Fall 2017

Story so far?

How a decision is presented affects people's actions

  • endowment effect
  • status quo bias
  • WTP \(\neq\) WTA
  • loss aversion

Lots of evidence from the laboratory, less from the "field".

Also, less evidence (though some) from "economically relevant" decisions.

What do haters say?

These phenomena are explained by…

  • bad experimental design
  • people who don't or haven't had time to learn

Enter Hossain & List

  • Data from a natural field experiment at Wanlida Group.
  • Wanlida \(\Rightarrow\) top 100 electronics manufacturers


  • "Reward/Gain": "positive frame" – if the week's average per-hour production reaches a threshold, a bonus is paid at the end of the pay period. Weekly bonus = RMB 80.
  • "Punishment/Loss": "negative frame" – employees provisionally given the bonus before the week begins, if the week's average per-hour production fails to reach a threshold, the bonus is taken away at the end of the pay period. Weekly bonus = RMB 80.
  • "Gift": just a control (also company wanted it): one-time four-week bonus of RMB 320

Main Results

  1. Bonuses (of either type) increase productivity for groups and individuals.
  2. Punishment/Loss frame outperforms the gain frame for groups and individuals (less robust) \(\Rightarrow\) an \(\uparrow\) of 1% of production (this is economically significant) at low marginal cost
  3. Effects are stronger for groups than individuals.
  4. Effects persist for whole period of study (long-term).

Implications and Significance

  1. Incentives are powerful & behavioral econ has a role in the workplace.
  2. Results complement industrial psch & expand toolset of practitioners.
  3. Social structure & institutions affect/constrain how people act in the workplace.

Institutional Detail

  • Teams of workers (14 people) engaged in relatively flexible production
  • Individual inspectors (of products) involved in incentive schemes too
  • Can compare pre- and post-experiment:
    • productivity in terms of output (DVD chips)
    • quality of productivity with error rate
  • 4 weeks in a treatment; 1 week after a treatment ended

Additional desing issues

  • all payments made to all teams at the same time
    • improves credibility
    • limits time preference concerns
  • compare within and between observations
    • same observation exposed to multiple treatments
  • workers didn't know about experiment
  • management wanted targets achieved in 60-80% of cases (morale)
    • targets calibrated with pre-experiment data
  • minimized information transmission between groups

Overall productivity

Defect rate

Group effects over time

Individual effect over time