Bureaucratic Investments in Expertise:
Evidence From a Randomized Controlled Field Trial
Accepted version
Short title for running header: Bureaucratic Investments in Expertise
Simon Calmar Andersen, Aarhus University
Donald P Moynihan, University of Wisconsin-Madison
Simon Calmar Andersen () is Professor at the Department of Political Science at Aarhus University, Denmark.
Donald P Moynihan () is Professor of Public Affairs at the La Follette School of Public Affairs at University of Wisconsin-Madison, WI, 53706.
Abstract
How can elected officials induce bureaucrats to invest in acquiring the expertise necessary to provide high-quality public services?To address this question, we test and extend aspects of Gailmard and Patty’s (2007, 2012) expertise model in the context of contemporary governance using a unique randomized controlled field experiment of school principals in Denmark. Consistent with the expertise model, we find that bureaucratic agents randomly assigned greater discretion in the allocation of personnel resources were more likely to acquire information on school performance. We extend the model in two ways. First, we show thatdiscretion effects are stronger when the information available aligns with bureaucraticgoal preferences.Second, we showthat institutionaldesign choices that improve the relative benefits of the information increase information acquisition.
Keywords: bureaucracy, expertise, performance, discretion, administration
Data and supporting materials necessary to reproduce the numerical results in the article are available in the JOP Dataverse ( .harvard.edu/dataverse/jop). Supplementary material for this article is available in the appendix in the online edition. Studies were conducted in compliance with relevant institutional guidelines. Support for this research was provided by the Graduate School at the University of Wisconsin-Madison, the Department of Political Science at Aarhus University and The Ministry of Education in Denmark.
Introduction
The quality of public services rests uponbureaucraticexpertise. In contemporary times governments have invested significant resources intended to foster bureaucratic acquisition of expertise towardmeasurable performance outcomes (Heinrich 2007; Moynihan 2008). While the political consequences of performance regimeshave been considered (Berry and Howell 2007; Bertelli and John 2010; Boyne, James, John and Petrovsky 2009; Kogan, Lavertu and Peskowitz 2015), elected officials struggle to encourage their use among target bureaucratic users.For example, overcoming bureaucratic resistance to seeking out and making use of information is a frequently cited problem in limiting the spread of evidence-based practice in medicine (Aarons, Hurlburt, and Horwitz 2011), environmental policy (Hall and Jennings 2011), and social policy (Boaz and Pawson 2005).
The creation of performance regimes, or parallel initiatives such as evidence-based policy movements, ask bureaucrats to amend traditional and well-settled ways of approaching policy decisions by seeking out data in new ways, with the knowledge that such data generally adds a layer of complexity to decision processes (Nathan 2005; Heinrich 2007). Reforms have floundered as bureaucrats have shown a limited, at best, willingness to engage in such behavioral change, and instead see performance regimes as burdensome distractions from their work (Lavertu, Lewis and Moynihan 2013) or coercive incentive structures to be gamed (Figlio and Getzler 2006; Heinrich and Marschke 2011; Jacob and Levitt 2003).
So, how do political principals encourage bureaucratic investments to acquire expertise?And what can political science theories contribute to solving the very practical problem of how to convert expensive performance regimes into governance toolsthat actually improve expertise? A number of theories propose that bureaucratic expertise is not a fixed stock, and can be encouraged by institutional design, but such investments come at the cost of delegating more authority to agents (Bawn 1995;Gilligan and Krehbiel 1987).This tradeoff between delegation and expertise is embedded in Gailmard and Patty’s (2007, 2012) model of bureaucratic investments in expertise (hereafter the expertise model). The model proposes that the interests of principals are best served if agents continually accrue expertise that improves policy outcomes. But principals are compelled to provide job security and discretion to give agents confidence they can make use of expertise to advance their own policy preferences. In short, better outcomes require loosening the reins.
In the next section we summarize the expertise model, and consider its relevance for contemporary governance. The practical difficulties of examining the effects of discretion on expertise acquisition have resulted in no systematic empirical testof the central causal claim of the model. Gailmard and Patty’s (2012) historical account of the administrative development and maintenance of expertise is intended to be an illustrative “proof of concept” (201). Performance regimesoffer a promising contemporary setting containing, to varying degrees, essential elements of the expertise model: they formalize policy-relevant facts, and sometimes delegate administrative autonomy to motivate agents to acquire and respond to that information.
To test the central component of the expertise model – the effect of granting discretion on bureaucrats’ active acquisition of policy information – weemploy a field experiment.The Danish government granted a group of school principals up to $25,000 to support teachers in the classroombut delegated varying levels of discretion in terms of how to administer the grant. While any specific setting provides limitations to external generalizability, the reliance on field experimental data reveals actual behavior by bureaucrats related to their substantive tasks, increasing the external validity of the study relative to research designs that rely on self-reported or intended actions provided by surveys or artificially-induced behavior associated with laboratory experiments.
