Agency Performance Challenges and Agency Politicization

Abstract

In this article we evaluate the relationship between political control and bureaucratic performance using information requested by researchers via Freedom of Information Act (FOIA) requests and Congress via congressional committee requests. The information requested was the same, and the timing of requests was similar. We find modest evidence of a relationship between agency politicization and a lack of responsiveness to requests for information from the public and Congress. Politicized agencies are slower to respond to requests even when controlling for agency size and workload. There is little evidence, however, that these agencies are more likely to respond poorly when they do respond. The difficulties in responding appear to be due to poor performance of the FOIA offices, either because political actors focus more on other agency activities or because of poorer management agency-wide. We conclude that efforts to make agencies responsive to elected officials may hurt management performance.

Questions of agency design are central to research on political accountability. From the moment they are created, agencies have different levels of political insulation that determine their level of responsiveness to elected officials (see, e.g., Lewis 2003; Moe 1989; Seidman 1998; Wood and Bohte 2004). Independent regulatory commissions, like the Federal Trade Commission or Nuclear Regulatory Commission, are designed for insulation from political control. Other agencies, such as executive departments, are not designed with political insulation as a primary goal and are therefore much more vulnerable to pressures from elected politicians.

Institutional design, however, involves tradeoffs (Moe 1989). Agency designs that facilitate political accountability to elected officials may help align agency policies with those of elected officials but allow politicization of agency tasks. Politicization, or the injection of politics into administration, can damage the ability of the agency to implement policy (see, e.g., Gailmard and Patty 2007; Light 1995; McCarty 2004). Designs that facilitate political accountability may introduce new layers of political review that hurt bureaucratic performance (Light 1995). Such designs may also accentuate the prioritization of political and immediate tasks, effectively crowding out attention and resources for other tasks (DeShazo and Freeman 2005; Dixit 2002; Holmstrom and Milgrom 1991; Wilson 1989). Finally, politicized agency designs may influence management quality by systematically increasing management turnover, emphasizing shorter time-horizons in planning, or altering worker incentives in the labor market in which agencies operate (see, e.g., Derlien 1996; Gailmard and Patty 2007; Lewis 2008; Suleiman 2003).

Of course, some scholars suggest that fully politicized structures (i.e., those that allow political influence into administrative operations) or at-will personnel systems are best for performance. Others suggest that a proper balance between political control and bureaucratic autonomy is optimal for performance (Bok 2003; Dunn 1997; Golden 2000; Krause, Lewis, and Douglas 2006; Maranto 1998). Unfortunately, the relationship between politicization and bureaucratic performance is difficult to evaluate because scholars disagree about its effects and comparative measures of bureaucratic performance are hard to find (see, e.g., Boyne 2003; Brewer and Selden 2000; Gilmour 2006; Moynihan et al. 2011; Radin 2006). Many existing measures of bureaucratic performance emphasize performance on policy-oriented tasks that further the agency’s mission, despite scholarly calls to reintroduce democratic values, like transparency, into performance evaluation (see, e.g., Piotrowski and Rosenbloom 2002).

In this article we evaluate the relationship between political control and bureaucratic performance using Freedom of Information Act (FOIA) requests. Our design exploits the FOIA process itself to compare agency performance on a non-mission, procedural task. We sent federal agencies identical FOIA requests and compared their responses to each other and to data provided by Rep. Darrell Issa (R-CA), former Chair of the House Government Oversight and Reform Committee. Since all federal agencies are subject to FOIA requirements, we measure agency performance by recording features of agency responses to our requests and a set of nearly-identical committee requests about FOIA sent by Rep. Issa to 107 agencies. Rep. Issa sent his request shortly before our requests, and his requests included the same information we requested. The cooperation of the House Government Oversight and Reform Committee provides a unique opportunity to measure performance and to compare responsiveness to political principals vis-à-vis the public.

We find modest evidence of a relationship between politicization and a lack of responsiveness to FOIA requests. Politicized agencies are slower to respond to requests even when controlling for agency size and workload. There is little evidence, however, that these agencies are more likely to respond poorly when they do respond. The difficulties in responding appear to be due to poor performance of the FOIA offices, either because political actors focus more on other agency activities or because of poorer management agency-wide, rather than intentional efforts to hide information or delay due to extra review. We conclude that efforts to make agencies responsive to elected officials may hurt management performance.

