position on foreign aid given the actual distribution of contract spending [alpha] (SPENDING) and hence may be endogenous.

(20.) Poole and Rosenthal's NOMINATE scores describe legislators' locations in a two-dimensional, vote-predicting space. These data provide an excellent predictor of the representative's votes in the absence of home district benefits. There is a 0.916 correlation between predicted AIDSCORE and actual AIDSCORE. Given the purpose of these data for this paper, we make no effort to ascertain the extent to which NOMINATE scores and predicted AIDSCORE reflect constituent preferences, so-called shirking by legislators, or constituent-independent preferences (e.g., personal ideology) held by legislators.

(21.) Likelihood ratio tests for joint significance yield p-values of 0.186 for the actual AIDSCORE variables and 0.372 for the predicted AIDSCORE variables.

(22.) Likelihood ratio tests for joint significance yield p-values of 0.404 for the actual AIDSCORE variables and 0.497 for the predicted AIDSCORE variables. We find no evidence that estimating quadratic or piecewise linear effects imposes overly restrictive functional forms. Estimating a spline function to allow a kink for a potential loyal-swing legislator division (as in Figure 1) produces statistically insignificant effects over the entire range of possible locations for the swing-loyal division; the specifications coming closest to statistical significance have p = 0.308 for AIDSCORE and p = 0.420 for predicted AIDSCORE. Furthermore, residual plots show no apparent relationship to support for foreign aid.

(23.) A linear specification yields t-statistics of 1.34 on AIDSCORE and 1.26 on predicted AIDSCORE.

(24.) Most surprising is the party variable, Despite Democrats' districts averaging twice as much contract spending as Republicans', party is not a significant determinant of contract spending once we control for qualifications.

(25.) The distributive politics literature refers to this type of something-for-everyone allocation as "universalism." Our test is based on the following logic. If USAID attempts to spread contract spending more evenly between states but still employs at least minimally qualified contractors, the qualifications of other districts within a state will have a negative effect on how much contract spending a district receives. Districts in less qualified states should receive more contracting dollars than equivalent districts in more qualified states. In some specifications, these estimated effects do provide weak evidence of state-level universalism: Spending in a district is lower if the other districts in the state have high percentages of nonprofit organization employment. However, results are weak at best and not robust.

(26.) Poole and Rosenthal generally interpret the first dimension (XCORD), the primary dimension of cleavage on roll-call voting, as party loyalty. The second dimension (YCORD) often captures how Congress votes on issues for which the division is not well described by party lines.

(27) Public opinion studies point to education and personal ties as important factors in building support for aid spending (USAID 1998b). Public and nonprofit-sector employees may be more supportive of government spending programs such as foreign aid.

(28) In the tobit model, one can view AIDSCORE as a scale indicating the degree of support for aid. A score of 100 indicates that the representative supports aid at least as much as the measures before the House allow, while a score of 0 indicates that the representative's level of support for aid was as low as or lower than the choices allowed. Since AIDSCORE is an average, the heteroskedasticity correction is [square root][n.sub.i] where [n.sub.i] is the number of votes cast by representative i.

Using a two-sided tobit, we reject the pooled equation (i.e., including Republicans and Democrats) at the 95% confidence level, Although estimating separate equations does not make use of interparty variation, we cannot distinguish Democratic support for aid as a party position and in support of the presidency from party support for aid due to the higher average level of spending in Democratic districts. Only the pooled equation requires a two-sided tobit because all those voting consistently for aid (106) are Democrats, and all those voting consistently against aid (48) are Republicans.

(29.) The results reported are robust to specification, variable definition, sample, and estimation method. Without BELTWAY, SPENDING is significant in all Republican specifications we examine. Using a bachelors degree criterion rather than advanced degree for education makes little difference. Other constituent characteristics, such as those related to income (e.g., per capita income or poverty), proved inconsequential. Dropping all observations with SPENDING [greater than or equal to] 100 or including the square root of spending does not change the results substantially. Restricting the sample to representatives who cast votes on all 10 measures does not change the results substantially; eliminating the heteroskedasticity correction also has little effect. Estimation by ordinary or weighted least squares produces similar results.

(30.) Dropping BELTWAY does little to the estimated effects of SPENDING, as do the other changes in sample and specification mentioned in the previous note.

