“Pay-for-performance” or “incentive pay” has been a top-of-mind topic for me in recent months. It’s a pretty pervasive industry (best?) practice, especially for executives and sales people, and many companies use it quite extensively beyond the bounds of those two functions. To develop a more first-principled point of view on this topic, I did some research aiming to understand the origins of the concept and the boundaries/contexts in which evidence suggest it may not be effective. I found several good resources, the most rigorous one was a paper by Ian Larkin, Lamar Pierce and Francesca Gino titled:
The Psychological Costs of Pay-for-Performance: Implications for the Strategic Compensation of Employees
But I’ve also used a slew of less rigorous resources, including:
- CultureAmp’s CEO explains why they’re not paying sales commissions (5 min talk) and his recent post on Remuneration and Culture
- Psychology Today: Why financial incentives don’t improve performance
- Inc: Why Sales commissions don’t work (in the long run)
- The Telegraph: Forget carrots and sticks, they don’t always work
- NYT: For some, paying sales commissions no longer makes sense
- Quartz: The dangers of sales commissions that work too well
- Think Growth: The best way to use money as a motivator is to take the issue of money off the table
Pay-for-performance or incentive pay is the practice of tying additional compensation to the achievement of a well-defined, measurable outcome. As opposed to a more permanent, long-lasting compensation change like a promotion.
- Closing a sales deal
- Completing a project on time
- Hitting a certain target for a metric
Agency Theory: the origins of incentive pay
Incentive pay became popular with the rise of Agency Theory in the late 1970s.
Agency Theory is based on a few core assumptions about companies and employees. Specifically that companies seek to maximize profits by motivating employee effort and attracting more highly skilled employees, while minimizing salary costs. And that employees seek to maximize utility by increasing income while minimizing efforts.
Agency Theory also takes into account some information asymmetries in the dynamic between companies and employees that give employees an advantage over companies. Specifically, that employees know their own effort exertion (while companies have imperfect information) and that employees know their skill level (while companies have imperfect information).
Taking those assumptions and information asymmetries into account, Agency Theory suggests that companies overcome these asymmetries by providing incentives for employees to exert effort and self-select by skill level. For example, by offering a low guaranteed salary with a large performance element, a company can incentivize higher effort from all employees, but it can also attract and retain employees with high skills, while ‘sorting away’ those with low skills.
Insights from Agency Theory:
- Employees work harder when their pay is based on performance.
- Companies are more likely to use performance-based pay when they have less information about actual employee effort.
- Companies are more likely to use performance-based pay when they have less information about employee skill level, and/or as employee skill level is more heterogeneous.
- Companies are more likely to use team-based performance pay vs. individual-based pay when coordination across employees is important, when free riding is less likely, or when monitoring costs are low.
Research in Psychology and Decision Research
However, since the 1970s research in psychology and decision research have painted a more nuanced picture of the dynamic between companies and employees. Two elements in particular, social comparison and overconfidence play a pivotal role in that dynamic.
Social comparison theory introduces considerable costs associated with individual pay-for-performance systems, because it argues that people evaluate their own abilities and opinions in comparison to referent others. Generally, people seek and are affected by social comparisons with people who are similar to them gaining information about their own performance.
People also tend to be overconfident about their own abilities and too optimistic about their futures. Overconfidence is thought to take at least three forms:
- People consistently express unwarranted subjective certainty in their personal and social predictions.
- They commonly overestimate their own ability.
- They tend to overestimate their ability relative to others.
People tend to be overconfident about their ability on tasks they perform very frequently, find easy, or are familiar with. Conversely, people tend to be underconfident on difficult tasks or those they seldom carry out. This tendency has strong implications for overconfidence in work settings, since work inherently involves tasks in which employees have strong domain expertise in.
The above refines the assumptions about companies and employees. Specifically, that Maximize profits for companies also requires minimizing non-wage costs (counter-productive work behaviors), and that maximizing utility for employees also requires minimizing perceived inequality.
The information asymmetries should also be refined to take into account the fact that employees perception of their effort and skill level are biased (while companies have imperfect information).
Insights from Psychology and Decision Research:
- Perceived inequity through wage comparison, compounded by overconfidence bias, reduces the effort benefits of individual pay-for-performance compensation systems.
- Perceived inequity through wage comparison, compounded by overconfidence bias, introduces additional costs from sabotage and attrition in individual pay-for performance compensation systems.
- Perceived inequity arising through random shocks in pay (economic downturn, weather, client going bankrupt) introduces additional costs from effort, sabotage, and attrition in individual pay-for-performance compensation systems
- Overconfidence bias reduces the sorting benefits of individual pay-for-performance compensation systems (low skill employees, will still self-select into a pay-for-performance scheme)
Alternatives to individual pay-for-performance
After incorporating insights from psychology and decision research, individual pay-for-performance seems less like the holy grail that Agency Theory made it to be. Are there better alternatives? Larkin, Pierce & Gino looked at a couple:
- Team-based compensation: additional compensation is tied to the achievement of team goals/objectives and is shared among the team members.
- Scaled wages: employees are compensated in relatively tight ‘bands’ based largely on seniority.
And concluded that:
- Team-based compensation reduces costs of social comparison when individual contribution is not highly heterogeneous within the team.
- Team-based compensation only resolves problems of overconfidence in individual pay-for-performance systems if the actual contribution of teammates is observable.
- Scaled wages have lower social comparison costs than team-based and individual-based compensation systems.
- Scale wages reduce costs of overconfidence in individual- and team-based pay-for-performance.
Just like many other issues pertaining to the complex problem of human collaboration the answer to whether pay-for-performance is effective, is not a definitive “yes” or “no”, but a more nuanced one, depending on the specific context in which pay-for-performance is used.
Pay-for-performance will be more effective if:
- Work requires low cognitive load
- Outcomes are very controllable (effort and outcome are highly correlated)
- Outcomes are easily attributable – it is easy to separate out individual contributions which led to a certain outcome
- Overconfidence bias is minimal or non-existent
- Social comparison is limited or non-existent
- Global optimum can easily be decomposed to pre-set individual outcomes
Pay-for-performance will be less effective if:
- Work requires high cognitive load
- Outcomes are not very controllable (effort and outcome are loosely correlated)
- Outcomes are difficult to attribute
- Overconfidence bias is meaningful
- Social comparison exists
- Global optimum cannot be easily decomposed to pre-set individual outcomes