Politicization and Performance

Scholars have long discussed politically “responsive competence” and “neutral competence” as two ideals for bureaucratic performance (Aberbach and Rockman 1994, Heclo 1975, Moe 1985). Politicization is the injection of politics into otherwise neutral administration. Important research shows the strategies presidents and Congress employ to control the bureaucracy and how such efforts have changed policy outputs (see, e.g., Moe 1982, 1985; Randall 1979; Stewart and Cromartie 1982; Weingast and Moran 1983; Wood and Waterman 1994). A separate literature describes how the politicization of the civil service in the United States and other countries creates performance problems (Cohen 1998; Dunn 1997; Durant 1992; Heclo 1975, 1977; Kaufman 1965; National Commission on the Public Service 1989, 2003; Newland 1983; Rosen 1983). In the United States the increased “presidentialization” of the administrative state is generally viewed in a negative light because of its effects on agency performance and “neutral competence” (Aberbach and Rockman 1994; Heclo 1975, 1977; Suleiman 2003; see, however, Moe 1985). Presidents have increased the number and penetration of appointees into the bureaucracy (Lewis 2008; National Commission on the Public Service 1989, 2003), extended their control of the selection of appointees to even those at the lowest levels and increasingly focused on loyalty as a criteria for selection of appointees (see, e.g., Edwards 2001; Mackenzie 1981; Moe 1985; Pfiffner 1996; Weko 1995). Presidents have also gained greater control over budgets and regulation through the reorganization of the Bureau of the Budget into the Office of Management and Budget and the centralization of regulatory review under the Office of Information and Regulatory Affairs (Heclo 1975; Lewis and Moe 2009).

Agencies also respond to political principals in the legislature, who create and oversee most agencies (see e.g., Aberbach 1990; Dodd and Schott 1979; McCubbins and Schwartz 1984; Weingast and Moran 1983). Legislative committees can hold agencies accountable by calling agency executives to hearings where they can be publicly chastised for activities that the people disagree with. The legislature controls the power of the purse and thereby appropriations, which allows it to reduce agency funding in response to public dissatisfaction with agency performance. The legislature can also threaten new legislation, threaten to grant or withhold authorization or appropriations, and otherwise creatively sanction executive officials through public sanction or elimination of agencies or programs valued by executives (Aberbach 1990; Dodd and Schott 1979; McCubbins and Schwartz 1984). All combined, the legislature enjoys tremendous leverage to influence the behavior of agency appointees, another avenue for political accountability of agencies.

When examining the relationship between politicization and performance some scholars argue that performance problems stem from too little rather than too much political control of the bureaucracy.¹ One strand of this research focuses on the virtues of political appointees for performance, arguing that appointees are more capable than their careerist counterparts and bring necessary energy, risk-taking, and responsive competence to federal management (Bok 2003; Krause, Lewis, and Douglas 2006; Maranto 1998; Moe 1985). Another strand focuses on how removing civil service protections can improve performance. In this view, a personnel system with fewer civil service protections would more easily allow political leaders to recruit, promote, and retain employees on the basis of merit (Bilmes and Neal 2003; Maranto 1998). Leaders would be able to incentivize high performance with greater control over pay. If political leaders could hire, remove and reward more easily, they would be able to energize the public service and improve performance, attracting to the public sector high performing employees who are now choosing the private or not-for-profit sectors over government work (National Commission on the Public Service 1989).

Although scholars have illustrated the means of political control and resultant changes on policy, they have been less successful evaluating the effect of such actions on agency competence and performance. This is largely due to difficulties measuring government performance across agencies and contexts. The most common methods of comparing agency performance include the use of surveys of employee attitudes or government-wide program management scores, such as the Bush Administration’s Program Assessment Rating Tool (PART) scores (see, e.g., Brewer and Selden 2000; Lewis 2008). Each of these measures has limitations. Scholars have sought alternative measures of performance, measuring agency performance on non-mission and procedural tasks that many agencies must complete, such as budget forecasting.² FOIA requests and congressional requests about FOIA are also non-mission tasks, and studying them allows us to include a larger number of agencies than analyses examining budget forecasting performance. Although politicization may affect performance on mission tasks, here, we limit our generalization to performance on non-mission tasks.

Even though FOIA is a non-mission, procedural endeavor, responding to a FOIA request requires four discrete tasks, some of which are more procedural than others. First, bureaucrats must confirm receipt of the request and assign a case number, a purely procedural task. Second, they must decide whether records exist that are responsive to the request, which can be procedural or more substantive in nature, depending on the content of the request. Our requests were highly procedural, which allowed us to make the same request across agencies. Nevertheless, responding even to a highly procedural request like ours is a more substantive endeavor than simply confirming the request. The third and fourth tasks require the bureaurcrat to make legal determinations about whether to claim exemptions and whether to charge fees. Criteria for both of these decisions are clearly laid out in the Administrative Procedure Act. The substance of the exemptions and fees decisions are not policy related, but nor are they rote, like confirming the request. The decision of whether to charge fees is more procedural than the decision of whether to claim exemptions, because fees are assessed depending on the nature, timing, and extent of the request. The exemption decision can be substantive, in that it requires legal determinations about the nature of the information at issue. The four measures we evaluate here—time to confirm, time to respond, number of exemptions, and whether the agency charged fees—therefore shed light on different aspects of performance on non-mission, procedural tasks.

How Can Politicization Hurt Responsiveness?