References

Bender, Bruce, and John R. Lott, Jr. 1996. Legislator voting and shirking: A critical review of the literature. Public Choice 87:67-100.

Coates, Dennis, and Michael Munger. 1995. Legislative voting and the economic theory of politics. Southern Economic Journal 61:861-72.

Congressional Quarterly. 1996. 1995 CQ almanac. Washington, DC: Congressional Quarterly.

Congressional Quarterly. 1997. 1996 CQ almanac. Washington, DC: Congressional Quarterly.

Cox, Gary W., and Mathew D. McCubbins. 1986. Electoral politics as a redistributive game. Journal of Politics 48:370-89.

Daniels, Walter M., ed. 1951. The Point Four program. New York: The H. W. Wilson Company.

Eggleston, Robert. 1987. Determinants of the levels and distribution of PL 480 food aid: 1955-79. World Development 15:797-808.

Fiorina, Morris P. 1989. Congress: Keystone of the Washington establishment. 2nd edition. New Haven, CT: Yale University Press.

Fleck, Robert K. 1999. Electoral incentives, public policy, and the New Deal realignment. Southern Economic Journal 65:377-404.

Goff, Brian L., and Kevin B. Grier. 1993. On the (mis)measurement of legislator ideology and shirking. Public Choice 76:5-20.

Grier, Kevin B., Michael McDonald, and Robert D. Tollison. 1995. Electoral politics and the executive veto: A predictive theory. Economic Inquiry 33:427-40.

Griffin, Keith, 1991. Foreign aid after the Cold War. Development and Change 22:645-85.

James, Jeffrey. 1995. The state, technology, and industrialization in Africa. London: Macmillan.

Jay, Keith, and Constantine Michalopoulos. 1989. Donor policies, donor interests, and aid effectiveness. In Aid and development, edited by Anne O. Krueger, Constantine Michalopoulous, and Vernon W. Ruttan. Baltimore: The Johns Hopkins University Press, pp. 72-82.

Jepma, Catrinus J. 1991. The tying of aid. Paris: Development Center Studies, OECD.

Kiewiet, D. Roderick, and Mathew D. MeCubbins. 1985. Congressional appropriations and the electoral connection, Journal of Politics 47:59-82.

Kamen, Al. 1996. Hearing AID. Washington Post, June 17, p. A-IS.

Krueger, Anne O., Constantine Michalopoulous, and Vernon W Ruttan, eds. 1989. Aid and development. Baltimore: The Johns Hopkins University Press.

Lippman, Thomas W. 1996. It's a smaller world after budget shrinkage. Washington Post, March 21, p. A-15.

Maizels, Alfred, and Machiko Nissanke. 1984. Motivations for aid to developing countries. World Development 12:879-900.

Meernik, James, Eric Krueger, and Steven Poe. 1998. Testing models of U.S. foreign policy: Foreign aid during and after the Cold War. Journal of Politics 60:63-85.

Morgan, Dan. 1995. Where foreign aid gets down to business. Washington Post, June 12, p. A-4.

Poole, Keith T, and Howard Rosenthal. 1997. Congress: A political-economic history of roll call voting. New York: Oxford University Press.

Poole, Keith T, and Howard Rosenthal. 1998. Web: http://voteview.gsia.cmu.edu/dwnl.htm.

Reeves, T. Zane. 1988. The politics of the Peace Corps & Vista. Tuscaloosa: University of Alabama Press.

Rice, Andrew, and Gordon Donald, Jr. 1984. A constituency for foreign assistance. In U.S. foreign assistance: Investment or folly?, edited by John Wilhem and Gerry Feinstein. New York: Praeger, pp. 342-61.

Romer, Thomas, and Howard Rosenthal. 1978. Political resource allocation, controlled agendas, and the status quo. Public Choice 33:27-44.

Rosenthal, Howard. 1990. The setter model. In Advances in the spatial theory of voting, edited by James M. Enelow and Melvin J. Hinich. Cambridge, UK: Cambridge University Press.

Stratmann, Thomas. 1992. Are contributors rational? Untangling strategies of political action committees. Journal of Political Economy 100:647-64.

Tendler, Judith. 1975. Inside foreign aid. Baltimore: The Johns Hopkins University Press.

