Bargaining Commitments and Executive Reputation: Legislative Response to Unilateral Action

Transcription

1 Bargaining Commitments and Executive Reputation: Legislative Response to Unilateral Action Jon C. Rogowski Harvard University June 16, 2017 Abstract When do executives achieve their legislative goals? Existing scholarship focuses largely on factors such as presidential popularity, the partisan composition of Congress, and exogenous events such as war. In this paper, I focus on the reputations executives develop with legislatures through the use of unilateral action and argue that unilateral action provides a signal to Congress about the president s commitment to legislative bargaining. Using data on proposed and enacted budgets in the United States from 1947 to 2003, I report evidence that unilateral action affects a president s future success in the outcomes of interbranch bargaining. Presidents who issue greater numbers of executive orders experience decreased success in bargaining with legislatures composed of large numbers of the president s partisan opposition, but increased success when bargaining with legislatures with large numbers of the president s copartisans. The results are robust to a number of modeling strategies and characterizations of the key dependent and independent variables. The findings provide evidence of important linkages across varieties of presidential behavior and have important implications for how presidents pursue their policy goals. I thank Alex Bolton, William Howell, Kenneth Lowande, and Sharece Thrower for sharing data used in this paper. I also thank Justin Fox, William Howell, Andrew Reeves, and Sharece Thrower for helpful comments. Assistant Professor, Department of Government, 1737 Cambridge Street, Cambridge, MA 02138; rogowski@fas. harvard.edu.

2 While running for president in 2008, then-senator Obama made no secret of his disdain for President George W. Bush s use of unilateral directives, including executive orders and signing statements. 1 Several years, after frustration had mounted with the Republican Congress s refusal to consider his proposals, the Obama administration unveiled its We Can t Wait initiative in fall 2011 (Lowande and Milkis 2014). The initiative announced a number of policies the Obama presidency would implement unilaterally through administrative rulemaking and executive orders. The shift in Obama s approach to governing led at least one media observer to comment that Obama s increasingly assertive use of executive action could result in pitched battles over the separation of powers were he to win a second term in office. 2 Though partisan and ideological differences are frequently a source of conflict between legislatures and executives, this account suggests that presidents use of power may also shape their relations with Congress. In this paper, I study how the exercise of unilateral powers affects an executive s subsequent success in bargaining with the legislature. Existing scholarship focuses largely on the capacity of Congress to pass legislation that overturns the policies a president has implemented with unilateral directives (e.g., Bolton and Thrower 2016; Howell 2003). I build upon this literature to argue that a president s exercise of unilateral powers affects Congress s willingness to negotiate with him in the future. Unilateral action informs Congress about the president s commitment to partisan and ideological principles or his willingness to accept legislative compromises. As a consequence, presidents who confront hostile congresses can expect to fare more poorly with Congress in the future when issuing unilateral directives, while unilateral action in the context of friendly congresses enhance the president s future bargaining position. My argument thus relates the dual notion of executive power (Belco and Rottinghaus 2017; Mansfield 1989) to the ways partisan legislatures interpret and subsequently respond to unilateral action. 1 Laura Meckler, Obama Shifts View of Executive Power, March 20, 2012, Wall Street Journal; available at https: // (accessed June 15, 2017). 2 Charlie Savage, Shift on Executive Power Lets Obama Bypass Rivals, April 22, 2012; /04/23/us/politics/shift-on-executive-powers-let-obama-bypass-congress.html (accessed June 15, 2017). 1

3 I test the argument outlined above using data on the president s budget requests and subsequent enacted appropriations for 77 agencies from fiscal years 1947 through Consistent with my account, I find consistent evidence that the effect of unilateral action on a president s future bargaining success is strongly conditioned by the partisan composition of Congress. Increased use of unilateral action significantly reduces a president s success in bargaining over the budget when Congress is composed of large numbers of legislators from the opposite party, yet unilateral action increases a president s bargaining successes when the partisan composition of Congress favors the president s party. These results are robust across a range of alternative model specifications, characterizations of key variables, and placebo tests. In addition, I provide evidence that unilateral directives in one policy domain affect a president s bargaining success over agency budgets in altogether different policy domains. The findings in this paper provide evidence to support Richard Neustadt s skepticism about the use of unilateral directives, who argued that [unilateral] action rarely comes at bargain rates. It can be costly to the aims in whose defense it is employed. It can be costly, also, to objectives far afield (Neustadt 1990, 28; emphasis added). Moreover, the results suggest that the downstream consequences of unilateral action for a president s future policy objectives may be an unappreciated component of a president s decision calculus. The Politics of Unilateral Action For presidents, few accomplishments are as important or as elusive as successfully convincing Congress of the merits of their legislative agendas. Securing passage of their key policy goals enables presidents to gain favorable media coverage, enact policies that stand to improve the country s fortunes and thus the president s electoral standing, demonstrate their commitments to key constituencies, and construct their legacies. For the most part, scholars have studied a president s success in bargaining with Congress 2

