The presidential pork barrel and the conditioning effect of term.

Taylor, Andrew J.

There have been numerous analyses of how presidential election rules and pressures shape campaign behavior. Shaw (1999) and Hill and McKee (2005) reveal, for example, that presidential candidates expend more resources in states with uncertain outcomes--these are sometimes called "battleground" states. Surprisingly little has been done on the effects of elections on presidential governance, however. Some anecdotal evidence suggests that presidents make policy that favors battleground states. Witness, for instance, the Bush administration's imposition of tariffs on steel imports during 2002-2003 that, many argued, was designed to attract support from the domestic industry's home states of Ohio, Pennsylvania, and West Virginia--all of them up for grabs in the 2004 presidential contest (Broder 2002). This is something Karl Rove called "asset deployment" (Solomon, MacGillis, and Cohen 2007). Other impressionistic evidence suggests that presidents advantage those states that provided them with the most support in their election. There have been reports that New York has received disproportionately little of federal homeland security money in the past few years because it is solidly Democratic (Newfield 2004).

Two hypotheses about federal government spending and presidential elections reflect these casual observations. The first is the battleground hypothesis. It asserts that states that are evenly balanced between the two parties will receive a disproportionate amount of federal largesse as presidents attempt to curry favor for the upcoming election. The second is the presidential-support hypothesis. It posits that presidents reward states whose voters supported them in disproportionately large numbers in the previous election.

This article is a contribution to the small but growing literature that tests these hypotheses and examines what might be called the presidential pork barrel. It adds value in two ways. First, it focuses on a category of spending, procurement, that has not been examined this way before. Procurement is particularly suited to a test of these hypotheses. It is, for instance, clearly distributive in nature. A problem with Larcinese, Rizzo, and Testa's (2006) recent study is that they use aggregate federal spending per capita to confirm the presidential support hypothesis. A significant proportion of this measure is in the form of direct payments to individuals--policy the president has little ability to alter unilaterally because it is established by prearranged statutorily derived formulae. Efforts to alter geographic patterns of this type of spending are likely to be shaped by partisan and ideological struggles.

Second, the article examines the conditioning effect of term. No other study of this issue has looked at presidential term, but it makes sense to assume that the vote-buying imperatives of the two hypotheses are greatly attenuated or completely absent in presidents' second terms. Forced from the ballot by the Twenty-Second Amendment, second-term presidents do not need to use distributive spending to buy popular votes.

Theoretical Underpinnings and Hypotheses

If distributive spending is used for electoral purposes, the fundamental assumption is generally that elected officials "buy" the votes of constituents by directing government largesse to them. The more spending their constituents receive, the greater the chances elected officials have of receiving their votes and winning the election.

The president's constituency is the nation. However, the Electoral College breaks this constituency into fifty-one discrete units--the fifty states plus the District of Columbia. Each state has the same number of electoral votes as it has members of Congress, and candidates must win a majority of these votes (at least 270) to become or remain president. What is more, with the exception of Maine and Nebraska, all states use the unit rule, which means they distribute their electoral votes en bloc to the candidate who wins a plurality of their popular vote. As a result, and because distributive spending is limited--dollars expended this way cannot be used for other programs or tax reductions or, alternatively, they add to the deficit--presidents must prioritize the states whose voters they wish to "buy."

There exist two main hypotheses about how presidents approach this prioritizing. The first is the battleground hypothesis. Here, presidents distribute spending to states that are competitive between the parties, believing that their behavior is a cheap way to affect the distribution of electoral votes in the coming presidential contest. In this model, all individual popular votes across the country are assumed to cost the same. The strategy is an inexpensive way to purchase electoral votes per capita because the electoral votes of states that are not considered battleground are either securely in the president's pocket (they have effectively been purchased already) or, in the case of states where the opposition party has much support, too expensive to acquire because on an electoral-vote-per-capita basis, the president would have to buy relatively large numbers of popular votes to obtain them.