In order to understand whether politicization helps or hurts performance it is important to specify the mechanisms by which politicization alters agency decisions and processes. The politicized structures and actions intended to increase agency responsiveness to the president and Congress may also have damaged agencies’ abilities to carry out important tasks. Among the possible mechanisms by which politicization hurts performance, three stand out in the case of political responsiveness. First, politicized agencies may add more political review to agency action. Additional political review can influence both the process and the objective quality of decisions. For example, presidents and political appointees may prefer to more frequently review agency decisions. The extra review for every agency decision can make more “political” agencies less nimble and less responsive. Giving a greater number of stakeholders access to agency decisions can generate internal agency friction and disagreement. Political review also provides opportunities for political actors to insert electoral calculations into agency policymaking. One motivation for the creation of insulated agencies is to limit politicians’ ability to manipulate monetary policy, shade budget forecasts, bias regulatory policies, or otherwise direct agency decisions in ways that help party electoral fortunes but not social welfare (see, e.g., Cukierman, Webb, and Neyapti 1992; Krause, Lewis, and Douglas 2013; McCarty 2004). More political intervention into administration could influence both the process (e.g., speed of decisions) and content of decisions (e.g., more political considerations).

Second, increased political influence in agencies could influence internal agency resource allocation in ways that could crowd out other activities. Because FOIA is a non-mission task, we already expect that the task will be starved for resources (Simon et al. 1991 ; Lee 2016; Wilson 1989). Politicization can make it worse by accentuating new administration priorities and augmenting the reordering of agency priorities away from non-mission tasks. If political appointees prioritize policy work, like rulemaking, over non-mission tasks, like FOIA, the allocation of time, personnel and resources across tasks will be influenced in predictable ways.³ Agency leadership is less likely to assign the agency’s best personnel to tasks that are not monitored by political officials or on the political agenda. Over time, the accumulation of choices to prioritize policy work shapes the internal labor market and the struggle for internal resources. Non-mission tasks can be seen as not-career enhancing (Wilson 1989). Parts of the agency gain reputations as “turkey farms” and have a hard time recruiting and retaining the best personnel from both inside and outside the agency. The poor reputations and second-best employees of the affected units can inhibit performance.⁴ The resulting shift is task-focused: once internal infrastructure develops around policy responsiveness over non-policy responsiveness, even elected officials will see less responsiveness to non-mission requests, like FOIA requests, in highly politicized agencies.

Third, more politicized agencies can suffer from generic management problems across agency tasks, mission or non-mission. Aggressive micromanagement can reduce incentives for agents to spend time and effort developing expertise and procedures that will benefit the agency and the public (see, e.g., Gailmard and Patty 2007; Lewis 2008; Richardson 2016; Stephenson 2011). High management turnover among executives reduces organizational memory, making it difficult for useful collected knowledge about agency politics, folkways, and processes to be transferred across administrations. This turnover and loss of organizational memory, combined with lower levels of executive competence, can compound management problems. Politics trumps competence for promotion considerations in highly politicized agencies, where top management positions are reserved for co-partisans, and civil servant pay is limited by federal pay guidelines (Suleiman 2003). All of these features combine to make it hard for highly-politicized agencies to recruit and retain the most expert career employees. Generic management problems will result in reduced performance.

Data and Analytic Strategy

To evaluate the performance of agencies, we examine citizen requests for information. Specifically, we conduct an original study, sending two FOIA requests to 132 agencies and evaluating the quality of the responses. We analyze confirmation and response times and the quality of responses. We also incorporate data from R-CA, who sent an almost-identical request to one of our requests a few months before our study.

An examination of the timing and quality of responses to identical FOIA requests provides a unique and comparable measure of agencies’ performance. FOIA requests are not the most important task agencies perform, but they are important and have the virtue of comparability across agencies and with data requested by Congress. FOIA has been instrumental in uncovering otherwise-hidden government information, like the Bush-era torture memos, documentation of the Iran-Contra affair, and the use of drones domestically. The FOIA provides clear benchmarks for performance. Under FOIA any person can request any agency record not covered by one of nine exemptions.⁵ Once the agency receives a request, FOIA requires that it confirm receipt within 10 business days and issue a determination within 20 business days. This determination can be to fill the request, fill it in part, decline the request, or indicate that the agency has no records responsive to the request. Agencies may ask for extensions to the statutory time frame or stop working on a request until the requester provides fees or clarifying information. After the agency responds to the request, if the requestor disagrees with the determination, she may file an appeal with the agency. If the appeal is denied or not resolved to the requestor’s satisfaction, her next remedy is in the courts.

On May 9, 2011, we submitted two distinct FOIA requests to 132 federal agencies, totaling 264 requests (132 of each type).⁶ Our requests, sent by different people on the same day, are located in Supplementary Appendix B. Each agency received a baseline and sensitive request. The baseline request asks agencies to provide “the agency’s FOIA log from 2010 or equivalent listing the FOIA request number, name of requester, and a description of the records being requested.” The politically sensitive request asks for a list of FOIA requests from 2010 about which a political appointee “personally made an inquiry, personally reviewed, or personally had a hand in the disposition of the request.” Each request was mailed with a return address of a private citizen, but the letter notes that the request for information is for educational purposes. We measure the time it took for the agency to confirm our request and respond to our request if they responded. We also note the number of exemptions claimed and the quality of the response for each request.

