Congressional Response to Presidential Signing Statements

Abstract

Popular and scholarly accounts argue that signing statements are important tools for presidents to shape the implementation of policy. Although signing statements might be important presidential tools, the legislative branch stands in the most immediate and direct competition with the executive for ultimate control of the bureaucracy. In this article, we assess whether congressional committees react to presidential signing statements with increased oversight. Using a data set that includes every oversight hearing held by the U.S. House between 1995 and 2007, we find evidence that congressional committees are sensitive to the number of objections raised by presidents in signing statements. As the president uses signing statements to object to a larger number of provisions in laws, the affected House committees respond with more oversight.

Keywords

congressional oversight signing statements time series cross-sectional zero-inflated Poisson

Introduction

Scholars have devoted considerable attention to presidents’ use of signing statements (e.g., Bradley & Posner, 2006; Cross, 1988; Cooper, 2002, 2005; Fisher, 2007; Garber & Wimmer, 1987; Kelley, 2006, 2007; Kelley & Marshall, 2008). Although many presidents have issued signing statements, President George W. Bush’s use of signing statements generated considerable discussion and controversy in both popular and scholarly accounts (e.g., Cooper, 2005; Pfiffner, 2009; Savage, 2006; Sonnett, 2006; Specter, 2009; Tribe, 2006). The American Bar Association even questioned whether Bush’s extensive use of signing statements was an abuse of power (Savage, 2009; Weisman, 2009). Pfiffner (2009, p. 254) argued that the “Bush administration used them with the clear purpose of expanding executive power at the expense of Congress and the courts . . . to accomplish goals it could not achieve through the legislative process.” Although as a candidate, Barack Obama spoke against the abuse of signing statements, as president he too has used signing statements in a fashion that frustrated Congress (Savage, 2009; Weisman, 2009). One Democratic House member stated “It’s outrageous. It’s exactly what the Bush people did” (Weisman, 2009, p. A6).

The scholarly literature suggests that signing statements are powerful tools at the president’s disposal as the president competes with Congress for control over policy implementation. To date, however, no one has assessed congressional reactions to signing statements. We argue that the effectiveness of presidential signing statements as instruments of presidential control over policy depends on the actions of the other branches in government. If signing statements are a means by which presidents can direct bureaucratic implementation, then congressional committees can use the directives in signing statements to guide their own oversight activities.

In the following sections, we explore some of the relevant scholarly literature, including work on signing statements, congressional oversight, and presidential control of the bureaucracy. Our main hypothesis is that presidential signing statements trigger congressional oversight. To test our hypothesis, we developed a data set that includes the oversight activity by month for all House standing committees from 1995 through 2007. We find that congressional committees respond to presidential signing statements with heightened oversight activity in the months following a signing statement as the number of objections in signing statements to specific provisions of law increases.

Signing Statements and Policy Implementation

A signing statement is a written presidential pronouncement regarding a piece of legislation that is issued at the time the president signs legislation into law. In addition to offering presidents an occasion to thank Congress for passing the bill or to criticize Congress for being penurious (or excessively generous), signing statements provide opportunities for presidents to articulate their view of specific provisions of the bill. The specific comments might raise, among other issues, policy concerns, exhortations for future legislative activity, guidelines for the implementation of the law by bureaucratic agents, or constitutional objections to provisions within the legislation.

Recent literature suggests that signing statements enhance presidential powers, especially in the executive-congressional competition to control bureaucratic agencies. For example, Sala argues that signing statements afford presidents “the last chance to frame legislative intent for bureaucratic action” (1998, p. 261). Cooper writes that signing statements were used to “reposition and strengthen the powers of the president relative to Congress” (2005, p. 515). “[T]he administrators [have] their marching orders, regardless of the language of the statute . . .” and they “must respond or face the wrath of the White House” (Cooper, 2002, p. 212). Pfiffner (2009, p. 251) extends the theme: “signing statements . . . provide authoritative guidance to . . . subordinates in the executive branch as to how they should carry out and execute the law.” In the context of the highly salient 2005 Detainee Treatment Act, Kelley and Marshall (2008) note that “. . . where the president could not get Congress to budge, he used the signing statement to turn a policy loss into a win” (p. 2) by directing agency implementation without regard to congressional preferences. In addition to privileging the president in policy disputes with Congress, Cooper (2005) argues that the use of the signing statement influences judicial actors as well. Thus, in the canonical view of the literature, the signing statement is an effective device available to presidents to rein in bureaucratic drift, move policy closer to presidential preferences, influence future judicial decisions, and ultimately assert presidential power in the coordinate construction of the constitution (cf. Halstead, 2008).

Congress might not be able to limit a president’s use of signing statements. For instance, Pfiffner (2009, p. 253 and p. 254) states that “there appears to be no easy way for Congress to compel . . . [the president] to stop.” Indeed, “there is not much the other branches can do if he [the president] is not acting in good faith” and “the only remedy seems to be self-restraint on the part of the executive” (Pfiffner, 2009, p. 254). Concern about the aggrandizement of executive power has generated numerous responses in the popular press and among social scientists (e.g., Pfiffner, 2009; Savage, 2006, 2008; Tribe, 2006). In addition, many members of Congress, like Senators Leahy (D-VT), Arlen Specter (then R-PA, now former), and Representatives Barney Frank (D-MA) and David Obey (D-WI, former) have voiced their frustrations with presidential signing statements (Savage, 2009; U.S. Congress, 2006, S8189-8190).

Given the claims in the extant literature, the cumulative effect of the hundreds of signing statements issued can be reasonably understood to expand the scope of executive power as long as other political actors accept the declarations within signing statements as beyond the reach of institutional checks and balances. Even in scenarios ostensibly so favorable to the president, agency responsiveness to presidential preference may be limited by the “watchful eye” of the congressional committees charged with executive branch oversight (Aberbach, 1990; Halstead, 2008). Indeed, as congressional preferences diverge from those of the implementing bureaucrats, we might expect Congress to “place tighter constraints on the use of delegated authority” (e.g., Epstein & O’Halloran, 1996, p. 374). No one, however, has assessed congressional sensitivity and responsiveness to signing statements.

