Principal-agent problem

From Academic Kids

In economics, the principal-agent problem in economics treats the difficulties that arise under conditions of incomplete and asymmetric information when a principal hires an agent. Various mechanisms may be used to try to align the interests of the agent with those of the principal, such as piece rates/commissions, profit sharing, efficiency wages, the agent posting a bond, or fear of firing. The principal-agent problem is found in most employer/employee relationships, for example, when stockholders hire top executives of corporations.



In economics, the problem of motivating one party to act on behalf of another is known as ‘the principal-agent problem’. The principal-agent problem arises when a principal compensates an agent for performing certain acts which are useful to the principal and costly to the agent, and where there are elements of the performance which are costly to observe. This is the case to some extent for all contracts which are written in a world of information asymmetry, uncertainty and risk. Here, principals do not know enough about whether (or to what extent) a contract is or has been satisfied. The solution to this information problem – closely related to the moral hazard problem – is to ensure (as far as possible) the provision of appropriate incentives so that agents act in the way principals wish them to. In terms of game theory, it involves changing the rules of the game so that the self-interested rational choices which the principal predicts the agent will make, coincide with the choices the principal desires. Even in the limited arena of employment contracts, the difficulty of doing this in practice is reflected in a multitude of compensation mechanisms (‘the carrot’) and supervisory schemes (‘the stick’).

Employment contract

In the context of the employment contract, individual contracts form a major method of restructuring incentives, by connecting as closely as is optimal the information available about employee performance, and the compensation for that performance. Because of differences in the quantity and quality of information available about the performance of individual employees, the ability of employees to bear risk, and the ability of employees to manipulate evaluation methods, the structural details of individual contracts vary widely, including such mechanisms as “piece rates, [share] options, discretionary bonuses, promotions, profit sharing, efficiency wages, deferred compensation, and so on.” (Prendergast 1999, 7) Typically, these mechanisms are used in the context of different types of employment: salespeople often receive some or all of their remuneration as commission, production workers are usually paid an hourly wage, while office workers are typically paid monthly or semimonthly (and if paid overtime, typically at a higher rate than the hourly rate implied by the salary). The way in which these mechanisms are used is different in the two parts of the economy which Doeringer and Piore called the “primary” and “secondary” sectors (see also dual labour market). The secondary sector is characterised by short-term employment relationships, little or no prospect of internal promotion, and the determination of wages primarily by market forces. In terms of occupations, it consists primarily of low or unskilled jobs, whether they are blue-collar (manual-labour), white-collar (e.g. filing clerks), or service jobs (e.g. waiters). These jobs are linked by the fact that they are characterized by “low skill levels, low earnings, easy entry, job impermanence, and low returns to education or experience.”

Non-financial compensation

Part of this variation in incentive structures and supervisory mechanisms may be attributable to variation in the level of intrinsic psychological satisfaction to be had from different types of work. Sociologists and psychologists frequently argue that individuals take a certain degree of pride in their work, and that introducing performance-related pay can destroy this “psycho-social compensation”, because the exchange relation between employer and employee becomes much more narrowly economic, destroying most or all of the potential for social exchange. Evidence for this is inconclusive – Deci (1971), and Lepper, Greene and Nisbett (1973) find support for this argument; Staw (1989) suggests other interpretations of the findings.

Team production

On a related note, Drago and Garvey (1997) use Australian survey data to show that when agents are placed on individual pay-for-performance schemes, they are less likely to help their coworkers. This negative effect is particularly important in those jobs that involve strong elements of ‘team production’ (Alchian and Demsetz 1972), where output reflects the contribution of many individuals, and individual contributions cannot be easily identified, and compensation is therefore based largely on the output of the team. In other words, pay-for-performance increases the incentives to free-ride, as there are large positive externalities to the efforts of an individual team member, and low returns to the individual (Holmstrom 1982, McLaughlin 1994). The negative incentive effects implied are confirmed by some empirical studies, eg Newhouse (1973) for shared medical practices (costs rise and doctors work fewer hours as more revenue is shared), and Leibowitz and Tollison (1980) find that larger law partnerships typically result in worse cost containment. As a counter, peer pressure can potentially solve the problem (Kandel and Lazear 1992), but this depends on peer monitoring being relatively costless to the individuals doing the monitoring/censuring in any particular instance (unless one brings in social considerations of norms and group identity and so on). Studies suggest that profit-sharing, for example, typically raises productivity by 3-5% (Jones and Kato 1995, Knex and Simester 1997), although there are some selection issues (Prendergast).

