Bureaucratic Capacity and Bureaucratic Discretion: Does Congress Tie Policy Authority to. Performance? Jason A. MacDonald

Transcription

1 Bureaucratic Capacity and Bureaucratic Discretion: Does Congress Tie Policy Authority to Performance? Jason A. MacDonald Department of Political Science Kent State University Kent, OH (Home) (Work) William W. Franko, Jr. Department of Political Science Kent State University Kent, OH (Home) (Work)

2 Biographical Information Jason A. MacDonald is an Assistant Professor in the Department of Political Science at Kent State University. William W. Franko, Jr. is a PhD candidate in the Department of Political Science at Kent State University.

3 Acknowledgements The authors would like to thank Steve Balla and anonymous reviewers for American Politics Research for comments that improved this manuscript. This research was supported in part by the University Research Council of Kent State Univeristy.

4 Abstract This paper assesses whether the managerial capacity of agencies influences the volume of policy authority that lawmakers delegate. By examining a sample of agencies whose managerial capacities were assessed along the same criteria, allowing for the comparison of performance across agencies, we observe that poorly performing agencies are more likely to lose policy authority. Our findings suggest that lawmakers promote effective policymaking by giving agencies the incentive to perform well and that models of discretion that do not account for performance underestimate the effect of another factor policy conflict between the legislative and executive branches on how much discretion agencies receive.

5 Modern democracies confront complex problems, often employing policies that combine scientific knowledge across disciplines from the natural sciences and engineering to economics and policy analysis. Given this complexity, it is unlikely that modern lawmakers with their backgrounds in law, business and public service will ever be the most well equipped individuals in government to design policy mechanisms to pursue favorable outcomes to these problems. Yet, electoral status confers on lawmakers both the legitimacy to make policy decisions and the incentive to balance competing societal values and interests successfully. In part, lawmakers manage this responsibility-capacity mismatch by delegating authority to make policy decisions to the bureaucracy, as has been documented in diverse research traditions from the political economy of institutional design (McCubbins, Noll, & Weingast, 1987; Moe, 1989) to American political development (Carpenter, 2001; James, 2000). A major area of research on delegation involves assessing why lawmakers vary the level of discretion provided to agencies, where discretion is defined as bureaucratic freedom to make policy decisions free from constraints, such as rulemaking requirements and other tools used by lawmakers to influence the substance of bureaucratic decisions (Epstein & O Halloran, 1999, chap. 5). A central finding of this research is that, as policy disagreement between lawmakers and agencies increases, lawmakers reduce the volume of discretion that agencies receive (Epstein & O Halloran, 1999; Huber & Shipan, 2002; Huber, Shipan, & Pfahler, 2001; Lewis, 2003; Potoski, 1999; Wood & Bohte, 2004). The theoretical basis for this finding is that policy disagreement leads lawmakers not to trust agencies to render policy decisions consistent with lawmakers priorities. In the language of the transaction cost approach taken by these studies, such disagreement increases the costs of delegation to lawmakers to the point at which it

6 becomes less costly for them to make policy themselves by writing detailed laws (see especially Epstein and O Halloran (1999) and Huber and Shipan (2002)). These studies constitute significant theoretical and empirical progress in understanding why agencies receive discretion to make policy. However, this literature does not address the question of whether lawmakers vary discretion based on the capacity of agencies to make policies that are effective in meeting policy goals. This issue is central to understanding whether lawmakers decisions in delegating policy authority contribute to the capacity of democratic governments to solve problems. Some agencies perform the tasks delegated to them effectively, solving problems that the architects of legislation place in their hands; other agencies flounder (Ingraham, Joyce, & Donahue, 2003). To be sure, theories of delegation stress that lawmakers provide greater discretion to agencies as policy complexity increases (Bawn, 1995; Epstein & O Halloran, 1999; Huber & Shipan, 2002). For any number of equally complex policy areas, though, the capacity of agencies to make effective policy can vary. Do agencies with greater/lesser capacities receive higher/lower levels of discretion? If the answer is affirmative, then there is reason to believe that democratic governments can create effective solutions to important problems; however, if lawmakers do not tie bureaucratic authority, at least in part, to bureaucratic capacity, students of government should be more sanguine about the ability of democracies to solve many problems. Agency Capacity and Bureaucratic Discretion Research on discretion, bureaucratic autonomy and agency termination provides a basis for the hypothesis that agency capacity affects the latitude agencies receive to make policy decisions, though no empirical research employs a systematic indicator of capacity that varies across agencies to assess the relationship. With respect to discretion, formal models predict that 2