Legislative–Executive Interaction and Policy Implementation

Many scholars explore the tension between congressional and presidential control of agency decision making. The presidential control literature suggests that in conflicts between the executive and Congress over issues of institutional design, presidents are favored in the “politics of structure” (Howell, 2003; Moe, 1989, 1993; Moe & Wilson, 1994). The congressional dominance literature, however, emphasizes congressional effectiveness in achieving desired policy results from the exercise of oversight, the design of administrative arrangements, and congressional budgetary authority (e.g., Epstein & O’Halloran, 1999). Some of the most recent scholarship in the area finds that control mechanisms attached to one branch can still yield a positive sum benefit for both branches. In his model of the White House’s Office of Information and Regulatory Affairs (OIRA) reporting and review requirements, Wiseman (2009) finds that there are instances in which both the White House and Congress benefit from the OIRA review.

In the competition between the legislative branch and the executive branch for control over policy, Moe (1993) and Howell (2003) argue that the president retains important residual decision rights that reflect the centralized control and unilateral action that presidents enjoy. Similarly, Huber and Shipan (2002) consider the president to be a privileged actor with respect to Congress, while Lewis (2003) and Howell (2003) note the advantages presidents have over policy outcomes because of their ability to shape the design and guide the actions of agencies.

Though many scholars emphasize presidential influence over bureaucratic implementation, others highlight the multiple means by which Congress can attempt to rein in bureaucratic drift. Senate confirmation of presidential nominees, appropriations, and reauthorizations can serve to limit bureaucratic discretion in policy implementation (Oleszek, 1995). In addition, McCubbins, Noll, and Weingast (1989) suggest that ex ante controls may secure bureaucratic compliance with congressional preference. Procedural requirements and statutory limitations within the 1935 Federal Register Act, the 1946 Administrative Procedures Act, notice and comment requirements, the 1966 Freedom of Information Act, and numerous amendments tied to these acts as well as reporting requirements increase the costs of noncompliant behavior on the part of the agency. By insulating policy, the enacting coalition of legislators is better able to ensure bureaucratic compliance with current legislative preferences (Bawn, 1995; MacDonald, 2007).

In addition to ex ante oversight, McCubbins and Schwartz (1984) argue that ex post oversight activities may serve to limit bureaucratic drift. They argue that police patrols and fire alarms permit Congress to restrict the scope of agency divergence. Although police patrols entail costly monitoring, the fire alarm approach to oversight shifts the monitoring costs from legislators to those individuals and groups most directly affected by agency actions. Presidential signing statements and congressional oversight are both connected to the political control of bureaucracies. We propose that congressional sensitivity to presidential signing statements leads an oversight committee to regard the particular objections noted by the president as an invitation for an agency to disregard congressional preferences. As Louis Fisher notes, “[m]embers of Congress and their staffs . . . know the content of signing statements. They are regularly included in the Weekly Compilation of Presidential Documents and are posted on the White House website” (2007, p. 209). Once a constitutional signing statement is published, the committee may choose to hold hearings to impose costs preemptively on bureaucratic noncompliance.

Of course, presidential preferences are also conveyed to Congress by Statements of Administration Policy, veto threats, and public pronouncements. However, the issuance of a signing statement remains the most proximate to congressional reaction and the most directly responsive to what Congress actually passed—rather than what was being deliberated. Therefore, the signing statement can convey new information, not found in other pronouncements.¹ In addition, the president might hope to influence future agency behavior and judicial decisions, as the signing statements are recorded by Westlaw as part of the (post) legislative history (see, for example, Kelley, 2007).

In addition to ex post oversight activities, Congress can limit bureaucratic compliance with presidential signing statements by introducing new statutory language to respond to the president’s intentions as expressed in those statements. For example, when Congress passed the Postal Accountability and Enhancement Act (Public Law 109-435), President Bush (2006) issued a signing statement in which he signaled that

[t]he executive branch shall construe subsection 404(c) of title 39, as enacted by subsection 1010(e) of the Act, which provides for opening of an item of a class of mail otherwise sealed against inspection, in a manner consistent, to the maximum extent permissible, with the need to conduct searches in exigent circumstances, such as to protect human life and safety against hazardous materials, and the need for physical searches specifically authorized by law for foreign intelligence collection.

Three weeks later, Sen. Olympia Snowe (R-ME) and six cosponsors responded to the president by introducing S. Res. 22, which proposed to reject “any interpretation of the President’s signing statement on the Postal Accountability and Enhancement Act that in any way diminishes the privacy protections accorded sealed domestic mail under the Constitution and Federal laws and regulations” (Library of Congress, 2009).

Although we do not explore the statutory response by Congress to presidential signing statements, responses like Snowe’s reflect a heightened sensitivity on the part of Congress to presidential attempts to move policy implementation closer to presidential preference. Congress is not indifferent to unilateral presidential actions. For example, Kelley (2007) documents an occasion of congressional sensitivity to a signing statement (and subsequent court action) issued by Reagan as he signed the Deficit Reduction Act of 1984. After an unfavorable district court ruling regarding the Administration’s constitutional objection to a provision of the act, the Administration refused to comply with the decision (Kelley, 2007, p. 297). As a result, the House Government Operations Committee compelled Attorney General Meese “to testify and explain why the administration would not enforce the decision . . . and voted to withhold funding for the Justice Department and OMB for the following two years unless the Administration relented” (p. 298).

Given the difficulty of crafting new statutes to explicitly limit bureaucratic drift, Congress is much more likely to exercise ex post oversight subsequent to the issuance of a signing statement that runs counter to congressional preference. Congressional oversight does not automatically negate presidential influence, but it does indicate congressional sensitivity to presidential declarations. The president may ultimately prevail in various contests to control bureaucratic implementation, but congressional interest is piqued. We do not argue that presidents have no influence on bureaucratic implementation. Rather, we suggest that through the issuance of signing statements Congress is alerted to areas of likely noncompliance by bureaucratic agents, and Congress can exercise ex post oversight to counteract drift away from its preferences. How constrained bureaucrats are by congressional sensitivity to unilateral presidential declarations is in part a function of the amount of oversight activity.

