Agency control through the appointed hierarchy: presidential politicization of unilateral appointees

Abstract

Schedule C and noncareer Senior Executive Service positions hold significant influence over policy outcomes, yet they have received limited scrutiny compared to advise and consent (PAS) appointments. Such appointments offer understudied avenues for presidential control over the bureaucracy. Through a comprehensive analysis of more detailed data than has been employed to date, we reveal that these appointments are responsive to broader political dynamics, particularly those relevant to PAS appointments, including inter- and intrabranch conflicts, agency ideology, Senate workload, and the political calendar. However, statutory constraints and agency characteristics – such as the managerial expertise of appointed agency leadership – also shape their utilization. While unilateral appointments provide an advantage to Presidents, executives are constrained when using them to overcome legislative opposition or reshape resistant agencies. These lower-level appointments reflect the wider political landscape, granting the President significant – but not unrestrained – opportunities to exert influence on both the bureaucracy and policy outcomes.

On April 6, 2007, Monica Goodling, a noncareer Senior Executive Service (SES) employee within the Department of Justice (DOJ), resigned her position. This was in response to evidence she and other department appointees had improperly considered political affiliations when deciding about civil servant hiring, firing, and promotion, including the midterm dismissals of seven United States Attorneys. Goodling had initially joined the DOJ as a Schedule C appointee, with her experience as an opposition researcher for the Republican National Committee serving as her primary qualification. Subsequent investigations uncovered similar instances where lower-level appointees intimidated career staff, censored government reports, and disclosed internal documents to external groups in pursuit of the administration’s objectives (Lewis and Waterman 2013).

While an extreme case, this scandal underscores the extent to which noncareer appointees can influence the inner workings of the American bureaucratic state – and, by implication, its policy outcomes. Often conduits between top appointees on the one hand and the civil service on the other, what determines their selection has considerable implications for politics and policy. Such choices take place in the context of a back and forth between competing political interests to influence the American bureaucracy (e.g. Shapiro and Wright 2011), with personnel a principal arena for separation of powers interactions over politicization.

As is well known, there are three principal types of political appointments, varying by agency hierarchy position and need for Senate approval: (1) Presidential appointees subject to Senate confirmation (PAS), top agency leaders whose appointments must be affirmed by the upper chamber; (2) appointed members of the SES, operating under the PAS appointees and alongside career civil servants; and (3) Schedule C appointees, working under the SES and often in advisory roles to other appointees (Cohen 1998). In the hierarchy of appointees, PAS positions are at the top, followed by the SES – no more than 10% of whom can be politically appointed overall ¹ with a limit of 25% in a given agency – with Schedule C employees at the bottom. Numerically, PAS and Schedule C are similar (approximately 1,200 and 1,400, respectively), with appointed SES about half that.

We now have more than 40 years to look back on how this hierarchy functions and politicization processes operate. While PAS appointments are associated with the 1883 Pendleton Act, Schedule C and SES positions are comparatively recent creations. The former originated in 1956 as a set of exempt positions the President could directly name. Despite shifting authority to the President, Schedule C as an institutional arrangement has not met much Congressional opposition, perhaps because these appointments are sufficiently low in the agency structure. The SES was part of the 1978 Civil Service Reform Act that reacted to Watergate by emphasizing merit-based pay and alleviating some problems with the relationships between civil servants and appointees (see Heclo 1977); although many of the forces viewed as the initial impetus for the SES have seemingly lost steam, the SES and the President’s right to appoint remain (e.g. Aberbach and Rockman 2000; Moynihan 2004). Beyond settling on whom to appoint, the chief executive must decide where to allocate the 10% of non-civil service positions available.

For understandable reasons – given their prominence and the tug of war between the President and Senate concerning their selections and approvals – PAS appointees have received more scholarly attention than their Schedule C and SES counterparts. How choices regarding these lesser roles compare to those for PAS appointments, and how such decisions affect bureaucratic politicization, is less studied. Much conventional wisdom implies direct intervention by the EOP for lower-level appointees is less, with such appointments made by higher ranked appointees and only reviewed and approved by the EOP. ² Key is the relationship of Schedule C positions to those above them. But others claim a more political dynamic underlies Schedule C selections and politically chosen SES appointments (Bonica et al. 2015; Lewis 2007, 2008; Moore 2018; but see West and Cheon 2019).

However, available data one would want to assess drivers of non-PAS appointments and their implications for politicization have been somewhat wanting. The cross-sectional breadth of agencies, types of appointments, and the temporal span covered have been limited. ³ Hence, an organic picture of agency politicization for the era following the 1978 Civil Service Reform Act including the full menu of Schedule C appointees, Presidentially chosen SES members, and Senate-approved PAS nominees has been lacking. How are the former two employed for different agencies and political contexts given that, unlike PAS positions, the chief executive’s influence over their allocation is more varied? ⁴

These questions are important given the potential of such appointees on policy outcomes. Schedule C appointees serve in policy or confidential positions within the executive branch and are typically involved in policy development, public affairs, and other roles aligned with administrative objectives. Noncareer SES appointees come from outside the civil service to high-level leadership positions. They bring specialized expertise and are responsible for driving initiatives and implementing policies within federal agencies (Cohen 1998).

