4.1. The problem of excess corporate compensation

A principal hires an agent to do a job that the principal could not do as well or as cheaply himself. The principal wants the agent to strive to do the best possible job at the lowest possible cost. In other words, he wants the agent's incentives to coincide with his own. But the agent is a self-interested person just like the principal. Unless the principal can evaluate and monitor the agent's performance with great accuracy and adjust the agent's compensation accordingly, the agent is unlikely to be perfectly faithful to the principal. Compensation is ‘excessive’ in an economic sense when it is greater than it would be if agency costs were zero.

When tasks are complex, the cost of monitoring can be prohibitive; it does not make economic sense to incur $2 in monitoring costs to avoid $1 in agency costs. Measures for reducing monitoring costs include inculcating workers (lawyers for example) with professional norms, in the hope that they will be motivated to comply with those norms even when there is no supervisor looking over their shoulder: contracts; tort remedies, such as suits for breach of fiduciary duty; and criminal laws against fraud. Some organizations try in addition to create a high-commitment environment, in which workers identify emotionally with the organization's goals and thus internalize the organization's agency costs.Footnote ⁹

None of these devicesFootnote ¹⁰ fully solves the agency-cost problems of large business corporations. One observes for example that wages usually vary across the employees of the same rank in the same company by less than the differences in their contributions to the company and that employees who do satisfactory work can expect inflation-adjusted annual increases in their wages throughout their career with the firm, even though their contribution will not be increasing that fast and eventually will not be increasing at all. The first phenomenon – horizontal wage compression – is designed in part to reduce friction among workers and encourage them to cooperate with each other; to the extent they are paid the same, all will benefit from sharing know-how with each other. Moreover, it is difficult to measure individual contributions to team production (Frazis and Loewenstein, Reference Frazis and Loewenstein2006: 3; Bishop, Reference Bishop1987: S36, S37–S41); however, despite this difficulty, it often is apparent (and to the other members of the team, as well as to management) that one member of the team is the best. So he is promoted and given a higher salary and in this way, at least, marked differences in worker contribution to profit are rewarded (see Gibbons and Waldman, Reference Gibbons and Waldron2006: 93). Nevertheless, there is much room for slack, influence activities, and other manifestations of agency costs.

The second phenomenon – vertical wage drift – is intended in part to match income with consumption over the life cycle and in part to solve the ‘last period’ problem.Footnote ¹¹ A worker who is ‘overpaid’ as he nears retirement has a diminished incentive to slack off (as he might otherwise be inclined to do, having not much of a future with the firm), since if he does and he is fired he will lose his windfall wage. Forfeitable pension rights are a parallel device. But no one supposes the device works perfectly to eliminate agency costs, especially given legal requirements for the vesting of pension rights after a specified number of years of employment.

The monitoring problem can be especially acute in the case of the chief executive officer of a corporation. He is the agent of a principal that consists of the owners of the corporation, but when ownership is widely dispersed and highly liquid, as in the modern publicly held corporation (which has many shareholders, and the stock of which is traded on a stock exchange and therefore readily marketable), none of the part owners is likely to have an incentive to expend resources (in time, in effort, and perhaps in expenditures on advisors) on careful selection and monitoring of the CEO. The gains from such effort would largely be reaped by the other, free-riding, part owners; the modest gain to the monitor would therefore be less than his cost. If he is dissatisfied with how the corporation is being managed, selling his stock is likely to be his preferred alternative to trying to improve management.

A CEO's performance is especially difficult to evaluate because of the uncertainty that surrounds success as a business leader (see Bertrand and Schoar, Reference Bertrand and Schoar2003) and because his only ‘supervisor’ is the board of directorsFootnote ¹² – itself, as we shall see, an unreliable agent of the principal (that is, of the shareholders). The second point is critical. Even if the literature on performance-based evaluation of corporate executivesFootnote ¹³ yielded a reliable method of evaluating the performance of CEOs of large corporations, boards of directors would be unlikely to force it on the CEO because of the directors’ own agency costs.

While it is often assumed that agency costs must be higher in government agencies and other nonbusiness organizations than in private, profit-making corporations because of lack of product market and capital market discipline, we shall see that those disciplines are overrated as controls on corporate agency costs. A neglected point is that governmental and other noncommercial organizations have tools for limiting agency costs that business firms lack.Footnote ¹⁴ Some such organizations, such as those related to national security and to educational and charitable work, may be abler than a business firm to create a high-commitment organization culture. Wages are kept down by the nonpecuniary rewards of work motivated by a sense of commitment. They are also kept down by monopsony when the governmental or other noncommercial employer faces weak competition for staff. If you want very much to be a soldier, an intelligence officer, a forest ranger, or a judge, you have limited employment options. And low pay, paradoxically, can be a screening device for quality by eliminating from the applicant pool persons who are not highly committed to the employer's mission.

Professional schools inculcate norms that guide and constrain the work of lawyers, judges, accountants, military officers, teachers, and other professionals who populate noncommercial enterprises, as they do business ones – but professionals employed by business firms are controlled by business executives, and business itself is not a profession. Although legislative oversight of government agencies is often ineffectual, this is often true with respect to corporate oversight by boards of directors. And congressional committees at least have large professional staffs; boards of directors do not.

The point is not that noncommercial organizations are more efficient than business firms, they are less so, because of politicization. Their lesser efficiency is due in part to the fact that government, especially in this era of privatization of many formerly public functions, is assigned jobs that the commercial sector is unable to perform satisfactorily − government is a kind of supplier of last resort of services. My point is: although there are methods of limiting agency costs that do not depend on product market competition or capital market competition, they are largely unavailable to business corporations. Therefore, if product and market competition do not prevent executive overcompensation, nothing else is likely to do so.

The compensation of American CEOs is substantially greater than that of foreign CEOs, and an explanation congenial to organization economics is that stock ownership is more concentrated abroad (Thomas, Reference Thomas2004: 1186–1189; Roe, Reference Roe1993: 1936–1941).Footnote ¹⁵ The less dispersed ownership is, the less the efforts of shareholders to monitor the performance of their firm's managers will be undermined by free riding.

The board of directors is supposed to be the solution to the problem of agency costs when stock ownership is dispersed. But dispersion weakens shareholders’ incentive to base their vote for directors on careful scrutiny of the slate of proposed directors, especially since there is no competing slate; the electoral process resembles that in dictatorships. Monitors who are not monitored are imperfect agents of their principal.