Hypotheses

Most contemporary scholarship suggests that the use of the signing statement privileges the president in bureaucratic policymaking. Our thesis is that Congress can choose to respond to signing statements with additional oversight, regardless of the extent to which signing statements might enhance executive control of bureaucracies. Congressional oversight might be spurred by any number of partisan, ideological, and institutional factors (Aberbach, 1990; Ogul, 1976). For instance, ideological and partisan differences between House committees and the executive branch may lead to more oversight. In short, signing statements are not the sole engines for oversight. The question—after controlling for a wide range of factors that might lead to more congressional oversight—is whether signing statements still play a role.

With the issuance of a signing statement the president has revealed preferences related to the legislation as actually passed and “shown his hand” to Congress.² Information and expertise asymmetries have long been linked to the abilities and incentives of legislators to engage in effective oversight (e.g., Ogul, 1976; Rourke, 1960). Legislators also remain sensitive to monitoring and transaction costs as well as credit claiming opportunities (e.g., McCubbins & Schwartz, 1984). As Congress comes to a collective judgment about how to spend resources to constrain bureaucratic drift away from congressional preference, signing statements provide important clues about executive branch implementation efforts. As Halstead (2007, p. 24) suggests, signing statements may “alert Congress to the universe of provisions that are held in disregard by the Executive branch, in turn affording Congress the opportunity not only to engage in systematic monitoring and oversight . . .” but also to “assert its prerogatives to counteract the broad claims of authority . . .” Lee (2008, p. 741) argues that “Congress cannot pursue effective oversight if it cannot differentiate between rhetorical claims of power and genuine constitutional reservations and objections.” Consequently, the more objections the president raises within a particular signing statement, the more clear the signal to Congress of potential executive noncompliance. In light of our thesis and the relevant literature on congressional-executive interactions and presidential signing statements, we anticipate that House committees engage in more frequent oversight activity as the number of objections contained in signing statements increases.

Data and Method

Our dependent variable is a measure of all House committee oversight activity, by committee and month, from 1995 through 2007. Using Lexis-Nexis Congressional, we identified all oversight hearings held by every committee in the House during each month of every year from 1995 through 2007. Any number of issues, including secular trends, might affect the underlying levels of committee oversight. The question is whether signing statements demonstrably alter the number of committee oversight hearings. Therefore, we aggregated the data by month, year, and committee.³ In this way, we created a time series, cross-sectional data set with 2,686 observations. All data are summarized in the appendix.

Our times series, cross-sectional model is

Y_{it} = β^{'} X_{it} + α_{i} + u_{it}

where the i index refers to the committee and t refers to the month. The α_i term indicates the intercept, which allows for unit effects. The X_it vector of independent variables may contain lagged values of Y and X. Of course, u_it represents the error term.

The total number of congressional oversight hearings during the period of study was 2,347, slightly less than one hearing per month by committee. However, this average number of hearings held by committee per month is deceptive because no oversight hearings were held in approximately two thirds of the observations. The large number of zero counts dovetails nicely with the longstanding impression that Congress rarely engages in active oversight (Ogul, 1976; Ogul & Rockman, 1990). Given the large number of zeros in the data, we ran a Vuong (1989) test, which strongly suggested a zero-inflated model. A zero-inflated Poisson (ZIP) model to test the hypotheses is indicated because the mean and variance of our dependent variable are close in value.⁴

ZIP models include two estimations. One accounts for the zeros and acts as a selection model for the second. In the first estimation, the zeroes are typically predicted by structural limitations that severely constrain positive counts. For our zero-inflated model, the independent variable is the number of days Congress is in session (by month) since congressional committees cannot engage in oversight activity when the chamber is not in session that month. The second estimation accounts for the positive counts. For the second, count estimation, the value of the dependent variable is conditional on the opportunity for oversight. As before, the vector of independent variables may include lagged values of X as well as lagged values of the dependent variable. The independent variables include data reflecting the signing statements themselves, divided government, the ideological preferences of Congress and the president, and other institutional factors. We include fixed effects (i.e., committee dummies) to control for varying levels of oversight across committees.

The full, zero-inflated Poisson specification of our model is

P (y_{it} = 0 | X_{it}) = ψ_{it} + (1 - ψ_{it}) e^{(- μ it)}

and

P (y_{it} | X_{it}) = (1 - ψ_{it}) [e^{(- μ i t)} μ_{i t}^{y i t} / y_{it}!] for y_{it} > 0 .

The µ_it substitution for β′X_it is standard in Poisson models. The ψ terms represent the probability estimate of zero oversight hearings.⁵ The last equation neatly shows how the probability of a positive outcome is tempered by ψ. Greene (2000) considers the last two equations as representing two different data generating processes or regimes. In other words,

Pr [y_{it} 0] = Pr [Regime 1] + Pr [y_{it} = 0 | Regime 2] Pr [Regime 2]

and

Pr [y_{it} = j] = Pr [y_{it} = j | Regime 2] \times Pr [Regime 2], j = 1, 2, \dots .

Signing Statements

To investigate the effects of signing statements on oversight activity, we identified all bills from 1995 through 2007 on which the president placed a constitutional signing statement (American Presidency Project, various years). We do not consider rhetorical signing statements, as those typically reflect presidential interest in “thank[ing] particular members of Congress for help on a bill, or . . . scorn[ing] Congress for sloppy or irresponsible work” (Kelley, 2007, p. 285). We have no reason to expect congressional oversight to be triggered by a rhetorical signing statement or a rhetorical objection in a signing statement. After reading every signing statement, we tied the signing statements to their relevant public laws. Each law was connected to its relevant committee, and committee data were gathered from the Legislative Information web site (2009).⁶

To assess whether the contents of signing statements might trigger committee oversight, we read every signing statement and recorded the number of objections included in the signing statements on bills from committees by month and year. Objections could include policy objections, expressions of dismay at the scope and cost of congressional action, or constitutional concerns. There were 387 signing statements issued from 1995 to 2007—257 by President Clinton, and 130 by President Bush.⁷ Signing statements averaged 8.42 objections (4.51 for Clinton; 11.42 for Bush).