4 one arena at a time. For instance, an important literature studies a president s ability to influence the content of legislation through veto bargaining (e.g., Cameron 2000), get their policy priorities on the legislative agenda (e.g., Bond and Fleisher 1990; Edwards 1989), and receive Senate confirmation for their appointment to executive branch and judicial positions (e.g., Black et al. 2007; Cameron, Cover and Segl 1990; McCarty and Razaghian 1999; Segal, Timpone and Howard 2000). By and large, these studies focus on how the features of the macro political environment, such as the partisan composition of Congress or presidential approval ratings, or characteristics of the president s initiative or nominee affect a president s success in achieving his goal with the help of Congress. As congressional action has slowed in recent years due to secular trends such as increased polarization across party lines, however, presidents have made greater use of unilateral powers for achieving their policy goals. By issuing executive orders, memoranda, and issuing appointments which do not require Senate consent, presidents can shift policy outcomes on their own without waiting for or first receiving approval from Congress. To do so, existing scholarship highlights the ways in which presidents need only consider the capacity for Congress to respond and overturn or replace a unilateral directive with a new policy closer to Congress s (and more distant from the president s) preferred policy outcome. Congress s willingness, however, to overturn an executive order may not be the only ways in legislatures can respond to the use of unilateral action. Instead, a president s use of unilateral action in one policy area may affect Congress s willingness to support the president in other policy areas. Existing scholarship highlights some specific examples of these dynamics. For instance, Neustadt (1990) attributes John F. Kennedy s reluctance to issue an executive order prohibiting discrimination in public housing to his desire to maintain the support of Southern Democrats for some of his other legislative priorities related to trade and foreign aid. In more recent work, Chiou and Rothenberg (2017, 22) posit that roll call voting patterns in Congress could be affected by executive orders and other unilateral actions which alienate its members. More generally, unilateral 3

5 action, like vetoes and veto threats (Cameron 2000), may help shape the president s reputation among members of Congress, which thus has implications for a president s future success when bargaining with the legislature. Presidential Reputation, Legislative Partisanship, and Interbranch Bargaining I argue that a president s use of unilateral power has consequences for future bargaining with Congress. Generally speaking, presidents do not regard unilateral action as the most-preferred means of accomplishing their policy goals. Virtually every presidential biography emphasizes the president s interest in working with Congress to pass major legislation that advances their key priorities. Frequently, however, particularly in the contemporary era of ever-increasing legislative gridlock, Congress will frustrate a president s legislative agenda or else other exigencies necessitate quick action. Presidents may inevitably contemplate unilateral action as a solution for achieving some measure of policy success, however short-term it may be. Here I informally sketch the intuition that links unilateral action to a president s subsequent legislative success. A president s decision to exercise unilateral powers reveals information about the president to Congress. In particular, presidents signal commitment to their personal, partisan, and ideological principles by issuing unilateral directives. Presidents who are more loathe to do so, however, signal a willingness to compromise those beliefs. These presidential types are roughly analogous to the recalcitrant and accommodating types described in Cameron (2000), in which Congress is relatively uncertain about the president s preferences and thus the circumstances under which the president will sign or veto a given piece of legislation. In the context of unilateral action, Congress is largely uncertain about the president s commitment to achieving what he identified as his key priorities, and thus the use of unilateral action conveys some information about the president s depth of commitment to those issues. 4

6 Consider a president who has exercised unilateral powers. A representative legislator s response is likely to depend on the legislator s partisan alignment with the president. Legislators from the party opposite the president are likely to interpret the president s use of power as an indication that the president is strongly committed to his partisan and ideological goals, and thus signal to the legislator the necessity of driving a harder bargain with the president so the president does not move government policy too far in the president s preferred direction (and opposite the legislator s). On the other hand, unilateral action should embolden a president s copartisan legislators because insofar as a president s goals are aligned with his party s, it indicates the president is willing to use the tools of his office to policy goals that are likely to be mutually beneficial for the president and his copartisan legislators. Altogether, this account suggests that the legislative response to unilateral action is conditioned by the partisan composition of Congress. When Congress is composed of many legislators who share the president s party, unilateral action is likely to benefit the president in subsequent legislative bargaining. But when Congress is composed of many legislators from the opposition party, unilateral action is likely to reduce the president s chances of succeeding in future bargaining situations. Though scholars have paid a great deal of attention to how the president s formal statements, primarily the State of the Union Address, affects a president s success in Congress, the effect of a president s formal powers on legislative success has received somewhat less attention. As noted above, the most important insights from this literature come from studies of vetoes and veto threats (e.g., Cameron 2000; Cameron and McCarty 2004; Conley 2003; Groseclose and Mc- Carty 2001; Hassell and Kernell 2016; Kelley and Marshall 2009; Kiewiet and McCubbins 1988), which finds that vetoes and veto threats can sometimes (but not always) lead Congress to modify legislation in ways more favorable to the president. No study of which I am aware study the effects of unilateral action on subsequent bargaining between presidents and Congress. However, in research more closely related to the present paper, Ainsworth et al. (2014) argue and provide 5

