The University of Chicago

THE UNIVERSITY OF CHICAGO

Booth School of Business

Business 33032 Canice Prendergast

Spring 2009

Topic 7

Subjective Performance Evaluation and Career Concerns

In this section of the course, we consider both subjective performance evaluation and the provision of incentives over a worker’s career in more detail. We begin by considering subjective performance evaluation, which often takes the form of a discretionary bonus, but also pervades promotion, as there are rarely completely objective criteria for determining the performance of an employee.

I. Subjective Performance Evaluation

We begin to study subjective performance evaluation by the Merck case. This is an example of a performance evaluation scheme that actually seems to work. Begin by considering Exhibit A1 on page 5 of the case. This measures the return on assets for Merck and on its largest competitors. What we see is a steady decline in the advantage of Merck relative to its competitors over a long period of time before 1985. In 1985, Merck CEO Roy Vagelos formed an Employee Relations Review Committee charged with redesigning personnel policies.

One of the characteristics of Merck that makes it worth looking at from our perspective is that a large proportion of the company consists of exempt salaried employees, covered by the firm’s Performance Evaluation and Salary Administration programs. Hence the company has significant discretion over its compensation schemes. Merck has typically paid well compared to the industry average. What advantages might this have?

Is paying wages above industry averages likely to be sufficient? What might it need to be supplemented with?

Performance Evaluation under the Old Plan

This is described on page 2. Workers were rated on an absolute scale of 1 to 5 (with pluses and minuses), with 5 designating exceptional performance and 1 indicating unacceptable performance. The rating distribution is given in Exhibit A2, as are average pay differences for various ratios and ratings.

Can you see any problems with this compensation scheme?

To see some potential problems, take a look at the comments of individuals on page 4. Note both incentive and adverse selections problems.

Question:

If there are easy and hard graders (as most organizations have), do these figures overstate the variation in ratings or understate the variation?

The New Scheme

The scale has been replaced by a five-point scale, with a targeted distribution. Note that the targeted distribution is not symmetric. Why?

Note the GM forced rankings scheme, and particularly its failure. Note that the difference here is that 70% of individuals are bunched in the HS category. What are the advantages and disadvantages of this?

Note the distribution of ratings in Exhibit B3. Here you see that there was some “fudging” on the target distribution, as even fewer were given bad ratings compared to the desired distribution. Also note that the wage spread is much greater than with the old scheme. Hence it is not only important to spot the good and bad performers but also it is necessary to reward them.

Problems:

1. These schemes should probably not be used with small groups. Merck decided to use it only when the group exceeded 100. What problems arise with small work groups?

Note the problems when teamwork is involved.

2. The “it’s her turn” phenomenon.

3. The monetary cost.

II. Other Problems with Subjectivity

1. Compression of Rankings

There is considerable evidence that managers like to avoid differentiating high from low performers, even when inflation is not possible.

An example: lots of B grades?

2. Reneging on Payments

When firms privately know output, they may lie about it to save on wage costs.

3. Rent-Seeking or Influence Activities.

The Basic Problem with Rent-Seeking

So far, we have assumed that the mistakes that manager’s make are random: just measurement error caused by things beyond their control. For example, in Topic 5 we described observed output as

q = e + noise

But there is considerable evidence that “noise” is not exogenous, but depends on the contracts that you write with your employees. Specifically, supervisors distort evaluations more when there is “more money on the line.”

· In other words, offering more pay for performance implies that subjective performance evaluations become less reliable.

An example: Juries with minimum sentencing requirements.

Costs

1. Wasted resources on Rent-Seeking.

2. Incorrect Allocations of Workers to Tasks.

Implications

1. Low-powered Incentives.

While it causes workers to exert less effort, it at least reduces rent-seeking and causes supervisors to reveal information more truthfully.

2. Bureaucracy

This refers to the use of rules which are ex post inefficient.

Examples

· Seniority

· Pay Scales

· Outside Reviews

3. Separating Pay from Evaluations.

· Important when training and feedback matter.

4. Getting multiple evaluations to “evaluate the evaluator”.

But there are problems: Figure Skating Judges.

III. Career Concerns

So far we have looked at cases where a worker’s effort affects his current output but not his career. Here we consider career concerns. Career concerns occur when a worker carries out some activity that affects how he is perceived, which has benefits over the worker’s career. This occurs through the worker’s reputation.

The uncertainty we are concerned with is over the worker’s ability. We can imagine two types of uncertainty: asymmetric information, where the worker has more information than the firm, and symmetric information, where nobody has information about the worker’s ability. We are concerned with the case of symmetric information and incentives that arise.