USAID. 1996. USAID 1995 Yellowbook (Web: http://gemini.info.usaid.gov/yellowbook/).

USAID. 1997. USAID 1996 Yellowbook (Web: http://gemini.info.usaid.gov/yellowbook/).

USAID. 1998a. Alabama (Web: http://www.info.usaid.gov/procurement_bus_opp/states/al.html).

USAID. 1998b. Direct economic benefits of U.S. assistance by state (Web: http://www.info.usaid.gov/procurement_bus_opp/states/).

USAID. 1998c. Why foreign aid? (Web: http://www.info.usaid.gov/about/y4naid.html).

U.S. Department of Commerce, Bureau of the Census. 1991. 1990 Decennial Census (on CensusCD by GeoLytics, Inc., 1996).

Wittkopf, Eugene. 1972. Western bilateral aid allocations: A comparative study of recipient state attributes and aid received. Beverly Hills, CA: Sage Publications.

Wood, Robert E. 1996. Rethinking economic aid. In Foreign aid toward the millennium, edited by Steven W. Hook. Boulder, CO: Lynne Rienner Publishers, pp. 19-37.

Wright, Gavin. 1974. The political economy of New Deal spending: An econometric analysis. Review of Economics and Statistics 56:30-8.

Zimmerman, Robert F., and Steven W. Hook. 1996. The assault on U.S. foreign aid. In Foreign aid toward the millennium, edited by Steven W. Hook. Boulder, CO: Lynne Rienner Publishers, pp. 57-73.

                                 Spending
                    Standard
              Mean Deviation Minimum Maximum
All            8.9      40.1       0     548
Republicans    5.9      21.8       0     210
 Non-Beltway   4.2      14.5       0     152
 Beltway     132.0      79.8      51     210
Democrats     12.5      54.1       0     548
 Non-Beltway   9.6      43.8       0     548
 Beltway     153.6     202.8      31     456
                        AIDSCORE
                   Standard
             Mean Deviation Minimum Maximum Count   Corr
All          48.6      38.8       0     100   426  0.149
Republicans  18.2      17.0       0      80   231  0.245
 Non-Beltway 17.5      16.0       0      80   228 -0.011
 Beltway     70.0      10.0      60      80     3  0.999
Democrats    84.7      23.1      10     100   195  0.128
 Non-Beltway 84.5      23.3      10     100   191  0.119
 Beltway     97.5       5.0      90     100     4  0.248
"Corr" indicates the simple correlation
between AIDSCORE and SPENDING.
                         SPENDING Tobit Estimation
                                       SPENDING as        SPENDING as
                                    Quadratic Function Quadratic Function
                                        of Actual         of Predicted
                                         AIDSCORE           AIDSCORE
AIDSCORE                                  -0.231
                                          (0.94)
[AIDSCORE.sup.2]                           0.0028
                                          (1.25)
Predicted AIDSCORE                                           -0.103
                                                             (0.34)
Predicted [AIDSCORE.sup.2]                                    0.0016
                                                             (0.60)
Dummy for AIDSCORE [less than] 40
Slope for AIDSCORE [less than] 40
Slope for AIDSCORE [greater than or
  equal to] 40
Dummy for predicted AIDSCORE
  [less than] 40
Slope for predicted AIDSCORE
  [less than] 40
Slope for predicted AIDSCORE
  [greater than or equal to] 40
BELTWAY                                   65.56 [**]         65.37 [**]
                                          (3.38)             (3.37)
EDUCATION                                628.8 [**]         638.6 [**]
                                          (8.42)             (8.46)
PUBLIC                                   470.1 [**]         495.6 [**]
                                          (2.23)             (2.35)
NONPROFIT                                239.3              244.2
                                          (1.04)             (1.05)
                                       SPENDING as          SPENDING as
                                     Piecewise Linear    Piecewise Linear
                                    Function of Actual Function of Predicted
                                         AIDSCORE            AIDSCORE
AIDSCORE
[AIDSCORE.sup.2]
Predicted AIDSCORE
Predicted [AIDSCORE.sup.2]
Dummy for AIDSCORE [less than] 40         11.87
                                          (1.00)
Slope for AIDSCORE [less than] 40         -0.119
                                          (0.39)
Slope for AIDSCORE [greater than or        0.174
  equal to] 40                            (1.35)
Dummy for predicted AIDSCORE                                  18.45
  [less than] 40                                              (0.87)
Slope for predicted AIDSCORE                                  -0.050
  [less than] 40                                              (0.15)
Slope for predicted AIDSCORE                                   0.255
  [greater than or equal to] 40                               (1.09)
BELTWAY                                   64.60 [**]          65.84 [**]
                                          (3.32)              (3.39)
EDUCATION                                630.8 [**]          636.7 [**]
                                          (8.44)              (8.42)
PUBLIC                                   474.4 [**]          494.8 [**]
                                          (2.24)              (2.34)
NONPROFIT                                240.9               242.7
                                          (1.04)              (1.04)
t-statistics in parentheses.
(*.)Significant at 90% confidence level.
(**.)Significant at 95% confidence level.
                         AIDSCORE Tobit Estimation
          Republicans Democrats
SPENDING     0.046       0.002
            (0.77)      (0.02)
BELTWAY     35.6 [**]  -17.1
            (2.89)     (-0.71)
XCORD      -84.4 [**] -128.8 [**]
          (-10.28)     (-9.07)
YCORD       37.3 [**]   29.1 [**]
            (9.91)      (3.46)
EDUCATION  117.2 [**]  248.2 [**]
            (2.34)      (2.30)
PUBLIC     -75.0       809.2 [**]
           (-0.64)      (2.77)
NONPROFIT  -32.2        18.4
           (-0.21)      (0.07)
FOREIGN    -38.4 [**]  -21.5
           (-2.00)     (-0.50)
BLACK      -14.7        -9.2
           (-0.86)     (-0.58)
HISPANIC    63.6 [**]    7.9
            (2.19)      (0.26)
ASIANLDC  -109.9 [**]   64.7
           (-2.13)      (1.04)
URBAN        9.4        -5.3
            (1.30)     (-0.39)
t-statistics parentheses. Tobit estimation with heteroskedasticity
correction of [square root][n.sub.i], where [n.sub.i]
is the number of votes cast by representative i.
(*.)Significant at 90% confidence level.
(**.)Significant at 95% confidence level.