7 evidence that statements of administration policy and signing statements signal the potential for bureaucratic non-compliance with congressional preferences, and subsequently trigger Congress to perform additional oversight investigations of the bureaucracy. In addition, the argument outlined above generates expectations that contrast with suggestions based on accounts in the literature that relate unilateral action to bureaucratic capacity and public opinion. Because unilateral directives are formulated within the executive branch, the increased use of unilateral action may reduce capacity of these offices to expend costly effort to accumulate expertise and engage in outreach efforts with members of Congress, both of which are important resources for aiding the president s success (Rudalevige 2002, 2015). The behavioral literature, meanwhile, indicates that presidents incur costs for the use of unilateral action (Christenson and Kriner Forthcoming; Kriner and Schickler 2016; Reeves and Rogowski 2016, Forthcoming). To the extent unilateral action diminishes a president s public standing, members of Congress may be less inclined to support the president s budgetary proposals. In addition, Madonna, Monogan and Vining (2016) show that controversial nominees who require Senate confirmation hurt a president s legislative agenda by consuming valuable Senate committee and floor time and thereby reducing the opportunity for the chamber to consider the president s policy proposals. While these accounts implicate quite different mechanisms, they both suggest that unilateral action poses a rather fixed cost on the president s ability to secure future bargaining successes regardless of the partisan composition of Congress. In contrast, my argument predicts that the partisan makeup of Congress conditions the legislative response to unilateral action, with presidents realizing increased [decreased] success when Congress is composed of greater [fewer] legislators from the president s party. 6

8 Data and Methods I test the argument outlined above in the context of presidents success in the appropriations process. Since the Budget and Accounting Act of 1921, presidents have submitted complete budgetary proposals to Congress. These proposals initiate the appropriations process in Congress by requesting funding levels not only for agencies and their subunits but also for the specific projects they administer and the personnel who serve. Though Congress is under no obligation to accept the president s proposals in fact, Congress frequently modifies the president s appropriations requests to suit their own spending preferences the president s budget is the most important document used by members of Congress to construct a final budget (Schick 2007). As previous scholarship details (Canes-Wrone, Howell and Lewis 2008; Howell, Jackman and Rogowski 2013; Howell and Jackman 2013; Kiewiet and McCubbins 1985, 1988; McGrath, Rogowski and Ryan Forthcoming), budgets confer several important advantages for studying executive success in bargaining with legislatures. Federal budgets are (at least in normal times) fashioned at regular annual intervals and require presidents to take public positions on their preferred agency and program spending levels in each instance. Budgetary data thus ameliorate concerns about strategic ambiguity or obfuscation that often accompany the normal legislative process. In addition, by comparing the president s proposal budget to Congress s enacted budget, the appropriations process provides a clear and continuous measure of the president s bargaining success. The closer the final budget hews to the president s proposal, the greater the president s success. Crucially for my purposes, this measure of bargaining success permits direct comparisons across agencies, policy domains, and presidents. I use appropriations data from fiscal years 1947 to 2003 to compare the president s budget requests and final appropriations across the 77 agencies studied in Kiewiet and McCubbins (1991). 3 Requests and appropriations are standardized to 1983 dollars. The agencies contained in the data 3 These data are reported in Howell and Jackman (2013). 7

9 span a wide range of policy domains, including the Food and Drug Administration, HUD Housing Programs, Environmental Protection Agency, Patent Office, and the National Park Service. About one-fifth of the observations are related to foreign policy or national defense, including the Administration of Foreign Affairs, the Coast Guard, Defense Procurement, and the Departments of the Air Force, Army, and Navy. Not every agency is represented in each of the 57 years spanned by the data. Some of the agencies were created, or merged with another unit at some point between 1947 and 2003, which accounts for the vast majority of missing data. Less commonly, agency data was simply not available for a given year. In addition, presidents generally do not submit budgets for the first fiscal year of their presidency, but instead budgets are proposed by the outgoing president. We therefore omit these observations. Altogether, the dataset contains 2,598 agency-year observations. Empirical Strategy Following Howell and Jackman (2013), I use the budget data to construct the primary dependent variable, Presidential success. This variable represents the natural log of the absolute value of the difference between the president s budgetary request and final appropriations, ln( Proposed i t - Enacted i t + 1), for each agency i in fiscal year t. This variable has a straightforward interpretation, as increasing values indicate greater differences between a president s budgetary preferences and the final appropriations outcome, while a value of zero indicates that Congress appropriated precisely what the president requested. In interpreting the coefficients for our independent variables, positive values indicate greater divergence between what the president proposed and what was appropriated (decreased presidential success) while negative values indicate that appropriations more closely tracked the president s proposal (increased presidential success). Recognizing that congressional control over the appropriations process may be asymmetric and come more in the form of a constraint rather than an accelerant on executive branch activity, 8

