4. Incentives Flashcards
What is the traditional economic view on how principals motivate agents?
Principals motivate agents through explicit incentive contracts, promotions and other forms of material reward or sanctions
Describe the standard agency model
Worker chooses effort e that costs c(e), disutility of effort is increasing in e and convex. Output x depends on effort e, skills s and luck ø with distribution m(ø). Firms observe x and pay wage w(x)
Describe the set up Gneezy & Rustichini 2000 experiment 1
-50 quiz problems IQ experiment
-160 students receive a show up fee of NIS 60 (roughly £10)
-4 Treatments of payment for a correct answer.
No payment
0.1 NIS
1 NIS
3 NIS
Results of Gneezy & Rustichini 2000 experiment 1
No payment 28.4 Qs
0.1 NIS 23 Qs
1 NIS 34.7 Qs
3 NIS 34.1 Qs
87% of principals chose to match with agents who were paid 0.1 NIS rather than no payment
Set up of Gneezy & Rustichini 2000 experiment 2
-180 high school students in Israel.
-Pairs of subjects go from house to house to collect donations for charities.
-3 treatments; no pay, 1% commission, 10% commission
Results of Gneezy & Rustichini 2000
Experiment 2
No payment 239 NIS
1% 154 NIS
10% 219 NIS
Performance doesn’t increase with pay. Again 77% of principals chose to be with groups with the 1% added incentive compared to no payment
Summary of findings from Gneezy & Rustichini 2000 experiment 1 and 2
- no monotonic relationship between incentive and performance (effort)
- low incentives can crowd out intrinsic motivation to perform a task
- if incentives are large enough the relative price effect dominates crowding out effect and higher payments result in higher effort
Possible reasons for undesired effect of incentives?
- incentives reduce self image motivation
- incentive may serve as a signal of unattractive mess of a task or of a negative implication
- small compensation might be perceived as insulting, people might be afraid to look cheap
- compensation of any amount can be perceived as insulting for particular activities
Gneezy & Rustichini 2000 exp 3 set up
-Field experiment at 10 daycare centres in Haifa.
-Control treatment- no fine for lateness.
-Fine treatment- w1-4 no fine, w5-16 fine, w17-20 no fine.
-Fine was $2.5 per child for being 10 or more mins late
Results of Gneezy & Rustichini 2000 exp 3
-In phase 1, no difference between groups.
-In phase 2, Fine group increases in late coming to 2x as high as before and of control group.
-Late comings remain high after fine is removed.
Explanation for results of Gneezy & Rustichini 2000 exp 3
-Change of perception.
-Before fine- looking after children after contracted time is a generous non market activity.
-After fine- service has a price and can be bought as needed “a fine is a price”
Examples of incentives being too high in field evidence
-Paserman 2010- grand slam tennis players could increase probability of winning by 25-30% if they could avoid choking.
-Apesteguia & Palacios Huerta 2010- team who shoot first in penalties win 60.5% of the time
-Dohmen Et al 2008 no evidence of choking on pens
Design of Ariely Et al 2009
-87 subjects played 6 games requiring different skills
-3 treatments of low(4), medium(40), high(400), rupees per game based on performance
-full payment for very good performance, half payment for good performance, nothing for poor performance
-Very good performance earns an amount equivalent to half of the mean yearly consumer expenditure in the village in high treatment
Results of Ariely Et al 2009
-No significant difference between low and mid (except for labyrinth)
-performance always lowest in high
-no difference of among categories (creativity, memory or motor skills)
-so yes people choke big time
What do standard Neoclassical models say about higher wages on labour supply?
It assumes forward looking subjects max discounted lifetime utility. So predicts workers should work more hours when wage is high ie positive wage elasticities
What is the empirical evidence from cabdriver studies on wage elasticities?
-Cab drivers can freely choose their hours.
-Wages vary across days due to demand shocks.
-High demand= less time searching for customers.
-Wages tend to be correlated within days and uncorrelated across days. All studies find negative elasticities
What is an explanation for the cab driver results?
-Reference dependent preferences.
-Workers have a daily target income and evaluate outcomes as losses and gains relative to this reference point.
-They are loss averse and have diminishing sensitivity.
-Alternatively quiet days could just be less stressful
Fehr & Goette 2007 experiment set up
-Randomised field experiment with bicycle messengers in Zurich.
-Bicycle messengers deliver parcels and are paid on commission.
-Treatments- 4 week increase in commission rate by 25%. Lab experiment measuring loss aversion of same workers used in tandem
Results of Fehr & Goette 2007
-Workers sign up for more shifts but pay increase causes a decrease in revenue/effort per shift by 6%.