As indicated, with some recent notable exceptions – Moore (2018), West and Cheon (2019), as well as others – research on Presidential strategies underlying the distributions of noncareer SES and Schedule C positions has been lacking and constrained by data availability. Thankfully, a large dataset now offers a means for assessing hypotheses about drivers of Presidential strategy – In 2017, BuzzFeed obtained and released a huge trove of information through Freedom of Information Act requests from the OPM, beginning with 1973 through first 2014 and then early 2017 and covering the vast swath of the bureaucracy. ⁵ As the data only separate out Schedule C appointments from the early 1980s, and we lack other requisite measures for most of 2017, we study the period from March 1984 to March 2017.

Our analysis generates insights into a world where Presidents require direct Senate support for some but not all appointments, and where the chief executive and higher-level appointees oversee lower-level appointees. For the 30-plus years for which we have non-PAS appointment data, we develop hypotheses about, and examine the relevance of, features drawn from the voluminous studies on PAS appointments and the more limited investigations into noncareer SES and Schedule C appointments: Partisan and ideological conflicts between the President and the Senate, the Senate’s inability to coordinate due to internal disagreements, the relationship between executive and agency preferences, the ideological relationship between the executive and relevant agency head, whether an agency is a central staff agency (i.e. Government Services Agency [GSA], Office of Management and Budget [OMB], or OPM), the progression of staffing within a given agency, the ability of appointed agency heads to manage their agencies, whether timing is early or late in an administration or near an election, and the business of the Senate’s agenda.

Our results are largely consistent with our hypotheses and show non-PAS appointments benefit Presidential efforts to influence politics and policy. Choices over all aspects of such appointments are sensitive to broader political forces, especially those germane to PAS appointments, yet statutory constraints also matter. Overall, unilateral appointments advantage Presidents, who use them to overcome legislative opposition or to move hostile agencies in more favorable ideological directions. But checks on unilateral authority are observable. Hence, the factors influencing lower-level appointment choices are consistent with their having political and policy importance and have implications for the President’s ability to steer the massive ship of the American bureaucracy.

Literature to date

Research on political appointments, particularly over the last decade, is considerable. As indicated, most analyses focus on PAS positions, with far less on lower-level appointments and their relationships to the overall appointment process. Much of the non-PAS literature studies Schedule C positions as proxies for Presidential politicization strategies (Bonica et al. 2015; Hollibaugh et al. 2014; Hollibaugh 2015a; Lewis 2007, 2008; Limbocker et al. 2022; Lowande 2019; Moore 2018), with Lewis (2008; see also Waterman and Ouyang 2020) including noncareer SES managers.

As insights from studies of PAS appointments help guide our empirical analysis, a brief review before turning to scholarship on unilateral appointments is helpful. This literature predominantly focuses on bargaining between the President and the Senate and determinants of observed outcomes (for theoretical perspectives, see Chiou and Rothenberg 2014; Hollibaugh 2015a; Jo 2017). A non-comprehensive list of empirical emphases includes determinants of the duration before appointments are acted on (Hollibaugh and Rothenberg 2018; McCarty and Razaghian 1999; Ostrander 2016); whether competence, ideology, or patronage are key (Hollibaugh et al. 2014; Hollibaugh 2015b; Krause and O’Connell 2016; Ouyang et al. 2017); how stages from vacancy to nomination to disposition interact (Hollibaugh and Rothenberg 2017, 2020); the roles agency characteristics and position placement within the hierarchy play (Chiou and Rothenberg 2014; Hollibaugh and Rothenberg 2017; McCarty and Razaghian 1999; Ostrander 2016); and how polarization and divided government structure the choice process (Chiou and Rothenberg 2014; Hollibaugh and Rothenberg 2017; Ostrander 2016).

From the sparser literature, statistically assessing politically appointed SES members or Schedule C positions emerges suggestions about politicization and the potentially relevant underlying factors. To reiterate, examining associated claims has been constrained by data unavailability.

All of this research, in contrast to much descriptive work, makes a number of implicit assumptions: unilateral appointments are policy-relevant, PAS appointee actions and agency policy can be influenced and constrained by their underlings, and the President finds making these choices strategically worthwhile. For example, in one of the few pieces explicitly focusing on noncareer SES appointees, Ban and Ingraham (1990) suggest noncareer appointees are typically placed into government because they (1) represent important constituencies or have strong ties to party power bases; (2) are experts who supported the party while out of office and are ready to move into positions when their party takes the White House (also see Light [1987]); or (3) constitute a means for Presidents seeking to change an agency’s or program’s direction to choose ideologically compatible appointees with the goal of “shaking things up.”

