The Wartime President: insights, lessons, and opportunities for continued investigation.

Howell, William G. ; Jackman, Saul P. ; Rogowski, Jon C. 等

In the course of sending soldiers off to fight and die, presidents wield the greatest powers available to leaders of sovereign nations. But do wars provide presidents with opportunities to go further still--that is, to use foreign crises as justification (some would say pretext) for advancing domestic policy initiatives, particularly when such initiatives only tangentially relate to the war effort itself?.

Viewed from one vantage point, one naturally inclines to the affirmative. During the early stages of foreign crises, the public regularly demands forthright action; inter-branch conflicts often subdue; and the exigencies of foreign crises may convince domestic interest groups to defer to Congress and the president, when in times of peace they might readily obstruct. As John Kingdon (2002) famously argues, crises constitute "focusing events" that pry open "windows of opportunity" for major policy change. And extending Kingdon's insights, David Mayhew notes, "wars seem to be capable of generating whole new political universes. They can generate new problems and open up policy windows, thus often fostering new policies, but they can also generate new ideas, issues, programs, preferences, and ideologies and refashion old electoral coalitions--thus permanently altering the demand side of politics" (2005, 473). These facts bode well for the president. As the individual primarily responsible for marshaling a response to the foreign crisis, the president is well situated to harness these forces in the service of his (someday her) policy agenda.

Consider, then, how the events of September 11, 2001, strengthened George W. Bush's influence at home. In the week following the attacks, Treasury Secretary Paul O'Neill publicly cajoled people to "buy American," Vice President Dick Cheney urged Americans to "stick their thumb in the eye" of the terrorists by purchasing stocks, and the president directed his officials to devise a plan to support the airline industry. (1) Within days, the House and Senate quickly fell into line, passing an airlines bailout bill by 356-54 and 96-1, respectively. During the brief congressional debate, House Minority Leader Dick Gephardt (D-MO) received a standing ovation for highlighting the tragic events that necessitated bipartisanship. Senator Hillary Clinton (D-NY) expressed strong support, explaining, "We need to look at transportation again as part of our national defense." (2) Her colleague Charles Schumer (D-NY) described the times as "a new era where everyone has to give a little bit." (3) With the start of the war in Afghanistan in early October and public approval ratings hovering around 90%, Bush moved swiftly to parlay such sentiments of unity to other items on his domestic economic agenda. He depicted economic growth as "part of the war we fight." (4) He characterized his stimulus proposal as an "economic security plan." (5) He transformed tax cuts into a test of patriotism, calling for legislators to "act quickly to make sure that the American people understand that at this part of our homeland defense, our country and the Congress is united." (6) Indeed, as one Democratic aide noted, "The president has so much power as a result of what happened he thinks he can use that to force huge concessions on a range of issues." (7)

Bush's strategy of linking domestic policy reforms to concerns about war and national security, however, may not be foolproof. Other facts about war may dampen the president's chances of advancing major policy initiatives. For instance, the sheer costs of war may introduce budgetary constraints that limit domestic policy initiatives. As wars become protracted and as death tolls mount, public support for the president may dwindle. And the time and efforts spent maintaining political support for ongoing military ventures may further reduce the resources needed to build the necessary coalitions for enacting domestic policy initiatives.

Contrast, then, Bush's first-term success at enacting economic reforms with Bush's second-term efforts to reform Social Security. In 2005, Bush designated Social Security reform as the centerpiece of his State of the Union address. Overhauling a massive national entitlement program would require money, voter support, political capital, and presidential attention. Precisely these resources, however, were being expended by ongoing efforts to enlarge the Defense Department and wage the increasingly unpopular war in Iraq. Reacting to the war's erosion of Bush's public approval ratings, congressional expert Ross K. Baker pointed out, "I think there is a very acute realization on the part of Republicans that they no longer can hitch their careers to his popularity. That, combined with the new aggressiveness by the Democrats, means you're seeing basically a Bush agenda that is largely being derailed." Others agreed. Political analyst Stuart Rothenberg characterized the Iraq War as "a cloud over everything." First-term Senator Jim DeMint (R-SC), who took office with hopes of revamping Social Security, lamented, "I feel like every morning, I wake up, get a concrete block and have to walk around with it all day. We can't even address the issues." (8) Preoccupied by Iraq and other foreign policy concerns, the president eventually quieted his calls for Social Security reforms, (9) realizing that foreign wars had led to his "circumscribed sway over Capitol Hill." (10)

Do wars regularly usher in new opportunities for presidents to advance policy change? Or do wars instead undercut presidents' policy initiatives? The short answer is that we do not know. As the above anecdotes suggest, the impact of war on the power that presidents wield at home is hardly obvious. And for all that has been written about wartime presidents--and lest there be any doubt, an extraordinary amount has been written (see Howell 2011 for a review)--few social scientists have built the data sets needed to systematically gauge the varying impacts of war on presidential power. Those who have, meanwhile, nearly universally have assigned war the status of an ancillary control variable in a regression meant to shed light on some other aspect of presidential power (for more on this point, see Howell and Johnson 2009).

While quantitative work on presidential power and war has run in short supply, efforts at theory building have been nonexistent. To be sure, a great deal has been written about the ways in which wars influence public support for the president (see Aldrich et al. 2006 on this point). And legal scholars, in particular, have made much of the ways in which presidents have used wars as a pretext for ever more expansive interpretations of their Article II powers (see Howell, 2009 for a review). We lack any theory, however, about how wars might increase a sitting president's actual influence over the content or implementation of public policy. While grand narratives about particularly audacious displays of wartime presidential power are ample, no one has developed theory with clear microfoundations that isolates the features of war that have the potential to turn legislative deliberations to the president's advantage.

In The Wartime President, we offer correctives to these empirical and theoretical deficiencies. Empirically, we collect data and suggest tests that are expressly designed to mitigate the standard identification and endogeneity concerns that pervade research on separations of powers issues generally and the American presidency in particular. Theoretically, we develop a formal model of interbranch relations that explores one mechanism through which wars might generate policy outcomes that better approximate presidential preferences.