Empirical evidence

There is however considerable empirical evidence of a positive effect of compensation on performance (although the studies usually involve “simple” jobs where aggregate measures of performance are available, which is where piece rates should be most effective). In one study, Lazear (1996) saw productivity rising by 35% (and wages by 12%) in a change from salary to piece rates, with a third of the productivity gain due to worker selection effects. Paarsch and Shearer (1996) also find evidence supportive of incentive and productivity effects from piece rates, as do Banker, Lee, and Potter (1996), although the latter do not distinguish between incentive and worker selection effects. Fernie and Metcalf (1996) find that British jockeys perform significantly better when offered prizes for winning races compared to being on fixed retainers. McMillan, Whalley and Zhu (1989) and Groves et al (1994) look at Chinese agricultural and industrial data respectively and find significant incentive effects. Kahn and Sherer (1990) find that better evaluations of white-collar office workers were achieved by those employees who had a steeper relation between evaluations and pay. Nikkinen and Sahlström (2004) find empirical evidence that agency theory can be used, at least to some extent, to explain financial audit fees internationally.

Four principles of contract design

Milgrom and Roberts (1992) identify four basic principles of contract design:

  • the Informativeness Principle,
  • the Incentive-Intensity Principle,
  • the Monitoring Intensity Principle, and
  • the Equal Compensation Principle.

Informativeness Principle

In the absence of a world of perfect information, Holmstrom (1979) developed what became known as the Informativeness Principle. This essentially states that any measure of performance that (on the margin) reveals information about the effort level chosen by the agent should be included in the compensation contract. This includes, for example, Relative Performance Evaluation – measurement relative to other, similar agents, so as to filter out some common background noise factors, such as fluctuations in demand. By removing some exogenous sources of randomness in the agent’s income, a greater proportion of the fluctuation in the agent’s income falls under his control, increasing his ability to bear risk. If taken advantage of, by greater use of piece rates, this should improve incentives. (In terms of the simple linear model below, this means that increasing x produces an increase in b.)

Incentive-Intensity Principle

However, setting incentives as intense as possible is not necessarily optimal from the point of view of the employer. The Incentive-Intensity Principle states that the optimal intensity of incentives depends on four factors: the incremental profits created by additional effort, the precision with which the desired activities are assessed, the agent’s risk tolerance, and the agent’s responsiveness to incentives. According to Prendergast (1999, 8), “the primary constraint on [performance-related pay] is that [its] provision imposes additional risk on workers…” A typical result of the early principal-agent literature was that piece rates tend to 100% (of the compensation package) as the worker becomes more able to handle risk, as this ensures that workers fully internalize the consequences of their costly actions. In incentive terms, where we conceive of workers as self-interested rational individuals who provide costly effort (in the most general sense of the worker’s input to the firm’s production function), the more compensation varies with effort, the better the incentives for the worker to produce.

Monitoring Intensity Principle

The third principle – the Monitoring Intensity Principle – is complementary to the second, in that situations in which the optimal intensity of incentives is high correspond to situations in which the optimal level of monitoring is also high. Thus employers effectively choose from a “menu” of monitoring/incentive intensities. This is because monitoring is a costly means of reducing the variance of employee performance, which makes more difference to profits in the kinds of situations where it is also optimal to make incentives intense.