Moore (2018), studying Schedule C appointments (from 1998 to 2013), finds such selections are employed with greater frequency for agencies ideologically aligned with the President and when legislative polarization is greater. While she emphasizes ideological concerns (though her focus is on intra-Senate polarization), partisan dynamics are a minor footnote. Moore (2018) notes this is “a departure from some previous literature, where divided government is frequently an independent variable” (84). This is especially curious if, as Moore argues, unilateral appointments provide an alternative route for Presidents to politicize agencies if they are unable (or less able) to achieve politicization through advise and consent channels. Thus, considering partisanship within the separation of powers context – and other determinants of PAS appointments – is likely essential for understanding unilateral appointment dynamics.

Bonica et al. (2015) also explore whether confirmation obstacles might affect the unilateral appointments process, claiming committee chairs block extreme PAS nominees in favor of moderates but Presidents counter by naming extremist Schedule C appointees. While not analyzing what determines the amount of excepted appointments, their theory predicts PAS and unilateral appointment processes are related.

Therefore, we build most directly on Moore (2018) and her claim that excepted positions functionally substitute for PAS appointments in some circumstances – as well as on Bonica et al. (2015) and their arguments about the effects of the PAS process on unilateral appointments. ⁶ While doing so, we acknowledge West and Cheon’s (2019) argument that appointments processes with and without Senate confirmation are unrelated, instead maintaining that most excepted positions are allocated consistent with the desire to serve agency heads’ managerial goals; in contrast, we allow political and managerial dynamics to be reinforcing and speculate that both types of dynamics are at play.

The need to dig deeper into unilateral appointments and possible relationships to Senate-approved appointees is underscored by the literature’s mixed findings regarding the germaneness of Presidents’ policy preferences versus other factors. For example, Ban and Ingraham (1990) argue that the Reagan Administration’s appointment patterns for noncareer SES positions were – at least in the first term – consistent with a “counter-staffing” model, wherein an administration selects ideologues who might help them “change radically the direction of an agency or a program” (110). Ingraham et al. (1995) note more heterogeneity in Presidential staffing patterns when investigating the Reagan and Bush I administrations. Reagan’s noncareer SES appointments focused on central staffing agencies – ostensibly to “achieve control over bureaucratic processes” (266) – and the aforementioned “counter-staffing model,” while the Bush I administration’s targeting of the General Services Administration and the Departments of Defense, Energy, and State might have been related to the “large amount of funds disseminated by these agencies for consulting contracts” (268). ⁷

More recent literature – West and Cheon (2019) aside – has emphasized the President’s primacy in the excepted appointments process. Lewis (2008) notes Clinton cabinet officials were explicitly told “the Senate-confirmed positions, the noncareer SES positions, and the Schedule C positions” were to be selected and appointed by the President himself (24). The Bush II administration continued this trend, with a rule “requir[ing] the presidential personnel office to approve all noncareer appointments, including those in the Senior Executive Service and all the Schedule Cs” (Patterson 2008, 101). More broadly, Lewis and Waterman (2013) argue one of the “most important trends in the administrative presidency include increases in lower-level appointees and more careful selection of appointees at these lower levels,” as these appointees play “crucial role[s] in presidential and agency politics and policy making” (37).

To reiterate, data constraints might confound these different findings, with limits in measures available for agencies and appointment type and the timeframe that can be covered. Our analysis differs in these respects. We include both categories of unilateral Presidential appointments, use measures improving on earlier analyses, and analyze a wide range of agencies for over 30 years of the personnel system put into place after 1978.

Theoretical and empirical expectations

Based on the foregoing, while other factors may also be important, our reading of the literature suggests excepted positions are, at least to some extent, used to further Presidential policy objectives (Ban and Ingraham 1990; Bonica et al. 2015; Hollibaugh 2015a; Hollibaugh et al. 2014; Ingraham et al. 1995; Lewis 2007, 2008; Limbocker et al. 2022; Lowande 2019; Moore 2018). Further, it indicates that Presidents maintain at least some control over their allocations (Lewis 2008; Light 1995; Patterson 2008; Waterman and Ouyang 2020). Together, these features suggest Presidents have the capacity and incentives to deploy such appointments more often when appointments not subject to such control are stymied. However, managerial concerns likely constrain the President’s ability and willingness to politicize (Krause and O’Connell 2016), and patterns across agencies likely reflect – at least in part – the preferences and abilities of those who might be tasked with managing such appointments (West and Cheon 2019).

Per the former, in line with speculations proffered by Bonica et al. (2015) and Kinane (2021) and others, we posit excepted appointments should be more common when conventional PAS appointments prove more difficult. Specifically, we derive two “political” hypotheses suggesting excepted appointments should be more common when (a) the President and the Senate are at ideological odds and (b) are controlled by different parties.