Some theoretical and empirical work along these particular lines has already been done. Lindbeck and Weibull (1993) and Dixit and Londregan (1996) provide formal models that show that, when groups have the same consumption preferences and parties are approximately equal in their capacity to direct benefits once in office, swing voters will be rewarded. Empirical vindication is presented by Herron and Theodos (2004), who show that Illinois government grants were disproportionately distributed to swing House districts in the run-up to the 2000 election, and Bickers and Stein (1996), who come to the same finding when looking at the pattern of Small Business Administration grants across U.S. House districts before the 1990 election. In presidential politics, Garrett and Sobel (2003) find that states the president has a roughly even chance of winning in the next election received a disproportionately large share of Federal Emergency Management Agency expenditures and disaster declarations. This is similar to how some scholars believe Franklin Roosevelt distributed New Deal spending (Wallis 1987; Wright 1974).

Derived from Cox and McCubbins's (1986) work, the presidential-support hypothesis, on the other hand, posits that presidents steer spending to states that provided them with relatively large proportions of their popular votes in the previous election. They do this because they believe it will yield a greater return of total popular votes in the future than if they steered spending to other, less supportive, states. Investment in battleground states or those that went solidly for the opponent last time may not provide presidents with a big enough return to win the next election. The individual popular votes of citizens who supported the president in the previous election are therefore considered less expensive than those belonging to voters who opposed him. In an urban setting, the result is a "machine politics" (Dixit and Londregan 1996) where, even without explicit on-the-stump pledges, groups of supporters receive patronage and city government business. This argument is consistent with Larcinese, Rizzo, and Testa's (2006) findings as well as analyses done on New Deal spending (Anderson and Tollison 1991; Couch and Shugart 1998).

Presidential attempts to distribute federal spending for electoral reasons ought to be conditioned by the Twenty-Second Amendment to the Constitution--a provision that prohibits presidents from running for a third term. Unable to run for reelection, presidents in their second term are considerably less likely to want to buy votes. Canes-Wrone and Shotts (2004) show that second-term presidents are much less responsive to public opinon than first-term ones. There ought, in other words, to be an electoral effect only in first terms. Interestingly, however, scholars have not examined presidential term as a conditioning effect on administration behavior. Garrett and Sobel (2003) and Lardnese, Rizzo, and Testa (2006), for instance, treat first and second terms identically.

The thinking about the Twenty-Second Amendment provides us with corollaries to the two hypotheses. I test these directly:

[H.sub.1] (battleground hypothesis): When a president is in his first term, the more competitive a state is between the parties, the more procurement spending it will receive.

[H.sub.2] (presidential-support hypothesis): When a president is in his first term, the larger the proportion of a state's vote the president won in the previous election, the more procurement spending the state will receive.

Note that in both hypotheses, I am assuming the relevant electoral incentives will have an effect in across-state distribution of spending by the administration in the first term. These electoral incentives do not provide us with an expectation of a commensurate reductive effect in the second term. The electoral effect will be severely attenuated or entirely absent in the second term.

Data

The dependent variable is the procurement dollars per capita that each of the fifty states received in each year between 1984 and 2004. (1) The data are taken from the Consolidated Federal Funds Report (CFFR) issued annually by the Census Bureau. (2) They are the total value of items in CFFR category procurement contracts (PC). The twenty-one-year span is the longest period for which I can be sure data are consistent. (3)

Procurement is the money spent by the federal government in contracted relationships with clients who are then obligated to provide deliverables. Most contracts are undertaken with the Department of Defense. The dollar amounts represent federal obligations incurred at the time the contract was awarded, not actual expenditures.

There are three reasons procurement spending provides an especially useful way to test the hypotheses. The first is that it represents important policy action. It cannot be seen as a symbolic incentive or reward to a state--as, for example, presidential visits and governmental appointments could be. It involves real and considerable resources. In FY 2004, for instance, the federal government spent nearly $340 billion on procurement. Throughout the period I study, procurement fluctuates from about 20 percent of federal expenditures in 1984 to, largely because the Cold War ends, about 13 percent from the mid-1990s until today.

Second, procurement is clearly distributive policy. Individual procurement contracts are concentrated benefits. They involve the directing of resources in separately granted increments to targeted and geographically grouped recipients. They can therefore be seen as an explicit incentive for electoral support.