Equal Compensation Principle

The final principle is the Equal Compensation Principle, which essentially states that activities equally valued by the employer should be equally valuable (in terms of compensation, including non-financial things such as pleasantness) to the employee. This relates to the problem that employees may be engaged in several activities, and if some of these are not monitored or are monitored less heavily, these will be neglected, as activities with higher marginal returns to the employee are favoured. This can be thought of as a kind of “disintermediation” – targeting certain measurable variables may cause others to suffer. For example, teachers being rewarded by test scores of their students are likely to tend more towards teaching ‘for the test’, and de-emphasise less relevant but perhaps equally or more important aspects of education; while AT&T’s practice at one time of rewarding programmers by the number of lines of code written resulted in programs that were longer than necessary – i.e. program efficiency suffering (Prendergast 1999, 21). Following Holmstom and Milgrom (1990) and Baker (1992), this has become known as “multi-tasking” (where a subset of relevant tasks is rewarded, non-rewarded tasks suffer relative neglect). Because of this, the more difficult it is to completely specify and measure the variables on which reward is to be conditioned, the less likely that performance-related pay will be used: “in essence, complex jobs will typically not be evaluated through explicit contracts.” (Prendergast 1999, 9). Where explicit measures are used, they are more likely to be some kind of aggregate measure, for example, baseball and American Football players are rarely rewarded on the many specific measures available (e.g. number of home runs), but frequently receive bonuses for aggregate performance measures such as Most Valuable Player. The alternative to objective measures is subjective performance evaluation, typically by supervisors. However, there is here a similar effect to “multi-tasking”, as workers shift effort from that subset of tasks which they consider useful and constructive, to that subset which they think gives the greatest appearance of being useful and constructive, and more generally to try to curry personal favour with supervisors. (One can interpret this as a destruction of organizational social capital – workers identifying with, and actively working for the benefit of, the firm – in favour of the creation of personal social capital – the individual-level social relations which enable workers to get ahead (“networking”).)

A linear model

The four principles can be summarised in terms of the simplest (linear) model of incentive compensation:

w = a + b(e + x + g×y)

where w stands for the wage, e for (unobserved) effort, x for unobserved exogenous effects on outcomes, and y for observed exogenous effects; while g and a represent the weight given to y, and the base salary, respectively. The interpretation of b is as the intensity of incentives provided to the employee.


The above discussion on explicit measures assumed that contracts would create the linear incentive structures summarised in the model above. But while the combination of normal errors and the absence of income effects yields linear contracts, many observed contracts are nonlinear. To some extent this is due to income effects as workers rise up a tournament/hierarchy: “Quite simply, it may take more money to induce effort from the rich than from the less well off.” (Prendergast 1999, 50). In addition, the marginal return to effort may increase: it is more important for a CEO to work hard than for a shop floor worker (eg Murphy 1998 highlights the importance of bonuses for executives). Similarly, the threat of being fired creates a nonlinearity in wages earned versus performance. Moreover, many empirical studies illustrate inefficient behaviour arising from nonlinear objective performance measures, or measures over the course of a long period (eg a year), which create nonlinearities in time due to discounting behaviour. This inefficient behaviour arises because incentive structures are varying: for example, when a worker has already exceeded a quota or has no hope of reaching it, versus being close to reaching it – eg Healy (1985), Oyer (1997), Leventis (1997). Leventis shows that New York surgeons, penalised for exceeding a certain mortality rate, take less risky cases as they approach the threshold. Courty and Marshke (1997) provide evidence on incentive contracts offered to agencies, which receive bonuses on reaching a quota of graduated trainees within a year. This causes them to ‘rush-graduate’ trainees in order to make the quota.

Objective and subjective performance evaluation

Objective performance evaluation

The major problem in measuring employee performance in cases where it is difficult to draw a straightforward connection between performance and profitability is the setting of a standard by which to judge the performance. One method of setting an absolute objective performance standard - rarely used because it is costly and only appropriate for simple repetitive tasks - is time-and-motion studies, which study in detail how fast it is possible to do a certain task. These have been used constructively in the past, particularly in manufacturing. More generally, however, even within the field of objective performance evaluation, some form of relative performance evaluation must be used. Typically this takes the form of comparing the performance of a worker to that of his peers in the firm or industry, perhaps taking account of different exogenous circumstances affecting that worker in particular (x in the linear model). Sometimes, the performance in the previous period is used as a standard. The problem with this is that there is potential for a ratchet effect, with standards being raised by employers over time. The threat of this may lead to severe negative consequences as workers respond to this by reducing output levels. A good example is the reluctance of Soviet managers to try to push output levels up, as this frequently resulted in “punishment” for having been lazy or corrupt in earlier periods, by even higher standards (targets) being set for the future (Milgrom and Roberts 1992, 233). It is often suggested that one of the reasons for the general union opposition to piece rates stems from the potential for a racket effect (where piece rates drift down over time as incentives make workers more productive, leaving them little better off). Where piece rates have been successful, for example at the Lincoln Electric Company, it has typically been through the establishment of a credible policy of only changing them due to new equipment or new work methods (Milgrom and Roberts 1992, 236).