Appendix A: Model Solution

We characterize the solution first for [omega] = 0, then for [omega] [greater than] 0.

The Case Where [omega] = 0

The solution is ([B.sup.SQ], [[alpha].sup.*]) where [B.sup.SQ] is the median value of [B.sub.i]. Since [omega] = 0, legislators vote according to their attitude toward aid spending ([B.sub.i]) and the efficiency with which aid is used--without concern for how much spending takes place in their home districts. The agency maximizes the budget that passes by following the efficient allocation ([[alpha].sup.*]). In this case, voting depends on political and constituent variables but not spending, and spending levels depend on contractor qualifications but not political variables. The line [omega] = 0 in Figure 1 reflects these characteristics.

The Case Where [omega] [greater than] 0

The solution (B, [alpha]) has the following characteristics:

1. Politics matter: [alpha] depends on [[alpha].sup.*] (a function of district qualifications) and ([B.sub.i]) (legislators' attitudes toward aid spending).

2. Swing Voters are rewarded: Legislators fall into one of three categories (Loyalist, Swing, Opponent), depending on their support for aid spending. Swing Voters' shares are disproportionately large compared with those of Loyalists and Opponents.

3. Swing Voters' rewards are inversely related to their support for aid. In other words, the Swing Voter section of Figure 1 is upward sloping.

The line [omega] [greater than] 0 in Figure 1 reflects these characteristics.