The inference that boards of directors are ineffectual monitors of CEO compensation is supported by a number of observations besides the salary gap already mentioned. One is that the board is likely to be dominated by highly paid business executives, including CEOs (Bebchuk and Fried, Reference Bebchuk and Fried2004: 33). They have a conflict of interest, since they have a financial stake in high corporate salaries, their own salaries being determined in part by the salaries paid to persons in comparable positions in other companies. They also have a natural psychological tendency to believe that the high salaries of corporate executives accurately reflect the executives’ intrinsic worth (Bebchuk and Fried, Reference Bebchuk and Fried2004: 33). People are inclined to exaggerate their own merit (O'Callahan, Reference O'Callahan2006; Koeppel, Reference Koeppel2006). Many people feel underpaid; no one feels overpaid.

Directors devote only a fraction of their time to the company. And if they are inside directors (that is, full-time employees of the firm), they have a palpable conflict of interest; if they are outside (independent) directors, they have less access to information about the company than insiders and less of a stake in the corporation's success – which explains why there is no persuasive evidence that corporate performance is positively correlated with the percentage of independent directors on a corporation's board (Bhagat and Black, Reference Bhagat and Black2002). Some outside directors are persons of distinction who have no business experience.

CEOs influence the selection of outside as well as inside directors, and there is evidence of mutual back scratching – the directors authorizing generous compensation for the CEO and the CEO supporting generous fees for the directors (Brick et al., Reference Brick, Palmon and Wald2006). Corporate executives, moreover, hire and pay the auditors who certify the correctness of the corporation's financial statements, dangle consulting contracts in front of auditors who also offer consulting services, and can influence securities analysts’ reports by steering underwriting fees to investment banks whose analysts give their companies glowing reports.

Still another reason to doubt that boards of directors ride herd on CEO compensation is that they can shield themselves from criticism, should the firm perform poorly, by pointing out that they paid top dollar for the CEO. Presumably, therefore, he was the best candidate for the post and so in picking him they made the best choice they could have made and should not be blamed because he turned out to be a lemon. The generosity of the compensation package they gave him thus becomes evidence that he was indeed the best choice ex ante.

The consulting firms that boards of directors hire to advise them on selection and compensation of a CEO play to the self-protective instincts of directors by typically recommending that the board hire as CEO someone who will demand compensation in the 75th or higher percentile of the CEOs of the firms that the consultant has identified as comparable to the board's firm. The result is an upward ratchet in CEO compensation: the old 75th percentile becomes the new 50th percentile, making the new 75th percentile higher than the old one. The compensation consultants’ conflict of interest is augmented by the fact that (just as with accounting firms) the consultants sell other consulting services to the firm (Morgenson, Reference Morgenson2007; Lublin, Reference Lublin2007b).

In March 2007, a retailer called ‘The Sharper Image’ daringly broke with the pattern and deliberately hired a ‘cheaper’ candidate than recommended to it, Steven Lightman. He lasted only ten months in the job, during which time the price of the company's stock plummeted; and the company is now bankrupt (Lublin, Reference Lublin2007a; Reuters, 2008; Yahoo! Finance, 2008; Ackerman, Reference Ackerman2008). Boards of directors of other companies are likely to take ‘The Sharper Image's experience as a compelling example of the wisdom of the conventional approach to CEO compensation.

Still another problem is that, unlike the board of trustees of a university, a corporate board of directors has responsibilities that go far beyond the selection and monitoring of the CEO. (That, along with fund raising and the management of endowment funds, is the principal function of a university's board.) The corporate board participates in shaping corporate strategy. This makes it complicit in the CEO's decisions, and reluctant − by firing him or cutting his pay − to acknowledge a mistake for which it may share responsibility.

The inference that boards of directors are not policing CEO compensation adequately is further supported by the fact that the most significant ‘incentive’ component of CEO incomes – stock options – are not well correlated with a CEO's contribution to the value of his company (Hall and Murphy, Reference Hall and Murphy2003: 58–61; Core et al., Reference Core, Guay and Larcker2003: 27). Many things move a company's stock besides the decisions of its CEO. To tie his income to the value of his company's stock is a bit like tying the salary of the President of the United States to GNP. The analogy is close in an industry like oil. The profits of an oil company are largely a function of the price of oil, over which the companies have little control. In addition, tying an executive's compensation to the value of the corporation's stock creates an incentive to manipulate the stock price, and this incentive has been responsible for a number of financial debacles (see Hannes, Reference Hannes2008 and references cited there).

The choice of stock options as the principal method of providing nonsalary compensation to CEOs may be related to the fact that the income generated by these options, unlike salary or bonus income, did not, until recently (see Financial Accounting Standards Board, 2004), have to be, and usually was not, reported as a corporate expense. Of course security analysts, and stockholders having a stake large enough to follow closely the affairs of a company in which they invest, can calculate the expense of stock options. But the ordinary public cannot; and there is evidence that capital markets are not as efficient at pricing securities at the best estimate of their companies’ discounted present value, as economists and finance theorists used to think (see, for example, Shleifer, Reference Shleifer2000; Shleifer and Summers, Reference Shleifer and Summers1990). So just as the corporation's internal control mechanisms cannot squeeze all the agency costs out of the enterprise, neither can the securities markets.

Criticism of stock options is not criticism of trying to tie the CEO's fortunes to those of the corporation in order to align the agent's incentives with his principal's. But this could be done by requiring the CEO to place all his financial assets in stock of the corporation (Core et al., Reference Core, Guay and Larcker2003: 38). That would mitigate the conflict of interest between CEO and shareholders that arises from his being less diversified and therefore more risk averse than the shareholders would like him to be. Of course, he would have to be compensated for bearing the additional risk, but the additional compensation would be unlikely to be as generous as it is when it takes the form of stock options, requiring no investment by the CEO. Instead, it is an arbitrary decision by the board (his board, as it were) on how many options to issue to him and at what exercise price. Indeed, the stock-option method of compensating CEOs has been found to induce them to take excessive risks because of the asymmetry of gain and loss; there is no ceiling on the potential gain, but the loss is truncated at the value of the stock options (Sanders and Hambrick, Reference Sanders and Hambrick2007). And sometimes there is no loss at all, because the options are repriced, enabling the CEO to exercise his options at a profit even though the corporation's stock price has fallen below the options’ original exercise price. Some companies even backdated the award of stock options in order to guarantee that the options would be in the money,Footnote ¹⁶ a practice that obviously cannot be justified on incentive grounds.