10 I employ a second characterization of the dependent variable. To do so, I distinguish appropriations in which Congress provides smaller amounts than the president requests from those where congressional appropriations exceed the president s proposal. This measure reflects the quantity ln( Proposed i t - Enacted i t + 1) if Proposed i t > Enacted i t, and zero otherwise, and thus treats appropriations that exceed the president s request the same way as appropriations that exactly meet the president s proposal. I measure the president s use of Unilateral action with data from Howell (2005), which describes the number of significant nonceremonial executive orders issued per year. Significant executive orders cannot generally be dismissed as trivial and thus are most likely to attract scrutiny from Congress and the public. This set of unilateral directives, therefore, provides the clearest incentives for a legislative response. Though executive orders are not the only form of unilateral directives issued by presidents, they are often the highest profile and most consistently substantive of them. 4 This variable is lagged by one year such that unilateral action in year t - 1 is used to predict budgetary success in year t. This specification captures the intuition a presidents use of power to achieve their goals has implications for Congress s future willingness to bargain with them. 5 Below I will also discuss results from models with different lag structures to examine the robustness of the findings presented below. From fiscal years 1947 to 2003, the mean annual number of significant executive orders was five and ranged from zero to 18. I characterize the Legislative context based on the percentage of the seats in each chamber held by the president s copartisans. For ease of interpreting the resulting regression coefficients, this variable is centered at zero to indicate a scenario in which the chamber is exactly split between the president s copartisans and legislators from the opposite party. Between 1947 and 2003, the president s party held an average of 52% of House seats, ranging from 33% to 77%. In the Senate, the 4 This may explain why recent presidents have relied increasingly on directives such as memoranda in recent years (see, e.g., Lowande 2014). 5 The fiscal year begins on October 1 of each year; thus, the structure of the models predicts the outcomes of budgetary bargaining that usually happens in the spring and summer of a year (to take effect on October 1) as a function of unilateral actions taken in the previous year. 9

11 president s party held an average of 54% of seats and ranged from 35% to 78%. I estimate separate models focusing on the composition of the House and Senate, though the partisan composition of the chambers is highly correlated. Using these data, and following the specification in Howell and Jackman (2013), I estimate the following model: Presidential success i t = β 0 + α i + δ p + β 1 Unilateral action i t + β 2 Legislative context i t + β 3 (Unilateral action i t Legislative context i t ) +XΩ i t + ɛ i t, where i, t, and p index agencies, years, and presidencies, respectively. Because Legislative context is centered at zero, the coefficient for β 1 characterizes changes in the president s success as he makes increasing use of unilateral action when Congress is evenly divided between parties. The key hypothesis is assessed with the estimate of β 3, which describes the interaction between Unilateral action and Legislative context. If, as the argument presented above expects, a president s success in bargaining with Congress is influenced by his use of unilateral powers, I expect to obtain a negative coefficient estimate for β 3. This result would indicate that presidents are penalized for the use of unilateral action when Congress is composed of greater numbers of legislators from the opposition party. Likewise, it would also indicate that the president succeeds in receiving final appropriations that more closely conform to his budget proposal as his party controls more seats in Congress. The model also includes agency fixed effects (α i ) to control for all observed and unobserved time-invariant attributes that may affect the politics of agency appropriations, and presidency fixed effects (δ p ) to account for president-specific sources of differences in presidential bargaining success. Given this specification, the coefficients of interest are identified with changes in the use of unilateral powers and the partisan composition of Congress that occur within a given presidency. I also include a matrix (X i t ) of other agency and contextual factors that may also 10