Third, the president and his administration can influence the distribution of procurement contracts greatly. To be sure, there are restraints. Federal procurement is governed by competitive bid, Davis-Bacon wage, "buy American," and minority business ownership rules. The laws that govern the basic contracting processes are also passed by Congress. Presidents have significant influence in the legislative process, however. They can manipulate the language in laws that guide universal decision criteria on contracts and negotiate with lawmakers over earmarks. As chief executive, the president has tremendous influence over the departments and agencies that make contracting decisions. The procurement bidding process and the general policies that shape it have been supervised since 1974 by the Office of Management and Budget's Office of Federal Procurement Policy--an agency controlled directly by the White House. Moreover, presidents have shaped procurement policies through executive orders--actions that are both unilateral and consequential (Howell 2003; Mayer 2001). (4) As dictated by law, the president also submits an annual budget that includes specific procurement requests. Finally, administration agencies oversee procurement contracts and report on contractor performance. This furnishes them with essential information and influence in the contract-granting process. The cumulative effect of all this, according to Nownes (2006, 165-81), is that procurement lobbyists spend a considerable amount of time trying to influence executive agencies.

Because congressional appropriations and executive spending decisions are generally made before the fiscal year in which the policy takes effect, the observations for all independent variables are from the year prior to the observation of the dependent variable) The first hypothesis is tested utilizing the absolute difference between the percentage of the national two-party vote the sitting president won in the previous election and the percentage of the two-party vote he won in the state in the previous election. (6) Because the president's goal is to secure a majority of the electoral vote in the cheapest possible way and all popular votes cost the same amount, he will order states from most to least supportive of him and steer procurement to those most likely to take him from below to above 270 in the Electoral College. In other words, he will focus on states with presidential election outcomes that tend to mirror the nation's in the aggregate. The score is then subtracted from 25 and the resultant variable labeled Battleground. (7) I undertake this manipulation so that the measure is directionally consistent with the indicator of Presidential Term--coded 1 for the first term and 0 for the second--with which it is interacted. (8) High values of the interaction should result in more procurement spending.

The second hypothesis is tested by subtracting the percentage of the national two-party vote the sitting president won in the previous election from the percentage of the two-party vote he won in the state in the previous election (labeled Presidential Support). (9) Again, high values of the resultant interaction should generate high values in the dependent variable. (10)

The literature suggests the need for congressional controls (Bickers and Stein 2000, 2004; Lee 2003; Lowry and Potoski 2004; Rundquist and Carsey 2002; Stein and Bickers 1995). I use six. The first two are the proportion of a state's House and Senate delegation that is affiliated with the chamber's majority party. The Senate and, especially, the House majority have significant proposal and gatekeeping powers (Gailmard and Jenkins 2007; Cox and McCubbins 2005). These allow the majority to regulate the substance of bills and the amount and type of federal spending (Balla et al. 2002; Bickers and Stein 2000).

The other congressional variables capture the proportion of a state's House and Senate delegation that is seated on the House and Senate Appropriations and Armed Services committees. (11) Adler (2002) and Carsey and Rundquist (1999; Rundquist and Carsey 2002) show that membership on these committees helps legislators direct federal spending to their constituents. I use Armed Services because in excess of two thirds of procurement is defense spending.

I use a series of other controls. Shepsle and Weingast (1981) argue--and others show (Bickers and Stein 2000; Levitt and Snyder 1995)--that spending on a congressional district has significant positive externalities for those close by and, therefore, the larger the state, the greater the incidence of such positive externalities. This should, in turn, encourage presidents to steer spending to larger states. I use a variable, Population, that is the state's population divided by 100,000. (12)

Because most contracts are generally issued by the Department of Defense, I employ Rundquist and Carsey's (2002) Gunbelt variable. This is a dichotomous variable used by Markusen et al. (1991) to identify states long associated with weapons manufacturing--states I score as 1.

I employ a final control. To estimate the amount of discretion a president will have over spending, I use a variable Divided Government, which is coded 1 when at least one chamber of Congress is controlled by the party that opposes him. Under divided government, we should expect Congress to increase the constraints on unilateral executive action and mitigate presidential electoral effects. Descriptive statistics and other details about the variables can be found in the Appendix.

Results

The data are annual observations for each of the fifty states. These are cross-sectional time-series data. I present analyses of a fixed-state-effects model and a generalized least squares estimation of a random-effects model, both with robust standard errors. I also report an ordinary least squares estimation using panel-corrected standard errors. Typically, analyses like this omit a random-effects model because states seem to have unique characteristics that are not interchangeable (Larcinese, Rizzo, and Testa 2006). For instance, Maryland and Virginia garner relatively large aggregate per capita dollar amounts of procurement because of their proximity to Washington, New Mexico because of the Los Alamos nuclear laboratory, and Missouri because it is where Boeing's defense business is headquartered. Here, however, Hausman and Breusch-Pagan tests for random effects suggest such a model is appropriate. I continue to show a fixed-effects model as a check of robustness.