Subjective performance evaluation

Subjective performance evaluation allows the use of a subtler, more balanced assessment of employee performance, and is typically used for more complex jobs where comprehensive objective measures are difficult to specify and/or measure. Whilst often the only feasible method, the attendant problems with subjective performance evaluation have resulted in a variety of incentive structures and supervisory schemes.

One problem, for example, is that supervisors may under-report performance in order to save on wages, if they are in some way residual claimants, or perhaps rewarded on the basis of cost savings. This tendency is of course to some extent offset by the danger of retaliation and/or demotivation of the employee, if the supervisor is responsible for that employee’s output. As an example, there have been numerous cases where net profits were apparently underreported on successful Hollywood films, where actors or writers had been promised a percentage of net profits – Cheatham, David, and Cheatham (1996).

Another problem relates to what is known as the “compression of ratings”. Two related influences – centrality bias, and leniency bias - have been documented (Landy and Farr 1980, Murphy and Cleveland 1991). The former results from supervisors being reluctant to distinguish critically between workers (perhaps for fear of destroying team spirit), while the latter derives from supervisors being averse to offering poor ratings to subordinates, especially where these ratings are used to determine pay, not least because bad evaluations may be demotivating rather than motivating. However, these biases introduce noise into the relationship between pay and effort, reducing the incentive effect of performance-related pay. Milkovich and Wigdor (1991) suggest that this is the reason for the common separation of evaluations and pay, with evaluations primarily used to allocate training.

Finally, while the problem of compression of ratings originates on the supervisor-side, related effects occur when workers actively attempt to influence the appraisals supervisors give, either by influencing the performance information going to the supervisor: multitasking (focussing on the more visibly productive activities – Paul 1992), or by working “too hard” to signal worker quality or create a good impression (Holmstrom 1982); or by influencing the evaluation of it, eg by “currying influence” (Milgrom and Roberts 1988) or by outright bribery (Tirole 1992).

Incentive structures: alternatives to direct pay for performance


Much of the discussion here has been in terms of individual pay-for-performance contracts; but many large firms use internal labour markets (Rosen 1982) as a solution to some of the problems outlined. Here, there is “pay-for-performance” in looser sense over a longer time period. There is little variation in pay within grades, and pay increases come with changes in job or job title (Gibbs and Hendricks 1996). The incentive effects of this structure are dealt with in what is known as “tournament theory” (Lazear and Rosen 1981). One of the predictions of this is that agents who fall behind in tournaments, whether in a sports context (Becker and Huselid 1992, in NASCAR racing) or in the broiler chicken industry (Knoeber and Thurman 1994), take riskier actions in order to improve the prospects of winning, which may be inefficient.

A major problem with tournaments is that since evaluation is relative to other individuals, people are less likely to help direct competitors when in need (Lazear 1989, Rob and Zemsky 1997). This is supported empirically by Drago and Garvey (1997). Why then are tournaments so popular? Firstly, because – especially given compression rating problems – it is difficult to determine absolutely differences in worker performance. Tournaments merely require rank order evaluation. Secondly, it reduces the danger of rent-seeking, because bonuses paid to favourite workers are tied to increased responsibilities in new jobs, and supervisors will suffer if they do not promote the most qualified person. Thirdly, where prize structures are (relatively) fixed, it reduces the possibility of the firm reneging on paying wages. As Carmichael (1983) notes, a prize structure represents a degree of commitment, both to absolute and to relative wage levels.