12 be related to presidential success. First, I control for the presence of major war, defined as the Korean War (calendar years ), the Vietnam War ( ), the Gulf War (1991), and the Afghanistan War ( ). Second, Second, I control for the size of the president s budget proposal, as Congress may be more acquiescent to smaller proposals than larger ones. Third, I include three economic indicators, as presidents may adjust their proposals and experience varying levels of success with changes in the nation s economic status: the average annual unemployment rate, growth of GDP, and the total budget deficit. And fourth, I include a time trend variable to capture secular trends in the use of unilateral action. Finally, β 0 is a constant term, Ω i t is a vector of coefficients for the control variables, and ɛ i t is a random error term, clustered on agencies. At the outset, I acknowledge the limitations of the chosen empirical strategy for cleanly estimating the effect of unilateral action on a president s subsequent bargaining success. In the ideal experiment, the president s use of unilateral directives would be randomly assigned across Congresses and time and thus allow for unbiased causal estimates of unilateral action. Of course, such an experiment is impossible, but the available data raise potential concerns about endogeneity and omitted confounders that generally accompany observational studies. To help assuage these concerns, below I discuss a variety of robustness checks and placebo tests whose results generally support a causal interpretation of the relationship between unilateral action and subsequent presidential success. Results The results are displayed in Table 1. As the coefficients in the first row indicate, unilateral action appears to have no implications for future bargaining when control of Congress is evenly split between Democrats and Republicans. The coefficients are all very small in magnitude, between 0.01 and 0.04, and none of them are statistically significant at conventional levels. Interestingly, the coefficients in the second row further provide little systematic evidence that the share 11

13 of the legislative seats held by the president s copartisans affect the president s bargaining success when that president has not made use of his unilateral powers. Three of the four coefficients are negative while the fourth is positive; moreover, the standard errors are all considerably larger than the coefficient estimates. Consistent with my argument, the results in the third row provide strong evidence that the effects of unilateral action on a president s subsequent bargaining success are conditioned by the partisan composition of Congress. Each of coefficients is negative and statistically significant. In addition, the magnitudes are fairly similar across chambers for both specifications of the dependent variable. And across both the House and Senate, the relationship is larger in magnitude when accounting for asymmetries in Congress s ability to constraint rather that accelerate spending initiatives within the executive branch. The remaining coefficient estimates are also of substantive interest and generally consistent with prior research. For instance, the state of the economy has important implications for a president s success in bargaining over the budget. Increased budget deficits reduce a president s bargaining success, while higher growth and unemployment rates both increase a president s bargaining success. 6 Similarly, a president s budget fares better in Congress during major wars, while presidents have greater success on smaller budget items than they do for their more expensive proposals. The coefficient estimates for the time trend variable, however, are inconsistently signed and not statistically significant. Table 1 goes here. The findings reported in Table 1 are robust across a number of alternative approaches for measuring the key independent variable. Recall that Unilateral action describes the number of significant executive orders issued in a given year. Significant executive orders are, following Howell (2005), non-ceremonial orders which received front-page coverage in The New York Times. 6 Howell and Jackman (2013) find the same patterns. 12

14 I estimated models that included alternative characterizations of the degree to which presidents exercised unilateral action using other measures in existing scholarship. First, using data from Bolton and Thrower (2016), I included the total of all non-ceremonial executive orders issued per year. 7 Second, Chiou and Rothenberg (2014) develop a continuous measure of executive order significance, which I use to characterize the total significance of the executive orders issued in a given year. Third, as discussed above, executive orders are not the only form of unilateral directive. Using data from Lowande (2014), I substituted the number of significant executive orders with the aggregate number of memoranda issued per year. And fourth, I summed the number of non-ceremonial executive orders and presidential memoranda to create a composite measure of the president s unilateral activity in a given year. The results of these additional models are shown in Table A.1 in the Supplementary Appendix. The patterns are strongly consistent with those shown above. Across each alternative characterization of unilateral action and every model specification, the interaction between Unilateral action and Legislative context is consistently negative. Unsurprisingly, the exact level of statistical significance varies somewhat, given differences in the years of coverage and varying levels of precision in the measures. In general, however, these findings support the robustness of the results in Table 1 to alternative characterizations of a president s use of unilateral action. The findings in Table 1 are also robust to other specifications of the dependent variable. I estimated models in which the dependent variable is measured by the absolute difference between the president s request and enacted appropriations, divided by the size of the president s request. This specification thus characterizes presidential bargaining success as a percentage of the president s proposal. I also estimated models that simply use the absolute dollar amount difference between proposed and enacted appropriations (again, standardized to 1983 dollars). The results are shown in Table A.2 in the Appendix. Both of these alternative specifications again 7 Presidents issued an average of 64 non-ceremonial orders per year, with a range of 31 (calendar year 1989) to 164 (calendar year 1945). 13