In this article, we summarize the main empirical and theoretical contributions of our forthcoming book. In the interests of space, we omit many of the model extensions and robustness checks to which we subjected our analyses. We also set aside the case studies, literature reviews, and efforts to distinguish our own theory from other plausible explanations about the relevance of war for interbranch negotiations. In an effort to reach as broad an audience as possible, moreover, we summarize our formal model without any mathematics. What follows, then, is intended as much to entice as to inform. Readers who harbor doubts about our claims or who want to see further analysis are encouraged to consult the book itself.

Evidence of a Wartime Effect on Presidential Power

Our empirical evidence is presented over four chapters--two quantitative followed by two chapters of case studies. In this section, we summarize some of the main findings from the two quantitative chapters. The first focuses on budgetary outcomes, while the second canvasses members' voting patterns on the broader universe of roll calls.

Wartime Appropriations

Appropriations present an especially profitable opportunity to assess Congress's variable willingness to support the president during war and peace. Every year, after all, presidents must issue a budget proposal, and Congress must subsequently enact a final set of appropriations. Unlike the traditional legislative process, the appropriations process does not permit presidents to remain silent on particularly controversial bills or members of Congress to refuse to cast judgment on presidential proposals. (11) Thus, by examining proposed and enacted budgets, we, as observers, have a basis upon which to gauge the difference.

Budgets, moreover, are not saddled by the same basic facts about the legislative process that make gauging presidential influence so difficult. Bills regularly undergo substantial amendments, with final enactments often yielding hazy renderings of original proposals. But with budgets, the president must take a public position on the amount of money to be allocated to an executive agency, and members of Congress must decide how much to award. The difference between these quantities, therefore, constitutes a clear and continuous measure of presidential success--and one, moreover, that can be readily compared across policy domains. The more that members of Congress wish to accommodate the president's proposal, the smaller the observed differences between proposed and actual appropriations will be.

To ascertain whether the observed differences between proposed and enacted budgetary allocations systematically vary in times of war and peace, we first require an operational definition of each. To minimize controversy, we focus on the largest and most consequential military ventures in the modern era. Following a host of scholars, including Clark (2006) and Epstein et al. (2005), we classify wars as including World War II, the Korean War, the Vietnam War, the Gulf War, and the wars in Afghanistan and Iraq that followed the attacks on September 11, 2001. By any measure, these constitute the most significant military actions taken by the United States in the last 75 years. Should we fail to find evidence of heightened deference to the president in these wars, we are not likely to do so in other, smaller military engagements.

With our definition of war, we then track budgetary proposals and allotments from 1933 to 2006 for the same 77 agencies found in Kiewiet and McCubbins (1991). (12) For each agency-by-year observation, we identify the president's budget proposal and the actual appropriations allotted to that agency (standardized to 1983 dollars). When considering all the observations in our data set, the president requests on average $4,382,662 per agency each year, while Congress allocates $4,368,323, a difference of $14,339. Over the course of their lifespans, 48 out of 77 agencies received less money from Congress per year than the president requested. Likewise, Congress granted fewer real dollars per agency than requested by the president in 55 out of 74 years.

Budgets routinely increase during times of war. On average, presidential budgetary estimates grow from $4.1 to $5.1 million per agency-year during war, and likewise Congress's approved budget increases from $4.1 to $5.1 million. Note further that in times of war, Congress allocates more money per agency than the president requests (by $19,691), while during peace Congress allocates less per agency than the president requests (by $26,629). The same trends hold when these data are disaggregated into defense agencies and nondefense agencies. Presidential budgetary estimates increase, on average, from $1.4 to $1.7 million during war for nondefense agencies and from $15.4 to $18.3 million for defense agencies. Likewise, congressional appropriations increase from $1.4 to $1.7 million for nondefense agencies and from $15.1 to $18.2 million for defense agencies. These figures indicate that presidents receive an additional $16,905 more than they requested during war years for nondefense agencies and that the gap between their request and Congress's appropriations decreases by just under $170,000 for defense agencies. Further disaggregating the data by agency, we find that 52 out of 76 agencies have higher presidential proposals during war, 53 out of 76 agencies have higher congressional appropriations, and that the gap between these two values (proposals and appropriations) is smaller during war in 42 out of 76 agencies.

The distribution of differences between proposed and final appropriations for each agency in each year reveals substantial skewness. As our dependent variable, therefore, we take the natural log of the absolute value of the difference between proposed and final appropriations for each agency in each year. Larger values of this variable indicate greater discrepancy between what the president requested and what Congress ultimately granted; and smaller values indicate less discrepancy. (13) When interpreting the effects of any particular covariate in our regression models, positive values indicate an expansion of the difference between proposed and final appropriations, while negative values indicate a contraction.

The likelihood that Congress accommodates the president's requests, of course, depends upon more than just the presence of peace or war. Most importantly, perhaps, it depends upon just how much money the president requests. At the margin, we expect that Congress will look more favorably upon smaller requests than larger ones. We therefore control for the logged value of the president's proposal for each agency in each year.

Congress's response to the president surely also depends upon the level of political support that he enjoys within its chambers. Presidents who confront congresses with large numbers of ideological or partisan supporters are likely to secure appropriations that more closely approximate their requests than presidents who face off against congresses dominated by the opposition party. Following Kiewiet and McCubbins (1985a, 1985b), we therefore control for the percent of House seats held by the president's party in each year.

We also include three economic indicators: the average unemployment rate during the year when appropriations are proposed and set, the national growth rate since the previous year, and the total budget deficit from the previous year. One might expect that presidents receive greater popular support when the economy is doing well and, further, that the economy might do better in times of war due to increased government spending. By controlling for these three economic indicators, we preclude their ability to bias the effect of war on presidential bargaining success.

Finally, all our statistical models include fixed effects that account for all observable and unobservable time-invariant characteristics of individual agencies and presidents. Identification in the model, therefore, comes from changes in appropriations within agencies and within presidential administrations.

Column 1 of Table 1 presents our preferred estimate of the impact of war on Congress's willingness to abide the president's budgetary requests. The effect is negative, substantively large, and statistically significant. During periods of war, differences between proposed and final appropriations attenuate substantially. Taking the inverse log of the point estimate, we find that during periods of war the average discrepancy between proposed and final appropriations for our sample of agencies decreases by roughly 26%.