Career concerns will be considered along three dimensions: (i) effort, (ii) project selection, and (iii) herd behavior.

Effort

Can reputation act as a sufficient incentive to exert effort? Consider a worker who is believed to have ability a0, who a firm would like to induce to exert effort, et over his career, which lasts from period 0 to period T. In the absence of explicit incentives, does a worker have incentives to exert effort?

If the worker has no reputation, then the answer is no. If other potential employers can observe his performance and can infer information about his ability from that, then he may, because high output signals that he is talented and is valuable to other firms. This is explained in the following way.

The worker’s output is assumed to be given by

yt = a + f(et) + lt,

where a is the worker’s true ability, et is the effort exerted in period t, and lt is the worker’s luck in period t, which has mean 0. If employers can observe output, they will “update” their impression of the manager on the basis of output in the past.

How do you update? Consider the period of the worker’s career, period 0. The firm places some weight on its original impression, a0. It also places some weight on current performance minus the productivity due to effort you think he exerted (since this measures his likely ability). In particular, if a1 is the expected ability of the worker at the beginning of period 1, it is derived by

a1 = w0a0 + [1 - w0](y0 - Ef(e0)),

where w0 is the weight placed on the initial observation in period 0 and Ef(e0) is the increase in output caused by the worker exerting effort in period 1. We can work out the expected ability of the worker in any period from a similar formula. For example, the expected ability of the worker in the 20th year of his career is

a20 = w19a19 + [1 - w19](y19 - Ef(e19)).

Thus everyone updates using the new observation on output and the opinion held before output is observed.

Question: Will w19 be more or less than w1?

In the absence of wage contracts, the firm will pay the worker his expected marginal product in each period. By exerting effort in period 1, the worker improves not only a1, but also all as in the future, because a2 depends on a1, and so on. Hence effort exerted early in a worker’s career affects his wage stream until he retires. What this means is that effort is high early on in a worker’s career but falls as he ages because (i) he is closer to retirement and has fewer years to amortize his investment, and (ii) the weight placed on new observations falls. Another relevant factor is the discount rate.

Efficiency requires that f¢(et) = C¢(et), independent of time. Hence the choice of effort by the worker bears little resemblance to the optimal effort level.

This model can explain why young workers initially work very hard but eventually slack off. It is important to realize that the employers can work out what the worker is up to and will discount the extra effort in equilibrium. This is why this phenomenon of working too hard early on is known as the rat race.

Evidence:

Suppose we considered the fraction of the compensation of an executive that is in the form of pay for performance (stocks and options). Would you imagine this fraction increases or decreases the longer that the executive is in the job?

Project Selection

Suppose a manager is employed for two periods and is good with probability 1/2 and bad with probability 1/2. Nobody knows the manager’s ability. The manager can carry out a project worth $1 million if the he is good and $0 if he is bad. In other words, the operation of the project reveals information about the manager. Should the project be undertaken? Clearly, yes, as its expected value is half a million dollars.

Now assume that only the manager can observe whether there is an investment opportunity, so he decides whether to undertake the project. Assume that the market pays a good manager $60,000 and pays a bad manager $20,000. Since nobody knows the ability of the manager, he can earn $40,000 in period 1 (if there are no contracts).

If the manager does not take the project in period 1 (which could be because there were no good opportunities), he gets $40,000 in period 2. If he takes the project, then with probability 1/2 he is good and gets $60,000, and with probability 1/2 he is bad and gets $20,000. Therefore if he is risk averse, he will not take the project as he prefers the $40,000 with certainty. Hence, career concerns can harm investment opportunities.

Solutions: (i) pay a bonus for investment projects undertaken; (ii) don’t penalize workers if projects go bad. Note that this is the Goldman Sachs solution, however.

Herd Behavior

Once again consider an investment decision being undertaken by lots of managers. (For example, they could be deciding what stocks to buy.) Assume that there are two kinds of managers: good and bad. Good managers get signals that reflect which stocks are the most profitable. Bad managers get lousy signals that reveal little information on profitability.

Suppose there are two stocks, A and B. The talented managers get a signal on which is the good stock so they tend to go to that stock. Suppose the first manager buys stock A. The second manager will look at the first manager and reckon that he may be talented, so that stock A could be the profitable one. As a result, he may override his own opinion (as he could be low ability) and also buy stock A, although he believes that stock B is the better one.

The reason for this is that by choosing a different stock than everyone else, he looks like a low-ability manager (as all the high-ability managers will tend toward the same stocks). Hence career concerns can imply that managers herd on the same stocks.