15 provide consistent evidence that the effect of unilateral action on subsequent bargaining success is conditioned by the size of the president s party in Congress. Figure 1 illustrates the substantive magnitude of the findings reported in Table 1. The plot shows the predicted differences between the president s proposed budget and enacted appropriations for a representative agency. The average agency budget proposal between fiscal years 1947 and 2003 was approximately $4.3 million (in 1983 dollars), which roughly corresponds to President George W. Bush s budget request for the National Park Service for FY2003. Holding constant this and the other model parameters (and setting the indicator for War equal to zero), the plot shows how the president s bargaining success varies across the range of values for percentage of House seats held the president s party. (Virtually identical results are obtained when focusing on the Senate.) More importantly, the two lines show how the predicted differences between the president s proposal and final appropriations vary with the use of unilateral action. The green line shows the predicted differences when the president issued one significant executive order in the previous year, which occurred in 1974, 1983, 1984, 1992, 1994, and 2000, and roughly corresponds to a value that falls one standard deviation below the mean. The purple line shows the predicted differences between proposed and enacted budgets when the president issued ten significant executive orders, as occurred in 1947 and corresponds to unilateral activity that is one standard deviation above the mean. The shaded regions are the 95% confidence intervals that correspond to the predicted values and the tick marks along the x-axis indicate the annual distribution of values of the president s copartisans in the House during this period. The figure shows that, first, presidential bargaining success is virtually unaffected by the partisan composition of the House when presidents make little use of unilateral powers. The slope of the green line is virtually flat across the entire x-axis. For instance, for the representative agency depicted by the figure, the estimates from the model imply that the difference between what the president requests and what Congress appropriates is about $142,000 when the president s co- 14

16 partisans control 40% of House seats, which represents about a 3.3% difference relative to the size of the president s proposal. Presidents are marginally more successful when their copartisans control 60% of House seats, in which the predicted difference between proposed and enacted appropriations is about $139,000, or about a 3.2% difference relative to the proposal. The differences in the president s success across these two conditions are neither statistically nor substantively meaningful. As the purple line demonstrates, however, presidential bargaining success is strongly conditioned by the partisan composition of Congress when presidents make frequent use of unilateral powers. When the president s party holds 40% of House seats, the difference between what the president proposes and what Congress appropriates is about $236,000, or 5.5% of the size of the president s proposal. But when the president s share of seats increases to 60%, the predicted difference between proposed and enacted appropriations is decreases by more than half to about $96,000, representing about 2.2% of the president s budget proposal. The key quantity represented by Figure 1 is the difference between the purple and the green lines for each value along the x-axis. This difference represents the marginal effect of a twostandard deviation increase in the use of unilateral action as the president s congressional seat share increases. Consistent with the results in Table 1, there is no difference in predicted presidential success when the partisan composition of Congress is split evenly at 50%. However, as the president s seat share declines, a two-standard deviation increase in the use of unilateral powers significantly decreases the president s bargaining success. On the other hand, as the president s seat share increases above 50%, presidents make greater use of unilateral powers experience significantly greater success in bargaining over the budget. Extrapolating these results, given that the annual U.S. budget is in the trillions of dollars, the differences in presidential success depicted in Figure 1 have enormous implications for the nation s financial status. Figure 1 goes here. 15

17 Consistent with my argument, the results in this section provide compelling evidence of an association between a president s use of unilateral powers and his subsequent success in bargaining with Congress. Moreover, the data indicate, this relationship appears to be strongly conditional, with unilateral action significantly increasing a president s budgetary success when confronting a Congress controlled by large numbers of copartisans, but experiencing less success on Capitol Hill when his party holds fewer legislative seats. Identification Issues and Placebo Tests As noted above, the observational data used here raise questions about endogeneity and omitted confounders. In this section I report results from several placebo tests in attempt to at least partially rule out the potential for reverse causality and endogeneity. I emphasize that no single result discussed here should be taken as dispositive; however, my goal is to present results from a range of models which, in totality, help to establish the plausibility of a causal association between unilateral action and subsequent bargaining success. Table A.3 in the Supplementary Appendix shows the results of these additional analyses. First, I estimated the models shown in Table 1 while also including the leading value of Unilateral action and its interaction with Legislative context. As the top panel shows, the inclusion of these additional covariates did not substantively change the estimates of Unilateral action, Legislative context, and its interaction shown in Table 1. Second, I substituted the leading value of Unilateral action for the measure included in the models displayed in Table 1. However, as the second panel of Table A.3 demonstrates, the estimates of β 3 are much smaller in magnitude than those in Table 1, with none of them reaching conventional levels of statistical significance. Together, these two sets of placebo tests weigh against the possibility of endogeneity and reverse causality. If a president s bargaining success (or lack thereof) affected unilateral action, or if both a president s bargaining success and use of unilateral powers were caused by some more general feature of the political environment, the results of these analyses would likely have been more similar to those 16