7 lawmakers provide greater discretion to agencies as policy complexity increases even when lawmakers expect that agencies will make policy decisions that stray from lawmakers priorities (Bawn, 1995; Epstein & O Halloran, 1999; Huber & Shipan, 2002). The basis for this result as that lawmakers need for policy solutions trumps the loss of utility they experience from bureaucratic shirking. The implication of this research for the relationship between capacity and discretion is clear. If lawmakers did not care about effective policy solutions, there would be no reason to delegate when agencies are likely to make decisions that stray from lawmakers priorities. Of course, if agencies reputations for effective policy making are poor, then lawmakers have reason to doubt that agencies will produce effective solutions, undercutting the rationale for ceding discretion. This literature, therefore, suggests that agency capacity to perform the tasks delegated effectively is positively related to discretion. Research on bureaucratic autonomy suggests the same relationship, though the causal mechanism differs. Examining the histories of three agencies during the late 19 th and early 20 th centuries, Carpenter (2001) argues that elected institutions cede policy authority after agencies evidence the capacity to solve policy problems. Briefly, this capacity was created by middle level managers, whose career longevities and institutional positions fostered policy learning and the ability to build support for policies they favored among diverse sets of interest groups. After securing interest group support for the policies they wanted to pursue in part because of sound reputations for the capacity to solve problems effectively these managers were able to place electoral pressure on members of Congress to give managers authority to create policies that managers favored. Bureaucratic autonomy, then, was largely the result of sound policy performance engineered by managerial leadership. 3

8 Research on agency termination also provides a basis for the hypothesis that discretion is due to the capacity to make policies effectively. Carpenter (2000) and Carpenter and Lewis (2004) argue that failure by an agency, accompanied by media coverage of the negative consequences of its bungling of the tasks to which it was assigned, imposes political fallout on the legislators that delegated authority. Such costs might include the loss of electoral support from key constituencies, the provision of an election issue to an opponent who could trace the lawmaker to the failure (Arnold, 1990) by credibly arguing that the lawmaker presided over the fiasco, and having to allocate scarce time on the legislative calendar in the future to revisit the policy. These costs increase the likelihood that legislators will exercise the ultimate act of political control (Carpenter & Lewis, 2004, p. 202), eliminating the agency an act that, to understate the point, reduces discretion. In summary, various research traditions on the authority that agencies receive either suggest, or explicitly state, that discretion increases with agency capacity. The basis for this relationship is that lawmakers observe the capacity of agencies to perform the policy-making tasks delegated to them through a variety of means, including media reports (Carpenter & Lewis, 2004) and information from interest groups and constituents dissatisfied with agencies (McCubbins & Schwartz, 1984) and react by manipulating discretion in the future. Yet, empirical support for this hypothesis is wanting. Neither research on discretion nor agency termination incorporate variables into empirical models of these phenomena to assess the influence of capacity on bureaucratic policy authority. Although Carpenter (2001) shows that the authority of three agencies was due in large part to their reputations for effective policy-making, this finding has not been extended to contemporary politics. 4

9 Data and Methods One reason why the relationship between capacity and discretion is not well understood empirically is due to the difficulty of calibrating capacity/performance in a valid manner and comparing it systematically across agencies. We take advantage of the availability of such a measure for a sample of 27 federal agencies. This sample was created by a joint effort, the Federal Performance Project (FPP), between scholars in Department of Public Administration at The George Washington University (GW) and correspondents for Government Executive, a biweekly periodical that provides specialized coverage of federal agencies. The project evaluated how well all agencies managed the tasks to which they were assigned, grading agencies in 1999, 2000 and To be clear, the FPP did not assess how well agencies achieved the policy goals assigned to them by laws enacted in the past. Rather, the project assessed how well agency managers Managed for Results based on the criteria employed by the researchers. 1 Managing for Results refers to administrating agencies in order to realize the outcomes they are charged with achieving. The researchers employed surveys of agency managers and employees, external reports on agency effectiveness and interviews with experts on the agencies inside and outside of government to gauge performance. 2 Researchers at GW and correspondents for Government Executive then evaluated this information across agencies and assigned grades based on the performance of agencies in relation to one another. Hence, according to the collective judgment of the researchers, agencies receiving A s performed better than agencies receiving B s on these criteria and so on. As such, the grades constitute assessments of the managerial capacity of agencies. A presence/lack of capacity indicates the success/failure of managers to administer agencies in a manner that allows them to realize the policy results they are charged with achieving by their political principals. 5