Deferred compensation

Tournaments represent one way of implementing the general principle of “deferred compensation”, which is essentially an agreement between worker and firm to commit to each other. Under schemes of deferred compensation, workers are overpaid when old, at the cost of being underpaid when young. Salop and Salop (1976) argue that this derives from the need to attract workers more likely to stay at the firm for longer periods, since turnover is costly. Alternatively, delays in evaluating the performance of workers may lead to compensation being weighted to later periods, when better and poorer workers have to a greater extent been distinguished. (Workers may even prefer to have wages increasing over time, perhaps as a method of forced saving, or as an indicator of personal development. eg Loewenstein and Sicherman 1991, Frank and Hutchens 1993.) For example Akerlof and Katz 1989: if older workers receive efficiency wages, younger workers may be prepared to work for less in order to receive those later. Overall, the evidence suggests the use of deferred compensation (eg Freeman and Medoff 1984, and Spilerman 1986 – seniority provisions are often included in pay, promotion and retention decisions, irrespective of productivity.)


In conclusion, the reason that employees are often paid according to hours of work rather than by direct measurement of results is that it is often more efficient to use indirect systems of controlling the quantity and quality of effort, due to a variety of informational and other issues (eg turnover costs, which determine the optimal minimum length of relationship between firm and employee). This means that methods such as deferred compensation and structures such as tournaments are often more suitable to create the incentives for employees to contribute what they can to output over longer periods (years rather than hours). These represent “pay-for-performance” systems in a looser, more extended sense, as workers who consistently work harder and better are more likely to be promoted (and usually paid more), compared to the narrow definition of “pay-for-performance”, such as piece rates.

It should also be noted that this discussion has been conducted almost entirely in terms of the analysis of self-interested rational individuals. In practice, however, the incentive mechanisms which successful firms use take account of the socio-cultural context they are embedded in (Fukuyama 1995, Granovetter 1985), in order not to destroy the social capital they might more constructively mobilise towards building an organic, social organization, with the attendant benefits from such things as “worker loyalty and pride (...) [which] can be critical to a firm’s success...” (Sappington 1991,63)


  • Alchian, Armen A., and Demsetz, Harold (1972), "Production, Information Costs, and Economic Organization", American Economic Review 62 (5), p777-795
  • Fukuyama, Francis, Trust: The Social Virtues and the Creation of Prosperity, Hamish Hamilton: London, 1995
  • Granovetter, Mark, 1985, “Economic Action and Social Structure: The Problem of Embeddedness”, American Journal of Sociology 91:3, p481-510
  • Milgrom, Paul, and Roberts, John, Economics, Organization and Management 1992, London: Prentice-Hall
  • Nikkinen, Jussi and Sahlström, Petri (2004): Does agency theory provide a general framework for audit pricing? International Journal of Auditing, 8: November, 253-262.
  • Prendergast, Canice, “The Provision of Incentives in Firms”, Journal of Economic Literature 37 (Mar. 1999), 7-63
  • Sappington, David E.M., “Incentives in Principal-Agent Relationships”, Journal of Economic Perspectives 5:2 (Spring 1991), 45-66
  • Williamson, Oliver E., Markets and Hierarchies: Analysis and Antitrust Implications, NY: The Free Press, 1975

See also



Academic Kids Menu

  • Art and Cultures
    • Art (
    • Architecture (
    • Cultures (
    • Music (
    • Musical Instruments (
  • Biographies (
  • Clipart (
  • Geography (
    • Countries of the World (
    • Maps (
    • Flags (
    • Continents (
  • History (
    • Ancient Civilizations (
    • Industrial Revolution (
    • Middle Ages (
    • Prehistory (
    • Renaissance (
    • Timelines (
    • United States (
    • Wars (
    • World History (
  • Human Body (
  • Mathematics (
  • Reference (
  • Science (
    • Animals (
    • Aviation (
    • Dinosaurs (
    • Earth (
    • Inventions (
    • Physical Science (
    • Plants (
    • Scientists (
  • Social Studies (
    • Anthropology (
    • Economics (
    • Government (
    • Religion (
    • Holidays (
  • Space and Astronomy
    • Solar System (
    • Planets (
  • Sports (
  • Timelines (
  • Weather (
  • US States (


  • Home Page (
  • Contact Us (

  • Clip Art (
Personal tools