Staffing alone: unilateral action and the politicization of the Executive Office of the President, 1988-2004.

Lewis, David E.

There is a burgeoning literature on unilateral action (see, e.g., Cooper 2002; Howell 2003; Mayer 2001). This literature frequently assumes that once orders are written, they are implemented without difficulty. As students of the policy process know, however, implementation post-enactment can be the key to determining whether policies succeed or fail (Pressman and Wildavsky 1974). There is often a significant slippage between what presidents or Congress intended and actual policy outcomes. Slippage can occur for a variety of reasons including resource constraints, difficulty observing outcomes, the complexity of joint action, and task difficulty. Importantly, bureaucratic resistance in implementation can also influence implementation.

In trying to ensure successful implementation of policy objectives, unilateral action once again comes into play. Presidents have significant unilateral influence over whether the managers who implement policies are chosen by the president or filled by the merit system. Presidential choices about whether federal managers will be appointees or career civil servants not only have significant consequences for the policy content of the measures enacted, but also the competence with which they are implemented (Gilmour and Lewis, forthcoming; Heclo 1975, 1977).

Presidents are frequently criticized for "politicizing" some part of the institutional presidency, from the national security bureaucracy to the budgetary process to environmental reports. (1) Implicit in these criticisms is the claim that presidential attempts to secure ideological fealty fundamentally damage the target of their attention. The National Security Council (NSC) staff fall victim to groupthink, the Office of Management and Budget's (OMB) economic forecasts become less reliable, or Council on Environmental Quality reports are widely perceived as biased or vacuous. In many cases, the harm to agency neutrality and competence hurts not only the agency but also seemingly the president himself. Given the potential harm to the agency and the president, when is it in the president's interest to politicize agencies in the institutional presidency? This is an important topic given that we cannot understand the modern presidency apart from the institutional structures that have grown up around the persons who fill the office. We cannot understand national security policy without an understanding of the NSC and its staff. We cannot understand the president's role in the budgetary process or environmental policy without an understanding of his relationship with the Office of Management and Budget and the Council on Environmental Quality.

Thus, the power to determine the number of appointed positions is an understudied and important aspect of presidential unilateral power. In this article, I will describe the mechanics of how presidents alter the number of political appointees and explain why this is important for politics. I look specifically at the agencies in the Executive Office of the President (EOP). The article is divided into five sections. In the first section I describe what we know about politicization in the EOP. In the second section I describe the mechanics of how politicization occurs and in the third section I explain when presidents want to politicize. In the fourth and fifth sections of the article I examine politicization in the EOP with new data from the Office of Personnel Management and discuss what I find.

The EOP and Politicization

The EOP, created in 1939, is the structural basis of the institutional presidency. It comprises a system of presidential agencies created primarily to help the president perform congressionally delegated or constitutional responsibilities. Currently there are eleven agencies in the EOP. (2) They vary from agencies such as the Office of the U.S. Trade Representative to the Council of Economic Advisers or the Office of Homeland Security. Some agencies in the EOP such as the OMB have a long history, established routines, professional expert staff, and substantial institutional memory that carries over from one administration to the next. Others, such as the White House Office, turn over a large percentage of their personnel with each presidential transition and by so doing eliminate an important source of continuing institutional knowledge. (3)

Scholars historically have expressed concerns about presidential attempts to politicize EOP agencies and make them more like the White House Office. (4) Heclo (1975) is perhaps the most forceful in his critique of politicization. He argues that Nixon administration attempts to politicize the Budget Bureau/OMB decreased impartiality, hampered communication, threatened OMB's brokerage function, and endangered cooperation among OMB units. It also threatened continuity and institutional memory to the detriment of both the agency and the presidency.

Moe (1985), on the other hand, suggests that presidents seek a more politically responsive staffing system. He is dubious of the responsiveness of career employees to presidential direction. He claims that the president is primarily a politician and is less concerned with effectiveness than with a staff structure that is responsive to his political needs. He cites the White House Office (all employees serve at the pleasure of the president) as an example of a structure that better meets the needs of the president than the Bureau of the Budget (later OMB). He also claims that while presidents largely inherit the basic institutional framework of the presidency, they try to make it more responsive by "manipulating civil service rules, proposing minor reorganizations, and pressing for modifying legislation ... to increase the number and location of administrative positions that can be occupied by appointees" (1985, 245).

Some scholars suggest that presidential politicization is a general pattern that extends beyond the Nixon administration to other presidencies (see, e.g., Burke 1992; Hart 1995; Wayne, Cole, and Hyde 1979). Burke (1992) finds that politicization occurred particularly in the Nixon and Reagan presidencies. Hart (1995) concludes that politicization is a more general pattern from the 1970s forward. What is likely, and perhaps closer to what these and other presidency scholars believe, is that presidents seek more appointees in some cases and none or fewer in others. The task, then, is to theorize about when presidents will seek to politicize and then look at the empirical evidence to see whether patterns follow expectations.

The Mechanics of Presidential Politicization

Appointees versus Career Civil Servants

In the most general terms there are two types of positions in the federal government, positions in the career civil service and positions in what is called the excepted service. At the heart of the first type of positions--civil service positions--is a series of rules and regulations about how people can obtain federal jobs and what their rights are with regard to promotion, removal, and other personnel actions. Merit system principles demand that persons be hired, promoted, and fired only on the basis of merit rather than on other factors such as party membership, gender, or race. Persons initially establish their merit through competitive examination or in some cases appropriate background qualifications. Once a person's qualifications have been established, a determination is made about their eligibility for both position and pay grade. Persons employed under the merit system have a series of rights defined in the Federal Personnel Manual, most notably rights to notification and appeal in cases of adverse personnel actions such as demotion or removal.