As Chair of the House Committee on Oversight and Government Reform, Rep. Issa emphasized his interest in ensuring agency responsiveness to FOIA requests. The most visible of his efforts in this regard was a request sent to 107 agencies about FOIA on January 25, 2011. In his request, Rep. Issa asked agencies for copies of agency FOIA logs from 2006 to 2011 and set a February 15, 2011 deadline. The content of his request is identical to the request for logs that we sent 4 months later, except that Rep. Issa requested significantly more data than we did. This means that by the time we sent our requests on May 9, 2011, Rep. Issa had already requested the same material from each agency (and most agencies had produced the material).⁷ The Committee recorded the responses to these requests and generously provided us the data. We analyze responsiveness to Rep. Issa separately from responsiveness to us.

Agencies confirm by letter or email that they have received a request, and they assign the request a unique case identifier. As past and present FOIA officers told us, after the request arrives to the FOIA office, the officer enters it into a spreadsheet and tracks the custody of the request as it travels between agency officials for review and response.⁸

We received the first agency confirmations and determinations May 12, 2011. The last data collected in response to these requests was received May 9, 2012, 1 year from the date we sent the letters. Out of 264 requests, we received confirmation of 219 (83%). The average time to confirmation, if we give all of the outstanding requests the maximum time of 1 year to respond, is 79 days (63 for baseline, 95 for sensitive). Out of 264 requests, agencies filled 195 (74%), and the average time was 122 days, giving non-responses the maximum time of 1 year to respond (96 for baseline, 147 for sensitive). In some cases, the confirmation and determination arrived in the same communication (i.e., we received your request and here is the determination). Supplementary Appendix C provides the list of agencies that had not filled either request after 1 year.

We coded poor responses where the response did not comply with the law or was not fully responsive. The poor responses asked for notarized proof of identity, which is not required by law, or sent the wrong information, including sending only a partial year’s logs. We also coded as a poor response any request to pay fees. Since the persons sending the requests were educational requesters, we did not expect to be charged fees unless the response was over 100 pages. According to the statute, agencies are prohibited from charging educational requesters for document search and review but are allowed to charge for duplication beyond 100 pages. Fees cannot be charged to public requesters if the response takes more than 20 days, unless the agency claims “exceptional” or “unusual” circumstances. All fees requested of us were either for responses that took longer than 20 days, or for agencies that provided fewer than 100 pages of documents. The only agency to claim exceptional circumstances did so for our request for FOIA logs. But the logs request did not present an exceptional circumstance, which is defined as a response that requires coordination across different components of the agency. By definition, the request for FOIA logs was handled entirely within the agency’s FOIA office itself. Supplementary Appendix D contains examples of agencies that provided both poor responses and whose responses exceeded statutory time limits.

As is expected and appropriate in a representative democracy, agencies were more responsive to Rep. Issa than to us. The average time to respond to Rep. Issa’s request for information was 29 days (an average of 8 days later than the deadline he set), whereas the average time to respond to our baseline request was 96 days. Even though most of the agencies we contacted had already provided Rep. Issa with the information we requested, they responded to us (private citizens) more slowly.

Measuring Politicization

It is difficult to directly observe and measure political influence into agency activities and resource allocation. Existing measures are proxies for politicization, measuring agency features that should correlate with the amount of political intervention into administration. Measures used in the existing literature fall into two categories: structural and evolutionary.

Structural measures are of stable differences in agency design that vary across agencies but less frequently across time, such as the ease or difficulty with which the President can remove appointees. Some structural features are mandated at the time of the agency’s creation (like fixed terms or removal protection) and others might be decided later by the agency, like the agency office that is in charge of an activity. Our first structural indicator is the agency type. Agencies can be loosely grouped into agencies in the Executive Office of the President (EOP; 0,1), executive departments (0,1), independent administrations (0,1), and independent commissions (0,1). Some agencies are also sub-components of larger agencies and others have no layers of bureaucracy above them (Bureaus; 0,1). Agency type reflects the degree of presidential influence since agencies in the EOP and the executive departments are designed to be most responsive to the president (although exceptions exist, such as the Internal Revenue Service). Appointees to independent agencies generally enjoy removal protections and are, by design, more protected from presidential influence (Barron 2008; Breyer et al. 2011; Lewis 2003; Lewis and Moe 2009; Selin 2015).

