Motivation: Performance measurement
Performance evaluation is a tricky business...
Joe Smith, my assistant programmer, can always be found hard at work in his cubicle. Joe works independently, without wasting company time talking to colleagues. Joe never thinks twice about assisting fellow employees, and he always finishes given assignments on time. Often Joe takes extended measures to complete his work, sometimes skipping coffee breaks. Joe is an individual who has absolutely no vanity in spite of his high accomplishments and profound knowledge in his field. I firmly believe that Joe can be classed as a high-caliber employee, the type which cannot be dispensed with. (… for my true assessment read only the marked lines (Project Leader)) Source: Milkovich, Newman and Gerhart (2014) Compensation, McGraw-Hill
Why evaluate performance?
To compute performance in performance pay schemes.
To recognize and promote talent in organizations.
To identify training needs and other personal improvements.
To signal bad job matches and weed them out.
To provide personalized feedback on (relative) performance. Some people value this feedback on its own right as a confirmation of status or a signal of whether they conform to group norms or not.
To communicate firm values in a salient way.
General rule
The general point here is that compensation can vary systematically only with things that the employer can observe. The employer cannot pay more to sales representatives who are particularly effective ... if it is impossible to tell who they are. In addition, even some observable indicators may not be suitable bases for compensation... To base a compensation on something that is not objectively measurable is to invite disputes and unhappiness among employees.
Source: Milgrom and Roberts, Economics, Organization, and Management, p. 215
Der allgemeine Punkt ist, dass die Vergütung nur bei Dingen, die der Arbeitgeber beobachten kann, systematisch variieren kann. Der Arbeitgeber kann Vertriebsmitarbeitern, die besonders effektiv sind, nicht mehr zahlen, ... wenn es unmöglich ist, zu erkennen, wer sie sind. Darüber hinaus sind selbst einige beobachtbare Indikatoren möglicherweise nicht als Grundlage für die Vergütung geeignet... Eine Vergütung auf etwas zu gründen, das nicht objektiv messbar ist, lädt zu Streitigkeiten und Unzufriedenheit unter den Arbeitnehmern ein.
Übersetzt mit www.DeepL.com/Translator (kostenlose Version)
Four principles of a good performance measure
1.
Monitoring intensity principle:
Be as precise as economically reasonable
Recall the formula for the optimal strength of incentives in the linear contract
It contains the term 𝜎 2 , the variance of the random noise component in the performance measure 𝑧 = 𝑒 + 𝜂. The more noisily performance is measured, the weaker the optimal incentives, the lower the effort, and the less effective incentives are.
—>It makes sense to use the best (i. e. most precise) measure of performance available on which to base incentives.
Measuring performance is costly: the more precisely you want to measure performance (i. e. lower 𝜎 2 ), the higher the measurement costs, 𝑀(𝜎 2 ), you must pay. 𝑀´(𝜎 2 ) < 0, i. e. noisier performance is cheaper 𝑀´´(𝜎 2 ) < 0, i. e. it is increasingly expensive to measure performance more precisely. —>At some point, the costs of producing a more accurate performance measure will exceed the output gains from sharper incentives this measure allows.
—>Thus there is a tradeoff between the strength of incentives and precision of performance measurement.
Monitoring costs: So far, we have ignored the monitoring costs of performance measurement. Yet, they do enter the profit equation.
2.
Informativeness principle:
Use all the available information about effort
As an alternative to reducing the noisiness 𝜎 2 of the performance measure by paying more to measure it better, P can use other indicators that are informative of A's effort.
The informativeness principle states that P should use every additional indicator about A's effort that is freely available, as long as this indicator is informative.
—>What is “informative"?
Issues with relative performance incentives (RPI)
RPI may be designed for heterogeneous agents (i.e., different costs of effort and risk preferences). In such case, optimality will require different 𝛼* , 𝛽*, 𝛽̃* for different agents. This unequal treatment may be hard to defend.
Additional performance data required to administer RPI may be expensive to obtain.
RPI may encourage collusion between different agents and sabotage.
RPI are also less desirable when externalities between outputs of different agents exist, e.g., in teams.
RPI are used for instance in executive compensation schemes.
3.
Equal compensation principle:
Reflect all the activities that are important for the P and reward them in proportion to their contributions
Suppose there is more than one dimension of effort that benefit P. Then, all effort dimensions should receive equal reward at the margin, i.e. marginal product of effort along all dimensions should be the same.
Otherwise the agent will focus on the most highly rewarded effort dimensions and ignore the rest.
The problem is, not all dimensions of effort have good performance measures. The badly measured ones will pull the optimal strength of incentives down, thus reducing effort and output on all dimensions.
—>Example: quantity (relatively easy to observe) and quality (difficult to observe)
4.
Alignment principle:
Be influenced by A's effort in the same way as the outcome desired by P
Suppose, as above, effort is multidimensional. For example: a desired outcome for P is long-term value of the firm, which depends on efforts spent stimulating sales, cultivating good customer relationships, maintaining good public image, quality standards, etc...
But the measure of performance P uses to evaluate and reward A is just, e.g., short-term accounting profits.
Even though the desired outcome (the long-term firm value) and observed measure of performance (the accounting profit) are both influenced by the same set of actions (e.g., time spent on sales vs. customer service), they may be influenced differently by the same actions.
—>The misalignment between the desired outcome and the available performance measure, which are both functions of the same actions, reduces efficiency and must be corrected.
Implications of the principle
A good performance measure p must be aligned with desired outcome 𝑦: 𝛿1 / 𝛾1 , must be close to 𝑦: 𝛿2 / 𝛾2 .
Simply a high correlation between p and y is not enough. To see this, suppose 𝑦 = 𝑒1 + 𝜂 and 𝑝 = 𝑒2 + 𝜂, so p and y are correlated via the shared noise term 𝜂, while p is clearly a useless measure of the effort important for y.
For example, accounting earnings (typical performance measure p) and the share price (a function of y) may be correlated not only because the same actions positively affect both, but also because of the random events such as business cycle.
—>Benchmarking helps to control for the effects of common shocks on p and y (recall the informativeness principle).
Last changed2 years ago