For the primary independent variables (number of objections), 1-, 2-, 3-, 4-, and 5-month lags are used. Lags are required because we anticipate the effect of signing statements on oversight activity to occur in the few months immediately after the signing statement is placed on a bill. In well-behaved time series, the impact of lagged independent variables should dissipate over time, so we expect naturally dissipating impact from the signing statements as the months pass.⁸

Institutional and Ideological Controls

The ideological relationship between the president and a House committee might affect oversight activity as the distance between the president and the committee’s median member increases (Huber, Shipan, & Pfahler, 2001). To test this possibility, we employed common space NOMINATE scores (Poole & Rosenthal, 1997) to measure policy preferences. From here, the absolute value of the difference between the president and each committee’s median member was computed.

A wide array of scholars considers the effects of divided government on legislative behaviors, including oversight (e.g., Aberbach, 1990; Mayhew, 2005). If Congress and the president share preferences, we expect Congress to be more likely to permit the executive broad discretionary authority in the implementation of statutes. Under divided government, however, Congress would likely be less inclined to cede that control (Ainsworth & Harward, 2009; Epstein & O’Halloran, 1999).

In addition, we test whether counterterrorism policy and the Iraq war affected oversight activity, since a large portion of the controversy over the use of signing statements centered on the 2005 Detainee Treatment Act (Kelley & Marshall, 2008; Savage, 2006). To investigate these effects, we constructed a dummy variable for those committees in the post 9-11 period (i.e., Armed Services, Foreign Affairs, and Homeland Security) that considered the bulk of the measures related to counterterrorism policy and the wars in Afghanistan and Iraq.

Three institutional controls are employed in the analysis. First, we controlled for the number of days the House was in session per month because the opportunities for oversight are diminished as the number of days in session per month decreases. Second, we controlled for the average number of pieces of legislation that each committee considered during the previous 3 months. Legislatively active committees have a larger number of opportunities to have signing statements placed on bills because these committees consider more bills. Committees that consider a larger number of bills might suffer from and react to more signing statements with more oversight. Alternatively, legislatively active committees might have little time left to conduct oversight regardless of the actions of the president. Third, we included a 1-month lag of our dependent variable to control for ongoing oversight activity and account for trends in which committees might conduct their oversight activity. Lagged dependent variables sometimes create small sample problems and often there is a downward bias in explanatory variables (Keele & Kelly, 2006). By most standards, our data set would not be considered small and downward biases work against our claims. Although trade-offs exist, Beck and Katz (1995) recommend the use of lagged dependent variables in time series, cross-sectional models. By capturing any remaining serial correlations or dynamic causality, lagged dependent variables insure the error terms are “well behaved.”

Finally, oversight hearings related to the same bill are unlikely to be fully independent from one another, so analyses are clustered by bill length. Time series, cross-sectional data often have nonindependent observations within clusters and independent observations across clusters. The clustering and fixed effects in the models address bill specific and committee specific concerns, and the clustering option ensures robust standard errors.

Results

Table 1 displays the results of the regression models. The models vary based on which measures of executive–legislative differences are used. Model 1 includes divided government while Model 2 replaces divided government with president−committee distance. Since the direct interpretation of coefficients in maximum likelihood estimation (MLE) models is tricky, Table 1 includes the predicted discrete changes in the dependent variable as changes in an independent variable are made. To compute these changes, an independent variable is moved from its minimum to its maximum while holding all other independent variables at their means.⁹

Table 1.

Signing Statements and Congressional Oversight Activity, 1995-2007

Independent variable	Model 1	Predicted change (Model 1)	Model 2	Predicted change (Model 2)
Signing statements
Objections_t−1	−.004 (.004)	−.18	−.005 (.004)	−.21
Objections_t−2	.009***(.001)	.84	.008*** (.001)	.61
Objections_t−3	.014*** (.001)	1.64	.012*** (.001)	1.21
Objections_t−4	.012*** (.001)	1.25	.011*** (.001)	1.09
Objections_t−5	.007** (.003)	.55	.005* (.003)	.39
Institutional & ideological factors
Divided government	.303*** (.013)	.14	—	—
President-committee median distance	—	—	.953*** (.024)	.47
Days in session	.074*** (.002)	1.28	.070*** (.002)	1.22
Committee activity	−.067*** (.017)	−.44	−.077*** (.015)	−.47
Iraq/counterterror committee	−.023 (.021)	−.01	−.132*** (.021)	−.06
Lagged oversight activity	.081*** (.004)	1.79	.073*** (.005)	1.44
Bill Length (millions of words)	−.679** (.221)	−.46	−.697*** (.218)	−.46
Constant	−2.888*** (.034)		−3.321*** (.034)
Zero inflate
Days in session	−.143*** (.006)		−.150*** (.007)
Constant	.946*** (.071)		.996*** (.071)
N	2,598		2,598
Log likelihood	−2,782.31		−2,766.31
Chi-square	638,975.90		697,551.29
Prob>chi-square	<.0001		<.0001

Note: The values in parenthesis are robust standard errors, clustered by bill length. Second, we include, but do not report dummies for each committee. Finally, the * denotes p < .05, ** denotes p < .01 and *** denotes p < .001, one-tailed tests.

The results across both models indicate that presidential signing statements affect oversight activity as the number of objections achieves statistical significance and has positive overall effects. Although the 1-month lags for objections are insignificant in Models 1 and 2, all other lags have strongly positive coefficients. The substantive impact of the signing statement variables is greatest at the 3-month and 4-month lags. For instance, the impact of the 3-month lag for the number of objections is 1.64 in Model 1 and 1.21 in Model 2. Given the paucity of oversight hearings, these are large substantive effects that rival the lagged dependent variable in substantive impact. By the 5th month, there is a clear dissipation of impact. As one would expect in a well-behaved time series, the independent variables lose impact as their lags increase.