The location of the FOIA officer is another structural feature of agencies that could predict responsiveness to political principals in their FOIA activities. The location of FOIA officers varies across agencies (e.g., general counsel’s office, public affairs office, management directorate, etc.) and is unrelated to the number of appointees in an agency (Clark 1967; Wozencraft 1967; Giannella 1971). About half (54%) of the agencies in our sample were created before the FOIA was adopted, so the decision on where to place the FOIA office was made at a different time than the decision of agency type. Including an indicator for location of the FOIA office in the executive secretariat (e.g., Office of the Secretary, Office of the Chairman) allows us to analyze whether proximity to political appointees influences FOIA office performance.⁹ If it does, then mechanisms of delay might include micromanagement and other management problems or increased layers of review. As agency size and FOIA workload increase, the probability that a FOIA office is located in the secretariat increases. We control for both, as we describe below.

The second type of politicization measure captures aspects of the evolution of administration. Here scholars focus on variation over time in the depth and penetration of appointees and variation in backgrounds of appointees, particularly the extent to which appointees are loyal and committed to extending the president’s influence down into the administration. Scholars also measure the layers of political review that check, cabin, and alter the content of agency decisions. These measures vary not only across agencies but also across time with the assumption that more appointees, more loyalty, and more political review correlate with more political intervention into agency policies and practices.

Finally, a limited number of works use survey data to measure the reported balance between appointees and career professionals in the decision-making process (Bertelli and Lewis 2012; Richardson 2016). The 2007–2008 Survey on the Future of Government Service (SFGS) asked top career federal executives “How much influence do the following groups have over policy decisions in your agency? [Political appointees, Senior civil servants].” Theoretically, both stable differences in agency structure across the executive establishment—the first type of measure—and different presidential strategies regarding appointees and review—the second type of measure—should predict differences in reported appointee influence in agencies, particularly in agencies where appointees do not share the preferences of career professionals.

We have conducted our main analysis using structural measures, which have the virtue of being available for all of our cases. Models using appointee depth are more difficult for us, since we asked about appointees themselves in our FOIA requests. Those models are in Supplementary Appendix E.¹⁰ The overlap between the SFGS survey data and our own is sparse, though estimates on time to respond and quality of response were consistent with our hypotheses. We present those models in Supplementary Appendix F.

Other Variables

We include other agency and FOIA-specific variables, in order to rule out competing explanations. The size of the agency could explain responsiveness because large agencies are more likely to have dedicated FOIA officers and large staffs. We measure agency employment as of March 2011. In order to avoid a situation in which responses to the sensitive request are delayed due to the number of appointees in the agency, we also include the number of appointees (i.e., more appointees means more potential matches between the request’s chain-of-custody list and the list of appointees in the agency who might intervene in FOIA responses).¹¹

We also measure the number of FOIA requests received in FY2010 since the number of requests could improve or hinder agency performance. An agency receiving a large number of requests has a lot of experience handling such requests but also a larger and probably more bureaucratized FOIA process.¹² Finally, we include an indicator for politically sensitive requests since such a request is less routine. In addition, we interact this variable with measures of politicization to determine whether more politicized agencies are particularly slow with more political requests. Table 1 summarizes these measures.

There is interesting variation in responsiveness by agency structure. The five agencies in the EOP took, on average, 13 days longer to confirm our requests and 26 days longer to respond to our requests than other agencies. They took 2 days longer to respond to Rep. Issa’s request, which had a tight, 22-day deadline. The pattern among the 52 agencies in executive departments is similar. Cabinet agencies took 8 extra days to confirm our requests and 39 days longer to respond to them than agencies outside the cabinet. Executive agencies’ responses to Rep. Issa were 9 days slower than non-cabinet agencies. Several aspects of FOIA workload might explain these tendencies. For example, combined, EOP and cabinet agencies received far more FOIA requests in the year leading up to our requests than independent administrative agencies and commissions (difference of 11,515, p = .000), but they have more people tasked with filling requests, with a FOIA office staffing difference of 77 people, on average (p = .000), resulting in a request-per-employee workload difference that is indistinguishable from zero (difference = 10, p = .6). So, EOP and Cabinet agencies are not disadvantaged, on a per-request basis, compared to independent agencies and commissions, but they underperform them in both confirming and responding to requests. The explanation could be more politicized review.

No EOP agency houses the FOIA officer in the secretariat, and cabinet agencies are no more likely than non-cabinet agencies to do so (difference = .003, p = .9). When we examine the average confirmation and response times by location of the FOIA offices themselves a few things stand out. Interestingly, agencies with FOIA offices in the office of a politically appointed agency head (“secretariat”) were quicker to confirm receipt of our request by about 20 days but slower to respond to it by more than 40 days on average. They were slower to respond to Rep. Issa by about 20 days on average. As with agency structure, there are a number of possible explanations for this pattern. FOIA offices in the secretariat receive more requests, on average, than FOIA offices outside of the secretariat (difference of 19,971, p = .000), but they have much larger staffs than FOIA offices outside of the secretariat (averaging 68 more employees per FOIA office, p = .007). So their overall workload is statistically indistinguishable from non-secretariat FOIA offices (147 requests per staff member versus 136, p = .7). The vast quantity of requests might mean that FOIA offices in the secretariat are best at automating confirmations, a highly proceduralized task, yet their location might mean that the offices have the most political review of requests.