Ideological Conflict Hypothesis: The rate of excepted appointments should be increasing in the ideological distance between the President and the Senate.

Partisan Conflict Hypothesis: Excepted appointments should be more common when the Presidency and the Senate are controlled by different parties.

Another hypothesis, consistent with Moore (2018), is intra-party polarization in the Senate should make unilateral appointments more appealing relative to PAS appointments. During times of high polarization, greater collective action problems must be overcome.

Polarization Hypothesis: Agencies should have higher rates of excepted appointments when intra-Senate polarization is higher.

Beyond the Senate’s internal conflict, its workload should matter. Previous research has found high Senate workload is associated with faster nomination (Hollibaugh and Rothenberg 2017, 2020, 2021) and confirmation rates (McCarty and Razaghian 1999; Ostrander 2016). Relatedly, we anticipate several administrative mechanisms should lead to lower excepted appointment rates during high workload periods: (a) finite administrative capacity to vet appointees, so Presidents and their administrations focus on where they are likely to achieve the most success most quickly; (b) a narrow pool of potential nominees who could theoretically be placed in either a PAS position or an excepted position, depending on the Senate’s ability and willingness to act; (c) legal requirements for excepted appointments – and their need for supervisory positions to be filled in many instances – incentivizing the President to hold off on politicizing until PAS appointments are filled; and (d) some combination of the above, or something else entirely. Regardless of the exact mechanism, there should be fewer excepted appointments with high Senate workloads, leading to our hypothesis:

Senate Workload Hypothesis: Agencies should have lower rates of excepted appointments when the Senate is experiencing higher workloads.

Our next hypotheses deal with the relationship between agency ideology and excepted appointments. However, the literature is mixed in terms of expecting if there is a linkage between the two and, if so, exactly the nature of the relationship. On the one hand, some have found Schedule C and other patronage-style appointments more concentrated in agencies more ideologically aligned with the President (Hollibaugh et al. 2014; Moore 2018), largely because these agencies tend to be more compatible with the skillsets and expertise of those receiving appointments (Lewis 2008, 2009). Conversely, Ban and Ingraham’s (1990) counter-staffing model suggests the opposite, with excepted appointments used to reorient agencies in ways more ideologically compatible with the President (implying targeting of ideologically misaligned agencies); West and Cheon (2019) also note excepted appointment patterns during the George W. Bush administration consistent with counter-staffing (though a significant number of excepted appointees went to agencies not needing such an approach). Of course, both dynamics may be occurring simultaneously, with the impacts of the supply of potential excepted appointees and of an administrative desire to exert control working at cross-purposes. We therefore derive two hypotheses:

Presidentially Aligned Agency Hypothesis: Agencies ideologically aligned with the President should have higher rates of excepted appointments.

Presidentially Opposed Agency Hypothesis: Agencies ideologically opposed with the President should have higher rates of excepted appointments.

Additionally, if Presidents care about performance, more capable managers should be allocated more Schedule C appointees, since these types of appointees generally exhibit lower levels of competence relative to other types (Waterman and Ouyang 2020), and more skilled managers should be better at compensating for such deficiencies. The relationship between noncareer SES appointments and the managerial expertise of agency heads is somewhat more ambiguous, since these appointees tend to have higher levels of competence than Schedule C appointees. Thus, we only hypothesize about the relationship for Schedule C appointees:

Expertise Hypothesis: Agencies led by those with higher levels of managerial expertise should have higher rates of Schedule C appointments.

An agency’s centrality for bureaucratic control may also induce excepted appointments. Ingraham et al. (1995) noted Reagan’s excepted appointment strategy was consistent with a focus on central staffing agencies, ostensibly to “achieve control over bureaucratic processes” (266). Thus, “central staffing agencies” (the OPM, the OMB, and the GSA) should have higher rates of excepted appointments.

Staffing Agency Hypothesis: Central staffing agencies should have higher rates of excepted appointments.

Finally, time in the political cycle should be relevant. At the dawn of a new Presidency, there is likely to be a focus on PAS appointments distracting attention from situating individuals in lower-level positions. With a looming election a lower rate of excepted appointments due to dwindling supply is likely, as potential appointees will be reticent to incur job change costs for what may prove a short-term position while those occupying positions are incentivized to leave as the administration’s final days approach (Bolton et al. 2021; Doherty et al. 2019). Thus, we have hypotheses about what to expect near the beginning and end of a four-year presidential term:

Transition Period Hypothesis: Transition periods should be associated with lower rates of excepted appointments.

Looming Election Hypothesis: The time period shortly before a Presidential election should be associated with lower rates of excepted appointments.