More than half of all federal jobs are now "excepted" from the competitive service system described above. The excepted service is a residual category, catching all jobs that are not subject to the provisions of Title 5 of the U.S. Code (which describes the civil service). (5) The excepted service includes the different types of political appointments and positions in agency-specific personnel systems (which make up the bulk of the excepted service). (6) The president's ability to determine the number of positions in the first category of excepted service positions (appointments) is the focus of this study.

The determination of which jobs are filled by appointment is made largely by administrative actors with a few exceptions including, notably, Senate-confirmed (PAS) positions. (7) In most cases, determinations about how to fill positions (merit versus appointee) are made administratively in response to agency requests to the Office of Personnel Management (OPM; or earlier, the Civil Service Commission) and OMB (or earlier, the Bureau of the Budget). Most experienced personnel officers know how to use the appropriate terms of art to ensure that requests for position classification turn out as the White House or the agency request. (8) The directors of both the OPM and the OMB and their subordinates serve at the pleasure of the president, easing the way for the White House to get its way. The official forms requesting reclassification of positions by agencies include space for White House opinion and input as a way of signaling White House opinion and intent.

Four Politicization Techniques

There are four primary ways that presidents politicize administrative agencies. The most obvious way is to change the appointment authority of a position from a merit-filled position to an appointee-filled position. For example, when President Eisenhower assumed the presidency, he changed the staff secretary position of the NSC from a career position to an appointed position. Presidents also often layer appointees on top of existing organizational structures. For example, one of the significant differences between the old Bureau of the Budget and the new OMB was the addition of four program associate directors on top of existing examining divisions. (9) New appointees can more carefully monitor the career managers and assume some of the careerists' policy-determining responsibilities including budget preparation, personnel decisions, and other administrative responsibilities. A third way of politicizing is to reorganize either through the creation of parallel structures to get around an existing bureaucracy or through other forms of reorganization. Finally, presidents also use reductions in force (RIF) to alter the careerist/appointee balance in an agency. According to a general rule of "save grade, save pay," those career employees with the least experience lose their jobs first during RIFs but those who stay with more seniority are bumped down in position and often assume new or different tasks from what they were doing before. They often have to do more work for the same amount of pay and the new tasks they assume are frequently jobs not performed by people in their pay scale. These ripple effects increase attrition beyond that caused by the initial RIF. Several recent episodes involving EOP agencies follow this pattern, including President Clinton's treatment of the Office of National Drug Control Policy. (10)

How Presidents Make the Politicization Decisions

By virtue of their unique position, presidents have always felt pressure to manage the executive branch and their administrative tasks (see, e.g., Dickinson 1997; Hart 1995; White 1948). The problem for presidents is that presidential responsibilities and expectations exceed their individual abilities. Historically, presidents hired relatives or close acquaintances to help them manage the business of the White House. Modern presidents, however, have to rely on professional staff and parts of the institutional presidency. As with any principal-agent relationship, problems can arise between the president (principal) and his staff agencies (agents) due to divergent preferences and variation in agency capacity. Ultimately, presidents care about the outcomes that result from agency actions. The extent to which agencies will produce outcomes that presidents prefer will partly be determined by the extent to which the agency shares the president's preferences and has the capacity to carry out the president's wishes.

Both the degree of preference divergence and agency capacity are a function of the percentage of political appointees in an agency. Presidents can work to change an agency's policy preferences or increase capacity but often these efforts have countervailing effects. You cannot politicize without endangering capacity and there is some capacity that cannot be achieved through appointments only. Most of the existing literature suggests that some expertise or competence is attached to agencies with continuing professional personnel and that, all else equal, a higher percentage of professional career employees provides more competence. Those units in the EOP that are filled completely with political appointees will be the most likely to have the same preferences as the president but they will probably not be the most competent. Similarly, those agencies in the EOP that are staffed with a high percentage of civil servants will have a measure of expertise and institutional memory that are valuable but they will also be less likely to completely share the president's preferences.

Presidents must choose an optimal level of political appointees, a process I describe more formally in the Appendix. This optimal level can vary from agency to agency because agencies in the EOP have important differences apart from their policy preferences, competence, and ease of monitoring. Given the dynamic described above, a number of predictions can be derived about the degree of politicization in different EOP agencies on the basis of the president's policy views and the president's concerns for competence. In the first case, presidents are more likely to add political appointees to an agency when the agency has different policy views from the president. An agency can have different policy views from the president either because of the continuing personnel in the agency or because the agency's mission has some policy content. In the first case, holdover personnel from other administrations are less likely to share the president's preferences than personnel the president chooses directly. This fact is starkly illustrated by Richard Clarke, a careerist holdover on the NSC staff who proved later to have preferences that diverged significantly from those of the president. Clarke was asked to stay on the staff of the NSC in 2001 because of his expertise in counterterrorism. After leaving the administration, however, Clarke was publicly critical of the administration and its handling of the war on terror. More generally, however, each new administration expresses concerns about how holdovers are influencing both the loyalty and policy of an agency whether it is the "Busbies" or the "Clintonistas." Both the Eisenhower and Kennedy administrations were concerned about the loyalty of the Budget Bureau because of long tenures under the opposition party's leadership. (11) After tenures under one party's control, the top managers in agencies are more likely to share the policy preferences of the president they serve under (Aberbach and Rockman 2000).