10 Although the sample of agencies is small, it represents the best data on agency capacity that is comparable across multiple agencies. 3 Hence, in evaluating the connection between capacity and discretion, it makes sense to use this data as a starting point. The 27 agencies were selected by FPP researchers due to their close interaction with the public. Therefore, the agencies do not represent a random sample of all federal agencies and our findings cannot be so generalized. Table A1 in the Appendix itemizes these agencies. The unit of analysis is the agency; that is, each agency has one observation and includes information on the grade received in the year that it was graded as well as the other independent variables and the dependent variable described below. In evaluating the agencies management of their tasks, the FPP employed a grade range from F to A. However, no agency received F to D- grades, limiting the range of the variables created to measure performance. Below, we assess the relationship between how much Congress limits the agencies discretion in the year after the grades were assigned and these grades. We do so by measuring capacity in three ways. First, we employ the plus/minus grades on a 1-13 scale with a grade of F coded as 1 and an A coded as 13 (in practice, the scale ranged between 4-13). Second, we employ a collapsed, traditional grade, version of this variable ranging from F (1) to A (5) to guard against the possibility that there is little difference between, for example, a B- and B (in practice, this scale ranged between 2-5). Finally, to account for the possibility that performance affects discretion in a non-linear manner, we create dummy variables for whether the agencies received grades of D, C, or B (1 if the agencies received these grades; 0 otherwise), with A as the reference category (since no agency received an F, no dummy variable for this grade was created). The basis for this variable is that Congress may not view the difference between an A and B performance as it 6

11 does a C and D performance. For example, Congress may take no action to limit the discretion of an agency when it receives a B instead of an A; although the agency is not performing in an excellent manner, its merely good performance probably will not cause political fallout. However, once an agency s performance becomes lackluster, e.g., the agency s performance is worthy of a D, political principals may limit discretion. Calibrating performance in this way allows us to observe such non-linear effects. To be clear, we offer no theory that specifies precisely how poorly an agency must perform to lose discretion; rather, we note that it makes sense theoretically that Congress will limit discretion after performance has become sufficiently poor. We leave how low capacity must be to reach this tipping point as an empirical question. 4 To assess the relationship between agency capacity and authority, we employ an indicator of how much authority is removed from agencies in the year after their managerial capacities were assessed by the FPP, the number of limitation riders (LRs) attached to the agencies appropriations. LRs forbid agencies from spending money for specific purposes. Importantly, agencies possess the authority to use funds for these purposes from past laws. However, when a LR is included in an appropriations bill that becomes law, agencies are prohibited from exercising that authority during the next fiscal year, limiting their discretion. For example, the fiscal year 2001 Labor, Health and Human Services and Education appropriations bill mandated that none of the funds... may be used by the Occupational Safety and Health Administration to promulgate, issue, implement, administer, or enforce any proposed, temporary, or final standard on ergonomic protection. Congress includes hundreds of such LRs in appropriations bills annually (MacDonald, 2006), giving it an annual opportunity to reign in agency authority. As 7

12 such, LRs constitute a tool that Congress employs to constrain agency authority regularly, making these tools of political control a valid indicator the limitation of discretion. 5 To measure how much Congress impinged on discretion, we counted the number of LRs in Congress s annual appropriation acts that applied to each of the 27 agencies in the year after their performances were evaluated by the FPP. 6 We examined LRs in the subsequent year because that was Congress s first opportunity to limit discretion after performance was assessed during the period for which grades were assigned. 7 Therefore, for agencies whose performances were assessed in 1999/2000/2001, we counted LRs imposed on these agencies in the appropriations bills passed in 2000/2001/2002. If discretion is reduced because of low capacity, the relationship between this dependent variable and the independent variables measuring capacity using the plus/minus and traditional grade scales should be negative and significant, whereas its relationship with the dummy variables indicating the presence of low capacity (e.g., the dummy variable for agencies who received the grade of D ) should be positive and significant. Since research demonstrates that Congress limits discretion when faced with a President of the opposite party (Epstein & O Halloran, 1999), a finding that holds for state governments (Huber & Shipan, 2002; Huber, Shipan, & Pfahler, 2001), we control for partisan conflict between the legislative and executive branches. To do so, we employ a dummy variable assuming the value of 1 when there is pure divided government, meaning that the presidency was controlled by one party and both chambers of Congress were controlled by its rival. Therefore, observations for agencies graded in 1999, for which we counted LRs applied in 2000 (under Democratic control of the presidency and Republican control of the House and Senate), were coded 1. Observations for agencies graded in 2000 and 2001, for which we counted LRs 8