Institutional Controls and Oversight Activity

The institutional controls play an important role as most were statistically significant. More specifically, divided government results in approximately .14 more days of oversight hearings. Furthermore, changing the distance between the president and the substantive committee from its minimum to its maximum value yields approximately .47 additional days of oversight hearings.

As the days in session variable moves from its minimum to its maximum value, between 1.22 and 1.28 additional days of oversight hearings are predicted. The results also indicate that legislatively active committees are less likely to engage in oversight activity. These results suggest that it is hard for committees to engage in both more legislative activity and more oversight activity. The results for bill length indicate that committees are less likely to oversee longer legislation. Longer legislation, which is apt to touch on many legislators’ interests, most likely receives greater scrutiny during the legislative process as a winning coalition is created. Not surprisingly, the lagged dependent variable is positively associated with additional oversight. The models predict between 1.44 and 1.79 additional days of oversight in the current month as the value of the dependent variable in the previous month moves from its minimum to the maximum.

The committees whose jurisdictions cover bills that deal with counterterrorism and the wars in Afghanistan and Iraq were less likely to engage in oversight activity subsequent to the September 2001 attacks. Although statistically significant in one of the two models, the substantive impact is dwarfed by the effects of other variables. Counterterrorism and the wars in Afghanistan and Iraq played a minor role with respect to oversight activity in these committees, never decreasing oversight by more than a twentieth of a day (Model 2).

As seen in Table 1, time series, cross-sectional models require large tables to present full results. The results from Table 1 are summarized in Table 2. Of course, some information is lost because Table 2 summarizes the results across both models.

Table 2

Summary of Model Results

Independent variable	Direction of statistical significance (p < .05 one-tailed or better)	Substantive significance
Number of objections in signing statements	Varies by lag, overall +	Varies by lag, overall strongly +
Divided government	+	Marginally +
President−committee median distance	+	Moderately +
Days in session	+	Strongly +
Committee activity	−	Moderately −
Iraq/counterterror committee	−	−
Lagged oversight	+	Strongly +
Bill length	−	Moderately −

That said, the results across the two models are quite robust and Table 2 succinctly portrays the statistical and substantive importance of the variables.

Robustness Checks

Times series, cross-sectional analyses require careful assessments of unit heterogeneity and dynamic structures (e.g., Beck, 2007; Stimson 1985; Wilson & Butler, 2007). Unit heterogeneity for this work simply refers to unknown differences across the committees. Various models were run with and without fixed effects for committees.¹⁰ In nearly all instances, variables neither lost nor gained the magical .05 level of significance. We retained the fixed effects for committees for two reasons. First, it is never possible to know that all effects have been measured. For instance, there might be differences across committees tied to legislators’ personalities. Second, using fixed effects to address unit heterogeneity “raises the bar for confirming our theories” (Wilson & Butler, 2007, p. 106), which strengthens our substantive claims.

Dynamic structures address the time series side of the analysis. Beck and Katz (1995) and many others simply employ a lagged dependent variable. However, Wilson and Butler (2007) show that “there is no logical reason why the LDV [lagged dependent variable] model championed by B&K should be considered more plausible than any of the other dynamic models,” including those with lagged independent variables (Wilson & Butler, 2007, p. 107). Here, the dependent variable and the signing statement variables were lagged for clear substantive reasons. Oversight activity during one month often bleeds into oversight activity in the following month. Similarly, signing statements do not elicit an immediate response, but congressional committees do respond over a relatively short period of time. Theory might allow one to predict lagged effects but theory seldom specifies the exact number of lags to anticipate. Often, increased lags are included simply to insure a proper model specification. The impact of right-hand-side variables must, however, at some point dissipate as their lags are increased. The lagged variables in the models here do indeed show proper dissipation. One final concern for dynamic panel models such as those presented here is the presence of bias due to a short time series. For most times series, T = 30 is large enough to secure a good specification. The models shown here have over 150 time periods. If bias due to T remains, it generally affects the coefficient of the lagged dependent variable to a much greater extent than the other right-hand-side variables.

The special robustness checks demanded by time series, cross-sectional data structures do not replace the more ordinary robustness checks. It is impossible to detail every robustness check because each model was tested under a variety of conditions to verify the strength of our findings. For instance, Model 2 was run under unified and divided government and the results remained largely unchanged. In addition, 1 year at a time was removed but the results did not change significantly. One committee at a time was removed but only minor changes resulted. Models for nonelection years were run but again only minor changes resulted. Multicollinearity checks were also run.

To investigate presidential effects, separate models for each president were run. The Republican-controlled House under President Clinton was slower to engage in oversight activity after signing statements. The coefficient for the number of objections variable was negative and significant at 1 and 2 months but was positive and significant at 3, 4, and 5 months. In contrast to the Clinton era, the Republican-controlled House under President Bush was quicker to engage in oversight activity. The coefficient for the number of objections in signing statements was positive and significant at 1, 2, 3, 4, and 5 months. In sum, oversight triggered by signing statements looked very similar under the Clinton and Bush presidencies over the 5-month period. Although there are some differences between the Clinton and Bush eras, those differences are not particularly compelling.

Omitted variable bias might be a concern for some readers if only because endogeneity problems are endemic in the social sciences. Such an omitted variable would have to be unrelated to our standard measures of interbranch competition. As a thought experiment, consider some measure of external controversy that leads to both signing statements and congressional oversight. The measure would need to capture controversy across a wide array of bills. Furthermore, that measure of controversy would have to be unrelated to presidential and congressional ideologies, which we already address. Controversy might be unique to the general issue or the specific legislation at hand. However, issues are generally tied to committees and we do use committee fixed effects. We also control for bill length, which is issue specific.

Discussion

Why would a president continue to use signing statements if Congress might be able to counter those statements with increased oversight? There are several issues to consider when approaching this question. First, the issuance of a signing statement bears little cost to the president.¹¹ The signing statement requires little effort from the president (a unitary actor), whereas a congressional response to the statement (through heightened oversight activity) requires coordinated, collective effort. Responding to signing statements forces members of Congress to engage a collective action problem, which by definition is costly for members.