Information acquisitionis measured by whether schoolprincipals downloaded online performance data on student wellbeing in their schools. For treatments, principals were randomly allocated different levels of discretion and certain types of performance information to allow us to test three hypotheses. The first hypothesis, directly derived from the expertise model, suggests that grants of administrative discretion will increase efforts to acquire expertise. The second hypothesis expands the model by proposing that the effect of discretion will be moderated by the preferences of bureaucrats between different outcomes in the same policy area.We find that not only are delegations of discretion associated with greater acquisition of performance data, but that this effect is indeed higher for school principals whoprioritize student wellbeing over other policy preferences. Third, we find thatprincipals can make investments in expertise more attractive by providing agents with tools that alter the relative costs and benefits of information acquisition. Specifically, if school principals were offered comparative performance data that helped to contextualize the performance data for their own school, they were likely to acquire more data.
The Expertise Model
The expertise model assumes that expertise is endogenous to the governance regime rather than a fixed stock of human capital that agents bring with them. It is also defined as relationship-specific, valued inside the bureaucracy at a higher level than in the general labor market. The development of expertise is costly, the result of active efforts by the agent.The potential for agents to make such investments therefore depends upon the combination of the underlying agent preferences and the incentives created by principals (Gailmard and Patty 2007, 2012).
In the expertise model heterogeneity of agent preferences creates two types of bureaucrats:zealots who care about policy outcomes and are therefore naturally inclined to increase their expertise, and slackers indifferent to policy outcomes and reluctant to expend costly investments. Principals can induce investments in expertise by creating conditions that increase zealot’s confidence that they can actually use the knowledge they develop to affect policy outcomes. The two primary ways in which principals can do this is by offering tenure, thereby giving a long-term time horizon for agents to apply their expertise, and by providing grants of discretion to agents so that they may have greater influence on policy outcomes. Such conditions will be less attractive to slackers, who will then tend to exit out of the public sector, due to outside payments assumed to be higher than government wages.
In coming to these conclusions, the expertise model joins other work (e.g., Brehm and Gates 1999; Miller and Whitford 2007) in amending some core tenets of principal agent theoryby arguing for the virtues of bureaucratic protections and discretion, and for suggesting that goal misalignment between principal and agent brings some benefits. The expertise model also conveys upon the principal the capacity to shape agent investment in expertise.
While the expertise model has been widely cited, there has been limited effort to test the basic claims inherent in the formal model. The expertise model predicts that principals will grant less discretion when agents share principal goals, or when agents deviate too greatly from principal goals. Consistent with this claim, Clinton et al. (2011) show a non-linear relationship between ideological distance between principals and federal agencies (using ideal point estimates) and the provision of discretion to agents (measured by the length of public laws). Consistent with the expertise model, agency-specific human capital and delegation of discretion to agency leaders is associated with turnover intention among US federal employees (Bertelli and Lewis 2012).
There has not been an effort to test what Gailmard and Patty (2007, 880) characterize as “the central theoretical concept” of their model, “the endogenous acquisition of expertise within a bureaucratic agency.” To test if grants of discretion result in greater investment in expertise we must address what expertise actually means.
Thus far, we have three basic parameters for expertise: it is relationship-specific, requires costly investments, and characterized by endogenous information. The most detailed definition of expertise offered by Gailmard and Patty (2012, 32)informs a fuller conceptualization: “The notion of expertise we adopt rests on the presumption that most individuals' preferences about policy decisions are induced by the “state of the world” prevailing in society at the time (and, perhaps, those expected to prevail in the future)…We define policymaking expertise as the knowledge of facts of this type. In other words, expertise is the ability to choose the government policy that best suits the prevailing conditions in the real world to achieve a particular policy outcome. In our theory, bureaucrats “acquire expertise" only if they invest individual effort in obtaining it, and this investment gives them knowledge of the facts whenever they are asked to make policy decisions in the future.” This conceptualization of expertise above is deliberately broad, not easilycaptured by any single measure but instead invites different operationalizations in empirical work.
As the definition above makes clear, expertise includes “knowledge” of policy-relevant “facts” that allow bureaucrats to make choices to achieve “particular policy outcomes.” A general and fixed stock of knowledge is not enough for two main reasons. The first is temporal change. The “state of the world” is not fixed. Underlying conditions change, as does the range of policy innovations to address these conditions.Knowledge of policy-relevant facts must evolve to reflect “prevailing conditions in the real world”,requiring an ongoing need to invest in expertise. The second is the specificity of policy-relevant facts. Even a general expertise needs actionable and detailed knowledge of how different policy choices and implementation processes function: how specific organizational changes, decisions, innovations, programs or resources affect “particular policy outcomes.”For many policy issues, what shapes policy outcomes are empirical questions. A general expertise about a policy area is made more relevant for affecting policy outcomes when combined with empirical knowledge about the “state of the world.”
For our purposes, expertise includes the ongoing accumulation of the policy-relevant information. If two bureaucrats have the same general training or education on a topic, the one that has acquired policy-relevant informationon the answer to empirical questions has more expertise.