13 applied in 2001 and 2002 respectively (under Republican control of the presidency and House and Democratic control of the Senate), were coded as 0. Although the Democratic Senate had the opportunity to influence the substance of appropriations bills during these years, its capacity to do so was limited to a greater degree than was the case for the Republicans, who controlled both chambers, in We expect this variable to be positively and significantly related to the number LRs imposed on agencies. We also account for congressional and presidential policy disagreement with agencies missions. Carpenter and Lewis (2004) show that agencies are more likely to be eliminated when the same party controls the U.S. House and the presidency but held minority status in the House and did not control the presidency when the agency was created. Agencies in this situation are in a precarious position because the current majority party is likely to object to the policy priorities embodied in the agencies missions. These conditions a unified government hostile towards agencies missions makes it more likely that such agencies will lose discretion. Therefore, following Carpenter and Lewis (2004), we create a variable assuming the value of 1 when these hostility conditions hold; 0 otherwise, and expect this variable to be positively and significantly associated with the imposition of LRs. Additionally, since research shows that the public salience of policies increases the likelihood that members of Congress make, rather than delegate, policy (Epstein & O Halloran, 1999, chap. 8; Gormley, 1986, 1989) and that Congress tries to influence bureaucratic policy decisions to a greater extent in salient policy areas (Rinquist, Worsham, & Eisner, 2003), we control for the salience of the policy area over which agencies have jurisdiction. This variable is a count of the number of stories appearing in the New York Times in which the agency was mentioned during the year before the agency was graded. 8 In general, the Time s coverage is 9

14 followed by other media outlets (Page, 1996). As such, this count approximates the degree to which the agencies missions are made salient to the public. We expect that it will be positively and significantly related to the imposition of LRs. 9 Finally, we control for the size of the appropriation bill in which the agencies were funded, since bigger bills may contain a higher volume of LRs. Table A2 in the Appendix provides summary statistics for all variables employed in the analysis. 10 Since the dependent variable is a count, we employ the poisson maximum likelihood estimator to assess the effects of the independent variables on the volume of LRs using the traditional grade scale specification and the dummy variable grade specification. In using the plus-minus grade scale specification, likelihood ratio tests indicated that the variance of the dependent variable exceeded its mean; therefore, we employed the negative binomial maximum likelihood estimator for this model. Findings Models 1, 2 and 3 of Table 1 present estimates for the influence of the independent variables on the volume of LRs, employing the traditional grade scale variable (Model 1), the plus/minus grade scale variable (Model 2) and dummy variables for agency grades (Model 3). The hypothesis that Congress limits discretion as capacity declines is supported in all three models. Measuring capacity using the traditional and plus/minus grades assigned to the agencies yields a negative and significant association between the number of LRs imposed on agencies and the coefficient for agency grades. Additionally, in Model 3, the coefficients of the dummy variables indicating that the agency received C s and D s are positively and significantly associated with the number of LRs with which agencies were burdened. This specification supports the interpretation that once agency performance drops below some adequate level at which Congress is willing to leave agencies alone, Congress will reduce 10

15 discretion. Empirically, Model 3 identifies this threshold level as the performance that merits a C based on the FPP s criteria. [Insert Table 1] Table 2 provides information on the magnitude of the influence of performance, as measured by the traditional grade scale, on LRs use, as estimated by Model 1. Setting the values of the independent variables to the modal or mean values, the model predicts that about 20 LRs will be imposed on agencies receiving a B. 11 However, when agencies receive an A, this prediction drops to approximately 17 LRs. Conversely, when agencies managerial capacities are graded at the C and D levels, the model predicts that agencies will be saddled with about 25 and 30 LRs respectively. Hence, a decline in performance from the maximum to minimum grades assigned by the FPP leads to almost a 2 standard deviation increase in the number of LRs that Congress is expected to impose on agencies. Simulations not presened based on the coefficients from Model 2 present a similar story. Simulations using the coefficients from Model 3 paint a more nuanced picture of the relationship between managerial performance and discretion. Setting the independent variables to their mean and modal values with the dummy variable for a B grade equal to 1, Model 3 predicts that Congress will impose about 21 LRs on agencies, a prediction that decreases to 18 if agencies receive A s, and increases to 24 if agencies receive C s. However, Model 3 predicts that Congress will attach about 41 LRs to appropriations language funding agencies programs during the next fiscal year if a D is assigned to their managerial performance. This finding suggests that there is a non-linear relationship between performance and discretion. It is also consistent with a bounded rationality explanation of political institutions (Jones & Baumgartner, 2005). Specifically, the finding suggests that Congress under-reacts to information stressing the performance of agencies is 11