The presence of random effects is suggested by Figure 1. It reveals the annual spending per capita on federal procurement in three states: California, Massachusetts, and Virginia. Note the different longitudinal patterns and the sometimes dramatic changes from year to year. California and Massachusetts, for example, suffered greatly from the post-Cold War decline in defense spending. These patterns also suggest why the data are not afflicted by serial correlation--as unreported AR1 models reveal. Annual changes for states are sometimes dramatic and not always in the same direction.

Table 1 reveals the results for separate models using the battleground and presidential-support interactions. The controls perform interestingly. There is a "largestate" effect that is consistent with the positive externalities argument. Bigger states get discernibly more procurement per capita.

The congressional controls, moreover, are often robust. In this regard, note two things especially. First, the Senate variables are frequently more robust than the House versions. Because our unit of analysis is the state, this makes some sense. Second, the committee variables are considerably more important than the party ones. There is some variation across estimation techniques, but the random-effects model testing the presidential support hypothesis shows that adding a senator to a state's delegation on Appropriations is worth about $42 per capita in procurement spending; to Armed Services it is worth about $77. This basic finding is consistent with the view of procurement as distributive policy. Distributive policy is not particularly important to party reputation and is generally crafted in a bipartisan process characterized by extensive logrolling (Adler 2002; Mayhew 1974; Shepsle and Weingast 1981). There is generally deference to the committee of jurisdiction.

[FIGURE 1 OMITTED]

More importantly, however, note the performance of the interactions. The battleground interaction does not have a statistically significant coefficient, regardless of the method used. The presidential support interaction, however, is robust and has a positive sign that is consistent with the term-conditioned hypothesis. There does seem to be evidence of an electoral connection here.

Table 2, however, rejects this assertion. It reveals the marginal effects of presidential support on procurement dollars per capita during both the first and second terms in the models presented in Table 1. (13) It is not that first-term presidents distribute a significantly disproportionate amount of contracts to states that gave them the most support in their election. To be sure, during the first term the marginal effect is positive, but it is not statistically significant. Instead, there seems to be a second-term effect. Note in Table 2 that the marginal effect of support for the president on procurement has a statistically significant negative coefficient during a president's second term. A glance back reveals that these coefficients and standard errors are those reported in Table 1. This is because the regression reports coefficients when Presidential Term = 0 or a president is in his second term (Brambor, Clark, and Golder 2006, 74). This is true for all estimation techniques. It reveals a counterintuitive second-term effect in that the states that gave the president less of their popular vote in his reelection received significantly more procurement dollars per capita in his second term. The magnitude of the effect differs across estimation techniques, but for fixed and random effects we can see that, during a president's second term, a 1 percent increase in the president's share of a state's two-party vote in the previous election results in a decrease of about $4.20 to $4.50 per capita in procurement spending.

Discussion

I do not find a presidential electoral connection to procurement spending, a category of distributive policy where, if such a connection existed, we should detect one. Plausible electoral effects are not even present when they are conditioned by presidential term--something scholars have not looked at before. These results cast significant doubt on the notion that presidents use pork barrel politics to get reelected. This is a central conclusion of the article.

Given the strong theoretical reasons to expect such behavior, the question is: why not? One possible answer is that the theory expects too much of the voter. Voters must see or feel the positive effects of procurement spending, attribute them to administration action, and then act upon this as they vote. Presidential general-election campaigns are generally about candidate positions on large foreign and domestic issues and programs with national rather than parochial implications.

Another possible reason for the lack of a presidential electoral connection is that, as Edwards (1989, 189-96) concluded after analyzing interbranch bargaining in his seminal book on presidential leadership, the White House cannot steer spending where it would like. Legislators, on the other hand, can. My results confirm previous studies that show that members of Congress direct distributive spending (Balla et al. 2002; Bickers and Stein 2000; Carsey and Rundquist 1999; Lowry and Potoski 2004; Rundquist and Carsey 2002; Stein and Bickers 1995). More specifically, I reveal the geographic distribution of procurement spending is largely a function of the composition of important congressional committees. To be more precise, the larger the proportion of its delegation on the Senate Appropriations and Armed Services committees, the more procurement dollars per capita a state gets.