Apart from personnel, presidents are going to have concerns about some agencies just on the basis of what they do. Some agencies have policy content embedded in their mission or enacting statute that makes them closer to one party's view than the other's. For example, President Reagan was suspicious of the Council on Environmental Quality because of its environmental focus. He politicized it by cutting its employment and adding a number of political appointees during his administration. President Kennedy was suspicious of the NSC apparatus set up by President Eisenhower and set about dismantling it during his administration. Whether through personnel or mission, some agencies in the president's orbit will be trusted more than others. When presidents fear that existing personnel and agencies do not have their interests in mind, they politicize.

H1: Presidents politicize when their preferences diverge from those of the agency.

While concerns about preference divergence increase the likelihood of politicization, decisions are also made with an eye toward preserving the competence of the president's staff. Indeed, the reason why Richard Clarke was originally kept on the NSC staff was his expertise in counterterrorism. Of course, some agencies are very sensitive to politicization while others can tolerate the addition of appointees without much consequence for performance. For some agencies, there is always a ready supply of qualified personnel to do the president's work. In others, qualified personnel are in short supply. The turnover associated with political appointees itself is enough to make presidents sensitive to adding too many appointees.

Sensitivity to politicization can vary across agencies and over time according to a number of important factors including labor markets and agency tasks. Sensitivity to politicization across agencies and across time should vary, among other things, according to the outside options available to agency employees. When outside options are numerous, presidents are less likely to politicize both because doing so will lead to the loss of high-competence careerists and it will be hard to find and retain qualified appointee replacements. For many persons, serving in government comes at some personal economic cost. These employees serve for the nonmonetary satisfaction of the job as well as for the money. Many could make more money in the private sector. Actions by the president to politicize can reduce both the monetary and the nonmonetary benefits of the job and make outside options look more attractive.

Politicization reduces the number of top policy and pay jobs available to careerists and reduces the nonmonetary benefits associated with policy influence. Of course, if there are no good outside options for careerists, presidents can politicize without as much concern for agency capacity because careerists are less likely to leave and a bad economy means there will be a larger pool of qualified replacements if career managers do decide to leave.

Another important factor in an agency's sensitivity to politicization is its task complexity. Agencies that perform mundane tasks such as coordinating paperwork are less sensitive to changing levels of political appointments because the jobs are easy enough that appointees or careerists can manage equally well. Agencies that perform difficult tasks are more sensitive to politicization because some jobs cannot be done or managed equally well by people from the outside of an agency. For example, some employees in the Office of Science and Technology Policy (OSTP) have to literally do or understand rocket science. Others such as the OMB are responsible for collecting data on the economy for use in complicated economic forecasting models. In some cases, agencies are sensitive to politicization precisely because the costs of task failure are extremely high. Very small variance in quality or the likelihood of making mistakes can have a large impact on outcomes.

This is not to say that important, complex, and sensitive jobs cannot be performed with aplomb by appointees. On the contrary, the personnel selection process can find excellent presidential staff to perform the roles they are assigned. Instead, I argue simply what presidents and presidency scholars have long known, which is that presidents need professional continuing career staff to help them do their job well. The extent to which they are willing to endanger the continuing institutional memory and expertise of the EOP depends upon factors such as preference divergence and the sensitivity of the different agencies at different times to politicization.

H2: Politicization decreases as the sensitivity of agency capacity to politicization increases.

Data, Variables, and Methods

In order to test the hypotheses above, I collected data on the number and percentage of political appointees in EOP agencies from 1988 to 2004. The data, which come from the OPM's Central Personnel Data File (CPDF), are only available reliably back to 1988. (12) Data from the CPDF include political appointment data on eight EOP agencies over this seventeen-year period--the OMB, the NSC, the Council on Environmental Quality, the Council of Economic Advisers, the Office of the U.S. Trade Representative, the OSTP, the Office of National Drug Control Policy, and the Office of Administration (OA). There are 134 agency-by-year observations on the eight agencies. The CPDF does not contain data on the White House Office, the Office of Policy Development, or the Office of Homeland Security. These three excluded agencies are staffed overwhelmingly by presidential appointees and, as such, figures used here will underestimate the true percentage of the EOP staffed by political appointees.

One other serious difficulty with the political appointments data is that while the OPM identifies the three main types of appointees--Senate-confirmed, non-career SES, and Schedule C--it does not identify what are commonly referred to as presidential appointment (PA) appointees, a separate category that dominates the staffing of the White House and the Office of Policy Development. This shortcoming of the OPM data is only relevant for analyses of the OA and the NSC, both of which are staffed substantially by PA appointees. Any counts of the number of appointees for these agencies will underestimate the true number of appointees. In this analysis, I focus primarily on the six EOP agencies without any PA appointments. I will, however, estimate some models that include the NSC and OA even given these shortcomings.

In Figure 1, I graph the average political appointee percentage in the EOP over time. No clear pattern exists across the whole EOP during this period. Politicization is neither increasing steadily nor dramatically influenced by the presence of Republican presidents. This is interesting given the common view that politicization is increasing over time and that Republican presidents politicize more than Democratic presidents.