16 merely average ( C grades), limiting discretion slightly. However, when Congress learns that agency performance is poor ( D grades), it decreases discretion greatly. [Insert Table 2] The hypothesis that divided government reduces discretion receives some support for the models presented in Table 1. The coefficient for the divided government variable is positively and significantly related to the number of LRs imposed on agencies in models 1 and 3, albeit at the.1 level. Turning to the hypothesis that a unified and hostile government leads to less discretion, the coefficient for a hostile and unified government is positive and significant in all three models, providing support for the hypothesis. Table 2 provides information on the magnitude of this relationship, showing that Model 1 predicts about 8 additional LRs (approximately 1 standard deviation of the dependent variable) for an agency supervised by a U.S. House and a President controlled by the opposite party that created the agency. Additionally, the number of stories in the New York Times mentioning agencies is positively and significantly related to the number of LRs imposed on agencies in all three models presented in Table 1. Table 2 shows that Model 1 predicts an additional 2 LRs for agencies mentioned by the Times at a standard deviation above the mean of that variable. This finding supports the hypothesis derived from the work of Gormly (1986, 1989) by Ringquist, Worsham, and Eisner (2003) that Congress will try to influence the bureaucracy s policy decisions to a greater degree in policy areas that the public views as salient. Ringquist, Worsham, and Eisner (2003) find that Congress both introduces and passes a greater number of new bills to reverse agency decisions when agencies preside over policy areas of salience to the public. This positive and significant coefficients for this variable, though not the main focus of the research presented in this paper, reinforces this finding. 12

17 One objection to the analysis presented above involves the ability to draw conclusions based on this small sample of agencies. What if the few agencies receiving very low grades happened to be burdened with relatively high numbers of LRs for idiosyncratic reasons and these reasons led to the significant relationships between performance and discretion observed in the models? If this is the case, a sample with more agencies, or an analysis including the entire population of federal agencies, would make it less likely that we would make a type I error and reject the null hypothesis that performance does not influence discretion when it is in fact true. To respond to this very real concern, we re-estimated models 1 and 2 without the one observation for which an agency received a D. In model 1, the coefficient for the Traditional Grade Scale variable remained positive and significant (p<.05). This finding held for Model 2, though the significance slipped to the.1 level. 12 In other words, it is not simply one observation driving the finding that Congress imposes a greater number LRs on agencies as the managerial performance of agencies decreases. 13 Conclusions The Federal Emergency Management Agency s (FEMA) bungling of the federal response to the devastation wrought by Hurricane Katrina, and the scurrying by members of Congress and President Bush to avoid the political fallout (e.g., VandeHei, 2005; Weisman & Goldstein, 2005), demonstrate that poor performance on the part of agencies managers can create substantial trouble for elected officials. 14 Additionally, formal theories of delegation stress that legislatures are willing to trade control over policy for the technical expertise agencies offer (Bawn, 1995; Epstein & O Halloran, 1999). It should be no surprise, then, that when agencies perform poorly that elected officials respond by stripping agency authority. 13

18 Yet, to our knowledge, no prior research has examined the relationship between agency capacity and discretion across a large sample of agencies. To be sure, research on agency termination emphasizes agency failure as a factor that increases the risk of termination (Carpenter, 2000; Carpenter & Lewis, 2004); however, this research does not show a relationship between these phenomena because of the lack of a measure of performance across the sample of agencies it examines. Using a unique dataset, we observe that lower levels of performance are indeed associated with the loss of discretion, controlling for the political environment, the salience of the agency, and the size of appropriations legislation through which lawmakers scale back discretion. These findings suggest the need to account for performance for a more complete understanding of why lawmakers grant discretion. Importantly, this emphasis on performance implies that lawmakers give agencies the incentive to make effective policies. If agencies want freedom to design the programs under their direction, as research on bureaucratic policymaking emphasizes (Carpenter, 2001), then convincing lawmakers of their effectiveness is one way to obtain this freedom. Empirical models of discretion without indicators of performance that show policy conflict reduces discretion communicate important features of inter-branch policymaking in an era of policy complexity. Nonetheless, empirical support for such models does not emphasize that lawmakers also promote performance. Our analysis, however, implies that lawmakers promote effective policy-making, a hopeful finding in an era in which many problems are complex and require bureaucratic expertise. As such, it is consistent with recent research stressing that Congress sometimes fosters effective policy-making by employing policy research (Esterling, 2004). 14