Second, if the purpose of a signing statement is to affect agency behavior, presidents have little disincentive to issue signing statements because they introduce a “dual principal” problem for agents, thereby potentially permitting bureaucratic drift away from congressional preference (Whitford, 2005). If the collective efforts within Congress fail to counter presidential preferences (as articulated through signing statements), agencies are left with only presidential directives to guide the placement of policy.

Third, signing statements that lead to specific agency behavior can have a “stickiness” that other methods of presidential influence (such as executive orders and directives) lack. Initial agency behavior establishes the status quo for implementation and the status quo is almost always favored over any other alternatives.

Fourth, like presidential directives, signing statements may serve to “make presidential intervention in regulatory matters ever more routine and agency acceptance of this intervention ever more ready” (Kagan, 2001, p. 2299). Signing statements, then, may serve the long-term interest of executives by creating a context where heightened authority in implementation is presumed. Even though signing statements reveal presidential preferences to Congress, they may be part of a coordinated strategy—along with directives and executive orders—to protect and expand institutional authority. All are compiled and published; none is unknown to agents, Congress, or the public. As Kagan (2001) writes of the Clinton Administration’s use of presidential unilateralism, “[t]he unofficial became official, the subtle blatant, and the veiled transparent” (p. 2299).

Fifth, signing statements might also affect judicial branch interpretations of agency actions (e.g., Cooper, 2005). Presidents have an incentive to articulate constitutional objections to provisions of bills in the hopes that future courts weigh presidential objections just as they weigh legislative histories. Modes of statutory construction that include legislative intent or legislative purpose might be particularly sensitive to presidential perspectives. Kelley (2007), however, argues that the Reagan Administration’s insistence that signing statements be included in the (post)legislative record was intended to drive the direction of implementation rather than judicial decision making.

Finally, signing statements can be an instance of a positive sum tool (Wiseman, 2009). Tools of bureaucratic control that appear to favor one branch can under various circumstances benefit both the executive and legislative branches. Presidents secure a firmer grasp of agency reins and Congress gains more detailed information about presidential directives. If signing statements are low cost, if congressional response is not certain (or perhaps complicated by collective action problems), if there is a chance to shape bureaucratic behavior over the long term and insulate policy from future executive action, if presidents can “soften the system” for regulatory intervention over time, or if there is a chance that a court’s construction of the statute might favor the president’s view, then the issuance of a signing statement can indeed benefit the president—even as the information gleaned by committees benefits Congress. Under such conditions, a president can reasonably conclude that the likelihood of heightened congressional oversight might be worth the potential gains in policy achieved when signing statements have an effect on implementation.

Most studies of congressional oversight focus on annual activity. We know of no large-N studies of oversight activity focused on monthly committee activity. We know of no other studies of oversight with NxT data structures. Given that so many scholars now accept the importance of trigger mechanisms inherent in the fire-alarm analogy for oversight activity, the annual assessments of oversight seem ill-suited for assessing potential triggers. Triggers suggest more immediate responses. Case studies often focus on a particularly unique trigger spurring congressional oversight. For instance, 100-year floods, record-breaking market collapses, and unprecedented terrorist attacks all prompt congressional oversight. The monthly assessment of oversight activity allows one to assess congressional committee responses to another sort of trigger mechanism. Objections in presidential signing statements provide clues about executive branch noncompliance, which congressional oversight is meant to correct or preempt. The results in the previous section fit nicely within this framework because the number of objections contained in signing statements acts as an oversight trigger, particularly in the months immediately after the issuance of the offending signing statements.

The results also indicate that the effect of that triggering mechanism dissipates over time. Congressional sensitivity to the number of objections is greatest in the 3rd and 4th months subsequent to the issuance of a signing statement, but by the 5th month, the level of a congressional response diminishes by 50%. (Note the third and fifth columns in Table 1.)

The extant theoretical and empirical results related to congressional oversight have failed to settle many debates. Many scholars lament the perceived lack of congressional oversight (e.g., Ogul, 1976) or suggest that researchers fundamentally misunderstand the incentives behind congressional oversight (McCubbins & Schwartz, 1984). McCubbins and Schwartz (1984) argue that police-patrol oversight is costly and offers few rewards. Responding to fire alarms allows for greater credit claiming and entails reduced monitoring costs.¹² How do police patrols and fire alarms dovetail with executive–legislative relations? The executive branch plays a silent role in the police-patrol and fire-alarm analogy, as if members of Congress choose an oversight strategy without regard to executive branch actions. Of course, executive–legislative relations are at the root of congressional oversight and many scholars, including Aberbach (1990) and Ogul and Rockman (1990), argue that a wider set of institutional relations deserve attention. Although not explicitly prescribed by them, signing statements provide the sort of bridge between the branches that Aberbach (1990) and Ogul and Rockman (1990) sought. How do signing statements dovetail with the police-patrol and fire-alarm analogy? The objections contained in signing statements reduce the costs of congressional oversight by reducing the information asymmetries between the branches. Monitoring and information-gathering costs make police patrols less attractive to legislators but the objections in signing statements reduce those costs.

Conclusion

The portrayals in the popular press and the recorded frustrations in the U.S. Congress reveal something of a puzzle. The common presumption has been that signing statements are powerful tools for the president—but no one has explored congressional responses to signing statements. Even after including the lagged dependent variable and controlling for a large number of factors that potentially affect congressional oversight, the number of objections in signing statements remains a statistically and substantively significant predictor of committee oversight.

Of course, signing statements are not the sole engines for monthly measures of congressional oversight. Opportunities and resources remain important for congressional oversight. Days in session provide an essential resource—time, but the signing statements remain crucial trigger mechanisms. The signing statement variables in the few months prior to the occurrence of oversight activity rival and at times overshadow the impact of the days in session. Indeed, the substantive impact of the signing statement measures also overwhelms the lagged dependent variable. There are also clear tradeoffs for a committee between legislative and oversight activity since more legislatively active committees are less likely to engage in oversight.