Granted, repricing options and backdating options must be distinguished; the latter is the concealed version of the former. Repricing options, like other measures that reduce the riskiness of CEO compensation, can be defended on the ground that the riskier the compensation, the higher the competitive wage; and one way to reduce that riskiness and thus lower the wage is to reduce the tight coupling of pay to the market value of a company's stock (see Aggarwal and Samwick, Reference Aggarwal and Samwick1999). But the empirical evidence for such decoupling is weak (Prendergast, Reference Prendergast2002: 1077), and there is a counterargument: in the presence of uncertainty, a principal cannot evaluate the quality of his agent's work directly; he can only infer quality from the firm's output. So he is likely to base compensation on that output, even though the output is risk-laden and so requires him to pay the agent a premium for bearing risk. But if generous CEO compensation is intended to motivate CEOs to take risks on behalf of shareholders whose diversified portfolios make them less risk averse than the CEO, it seems odd to reduce the riskiness of the CEO's compensation by repricing his stock options in order to increase his compensation. How can the board have it both ways?

The most questionable compensation practice is giving the CEO an employment contract entitling him to generous severance pay (the ‘golden parachute’, as it is aptly called). This reinforces his incentive to take excessive risks by truncating the downside risk to him. At the same time, it signals his lack of self-confidence. The fact that a CEO would ask for a contract, rather than demanding higher pay to compensate him for forgoing it, should be a negative signal to the board.

A CEO cushioned against loss has an incentive to take high risks in order to maximize the expected value of his stock options. This may be a factor in the current financial crisis in the United States, a result in part of excessive risk-taking by corporate executives in the financial services industry. The greater the compensation paid the CEO, moreover, the greater his incentive to maximize the corporation's short-run profits, since he will make a great deal of money in the short run. Even if the firm later collapses, he will not have to repay the money he has been paid, unless much of his compensation was backloaded. In the wake of the financial debacle, corporations are now beginning to do this on a much larger scale than before the debacle shattered public faith that the market in executive compensation could be left to regulate itself.

One might think that competition in the corporation's product markets would constrain managerial greed, because that greed increases the corporation's costs. But the problem of agency costs is inherent in the structure of any large firm or other large enterprise; that is a basic insight of organization economics. Therefore, it is likely to plague all major competitors in a market and thus not be eliminated by competition, even if the markets in which the firms sell their products or services are highly competitive. Moreover, the effect of excess CEO compensation on the firm's welfare is indirect. The direct effect is merely to transfer wealth from shareholders to managers, although the indirect effect (apart from any distortions of managerial behavior) is to increase the firm's cost of obtaining new equity capital.

Another imperfect, though not wholly negligible, control of corporate agency costs is public opinion. Companies cannot afford to ignore it completely, because adverse public opinion can drive legislative or regulatory measures harmful to a company or an industry, as we are now seeing. There is no doubt that CEOs want to avoid media criticism (Dyck and Zingales, Reference Dyck and Zingales2002. See, for example, Morgenson, Reference Morgenson2006). A spate of newspaper and magazine articles has explained the ingenious devices (see Bebchuk and Fried, Reference Bebchuk and Fried2003: 79–81) by which CEO compensation − that would strike the average person as grossly excessive − is concealed from the public. This suggests that what the designers of CEO compensation packages are hiding from is not the shareholders, but the general public (Gordon, Reference Gordon2005: 679). A related suggestion is that the gap between European and American CEO compensation is due not to differences in governance structure, but to the fact that there is more envy in European societies. However, there is not zero envy in American society, and this has to influence corporate policy toward the level of CEO compensation, unless that compensation can be concealed from the public and the media.

4.2. The US national intelligence security community

An intelligence agency is a very different animal from a business firm. The difficulties inherent in optimizing an organization's performance are compounded when its output, not being sold in a market, is difficult to value. And this is a particularly serious problem for an intelligence agency. While market valuations are not the only ones on which rational decisions can be based – for example, the loss of a battle yields valuable information about an army's performance – the information about performance that a failure of intelligence yields tends to be equivocal. Even the best intelligence service will be surprised from time to time because the only way to ensure against ever being surprised is to ignore the costs of false alarms and thus bombard action-level officials with dire warnings, which, because of their frequency, are ignored until it is too late (‘the boy who cried wolf’ phenomenon). This does not mean that preventive efforts can never be evaluated. But there is a critical difference between competitive suppliers dealing with frequent threats, such as burglar alarm services, so that there is information on which to base an objective comparison of performance, and a monopoly supplier dealing with infrequent threats, which is an approximate description of a nation's intelligence apparatus. In the latter case, comparative performance evaluations will rarely be possible.

The difficulty of valuing an intelligence service's output translates into difficulty in valuing the performance of individual intelligence officers. This in turn retards efforts to prevent self-interested behavior that harms the enterprise – behavior such as warping intelligence estimates to conform to superiors’ preconceptions, making overly vague predictions, and, relatedly, shying away from issuing warnings – or overwarning in panic situations following a surprise attack. Think of what a joke the color-coded warnings issued by the Department of Homeland Security rapidly became. Too many warnings can, paradoxically, make people feel too safe.

These pathologies are compounded by the difficulty of recruiting able people to work in intelligence. Intelligence is a controversial, indeed a widely unpopular, career in a democratic society. The emphasis on secrecy, on leading a double life, turns off most people. The job pressures are extreme, there is no tenure protection, and melding a career in intelligence with normal family life is difficult. Promotion opportunities are truncated because the top layers of government are occupied mainly by political appointees rather than by career civil servants. The salary ladder is compressed, moreover, which makes it difficult to reward intelligence officers for taking risks, such as by making bold predictions.

The salary ladder is compressed partly by democratic political resistance to compensating government employees generously; partly because the government is, as I already said, a monopsonist hirer for jobs that do not have good substitutes in the private sector; and partly because, as in the military, promotions to senior positions (below the political level) in an intelligence agency are generally made from within the agency rather than by lateral hiring. Salaries must therefore be set high enough to attract people able enough to eventually fill the senior slots. It might seem that because those individuals cannot be identified at the outset, all beginning employees would be ‘overpaid’, making it difficult, given budget constraints, to pay appropriately high salaries to the senior employees. But, in fact, entry-level salaries in intelligence services are modest, suggesting that nonpecuniary income is a major element of the full income of an intelligence officer. (A further indication is that, despite the modest salaries, the Central Intelligence Agency receives almost ten times as many job applications as it has jobs to fill.) But this may increase the fragility of intelligence officers’ morale. The monetary salary may not be high enough to compensate for a loss of nonpecuniary income as a result of public criticisms, frequent shakeups in the managerial ranks, increased job pressures, or other adverse features of the work environment.