Variables

My expectation is that increases in preference divergence between the president and an agency will increase the percentage of an agency's employees who are appointees. Measuring policy disagreement between the president and an agency is difficult because there is no direct way to measure agency preferences. Secondary indicators of likely disagreement between the president and the EOP do exist, however. Specifically, we should see an increase in preference divergence between the president and the EOP when there is party (or preference) turnover in the White House. (13) Each president influences the shape and composition of his staff agencies. When new presidents assume office, it is only natural that the existing structure and personnel be viewed with some suspicion. New presidents perceive that their preferences diverge from those of the continuing institutional presidency. In order to test whether preference divergence stemming from party turnover increases the percentage of appointees, I code 1993-1996 and 2001-2004 with a 1 and all other years a 0. These are presidential terms where party turnover has occurred.

My other primary expectation is that those agencies that are most sensitive to politicization will have the fewest appointees. Sensitivity to politicization varies over time according to the availability of good private sector options for agency employees. If agency personnel have good outside options, small changes in politicization can have a significant impact on the decision of careerists to stay in an agency or leave. Politicization can decrease the material and nonmaterial benefits of work, particularly for those who do not share the president's preferences. To measure the quality of outside options, I use real private sector business and professional wages. (14) As private sector wages increase, politicization should decrease. To measure policy area or task complexity, I include indicators (0,1) for each agency. I expect to see agencies with more complex tasks having fewer appointees.

Of course, there are a number of other factors that could influence changes in the percentage of appointees. I include a control for unified government (0,1) to account for the possibility that presidents politicize more freely when they have a majority in Congress. As suggested above, I include an indicator for whether or not the president is a Republican (0,1). It is possible, despite the evidence presented above, that politicization is a partisan rather than institutional phenomenon. Similarly, I include a trend term to evaluate whether politicization is increasing over time. I control for the dynamics within administrations by including indicators for year of the president's term. (15)

Methods

Estimating models on time-series-cross-sectional data using ordinary least squares (OLS) can be problematic because the errors can be correlated over time (errors in agency i at time t are correlated with errors in agency i at time t - 1), correlated across agencies (errors in agency i at time t are correlated with errors in agency j at time t), and different panels can have different error processes. Rather than estimating models where the levels of the independent variables predict the level of the dependent variable, I estimate a model of first differences,

[DELTA][y.sub.i,t] = [alpha] + [DELTA][x.sub.i,t] [beta] + [[epsilon].sub.i,t],

with appropriate lags in the differences where necessary. (16) The dependent variable is a difference and first differenced series are often stationary where normal time series are not. To account for contemporaneously correlated errors across panels, I report robust standard errors clustered on years. (17)

Results

In Table 1, I include OLS estimates of several models of presidential politicization. The estimates are generally supportive of the theoretical argument I put forward. Care should be taken in interpreting the results for a number of reasons, however. First, there is a relatively small sample size for a large number of parameters. The models are sensitive to changes in specification and technique. (18) Second, using data from the 1988-2004 period has the disadvantage of including only one Democratic president, two party changes in the White House, and two second-term presidents. While there are numerous cases, some of the variables only vary across years and not agencies. With these caveats in mind, I include multiple specifications of simply specified models.

Preference Divergence

In Table 1, party change in the White House has a consistent effect on percentage of appointees in EOP agencies. The coefficient on difference in party change is negative and significant in all of the models, indicating that politicization decreases in the first year after a party change. This is not surprising because it likely takes new presidents longer to fill out their teams than continuing presidents. This was certainly the case with President Clinton whose transition suffered from a number of missteps. The coefficient on the lag of the difference is positive and significant in all of the models, suggesting that politicization is significantly higher in the second year of a president's term if that president is from a new party. The estimates from Model 5 suggest that politicization will also be higher in the third and fourth year of a president's term after a party change. If, for example, a Republican succeeds a Democrat (or vice versa), the percentage of political appointees will be higher in years 2-4 than it would have been had a Republican succeeded a Republican. This is illustrated in Figure 2, which compares the cumulative change in appointee percentage over a president's term by whether or not there is party change in the White House. The dashed line tracks cumulative politicization after a party change relative to a baseline of 10 percent and cumulative politicization with no party change.

When party change has occurred, estimates from Models 1 and 5 suggest that EOP agencies will have about 1-7 percent more appointees. In a small agency with less than 50 employees such as the Council of Economic Advisers or the Office of Science and Technology Policy, this amounts to 1-4 more appointees at the management level. In a 100-person agency such as the Office of National Drug Control Policy, this is 1-7 more appointees. In a 500-person agency such as the OMB, this is 10-35 new appointees.

It is important to note one other aspect of these estimates. When no party change has taken place, the percentage of appointees actually decreases. This helps explain why there has been no long-term increase in the percentage of political appointees in the EOP during this period. During second terms or long rule by one party, presidents have fewer concerns about preference divergence between themselves and the staff that serves them. As a consequence they reduce the number of appointees.

Sensitivity of Agency Performance to Politicization

Model estimates partly support the expectation that the percentage of appointees in an agency will vary according to the sensitivity of the agency to politicization. Changes in private sector business and professional wages alter the president's politicization choices. When private sector wages get higher, presidents appear less likely to politicize. The coefficients on the private sector wage equations are all negative and significant in two of the models. Estimates suggest that a one-year increase in private sector wages of $0.5 leads to an immediate decrease in appointee percentage of 0.40 to 0.85. When one considers that private sector wages increased by $0.5 per year on average during this period, this is a large effect. If private sector business and professional wages went up $2 during a four-year term, the percentage of appointees in an agency with 10 percent appointees would be reduced to 8 percent by the end of the term.

