This dissertation argues that early land reforms had consequential long-run impacts on the development of the American state. The state-building role of land reform, frequently discussed in the context of comparative political economy, is often overlooked in the study of American political development. I provide evidence that mid-nineteenth century homestead acts had significant long-run impacts that can help explain contemporary differences in state capacity. I find that homestead policy --- or the homestead entries authorized by homestead policy --- had significant and negative effects on the size of state governments lasting a century after passage of the 1862 Homestead Act. This finding implies that in the absence of homestead policy, the average size of frontier state governments would have exceeded the average size of governments in states that were not directly affected by the policy. Potential causal mechanisms include land inequality, which I show to be positively correlated with state capacity, especially at higher levels of inequality. This empirical relationship is consistent with median voter theories of inequality and redistribution.

The dissertation contributes to a new generation of data-driven machine learning methods for estimating causal impacts of policy interventions on panel data. Machine learning approaches are capable of automatically choosing appropriate predictors without relying on pre-intervention covariates, limiting the 'researcher degrees of freedom' that arise from model specification choices. The proposed method based on recurrent neural networks can learn nonconvex combinations of predictors, are specifically structured for sequential data, and outperform regression-based estimators in high-dimensional data settings. The dissertation also introduces a matrix completion method that is adaptable to settings with staggered treatment adoption and outperforms other regression-based estimators in small-dimensional data settings.

Political scientists typically view unilateral action as the president “going it alone” in opposition to Congress. However, implementation of unilateral action relies on the cooperation of administrative agencies. In this dissertation, comprising three papers, I start with this premise that unilateral actions should be considered acts of delegation. First, I conceptualize delegation and discretion for the executive branch context and provide novel measures of those concepts to facilitate empirical research. I then turn to applying these measures to empirical questions, studying both potential determinants of variation in delegation and discretion, and consequences of this variation for inter-branch politics.

The first paper, “The Administrative Politics of Unilateral Action: Measuring Delegation and Discretion in the Executive Branch,” introduces novel measures of executive-branch delegation— the extent of agents’ authorization for action—and discretion—the flexibility offered to agents in defining how to use that authority. The paper provides evidence of the reliability and validity of these measures, which are substantive and rely on textual analysis rather than structural proxies. The main contribution of the paper is the introduction of the Administrative Delegation Dataset, which provides scores for 1,641 presidential unilateral directives; including executive orders, memos, and proclamations; issued between 1936 and 2021. To my knowledge this is the first such measure of executive-branch delegation, and it promises to be a useful tool for researchers of executive-branch politics.

In the second paper, “Delegation by Executive Order: Discretionary Windows in Presidential Unilateralism,” I leverage these new scores to consider the president’s ability to constrain bureaucratic behavior in the context of executive orders. In particular, I examine the relationship between the role of agencies in the formulation of an executive order, and the order’s associated discretionary window. The paper uses data from Rudalevige (2021), identifying the origins of more than 500 executive orders as either emanating from within the White House, or originating in executive branch agencies. Using my delegation and discretion scores, I find that agency development of an order is associated with a narrower grant of discretion relative to development by White House staff, even when controlling for order significance and extent of delegated authority. Moreover, this relationship does not seem to have changed over time, and does not appear to vary by party. This finding sheds new light on the use of discretion within the executive branch.

In the final paper, “Participatory Unilateralism: Understanding Congress’s Role in Presidential Unilateral Policymaking,” I advance my principal argument: that Congressional committees can use oversight to shape the implementation of executive orders by administrative agencies. Using a dataset of 1,420 executive orders issued between 1970 and 2021, as well as all Congressional committee hearings pertaining to oversight of executive orders conducted over the date range (n = 362), I find that higher delegation and discretion scores are associated with more Congressional hearing activity. This relationship is generally not conditioned by partisan division but persists during periods of both unified and divided government. The implication is that when agents are given the authority and latitude to execute unilateral directives, Congress has an opportunity to shape the implementation of those orders through committee oversight.

This thesis contributes to our understanding of government corruption and the transparency laws designed to prevent and expose it. Specifically, it analyzes the political conditions under which we expect to see transparency laws emerge, the strength of the laws we expect under different mixes political power, and the economic and political effects that result from transparency laws. It analyzes three cases of transparency laws: whistleblower laws; bounty laws, in particular the false claims act; and campaign finance transparency laws. These laws vary in terms of the burden and risk assumed by individuals availing themselves of the information provided by the laws. It finds, using quantitative methods supplemented by expert interviews, that the high burden, high risk laws are passed for politically predictable reasons, though they are less likely to have an actual effect in terms of political or economic outcomes. Low burden, low risk laws, however, have a clearer impact.

