The principles which have been outlined in the first five sections of this chapter have applications which are more wide ranging than the sharecropping case which we have thus far been using for illustrative purposes. Harris and Raviv (1978) provide a number of interesting examples.
1. Health and Motor Insurance
Moral hazard in insurance markets is a classic problem, as was seen in Chapter 2. In terms of the content of Chapter 5, however, what kind of insurance contracts are likely to be observed and in what situations? Let θ, the state of the world, stand for ‘degree of illness’ instead of weather conditions. Let e stand for ‘health effort’ (for example, not smoking, taking safety precautions in dangerous tasks and so forth). Finally, let ^ represent ‘amount of healthcare used’; that is, the outcome.
A contract which depends on the outcome alone (that is, simply on the amount of healthcare someone consumes) will, according to the earlier discussion, be inefficient relative to a contract which depends on both the outcome and the state if the latter is observable. Thus, if ‘degree of illness’ can be observed, we would expect contracts to specify payments conditional upon the degree of illness, just as payments to the labourer in the earlier example depended if possible on the weather and not merely the harvest. Assuming ‘degree of illness’ is unobservable, information about ‘health effort’ should be valuable. If health effort is observable, again we would expect this to be reflected in contracts; for example, lower premiums for non-smokers, higher premiums for those who refuse to wear seat belts in cars, and so on.
In the case of motor insurance, θ could represent ‘difficulty of driving conditions’, e ‘safety effort’, and π ‘damage to the insured and to third parties’. Clearly, a contract which depends only on damage will be inferior to one based on both π and θ if the latter is observable. Thus, we might expect insurance contracts to vary by geographical area if ‘driving conditions’ differ. If θ is not observable, information on safety effort will be valuable. This too may be very difficult to observe, but if driving care is perfectly correlated with age, for example, we would expect to see insurance contracts varying with the age of the insured.
2. Law Enforcement
Here the principal is seen as the political representative of the general public, while the agent is the policeman whose task it is to detect and punish crime. Suppose the outcome ^ in this case is revenue generated by fines. Effort e will be ‘policing effort’ such as patrols, observations and whatever other activities compatible with the law raise the total of fines collected. 0 represents the ‘state of crime’. It reflects the type and seriousness of crimes committed and, to fit into the principal-agent framework of this chapter, it is assumed that this ‘state of crime’ is independent of policing effort e.
Applying the sharecropping results to this new situation, an efficient contract would be expected to involve both the outcome, ‘fines generated’, and the ‘state of crime’. The government would receive a payment from the police which depended only on the state of crime and the police would keep the residual fines. Risk could then be distributed in accordance with the principles of section 3. If the government were risk neutral and the police were risk averse, the residual fines in each ‘state of crime’, given that the standard policing effort had been applied, would ideally be constant. On the other hand, if the police were risk neutral, no information about effort or state would be necessary for the efficient contract. The police would pay a flat fee to the government independent of the state of crime and would keep whatever revenue in excess of this fee they succeeded in generating. Effectively, the police would be paying a certain sum for the ‘policing franchise’ of an area and would be rewarded by their success in levying fines. Unsettling though such arrangements undoubtedly appear, there seems little doubt that they are well designed to produce a dedicated police force (dedicated, that is, to collecting fines). A more detailed discussion of ‘franchising’ as an incentive device and form of business organisation occurs in Chapter 7, while Chapter 15 contains a discussion of the role of franchising in government regulation.
Assuming that the state of crime is unobservable, information on effort is valuable if the police are risk averse. Even if the information is ‘ noisy’, it may be used to improve incentives, as we have seen in section 5. A policeman’s contract might, for example, be designed so that effectively he ‘posted a bond’ which would be returned to him providing that malfeasance remained undetected over a specified period. Corruption or negligence would result, if observed, in the loss of the value of the bond. In other words, loss of the bond would be equivalent to the term —8θ which appears in the monitoring gambles of section 5.
3. Employment Contracts
Consider first a case in which the employee’s output is observable, whilst effort and state are unobservable. This case is identical with the sharecropping example discussed in section 4. The employee must take some risk if effort is to be induced, and this will involve inefficient risk sharing unless the employee is risk neutral. In the latter case, we once more observe the ‘franchising’ solution, with the ‘employee’ paying a fee to the ‘employer’ for permission to use specified resources for a given period of time.
Where employees are risk averse, a contract involving both outcome and state will be preferred to a contract involving just the outcome. The state variable 0 might represent ‘market conditions’, in which case, if these were ‘observable’, contracts might be expected to link remuneration to some economic indicators of these conditions. Employees could be envisaged to pay a fee to the employer conditional only upon the state. This fee would presumably be lower in times of depression and higher in times of prosperity if workers were risk averse. The employer effectively offers insurance to the workers and takes the brunt of economic fluctuations. In the extreme case of a perfectly observable state variable θ, and a known function π = π (θ, e), the worker’s remuneration would be constant in both prosperity and depression. Although practical examples of contracts explicitly written in this way are not easy to find, it has been argued that many contracts of employment are implicit.9 The employer accepts an implicit unwritten obligation to insure the workforce against fluctuations in the way described. It is suggested that this implicit obligation helps to explain the ‘stickiness’ or ‘inflexibility’ of wage rates over the business cycle. Along with supplementary assumptions about the type of social security system operating, or the nature of information about θ (it may be asymmetrically distributed and available to employers but not employees), the implicit contracts literature has attempted to explain the existence of levels of employment greater than or less than would be observed under symmetric information. We shall discuss some of the ideas of the implicit contracts literature in a little more detail in Chapter 6.
Another application of the principal-agent results to employment contracts concerns the use of educational qualifications. Suppose that output depended not only on effort e but also on ‘native ability’ (Harris and Raviv, 1978). The state variable 0 would now refer to the ‘native ability’ of the worker. Clearly this is not easy to observe, but if employers believed that ‘native ability’ was reflected in educational attainment we would expect remuneration to depend on qualifications and not just on the outcome. According to this view of things, education does not necessarily equip people with specific skills, but provides them through a series of tests, no matter how pointless in other respects, with evidence about their ‘native ability’. Education is a ‘screening procedure’10 which tests for a particular type of attribute. This evidence can then be used by employers when contracting with employees.
Where the state variable is unobservable and workers are risk averse, information about effort is valuable. Perfect observability of the employees’ effort by the employer would simply result, of course, in a contract dependent on effort alone. The risk-averse employee would again be assured of the outcome, and the risk-neutral employer would take the risk. This, it will be recalled, is the principal and agent result considered in Chapter 4, where the ‘single proprietorship’ was discussed. There, however, it was assumed that monitoring was costly and there was no presumption that effort was perfectly observable. It is worthwhile remembering, therefore, why it was that, even with costly monitoring, the contract did not involve the outcome and depended on effort alone. Conditions of team production implied that output could not be ascribed to particular members of the team, and the incentive effects of linking individual rewards to the collective outcome are minimal for large teams. In terms of our discussion of this chapter, therefore, contracts dependent upon the outcome will not induce effort under conditions of team production, because the probability of a preferred outcome is perceived to be only very weakly related to individual effort. A large individual effort (distance 0a in Figure 5.4) will not greatly affect the slope of the indifference curves so that the set of contracts xyw in Figure 5.5 disappears and there are no Pareto improvements on θ. From an individual point of view, the ‘efficiency of effort’ is small, even though jointly it may be very high.
Source: Ricketts Martin (2002), The Economics of Business Enterprise: An Introduction to Economic Organisation and the Theory of the Firm, Edward Elgar Pub; 3rd edition.