Based on the simple, bivariate, descriptive data on FOIA responses, the most politicized agencies appear slower at responding to citizen requests for information than less politicized agencies. As we turn to statistical modeling, we bear in mind this basic relationship as well as the three mechanisms that might explain it: highly-politicized agencies might have more layers of political review, allocate resources in a way that crowds out non-mission tasks, and experience generic management difficulties, more than less-politicized agencies.

Models of FOIA Response

To account for potential confounders and right-censoring on the agencies that never responded we estimate Accelerated Failure Time (AFT) models with the following basic specification:

for agency a, where Y_a is the number of days it takes an agency to fill a FOIA request, X_a includes agency-specific controls, p_a is agency politicization (agency structure or FOIA office in secretariat), R_a is request type (baseline or sensitive), and ε_i is the random error term modeled to have a density function f(ε), in this case assumed to be log-gamma.¹³ We report robust standard errors clustered on agency since each agency received two requests and those observations are nonindependent.

Results

Do structural features that enhance political accountability hinder FOIA performance? In table 2 we model the time it took for agencies to confirm receipt of our FOIA requests. In table 3 we include models of the time it took for agencies to fill FOIA requests to us and respond to Rep. Issa. A positive coefficient indicates that it took agencies a longer time to respond to the FOIA requests.¹⁴ Each table includes basic specifications and specifications that interact politicization and the type of request (baseline versus sensitive) to test whether more politicized agencies are particularly slow at responding to politically sensitive requests (they were not). Overall, model estimates suggest that agencies with less structural independence and FOIA officers in agency secretariats (i.e., more politicized), took longer to confirm and fill our FOIA requests. More politicized agencies, however, were no more likely to produce a bad response if they responded. More politicized agencies also took longer to respond to requests from Rep. Issa.

Confirmation of FOIA Requests

The results across models in table 2 indicate that politicization affects responsiveness. If politicized agencies perform less well at non-mission activities like FOIA, then agencies with FOIA housed in the office of the appointed agency head, such as the Office of the Secretary, Director, or Chair (all coded “Secretariat”), rather than management offices, general counsels’ offices, or public affairs offices, should respond more slowly than agencies outside of the Office of the Secretary. Indeed, in models with full controls, agencies with FOIA officers in the Office of the Secretary or Chair are estimated to respond more slowly. Agencies with the FOIA officer in the secretariat were estimated to take 38 days to confirm a request compared to 20 days (i.e., almost twice as long) to confirm in agencies with FOIA officers in another part of the agency. This is important evidence that more politicized agencies or politicized FOIA processes lead to worse FOIA performance.

Using agency location as a measure of politicization, agencies in the EOP are estimated to be significantly slower in confirming FOIA requests than the base category, independent commissions. Median confirmation times were about 30 days slower for agencies in the EOP, even when controlling for the percentage of appointees. The slow confirmation times could stem from the fact that political oversight from the White House slows down response to records requests, or it could be that EOP agencies simply do not prioritize FOIA relative to other tasks.

Interestingly, the type of request (baseline or sensitive) did not influence confirmation times, even in the most politicized agencies. The fact that the most politicized agencies treat politically sensitive requests like other requests suggests that poor performance by politicized agencies has less to do with more layers of oversight or efforts to hide information.

Filling FOIA Requests

Ultimately, what is important in FOIA response is how long it takes for agencies to fill or make a determination in a case. That is, how long did it take for the agency to give us our FOIA logs in various forms (i.e., full log for baseline request or log of appointee-relevant FOIA cases) or determine that our request produced no records? The results for filling FOIA requests mirror those for confirming requests except that the coefficients for FOIA officers in the secretariat are estimated less precisely. Substantively, response times are estimated to be more than twice as long for agencies with FOIA officers in the secretariat, although we cannot reject the null that FOIA office location has no effect except for the requests from Rep. Issa.

As with confirmation times, agencies in the EOP are estimated to be about 34 days slower in filling requests than the base category of agencies, independent commissions. Estimates also indicate that agencies took 9 days longer with the politically sensitive requests. Again, politicization is correlated with slower response times.

Responses to Rep. Issa’s Request for FOIA Information

The patterns of response to Rep. Issa’s request for information about agency FOIA processes are similar to those of agency responses to our FOIA requests. Agencies with FOIA officers in the secretariat are significantly slower than agencies with FOIA officers in other parts of the agency. They are estimated to be about 10 days slower than other agencies. Given the tight timeline requested by Rep. Issa, a 10-day delay is substantively significant. Independent administrative agencies and independent commissions responded, on average, after 19 days. Agencies in the EOP and Cabinet were slower to respond, by 3 and 10 days, respectively. Beyond that, large agencies were slower to respond to Rep. Issa. The median response time for smaller agencies (i.e., bottom quartile) is estimated to be 17 days compared to 29 days for larger agencies (i.e., top quartile).