Substantively, presidents appear to take into account the competence of their staff agencies when making politicization decisions. When outside options are available, they are more reticent to politicize, knowing that their actions could lead to harmful decreases in agency performance. Presidents are less likely to politicize when it could lead talented people to leave and when it will be hard to replace those competent employees with other high-capacity personnel. While presidents are sometimes willing to trade competence for responsiveness, they make politicization decisions constrained by concerns for competence.

Evidence from agency tasks is more equivocal. My expectation was that agencies dealing with complex policy areas such as regulation or science would be less likely to be politicized. Adding appointees to agencies with complex task environments would hurt these agencies more than other agencies. In Table 2, I list the average percentage of appointees in the EOP. The EOP agency with the lowest appointed percentage is the OMB while, somewhat surprisingly, the OSTP has the most. There is no discernible pattern in these percentages. Science and regulatory agencies such as the Council on Environmental Quality and the OSTP, which arguably have the most complex policy areas, have fewer appointees than the OA and the White House Office but more appointees than other agencies such as the Office of the U.S. Trade Representative and the Council of Economic Advisers. One possible explanation for the lack of a clear pattern is that the work of these agencies is really no more complex than the work of the OMB or Council of Economic Advisers. It is also possible that the careful selection of fewer appointees can produce the same effect as adding more appointees.

Interestingly, model estimates suggest that the percentage of appointees increases during periods of unified government, and partisanship, if it matters at all in the 1988-2004 period, leads Republicans to prefer fewer appointees than Democrats. The increase in the percentage of appointees during periods of unified government suggests that presidents may have more flexibility in staffing when their party controls Congress. The evidence that partisanship does not matter for politicization choices is surprising given what we know about the politicization of the Nixon and Reagan administrations (Heclo 1977; Nathan 1975). What is likely true and is suggested by the coefficient estimates of the party change variable is that politicization is a tactic used by presidents of both parties. What determines their use of this tactic at any given time is the extent to which their preferences diverge from those of the agency and how sensitive the agency's performance will be to politicization.

In total, the model estimates are mixed but noticeably supportive of the two hypotheses, particularly given the weakness of the measures employed. The president's perception of agency preferences systematically drives presidential politicization choices. Presidents are more likely to politicize agencies with divergent ideologies embedded in law or personnel. Politicization also increases when agencies' sensitivity to politicization decreases, such as when outside wages decrease.

Some of the difficulties with the data analysis could be improved in future research. First, some of the uncertainty arises from the small number of agencies in the sample. This can be resolved by looking outside the EOP to the larger federal bureaucracy (Lewis 2005). A larger set of agencies will allow for a more reliable coding of agency task complexity, for example. It will also allow for the examination of how party change in the White House affects a larger set of agencies. Second, some accounting should be made of how the appointment process itself influences the politicization decision. At the same time that presidents are making decisions about the number of appointees, they are selecting the appointees who fill appointed slots. Presidents screen appointees for loyalty but they also screen appointees for competence, particularly for key slots. One of the key factors in the politicization decision is how large the pool of acceptable appointees is. When the pool of qualified, loyal people is small, politicization is less attractive.

Conclusion

This article has examined an underappreciated source of presidential unilateral power, control over the number and depth of political appointments. Scholars are increasingly realizing that presidents unilaterally make choices that can have dramatic consequences for policy outputs. The influence of unilateral action is evident not only in policy formation but also implementation through choices about who will be responsible for the policies after enactment.

Choices presidents make about their staffing systems have consequences both for the policy content of decisions coming out of their staff agencies but also the quality of the decisions. Adding or subtracting the number of political appointments has affected the three agencies mentioned at the start of the article, the NSC, the OMB, and the Council on Environmental Quality. Each agency has had the number of political appointees increased and decreased at different points in their existence, with consequences for the credibility and usefulness of the agency. President Kennedy notably rearranged Eisenhower's NSC system, perhaps to his detriment, in his handling of national security early in his term. President Nixon famously reorganized the OMB and not long after, Congress created the Congressional Budget Office. President Reagan so eviscerated the Council on Environmental Quality that it practically lost its voice in environmental politics.

This has important implications for our understanding of the presidency. Heclo's (1973) concern in his landmark article was that the president's decision to politicize the OMB would lead to a loss of bureaucratic capacity. One important implication of this article is that presidents might still choose to politicize even though they understand that doing so hurts bureaucratic capacity.

A second implication is that decreasing bureaucratic capacity in the permanent institution of the presidency makes personnel selection critically important. Presidents can partly make up for losses of capacity through the careful selection of appointed personnel, particularly personnel who have past experience appropriate for their chosen job. This may be one reason why the recycling of personnel from past administrations is common practice. If presidents fail in their staffing decisions, however, the politicization of the modern presidency can have deleterious consequences for presidential policy making in many areas, including national security, economic, and trade policy.