This reaffirmation of the importance of committee membership to the geographic distribution of government spending is important. Despite the discussion of the rise of party and the centralization of power in the House especially (Cox and McCubbins 2005; Rohde 1991), procurement spending fits traditional distributive models in which committees are critical and outcomes are not particularly partisan.

The counterintuitive second-term effect I find could be evidence of the presidential steering of procurement, however. Let me emphasize that this explanation is highly speculative and requires testing. It may be that lame-duck presidents, especially, use procurement to buy legislative votes--rather than popular ones--in support of their agenda. They steer procurement to members from states that gave them a smaller proportion of their popular votes in the previous election because these members are predisposed to oppose the administration. Legislators are receptive to this because voters credit them, not presidents, for securing distributive policy. What is more, congressional opposition to the president is greater in the second term when he does not have the resource of being on the top of the ticket in the next election. During the first term, it may be that presidents--who enjoy an incumbency advantage and must be considered the favorite to win the next election (Mayhew 2005)--have reputational resources they can use to leverage congressional support for their proposals. Legislators can benefit from being associated with the president the next time both are on the ballot. Such resources evaporate in the second term and must be compensated for, especially with members who have few other reasons to assist the administration.

Appendix

The Variables

Variable Minimum
 How Measured Value

Procurement Procurement per 113.75
 (dependent capita in dollars
 variable)
Battleground 25--absolute difference 2.30
 of president's share
 of state two-party vote
 and national two-party
 vote in previous election
Presidential President's share of state -17.30
 Support two-party vote--
 president's share of
 national two-party
 vote in previous
 election
Presidential Is president in first 0
 Term (coded 1) or second
 (coded 0) term?
House Proportion of state's 0
 Majority delegation in majority
Senate Proportion of state's 0
 Majority delegation in majority
House Proportion of state's 0
 Appropriations delegation on committee
Senate Proportion of state's 0
 Appropriations delegation on committee
House Armed Proportion of state's 0
 Services delegation on committee
Senate Armed Proportion of state's 0
 Services delegation on committee
Population State's in year x / 100,000 4.50
Gunbelt Is state in gunbelt? (Rundquist 0
 and Carsey 2002) Yes = 1
Divided Government Is party of president different 0
 from majority in at least one
 chamber of Congress? Yes = 1

The Variables

Variable Maximum Standard
 Value Mean Deviation

Procurement 4735.30 712.52 549.06
 (dependent
 variable)
Battleground 25.00 19.29 4.43
Presidential 22.70 0.58 7.21
 Support
Presidential 1 0.62 0.49
 Term
House 1 0.57 0.29
 Majority
Senate 1 0.54 0.39
 Majority
House 0.50 0.12 0.12
 Appropriations
Senate 1 0.29 0.28
 Appropriations
House Armed 1 0.12 0.15
 Services
Senate Armed 1 0.21 0.25
 Services
Population 354.84 52.06 56.70
Gunbelt 1 0.40 0.49

Divided Government 1 0.57 0.50

N = 1050.

AUTHOR'S NOTE: An earlier version of this article was presented at the 2005 annual meeting of the American Political Science Association. I would like to thank Bird Loomis and Brendan Doherty for their comments and Matt Golder for assistance with data analysis. The data were analyzed using Stata 9.0.

References

Adler, E. Scott. 2002. Why congressional reforms fail: Reelection and the House committee system. Chicago: University of Chicago Press.

Anderson, Gary M., and Robert D. Tollison. 1991. Congressional influence and patterns of New Deal spending. Journal of Law and Economics 34: 161-75.

Balla, Steven J., Eric D. Lawrence, Forrest Maltzman, and Lee Sigelman. 2002. Partisanship, blame avoidance, and the distribution of legislative pork. American Journal of Political Science 46: 515-25.

Bickers, Kenneth N., and Robert M. Stein. 1996. The electoral dynamics of the federal pork barrel. American Journal of Political Science 40: 1300-26.

--. 2000. The congressional pork barrel in a Republican era. Journal of Politics 62: 1070-86.

--. 2004. Interlocal cooperation and the distribution of federal grant awards. Journal of Politics 66: 800-22.

Brambor, Thomas, William Roberts Clark, and Matt Golder. 2006. Understanding interaction models: Improving empirical analyses. Political Analysis 13: 1-20.