We find these characteristics of the solution by following the agency's strategy starting from the status quo point. Set [alpha] = [[alpha].sup.*] and identify the [[B.sup.0].sub.i] that maximizes utility for each legislator i: [[B.sup.0].sub.i] = [B.sub.i] + [omega][[alpha].sup.*].sub.i]/2. Order the legislators so that [[B.sup.0].sub.1] [greater than or equal to] [[B.sup.0].sub.2] [greater than or equal to] ... [greater than or equal to] [[B.sup.0].sub.n]. Let m = (n + 1)/2 for odd n and m = n/2 + 1 for even n. The combination ([[B.sup.0].sub.m], [[alpha].sup.*]) is the Condorcet winner in a pairwise competition against any other combination (B, [[alpha].sup.*]). We treat this as the "status quo" (i.e., [B.sup.SQ] = [[B.sup.0].sub.m]) since it could be offered by an agency opponent who cannot control [alpha]. Now consider the agency increasing B and altering [alpha] to maintain majority support. If B is increased by any amount, the agency must compensate legislator m by increasing [[alpha].sub.m]. If no other constraints bind, this can be done most efficiently by increasing [[alpha].sub.m] by [delta][alpha] while decreasing every other share by [delta][alpha]/(n - 1). Thus, we can rewrite the problem as choosing [delta][alpha] to maximize B subject to [delta][U.sub.m] = 0. The quadratic form of the legislator's utility function guarantees a finite solution ([B.sup.m], [[alpha].sup.m]), which is a function of the number of representatives (n), the direct impact of spending on voting ([omega]), district qualifications ([[alpha].sup.*]), and representatives' attitudes toward aid spending ({[B.sub.i]}). In this case, [[alpha].sub.m] [greater than] [[[alpha].sup.*].sub.m] and [[alpha].sub.i[neq]m] [less than] [[[alpha].sup.*].sub.i].

For some values of n, [omega], [[alpha].sup.*], and {[B.sub.i]}, other constraints bind before we reach [B.sup.m]. It is straightforward to show that [U.sub.i](B, [alpha]) - [U.sub.i]([B.sup.SQ], [[alpha].sup.*]) [greater than or equal to] [U.sub.j](B, [alpha]) - [U.sub.j]([B.sup.SQ], [[alpha].sup.*]), for i [less than] j [less than or equal to] m and B [greater than] [B.sup.SQ] (i.e., if j supports the bill, then i [less than] j will also support the bill). Legislators divide into three groups: i = 1 to l, Loyalists, who support the bill but receive unfavorable treatment [[U.sub.i](B, [alpha]) [greater than or equal to] [U.sub.i]([B.sup.SQ], [[alpha].sup.*]), [[alpha].sub.i] = [[[alpha].sup.*].sub.i] - [delta][alpha]]; i = l + 1 to m, Swing Voters, who support the bill and receive favorable treatment [[U.sub.i],(B, [alpha]) = [U.sub.i]([B.sup.SQ], [[alpha].sup.*]), [[alpha].sub.i] [greater than] [[[alpha].sup.*].sub.i] - [delta][alpha]]; and i = m + 1 to n. Opponents, who vote against the bill and receive unfa vorable treatment [[U.sub.i](B, [alpha]) [less than] [U.sub.i]([B.sup.SQ], [[alpha].sup.*]), [[alpha].sub.i] = [[[alpha].sup.*].sub.i] - [delta][alpha]]. l ranges from 0 to m - 1, depending on the values of the model parameters.

The final property of the solution is the upward slope in the Swing Voter section of Figure 1, or, in algebraic terms, [[alpha].sub.i] - [[[alpha].sup.*].sub.i] [less than] [[alpha].sub.j] - [[[alpha].sup.*].sub.j] for l [less than or equal to] i [less than] j [less than or equal to] m. To demonstrate this, note that the solution has [U.sub.i](B, [alpha]) - [U.sub.i]([B.sup.SQ], [[alpha].sup.*]) = 0 and [U.sub.j](B, [alpha]) - [U.sub.j]([B.sup.SQ], [[alpha].sup.*] = 0. Setting these two equations equal to each other and simplifying yields ([[alpha].sub.j] - [[[alpha].sup.*].sub.j]) - ([[alpha].sub.i] - [[[alpha].sup.*].sub.i]) [varies] B - B where B is the average of [[B.sup.0].sub.i] and [[B.sup.0].sub.j]. It is easy to show that B [greater than] B for [U.sub.i](B, [alpha]) = [U.sub.i]([B.sup.SQ], [[alpha].sup.*]), so we have [[alpha].sub.i] - [[[alpha].sup.*].sub.i] [less than] [[alpha].sub.j] - [[[alpha].sup.*].sub.j]. (Details available on request.)