Scholars usually think that more politicized agencies are designed to be more responsive to elected officials, but in this case more politicized agencies were actually less responsive to an elected official, albeit one in Congress. The fact that the same factors that predict performance in response to our FOIA request also predict performance in response to Rep. Issa’s request suggests that politicization has a general effect on federal management performance.

In sum, not only did agency politicization reduce responsiveness to the public, agency politicization reduced responsiveness to Rep. Issa. These findings present new evidence that agencies with the features of political accountability (i.e., located in EOP, FOIA located near the most political part of the agency) are less responsive to requests for information. Features of administrative agencies that enhance political accountability may decrease both direct democratic accountability and accountability to a political principal.

Quality of Response

The time it takes an agency to respond to a FOIA request is only one measure of performance, although arguably the most important one. Table 4 includes other estimates of response quality, whether or not an agency claimed exemptions in filling the request or otherwise produced a poor response to the request. These estimates are based upon only cases where we actually received a response. This data restriction leads to potentially underestimating the effect of politicization since the worst performing cases (i.e., those that did not respond to us at all) are excluded from these analyses. The excluded cases are also those with higher levels of politicization since high-politicization cases are among the least likely to respond in the first place. Nonetheless, it is worth looking at the quality of the responses we received.¹⁵

The results here are inconclusive, generally indicating that agencies designed to be more responsive to political principals are no more likely to claim exemptions or otherwise produce a poor response. We could not reject the null hypothesis that our measures of politicization had no influence on the number of exemptions claimed or the quality of the response if a response was sent. And depending on the measure used, politicization could create more (secretariat) or fewer (executive department) poor responses.

Other coefficient estimates suggest that overworked FOIA offices—in agencies with a large number of requests—are the most likely to produce a poor response. Increasing the number of requests to FOIA staff by 1% is estimated to increase the probability of a poor response by 5%. Interestingly, the responses to our (baseline) request for logs were significantly more likely to lead to poor responses, despite the fact that our sensitive request was less common, more complicated, and had not previously been made by Rep. Issa. Although the evidence suggests having FOIA offices close to political officials may lead to slower responses, it is also the case that the sheer size of the task and relative workload influence the quality of the response, even though it does not affect the speed of the response.

Summary

In total, the analysis of timing with controls suggests a relationship between politicization and a lack of responsiveness to FOIA requests. Politicization is correlated with poor performance. Politicized agencies are slower to confirm or fill requests even when controlling for agency size and workload. There is little evidence in the full dataset, however, that these agencies are more likely to claim exemptions or respond poorly when they do respond. What evidence there is suggests that agencies with heavy workloads are the most likely to produce a response that is less satisfactory to the requester (i.e., produce a poor quality response). Difficulties in responding may also be due to poor performance of the FOIA offices, either because political actors focus more on other agency activities or because of poorer management agency-wide.¹⁶

Discussion

The results of our study suggest that the effects of politicization in agency design can hamper performance on non-mission tasks. We find relatively stable relationships between design features intended to enhance political responsiveness and slow agency responses to political principals in non-mission tasks. The relationship persists when we examine agency responsiveness to the public, a less surprising, but still politically consequential result. Indeed, one interesting implication these results may be that agencies designed to be more responsive to elected officials may be less responsive to the public. These results can be added to a growing list of studies that suggest that too much politicization can be harmful for performance (Cohen 1998; Dunn 1997; Durant 1992; Heclo 1975, 1977; Kaufman 1965; National Commission on the Public Service 1989, 2003; Newland 1983; Rosen 1983).

Our design allowed us to examine additional delay for politically sensitive requests, which are exactly the kind of request we would expect the extra layers of review to affect. Interestingly, more politicized agencies were no slower in responding to the politically sensitive requests. As such, there was little support for the idea that politicization harms performance because of increased layers of political review. One explanation for this null finding is that the sensitive requests were not sensitive enough to trigger additional review by appointees. We cannot rule out this possibility. Nonetheless, although all agencies were slower responding to political sensitive requests, this was uncorrelated with the degree of agency politicization.

Another explanation for the effects of politicization on performance is that increased attention to politically relevant tasks crowds out attention to non-mission tasks such as FOIA. The results here are consistent with this hypothesized mechanism for poor performance. Indeed, politicized agencies systematically provide fewer resources to FOIA offices than other agencies. One way to measure this is the percent of employees in the agency FOIA office, relative to the FOIA workload. With the exception of EOP agencies, which probably have a floor effect (i.e., every agency has to have at least one FOIA person), we see the expected pattern hold. Agencies dedicate fewer resources proportionally in cabinet agencies and more in independent agencies. In addition, in recent a recent survey of federal executives, respondents in EOP and Cabinet agencies were less likely to name information transparency (i.e., FOIA, data availability) as an important agency task relative to other tasks such as rulemaking, analysis, or inter-agency collaboration.¹⁷ This indicates that policy-related tasks are more likely to crowd out non-mission tasks in politicized agencies.