In addition to using large datasets on political advertising, county and local bond ratings and yields, and campaign finance disclosure laws, the project contributes original data on state-level whistleblower laws and local-level government corruption, as well as providing an update to existing data on state-level political competition.

The U.S. District Courts resolve the vast majority of cases in the U.S. federal legal system, but we know much less about them than we do about the U.S. Courts of Appeals or the U.S. Supreme Court. The disproportionate focus on appellate courts has reinforced the perception that courts are primarily policymaking institutions. However, district courts rarely make new law, and are mostly at the front line of law enforcement. In the aggregate, the quality of their decisions on individual cases determines the degree to which laws are applied fairly and consistently. However, their work is not trivial, and the quality of outcomes depends on whether district judges have sufficient incentives to produce high quality decisions. Indeed, district courts are the primary fact-finders in legal disputes. Judges often have to cajole litigants, witnesses and attorneys into being forthcoming about what they know. They have to rule on a flurry of motions about what information can be discovered or admitted as evidence. District judges are therefore intricately involved in creating the case records, which means that they must engage in some costly effort to figure out the facts of cases they oversee.

A district judge's work on any given case constitutes a process of "learning through litigation." The importance of this process for the enforcement of laws is the point of departure for this dissertation. In a series of three papers, I analyze the agency problem confronting appellate courts: how can district judges be induced to undertake maximum effort to learn the facts and the law pertaining to the legal cases before them when this effort is costly, but imperfectly observed by principals in the appellate courts? Moreover, what are the consequences for litigants (and society more broadly) of district court institutions and practices that serve to induce effort and compliance by district judges?

To address these questions, I analyze three original models of district court decision making. The models present three main takeaways: (1) due to their informational advantage, deferring to district judges is often optimal for appellate courts, even when those judges are biased; (2) dispassionate judges are often harder to motivate to exert effort to resolve cases than either activist or biased judges; (3) endogenous appeals by litigants can provide perverse incentives to unbiased "legalist" judges to tilt judgments toward powerful litigants. Moreover, the models I study in these papers offer observationally equivalent explanations for commonly observed empirical phenomena, such as compliance with legal doctrine and in-group bias. Taken together, the models underscore the important role that courts play in enforcing laws effectively and accurately. Indeed, if there is significant variation in district judges' effort to resolve cases, then laws will be enforced inconsistently. Perhaps more insidiously, the results suggest that low quality decision making by district judges may affect different kinds of litigants differently.

Public-employee pensions have grown significantly in cost in recent years, and will continue to be one of the central policy problems that state and local governments confront in the 21st Century. Pensions show United States' federalism at work: they vary tremendously across and within states. I examine the causes and consequences of this variation. Specifically, I ask how democratically-elected politicians can commit to pension performance, and what are consequences do pensions and politicians' decisions have?

This dissertation presents three separate empirical papers using data from 103 plans across all 50 states from 2001-2011. I also include introductory, transitional, and concluding sections. As discussed both here and in prior literature, politicians have clear incentives to compensate employees through pensions, while disguising expenses and skipping payments. This has led to significant costs and increased attention to pensions.

The first paper focuses on the factors leading to variation in pensions' financial performance, or funding ratios. Centrally, I find that as management board personnel is increasingly insulated from political control, performance improves. The second paper focuses on politicians' contributions into funds. There, I uncover evidence that politicians prefer to commit to pensions covering police and fire employees. They also tend to skip payments when prior economic performance and investment returns are higher. This shows that elected officials prefer to keep taxes low when possible, even at the cost of underfunding pensions.

The third paper pivots to examine a consequence of pensions: how they affect public-employee retirement rates. I present a simple theoretical model showing how reductions in commitments from employers might make public-employees less likely to believe they will receive their full pension benefits, and in turn, more likely to retire. I then present empirical evidence that as pensions are more controlled by politicians, retirements increase. Poorer funding from the prior year also is associated with increased retirements, providing evidence that performance generates feedback effects for human capital. Jointly, my papers provide broad and novel insight into the causes and consequences of variation in public pensions.

Using methods of causal inference, computational social science and careful qualitative analysis, this thesis examines the roles that race and gender play in three key areas of modern American political life: political polarization, immigration policy and political participation.