Data, methods, and results

Key for assessing our hypotheses is the BuzzFeed Freedom of Information Act data. ⁸ Specifically, we use these data to measure variations in position type from March 1984 to March 2017, focusing on appointments to cabinet departments, the EPA, and the central staffing agencies. ⁹ We aggregate the data to the agency-quarter level to measure the prevalence of Schedule C and appointed noncareer SES personnel, defining two dependent variables distinguishing between unilateral appointment type: the proportion of total positions filled by those with Schedule C appointments and the proportion of SES positions filled by those with noncareer assignments. ¹⁰

Figure 1 displays the broad temporal trends for both position types. ¹¹ Patterns suggest several features needing to be accounted in our multivariate analysis. One is that, while the proportions of such positions are stable for the most part – the average proportion of noncareer SES positions hovering around the statutory cap of 10% of all SES positions (the bottom dotted line in the SES panel), the agency-specific number of SES positions usually capped at the statutory 25% cap (the top dotted line in the same panel), and Schedule C positions typically around one-tenth of 1% of all positions – there are “transition” periods of higher turnover when a Presidential election is near as well as when there is a new President (1989, 1993, 2001, 2009, and 2017), with turnover exacerbated when the new President comes from a different party. These dynamics are consistent with our Transition Period and Looming Election Hypotheses. Another is proportions across agencies vary significantly. Finally, some agencies appear to have some turnover when Presidents are reelected, though the relevant magnitudes are small compared to when executive branch partisan control changes.

Figure 1. Noncareer SES and Schedule C positions over time (March 1984 to March 2017).

For independent variables, we include features mentioned in our previous discussion. Two measures capture ideological differences between the President and the Senate: President–Senate Distance, the (normalized) absolute difference in Common Space (Poole 1998) scores between the President and the Senate median, and President–Filibuster Distance, the (normalized) absolute difference in Common Space scores between the President and the filibuster pivot. We incorporate partisan conflict by interacting each of these with Divided Government, the latter equaling one if the President and Senate median are of different parties, and zero otherwise.

We follow Moore (2018) in measuring Senate intra-chamber conflict by defining Polarization as the mean ideological distance between members of opposing parties in the Senate (measured at the Congress-level). Higher distance signifies greater conflict (and potentially greater collective action problems limiting the Senate’s ability to bargain with the President or confirm nominees).

We capture alignment between the President and each agency by contrasting their respective ideal points, although measurement is admittedly a bit crude. Specifically, President–Agency Alignment uses Clinton and Lewis (2008) agency ideology scores and Presidential partisanship. The former are intended to assess whether agencies have “policy views [that] … can be characterized as liberal or conservative” (Clinton and Lewis 2008, 5). Negative and positive agency ideology estimates correspond to liberal and conservative agencies. Thus, we calculate President–Agency Alignment by examining whether the scores’ 95% credible intervals contain 0; if they do, we set President–Agency Alignment to 0 and if not we either set President–Agency Alignment to 1 if the agency has a positive [negative] score and the President is a Republican [Democrat] or −1 if the agency has a positive [negative] score and the President is a Democrat [Republican]. Thus, 1 indicates President–agency ideological alignment, −1 ideological divergence, and 0 ideological ambivalence. ¹²

For agency head managerial expertise, we employ Krause and O’Connell’s (2016) Managerial Expertise measure. This variable captures public sector managerial skills inferred from appointees’ backgrounds at the times of their nominations and includes information on previous management experience and previous experience as an agency appointee, among other traits. Higher values should indicate higher levels of managerial expertise. ¹³

For Senate workload, we use Quarterly Roll Calls, the number of Senate roll calls during the relevant quarter. To reiterate, while workload might not seem an obvious factor for unilateral appointments, it may if PAS and unilateral appointments are intermeshed, as the chamber’s activity level affects the speed of PAS nominations and confirmations. ¹⁴

Central Staff Agency equals one if the agency in question is the OPM, OMB, or the GSA. As previously mentioned, Ingraham et al. (1995) note the Reagan administration specifically sought to politicize these agencies, and later administrations employed varying strategies toward them.

Per our hypotheses as well as Fig. 1, we account for transition periods with a series of dummy variables. New President equals 1 if the indicated report is for the first or last quarter of a President’s administration; New Presidential Party equals 1 if the report is for the first [last] quarter of a President’s administration and the previous [incoming] President is from the opposing party; and (to account for possible staff shuffling around reelection time) Presidential Reelection equals 1 if the report is for the last quarter of the President’s first term (except for George H. W. Bush’s) or the first quarter of the President’s second term. Looming Election equals 1 if the report is the last one (i.e. the September report) before a presidential election in an election year, and 0 otherwise.