Typically, trigger mechanisms have been linked to fire-alarm oversight. The results in this work have uncovered the signing statement as a hitherto overlooked trigger mechanism for oversight. The canonical view of presidential signing statements—as powerful tools of presidential control over policy—might not capture the totality of the policymaking context for bureaucratic agents. While future work might consider the degree to which bureaucracies are sensitive to signing statements, we uncover evidence that congressional committees pay attention to the objections in signing statements. In the competition over the political control of the bureaucracy, presidential signing statements often trigger congressional oversight.

Footnotes

Appendix

Summary Statistics for Oversight Variables, N = 2,686

Variable	M	SD	Minimum	Maximum
Number of oversight hearings	.872	1.853	0	20
Number of objections	.476	3.962	0	110
Divided government	.6875	.464	0	1
President−committee median distance	.687	.188	.333	1.257
Days in session	10.958	5.943	0	21
Committee activity	1.381	2.298	0	28
Iraq/counterterror committee	.063	.242	0	1
Bill length	3,887.73	25,763.90	0	472,010

Acknowledgements

We would like to thank Ryan Bakker, Jamie Carson, Bob Grafstein, Gbemende Johnson, Tony Madonna, Jamie Monogan, Keith Poole, Jason Roberts, and the APR editor and reviewers. We also thank Chelsea Harris and Brandon Rahn for their research assistance.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: Research support was provided through a grant from the Undergraduate Research and Creative Activities fund, Southern Illinois University.

Notes

Bios

Scott H. Ainsworth is Professor of Political Science in the School of Public and International Affairs at the University of Georgia. His scholarship focuses on lobbying and strategic interactions between lobbyists and government officials and between officials in different branches of government.

Brian M. Harward is Associate Professor of Political Science and Director of the Center for Political Participation at Allegheny College. His recent publications explore issues related to federalism, the presidency, and congressional oversight.

Kenneth W. Moffett is assistant professor of political science at Southern Illinois University Edwardsville. His research focuses on Congress, the presidency, and the courts.

References

Aberbach

(1990). Keeping a watchful eye: The politics of congressional oversight. Washington, DC: The Brookings Institution Press.

Ainsworth

S. H.

Harward

B. M.

(2009). Delegation and discretion in anticipation of coalitional drift. American Politics Research, 37, 983-1002.

American Presidency Project. (2009). Web site maintained by Woolley, J., & Peters, G. Retrieved from http://www.presidency.ucsb.edu/index.php

Balla

Deering

(n.d.). Police patrols and fire alarms: An examination of the legislative preference for oversight (Working Paper at George Washington University). Retrieved from home.gwu.edu/~rocket/oversight.doc

Bawn

(1995). Political control versus expertise: Congressional choices about administrative procedures. American Political Science Review, 89, 62-73.

Beck

(2007). From statistical nuisances to serious modeling: Changing how we think about the analysis of time-series—cross-sectional data. Political Analysis, 15, 97-100.

Beck

Katz

(1995). What to do (and not to do) with time-series cross-sectional data. American Political Science Review, 89, 634-647.

Bradley

Posner

(2006). Presidential signing statements and executive power. Constitutional Commentary, 23, 307-364.

Bush

G. W.

(2006). Statement on signing the postal accountability and enhancement act. Retrieved from http://www.presidency.ucsb.edu/ws/index.php?pid=24392

10.

Chatfield

(2004). The analysis of time series: An introduction. London, UK: Taylor & Francis.

11.

Cooper

(2002). By order of the president: The use and abuse of executive direct action. Lawrence, KS: University of Kansas Press.

12.

Cooper

(2005). George W. Bush, Edgar Allan Poe, and the use and abuse of presidential signing statements. Presidential Studies Quarterly, 35, 515-532.

13.

Cross

F. B.

(1988). The constitutional legitimacy and significance of “presidential signing statements.” Administrative Law Review, 40, 209.

14.

Enders

(2009). Applied econometric time series. Hoboken, NJ: John Wiley.

15.

Epstein

O’Halloran

(1996). Divided government and the design of administrative procedures: A formal model and empirical test. Journal of Politics, 58, 393-417.

16.

Epstein

O’Halloran

(1999). Delegating powers: A transaction cost politics approach to policy making under separate powers. Cambridge, UK: Cambridge University Press.

17.

Fisher

(2007). Signing statements: Constitutional and practical limits. William & Mary Bill of Rights Journal, 16, 183-210.

18.

Garber

Wimmer

(1987). Presidential signing statements as interpretations of legislative intent: An executive aggrandizement of power. Harvard Journal on Legislation, 24, 663-366.

19.

Gormley

W. T.

Jr. Balla

S. J.

(2008). Bureaucracy and democracy: Accountability and performance (2nd ed). Washington, DC: CQ Press.

20.

Greene

W. H.

(2000). Econometric analysis. Upper Saddle, NJ: Prentice Hall.

21.

Halstead

T. J.

(2007). Presidential signing statements: Constitutional and institutional implications. Retrieved from www.gao.gov/new.items/d08553t.pdf

22.

Halstead

T. J.

(2008). Presidential signing statements: Executive aggrandizement, judicial ambivalence and congressional vituperation. Government Information Quarterly, 25, 563-591.

23.

Heclo

(1977). A government of strangers: Executive politics in Washington. Washington, DC: Brookings Institution Press.

24.

Howell

W. G.

(2003). Power without persuasion: The politics of direct presidential action. Princeton, NJ: Princeton University Press.

25.

Huber

Shipan

C. R.

(2002). Deliberate discretion? The institutional foundations of bureaucratic autonomy. Cambridge, UK: Cambridge University Press.

26.

Huber

J. D.

Shipan

C. R.

Pfahler

(2001). Legislatures and statutory control of bureacracy. American Journal of Political Science, 45, 330-345.

27.

Kagan

(2001). Presidential administration. Harvard Law Review, 114, 2245-2385.