Appendix

Formalization of Theoretical Argument

Assume that presidents care about policy outcomes and want agencies to act in ways that will produce outcomes that presidents prefer and assume that the president and the agency have single peaked and quadratic preferences over policy outcomes in a unidimensional policy space so that

[u.sub.i] = [-(x - 1).sup.2]. (1)

Policy outcomes are determined partly by the outputs produced by the agency. Agency outputs are a function of the agency's policy preferences and its competence. Specifically, the agency will try and set policy (e.g., prepare budget, write policy memo, negotiate trade agreement) exactly at its own induced ideal point, a', but cannot do so perfectly. The agency will try to implement a' but the actual outcome will be a" - o, where o) is a random variable so that

x = a' - [omega](q) (2)

and O) has a mean 0 and a variance defined partly by q, the percentage of appointments. There are a number of possible ways to model how q influences the variance but for simplicity assume var([omega])= mq. Substantively, this means that errors increase as q increases.

The agency's induced ideal policy, a', is a function of its inherent ideal policy, a, the president's ideal policy (p), and the percentage of political appointees (q):

a' (q, p) = [gamma](q) p + (1 - [gamma](q))a (3)

where [gamma](q) is increasing in q. The amount of presidential influence in the agency's ideal policy is a function of the percentage of political appointees in the agency.

Substituting the induced ideal point into the president's utility function gives

[u.sub.p] = -[[(1 - [gamma](q))(a - p)].sup.2] (4)

and this means that the president's expected utility is

E([u.sub.p]) = -[[(1 - [gamma](q))(a - p)].sup.2] - mq (5)

where mq is the expectation of the variance. The separability of the mean and the variance in the utility function is a property of the quadratic utility function, mq reflects the errors that agencies can make in trying to carry out their responsibilities. The variance is a function of q because I argue that bureaucratic capacity is partly determined by the career/appointee percentage. Presidents have utility for mq because they prefer that agencies have more capacity rather than less.

Presidents choose the optimal percentage of political appointees, [q.sup.*], so that the marginal benefits of decreasing preference divergence between the agency and the president equal the marginal costs of decreasing competence. The partial derivative with respect to q is

[partial derivative] u/[partial derivative] q = 2[(1 - [gamma])(a - p)][gamma]' (a - p) - m (6)

which when solved is:

[(1 - [gamma])[(a - p).sup.2]][gamma]' = m/2, (7)

This nicely sets the impact of preference divergence between the president and the agency against presidential concerns for agency capacity if m is understood as an agency's sensitivity to the percentage of appointees. If preference divergence increases, the optimal q increases. To see this consider that

[(1 - [gamma])[(a - p).sup.2]][gamma]' - m/2 = 0. (8)

By the implicit function rule

[partial derivative] [q.sup.*]/[partial derivative][(a - p).sup.2] = -[gamma]' (1 - [gamma]/[(1 - [gamma]) [(a - p).sup.2]][gamma]" - [gamma]'[(a - p).sup.2]

This suggests that [q.sup.*] increases as (a - p) increases. [gamma]' is > 0 because [gamma](q) is increasing in q. (1 - [gamma]) is > 0 by definition. Together this implies that the numerator is < 0. In the denominator the quantity is negative if [gamma]" < 0 because [(a - p).sup.2], (1 - [gamma]), and [gamma]' > 0 . (19) Both the numerator and denominator are negative, giving a positive partial derivative. As the distance between the agency and the president increases, so does the equilibrium level of political appointees (Hypothesis 1).

If the agency's sensitivity to the number of appointees increases, the optimal q decreases. To see this consider that

[partial derivative] [q.sup.*]/[partial derivative] m = - 1/2 [(1 - [gamma]/[(1 - [gamma])[(a - p).sup.2]] [gamma]" - [gamma]'[(a - p).sup.2] (9)

by the implicit function rule. When solved, this is

[partial derivative] [q.sup.*]/ [partial derivative] m = 1/2[[(a - p).sup.2] ([gamma]"(1 - [gamma]) - [gamma]']. (10)

The denominator is negative if [gamma]" < 0 because (a - p)2, (1 - [gamma]), and [gamma]" > 0. The last of these expressions is positive because [gamma](q) is increasing in q. If an agency's sensitivity to q decreases, the president's optimal level of appointees will increase (Hypothesis 2).

AUTHOR'S NOTE: I gratefully acknowledge the help of Nolan McCarty who got me started on this project and helped me to think more rigorously about it. Chris Achen, Larry Barrels, Joshua Clinton, Fred Greenstein, Shigeo Hirano, Will Howell. George Krause, Adam Meirowitz, Tom Romer, and the Breakfast Club at Princeton University spent more time than I deserve talking about this project and teaching me how to do it better Seminar participants at Princeton University and Harvard University provided many useful comments. The errors that remain are my own.

References

Aberbach, Joel D., and Bert A. Rockman. 2000. In the web of politics: Three decades of the U.S. federal executive. Washington, DC: Brookings Institution.

Berman, Larry. 1979. The Office of Management and Budget and the presidency, 1921 - 1979. Princeton, N J: Princeton University Press.

Burke, John E 1992. The institutional presidency. Baltimore: Johns Hopkins University Press.

Cooper, Philip. 2002. By order of the president: The use and abuse of executive direct action. Lawrence: University Press of Kansas.

Dickinson, Matthew J. 1997. Bitter harvest: FDR, presidential power and the growth of the presidential branch. New York: Cambridge University Press.

Fisher, Louis. 1975. Presidential spending power. Princeton, NJ: Princeton University Press.

Gilmour, John B., and David E. Lewis. Forthcoming. Political appointees and the competence of federal program management. American Politics Research.

Hart, John. 1995. The presidential branch. New York: Chatham House.