Finally (though not directly related to our hypotheses), we include two control variables to account for potential staffing limitations – Previous Quarter Executive Schedule Percentage and Previous Quarter Noncareer SES Gap. The former constitutes the proportion of a given agency’s staff on an executive (EX) pay plan in the previous quarter (most of whom are subject to Senate confirmation). This roughly captures the idea that the Schedule C and noncareer SES processes are related to the PAS appointments process, while allowing us to remain agnostic about its expected impact; negative values might suggest PAS appointees and Schedule C/noncareer SES positions are substitutes, and positive values might reflect complementarities associated with high-level confirmed officials being statutorily able to supervise such lower-level positions. ¹⁵ The latter variable is defined as ten minus the percentage of SES employees in an agency (in the previous quarter) who are noncareerists; this roughly accounts for the 10% cap on such appointees. ¹⁶

We estimate our specifications with a binomial logistic regression model with random effects for agency and President. ¹⁷ Relative to alternative generalized linear model techniques, this method better accounts for varied agency sizes across the government and over time. ^{18 , 19} Table 1 presents the estimated results.

Table 1. Determinants of politicized positions, March 1984 to March 2017

Note: Binomial logistic coefficients. Observations are at the agency-quarter level and include random effects for agency and President. For Models 1 through 4, the dependent variable is the proportion of positions categorized as having a Schedule C appointment type, out of all positions. For Models 5 through 8, the dependent variable is the proportion of Senior Executive Service positions that are noncareer in nature. As these are binomial regression models, the agency-quarter observations are weighted by the total number of positions (for Models 1 through 4) or the total number of SES positions (for Models 5 through 8). Positive [negative] coefficients indicate the covariate is associated with higher [lower] proportions of Schedule C positions (for Models 1 through 4) or noncareer SES positions (for Models 5 through 8). Standard errors in parentheses. Two-tailed tests.

***p < 0.01,

**p < 0.05,

*p < 0.1.

Before turning to what these results indicate about our hypotheses, a few general comments are in order. Notably, across models there are consistent results pointing to the broad importance of outside political forces and context.

Consider Fig. 2, which shows the effects of one standard deviation changes in President–Senate Distance and President–Filibuster Distance on the predicted proportions of position type. In the latter, President–Senate Distance and President–Filibuster Distance are varied while other variables are kept at their observed values (in line with Hanmer and Kalkan’s [2013] observed value approach); 90 and 95% confidence intervals (denoted by thick and thin vertical bars about the point estimates) are also provided. ²⁰ As can be seen, President–Senate Distance and President–Filibuster Distance are consistently positive and significant, though the effect of ideological divergence between the President and key members of Congress depends on Senate partisan control. Under unified government, the expected proportion of politicized positions – of both types – generally increases with divergence. Results are less consistent with divided government, with politicization only slightly affected when President–Senate Distance is the ideological measure of interest, but decreasing considerably when President–Filibuster Distance is used.

Figure 2. Effects of President–Senate conflict on the proportion of politicized appointments.

Moving from Fig. 2’s predicted proportions to Fig. 3’s predicted number of appointees of a given type provides additional substantive context into how political forces operate. Note that the quantities in Fig. 3 are generated by calculating the predicted proportions when the indicated President–Senate Distance is set to one standard deviation above its mean, subtracting from it the predicted proportions when the indicated ideological measure is set to one standard deviation below its mean, and multiplying by the relevant number of agency positions at the time (i.e. the number of total or SES positions). ²¹ While this approach has the benefit of contextualizing the results in more concrete ways, it suffers from the limitation of large prediction intervals due to significant heterogeneity in agency size. Nonetheless, the results show the conditioning effects of unified versus divided government. For example, under unified government, increasing President–Senate Distance from one standard deviation below its mean to one standard deviation above increases the predicted number of Schedule C positions by approximately 16.14 and the predicted number of noncareer SES positions by about 4.44; when President–Filibuster Distance is used as the ideological measure of interest, the predicted increases are about 20.66 and 1.08, respectively. Under divided government, the results are somewhat less consistent – a result in line with Fig. 2 – with Schedule C positions either increasing by 2.23 or decreasing by 9.02 positions, depending on which ideological measure is used, and noncareer SES positions either increasing by about 2.96 or decreasing by about 4.57. ²²

Figure 3. Effects of ideological conflict on the number of excepted appointees of a given type, conditional on ideological divergence and partisan control of government.

Finally, Fig. 4 provides scope conditions for divided government’s effect. Specifically, it presents the predicted proportions under divided government minus corresponding proportions under unified government, all else equal. Therefore, it shows the marginal effect of divided government on the proportion of politicized appointments, conditional on ideological divergence and appointment type. Beyond again displaying 90 and 95% confidence intervals after using the observed value method to set covariates, we include two dashed vertical lines in each facet – the leftmost indicating the lowest divergence level under divided government, and the rightmost indicating the highest divergence level under unified government. By defining the ranges of ideological divergence under which switches from unified to divided government – all else equal – might be consistent with the data, these lines provide the scope conditions for divided government’s effect. They show that divided government’s effect is almost always negative when President–Senate Distance is used as the measure of interest, negative most of the time when President–Filibuster Distance is employed and the dependent variable is the proportion of Schedule C positions, and negative at the upper end of the overlapping range when President–Filibuster Distance is utilized and noncareer SES positions are examined. Hence, divided government usually inhibits Presidents’ abilities to politicize the executive branch, especially when the Senate and the President are at ideological loggerheads. Although this may seem surprising given the lack of required Senate confirmation, as Moore (2018) notes, this may represent Presidential reluctance to overstep unilateral authority for fear of Congressional blowback.