19 An additional implication of this study is that models of discretion that do not control for performance risk underestimating the influence of legislative-executive policy conflict on discretion. Consider one hypothetical case during unified government when legislative-executive conflict is low, and, all else equal, lawmakers are prone to grant discretion (Huber & Shipan, 2002; Epstein & O Halloran, 1999). Nevertheless, the agency that would implement a law under consideration has a poor record of performance, leading lawmakers to scale back the agency s discretion. In contrast, a second case occurs during divided government when lawmakers, all else equal, lean toward slashing discretion. In this case, however, the agency that would receive authority has a record of good performance, enticing lawmakers to take advantage of its technical capacity (Bawn, 1995) by providing more discretion than they would have if the agency had a mediocre reputation. Ignoring considerations about performance, theory would predict a relatively high level of discretion for the former observation and a relatively low level of discretion for the latter observation. However, considerations about performance attenuate this relationship. Any model using these observations to probe the relationship between legislativeexecutive conflict and discretion that did not control for performance, then, would underestimate the magnitude of the relationship between inter-branch policy conflict and discretion. Therefore, the findings of this research suggest that prior studies of discretion that do not control for performance, observe a weaker connection than exists. Of course, our findings are based on a relatively small number of agencies whose inclusion in the sample was based on their close interaction with the public. Do lawmakers limit discretion based on performance generally? Or is this relationship conditional on the nature of agencies interaction with the public? Future research on the link between capacity and discretion should focus on developing measures of performance across more, and different types 15

20 of, agencies to answer these questions. This is especially the case give that there is good reason to believe that many of the factors that influence the volume of discretion agencies receive are conditional. For example, Volden (2002) theorizes that inter-branch policy disagreement influences how much discretion agencies receive conditional on the existence of an executive veto. Similarly, Huber and Shipan (2002) theorize that the professional capacity of legislatures to craft effective policies conditions whether inter-branch disagreement affects how much discretion agencies receive and provide empirical support for this perspective. Additionally, although our findings support the hypothesis that (a lack of) capacity (reduces) increases the policy-making authority that agencies receive, there are several causal mechanisms that can account for this relationship. A goal of future research should be to explore the underlying cause. Do lawmakers cede (reduce) authority based on assessments of how likely the agency s actions are to lead to negative political fallout? Are lawmakers concerned intrinsically about the quality of policies created by agencies when providing authority? 16

21 Notes 1 Briefly, the criteria included how well managers defined and measured success, managed agency resources, took measures to ensure that managers were held accountable for decisions and performance, communicated effectively and honestly with stakeholders and political principals, ensured that employees possessed information necessary to perform tasks to achieve results, staffed individuals with appropriate skills for the tasks that they performed, possessed the physical infrastructure necessary to achieve the results, and provided sound fiscal management. 2 The FPP interviews included groups such as congressional oversight and appropriations committees, the GAO, the OMB, think tanks, the press, client and advocacy groups, academic institutions, and government commissions (Laurent, 1999). More detailed information about the FPP can be found on the project s website: Also see Ingraham, Joyce, and Donahue (2003). 3 Another source on agency performance is the Performance Assessment Rating Tool (PART) through which the Office of Management and Budget (OMB) evaluates the performance of federal programs. Using PART data would increase the size of the sample of agencies. However, given that the OMB works directly for the President within the Executive Office of the President (EOP), and given that presidents politicize the work of the EOP (Moe, 1985; Lewis, 2005), we believe that PART ratings are likely to be endogenous to the political priorities of the President, making them an invalid measurement of agency performance. 4 The grades assigned to agencies by the FPP can be found in Laurent (1999, 2000, 2001). 5 Measuring the volume of authority stripped from agencies in this way has an advantage. It is possible that agencies that possess high levels of discretion develop greater capacities than agencies that possess low levels of discretion. If this is the case, employing a dependent variable that measures the volume of discretion that agencies possess would preclude unbiased estimates of the influence of capacity on discretion since higher levels of discretion also cause high levels of capacity. However, our indicator measures the volume of authority taken away from an agency in year t + 1. Authority taken away in the future cannot affect agency capacity in the present. Additionally, existing research on the link between agency capacity and authority emphasizes that capacity leads to authority (Carpenter, 2001) rather than the other way around, suggesting that, at any rate, managerial capacity is not endogenous to the volume of discretion agencies possess. 17