The other variables in the model generally behave as expected. Presidents who confront congresses with larger numbers of House copartisans enjoy higher levels of budgetary success than do presidents who must work with larger numbers of partisan opponents--an effect that is substantively large and statistically significant. Congress demonstrates greater accommodation to the president's proposed budget when national growth rates are large, and less accommodation when available revenues (as measured by budget deficits) are relatively scarce. Interestingly, though, presidents also experience more accommodation from Congress when unemployment rates are high. We also find that Congress appropriates monies that more closely approximate smaller budgetary requests than it does for larger ones--another effect that is substantively large and highly statistically significant. Finally, the agency and presidential fixed effects, which are not reported in order to conserve space, are both jointly significant.

Subsequent columns of Table 1 report the recovered estimates of models that recharacterize the dependent variable. In the second column, we predict the logged absolute value of the difference between proposed and final appropriations as a percentage of the president's proposal for each agency in each year. Once again, we find that final appropriations more closely approximate presidential proposals during war than during peace. Moreover, the other covariates generate comparable estimates to those observed in our preferred specification.

We also explore characterizations of the dependent variable that distinguish instances when Congress appropriates larger amounts than the president's request from instances when Congress appropriates smaller amounts. When Congress refuses to appropriate the full amount of money requested for a specific agency, it clearly constrains the agency's ability to either perform at a level of activity that president would like or to pursue specific policy functions that constitute presidential priorities. But given the president's ability to influence, ex post, how monies are spent--whether by discouraging bureaucrats from vigorously enforcing their mandate, reprogramming or transferring funds from one account to another, or simply impounding funds, as was done for much of the period under investigation--Congress may have a difficult time inducing agencies to either more vigorously pursue their mandate or to administer a larger number of policy activities. Congressional influence, under this account, primarily acts as a constraint rather than a stimulant to executive activity. Congress can readily impede executive functioning, but it has a much more difficult time either galvanizing existing executive functions or jumpstarting altogether new ones. To account for this asymmetry, we generate a dependent variable that continuously measures final appropriations that are lower than the president's proposal, but that treats appropriations that exceed proposals as equivalent to ones that exactly meet them. The results, presented in column 3, compare well with those observed in our preferred specification. Once again, we find a negative, substantively meaningful, and statistically significant relationship between war and the observed discrepancy between proposed and final appropriations. In fact, the magnitude of the effect of war increases significantly in this specification of the dependent variable, such that war increases presidential bargaining success by approximately 35%.

We also have estimated models that employ still more characterizations of the dependent variable, and in each instance we found comparable results. For instance, we have examined the dependent variable as the proportion of the president's proposal that is enacted into law. In an effort to address the possibility of asymmetric effects associated with under- and overappropriations, we also have set an upper limit on this proportion at one. We have utilized a measure developed by Brandice Canes-Wrone (2006), wherein we calculate the difference in annual percentage changes in presidential proposals and annual percentage changes in final appropriations. And finally, we have estimated models that characterize the differences between proposed and enacted budgets in raw, as opposed to logged, terms.

Just as the main results associated with war are robust to a variety of different characterizations of the dependent variable, so too do they hold across a host of alternative model specifications. We have estimated models that omit the president's proposal from the regressors; that include the raw value of the president's proposal; that include subsets of the economic variables; that account for the partisan composition of the Senate either as well as or in lieu of the House; that exclude the substantial number of president and agency-specific fixed effects; that include measures of the public's demand for government services; and that include controls for periods of unified and divided government, election years, each agency's budget authority from the previous year, and the president's term in office. In every instance, the main results hold.

We also have investigated the possibility of strategic proposal making on the part of the president. What presidents propose surely depends upon what they expect Congress will grant. And if these expectations themselves are a function of war, then our recovered estimates may be either upwardly or downwardly biased. As it turns out, however, the average peacetime and wartime agency requests by those presidents who served in both war and peace, and who therefore contribute to our estimates of the effect of war, are nearly identical. And when estimating systems of equations that attempt to leverage exogenous variation in presidential requests, we again recover negative and statistically significant effects associated with war.

Voting in War and Peace

In the book's second empirical chapter, we evaluate whether members compile voting records during wartime that better reflect the president's preferences than in peacetime. A basic intuition structures these analyses: if a conservative president takes the nation to war, we ought to observe a shift in members' voting records to the ideological right. And when this war ends, we then should witness a shift to the ideological left. When a liberal resides within the White House, meanwhile, members of Congress should shift in exactly the opposite directions as the nation enters and exits wars.

To investigate such possibilities, ideally we would observe identical members of Congress voting on identical bills immediately before and immediately after the outbreak of war, when all other potential confounders might be assumed to be constant. Lacking such a natural experiment, we exercised the next best option and split the universe of roll calls based on the presence or absence of war. Using the raw congressional roll call data, we use a Bayesian approach to ideal point estimation (Clinton, Jackman, and Rivers 2004) that enables us to recover estimates of each legislator's voting record. However, we cannot simply compare peacetime and wartime ideal points that have been estimated independently because the scales may have shifted or stretched. (14) To solve this problem, we use two interest groups--the American Conservative Union (ACU) and Americans for Democratic Action (ADA)--as "bridge" observations (see Bailey 2007; Bailey and Chang 2001 for a discussion of bridges in ideal point estimation) that link the peacetime and wartime roll call voting record. In doing so, we make two assumptions: first that the ideological scales across the peacetime and wartime periods remain fixed and, second, that the interest groups took ideologically consistent positions across both periods.

Operationally, we split the set of roll calls within any given Congress at the date on which a war began. We then construct separate matrices for the peacetime and wartime roll calls in both the House and Senate in which the rows correspond to legislators and the two interest groups. Column entries indicate whether the legislator or interest group supported or opposed each bill. We then generate member ideal points for the roll call votes that occurred prior to the outbreak of war, where negative ideal points identify more liberal members, and positive ideal points identify more conservative members. After recovering the prewar estimates, we estimate the postwar ideal points for all members of Congress while constraining the estimates of the ACU and ADA to equal their prewar estimates. We then compare the peacetime and wartime estimates at both the individual and chamber levels.