The expertise model specifies that investments in expertise should be costly to the bureaucrat. If it is not costly, all bureaucrats are expected to acquire it because it would, all else being equal, enhance their chances of influencing policy outcomes. What does a costly investment mean in practice? Gailmard and Patty do not offer many specific examplesbeyond “education and/or on the job training” (2012, 132).What is common to both examples is that they require the agent to invest time and effort. A willingness to acquire and make use of policy-relevant facts on an ongoing basis is costly to the agent, at least in contrast to settling for a less demanding life of not seeking out such information. While acquiring a single piece of information may require little effort, the cumulative effort to seek out such information is a real cost (Wilson 1999). As we describe in the next section, governments have made institutional design choices to encourage such information acquisition, but have been frustrated by bureaucratic unwillingness to adjust behavior in response to these design choices.
We do not contend that information acquisition is the only parameter of expertise. Studies of decision-making suggest that expertise also includes how individuals conceptualize and solve complex tasks (Lipshitz et al. 2001).But we suggest that it certainly is one important element that fits well into the “decidedly information-centered approach” of Gailmard and Patty (2012, 32). In the following sections we consider how principals can induce the acquisition of policy-relevant information.
The Relevance of the Expertise Model for Contemporary Governance
We seek to move beyond the historical perspective of Gailmard and Patty (2012) to demonstrate the relevance of the expertise model in informing contemporary tools wielded by political principals. We examine the application of these tools in the context of performance regimes, focusing on the field of education. One of the most marked trends in public sector reform across the globe has been the turn toward performance measurement systems that track government outputs and outcomes (Moynihan 2008).
The widespread adoption of such reforms is an example of an institutional design choice intended to encourage bureaucrats to learn more about what facilitates the effectiveness of the organization or program, how bureaucratic efforts contribute to these outcomes, and to alter policy and implementation choices based on that knowledge. The creation of performance regimes involves the political principal establishing a series of rules that require the creation and dissemination of reliable and frequent points of data, and may even explicitly direct agents to devote time and effort into examining data (Moynihan 2008).
Political principals remain wedded to using performance regimes, but acknowledge their struggles to design them in ways that actually facilitate the development of expertise. For example, in 2010 the US government saw the need to revise the 1993 Government Performance and Results Act with the GPRA Modernization Act of 2010. The original legislation required performance reporting, but was perceived to have failed in changing bureaucratic behavior (Moynihan and Lavertu 2012). The new legislation put in place more specific provisions to encourage bureaucratic use of performance data, such as requiring bureaucrats to hold quarterly meetings to discuss performance measures. The desire to increase bureaucratic consideration of information was explicitly identified by policymakers for the Modernization Act. For example, a Senate report on the legislation notes that the lawis “aimed at increasing the use of performance information to improve performance and results…Agencies are collecting a significant amount of information, but are not consistently using that information to improve their management and results” (U.S. Senate Committee on Homeland Security and Governmental Affairs, 2010, 11-12). Such policymaker goals are neither new nor unique – almost equivalent language can be found to justify the Government Performance and Results Act, and George W. Bush-era performance regimes (Moynihan and Lavertu 2012) – reflecting the limited progress of performance regimes.
While both the political consequences (Berry and Howell 2007; Bertelli and John 2010; Boyne, James, John and Petrovsky 2009; Kogan, Lavertu and Peskowitz 2015) and practical problems of performance regimes have been well documented (Figlio and Getzler 2006; Heinrich and Marschke 2010; Jacob and Levitt 2003; Moynihan 2008; Moynihan and Lavertu 2012), the expertise model offers an untapped framework for understanding how to restructure them to focus bureaucratic attention towards the acquisition of expertise. To examine how principals might better induce investments in expertise we used a randomized controlled field experiment in the context of the Danish public education system. In this setting, as in the expertise model, political principals have a demonstrated desire that agents invest in acquisition of performance information, and have made institutional design choices (the creation and provision of performance data) to facilitate that acquisition.
The Danish government has adopted both sophisticated value-added estimates of school test scores, which are widely covered in the media, and passed legislation that requires national reporting on measures of student wellbeing.Since 2015 all students in public schools annually respond to a survey measuring different aspects of their wellbeing. School principals receive reports for their own school. National results are reported and discussed in the media. Our experiment occurred prior to this legislation being implemented, but survey data shows that student wellbeing was salient to school principals at this point, who ranked it as more important than test scores (more detail below).
In a way that was not the case a generation ago, school bureaucrats in Denmark are expected to be knowledgeable of performance metrics, and appear to make use of this data when making management decisions (Nielsen 2014a). The performance data in our experiment are measures of studentwellbeing, which is factual, relationship-specific, policy-relevant, and can inform the achievement of policy outcomes, therefore fitting with the definition of central aspects of expertise laid out above. Such data can inform bureaucratic identification of problems (whether problems exist on student wellbeing in their schools), the nature of those problems (which aspects of wellbeing are high and low), and where to direct effort (if student wellbeing varies by classroom or sub-group). This actionable information, in turn, may inform the search for solutions. The expertise model also calls for agents to have some measure of tenure protection if delegations of authority are to generate investments in expertise. This is the case for our subjects, Danish school principals, who enjoy relatively high job security compared to many counterparts in, for example, the United States or the United Kingdom.