6.1 (How) incentives work: An overview of empirical studies

• This lecture offers a review of important empirical research that tests theories of incentives (previous lectures) and give lessons for better compensation and benefits practices within firms.

• Types of incentives we study today:

  1. Individual incentives

  2. Incentives for managers (not for workers!)

  3. Bonus vs. malus

Individual incentives: Study by Lazear (2000)

• Reference: Lazear (2000) Performance pay and productivity, American Economic Review, 90(5): 1346-1361.

• Safelite Glass Corporation implemented a piece rate compensation scheme in January 1994, affecting 2,755 workers.

• The scheme was phased in over 19 months, starting from the headquarter town. Thus, most workers worked under both hourly and piece rate during the study period.

—>Piece rate: 20 USD per unit, guaranteed hourly rate: 11 USD per hour

How incentives work: incentive and sorting effects

• We know from earlier lectures that incentives stimulate effort by affecting its marginal product (see the “incentive effect")

• We also know that a given incentive scheme brings higher utility to workers with lower costs of effort and lower risk aversion (see the “sorting effect")

—>How to estimate these two effects?

Estimating incentive and sorting effects

• Statistically, the total effect of incentives on performance is estimated from ln 𝑦𝑖𝑡 = 𝛿𝑙𝑖𝑡 + 𝑐𝑜𝑛𝑡𝑟𝑜𝑙𝑠𝑖𝑡 + 𝜀𝑖𝑡 where 𝑦𝑖𝑡 is a measure of performance of agent 𝑖 in period 𝑡, 𝑙𝑖𝑡 is 1 when 𝑖 works under an incentive scheme in 𝑡, and 0 otherwise, controls are variable controlling for other factors that affect performance, e.g., weather, and 𝜀 is the error term reflecting the noise to performance measure that is independent of i 's effort.

• Coefficient 𝜹 measures the effect of incentives on log performance. That is, performance increases by a factor of exp(𝛿) − 1 as a result of incentives.

  —>Notice: 𝜹 is the total effect: the combination of the incentive and sorting effects.

• To estimate the pure incentive effect, we must i) focus on the agents who worked before and after incentives were introduced, and ii) control for their characteristics relevant for the sorting effect, such as ability, costs of effort and risk aversion.

• The easiest way to do so is by using individual fixed effects 𝑢𝑖 , i.e. person-specific dummy variables in the regression equation. They absorb all individual-specific and time-invariant factors affecting productivity. So, assuming ability, costs of effort and risk aversion are constant in time, the pure incentive effect is the coefficient 𝜸 in ln 𝑦𝑖𝑡 = 𝛾𝑙𝑖𝑡 + 𝑐𝑜𝑛𝑡𝑟𝑜𝑙𝑠𝑖𝑡 + 𝜀𝑖𝑡

—>Assuming that the incentive and sorting effects are additive to each other, the sorting effect is estimated as 𝜹 − 𝜸

Incentives for managers

• Bandiera, Barankay and Rasul (2007). Incentives for managers and inequality among workers, Quarterly Journal of Economics, 122(2): 729-773.

—>Does incentive pay for managers affect the performance of their subordinates?

—>Different from Lazear (2000), since managers themselves do not produce output. On the other hand, rewarding managers, not workers, may be an effcient way of creating incentives via targeting and selection effects.

Targeting and selection effects

Managers can affect the productivity of their subordinates in two ways:

1. Take actions affecting the productivity of existing workers (targeting effect). Recall the incentive effect of piece rate. Given the limited span of attention, abler workers are targeted more.

1. Choose the most able workers into employment (selection effect). Think of the sorting effect of piece rate as an analogy.

—>Incentivising managers should lead to an increase in the average performance of their subordinates, but also in an increase in its variation.

Implications: Incentives for managers

• The targeting and selection effects reinforce each other, rather than just add up, since the workers with larger productivity gains after the bonus are also more likely to be working.

• Giving bonuses to managers, rather than to workers, may be more efficient way of incentivising the workers when the manager is able to observe worker effort better than the principal (the firm owner).

• This is even more so, when the worker pay is constrained by pay scale, collective labor agreement, etc.

• Disadvantage: Bonuses to management increase earnings inequality among the workers. This may backfire in the long run by undermining team spirit.

Reward vs. punishment: Incentive pay with a twist

Two questions:

1. How much would you give for a chance of winning 100 EUR?

2. How much would you pay to avoid the same chance of losing 100 EUR?

—>Loss hurts more than gain gratifies. Most people are loss-avers

—>Implication: people would do more to avoid a loss than to receive the same value of the gain. A monetary incentive framed as a loss will induce more effort to avoid it than will the same amount of incentive framed as a gain.

Framing experiment

Depending on the task, workers are organized in teams (or act individually).

• The “punishment treatment":

  For every week in which the weekly production average of your team is below K units/hour, the salary enhancement will be reduced by RMB 80 …

• The “reward treatment":

  You will receive an RMB 80 bonus for every week the weekly production average of your team is above or equal to K units/hour.

• The “gift treatment":

  For the next 4 weeks from July 28 to August 23, in addition to your standard salary, you will receive a one-time salary enhancement of RMB 320

• The control group: Fixed wage

• The bonuses are large, more than 20% of wages. A strong response to incentives is expected.

• Productivity data available only at the end of the week, so loss aversion is exploited to its full. There is no feedback on performance to adjust effort.

• All participants took part in all experimental treatments. Thus, comparisons possible between treatments, teams, and individuals.

Why is framing manipulation more powerful for groups than individuals?

• Peer pressure: a worker does not want to be the one who brings punishment to the whole team.

• Individual loss aversion is magnified by team size: if there is one very loss averse worker, s/he will stress the whole team out.

• Individuals are older and have longer tenure at Wanlida, and also more likely men, all these characteristics reduce the strength of the framing effect.

• Incentives work by affecting the effort of individual workers: The incentive or targeting effect

• Incentives work by changing the composition of workers in the firm: The sorting or selection effect

• Some evidence that framing plays a role, especially in teams.

  —>In theory, both incentive and sorting effects increase with the strength of incentives.

  —>Does it mean that stronger incentives will always lead to better performance? More next week.