More generally, based on the nature of some of the slow responses even to confirm requests, it seems that general management challenges could impede the FOIA process (Gilmour and Lewis 2006; Heclo 1977; Lewis 2008). The delays in responding to Rep. Issa’s requests followed a pattern similar to ours, indicating that, from the standpoint of political responsiveness, politicized agencies were bad at both confirming and responding. These findings are consistent with previous work demonstrating that politicized management teams can hurt program and agency performance (Gilmour and Lewis 2006; Gailmard and Patty 2007; Lewis 2008). Politicization leads to higher turnover, systematic differences in background qualifications, and difficulties in the recruitment and retention of the best and the brightest. These management difficulties influence a whole range of agency activities including, but not limited to, FOIA.

We should be clear. Our research design involves tradeoffs and our argument is limited. These data have the virtues of replicating the kind of experience normal FOIA users have, and they do not rely on agency self-reports. In addition, they provide a wonderful and unique opportunity to compare the citizen request to the request that Rep. Issa made. The data have the advantage of being 264 draws from agency response times but the disadvantage of only being two draws from each agency’s distribution of agency confirmation or response times. In addition, politicization is not randomly distributed among agencies. It is possible that politicization is correlated with omitted agency features correlated with both politicization and performance. If our efforts at identifying the appropriate agency controls are ineffective, our coefficient estimates on the percentage of appointees or FOIA office location could be biased. We have tried to identify the relevant covariate controls but these data have the same weakness as most cross-sectional studies.

We further recognize that the task of providing information is only one type of agency role, fitting in a broader context of agency activities, primary of which are rulemaking, adjudication, enforcement and policy implementation. Our findings show that there is variation in direct public accountability and accountability to political principals. The variation in accountability should not be ignored or assumed not to transfer to the agency roles more central to their core tasks. If an agency is unaccountable in the role that is most directly associated with citizen contact, might it also be unaccountable in rulemaking, adjudication, or other policymaking? The possibility remains, and it is highlighted by our findings.

Conclusion

This article evaluated the relationship between political control and bureaucratic performance using requests for information about FOIA, partly through the FOIA process itself. It finds evidence of a relationship between politicization and a lack of responsiveness to FOIA requests. It provides new evidence to help arbitrate an ongoing debate about the effects of politicization on agency performance.

Bureaucratic accountability—bureaucrats’ obligation to answer for their actions—is a longstanding concern in democratic politics. It has become an increasingly important concern in the post-war and post-Watergate era, as the public has demanded transparency in an effort to facilitate accountability. The transparency provided by FOIA allows the public to hold agencies accountable. FOIA is a potentially powerful tool, made more powerful when agencies respond to FOIA requests both quickly and completely. FOIA requests help private citizens, the press and civil society groups examine government actions, an essential component of electoral accountability in a free society. Yet some features of agency design might impede agencies’ ability or willingness to respond to FOIA and other “non-mission” requests in a timely manner.

The implications of the findings go well beyond their contribution to the literature on politicization and performance. Although in many ways a mundane task for agency employees, the FOIA process is crucial to agency accountability. It has recently become the target of oversight, not only from Rep. Issa, but from the Obama White House as well. The White House Open Government Initiative featured a specific focus on information availability directly to the public. From creating centralized government Web sites that ease information searches to de-classifying documents and ramping up FOIA responsiveness, the Obama Administration reported that it made progress in opening up government and encouraging agencies to be more directly responsive to citizens. Although increased transparency is generally popular with the electorate, any President faces an uphill battle to improve the FOIA process. The structures that any president relies upon for political responsiveness—including locating even non-mission agency tasks in the Secretariat—correlate with less openness in government and weaker FOIA responsiveness in particular.

Supplementary Material

Supplementary data is available at the Journal of Public Administration Research and Theory online.

Funding

This work was supported by the Berkeley Empirical Legal Studies Fellowship..

Acknowledgments

Previous versions of this article were presented at the 2012 annual meeting of the Midwest Political Science Association, April 12–15, Chicago, IL, the Political Economy and Public Law meeting, May 30–31, 2012, Charlottesville, VA, and the Southern California Law and Social Science Forum, March 27, 2015. We thank Joshua Clinton for the original idea to use FOIA responses as data, Erin Hartman and Jas Sekhon for helpful conversations around research design. Ben Goldberg, Evan Haglund, John Hudak, Gbemende Johnson, Tegan Milspaw, Janna Rezaee, Saskia Lewis, and Jennifer Selin provided helpful research assistance. We thank Scott Altman, George Krause, David Nixon, anonymous reviewers, and seminar audiences at the London School of Economics, UC Berkeley, the University of Exeter, and Yale University for generous and helpful feedback. The errors that remain are our own.

References