We know that hierarchy slows and distorts the information flow within an organization, and since information is the business of an intelligence agency and qualities of originality and imagination do not flourish in hierarchical, bureaucratized organizations, we might expect pressure to adopt a decentralized organizational structure for intelligence. The ultimate decentralization would be to transform the divisions of a decentralized organization into separate organizations, a process visible in the history of the intelligence services of the United States and other countries. The more heterogeneous the activities of the different divisions of an agency, the more likely they are to evolve into freestanding agencies. The design and launching of spy satellites, for example, require such different skills, procedures, and facilities from those required for intelligence analysis that combining the two functions, each essential to intelligence, in one agency (let alone one division) would yield no economies of scale or scopeFootnote ¹⁷ but merely increase the diseconomies of heterogeneous organizations. This point is overlooked in proposals to consolidate government functions in fewer but larger agencies.

The issue of the optimal degree of centralization is often presented as a choice between the ‘U-form’ (‘U’ for unitary) and the ‘M-form’ (‘M’ for multidivisional) organization. In the first format, functional divisions, such as research, production, and sales, are coordinated at the top of the organization. In the second, each division operates as a largely self-contained producer, such as the Chevrolet and Buick divisions of General Motors (Sloan, Reference Sloan1963); top management does not coordinate functions directly, but maintains control by monitoring the costs, revenues, and profits of each division and intervening in a division's operations when those indicators reveal inadequate performance.Footnote ¹⁸ The U-form is less nimble because information must travel up the hierarchy from the different functional divisions to be integrated at the top, and, after being converted there into directives, must travel back down again to the divisions for implementation. Hence the U-form is preferable for routinized activities that can be supervised by a handful of highly competent executives, the M-form for activities that require a high level of competence and initiative at the operating level as well. A pertinent risk created by the U-form, as by dictatorship, is deferring excessively to the judgment of a single person – the lesson of the failure of Israeli intelligence to anticipate the 1973 surprise attack by Egypt and Syria (Bar-Joseph, Reference Bar-Joseph, Betts and Mahnken2003).

The M-form seems the superior model for intelligence, but the US intelligence system cannot be organized as a pure M-form organization – a collection of intelligence services equivalent to the automotive-brand divisions of General Motors – because some of the services exist in the relation of supplier (or customer) to one another rather than producing similar outputs: the National Reconnaissance Office, which designs, builds, and launches spy satellites, is a supplier to the National Security Agency (the agency that spies on electronic communications), which in turn provides data to the CIA. Some of the intelligence agencies, however, produce similar, indeed competitive, outputs. For example, the CIA, the Defense Intelligence Agency, the National Intelligence Council, and the State Department's Bureau of Intelligence and Research are all engaged in what has aptly been termed ‘competitive analysis’ (Lowenthal, Reference Lowenthal2000: 13, 89).

The thrust of the reorganization of the intelligence system in the wake of the terrorist attacks of September 11, 2001, was toward more centralized control of the intelligence system and thus toward the U-form of organization. The office of Director of National Intelligence, created by congressional action in 1984 to manage the entire US intelligence system with the aid of a staff that has now reached 1,500, is a new layer of control placed over the existing conglomeration of organizations that constitutes the intelligence community. It was a problematic reform. In every organization, information is filtered as it rises in the organization's hierarchy (Arrow, Reference Arrow1974: 53–54) – the top would be flooded if it received all the information generated at the lowest level. The more layers there are, the less information will reach the top, where the decisions are made. Given the normal reluctance of intelligence officers to issue warnings of a surprise attack, the insertion of an additional layer is likely to cause a dilution of warning signals, impairing the ability of the policymakers to take timely action to prevent attacks.

An additional layer also lengthens the distance between the producers and the consumers of intelligence, threatening to cause dangerous delays in information flow, even if there is no information filtering by the additional layer. The additional supervisory layer also exacerbates the tendency of local military commanders – vital consumers of intelligence, for they may be the only officials in a position to thwart an attack – to ignore warnings from Washington. The greater the distance (cultural, organizational, information-transmitting) between center and periphery, the likelier the consumers of intelligence are to dismiss the analyst's assessment on the ground that he is too remote to be knowledgeable about local conditions.

It would be a mistake, however, to conclude that hierarchy in an intelligence system is a source only of problems, and not of solutions. Hierarchy enables the aggregation of information, though it can also delay and distort the flow. Each agent processes a bit of information, summarizes what matters or combines it with other information, and passes the summary up to the next level for the bits of information and analysis to be combined and assessed. Hierarchy also enables experts’ knowledge to be reserved for situations in which it is especially valuable. A bit of information from the field is, if routine, dealt with by a field officer. If the bit of information is unusual, it goes up the organizational ladder to a more knowledgeable officer who decides what action to take. Information that is truly exceptional continues up to the top of the hierarchy. This ‘management by exception’ allows for the optimal matching of problems with expertise.

Analysts and consumers of intelligence often require several pieces of information that confirm each other before they will believe a particular claim. But the analyst may not know whether what appear to be distinct pieces of information are truly corroborative or whether the same piece of information is being passed along through several different channels. In concluding before the invasion of Iraq that Saddam Hussein had weapons of mass destruction, the intelligence agencies relied on (among other things) multiple reports from Iraqi exiles. But rather than reflecting on independent sources of information, the reports originated from a single source – the Iraqi National Congress, an exile organization. Organization economists describe this phenomenon as ‘herding’.Footnote ¹⁹ The analyst may think he is confronting a consensus judgment, based on many sources, yet unbeknownst to him the judgment may be based on the same ultimate source and simply have been communicated to him through several channels. He may be acting rationally in agreeing with the consensus, even if it contradicts his own information, since he has only one data point, while the information that he is receiving from multiple sources consists of at least one data point and possibly of many.

A related phenomenon is ‘yes man’ behavior (see Prendergast, Reference Prendergast1999). If an employee's career depends on evaluations by his superiors, he will have an incentive to echo their opinions. Given this tendency, overt politicization of intelligence – that is, direct pressure from superiors – is not required for intelligence estimates to be produced that conform to political superiors’ possibly erroneous preconceptions. All that is required is that career rewards depend on performance evaluation by superiors and that the biases of superiors be known to their subordinates.

Centralization of the intelligence system is likely to exacerbate the ‘yes men’ problem by creating a tighter hierarchy, whereas in a system in which there are many bosses − even if the subordinates of each boss echo his views − there will still be many different views, as long as the bosses have different preconceptions. But one must be careful to identify who the ultimate bosses are. They are not the heads of the intelligence agencies, but the President and other top policymakers. If they have homogeneous views, the fact that the intelligence system is decentralized cannot be relied on to produce a diversity of views; all the agency chiefs will have an incentive to conform their advice to the policymakers’ preconceptions, whatever their private views.