Broder, David. 2002. Tailoring policy to electoral votes. Washington Post, March 10, p. B9.

Canes-Wrone, Brandice, and Kenneth W. Shotts. 2004. Leadership and pandering: A theory of executive policy making. American Journal of Political Science 45: 532-50.

Carsey, Thomas M., and Barry Rundquist. 1999. Party and committee in distributive politics: Evidence from defense spending. Journal of Politics 61: 156-69.

Couch, Jim E, and William F. Shugart II. 1998. The political economy of New Deal spending. Cheltenham, UK: Edward Elgar.

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

--. 2005. Setting the agenda: Responsible party government in the U.S. House of Representatives. New York: Cambridge University Press.

Dixit, Avinash, and John Londregan. 1996. The determinants of success of special interests in redistributive politics. Journal of Politics 58: 1132-55.

Edwards, George C. III. 1989. At the margins: Presidential leadership of Congress. New Haven, CT: Yale University Press.

Gailmard, Sean, and Jeffrey A. Jenkins. 2007. Negative agenda control in the Senate and House of Representatives: Fingerprints of majority party power. Journal of Politics 69: 689-700.

Garrett, Thomas A., and Russell S. Sobel. 2003. The political economy of FEMA disaster payments. Economic Inquiry 41: 496-509.

Hauk, William R., and Romain Wacziarg. 2007. Small states, big pork. Quarterly Journal of Political Science 2: 95-106.

Herron, Michael C., and Brett A. Theodos. 2004. Government redistribution in the shadow of legislative elections: A study of the Illinois member initiatives grant program. Legislative Studies Quarterly 24: 287-312.

Hill, David, and Seth C. McKee. 2005. The Electoral College, mobilization, and turnout in the 2000 presidential election. American Politics Research 33: 700-25.

Howell, William G. 2003. Power without persuasion: The politics of direct presidential action. Princeton, N J: Princeton University Press.

Larcinese, Valentino, Leonizo Rizzo, and Cecilia Testa. 2006. Allocating the U.S. federal budget to the states: The impact of the president. Journal of Politics 68: 447-56.

Lee, Frances E. 2003. Geographic politics in the U.S. House of Representatives: Coalition building and the distribution of benefits. American Journal of Political Science 47: 714-28.

Lee, Frances E., and Bruce I. Oppenheimer. 1999. Sizing up the Senate: The unequal consequences of equal representation. Chicago: University of Chicago Press.

Levitt, Steven D., and James M. Snyder, Jr. 1995. Political parties and the distribution of federal outlays. American Journal of Political Science 39: 958-80.

Lindbeck, Assar S. N., and Jorgen W. Weibull. 1993. A model of political equilibrium in a representative democracy. Journal of Public Economics 51: 195-209.

Lowry, Robert C., and Matthew Potoski. 2004. Organized interests and the politics of federal discretionary grants. Journal of Politics 66: 513-33.

Markusen, Ann, Peter Hall, Scott Campbell, and Sabina Deitrick. 1991. The rise of the gunbelt. New York: Oxford University Press.

Mayer, Kenneth R. 2001. With the stroke of a pen: Executive orders and presidential power. Princeton, NJ: Princeton University Press.

Mayhew, David R. 1974. Congress: The electoral connection. New Haven, CT: Yale University Press.

--. 2005. Incumbency advantage in presidential elections: The historical record. Manuscript.

Newfield, Jack 2004. Bush to city: Drop dead. The Nation, April 19, pp. 11-13.

Nownes, Anthony J. 2006. Total lobbying: What lobbyists want (and how they try to get it). New York: Cambridge University Press.

Rohde, David W. 1991. Parties and leaders in the postreform Congress. Chicago: University of Chicago Press.

Rundquist, Barry S., and Thomas M. Carsey. 2002. Congress and defense spending: The distributive politics of military procurement. Norman: University of Oklahoma Press.

Shaw, Daron R. 1999. The methods behind the madness: Presidential Electoral College strategies, 1988-1996. Journal of Politics 61: 893-913.

Shepsle, Kenneth, and Barry Weingast. 1981. Political preferences and the pork barrel: A generalization. American Journal of Political Science 25: 96-111.

Solomon, John, Alec MacGillis, and Sarah Cohen. 2007. How Rove directed federal assets for GOP gains. Washington Post, August 19, p. A1.