-Net effect is still positive
-Those with higher degree of loss aversion have stronger negative impact of wage increase in effort per shift
-Those who weren’t loss averse in lottery choice task don’t show a significant effort reduction
Falk & Ichino 2006 experiment set up
-24 high school students in Switzerland prepare mailing of a questionnaire conducted by uni of Zurich.
-4 hrs for £60. Single treatment- each subject works alone, pair treatment- two subjects work in the same room
Results of Falk & Ichino 2006
-Positive peer effects.
-Standard deviation is significantly smaller than between pairs.
-Average output is 16.3% higher in pair treatment.
-Less productive workers show stronger reaction than high productivity worker (“bad apples” gain from “good apples”)
Mas & Moretti 2009 experiment design
Scanner data from a US supermarket chain. 6 stores over 2 years, data from 370 cashiers. Measured avg number of items scanned over 10 min period
Results of Mas & Moretti 2009
-Significant productivity spill overs- the presence of high productivity workers raises the productivity of other workers especially ones who are less productive.
-The effect manifests itself at the time of entry of more productive co workers and is persistent.
Possible explanations for Mas & Moretti 2009 results
Social pressure - shame reputational concerns, fear of sanctions
Pro-social preferences - altruism, guilt
Data supports social pressure as workers aren’t affected by presence of a highly productive co-worker when the co-worker can’t observe them.
What is the multi-tasking problem?
Workers will put effort only in those activities that are incentivised (Holmstrom & Milgram 1991)
What are incomplete contracts?
Contracts where tasks and duties aren’t fully specified. Relational contracts are based upon a relationship of trust between the parties
What does Akerlof 1982 find when trying to find if workers reciprocate fair or high wages with higher effort in one shot interaction?
Lab- workers reciprocate higher wages with higher effort but effect is small
Field- confirm lab results, small sample size, also strong impact of negative reciprocity
Describe design of Kube Et al 2012
Field experiment, Students recruited for data entry task not aware of experiment. Treatments:
Baseline- €36 for 3hrs (one-shot)
Money- surprise 20% pay increase €43
Bottle- thermos flask wrapped as gift
Price tag- same bottle with price attached
Choice- subjects choose between money and bottle
Oragami- money presented as oragami
Results of Kube Et al 2012
-Orgagami 29.3%, bottle 25%, choice 24.5%, price tag 21%, increase in performance compared to baseline
-80% choose money over bottle, money has no significant impact on workers effort 5.2%
Describe the design of Brown, Falk, Fehr 2004
Experimental labour market, 15 periods.
-7 firms, 10 workers, firms offer contracts, workers accept or reject then choose effort level. Treatments
1. Complete contract condition: contracted effort e is automatically enforced
2. Incomplete contract condition (ICF)- fixed IDs
3. Incomplete contract condition (ICR)- random IDs
Results of Brown, Falk & Fehr 2004
-Wages always higher in ICF than ICR
-Wages fall in C over time
-high effort if effort is enforceable
-if effort isn’t enforceable: reputation matters- effort always higher in ICF than ICR but effort drops sharply in final period of ICF
-high effort is rewarded with contract renewal
-in ICF total surplus higher than ICR, rents shared almost equally, higher rents in longer relationships
What explicit incentive systems might principals use?
Fines/ penalties
Monitoring and controlling systems
Describe the set up of Fehr & List 2004
-Principal and agent both receive endowment of 10 ECU
-Principal can transfer x to agent, agent receives 3x principal announces desired back transfer.
-Agent is informed about desire and decides on actual back transfer. -Treatments of normal game and with punishment where principal can fine the agent 4 tokens
126 subjects of uni of Costa Rica
76 CEOs from coffee industry
Results of Fehr & List 2004
-refraining use of punishment option nearly doubled the actual payback from CEO- less pronounced effect on students
-use of punishment crowds out trustworthiness
-overall efficiency is highest if punishment is available but not used
-60.5% of CEOs and 79.4% of students use the punishment
Describe the design of Falk & Kosfield 2006
Principal agent game. Agent has initial endowment of 120, principal endowment is 0, agent chooses effort level x and has costs c(x)= x
Payoff P= 2x
A= 120-x
Before agent decides on x, the principal determines the agents choice set.
Treatments
Low control c5
Medium control c10
High control c20
Prediction of Falk & Kosfield 2006
Selfish agents will always choose minimum. Principals who want to maximise their payoff should always control the agents choice set
Results of Falk & Kosfield 2006
-Hidden cost of control outweigh the benefits in all treatments.
