Bengt Holmström Prize Lecture: Pay for Performance and Beyond

434 

The Nobel Prizes

is partly incentivized for the later periods because of the delayed information, 

and from this point of view it would be efficient to let him continue.

C. Career Incentives

People want to be appreciated for their work. Career concern models (Holm-

ström 1999a; Dewatripont, Jewitt, and Tirole 1999) can capture the desire for 

appreciation through a dynamic signaling mechanism. Employers are learn-

ing about the value of employees from past performance. The value is initially 

unknown (also for the employee) but will be revealed at least partly over time. 

Employees try to make a favorable impression because better performance can 

lead to promotions, higher pay, more status, and other rewards in the future.

Career concerns can be expected to be more powerful in firms than in mar-

kets, because in firms the employees typically know whom they should impress. 

Firms can exploit this motivation by indicating what is desired from an employee. 

In employee surveys, one of the most common complaints is that the employees 

would like a clearer understanding of what is expected of them. If they only knew 

what the boss wants, they could work on making a good impression in those 

dimensions. The craving for appreciation and the desire to impress superiors 

explains why a mere change in the accounting system can have a big impact on 

the behavior of employees. Key performance indicators are important because 

they signal what management wants. Alone or combined with relatively modest 

bonuses, they can have a surprisingly strong effect. Putting money behind a mea-

sure conveys a stronger message of what is expected. Employees also respond to 

mission statements as in Dewatripont, Jewitt and Tirole (1999).

Career incentives can be so strong as to obviate any need for financial incen-

tives. In fact, they can be too strong and lead to wasteful pandering and influ-

ence activities (Milgrom and Roberts 1988; Holmström and Ricart i Costa 1986; 

Prendergast 1993). Just like financial incentives, career incentives can be imbal-

anced, because some indicators of performance are more visible and therefore 

more salient than others. The firm can regulate visibility through the access the 

employee has to superiors and they can institute pay and promotion practices 

that do not respond to performance as strongly. Basing salary increases and pro-

motions on seniority mutes career incentives that are perceived to be too strong 

or lead to short-sighted behavior. In general, the firm’s overall hierarchy and its 

promotion policies are powerful incentive drivers without direct counterparts 

in the market.

21

The problems of multitasking tend to manifest themselves in much the 

same way when career concerns are the driver of employee behavior in place 

of explicit bonuses. This is why the earlier discussion of multitasking assuming 

---

Pay For Performance and Beyond 

435

commissions as incentive instruments provides useful guidance even though 

commissions are relatively less used by firms.

VI. TWO INCENTIVE SYSTEMS: EMPLOYMENT VERSUS CONTRACTING

The preceding discussion makes clear the point that firms have access to a variety 

of incentive instruments that can compensate for low-powered pay-for-perfor-

mance incentives. In this section I want to highlight two implications for the 

design and use of incentive systems.

The first is that structuring efficient incentives within the firm requires a 

concerted use of all available incentive instruments. This parallels the earlier 

lesson from multitasking that commissions for all tasks need to be considered 

in concert. The second point is that incentives in firms and in markets form two 

logically coherent systems, each with its distinct comparative advantage.

In Holmström and Milgrom (1994) we illustrate both points by applying our 

multitask model to the choice between employment and contracting. We want to 

explain the findings of Anderson and Schmittlein (1984) and Anderson (1985), 

who studied how industrial selling is organized in the electronics industry. Some 

sales agents work as independent representatives, others as employees. A firm often 

uses both forms of sales organization. What determines the choice between the 

two forms of sales and how are the incentives structured within each alternative?

Anderson and Schmittlein find that the most important measures determin-

ing the choice between employment and independent contracting were “the dif-

ficulty of evaluating performance” and “the importance of non-selling activities.” 

Performance evaluation is difficult for complex sales, while non-selling activities 

are important when cooperation with other sales agents is desired. Increases in 

both of these measures made employment more likely. Moreover, independent 

representatives were compensated entirely by commission and they were allowed 

to represent other manufacturers, while sales employees were on fixed wages and 

could not sell the products of other manufacturers.

These findings can be explained in our standard multi-tasking setting where 

tasks compete for the sales agent’s attention. The fact that independent agents are 

free to sell the products of other manufacturers and have strong incentives par-

allels the earlier argument about giving an agent the freedom to pursue private 

tasks if the commission rate is high, but not if the commission rate is low. The 

difficulty of evaluating performance drives the commission rate down, as does 

the desire to have the agent allocate time to non-selling (cooperative) activities.

The model includes two types of selling activities: direct selling, which affects 

short-term returns and is influenced by commissions, and indirect selling which 

---