Appendix B: Data

Definitions of Variables

AIDSCORE: For each member of the 104th House, the pro-aid voting percentage on the 10 votes listed in Appendix B2, that is [100.sup.*](pro-aid votes cast)/(total votes cast). Roll-call data are from Poole and Rosenthal (1998). Six votes came from the debate on the FY96-97 Foreign Aid and State Department Authorization Bill. Two are on amendments to the FY96 Foreign Operations Appropriations Bill. Another is an amendment to cut spending on Food for Peace (P.L. 480) in the FY96 Agriculture Appropriations Bill. The last is on an amendment to the FY97 Foreign Operations Appropriations Bill.

Predicted AIDSCORE: Ordinary-least-squares fitted values using X and Y dimension NOMINATE scores as explanatory variables. There is a .916 correlation between predicted AIDSCORE and actual AIDSCORE.

SPENDING: USAID contract spending, covering all contracts that were active in either fiscal 1995 or 1996 and were with firms based in the congressional district. Data are annualized dollar amounts in per capita terms. Derived from contract-level data in USAID (1996, 1997).

BELTWAY: Dummy variable equal to one for districts in the Beltway area (i.e., Washington, D.C., area): Maryland districts 3, 4, 5, and 8 and Virginia districts 8, 10, and 11. Maryland district 7 also qualifies geographically but is dropped from the sample because of a change in representative midway through the period.

XCORD: X-dimension NOMINATE scores for members of the 104th House (Poole and Rosenthal 1998).

YCORD: Y-dimension NOMINATE scores for members of the 104th House (Poole and Rosenthal 1998).

EDUCATION: Fraction of the 25-or-older population with postgraduate education, including professional degrees, by district (U.S. Department of Commerce 1991; hereafter 1990 Census).

PUBLIC: Fraction of population employed in the public sector, by district (1990 Census).

NONPROFIT: Fraction of population employed in nonprofit organizations, by district (1990 Census).

FOREIGN: Fraction of population foreign born, by district (1990 Census).

BLACK: Fraction of population African-American, by district (1990 Census).

HISPANIC: Fraction of population of Hispanic origin, by district (1990 Census).

ASIANLDC: Fraction of population of Asian/Pacific origin, excluding Japan and Korea, by district (1990 Census).

URBAN: Fraction of population living in urban areas, by district (1990 Census).

Descriptive Statistics and Additional Information on Variables

Appendix B1 presents some basic summary statistics for the NOMINATE scores and constituency data. Appendix B2 describes the roll-call votes used in the empirical analysis. Appendix B3 presents a vote-level breakdown by party. The partisan split is immediately apparent. On all but one motion, the majority of Republicans take the anti-aid position. On every motion, the majority of Democrats take the pro-aid position. The only item that has broad bipartisan support is Vote 547, funding for Food for Peace (P.L. 480), a well-established program of agriculture surplus disposal that has clear economic benefits for many constituencies, including farmers, shippers, and food processors. Appendix B4 separates out representatives from Beltway districts (those in the immediate vicinity of Washington, D.C., and USAID Headquarters). On most issues, the three Beltway Republicans take the pro-aid position, in sharp contrast to their non-Beltway counterparts. Beltway Democrats are also more supportive of aid, though they diff er less from non-Beltway Democrats.