Coordination among the different intelligence agencies is essential, but it is retarded because the agencies view their relationship as a competition for limited budgetary resources. In a competitive market, a firm thrives by satisfying consumer demands better than its competitors do. But a government agency thrives by persuading politicians to increase its budget. A bureaucrat who does not advocate forcefully for his agency risks stagnation or decline in the agency's budget, and even if his salary and job security are not affected, career satisfaction – a major form of reward for ambitious bureaucrats – will deminish. Hence, the phenomenon of the turf war.Footnote ²⁰ It is not unique to government, but it tends to be more virulent there because it is competition in advocacy rather than in selling better products at lower prices. If your competitor is selling a better product, passionate advocacy of your product, politicking, indignant denunciation of the competitor, and a blind commitment to your way of doing business will not bring you success. But if you are a bureaucrat, such behavior may win you a turf war.

Turf wars are not an entirely bad thing. One way to win a turf war is to convince, by honest argument and truthful evidence, the higher authority responsible for carving up the turf that your agency should be given the additional resources or power it seeks. Often, however, potential turf warriors decide they are better off colluding than competing, presenting higher authority not with a choice but with the bureaucratic equivalent of a division of markets. On this view, turf wars, like price cuts designed to discipline competitors who defect from a cartel, are punishments for breakdowns of collusion among competing agencies. This may explain the remarkable fact that, although military technology and threats keep changing, the shares of the US defense budget of the three major uniformed services are almost identical year after year.

Turf wars in government resemble turf wars between street gangs. In both cases, the competitors cannot rely on the usual market- and law-based methods − such as the creation and enforcement of property and contract rights − of obtaining resources, so they are forced to turn to alternatives that cost more (Levitt and Venkatesh, Reference Levitt and Venkatesh2000: 780–781).

A parallel phenomenon to the turf war, further illustrating the problem of serious tensions within an organization, is the diseconomies of scope encountered when inconsistent missions are combined in a single organization – intelligence and counterintelligence, for example. An intelligence officer uses spies or other means of penetration or surveillance to obtain what he hopes is truthful information about a potential threat, while the counterintelligence officer tries to show that the spy is a double agent whose information is disinformation. The counterintelligence officer is, in an expressive bit of unofficial bureaucratic lingo, the ‘office asshole.’ Another example of mission incompatibility within the same agency is the placement of both criminal investigations and domestic intelligence in the FBI, discussed in the next section of this paper.

There is mission incompatibility even in combining intelligence collection and intelligence analysis. An agent is reluctant to disclose his sources because disclosure might reveal a spy's identity, jeopardizing the spy's life should the identification leak, or to reveal that an enemy's code has been broken, which if discovered by the enemy would lead to the code being changed. Moreover, the type of person who operates undercover, runs spies, organizes coups, and engages in or manages other risky clandestine activities is likely to have a different psychology from a deskbound intelligence analyst: he is the man of action versus the man of thought; the warrior versus the intellectual. The difference is likely to breed distrust and misunderstanding, impeding communication and cooperation.

Code breakers and other technicians likewise have little in common with clandestine operators, who are likely moreover to dominate the agency (that has been the CIA's experience) (Szanton et al., Reference Szanton1976: 195–196) – and not simply because they are more aggressive. They are likely to be abler. People with a high level of analytic skills can find all sorts of attractive jobs in the private sector, making it difficult for an intelligence service to compete for them. People with a proclivity for clandestine operations can find good jobs in the private sector too (in part because of the demand by defense contractors for employees who are cleared to handle classified materials), but they cannot easily find jobs that yield the same nonpecuniary income that they derive from their intelligence career − that is an aspect of monopsony hiring.

4.3. The FBI

The question of the optimal combination of tasks within a single organization is well posed by the Federal Bureau of Investigation, which has two distinct tasks: investigating federal crimes (in this regard the Bureau is essentially a national detective bureau) and domestic intelligence, consisting of counterespionage and counterterrorism.

The Bureau is geographically decentralized; most of its employees are scattered among 56 field offices. A geographically decentralized organization is likely to be optimal for crime fighting as long as most crime is local or regional rather than national or international, since such a system is best at utilizing local knowledge. Most federal crime is local or regional. Decentralization motivates FBI agents to invest in local knowledge, confident that agents from headquarters will not swoop down and take credit for arrests and prosecutions. It also enables ‘yardstick competition’ – performance can be assessed by comparing the output of different offices.

But at the same time, decentralization undermines control: when agents can appeal to local knowledge as the basis of their decisions, the center, which lacks that knowledge, cannot monitor their decisions effectively. When the task is criminal investigation, this problem is not acute, because the outputs of FBI agents engaged in criminal investigation – number of arrests, prosecutions, convictions, length of sentences, and amount of property recovered – are quantitative, relatively hard to manipulate (at least legally), and therefore feasibly measured. FBI agents can thus be motivated by ‘high-powered’ incentives, that is by basing promotion and other career benefits on objectively measured, individual performance at the field-office level.

But a geographically decentralized organizational structure is suboptimal for intelligence. Foreign espionage and international terrorism are not local problems. Solving them requires the sharing of information across geographical boundaries when career incentives discourage sharing information with other offices that are viewed as competitors. Agents will resist improvements in organizational performance that decrease their own measured output (Dessein et al., Reference Dessein, Garicano and Gertner2007). The pre–9/11 FBI did not fail in the intelligence mission because of a lack of initiative by field offices, but because of a failure to coordinate the information generated in the different offices.

Moreover, arrests and convictions are a good measure of a detective bureau's performance, but not of an intelligence agency's performance. Intelligence tries to detect plots and threats before they reach the level at which they are prosecutable crimes. It thus aims at long-term penetration of suspect groups. And it doesn't worry much about whether the data it collects would be admissible as evidence in a criminal trial, but instead assumes that once subversive activity is discovered it can be disrupted by disclosure, disinformation, deportation, bribery, threats, and blackmail. Intelligence work actually disfavors criminal prosecutions, because they reveal to the enemy what the government knows and because trials can force the disclosure of secret information.

In fact, objective output measures of any kind are unlikely to be highly correlated with the desired output, which is successful intelligence work as just defined. Acts of terrorism are few and far between, and the earlier in the planning stage a terrorist project is interrupted, the less likely it is that, had the planning been allowed to continue, the project would actually have been carried through rather than abandoned. (Much terrorist ‘planning’ turns out to be just talk.) In addition, effective intelligence depends on the interlocking performance of many agents and organizations, rather than on an individual or a small team, as in the case of most criminal investigations; the individual's contribution is therefore very difficult to evaluate. Finally, the social benefits derived from intelligence activities consist not just in information obtained, but also in raising the costs of planning and organizing terrorist activity. Such benefits are exceedingly hard to assess, let alone to assign to particular officers or even agencies.