Heclo, Hugh. 1975. OMB and the presidency--The problem of "neutral competence." Public Interest 38(Winter): 80-98.

--. 1977. A government of strangers: Executive politics in Washington. Washington, DC: Brookings Institution.

Howell, William G. 2003. Power without persuasion: The politics of direct presidential action. Princeton, NJ: Princeton University Press.

Lewis, David E. 2005. Presidents and the politicization of the United States federal government, 19882004. Paper presented at the annual meeting of the Midwest Political Science Association, April 7-10, in Chicago.

Mayer, Kenneth. 2001. With the stroke of a pen: Executive orders and presidential power. Princeton, NJ: Princeton University Press.

McCarty, Nolan M., and Keith T. Poole. 1995. Veto power and legislation: An empirical analysis of executive and legislative bargaining from 1961 to 1986. Journal of Law, Economics, and Organization 11(2): 282-312.

Moe, Terry M. 1985. The politicized presidency. In The new direction in American politics, edited by J. E. Chubb and P. E. Peterson. Washington, DC: Brookings Institution.

Nathan, Richard E 1975. The plot that failed: Nixon and the administrative presidency. New York: Wiley.

Patterson, Bradley H., Jr. 2000. The White House staff Inside the West Wing and beyond. Washington, DC: Brookings Institution.

Poole, Keith. 1998. Estimating a basic space from a set of issue scales. American Journal of Political Science 42(July): 954-93.

Pressman, Jeffrey L., and Aaron Wildavsky. 1974. Implementation: How great expectations in Washington are dashed in Oakland. Berkeley: University of California Press.

Relyea, Harold, ed. 1997. The Executive Office of the President: A historical, biographical, and bibliographical guide. Westport, CT: Greenwood Press.

Tomkin, Shelley Lynne. 1998. Inside OMB: Politics and process in the president's budget office. Armonk, NY: M. E. Sharpe.

Wayne, Stephen J., Richard L. Cole, and James E C. Hyde, Jr. 1979. Advising the president on enrolled legislation: Patterns of executive influence. Political Science Quarterly 94(2): 303-16.

White, Leonard D. 1948. The federalists. New York: Macmillan.

DAVID E. LEWIS

Princeton University

(1.) For example, see Burke (1992) (National Security Council); Berman (1979), Fisher (1975), and Tomkin (1998) (Office of Management and Budget); and Lawrence Mosher, "Environmental Quality Council Trims Its Sails in Stormy Budget Weather," National Journal, July 24, 1982 (Council on Environmental Quality).

(2.) As defined by Relyea (1997). They include the White House Office, the OMB, the NSC, the Council on Environmental Quality, the Council of Economic Advisers, the Office of the U.S. Trade Representative, the Office of Science and Technology Policy, the Office of Homeland Security, the Office of National Drug Control Policy, the Office of Policy Development, and the Office of Administration. Some agencies we often think of as independent, such as the Office of Faith-Based and Community Initiatives and the USA Freedom Corps Office, are actually located inside the White House Office.

(3.) Technically, all persons employed in the White House Office serve at the pleasure of the president but many in the subsidiary offices, such as the Office of the Executive Clerk or the Correspondence Office, remain from one administration to the next (Patterson 2000).

(4.) Politicization can be defined by a number of related but distinct phenomena including (Heclo 1975): (1) the addition of political appointees on top of existing career civil service employees or the practice of placing loyal political appointees into important bureaucratic posts formerly held by career professionals, (2) choosing appointees only on the basis of loyalty, (3) involving civil servants in political fights, and (4) making promotion decisions on the basis of political attitudes. I focus here on the first definition of politicization.

(5.) There are five main categories of excepted jobs, positions requiring presidential nomination and Senate confirmation (PAS); positions requiring just presidential appointment (PA); jobs filled by persons in the Senior Executive Service (SES); positions in what are known as Schedules A, B, and C; and positions in agency-specific personnel systems. Out of the excepted service, PAS, PA, politically appointed SES, and Schedule C positions are politically appointed. The SES is a corps of federal managers that serve in management positions between PAS appointees and the traditional civil service. It is made up of a mixture of career managers (~90 percent) and political appointees (~10 percent).

(6.) Technically, the Senior Executive Service is part of the "competitive service" but the Office of Personnel Management counts SES positions as excepted. As defined by Section 2103 of Title 5, the "'excepted service' consists of those civil service positions which are not in the competitive service or the Senior Executive Service." Because the SES system is distinct from the provisions of the classical competitive system, it probably makes more sense to consider them as something distinct from this system. See U.S. General Accounting Office, The Excepted Service: A Research Profile (Washington, DC: U.S. General Accounting Office, 1997).

(7.) As with other areas of unilateral action, however, Congress can and does intervene on occasion. They can designate in statute how positions will be filled, limit the number of persons paid at executive levels, or generally write more restrictive language limiting presidential discretion. Even with Senate-confirmed positions, however, presidents have increasing authority, because many statutes use language authorizing a certain number of Senate-confirmed positions but not designating which positions must be so filled. This allows a president the ability to choose which positions he wants subject to Senate confirmation.

(8.) See the "Federal Political Personnel Manual" reprinted in Presidential Campaign Activities of 1972, vol. 19, Watergate and Related Activities, Use of Incumbency-Responsiveness Program. "Executive Session Hearings before the Select Committee on Presidential Campaign Activities," 93 Cong., 2d Sess (Washington, DC: Government Printing Office), 8903-9050.