Stein, Robert M., and Kenneth N. Bickers. 1995. Perpetuating the pork barrel: Policy subsystems and American democracy. New York: Cambridge University Press.

Wallis, John J. 1987. Employment, politics and economic recovery in the Great Depression. Review of Economics and Statistics 69: 516-20.

Wright, Gavin. 1974. The political economy of New Deal spending: An econometric analysis. Review of Economics and Statistics 56: 30-38.

ANDREW J. TAYLOR

North Carolina State University

(1.) The District of Columbia is excluded for two reasons. First, its location means that it will receive a massively disproportionate amount of procurement. Seconds it has no values for the congressional control variables.

(2.) CFFR reports from 1993 to 2004 are available from http://harvester.census.gov/cffr/index.html. Earlier data were taken from Consolidated Federal Funds Report, Volume 1, County Areas (Washington, DC: U.S. Department of Commerce, Bureau of the Census, various years).

(3.) CFFR's categorization of procurement changes too much prior to 1983 to make the data comparable with those of the twenty-one-year period I examine.

(4.) In 1993, for example, President Bill Clinton issued an executive order changing federal acquisition standards (Order 12874).

(5.) A fiscal year starts on October 1 of the previous calendar year.

(6.) The assumption here is that to get to 270 electoral votes, a president needs to buy popular votes from among battleground states. This is the case in all of the elections except, perhaps, 1984, when Reagan's landslide meant that it was in only very strong Democratic states that both candidates were competitive and the outcome was uncertain as the election approached. Still, even in this instance, no one could predict the election would turn out like this in the first few years of Reagan's term.

(7.) That is, Battleground = 25 - abs ([n.sub.i] - [s.sub.i]), where [n.sub.i] = the president's national two-party vote in election i and [s.sub.i] = the president's share of the two-party vote in the state in election i. To provide an example, in 1992, Bill Clinton received 51.2 percent of the two-party vote in Ohio and 53.5 percent of the two-party vote nationally. The score for Ohio in the four years of Clinton's first term is, therefore, 25 - (53.5 - 51.2), or 22.7.

(8.) I prefer this measure of a state's perceived battleground status over that used by Shaw (1999) and Hill and McKee (2005). They use a dichotomous version of battleground status based upon data obtained from individuals working on presidential campaigns. Unfortunately, the measure is unavailable for the entire period I study.

(9.) That is, using the terms above, Presidential Support = [s.sub.i] - [n.sub.i].

(10.) The battleground and presidential support variables are clearly independent. Correlated with one another, r = -.109. Still, the measures are constructed from the same basic indicator, so I present the tests of the hypotheses in separate models.

(11.) The House changed the name of Armed Services to National Security from 1995 to 2000.

(12.) Lee and Oppenheimer (1999) have shown that, at least for federal grants, spending is subject to state minimums and that smaller states therefore tend to receive more per capita. Hauk and Wacziarg (2007) argue that Senate malapportionment explains why smaller states received disproportionately more of the spending in the 2005 highway bill.

(13.) The program used to calculate such marginal effects in Stata was kindly provided by Matt Golder.

Andrew J. Taylor is a professor and chair of political science in the School of Public and International Affairs at North Carolina State University. His research on American governmental institutions has been published in the American Journal of Political Science, Journal of Politics, Political Research Quarterly, and Legislative Studies Quarterly.

TABLE 1
A Test of the Electoral Conne Hypotheses

Battleground Model

 Random
 Fixed Effect Effects

Battleground -10.88 ** -10.78 **
 (3.12) (3.07)
Presidential Term 142.12 * 149.20 *
 (58.92) (61.22)
Battlegmund x -6.03 -6.26
 Presidential Term (3.21) (3.31)
House Majority 49.05 44.01
 (30.35) (30.39)
Senate Majority -28.85 -29.96
 (23.62) (23.81)
House Appropriations 82.94 85.96
 (59.87) (61.21)
Senate Appropriations 119.65 * 121.72 *
 (58.21) (60.09)
House Armed Services 29.16 12.35
 (93.20) (95.98)
Senate Armed Services 135.37 ** 130.35 **
 (50.85) (49.98)
Population 51.97 ** 19.99 **
 (16.40) (7.16)
Gunbelt 329.12 *
 (146.44)
Divided Government 16.28 2.16
 (17.58) (17.02)
Constant 575.54 ** 611.43 **
 (113.49) (98.44)
[R.sup.2] (within, .06, .00, .00 .05, .10, .09
 between, overall)
F (for FE)/Wald 4.21 ** 162.97 **
 [chi square]
Hausman 13.88
Breusch-Pagan 5051.79 **
T 21 21
N 50 50