                                Appendix B1
                            Summary Statistics
                Standard
          Mean  Deviation Minimum Maximum
XCORD     0.165   0.584   -0.940   0.996
YCORD     0.013   0.384   -0.950   0.995
EDUCATION 0.071   0.033    0.018   0.240
PUBLIC    0.022   0.012    0.006   0.112
NONPROFIT 0.031   0.011    0.012   0.093
FOREIGN   0.079   0.097    0.002   0.585
BLACK     0.115   0.157    0.001   0.739
HISPANIC  0.073   0.130    0.002   0.814
ASIANLDC  0.060   0.088    0.001   0.579
URBAN     0.750   0.220    0.131   1.00
                                Appendix B2
                           Description of Votes
Congressional
Quarterly No.                    Vote    Date    Sponsor
HR 1561: Fiscal 1996-1997
Foreign Aid and Statement
Department Authorization Bill
352                              121-292 5/24/95 Wynn
                                                 (D-md.)
354                              135-275 5/24/95 Hastings
                                                 (D-Fla.)
360                              172-230 6/7/95  Ackerman
                                                 (D-N.Y.)
363                              179-231 6/8/95  Burton
                                                 (R-Ind.)
365                              174-234 6/8/95  Hamilton
                                                 (D-Ind.)
366                              221-185 6/8/95
HR 1868: Fiscal 1996 Foreign
Operations Appropriations Bill
420                              198-215 6/27/95 Gilman
                                                 (R-N.Y.)
423                              235-179 6/27/95 Burton
                                                 (R-Ind.)
HR 1976: Fiscal 1996 Agriculture
Appropriations Bill
547                              330-83  7/21/95 Hoke
                                                 (R-Ohio)
HR 3540: Fiscal 1997 Foreign
Operations Appropriations Bill
212                              182-227 6/5/96  Burton
                                                 (R-Ind.)
Congressional
Quarterly No.                    Description
HR 1561: Fiscal 1996-1997
Foreign Aid and Statement
Department Authorization Bill
352                              Amendment to increase debt relief
                                  in Latin  America and the Caribbean
                                  by $12 million in  each of
                                  FY96 and FY97.
354                              Amendment to increase Development
                                  Fund for Africa by $173 million
                                  in each of FY96 and FY97.
360                              Amendment to require a cost-benefit
                                  analysis before USAID, USIA,
                                  and ACDA could be folded
                                  into the State Department.
363                              Amendment to cut USAID operating
                                  budget by an additional $69 million
                                  in FY96 and $22.4 million in FY97.
365                              Motion to recommit bill to
                                  committee with instructions to
                                  reverse abolishment of USAID.
366                              Passage of bill including
                                  abolishment of USAID.
HR 1868: Fiscal 1996 Foreign
Operations Appropriations Bill
420                              Amendment to cut the Development
                                  Assistance Fund by $24 million.
                                  (The Fund is administered by USAID)
423                              Amendment to eliminate a $29.9
                                  million fund  to cover mandated
                                  downsizing; expenses to be
                                  covered by USAID's regular budget.
HR 1976: Fiscal 1996 Agriculture
Appropriations Bill
547                              Amendment to cut $113 million
                                  From for Peace (PL 480). (The
                                  program is adiminstered by USAID).
HR 3540: Fiscal 1997 Foreign
Operations Appropriations Bill
212                              Amendment to cut the USAID
                                  operating budget by $47 million.
Voting data from Poole and Rosenthal (1998). We include only
representiativers who remained in office and did not switch
parties durign this period.
                                Appendix B3
                         Individual Votes by Party
               Republicans        Democrats
Congressional   Pro-Aid-    Not   Pro-Aid-   Not
Quarterly. No.  Anti-Aid   Voting Anti-Aid  Voting
352               0-225      6     121-67     7
354               1-223      7     134-52     9
360               5-214     12     167-16    12
363 [a]          72-154      5     159-25    11
365               3-222      6     171-12    12
366 [a]          16-210      5     169-11    15
420 [a]          47-179      5     168-19     8
423 [a]          37-190      4     142-45     8
547 [a]          151-72      8     179-11     5
212 [a]          67-157      7     160-25    10

Voting data from Poole and Rosenthal (1998). We include only representatives who remained in office and did not switch parties during this period.

(a.)Indicates an anti-aid measure; we counta vote against the measure as a pro-aid vote in this table.

                                Appendix B4
                            Beltway Comparison
                Republicans               Democrats
              Pro-Aid-Anti-Aid         Pro-Aid-Anti-Aid
Congressional     Outside      Inside      Outside      Inside
Quarterly No.     Beltway      Beltway     Beltway      Beltway
352                0-222         0-3        117-67        4-0
354                1-220         0-3        130-52        4-0
360                2-214         3-0        164-15        3-1
363 [a]           69-154         3-0        155-25        4-0
365                0-222         3-0        167-12        4-0
366 [a]           15-208         1-2        165-11        4-0
420 [a]           44-179         3-0        164-19        4-0
423 [a]           34-190         3-0        138-45        4-0
547 [a]           149-71         2-1        175-11        4-0
212 [a]           64-157         3-0        156-25        4-0

Voting data from Poole and Rosenthal (1998). We include only representatives who remained in office and did not switch parties during this period. "Inside Beltway" includes Maryland districts 3, 4, 5, and 8 and Virginia districts 8, 10, and 11. Maryland district 7 also qualifies geographically but is dropped from the sample because of a change in representative midway through the period.

(a.) Indicates an anti-aid measure; we count a vote against the measure as a pro-aid Vote in this table.