Chapter 2 reviews the bureaucratic political control literature across both political science and public administration. It discusses the gaps in the existing literature and reviews the overall research question this study seeks to address. It details the central assertion in the study that the variation in political control of the bureaucracy may also be a function of internal characteristics of agencies themselves; specifically, differences in the bureaucracy’s workforce organization, composition, and responsibilities. These differences, I argue, lead to variation across the bureaucracy in how well different principals can control bureaucratic activity. For the purpose of this analysis, I am defining political control as “the extent that external political actors can control internal bureaucratic activity.”1 I propose a modification of the traditional principal–agent theory, which I call the agent-principal model to examine bureaucratic control from an agency-centered approach.
In working to overcome the problem of uniform sampling in the bureaucratic literature, I utilize two common metrics every single agency contains: a budget and workforce counts. Budgeting is discussed in chapter 3 and workforce counts in chapter 4. Budgeting is a shared tool between the president and the Congress, and agency budgets are appropriated on an annual basis. Typically, budgets are not assigned anew each year; rather, policy-makers use the current year’s budget as a starting point for next year’s budget plus or minus some small interval (Fiorina 1989). Chapter 3 discusses the incremental budgeting literature developed by Aaron Wildavsky, among others. It details the decision rules that policy-makers use when appropriating agency budgets every year, and some of the inherent problems in budgeting.
In chapter 4, I propose a relaxation of central assumptions in traditional principal–agent theory to examine bureaucratic control from an agency-centered orientation. It is a bottom-up perspective to study the bureaucracy examining how differences in the internal characteristics of different agencies, such as variation in workforce organization and composition, may affect the degree of political control. First, I argue that agencies are not organized uniformly. While all agencies are, to some degree, hierarchical, there is a range within hierarchical organizations whereby many employees rely more on their coworkers, rather than supervisors, for their job directives (Brehm and Gates 2008). Some organizations, in contrast, may rest on an organizational continuum from strict hierarchies to more heterarchical (flat) forms. The structure of the organization will have an effect on different principals’ political control because of inherent differences in the workforce organization for different agencies (Kennedy 2014; Bolton 2015: Selin 2015; Hollibaugh, Horton, and Lewis 2014). Second, the bureaucracy’s workforce composition is divided between political appointees and career bureaucrats who jointly run the agency. I argue that the incentive structures for careerists and political appointees are inherently different, which may account for variation in political control across different principals (Lewis 2009). Lastly, agencies have different workforce responsibilities across the federal bureaucracy. Many of them are transaction oriented, meaning that they are a processing center for checks or clearing houses for grants; however, others make and implement policy. Therefore, I expect that political control may operate differently across contexts.
The central thesis in this study is that political control is contingent on characteristics of the bureaucratic agency itself. Broadly, I expect the president and Congress to exert greater control when there is a higher concentration of political appointees, a more hierarchical organization, and when the agency is more aligned to policy production. Conversely, I expect an absence of political control when there are few to no political appointees in the agency’s workforce, the structure is flatter, and the agency performs either technically advanced or mundane tasks.
These hypotheses are evaluated in chapters 5 and 6 which are empirical studies of the theoretical propositions developed in the previous chapters and they also assess the agent-principal model discussed in chapter 4. Specifically, chapter 5 provides a broad test for the agent-principal model and works to overcome all three problems afflicting the bureaucratic control literature: the problem of uniform sampling, the problem of using case studies to make generalizations, and the problem of an overreliance on the principal-agent framework. This chapter utilizes a unique data set of 1,921 observations across 139 different bureaus in the U.S. federal bureaucracy from 2000 to 2014. They range from smaller bureaus with limited budgets such as the American Battle Monuments Commission to larger bureaus with more expansive budgets and workforce complements such as the Employment and Training Administration within the Department of Labor. Utilizing a time-series cross-sectional approach for the pooled panel data, the analysis spans a fifteen-year period. The main findings are that both the president and the Congress control the bureaucracy, but through different mechanisms. Second, internal characteristics of the bureaus affect how political principals control bureaus. This suggests that the nature of control operates differently across institutional settings and that internal agency characteristics condition how principals interact with bureaus. There is substantial support for the agent-principal model developed in chapter 4 which implies that further areas of inquiry into the bureaucracy’s environment are warranted beyond those few scholars who have examined agencies’ organizational elements as a means to explain political control (see Bolton et al. 2015; Kennedy 2014; Selin 2015; Lowande 2019).
Next, chapter 6 takes a more traditional approach in its research design in using a case-study methodology on the dynamics of education policy change in the United States over a sixty-five-year period. This is a more typical format (case studies) for studying a bureau in taking a deep look into a specific area and then teasing out more general trends as they apply more broadly to the bureaucracy. Chapter 6 asks a similar question to the rest of the book: Which political actor is driving [educational] policy change in the United States? Do the internal characteristics of the Department of Education condition how the principals interact with the department? Also using a cross-sectional time-series methodological analysis, the main findings from this chapter are that education policy is elite-driven but that internal characteristics of the department, specifically variation in workforce responsibilities, workforce composition, and workforce organization, can influence the direction of public policy, but it varies by party. This case study also finds support for the agent-principal model developed in chapter 3.
These two different empirical studies in chapters 5 and 6 are designed differently in order to provide insights into the nature of control. In chapter 5, I specifically move away from the case-study approach that has largely characterized the literature in order to make generalizations across the bureaucracy (Coppedge 1999). Using a cross-sectional approach across 139 different bureaus provides a broad view of bureaucratic activity for Congress and the president to manage. The results from chapter 5 indicate that there are differences in how political principals exert influence over the bureaucracy not only across dramatically different bureaus but also across time (Geddes 2004). This suggests that, ceteris paribus, the presidency and the Congress impact the bureaucracy through different mechanisms and if internal characteristics of bureaus dramatically change, the nature of influence may as well. The large-N study in chapter 5 enables generalization for the theoretical approach and overall findings. It also has higher external validity (King, Keohane, and Verba 1994). The shortfalls of this approach are that it assumes homogeneity of the data (Brady et al. 2004) by comparing disparate bureaus in the government and it also loses contextual knowledge gained from a specific policy areas (Brady et al. 2004). This chapter was specifically designed to confront the myriad challenges of studying the bureaucracy that has beset scholars.
By contrast, the in-depth case study in chapter 6 remedies the shortfalls of the large-N approach. It provides a case study on specific policy domain, education. This design provides a context to understand the theoretical approach advanced in the agent-principal model (Rogowski 2004) where we can gain more understanding about the political environment affecting a specific policy question (Fearron and Laitin 2009; Geddes 2003). This chapter helps gain traction on the question about the nature of policy change and which actors in the political ecosystem are more or less influential in initiating change.