Given our empirical approach, it would not be surprising to find null war effects. By virtually all accounts, after all, congressional voting behavior is remarkably stable over time. As Keith Poole remarks, "based upon the roll call voting record, once elected to Congress, members adopt an ideological position and maintain that position throughout their careers--once a liberal or moderate or conservative, always a liberal or moderate or conservative" (2007, 435). Comparisons of individual members over short periods of time, and especially those that focus on trends within a single congressional term among members who did not switch parties, can be expected to reveal overwhelming stability in individual voting behavior; moreover, no external event, very much including war, is likely to dislodge members from their chosen ideological footings. Accordingly, the empirical tests that follow constitute a stringent test of the proposition that members' voting behaviors differ during war and peace.

We have the greatest confidence in estimates associated with the 107th Congress. The attacks of September 11, 2001, occurred roughly midway during a congressional session, offering plenty of peacetime and wartime votes to analyze. The transition from war to peace during this period, moreover, occurred in very short order, reducing concerns about treatment contamination. And precisely because they caught the nation off guard, the attacks can be considered plausibly exogenous to the prior legislative agenda. Given these and other technical strengths, we focus on the results from the pre- and post-9/11 periods.

Figure 1 plots the densities of members' estimated peace- and wartime ideal points in the 107th Congress. The dashed lines indicate the peacetime scores, and the solid lines indicate the wartime scores. Larger values indicate more conservative voting behavior. In both the House (left panel) and Senate (right panel), we see marked shifts to the right. In the House, the movement appears to be concentrated among Republicans, while in the Senate both parties shift rightward. The mean and median shifts in the House were +0.54 and +1.28, respectively; in the Senate, they were +1.21 and +1.26.

Figure 2 again plots the pre- and post- scores for every member of the 107th House and Senate. This time, however, individual scores before the outbreak of war are aligned on the x-axis, and wartime scores appear on the y-axis. If a member's voting behavior did not change at all, then she will locate right on the 45-degree line. Scores above the 45-degree line indicate movement in the conservative direction, and scores below the 45-degree line reveal movement in the liberal direction. Observations whose peace- and wartime scores are significantly different from one another at p < .001 are solid, and the rest are shaded. Plainly, the vast majority of members in both chambers appear above the line. Indeed, all 95 members of the Senate and 323 of the 362 members of the House who reveal statistically significantly different pre- and post- scores document movement in the conservative direction.

[FIGURE 1 OMITTED]

These results are robust to a wide variety of alternative empirical approaches. More specifically, these shifts do not appear to be an artifact of changes in the agenda, party control, the electoral calendar, rising conservatism across all levels of government, factors that preceded the terrorist attacks on 9/11 and precipitated the war, or our choice of identification strategies. We also find similar effects when limiting the sample of votes to purely domestic legislation or bills that were the most visible or highly contested.

In examining prior wars, we face a variety of new empirical challenges. For instance, interest group positions are unavailable for World War II and the Korean War. And because the Vietnam and Gulf wars began and/or ended early in a congressional term, we must pool across congresses. Employing a range of techniques to address these and other issues, we find that the United States' entry into World War II coincided with a significant shift to the ideological left; the end of the war, however, evoked a significant shift to the ideological right, away from the orientation of the president then in office. We find limited evidence that the beginning of the Korean War induced Congress to vote in ways that better reflected the ideological leanings of President Harry Truman, while the end of the war coincided with a significant shift away from President Dwight Eisenhower. We find no consistent evidence that the beginnings of the wars in Vietnam and the Persian Gulf induced members to vote in ways that better reflected the preferences of the presidents then in office, while the end of the Vietnam War yielded a Congress less inclined to reflect the ideological orientation of its president. Finally, we find no consistent evidence of movement away from President George H.W. Bush upon the end of the Persian Gulf War.

[FIGURE 2 OMITTED]

It appears, then, that some wars increased presidential success more than others. At the outset of World War II and the Afghanistan War, members of Congress began to vote in ways that better reflected the ideological orientations of the presidents then in office. But evidence for the beginnings of the Korean, Vietnam, and Persian Gulf Wars is more mixed, as members of Congress did not line up behind the president in consistent ways. And with the termination of every war we analyzed, members of Congress shifted away from presidents--suggesting that whatever "public treasures," "honours and emoluments," "patronage" and "laurels" had to offer the president, to borrow from James Madison in the Pacificus-Helvidous Debates, they were promptly withdrawn when these wars ended.

From What to Why: Isolating a Causal Mechanism

Our empirical findings suggest that at least some wars substantially increase the president's influence over policy. But what is it about war, exactly, that causes members of Congress to vote with the president in war when, in peace, they would oppose him? Why, that is, do transitions from peace to war and then back to peace again encourage at least some members of Congress to change their voting behavior?

Wars, of course, induce all sorts of changes to the domestic polity, altering the economic, social, and political landscapes in subtle and not-so-subtle ways. Rather than survey all possible dimensions of war, however, we isolate one: the ways in which wars alter the terms of public debate about policy change, generally, and the possibility that in these debates national (as opposed to parochial) considerations will predominate, in particular. Wars, we stipulate, have the potential to bring the nation's health and security into stark relief; affirm citizens' shared status as Americans; and underscore Americans' common heritage, values, and shared fates. And so doing, wars encourage members of Congress to prioritize national political outcomes over policy outcomes that favor their local constituencies.

To investigate the relevance of the nationalization of policy debates for legislative processes, we take a step back from the historical details--the personalities, contingencies, and cultural norms--that define particular wartime episodes in American history. Instead, we venture into the more abstract environment of formal theory, which has made only a selective appearance in the presidency subfield. For our purposes, however, it is ideally suited for the task at hand. Through formal models, after all, we can discern the particular ways in which changes in how members of Congress think about public policy have consequences for the propensity of these members to side with the president. Having specified the parameters of the political environment that influence negotiations between the president and Congress, we can readily identify how war--or more exactly, the ways in which war increase the nationalization of politics--induces heightened congressional deference to presidential prerogatives.

A complete presentation of our model is available in our book. The intuition behind the model, however, can readily be grasped without any formal training in game theory. The model investigates the bargains that are struck between a president and representative member of Congress. Ex ante, neither of these players is assumed to agree with the other. Indeed, the whole point of this exercise in theory-building is to identify features of politics that can be linked to war and that bear upon the president's ability to advance his policy agenda, even if they should disagree with each other.