Given the inappropriateness of objective output measures, incentives that link promotion to success are likely to have perverse effects. Individuals will change their behavior in response to such incentives, but with results different from those sought by the principal if the performance rewarded is insufficiently close to the performance desired. Such dysfunctional responses are observed in intelligence. Intelligence analysts are frequently rewarded on the basis of the sheer number of intelligence pieces produced, which encourages them to focus on short-term issues that can be analyzed quickly (Commission on WMDs, 2005: 175), while operations officers are often rewarded on the basis of how many spies they recruit, which encourages them to focus on ‘foot soldiers’ who are easier to recruit than higher-ups and as a result unlikely to have good access to valuable information (Commission on WMDs, 2005: 159).

To avoid these effects, a rational organization will focus on monitoring the quality of its inputs rather than trying to link incentives to output. For an intelligence agency, this means an emphasis on careful hiring, but, once hired, intelligence officers who perform adequately in an unavoidably rather subjective sense should probably advance according to seniority. Departures from seniority should be limited to extreme cases of incompetent or superlative performance, since extremes can often be observed even when means cannot be.

So if criminal investigation requires a geographically decentralized organization with limited sharing of information and high-powered individual incentives based on outputs, but national security intelligence requires a geographically centralized organization with extensive sharing of information and careful screening of inputs with only low-powered incentives, the FBI's geographical, incentive, and information-sharing structure will assure suboptimal performance of one of the two missions. And when an organization has two career tracks, one of which bases advancement on objective criteria (as are feasible for criminal investigation), and the other does not, the abler employees will gravitate to the first track. For they know they will do well if evaluated according to objective criteria, but they cannot know with any confidence how well they will do if the criteria are subjective, since use of subjective criteria enables personal contacts and other ability-independent considerations to influence advancement. This is not a problem if the tracks are completely separate (for example, criminal investigation and orthopedic surgery), but criminal investigation and intelligence are similar enough to enable choice, especially when the Bureau's policy (probably a mistaken one) is to train all its special agents in both criminal investigation and intelligence.

4.4. The US federal judiciary

I turn my attention now to the organization of court systems, beginning with the appointed, life-tenured US federal judiciary. There are three principal tiers of federal judges: district judges (first-tier decision makers who, for example, conduct trials); circuit judges, who are intermediate appellate judges; and Supreme Court Justices. I shall focus on the first two tiers.Footnote ²¹

For the typical district judge, the appointment is final; the judge is unlikely to be promotedFootnote ²² and, therefore, unlikely to be constrained in his rulings by desire for promotion. He is also unlikely to resign and go into the practice of law or into some other line of work, even after reaching retirement age; most federal judges prefer upon reaching retirement age to take (as they are permitted to do) senior status, which allows them to judge part time yet receive their full salary. And only criminal or other egregious misconduct or incapacity will get the judge removed involuntarily. Moreover, not only is the judicial salary the same for all district judges – there are no bonuses for outstanding performance or financial penalties for poor performance – but a judge's ability to cash in on his judicial reputation by moonlighting as a teacher or lecturer is very limited, as there are low caps on outside earned income other than book royalties. It seems, then, that the federal judicial career has been carefully designed to insulate the judges from the normal incentives and constraints that determine the behavior of rational actors, except for the relative handful of judges who are ambitious for promotion to the court of appeals; and I shall ignore them. So it is as if the federal judicial career had been configured to perplex organization economists!

The mystery is at the practical rather than the theoretical level. Most decisions that a person makes are as limited in their impact on his utility as a judge's decision is limited in its impact on his utility. A person faced with a choice between two nearly identical items on a menu, such as two flavors of ice cream, cannot base the choice on the effect on his income or job security, yet his choice will be rational: it will be the choice that generates a larger net increment in his utility. But it may be very difficult to figure out why that particular choice is the one that has that consequence. And so it is with judges. The decision of a federal district judge will not affect his income or job security, but it will affect his utility in some other way – the question is in what way. The difference between this and the ice cream case is that the effect of the judicial decision on the judge's utility cannot be reduced to a single dimension, such as taste. Deciding a particular case in a particular way might increase the judge's utility just by the satisfaction that doing a good job produces. But it might also do so by advancing a political or ideological goal, economizing on the judge's time and effort, inviting commendation from people whom the judge admires, benefiting the local community, getting the judge's name in the newspaper, pleasing a spouse or other family member or a friend, galling a lawyer whom the judge dislikes, expressing affection for or hostility toward one of the parties – the list goes on and on.

Not only is there no effective mechanism for punishing a judge who yields to such temptations; that he has yielded can in most cases not even be detected. A trial judge's broad discretion in managing the timing and scope of the litigation before him, in ruling on objections to evidence, and in resolving factual disputes enables him to shape the factual premises of his decision so that the decision will appear to proceed logically from the facts. The process by which some facts are highlighted and others ignored or given little weight need not constitute a conscious program for making the judge look good. It may simply be the consequence of uncertainty. If an arresting officer says one thing and the person he arrested says something else, the judge's decision as to which one to believe is likely to be influenced, sometimes decisively, by the judge's background (was he a prosecutor before he became a judge? A defense lawyer?). Similarly, if one thinks back to the extraordinary variance in federal sentences that prevailed before the promulgation of the federal sentencing guidelines, one will find it difficult to resist the inference that the most important considerations in fixing a defendant's sentence within the limits permitted by law had nothing to do with legal analysis but everything to do with the judge's attitudes toward personal responsibility and the deterrent effects of criminal punishment.

As I noted in discussing intelligence organizations, an employer who cannot evaluate an employee's output directly will tend to base hiring, salary, promotion, and firing decisions on observable inputs instead, ex ante signals of quality. So one expects – and finds – that the more secure the judges’ tenure and hence the more difficult it is to control their behavior, the more carefully will candidates be screened by the appointing authorities; there will also be more competition for a judgeship, which will widen the field of selection. Assuming that people have generally stable preferences and that behavior has a strong habitual element, the older a person is when he is appointed to a job, the more predictable his performance in it will be. If a lawyer has performed successfully for many years in practice, demonstrating qualities of sobriety, good judgment, integrity, and other attributes that are important in a judge, the probability is greater that he will continue to display those qualities, even when the carrots and sticks of a legal practice are withdrawn. It is like firing a gun: the position and rifling of the gun's barrel impart direction to the bullet, but momentum takes over in guiding the bullet once it leaves the barrel, though wind or other environmental disturbances may deflect it from its initial path.