18 shown in Table 1. I also estimated models that more fully explored the lag structure of the association between unilateral action and bargaining success. The third panel of Table A.3 shows results when estimating models that also include the number of significant executive orders that were issued two years before the budget bargaining outcome and its interaction with Legislative context. As above, the inclusion of these lagged measures do not materially affect the relationship between unilateral action and bargaining success as displayed in the primary models above. In addition, I note that the interaction between Unilateral action, lagged by two years, and Legislative context is small in magnitude (though still negative) and not statistically significant. This result suggests that Congress s memory when it comes to unilateral action may be relatively short, responding more to recent exercises of unilateral power rather than the accumulation of such actions within a given presidency. Finally, I substituted the contemporaneous number of unilateral actions and estimated its relationship with Legislative context, the results of which are shown in the bottom panel of Table A.3. The estimates for the interaction terms continue to be negative, though they are smaller in magnitude relative to the models in Table 1 that lag this measure by one year. This is quite possibly due to the fact that some number of these unilateral directives were issued after the budget had been passed. Nevertheless, these results provide further evidence in support of the lag structure estimated in the primary models shown above. Unilateral Action, Interbranch Bargaining, and the Water s Edge I have argued that unilateral action provides information to the legislature about the president s type and is suggestive of his commitment to the legislative bargaining process as a means for accomplishing his goals. The account I have offered largely sidesteps the question, however, about whether a president s bargaining success in one policy domain would reflect the exercise of unilateral powers to shape policy in an altogether unrelated domain. If interbranch bargaining 17

19 occurs one-policy-at-a-time, we may not expect a president s use of unilateral power in one area to affect bargaining outcomes in another. If, however, unilateral power conveys more general information about the president s reputation, it may be more likely that a president s actions in one affect legislative outcomes in another. Nonetheless, distinguishing whether presidents need anticipate (or fear) congressional reaction to unilateral action within or across policy domains is important for characterizing the more general set of incentives and constraints associated with unilateral power. To address this question, I examine whether a president s use of unilateral powers in the foreign policy arena affects his bargaining success in agencies chiefly concerned with domestic policy. This case provides a difficult test for evaluating how unilateral action affects bargaining success across domains. Previous research details a number of ways in which presidents are more successful realizing their goals in foreign policy than they are in domestic policy (e.g., Canes-Wrone, Howell and Lewis 2008), in part because Congress may grant the president greater deference due to informational advantages that accompany the president s status as commanderin-chief and because he is well-positioned to receive intelligence and other classified information that may affect his decisionmaking (Howell, Jackman and Rogowski 2013). The findings from this research suggest that foreign policy unilateralism poses a hard case for detecting its effect on subsequent bargaining outcomes. From the set of significant executive orders issued between 1945 and 2001, I identified the directives that clearly involved foreign policy. While the bulk of these directives concerned defense and security, a number of them concerned other foreign policy issues including trade and immigration. Altogether, fully half (49.7%) of the significant executive orders issued during this time period concerned foreign affairs. Using these data on foreign policy executive orders, I estimate the regression model shown above for only those agencies that concern domestic policy. About three-quarters (57) of the 77 agencies in the budget data are primarily concerned with domestic policy. 18

20 The results are shown below in Table 2 and provide clear evidence that a president s use of unilateral action to set foreign policy have implications for his bargaining success for agencies concerned with domestic policy. Consistent with the findings in Table 1, the coefficient estimates of β 3 are all strongly negative and statistically significant. Table 2 goes here. The results from Table 2 confirm and extend the results from the main set of analyses. How presidents wield power has implications for subsequent success in achieving their legislative goals, with the nature of these implications dependent upon the presence of the president s partisan allies in Congress. Somewhat more speculatively, the findings in Table 2 suggest that presidents and Congress do not bargain policy area by policy area; rather, a president s actions in one policy domain have implications for the legislative response in another policy domain. Furthermore, and consistent with other research on the domestic politics of foreign policymaking (Howell and Pevehouse 2007; Kriner 2010; Milner and Tingley 2015), Congress does not grant the president a blank check when it comes to setting the country s policies in national security, defense, and trade. Rather, Congress s response to unilateral action in the area of foreign policy appears to be quite similar to its response to unilateral action on purely domestic policy domains. Conclusion Unilateral action is one of the defining characteristics of the modern presidency (Moe and Howell 1999). To date, scholarship on the politics of unilateral action has focused closely on the institutional incentives and constraints that influence a president s decision to create new policies by striking out on his own. This scholarship pays particular attention to the capacity of Congress to intervene and undo the policies created unilaterally by presidents. But persistently high (and increasing levels of gridlock) and the overwhelming absence of legislation that reverses 19