The model begins with the president making a policy proposal. To be clear, the president's proposal need not be a formal legislative proposal that Congress amends or votes down. It can be as informal as a speech or press release. In the budgetary process, the proposal is formally offered, and the president's annual budget proposal necessarily precedes the final appropriations allocated by Congress. But even in legislative processes, presidents have ample ways of seeing that their policy initiatives receive a hearing in Congress.

Upon receiving the president's proposal, the representative member of Congress is free to pass a final bill, as she so chooses. The policies generate outcomes, from which the president and Congress derive their utility. Nothing about the president's proposal is intrinsically binding; Congress is free to adopt any policy it would like. And as a consequence, nothing about the model presumes the existence of a presidential advantage over policy making. Quite the opposite, in fact. Having stripped the president of any veto or unilateral powers, our model grossly understates the actual influence that he wields over policy outcomes.

Three features of our model help explain why the president, whose proposal constitues more cheap talk, can extract policy concessions from a disagreeable Congress. First, our theory explicitly recognizes that a single policy can generate different outcomes for different constituencies. Policies routinely yield outcomes for a particular district or state that look quite different from the outcome for the nation as a whole. Reducing steel tariffs, cutting farm subsidies, or increasing insurance regulations may improve the standard of living for the country as a whole, just as they devastate local economies in Pennsylvania, Kansas, and Connecticut. Conversely, federal grants and aid may materially improve lives in specific communities without having much of an impact at all on the country in the aggregate. Thus, when we think about how policies manifestly change the world, we allow for the possibility that a single policy can affect a particular state or district in ways that look quite different from the effect on the nation in its entirety. Moreover, just as foreign policy yields outcomes at the local and national levels, so to does domestic policy.

Our theory further recognizes that individual members of Congress serve different constituencies than do presidents. Whereas members of Congress serve districts or states, presidents serve the entire country. This basic fact has crucially important consequences for the deliberations that occur across branches of government. When choosing among policy alternatives, politicians prioritize those that yield outcome favored by the constituents whom they have been elected to serve. As a consequence the distinct constituencies of members of Congress and the president lead the two branches of government to evaluate policies on different grounds. Precisely because they serve the nation as a whole, presidents view national outcomes as paramount, and a representative member of Congress, meanwhile, must make a more complicated decision when enacting policy. Individually, each member of Congress is expected to provide benefits to one particular district. Simultaneously, members of Congress are evaluated as part of a collective branch of government intended to represent the entire country. As such, each legislator is torn between what is good for the country as a whole and what is good for their own constituents--and, hence, members of Congress may assign at least some value to the former category, as distinct from the latter. Foreshadowing our later conclusions, one might imagine that the weight legislators place on national outcomes relative to local outcomes shifts markedly during transitions to war.

Third, and finally we draw on a large American political institutions literature in assuming that there is a distinction between the policies that Congress adopts and the associated outcomes they generate. All of the political actors in our story initiate and pass policies, but when they do this, their primary interest lies not in the policy itself, but rather in the outcome of the policy--the way in which policy translates into changes in the material world. As Keith Krehbiel (1992, 66) elaborates, "utility is determined by outcomes--not by policies. That is, passage of a given policy has no bearing on a legislator's utility apart from the outcome associated with that policy." As a result, politicians cannot always shape the world as they would like. They cannot simply dictate that inflation will stabilize or that unemployment rates will drop. Rather, politicians must design policies that they hope will induce desired changes to the material universe. They must learn how specific policy instruments (over which politicians have direct control) are connected to specific policy outcomes (from which politicians derive their incentives). And the mapping of policies into outcomes, it bears emphasizing, can be remarkably complex.

Political disagreements persist, at least in part, because the acquisition of information about how policies translate into outcomes is costly--sometimes prohibitively so. To discern how increases in the minimum wage, heightened oversight of financial industries, and deeper investments in human capital manifest in local and national outcomes, members of Congress and presidents require information. And the required information is not freely available. Expertise comes at a cost. A great deal of politics generally, and the design of political institutions in particular, therefore involves creating incentives for politicians to pay these costs, collect reliable information about the likely outcomes of different policy initiatives, and then communicate this information to others.

In our story, the nature of uncertainty about how policies translate into outcomes is slightly different than in the conventional construction. While we do build on the extant literature, our conception of the mapping of policies into outcomes is more nuanced. Assuming the president has acquired expertise on how policies translate onto national outcomes, and moreover, that he makes a policy proposal to a Congress that has not acquired the relevant expertise, Congress is able to extract some information about the mapping of policies into outcomes from the proposal. Unlike in Krehbiel's seminal work, though, the amount of information that can be extracted is limited and decreasing, as Congress's enacted policy deviates farther from the president's proposal. This captures an idea that Callander (2008) refers to as proportional invertibility. To see the intuition behind it, consider a simple example: An expert reveals how a change of the minimum wage to $7.25 translates onto the national unemployment rate. Knowing this single translation and that the mapping process is proportionally invertible, the layperson can readily assess how a minimum wage of $7.24 or $7.26 might influence the national unemployment rate. Similarly, albeit with less confidence, the layperson can conjecture how a minimum wage of $7.00 or $7.50 is likely to translate onto a national unemployment rate. But this single translation yields relatively little additional insight about the effect of a minimum wage of $15 and nearly none whatsoever about the outcome associated with a minimum wage of $50.

With these three key insights--that policies generate multiple and distinct outcomes, that the political relevance of these outcomes differs for members of Congress and the president, and that expertise is required to ascertain the relationship between any policy and any outcome--we are able to explain many features of interbranch negotiations. Most important to us, our theory yields clear and crisp predictions about presidential success. Each branch of government can be expected to acquire expertise about the relationship between policies and those outcomes that are politically most salient to them. Hence, presidents obtain an informational advantage with regard to national outcomes, while members of Congress hold the advantage when considering district level outcomes. But when the evaluative criteria of either branch of government change, the content of enacted policies may shift dramatically. As long as they stand by the parochial interests of their constituents, legislators can be expected to vote against the president. But when they put the interests of the country before the interests of their constituents, legislators can be expected to look more favorably on the president's policies. Legislators, after all, remain relatively uncertain about the national outcomes of policies, and the cost associated with this uncertainty increases the more that legislators amend a proposal issued by a president who has invested in this policy expertise. As a result, even when politicains' ideal points and uncertainty are held constant, policy outcomes better reflect the president's preferences when legislators assign greater importance to the set of outcomes over which the president holds an informational advantage.