22 6 All of the agencies graded by the FPP receive funding from these bills. We examined the versions of bills that became public law rather than House/Senate committee passed or House/Senate passed versions. 7 To count LRs, we first identified the bill in which the agency was funded. Next, we counted the number of instances stating that no funds or none of the funds could be spent for specific purposes that applied to the agency. LRs applyied to the agency if they were located in the specific section funding the agency or a general provisions section or title that applied to the agency. Using counts of the number of LRs in the year during which, and the year before, performance was graded did not changed the findings reported below. 8 This count was obtained through a Lexis/Nexis search of agency names, as they appear in Table A1, in the titles and lead paragraphs of New York Times stories in the year before the agency s managerial performance was graded. 9 Rinquist, Worsham, and Eisner (2003) also examine the relationship between complexity and congressional efforts to direct the bureaucracy, finding no direct relationship between complexity and such efforts even though complexity does condition the relationship between salience and such efforts. Since the primary focus of our analysis does not involve the relationship between policy type and discretion, since these authors found no direct relationship between complexity and such efforts, since there is no readily available measure of complexity across agencies, and since the degrees of freedom in our analysis is small to begin with, we opt not to control for complexity in our analysis. 10 Huber and Shipan (2002) identify factors, such as legislative capacity, that vary across legislatures that influence the level of discretion agencies receive. In the analysis below, we focus on cases in which a single legislature delegates authority. Therefore, these factors are constant and cannot explain variance in our dependent variable. 11 The note in Table 2 provides information on these baseline values. Predictions were calculated using Clarify 2.1 (Tomz, Wittenberg, & King, 2003). 12 These results are available from the authors upon request. 13 The results of this estimation are available from the authors upon request. 14 The title of VandeHei s article in The Washington Post illustrates the alacrity with which members of Congress and the President sought to avoid the political costs of FEMA s failure: Officials Deal with Political Fallout by Pointing Fingers. 18

23 References Arnold, R. D. (1990). The logic of congressional action. New Haven: Yale University Press. Bawn, K. (1995). Political control versus expertise: Congressional choices about administrative procedures. American Political Science Review, 89(March), Carpenter, D. P. (2000). Stochastic prediction and estimation of nonlinear political durations: An application to the lifetime of bureaus. In D. Richards (Ed.), Political complexity: Nonlinear models of politics ( ). Ann Arbor: University of Michigan Press. Carpenter, D. P. (2001). The forging of bureaucratic autonomy: Reputations, networks and policy innovation in executive agencies, Princeton: Princeton University Press. Carpenter, D. P., & Lewis, D. E. (2004). Political learning from rare events: Poisson inference, fiscal constraints, and the lifetime of bureaus. Political Analysis, 12 (Summer), Esterling, K. M. (2004). The political economy of expertise: Information and efficiency in American national institutions. Ann Arbor: University of Michigan Press. Epstein, D., & O Halloran, S. (1999). Delegating powers: A transaction cost politics approach to policy making under separate powers. New York: Cambridge University Press. Gormley, W. (1989). Taming the bureaucracy. Princeton: Princeton University Press. Gormley, W. (1986). Regulatory issue networks in a federal system. Polity, 18(4), Huber, J, D., & Shipan, C. R. (2002). Deliberate discretion: The institutional foundations of bureaucratic autonomy. New York: Cambridge University Press. Huber, J. D., Shipan, C. R., & Pfahler, M. (2001). Legislatures and statutory control of the bureaucracy. American Journal of Political Science, 45(April),

24 Ingraham, P., Joyce, P. K., & Donahue, A. K. (2003). Government performance: Why management matters. Baltimore: Johns Hopkins University Press. Jones, B. D., & Baumgartner F. R. (2005). The politics of attention: How governments prioritize problems. Chicago: University of Chicago Press. Laurent, A. (1999, February). Stacking up. Government Executive Laurent, A. (2000, March, 10). Measuring up. Government Executive. Laurent, A. (2001, April 1). Managing for results. Government Executive. Lewis, D. E. (2003). Presidents and the politics of agency design: Political insulation in the United States government bureaucracy, Stanford: Stanford University Press. Lewis, D. E. (2005). Staffing alone: Unilateral action and the politicization of the executive office of the president, Presidential Studies Quarterly, 35(Sept), MacDonald, J. A. (2007). Competing agents: Managing policy through competition in the U.S. House of Representatives. Kent State University, Typescript. McCubbins, M. D., Noll R. G., & Weingast, B. R. (1987). Administrative procedures as instruments of political control. Journal of Law, Economics, and Organization, 3, McCubbins, M. D., & Schwartz, T. (1984). Congressional oversight overlooked: Police patrols versus fire alarms. American Journal of Political Science, 2, Moe, T. M. (1989). The politics of bureaucratic structure. In J. E. Chubb & P. E. Peterson (Eds.), Can the government govern ( )? Washington, DC: The Brookings Institution. Moe, Terry M. (1985). The Politicized Presidency. In J. E. Chubb & P. E. Peterson (Eds.), New directions in American politics ( ). Washington, DC: Brookings. 20