Moreover, when gross incentives and constraints are removed, a space is created for ones normally of only minor significance to determine the individual's behavior. This point has an upside as well as a downside, as far as constraining judicial behavior is concerned. People care about their reputation, which is often a function of the performance of other people; that is why rank orderings and prizes have psychological effects distinct from any career effects of being singled out from one's fellows. So, for example, federal district judges are sensitive to the quarterly statistics compiled by the Administrative Office of the US Courts showing how many cases the judge has had under advisement for more than a specified length of time – so sensitive that judges will sometimes dismiss a case at the end of a reporting period, with leave to reinstate it at the beginning of the next reporting period, in order to improve their statistics. Judges also don't like to be reversed (Stuntz, Reference Stuntz2001: 541; Caminker, Reference Caminker1994: 77–78; Higgins and Rubin, Reference Higgins and Rubin1980: 130), even though a reversal has no tangible effect on a judge's career if he is unlikely to be promoted to the court of appeals in any event.Footnote ²³

As judges are sensitive both to backlog and to reversal, they will not allow their backlog to grow to an inordinate length merely to reduce the probability of reversal, or their reversal rate to soar merely to eliminate their backlog by making precipitate rulings because then his reversal rate will rise to embarrassing level − they are constrained to be both careful and diligent, or, in short, conscientious. Backlog pressure keeps the judge working hard; reversal threat keeps him working carefully.

The analysis of federal circuit judges is broadly similar to that of district judges (see, for example, Georgakopoulus, Reference Georgakopoulus2000: 205–206), in particular with regard to promotion prospects, which indeed are even poorer for most circuit judges (the qualification in ‘most’ is explained below) than for most district judges, because the ratio of circuit judges to Supreme Court Justices (more than 20 to 1) is much higher than the ratio of district judges to circuit judges (less than 4 to 1). Since promotion is so limited at both lower levels of the judiciary (district judge and circuit judges) − even the lowest level is a highly responsible job requiring considerable experience to be performed competently − the federal judiciary is a lateral-entry organization. The average age of appointment to the district court has, since Harry Truman's Presidency, varied from 49 to 53, and of circuit judges from 50 to 56 (Yoon, Reference Yoon2003: 1047–1048, n. 70). A lawyer first appointed to a judgeship in his 40s or 50s is embarking on a second career.

Several differences, from an organization-economic perspective, between district and circuit judges should be noted. First, the dual constraints imposed by backlog pressure and reversal threat are attenuated. The caseloads of circuit judges are lighter than those of district judges, and so the threat of an unmanageable caseload looms less ominously. And once a case has been argued to the appeals panel, there will be no further activity in the case until it is decided, which means that backlog does not affect the workload of the judge as it does in the district court. In addition, so few court of appeals decisions are reviewed by the Supreme Court that the threat of reversal cannot operate as a significant constraint on circuit judges’ decisions – and for the additional reason that many reversals by the Supreme Court reflect ideological differences rather than error correction and therefore explicit or implicit criticism; this is less true of reversals of district judges.

Second, because appellate judges sit in panels rather than by themselves, there is a premium placed on cooperative behavior; the downside is the risk of factions and (although this is rare in the federal judiciary) of log rolling (i.e., vote trading).

Third, appellate judges have a greater opportunity to influence the direction of the law than trial judges do. One reason − apart from the obvious one that appellate adjudication focuses far more than at the trial level on general issues of law rather than on factual or procedural issues specific to the particular case − is that the Supreme Court reviews such a small percentage (currently less than 1%) of court of appeals decisions. Entire fields of law are left mainly to the courts of appeals to shape. Many court of appeals judges are not ambitious to influence the direction in which the law will evolve. Because the risk of reversal is small, these judges are likely to weight leisure more heavily, and to vote on the basis of their personal preferences more often than district judges do since the risk of reversal is so much greater, and the reward for creative legal thinking less, at the district court than at the court of appeals level. Like trial judges, and indeed more easily, appellate judges can conceal the role of personal preferences in their decisions by stating the facts selectively so that the outcome seems to follow ineluctably, or by taking liberties with precedents, where there are often interpretive leeways.

As for the minority of ambitious appellate judges, the principal constraint is stare decisis, that is the practice of adhering to precedent. In a rational-actor analysis, the constraining effect of precedent comes not from the fact that stare decisis is a sound policy but from the fact that a judge's influence is dependent to a significant degree on his decisions being treated as precedents by other judges. If he is cavalier about adhering to precedent in his own decisions, he undermines the practice of stare decisis in general and the likelihood that his own decisions will be followed by other judges in particular. What is more, deciding a case on the basis of a precedent rather than by reference to first principles is a time saver, which therefore serves a judge's leisure preference or effort aversion.

Another though overlapping minority of court of appeals judges should be mentioned: those who by reason of prominence, political connections, race or ethnicity, or other factors have a substantial prospect of being appointed to the Supreme Court. It is true that with rare exceptions the probability that a given court of appeals judge, however well placed he or she seems in the competition to be appointed to the Supreme Court, will actually be appointed is low. But if the judge attaches enormous value to being a Supreme Court Justice, the expected utility of such an appointment (most simply, the uitility of the appointment multiplied by its probability) may influence behavior. Thus, a study found that after Robert Bork's nomination to the Supreme Court failed, in part because of his extrajudicial writings (the largest component of the ‘paper trail’ that led to his defeat), the publication rate of court of appeals judges, after adjustment for other factors, declined precipitously (Gaille, Reference Gaille1997).

Criticism of judicial opinions by lawyers and law professors might seem an effective constraint on the basis of my earlier point that when conventional incentives and constraints are absent, weaker ones may have a significant effect in shaping behavior. But such criticism is easily dismissed as motivated by envy, politics, or simply a misunderstanding of the conditions under which judges work. The ineffectuality of such criticism has spurred the development of quantitative criteria of judicial performance, which are less easy to dismiss, because they are (or at least aspire to be) objective. But, like performance criteria for intelligence officers, the quantitative performance criteria for judges have limitations. For example, the performance criteria for circuit judges that have received the most attention, such as out-of-circuit citationsFootnote ²⁴ or the more inclusive quantitative ranking scheme developed and applied by Choi and Gulati (Reference Choi and Gulati2004a; see also Choi and Gulati, Reference Choi and Gulati2004b), measure influence and prominence (for example, one of their performance criteria is the number of times a judge is mentioned by name (see also Klein and Morrisroe, Reference Klein and Morrisroe1999), and in another study Gulati ranks judges by the number of their opinions published in casebooks (Gulati and Sanchez, Reference Gulati and Sanchez2002), and so implicitly treat judicial creativity as a desirable characteristic of circuit judges.