Figure 4. Effect of divided government on the proportion of politicized appointments (conditional on ideological conflict).

With these results, we can turn to assessing our hypotheses. Overall, there is some support for the Ideological Conflict Hypothesis, at least under unified government (and under divided government when President–Senate Distance is the measure of interest), with higher levels of politicization present when the President and Senate are at ideological odds. Conversely, there is no support for the Partisan Conflict Hypothesis.

Figure 5, which presents the predicted effects on the number of excepted positions (using the observed value method and results from Models 2, 4, 6, and 8) for all other independent variables significant in at least one model, allows us to assess our other hypotheses. ²³ For binary independent variables (New President, New Presidential Party, Presidential Reelection, and Upcoming Presidential Election), the results reflect predicted changes induced by a change from 0 to 1; for President–Agency Alignment, the results reflect predicted changes induced by a change from −1 to 1; and for all other variables, the results reflect predicted changes induced by a shift from one standard deviation below the mean to one above it.

Figure 5. Effects of other independent variables on the number of excepted appointees of a given type.

First, we find evidence for the Polarization Hypothesis, as our findings indicate that internal Senate conflicts induce the President to turn to unilateral appointments. While the President may dislike antagonizing the Senate, she acts when the Senate appears unable to do so. Thus, within-Senate polarization is associated with higher executive politicization, at least for Schedule C positions (with smaller and more inconsistent effects for noncareer SES appointees); regardless of whether the model is estimated with President–Senate Distance or President–Filibuster Distance as an independent variable, increasing Congressional Polarization from one standard deviation below its mean to one above increases the predicted number of Schedule C positions by between six and seven. Hence, our findings are consistent with the Polarization Hypothesis and Moore’s (2018) argument that unilateral appointments are more likely when such conflict is high, probably because of the reduced likelihood of successful advise and consent appointments. Having non-PAS political appointees in place becomes more important to the President.

Ideological compatibility and higher politicization are also associated with excepted appointments. Shifting the Clinton-Lewis version of President–Agency Alignment from −1 (President–agency misalignment) to 1 (alignment) increases the predicted number of Schedule C positions by about 2.9 and decreases the predicted number of noncareer SES positions by about 0.2. These results partially support the Presidentially Aligned Agency Hypothesis (for Schedule C appointees) as well as the Presidentially Opposed Agency Hypothesis (for noncareer SES appointees). Presumably, the distinction is due to the higher levels of competence of SES appointees (Waterman and Ouyang 2020), since they might be more likely to be deployed to agencies where career employees are likely to resist the President’s agenda; agencies whose missions align with the President are conversely seen as better targets for Schedule C appointees since career employees are less likely to resist Presidential directives (Hollibaugh et al. 2014). Thus, politicization is not purely a means of bringing in reinforcements to agencies the President finds problematic.

The Expertise Hypothesis also receives support. Agency heads with higher levels of managerial expertise are associated with higher politicization rates. Moving from one standard deviation below the mean value of Managerial Expertise to one standard deviation above increases the predicted number of Schedule C appointees by about 2 and the predicted number of noncareer SES appointees by about 0.7. ²⁴

Corresponding to the Transition Period and Looming Election Hypotheses, place in the political cycle also matters. Transition periods are consistently associated with lower politicization rates, as New President, New Presidential Party, Presidential Reelection, and Upcoming Presidential Election are consistently negative and significant. Regardless of which President–Senate conflict measure is used, shifting New President from 0 to 1 leads to a predicted decrease of about five excepted appointments (regardless of type) and shifting Presidential Reelection analogously leads to a predicted decrease of between two and three appointments (again, regardless of type); shifting New Presidential Party from 0 to 1 leads to predicted decreases of about 19 Schedule C positions and five noncareer SES positions, and doing the same for Upcoming Presidential Election leads to predicted decreases of about six or seven Schedule C positions and three noncareer SES positions.

Per the Senate Workload Hypothesis, when the Senate’s docket is full politicization rates are lower. Shifting Quarterly Senate Roll Calls from one standard deviation below its mean to one above leads to a prediction of about two fewer Schedule C positions per agency as well as one fewer noncareer SES position. Presidents may focus their strategies on PAS appointments when the Senate is working efficiently to “strike while the iron is hot,” and to concentrate on alternative strategies at other times. This logic not only corresponds to the workload hypothesis, but it is consistent with the finding that Presidents politicize more when intra-Senate polarization is high.