25 Page, B. I. (1996). Who deliberates? Mass media in modern democracy. Chicago: University of Chicago Press. Potoski, M. (1999). Managing uncertainty through bureaucratic design: Administrative procedures and state air pollution control agencies. Journal of Public Administration Research and Theory, 9(October), Rinquist, E. J., Worsham, J., & Eisner, M. A. (2003). Salience, complexity, and the legislative direction of regulatory bureaucracies. Journal of Public Administration Research and Theory, 13(2), Tomz, M., Wittenberg, J. & King, G. (2003, January 5). Clarify: Software for interpreting and presenting statistical results (version 2.1). Stanford University, University of Wisconsin, and Harvard University. Retrieved from VandeHei, J. (2005, September 5). Officials deal with political fallout by pointing fingers. The Washington Post, p. A17. Volden, C. (2002). A formal model of the politics of delegation in a separation of powers system. American Journal of Political Science, 46(January), Weisman, J., & Goldstein, A. (2005, September 8). Bush requests $51.8 billion more for relief; GOP leaders launch inquiry on Katrina preparation and response. The Washington Post. pp. A1+. Wood, B. D., & Bohte, J. (2004). Political transaction costs and the politics of administrative design. Journal of Politics, 66(February),

26 Table 1. Poisson Regression Models of the Imposition of Limitation Riders on the FPP Sample of Agencies in the Year after Agency Performance Was Graded Independent Variables Model 1 Model 2 Model 3 Traditional Grade Scale -.191*** (.056) Plus/Minus Grade Scale -.060*** (.024) B.127 (.145) C.277* (.141) D.784*** (.230) Divided Government (.144) (.132) (.121) Unified & Hostile.298*.321*.358** Government (.141) (.179) (.152) No. of NY Times Stories.0010* * (0005) (.0007) (.0005) No. of Pages in Bill.005*.006*.004* (.002) (.003) (.002) Constant (.270) (.320) (.188) Log Likelihood Chi-Square 31.30*** 14.49*** 33.63*** N Notes: Coefficients are unstandardized. Standard errors are in parentheses. p <.10, * p <.05, **p <.01, ***p<.001 (one-tailed tests). The estimates for models 1 and 3 are poisson maximum likelihood estimates, since likelihood ratio tests did not reject the null hypothesis that the mean and variance of LRs were equal. The estimates for Model 2 are negative binomial maximum likelihood estimates, since the null hypothesis that the mean and variance of LRs were equal could be rejected (p<.05) and the alpha statistic was positive. Models 1 and 3 were also analyzed using the negative binomial estimator; the only change in the significance of the coefficients was that, for both models, the no. of NY Times stories variable was significant at the.1, rather than.05, level. 22

27 Table 2. The Predicted Number of Limitation Riders by Agency Characteristics (Model 1 Estimates) Agency Characteristics Predicted Values for Model 1: Traditional Grade Scale Performance A B (Baseline) C D Policy Disagreement with Agencies Agency Has Support in at Least One Branch (Baseline) Unified and Hostile Government No. of NY Times Stories Mean (Baseline) Std. Dev Notes: Predicted values were calculated using Clarify (Tomz, Wittenberg, and King 2003). The predicted values are compared to a hypothetical, or baseline model, where the variables are set to the modal or mean categories. The baseline was calculated using an agency operating under divided government that received a B, was created during a period when the presidency and the House were not controlled by a unified and hostile party, and was covered by the New York Times at the mean level for all agencies (58.63 stories). The bill in which the baseline agency was funded spanned pages. 23

28 Appendix Table A1. Federal Agencies Graded by the Federal Performance Project Agency Year of Grade Grade Coast Guard 2000 A National Weather Service 2001 A Social Security Administration 1999 A Postal Service 2001 A- Administration for Children and Families 2001 B Army Corps of Engineers 2000 B Federal Emergency Management Agency 1999 B Food and Drug Administration 1999 B Food and Nutrition Service 1999 B Food Safety and Inspection Service 1999 B NASA 2001 B Veterans Health Administration 1999 B Environmental Protection Agency 1999 B- Federal Housing Administration 1999 B- Occupational Safety and Health Administration 1999 B- Patent and Trademark Office 1999 B- Veterans Benefits Administration 2000 B- Bureau of Consular Affairs 2001 C Customs Service 1999 C Federal Aviation Administration 1999 C Forest Service 2001 C Health Care Financing Administration 1999 C Internal Revenue Service 1999 C National Park Service 2000 C Office of Student Financial Assistance 2000 C Immigration and Naturalization Service 1999 C- Bureau of Indian Affairs 2001 D 24

29 Table A2. Descriptive Statistics Variable Minimum Maximum Mean Std. Deviation Limitation Riders Grade Dummies: B C D Traditional Grade Scale Plus/Minus Grade Scale Divided Government Unified & Hostile Government No. of NY Times Stories No. of Pages in Bill