(9.) See Heclo (1975); Heclo (1977, 78-81); U.S. Congress, Senate. Committee on Post Office and Civil Service, Policy and Supporting Positions, 86th Cong., 2d Sess. (Washington, DC: Government Printing Office), 19.

(10.) See Michael Isikoff, "Under Clinton, Drug Policy Office's Hot Streak Melts Down," Washington Post, February 10, 1993, p. A14. Of course, there were other intervening factors in the Office of National Drug Control Policy case because a reduction in employment in the EOP was connected to the National Performance Review. It is not surprising, however, that the administration chose the Office of National Drug Control Policy with its policy preferences to first cut, then change personnel, and then build back up again.

(11.) Heclo (1975).

(12.) According to the OPM, the CPDF codes changed during this period but many agencies did not adjust their self-reporting behavior consistent with the code changes, so there may be miscounting. An employee in the Office of Federal Civilian Workforce Statistics explained to me, "For any data submitted by agencies in 1988, there would probably be some discrepancy. Because the coding changed in the CPDF as of 1986, some agencies may not have changed their coding" (E-mail correspondence with British Morrison, Office of Federal Civilian Workforce Statistics, August 4, 2008).

(13.) I talk exclusively here in terms of party and partisanship, but have estimated models using preference measures in place of partisan measures and the results mimic what I find with the partisan measures (McCarty and Poole 1995; Poole 1998).

(14.) Source: Bureau of Labor Statistics (available from http://www.bls.gov). The series is average hourly wage, seasonally adjusted, for professional and business occupations.

(15.) I have also estimated models with a variety of different controls including the natural log of agency employment and the levels (rather than differences) of several of the variables included. These specifications generally confirm what is reported here.

(16.) I have estimated models primarily with differences or lags of differences where appropriate and chosen the models that fit the best, balancing the need to preserve degrees of freedom.

(17.) In all models with agency indicators, I cannot reject the null hypothesis that the coefficients on the indicators are 0.

(18.) While I have tried to let theory drive the model estimation, some judgment is always necessary. This is particularly true in determining which measures to use, the number of lags, etc. Readers are encouraged to obtain the data from me and examine it themselves.

(19.) [gamma]" < 0 implies diminishing marginal returns to politicization.

David E. Lewis is an assistant professor of politics and public affairs at Princeton University. He is the author of Presidents and the Politics of Agency Design (Stanford University Press, 2003) and a number of articles on American politics, public administration, and management.

TABLE I
Model of Presidential Politicization of the EOP, 1988-2004

Variables Model 1 Model 2 Model 3

Preference divergence
 [DELTA] Parry change in
 last election -0.85 ** -0.80 ** -0.81 **
 [DELTA] Party change in last
 [election.sub.t-1] 1.91 ** 1.64 ** 1.97 **
 Party change in last election (0, 1) -- -- --

Sensitivity to politicization
 [DELTA] Business and professional -- -1.25 --
 hourly wages
 [DELTA] Business and professional -2.73 ** -0.88 -1.40
 hourly [wages.sub.t-1]

Controls
 Year 1 -2.52 * -2.47 * -2.59 *
 Year 2 2.92 * 2.87 * 2.86 *
 Year 3 -1.46 -1.47 -1.36
 [DELTA] Unified government -- -- 0.45 **
 [DELTA] Republican president -- -- --
 Trend 0.09 -- --
 Constant 0.41 0.95 0.7

N (Obs, Groups) 82,6 82,6 82,6

Variables Model 4 Model 5 Model 6

Preference divergence
 [DELTA] Parry change in
 last election -0.81 ** -2.57 ** -0.54 **
 [DELTA] Party change in last
 [election.sub.t-1] 1.97 ** -- 1.41 **
 Party change in last election (0, 1) -- 2.42 ** --

Sensitivity to politicization
 [DELTA] Business and professional -- -- --
 hourly wages
 [DELTA] Business and professional -1.40 -- -1.26 *
 hourly [wages.sub.t-1]

Controls
 Year 1 -2.49 * -1.15 -1.79 *
 Year 2 2.84 * 2.80 ** 2.07 *
 Year 3 -1.37 -2.27 ** -1.01
 [DELTA] Unified government -- -- --
 [DELTA] Republican president -0.28 -- --
 Trend -- -- --
 Constant 0.58 -0.84 0.53

N (Obs, Groups) 82,6 82,6 110,8

Note: Dependent variable is [DELTA] political appointee
percentage. Model 6 includes executives in the excepted
service in counts of political appointees.

* Significant at the .10 level in one-tailed test.

** significant at the 0.05 level in one-tailed test.
Models with agency indicators are excluded. In all
such models, I could not reject the null hypothesis
that the coefficients on the indicators were equal
to 0.

TABLE 2
Percentage of Agency Employees that Are Appointees
(Senate-Confirmed, Senior Executive Service, or
Schedule C), 1988-2004

 Standard
Agency Mean Deviation Minimum Maximum

Council of Economic Advisers 16.65 4.89 6.67 25.93
Council on Environmental
 Quality 21.16 5.92 11.76 31.58
Office of Management
 and Budget 5.36 1.05 4.22 8.53
Office of National Drug
 Control Policy 19.90 11.40 2.65 45.05
Office of Science
 and Technology Policy 26.31 9.64 5.77 39.39
Office of the U.S.
 Trade Representative 9.14 1.71 6.47 11.76

Note: Excludes presidential appointments (PA).