Presidential-Support
Model

 Random
 Fixed Effects Effects

Presidential Support -4.21 * -4.52 *
 (1.75) (1.76)
Presidential Term 23.91 26.71
 (16.67) (16.51)
Presidential Support x 8.74 ** 9.17 **
 Presidential Term (2.04) (2.07)
House Majority 20.95 13.93
 (34.31) (34.76)
Senate Majority -34.74 -36.46
 (24.69) (24.87)
House Appropriations 65.56 83.21
 (38.21) (46.78)
Senate Appropriations 71.83 * 84.67 *
 (32.84) (34.84)
House Armed Services 38.09 21.04
 (96.51) (99.40)
Senate Armed Services 160.51 ** 154.51 **
 (52.11) (50.85)
Population 54.12 ** 19.11 **
 (16.28) (7.06)
Gunbelt 340.21 *
 (144.74)
Divided Government 11.57 -3.63
 (18.37) (17.83)
Constant 367.89 ** 419.29 **
 (95.69) (82.26)
[R.sup.2] (within, .04, .00. .00 .04, .12, .10
 between, overall)
F (for FE)/Wald 4.19 ** 163.74 **
 [chi square]
Hausman 6.03
Breusch-Pagan 5074.77 **
T 21 21
N 50 50

Battleground Model

 Panel-
 Corrected

Battleground 3.52
 (3.02)
Presidential Term 126.03
 (88.41)
Battlegmund x -5.48
 Presidential Term (3.74)
House Majority 68.32
 (47.65)
Senate Majority -50.30
 (26.47)
House Appropriations 244.85 *
 (96.71)
Senate Appropriations 175.73 **
 (37.46)
House Armed Services 317.23 **
 (93.57)
Senate Armed Services 394.62 **
 (65.54)
Population 7.09 **
 (1.25)
Gunbelt 368.78 **
 (19.00)
Divided Government -9.25
 (51.16)
Constant 315.66 **
 (78.78)
[R.sup.2] (within, .184
 between, overall)
F (for FE)/Wald 720.82 **
 [chi square]
Hausman
Breusch-Pagan
T 21
N 50

Presidential-Support
Model

 Panel-
 Corrected

Presidential Support -7.07 *
 (2.77)
Presidential Term 14.77
 (49.79)
Presidential Support x 12.19 **
 Presidential Term (3.40)
House Majority 33.35
 (50.64)
Senate Majority -60.50
 (39.19)
House Appropriations 250.89 *
 (97.21)
Senate Appropriations 169.37 **
 (37.16)
House Armed Services 311.73 **
 (93.38)
Senate Armed Services 404.00 **
 (64.84)
Population 6.78 **
 (1.07)
Gunbelt 369.86 **
 (19.14)
Divided Government -5.60
 (48.82)
Constant 405.97 **
 (52.58)
[R.sup.2] (within, .189
 between, overall)
F (for FE)/Wald 603.15 **
 [chi square]
Hausman
Breusch-Pagan
T 21
N 50

Note: Robust or panel-corrected standard errors are in parentheses.
* indicates coefficient is statistically significant at p < .05,
**p < .01. "Fixed effects" is the fixed-state-effects model.
"Random effects" requests the generalized least squares estimator
of the random-effects model. The third model is ordinary least
squares with panel-corrected standard errors. In the Hausman test,
[H.sub.0] = random effects consistent and efficient. When
significant, the Breusch-Pagan test reveals the presence of random
effects.

TABLE 2
The Marginal Effect of Presidential Support on Procurement Dollars
Per Capita in First and Second Terms

 First Term Second Term

Fixed effects 3.13 (3.19) -4.21 * (1.75)
Random effects 3.21 (3.31) -4.52 * (1.76)
Panel-corrected 4.20 (10.01) -7.07 * (2.77)

Note: Robust or panel-corrected standard errors are in parentheses.
* indicates coefficient is statistically significant at p < .05.
Fixed effects is the fixed-state-effects model. "Random effects"
requests the generalized least squares estimator of the random-effects
model. The third model is ordinary least squares with
panel-corrected standard errors.