28.

Keele

Kelly

N. J.

(2006). Dynamic models for dynamic theories: The ins and outs of lagged dependent variables. Political Analysis, 14, 186-205.

29.

Kelley

(2006). The significance of the presidential signing statement. In Kelley

C. S.

(Ed.), Executing the constitution: Putting the president back into the constitution. Albany: State University of New York Press.

30.

Kelley

(2007). A matter of direction: The Reagan administration, the signing statement, and the 1986 Westlaw decision. William & Mary Bill of Rights Journal, 16, 283-306.

31.

Kelley

Marshall

(2008). Assessing presidential power: Signing statements and veto threats as coordinated strategies. American Politics Research, 37, 508-533.

32.

Lee

(2008). Reorienting the debate on presidential signing statements: The need for transparency in the president’s constitutional objections, reservations, and assertions of power. UCLA Law Review, 55, 705-744.

33.

Lewis

(2003). Presidents and the politics of agency design. Palo Alto, CA: Stanford University Press.

34.

Library of Congress. (2009). Thomas: Legislative information on the internet. Retrieved from http://thomas.loc.gov

35.

Light

P. C.

(1982). The president’s agenda: Domestic policy choice from Kennedy to Carter. Baltimore, MD: Johns Hopkins University Press.

36.

Long

J. S.

(1997). Regression models for categorical and limited dependent variables. Thousand Oaks, CA: SAGE.

37.

Long

J. S.

Freese

(2006). Regression models for categorical dependent variables using Stata. College Station, TX: Stata Press.

38.

MacDonald

(2007). Agency design and postlegislative influence over the bureaucracy. Political Research Quarterly, 60, 683-695.

39.

Maltzman

(1998). Competing principals: Committees, parties, and the organization of congress. Ann Arbor: University of Michigan Press.

40.

Mayhew

D. R.

(2005). Divided we govern: Party control, lawmaking and investigations, 1946-2002. New Haven, CT: Yale University Press.

41.

McCubbins

M. D.

Noll

R. G.

Weingast

B. R.

(1989). Structure and process, politics and policy: Administrative arrangements and the political control of agencies. Virginia Law Review, 75, 431-482.

42.

McCubbins

M. D.

Schwartz

(1984). Congressional oversight overlooked: Police patrols versus fire alarms. American Journal of Political Science, 28, 165-179.

43.

Moe

T. M.

(1989). The politics of bureaucratic structure. In Chubb

Peterson

(Eds.), Can the government govern? Washington, DC: Brookings Institution Press.

44.

Moe

T. M.

(1993). Presidents, institutions, and theory. In Edwards

G. C.

III. Kessel

J. H.

Rockman

B. A.

(Eds.), Researching the presidency: Vital questions, new approaches (pp. 337-385). Pittsburgh, PA: University of Pittsburgh Press.

45.

Moe

T. M.

Wilson

S. A.

(1994). Presidents and the politics of structure. Law and Contemporary Problems, 57, 1-44.

46.

Ogul

M. S.

(1976). Congress oversees the bureaucracy: Studies in legislative supervision. Pittsburgh, PA: University of Pittsburgh Press.

47.

Ogul

M. S.

Rockman

B. A.

(1990). Overseeing oversight: New departures and old problems. Legislative Studies Quarterly, 15, 5-24.

48.

Oleszek

W. J.

(1995). Congressional procedures and the policy process. Washington, DC: CQ Press.

49.

Pfiffner

J. P.

(2009, March/April). Presidential signing statements and their implications for public administration. Public Administration Review, 69, 249-255.

50.

Poole

K. T.

Rosenthal

(1997). Congress: A political-economic history of roll call voting. New York City, NY: Oxford University Press.

51.

Rourke

F. E.

(1960). Administrative secrecy: A congressional dilemma. American Political Science Review, 54, 684-694.

52.

Sala

(1998). In search of the administrative president: Presidential “decree” powers and policy implementation in the United States. In Carey

J. M.

Soberg Shugart

(Eds.), Executive decree authority (pp. 254-273). Cambridge, UK: Cambridge University Press.

53.

Savage

(2006, April 30). Bush challenges hundreds of laws. The Boston Globe, p. A1.

54.

Savage

(2007). Takeover: The return of the imperial presidency and the subversion of American democracy. New York, NY: Little, Brown.

55.

Savage

(2009, April 9). Obama’s embrace of Bush tactic riles Congress. New York Times, p. A16.

56.

Sonnett

N. R.

(2006). ABA taskforce report on presidential signing statements and the separation of powers doctrine. Retrieved from http://www.americanbar.org/content/dam/aba/migrated/leadership/2006/annual/dailyjournal/20060823144113.authcheckdam.pdf

57.

Specter

(2009, May 14). The need to roll back presidential power grabs. New York Review of Books, 56(8).

58.

Stimson

J. A.

(1985). Regression in time and space: A statistical essay. American Journal of Political Science, 29, 914-947.

59.

Tribe

(2006, August 9). “Signing statements” are a phantom target. The Boston Globe, p. A9.

60.

U.S. Congress, Senate. (2006, July 25). Congressional record (109th Congress, Second Session). Washington, DC: Government Printing Office.

61.

U.S. House of Representatives. Office of the Clerk. (2009). Historical Information about the United States House of Representatives. Retrieved from http://clerk.house.gov

62.

Vuong

Q. H.

(1989). Likelihood ratio tests for model selection and non-nested hypotheses. Econometrica, 57, 307-333.

63.

Weisman

(2009, July 15). President’s signing statements anger lawmakers. Wall Street Journal, p. A6.

64.

Whitford

A. B.

(2005). The pursuit of political control by multiple principals. Journal of Politics, 67, 29-49.

65.

Wilson

S. E.

Butler

D. M.

(2007). A lot more to do: The sensitivity of time-series cross-sectional analyses to simple alternative specifications. Political Analysis, 15, 101-123.

66.

Wiseman

A. E.

(2009). Delegation and positive-sum bureaucracies. Journal of Politics, 71, 998-1014.