Not everyone will agree, however, and furthermore any multidimensional numerical rankings are questionable because the weighting of the different dimensions is critical yet tends to be arbitrary. The problem is particularly serious with respect to the ranking of judges, because there is no agreement on what are the most important dimensions of judicial performance. If you happen to think that lucidity is an extraordinarily important virtue of a judicial opinion, this will affect your weighting relative to someone who thinks that explaining carefully to the losing party why he lost, or discussing even minor issues, or stating the facts in comprehensive detail, is a more important virtue in a judicial opinion.

The federal judiciary should be of particular interest to organization economists because it is an example of extraordinary independence of agents from their principal. There is very little that a federal judge's principal (defined realistically as a combination of the legislative and executive branches of the federal government) can do to prevent the judge from exercising discretion in a way that the principal disapproves of. Little is not nothing. Judicial behavior can become a political issue, with the result that different types of judge may be appointed who may gradually isolate the judges whose rulings created the political issue, as happened in reaction to perceived liberal excesses of federal judges during the era when Earl Warren was Chief Justice. Another way to isolate or dilute the influence of judges deemed disloyal to the (current) principal is to expand the size of the federal judiciary.

Even so, judicial independence remains a striking example of agency costs, though agency costs are distinct from those found in commercial organizations. The ideological stakes in judicial decision making, in contrast to the strictly financial stakes of a business firm, implies that judicial agency costs will take the form not only of shirking (leisure preference or, equivalently, effort aversion), but also of ‘sabotage’. That is, judges will have a tendency to redefine the judiciary's mission to coincide with their personal ideological goals rather than those of the legislators whose handiwork the judges are supposed to interpret and apply neutrally (see, for example, Brehm and Gates, Reference Brehm and Gates1997; Kaiser, Reference Kaiser2003).

But there are offsetting benefits to the high agency costs of an independent judiciary – not only the obvious one of securing the impartial protection of property and personal rights, but also the subtler one of enabling the elected branches of government to duck many controversial issues because those issues have been confided to the independent judiciary. The elected branches avoid the obloquy of, say, acquitting an unpopular criminal defendant (who is however innocent) by pointing out that the defendant's fate is the hands of officials – the judges – whose decisions the elected branches cannot overrule.

A final economic observation to be made about the organization of the federal judiciary is that beginning in 1960, there was a rapid and steep increase in the amount of federal court litigation. The number of judges rose, but at a low rate; the slack was taken up by a disproportionately great increase in the number of supporting staff. This led to a change in output: a smaller percentage of cases disposed of after a trial. Trials are judge-intensive; you can't have a trial without a judge presiding over it. But cases decided without a trial, as by summary judgment or settlement, can often be handled by ancillary staff, with the judge's participation minimal. So those became more common methods of decision. This is an illustration of a general tendency for organizations to respond to increased demand by changing the character of their production.

4.5. Civil law judiciaries

The career judiciaries found in countries whose legal systems (‘civil law’ systems) do not have an English origin (‘common law’ systems) are, as the term ‘career judiciary’ implies, systems manned by lawyers who make an entire career of being a judge, as distinguished from the lateral-entry US federal judiciary. Appointment and promotion are by merit. Promotion is a critical feature of career judiciaries because a fresh law school graduate will naturally occupy the lowest rung of the judicial ladder when first appointed and will expect to rise to more responsible positions as he gains experience. There is little difference between a career judiciary and any other professional civil service, such as the diplomatic service or the armed forces. So the analysis of judicial behavior in a career judiciary should be essentially the same as the analysis of the behavior of public bureaucracies in general.

An essential control mechanism in public bureaucracies (since they cannot use monetized output to monitor, evaluate, and motivate employees) is clear rules to guide employees. The rules serve not only to maintain hierarchical control and thus reduce agency costs by reducing agents’ discretion, but also to provide a basis for evaluation and promotion. In the case of the career judiciaries in civil law systems, the rules are rules of law set forth in detailed codes. Detailed legal codes reduce judicial discretion by enabling and directing the judges to decide the vast majority of cases by a simple comparison of the text of the applicable code with the facts of the case. Judicial performance of this straightforward task can be objectively monitored by the judge's superiors as a basis for promotion because conformity to a rule is easier to determine than whether the judge is innovative or imaginative. That is why career judiciaries tend to be found in legal systems that rely heavily on detailed codes rather than on the looser standards that are characteristic of common law systems. The causation may go both ways: the decision to have a career judiciary may lead to the adoption of detailed legal codes.

Because of the hierarchical control necessary in a career judiciary (necessary because abundant promotion opportunities must be offered to persons spending their entire career as a judge, and desire for promotion places one at the mercy of one's superiors), the individual judge lacks the independence of a US federal judge. But the system as a whole is likely to have a considerable measure of independence from the other branches of government in recognition of the social and political value of judicial independence. As long as the judge's ‘dependence’ is limited to dependence on superiors who base promotion on nonpolitical criteria (accuracy, expedition, affirmance rate, and so on), judicial independence is adequately secured.

A career judiciary can be expected to be methodologically conservative and therefore unadventurous. Promotion in a career judiciary as in any other branch of the civil service depends ultimately on one's being able to perform to the satisfaction of one's superiors, and it is difficult to see how the supervisors in a career judiciary will benefit in their own careers from having bold, experimentally minded subordinates. It is not like a business firm, in which a division head's hard-driving, innovative subordinates may produce increases in revenues and profits that will redound to his credit for having selected and encouraged those subordinates. Furthermore, a career judge, having less exposure to the world outside the courtroom than a lateral-entry judge, is likely to lack the experience that he would need to play at rulemaking, as distinct from a rule-applying, role. The judge in a common law system is a rulemaker as well as a rule applier because common law is judge-made law, and though nowadays most of the law applied by common law judges is statutory law, often statutes are vague, especially American statutes, requiring creative judicial interpretation.

We can expect the output of a career judiciary to display low variance, to be of uniformly professional quality, but to be uncreative.Footnote ²⁵ It is no surprise that legal academics in nations that have career judiciaries are treated not merely as commentators on the law, as in Anglo-American legal systems, but as sources of law; the judges are unwilling to play such a role. With the structure of the judicial career and the heavy reliance placed on legal codes and treatise writers as sources of law, performance criteria that emphasized intellectual creativity and independence, political acumen, or even pragmatic insights into ‘law in action’ would be misplaced. A judge who excelled in such dimensions would be stepping out of his designated area.

The civil law approach to organizing the judiciary would be unworkable in the American legal culture. America's undisciplined legislatures can rarely produce legal codes that are sufficiently clear and detailed to minimize judicial discretion, and this makes lateral entry a more attractive method of staffing a judiciary than career appointments; lateral entrants are more likely to have a breadth of experience that enables them to make sensible discretionary judgments. The common law and civil law judiciaries are very different types of organization, but both may be well adapted to the legal systems in which they are found.