21 presidential unilateral actions have led scholars to question why unilateral action is not more common than it is (Christenson and Kriner 2015). This paper provides a novel yet partial answer. The potential legislative constraints on unilateral action by presidents do not end once congressional action (or inaction) sustain a unilateral directive. According to my argument, Congress considers how presidents wield power. The exercise of unilateral powers provides information about the president s commitment to his partisan and ideological policy goals and, by extension, his commitment to achieving them through the legislative process. When confronted with a legislature filled with his partisan opponents, presidents can expect their use of unilateral action to reduce their chances for future legislative success. On the other hand, when the legislature is filled with his partisan allies, unilateral action improves the prospects for further success. The legislative response to the president may therefore structure the incentives for presidents to exercise unilateral power. To the extent presidents value future legislative success at the same level they value immediate policy victories achieved through unilateral directives, presidents under divided government may be deterred from exercising unilateral powers. Given the prevelance of divided government in the contemporary United States, the account sketched in this paper may help explain the relative paucity of unilateral directives. I close by noting some limitations of the current paper and opportunities for further research. First, though I have outlined a set of factors that may influence how unilateral actions from presidents are viewed by Congress, further theoretical scholarship is needed to more rigorously define the incentives for legislators to respond to unilateral action. In addition, elaborating on comments offered above, further theoretical scholarship could also be useful to better characterize how the president s expectations about the future affect his calculations at the present. Existing theoretical research on unilateral power assumes perfect information from both presidents and Congress (Howell 2003). Just as the introduction of incomplete information has enriched theories of interbranch bargaining in the context of vetoes (Cameron 2000), similar refinements to scholarship 20

22 on unilateral action could bear substantial fruit. More generally, research that takes seriously both the institutional and behavioral incentives for presidents and legislators can make important contributions to unilateral power and interbranch bargaining in the U.S. and presidential systems around the world, particularly Latin America. Finally, though the empirical research are broadly consistent with the argument I presented, future scholarship would do well to consider opportunities for leveraging more exogeneous sources of variation in helping to identify the effects of unilateral action on future interbranch bargaining. 21

23 References Ainsworth, Scott, Brian Harward, Ken Moffett and Laurie Rice Congressional Response to Statements of Administration Policy and Presidential Signing Statements. Congress & the Presidency 41: Belco, Michelle and Brandon Rottinghaus The Dual Executive: Unilateral Orders in a Separated and Shared Power System. Stanford: Stanford University Press. Black, Ryan C., Anthony J. Madonna, Ryan J. Owens and Michael S. Lynch Adding Recess Appointments to the President s Tool Chest of Unilateral Powers. Political Research Quarterly 60: Bolton, Alexander and Sharece Thrower Legislative Capacity and Executive Unilateralism. American Journal of Political Science 60: Bond, Jon and Richard Fleisher The President in the Legislative Arena. Chicago: University of ChicagoPress. Cameron, Charles M Veto Bargaining: Presidents and the Politics of Negative Power. New York: Cambridge. Cameron, Charles M., Albert D. Cover and Jeffrey A. Segl Senate Voting on Supreme Court Nominees: A Neoinstitutional Model. American Political Science Review 84: Cameron, Charles M. and Nolan M. McCarty Models of Vetoes and Veto Bargaining. Annual Review of Political Science 7: Canes-Wrone, Brandice, William G. Howell and David E. Lewis Toward a Broader Understanding of Presidential Power: A Reevaluation of the Two Presidencies Thesis. Journal of Politics 70:

24 Chiou, Fang-Yi and Lawrence S. Rothenberg The Elusive Search for Presidential Power. American Journal of Political Science 58: Chiou, Fang-Yi and Lawrence S. Rothenberg The Enigma of Presidential Power: Parties, Policies and Strategic Uses of Unilateral Action. New York: Cambridge University Press. Christenson, Dino P. and Douglas Kriner. Forthcoming. Mobilizing the Public against the President: Congress and the Political Costs of Unilateral Action.. Christenson, Dino P. and Douglas L. Kriner Political Constraints on Presidential Unilateral Action. Case Western Reserve Law Review 65: Conley, Richard S George Bush and the 102d Congress: The Impact of Public and Private Veto Threats on Policy Outcomes. 33: Edwards, George C. III At the Margins: Presidential Leadership of Congress. New Haven: Yale University Press. Groseclose, Tim and Nolan McCarty The Politics of Blame: Bargaining before an Audience. American Journal of Political Science 45: Hassell, Hans J. G. and Samuel Kernell Veto Rhetoric and Legislative Riders. American Journal of Political Science 60: Howell, William G Power without Persuasion: The Politics of Direct Presidential Action. Princeton: Princeton University Press. Howell, William G Unilateral Powers: A Brief Overview. Presidential Studies Quarterly 35: Howell, William G. and Jon C. Pevehouse While Dangers Gather: Congressional Checks on Presidential War Powers. Princeton: Princeton University Press. 23