In the first essay entitled "The Big Sort(s): Diversity, White Flight and Polarization in Neighborhoods and Cities,'' I develop the Migration-Flight-Polarization (MFP) hypothesis to explain how changes in diversity brought about by internal migration and immigration hold the key to understanding the connection between residential choice decisions and geographic polarization along partisan and ideological lines. Using an original agent-based modeling simulation and Hurricane Katrina evacuee data collected from schools and neighborhoods in Houston, Texas, I demonstrate that changes in diversity and "white flight'' responses to these changes are responsible for the growing partisan divide in Houston neighborhoods and the City of Houston as a whole.

My second essay entitled "Not in My Backyard: The Effect of Immigrant Race and Proximity on Immigration Policy Preferences,'' examines the extent to which immigrant race and proximity to a respondent informs immigration policy opinion. Using a survey experiment which employs blurry images of a fictional undocumented Mexican immigrant and respondent Internet Protocol addresses, I randomly manipulate immigrant skin tone and perceived distance between respondents and the immigrant. I find that the effect of race on immigration policy opinion depends upon the perceived distance between the immigrant and respondents. When respondents believe that the immigrant lives nearby, the darker immigrant elicits more anti-immigration responses to immigration policy questions. Conversely, when no immigrant location is provided, the darker immigrant elicits greater pro-immigration responses to the same questions. I also find that attitude polarization on immigration policy increases when respondents believe that the immigrant lives near them. These findings help explain the paradoxical divide between support for pro-immigration policies at the national level and anti-immigration policies at the state level.

My third essay with Morris Levy entitled "Estimating the Gender Penalty in the House: A Regression Discontinuity Approach,'' brings a novel regression discontinuity design to bear on the question of whether net voter bias against female candidates for office can help explain the limited growth of female election to the House of Representatives. Using house primary vote share as a forcing variable, we estimate the causal effect of a major party nominee's gender on that candidate's general election vote share. Our period of study encompasses all Congressional elections since 1982. Our findings suggest that female Republican candidates that barely win two-person House primaries against males receive a substantial boost in general election vote share. A similar effect among female winners of close Democratic primaries is not found.

I argue that understanding executive action requires examining the conditions that precede and motivate that action. In a series of three papers, I explore those conditions. The first paper studies how Congressional anticipation of presidential unilateralism conditions the production of legislation. If members of Congress can anticipate unilateral action, their failure to legislate cannot be explained by ``gridlock intervals'' in a standard spatial model. I argue instead that they may willingly surrender authority to the president to head off potential attacks from voters or interest groups. This helps to explain the president's accumulation of authority over time. More broadly, I argue that just as a large literature has examined outside pressure on Congress in isolation, we should examine its influence in the presence of the president's unilateral powers.

The second paper examines credible political communication, which may include symbolic executive orders. I present a cheap talk model in which a politician (sender) is aligned with one of two opposed groups (receivers) and seeks to communicate her preferences to win support. An increase in the weaker group's capacity may enable credible communication by the opposed type of politician, ironically making the weaker group worse-off. I illustrate the model with the case of the 2016 presidential campaign in the United States, in which Donald Trump credibly communicated alignment with anti-immigration groups through harsh messaging against immigrants, whose power was increasing. More broadly, the model and case show how shifts in relative group power can enable politicians to use messaging to assemble novel political coalitions.

The third paper (coauthored with Joseph Warren) presents a novel explanation for the origins of bureaucratic agencies, which are an important venue for executive action. Early support for expert policymaking through administrative agencies was rooted in concerns over political power. We argue that potential policy feedback effects made an anti-business coalition between liberals and populists in the late 19th-century United States unachievable. However, agencies diminished feedback effects and facilitated a successful progressive-liberal coalition. The strategic dilemma created by a changing distribution of power thus explains the development of broad political support for bureaucratic agencies.

Paper 1: While scholars have identified ways that racial conservatives exerted out-sized influence on criminal justice policies, little attention has been paid to whether police departments have incentives to learn about and adopt reforms that reduce racial disparity.

I present a game of imperfect information between residents and a municipal police chief to show that a chief's inability to prevent officer behavior that residents perceive to be abusive, coupled with resident unwillingness to assist police in the aftermath of this behavior, creates an incentive for the chief to choose and learn about new policing strategies that rely less on resident assistance. This induces a bias in favor of aggressive over collaborative tactics in the police chief’s policy selection and learning decisions.

Segregation and discrimination ensured that many Black Americans lived in conditions that produced this result in the later twentieth century; thus the model shows structural racism in local police policy making.