Though this theory is broadly applicable to interbranch bargaining, it speaks directly the role of war in domestic politics. When the country enters war, members of Congress become more concerned with national policy outcomes. Recognizing that the president knows more about national outcomes, members then defer more to presidential proposals, for the uncertainty cost of deviating from the president's proposal rather suddenly is magnified. Thus, even if the president and Congress disagree ideologically about the ideal national outcome, the legislator will still defer more to the president in war than in peace in order to avoid the perils of guessing the likely national outcomes of public policy. In this sense, the president's relative expertise about the national outcomes of policies serves as a special source of influence when the nation initiates war. (15)

With our theory, we have a basis for making sense of why wars might augment presidential success. When the nation transitions from peace to war, legislators can be expected to assign greater value to the national implications of policies and relatively less to the local implications of policies. During war, therefore, both the president and Congress fixate on the national outcomes of policies, which encourages legislators to more closely abide the expertise that presidents retain over how policies map onto national outcomes. As a result, presidential proposals during war--whether involving foreign or domestic policy--can be expected to receive a more favorable hearing than they do in peace.

In addition to identifying a plausible mechanism by which wars may augment presidential influence, our research illuminates two important aspects of the empirical record. Recall that our estimated effects of war were particularly prominent in World War II and the post-9/11 period. For reasons we discuss at some length in the book, these wars also did the most to nationalize policy debates. Though the Korean, Vietnam, and Persian Gulf Wars all involved large troop deployments, sustained substantial casualties, and drew upon substantial economic resources, they did not reshape the ways in which politicians evaluated public policy to the same degree as did the two largest wars in our study. It is no accident, then, that the evidence we uncover for these wars appears more muted.

In both the budgetary and roll call analyses, the estimated effects of war appear just as strong for domestic policies as they do for foreign policies. In some instances, we observe larger war-related effects for domestic policies than foreign ones. Once again, our theory helps explain why. Our theory highlights the importance of marginal changes in the relative weights that legislators assign to national versus local outcomes. Precisely because legislators can be expected to value the national implications of foreign policies during peace, the transition to war may not be especially consequential. On domestic policies, however, the transition to war can fundamentally shift the terms of legislative debate. As we document in the book extensively, with the outbreak of some major wars, debates about purely domestic policies suddenly became infused with concerns about national security and identity. And this turnabout in how politicians evaluate policy has important consequences for the president's ability to advance his policy agenda.

Implications for Research on the American Presidency

By suggesting avenues for further inquiry and offering rationales for findings that, ex ante, one might not expect, our research on wartime presidents illustrates ways in which empirical analyses can profitably inform (and be informed by) theory. Our real hope for the book manuscript, however, lies not in offering definitive answers to longstanding questions, but rather in setting the field on more profitable footing to investigate issues involving presidential power. The issue of war is too important for presidency scholars to leave to lawyers and historians. The success of this book, from where we sit, will be gauged by what it begins rather than what it settles.

The presidency subfield of American politics would do well to investage several important areas of continued empirical study. Our core empirical tests, after all, examine policy outcomes and the voting habits observed within Congress as a whole. Future work should do more to distinguish the variable impacts of war within Congress and, further, to do so in ways that draw upon the insights from our theory. Nothing about the model precludes the possibility of heterogeneous wartime effects among members of Congress. If, for instance, members only reconsider the criteria by which they evaluate policy alternatives when they support the broader aims of a war effort, then evidence of wartime changes in voting behavior should be confined to this subset of the House and Senate. Similarly, if members face particularly strong pressures to deliver constituency benefits during election years, then the effects of war may vary for the subset of Senators who face reelection relative to those who can count on remaining in office for a longer period of time. Surely, other possibilities exist.

Future empirical work also should investigate whether events beyond war might augment the salience of national outcomes and thereby bolster presidential influence. And when looking for such events and such influence, scholars should remain especially attuned to the possibility that the effects will be more localized than those observed for large-scale military ventures. As we report in our book, we found no evidence that smaller military ventures in the 1990s, presidential scandals, or the Iran hostage crisis altered members' general voting behavior. One can well imagine, however, that the effects of economic crises might be confined to economic policy, in much the same way that concerns about international drug trade might augment the president's influence over just drug policy. In this vein, scholars also might consider an interesting corollary to our central argument: as politics become more localized, congressional support for the president should decline. When a legislator finds cause to prioritize local over national outcomes--consider, for instance, the sectional pressures members of Congress face when their district or state has just experienced a massive flood or a deadly crime spree--she will grant less support to the president.

In addition to continued empirical scrutiny, the book invites a variety of theoretical extensions. As we have already noted, our theory presents a rather anemic presidency, one whose capacity to influence public policy is restricted entirely to proposal making. Future work ought to incorporate some of the other tools available to the president, some of which are enshrined in the Constitution (e.g., the veto) and others of which are the joint product of presidential initiative and congressional delegation (e.g., the opportunity to forego the legislative process and set policy unilaterally). While these richer models can be expected to yield heightened assessments of presidential influence, it remains to be seen whether they materially alter our assessments of the marginal impact of war on presidential bargaining success. Additionally, these models may profitably enhance our understanding not only of the effect of war on public policy, but also its effects on the strategies presidents employ to advance their policy interests.

Scholars also would do well to investigate the conditions under which multiple members of Congress, as opposed to a single representative member, invest in expertise about the intermittently discrete and overlapping mapping processes of policies into national and local outcomes. A substantial amount of existing scholarship explores the ways in which congressional committees help solve the basic informational problems faced by Congress as a whole. What remains unclear, though, is how committees, or any other congressional institution, affect the president's calculus about whether to acquire expertise and whether, in turn, the existence of such institutions in any way alters our predictions about war and presidential success.