Notably, Table 1 indicates Central Staffing Agency is never significant (also generally true in the models in the Appendix), thus providing no support for the Staffing Agency Hypothesis.

Finally (though not related to our hypotheses), for noncareer SES positions the results suggest a “stickiness.” Agencies closer to the 25% noncareer SES cap in one quarter remain closer to the cap in the subsequent quarter. More specifically, a move from one standard deviation above the mean of noncareer SES staffing to one standard deviation below is associated with about 24 fewer noncareer SES positions in an agency. ²⁵

Conclusions: politicization in the bureaucratic trenches

As recent years have borne out, perhaps more dramatically than ever, bureaucratic politicization is central for understanding policy choices and control. In the USA, much of the process involves personnel. But how unilaterally appointed agency positions are employed as a part of a broader Presidential strategy for different agencies and in varied political contexts has remained enigmatic until now.

Our analysis, using more comprehensive data than employed previously, suggests politicization not only involves PAS appointments but even those lower in the hierarchy whom the President can name unilaterally. In fact, PAS appointments and unilateral appointments are intertwined. Presidents use noncareer SES and Schedule C appointments to promote their interests but, as with many unilateral choices, feel constrained politically.

We demonstrate politicization varies for predictable reasons, with multiple factors influencing the process. These include differences between the President and the Senate, internal Senate dynamics, partisan conflict, President–agency relations, the managerial expertise of PAS appointees, and the timing of political events. The appointment of non-PAS positions is related to these political conditions and choices, with significant policy implications. Overall, our findings align more closely with Moore’s (2018) perspective on the political process shaped by the separation of powers rather than Ban and Ingraham’s (1990) or West and Cheon’s (2019) emphasis on internal administrative preferences, though the general support of the Expertise Hypothesis suggests these latter factors still possess some influence.

Emblematic are President–Senate relations. These are key for politicization, with higher ideological divergence between the President and the Senate’s key actors resulting in greater politicization, at least as measured by the concentration of Schedule C positions within the federal workforce and the relative proportion of noncareer SES appointees to the entire SES. This relationship is found under both unified and divided governments when the ideological distance between the President and the Senate median is of interest, as well as under unified government when we examine the distance between the President and the filibuster pivot. We do uncover countervailing results with divided government and when the ideological measure of interest is President–Filibuster Distance. Moreover, we generally find divided government reduces politicization, though the results are again more consistent when President–Senate Distance is used instead of President–Filibuster Distance. Nonetheless, our results suggest Presidents respond to higher ideological divergence levels between themselves and the Senate by engaging in politicization, but they can be disincentivized from doing so by partisan conflict (as the latter may heighten the likelihood of Congressional backlash and negatively impact other Presidential priorities). ²⁶

Alternatively, the President politicizes when the Senate is at war with itself. When the upper chamber functions poorly – casting few roll call votes or exhibiting high interparty polarization – the President’s cost of politicizing via direct appointment (and potentially other means not studied here) is reduced and corresponding benefits are raised (though polarization is only relevant for Schedule C positions). Conversely, when the Senate is operating vigorously the executive holds back on unilateral politicization, presumably taking advantage of the chamber’s willingness to move things along, including PAS appointments.

As for where politicization occurs, one answer appears to be where, in some sense, it is unneeded – in bureaucracies aligned with the President. To reiterate, while this might seem counterintuitive, it corresponds to a world where politicization is conditioned by the quality of potential appointee pools.

Finally, “when” matters: Times when the President is preoccupied with other things – getting PAS appointments through, for example – or when a unilateral appointment might be less valuable (and the appointee may be less enthusiastic about accepting) produce lesser unilateral politicization.

Thus, to reiterate, our analysis – with more detailed data over a broader timeframe than previously available – demonstrates contextual forces matter even for appointments in the proverbial bowels of the bureaucracy. Unilateral appointments are a Presidential asset but a fettered one. The President takes Congress into account even for unilateral actions. ²⁷

Future research is needed to detail this process further, e.g., by delving deeper into whether different administrations employ different strategies toward lower-level appointees, as reported in qualitative studies (see also Waterman and Ouyang 2020). Another possibility would be to match specific supervisors with specific excepted positions and see if characteristics of the former help determine choices for the latter.

Additionally, it remains to be seen how the dynamics discussed herein affect policy implementation. Indeed, it is intriguing – and, perhaps, concerning – that determinants of excepted personnel positions are so closely related to those of Senate-confirmed appointments. Given our findings and that these positions play “crucial role[s] in presidential and agency politics and policy making” (Lewis and Waterman 2013, 37), it seems likely similar dynamics affecting the PAS process ultimately influence policy administration and implementation. While additional analyses are needed, our results indicate bureaucratic politicization’s effects run deeper than previously thought.