Paper 2: Scholarship on racial inequality in policing has largely focused on factors that would cause individual officers to rely on race when deciding whether to make an arrest. Extant explanations suggest that if officers chose not to discriminate and managed to eliminate the influence of stereotypes and implicit associations on their behavior, any remaining racial disparity would be statistical discrimination. I identify an additional source of officer bias: strategic limits on information transmission. I show that when dedicated officers are uncertain whether their police chief is independent of political pressures, a chief who cares about crime control cannot credibly communicate the reason for their policy choices. Therefore if the chief is better informed than the officers about what arrest intensities would be optimal for reducing crime, chief policy choice will give officers exaggerated or understated posterior beliefs about the probability that individuals within a particular group should be arrested. Under certain conditions this would lead to the endogenous development of taste-based discrimination.

Paper 3: This paper examines the conditions under which municipal police chiefs in the latter half of the 20th century would have adopted policies that reduced the rate of arrests among Black residents. I use a two player game of incomplete information, in which the commonly valued outcome depends upon non-contractible effort from a less informed subordinate, to show that officer uncertainty about their chief's competence and skepticism of policies that reduce rates of arrest among Black people could have prevented chiefs from adopting such policies in equilibrium. This conservatism arises even when both chief and officer are not individually racist and the chief knows that the new policy would be more effective at controlling crime.

Given gridlock in Congress, presidents and subunits of Congress are increasingly pursuing their policy goals in the executive branch. Yet we know little about how they go about doing so. In Chapter 1, I introduce the broad notion of positive agenda control by the president. Up until this point, the existing literature has been narrowly focused on presidents' negative agenda control over agencies (Moe 1985; Nathan 1983) and has only recognized positive agenda control in terms of executive orders (Moe and Howell 1999; Howell 2003). I argue that presidents direct their resources toward allied agencies to strengthen, pass, and implement policies the president supports. Using a newly collected dataset, I look at the Office of Information and Regulatory Affairs, the most important innovation in presidential power in 35 years, and I show that Presidents Clinton and Obama were disproportionately likely to use OIRA to review the policies of relatively more liberal agencies and President George W. Bush was disproportionately likely to use OIRA to review the policies of relatively more conservative agencies. This finding differs from the conventional wisdom that OIRA is a presidential watchdog and it changes how we think about presidential power in administrative oversight.

In Chapter 2, together with Abby Wood and Sean Gailmard, I turn to congressional committees and their pursuit of policy goals in the executive branch. We introduce the theory that congressional committees use oversight of bureaucracy as a means to mitigate agency problems they face with the bureaucracy. Up until this point, scholars have not systematically connected the oversight activity of Congress to the democratic problems created by delegation. The landmark analysis of Aberbach 1990 placed oversight in a framework centered on Congress and the broad contours of the national policymaking environment, without taking into account agency policymaking activity. In contrast, we present a principal-agent model of oversight and find support in a newly-collected dataset of on-the-record legislative hearings for the model's three observable implications: legislative oversight is increasing in agency policymaking activity, increasing in the amount of ideological conflict between committees and agencies, and increasing in the amount of ideological conflict between House and Senate committees during divided Congresses. Our results provide evidence that committees oversee agencies not merely to position-take for the next campaign cycle, but to influence actual policy outcomes.

In Chapter 3, I return to a focus on the president and I treat limiting lobbyist influence as a policy goal in its own right. I argue that lobbyists going "offensive" in their lobbying strategies combined with underwriting some of the work of under-resourced agencies puts lobbyists in a position to set the agenda in the executive branch. I contribute a bargaining model in which an agency, the president, and a lobbyist each have agenda-setting power some of the time. I derive the conditions under which the president, so long as he or she is not perfectly ideologically aligned with the agency, can limit the influence of the lobbyist on the agency's policies by acting as a "tough" bargaining agent. This model reveals a tradeoff from the standpoint of the agency between working with an ideologically aligned president (an ally on policymaking goals) and working with an ideologically unaligned president who can help limit lobbyist influence.

If there is no expert but bureaucracy in a policy area, expertise may be sufficient for an agency to acquire its discretion. However, it is hard for an agency as an expert to have monopolistic status in any policy area, due to the growth of interest groups. Delegation in recent decades depends not only on its expertise, but also on the other bureaucratic capability to affect the behaviors of the outside experts. Even the most expert agencies cannot maintain their discretionary authorities without sufficient political capacities to affect the behaviors of interest groups. In particular, "brokerage capacity" that allows agencies to link interest groups and to resolve their interest conflicts, has become important in bureaucratic discretion. By mediating conflicts of interest and minimizing unnecessary contingencies, agencies are able to reduce interest groups' incentives to provide information to the legislature, thereby indirectly affecting legislators' decisions on delegation and oversight.