We also recognize that opportunities for presidential influence extend well beyond the formal lawmaking process. And for all intents and purposes, the empirical and formal literatures in American politics have paid barely any mind to questions about the broader relationships between war and presidential power. How do wars influence trends in domestic unilateral policy making (e.g., the issuance of executive orders, proclamations, national security directives), signing statements, and foreign policy making (whether by treaty or executive agreement)? Does Congress tend to delegate broader powers to the president during war? And if so, do these powers extend beyond military matters? When issuing emergency legislation, does Congress facilitate the exercise of presidential power long after a particular crisis has passed and in domains that only tangentially relate to the crisis itself?. How do wars affect both the design and presidential control over the federal bureaucracy? A handful of empirical studies do investigate these questions. Few, however, explicitly address the nettlesome research design problems associated with this line of work. And none is grounded in deep theory. Instead, and as we detail in our book, the preponderance of scholarship on these topics consists of sweeping historical narratives and normatively driven legal studies.

Our book draws attention to still broader themes of the American presidency, on which we hope others will elaborate. For starters, it places at the very center of its analysis a fact that presidency scholars too often observe only idly, and that game theorists who try to formally represent policy-making processes altogether ignore: namely, that presidents sit in a very different place of power than either judges or legislators; and whereas members of Congress work doggedly on behalf parochial interests of their local constituents, presidents represent the entire country. The difference, of course, is one of degree rather than kind. Presidents occasionally pursue a particularistic agenda, just as individual members of Congress periodically set their sights on the nation's welfare. In the main, however, this distinction lends important insights into the modern presidency in its own right and into the kinds of policies that the federal government as a whole will produce and sustain. By constructing a general theory of interbranch relations that accommodates this fact, we are able to step beyond the role of war to consider how shifting constituencies can effect political negotiations and activity.

Beyond their intrinsic import, wars also reveal ways in which events can reshape our politics. And our subfield's treatment of events, for the most part, follow the dictates more of journalism than of social scientific inquiry. As David Mayhew (2005) lamented just a few years after the attacks of 9/11, political scientists too rarely include events within the class of explanatory variables. And among events that cry out for careful examination, wars reside very near the top of the list.

Quantitatively oriented scholars of the American presidency are quite accustomed to examining the political consequences of elections. We now have an extraordinary body of research showing that presidential power surges and recedes as more or fewer likeminded folk reside in Congress. Elections, however, are not the only factor that systematically alters the various checks on and resources for presidential power. Events, many of which do not originate within politics, matter as well; and they do so not merely for the upcoming election, but for the behavior of politicians who already sit in office. Shocks to the American landscape generated by wars, economic crises, and the like can fundamentally reorient the kinds of negotiations that occur within and across the branches of government. It is time that we subjected them to the same kind of careful empirical scrutiny that we do to presidents' public appeals, vetoes, appointments, and unilateral activities.

References

Aldrich, John, Christopher Celpi, Peter Feaver, Jason Reifler, and Kristin Thompson Sharp. 2006. "Foreign Policy and the Electoral Connection." Annual Review of Political Science 9: 477-502.

Bafumi, Joseph, Andrew Gelman, David K. Park, and Noah Kaplan. 2005. "Practical Issues in Implementing and Understanding Bayesian Ideal Point Estimation." Political Analysis 13 (2): 171-187.

Bailey, Michael. 2007. "Comparable Preference Estimates Across Time and Institutions for the Court, Congress, and Presidency." American Journal of Political Science 51 (3): 433-48.

Bailey, Michael, and Kelly Chang. 2001. "Ideal Point Estimation with a Small Number of Votes: A Random-Effects Approach." Political Analysis 9 (3): 192-210.

Callander, Sreven. 2008. "A Theory of Policy Expertise." Quarterly Journal of Political Science 3 (2): 123-40.

Canes-Wrone, Brandice. 2006. Who Leads Whom? Presidents, Policy, and the Public. Chicago: University of Chicago Press.

Clark, Tom. 2006. "Judicial Decision-Making During Wartime."Journal of Empirical Legal Studies 3 (3): 397-419.

Clinton, Joshua, Simon Jackman, and Doug Rivers. 2004. "The Statistical Analysis of Roll Call Data: A Unified Approach." American Political Science Review 98 (2): 355-70.

Epstein, Lee, Daniel Ho, Gary King, and Jeffrey Segal. 2005. "The Supreme Court During Crisis: How Wars Affect only Non-War Cases." New York University Law Review 80 (1): 1-116.

Ferejohn, John, and Keith Krehbiel. 1987. "The Budget Process and the Size of the Budget." American Journal of Political Science 31 (2): 296-320.

Hamilton, Alexander, and James Madison. 2007. Letters of Pacificus and Helvidius on the Proclamation of Neutrality of 1793. Indianapolis, IN: Liberty Press.

Howell, William G. 2009. "Wartime Judgments on Presidential Power: Striking Down but Not Back." Minnesota Law Review 93 (5): 1778-1819.

--. 2011. "Presidential Power in War." Annual Review of Political Science 14: 89-105.

Howell, William G., and Tana Johnson. 2009. "War's Contributions to Presidential Power." In G. Edwards and W. Howell, eds., The Oxford Handbook of the American Presidency. New York: Oxford University Press, 724-46.

Jackman, Simon. 2001. "Multidimensional Analysis of Roll Call Data via Bayesian Simulation: Identification, Estimation, Inference, and Model Checking." Political Analysis 9 (3): 227-41.

Kiewiet, D. Roderick, and Mathew D. McCubbins. 1985a. "Congressional Appropriations and the Electoral Connection." Journal of Politics 47 (1): 59-82.

--. 1985b. "Appropriations Decisions as a Bilateral Bargaining Game between President and Congress.'" Legislative Studies Quarterly 10 (2): 181-202.

--. 1988. "Presidential Influence on Congressional Appropriations Decisions." American Journal of Political Science 32 (3): 712-36.

1991. The Logic of Delegation: Congressional Parties and the Appropriations Process. Chicago: University of Chicago Press.

Kingdon, John W. 2002. Agendas, Alternatives and Public Policies. 2nd ed. New York: Longman.

Krehbiel, Keith. 1992. Information and Legislative Organization. Ann Arbor: University of Michigan Press.

Martin, Andrew, and Kevin Quinn. 2002. "Dynamic Ideal Point Estimation via Markov Chain Monte Carlo for the U.S. Supreme Court, 1953-1999." Political Analysis 10 (2): 134-53.

Mayhew, David R. 2005. "Wars and American Politics." Perspectives on American Politics 3 (3): 473-93.

Poole, Keith T. 2007. "Changing Minds? Nots in Congress!" Public Choice 131 (3-4): 435-51.

WILLIAM G. HOWELL

University of Chicago

SAUL P. JACKMAN

Vanderbilt University

JON C. ROGOWSKI

Washington University

(1.) Glenn Kessler and Mike Allen, "Bush Signals Support for New Tax Cut; President Also Backs Major Bailout of Airline Industry in Bid to Ward Off Recession," The Washington Post, September 18, 2001.

(2.) Previous quotes found in Frnk Swobodaand Martha McNeil Hamilton, "Congress Clears $15 Billion Bill To Aid Airlines; Bailout Has Grants, Loan Guarantees," Washington Post, September 23, 2001.

(3.) Lizette Alvarez and Stephen Labaton, "An Airline Bailout," New York Times, September 22, 2001.

(4.) Dana Milbank, "Bush Campaigns for More Tax Relief; Some in White House Look to Senate to Cut House-Passed Measure," Washington Post, October 25, 2001.

(5.) David E. Sanger, "Bush, on Offense, Says He Will Fight to Keep Tax Cuts," New York Times, January 6, 2002.

(6.) Dana Milbank, "Bush Campaigns for More Tax Relief; Some in White House Look to Senate to Cut House-Passed Measure," Washington Post, October 25, 2001.

(7.) Juliet Eilperin, "Bipartisan Approach Faces Tests; As Congress Tackles Tough Issues, Leadership Is Questioned," Washington Post, September 27, 2001.

(8.) All quotes can be found in Joseph Weisman and Charles Babington, "Iraq War Debate Eclipses All Other Issues; GOP Flounders as Bush's Popularity Falls; Democrats Struggle for a Voice," Washington Post, November 20, 2005. DeMint later clarified that the Democratic Party, rather than the Iraq War, functioned as the figurative "concrete block." See http://www.washingtonpost.com/wp-dyn/content/article/ 2005/12/05/AR2005120501989.html (accessed August 22, 2012).

(9.) Dan Balz and Jim VandeHei, "Lowered Expectations Reflect Political and Fiscal Realities," Washington Post, February 1, 2006.

(10.) Bruce J. Schulman, "A New Frontier; Being President in an Age of Limits," Washington Post, October 2, 2005.

(11.) Though Congress can delay formal approval of the budget (and frequently has done so), it ultimately has always passed an annual budget, even if late.

(12.) In order to extend the data set back to 1933, we also include a handful of the predecessors to agencies in the Kiewiet and McCubbins data set. None of the key findings depend upon their inclusion.

(13.) This characterization comports with other spatial models of the budgetary process, including Canes-Wrone (2006), Kiewiet and McCubbins (1988), and Ferejohn and Krehbiel (1987). In each of these models, the president is assumed to have an ideal appropriation amount for a given unit at a given time period, and the utility he derives from actual appropriations is decreasing in their distance from his ideal point.

(14.) Bafumi et al. (2005), Clinton, Jackman, and Rivers, (2004), Jackman (2001), and Martin and Quinn (2002) discuss this issue in greater detail.

(15.) One could imagine a technically simpler model that did not rely upon expertise but instead ideal point convergence. Consider, then, a theory that assumes exactly the same form as our own, except that there is no uncertainty about the mapping of policies onto either national or local outcomes. In this alternative model, increasing the salience of national outcomes will lead to heightened presidential success, but only if the president and legislator have more ideological comity regarding national outcomes than they do regarding local outcomes. On the other hand, if the president and legislator disagree more about national outcomes than they do about local ones, then increases in the salience of national outcomes necessarily leads to policy enactments that make the president worse off. By focusing on expertise, then, our model altogether avoids these entanglements, generating clean predictions that do not rely upon strong (and perhaps unwarranted) assumptions about ideological convergence.

William G. Howell is the Sydney Stein Professor in American Politics at the University of Chicago, where he holds appointments in the Harris School of Public Policy, the department of political science, and the college.

Saul P. Jackman is a postdoctoral fellow in the department of political science at Vanderbilt University.

Jon C. Rogowski is an assistant professor in the department of political science at Washington University in St. Louis.

TABLE 1 Comparing War and Peace

 Column 1 Column 2 Column 3

War -0.298 ** -0.016 ** -0.425 **
 (0.123) (0.007) (0.188)
House Seat Share -1.863 ** -0.145 ** -3.117 ***
 (0.774) (0.055) (1.096)
Unemployment -0.059 *** -0.004 *** -0.082 ***
 (0.020) (0.001) (0.027)
Real Deficit 0.082 -0.009 0.288 **
 (0.056) (0.010) (0.133)
Real Gross Domestic -2.854 *** -0.156 ** -3.745 **
 Product Growth (1.014) (0.061) (1.428)
in(Proposal) 1.047 *** -0.042 * 0.972 ***
 (0.106) (0.024) (0.151)
(Intercept) -3.266 ** 0.738 ** -5.928 **
 (1.536) (0.346) (2.485)
 3201 3201 3201
R-squared 0.74 0.26 0.31
Mean Squared Error 2.11 0.17 3.84

Entries are linear regression coefficients with standard errors
shown in parentheses. In column 1, the dependent variable is
ln([absolute value of [Prop.sub.it] - [Approp.sub.it]] + 1). In
column 2, the dependent variable is ln(([absolute value of
[Prop.sub.it] - [Approp.sub.it]]/[Prop.sub.it]) + 1). In column
3, the dependent variable is ln([absolute value of [Prop.sub.it]
- [Approp.sub.it] + 1) if [Prop.sub.it] > [Approp.sub.it], and
zero otherwise. Though not reported, all models include president
and agency/program fixed effects. *** indicates p < .01, two-tailed
test; ** p <.05; * p <. 10.