-Average performance is higher if principle doesn’t control.
-Strong heterogeneity among agents. -The majority of principals don’t control the agent.
-Expectations coincide with agents average performance in most of the cases
Evaluate work from policies
Potential hindering of positive peer effects and creativity but it signals trust to workers. Firms positively surprised by WFH (Barrero Et al 2020)
In Brown Et al 2004, in ICR what % of workers give min effort?
43% of workers
In Brown Et al 2004, in ICF what % of workers give max effort?
34% of workers
What happens when there is an enforcement problem in Brown Et al 2004?
-Markets resemble bilateral trading islands rather than competitive markets.
-Firms voluntarily restrain the set of trading partners by making private offers to the previous employee.
-Rent sharing and long term relations prevail
In Fehr 2004 how do CEOs compare to students?
CEOs exhibit more trust than students and reach higher efficiency levels
In Falk 2006 how do agents view restriction?
As a negative signal of distrust, may also assume principal has low expectations.
Main message from Falk & Kosfield 2006
That control and explicit incentives entail hidden costs which should be taken seriously. The message isn’t that it is always better for principles to trust rather than control
Falk 2006 what findings are there for 3rd parties?
Agents perform better in EX10 treatment where control was implemented exogenously. This finding corrugated the important role for 3rd parties like consultancies
When does the principal agent theory have little predictive power?
If factors other than money and effort play a role. Explicit incentives might crowd out implicit incentive
Peer effects
When output/productivity of individual I increases when output/productivity of j increases and nothing else changes
What is the problem with peer effects in happenstance data?
Peer effect could be the result of endogenous sorting according to local or personal attributes
Payoff functions in Brown, Falk & Fehr 2004
Firms payoff 10e-w if contract was concluded
0 if no contract was concluded
Workers payoff w-c(e) if contract was concluded
5 if no contract was concluded
Payoff functions in Fehr & List 2004
Principal payoff: 10-x+y
Agent payoff: 10+3x-y (-4 if principal chooses to fine y<ŷ
Game theoretic predicts in both treatments in Fehr & List 2004
- Trust game x=0, y=0
- Trust game with punishment
X=1(2), ŷ=3(4), y=3(4)
Stats associated with Mas & Moretti 2009
-10% increase in average co-worker permanent productivity is associated with a 1.7% increase in workers effort
-workers are unresponsive to the presence of co-workers who they infrequently overlap with
-the optimal mix of a given shift is the one that maximises skill diversity
-optimising shifts may not result in high profits because it may involve paying higher wages as compensation for workers not being able to pick their shifts
What do Fehr & Goette 2007 suggest should be taken more seriously?
Heterogeneous preferences since only those workers with loss aversion reduce effort per shift when paid a high wage
What model fits the data well in Fehr & Goette 2007 and how do their results compare to previous findings?
-Neoclassical model with loss averse preferences is consistent with observed decrease in effort per shift.
-it seems loss aversion drives the negative effect of wages on effort
-the results contrast sharply with small and insignificant substitution effects from previous studies
What % of bicycle messengers say they have a dollar amount in mind for the day after which earning is less urgent?
73%
Which model does Goette 2004 prefer?
RDP model preferred to nonseperabilities in labour supply model
What is a weakness of the Falk Et Ichino 2006 experiment?
There is no measure of peer effects over time- eg does it start off strong then diminish?
How can we interpret the effects measured in the Falk & Ichino 2006 study?
Effects measured in study are likely to be lower bound for effects in actual labour relations since subjects don’t know each other and only interact once
Give an example of incentives seemingly depending on prior achievement
Leuven Et Al 2010 found that providing incentives to 1st year uni students at Amsterdam for passing all first year requirements had a positive effect on performance for most able but a negative impact on low ability students. Bettinger 2010 had similar results
What do Ariely Et Al 2009 find about public and private donations with incentives?
If individuals behave pro socially in private, incentives work well. However, in cases in which desire to behave prosocially is due to image motivation being paid in public crowds out prosocial behaviour
How does Gneezy Et Al 2011 evaluate the affects of incentives over time?
Large enough incentives work in changing habits in short and medium run but in the long run the desired change in habits can disappear
What is the technical explanation of the results found in Gneezy & Rustichini 2000 in exp 3?
Contracts, social or private, are usually incomplete and regulate an interaction in a situation of incomplete information. The introduction of a reward or punishment modifies some of the terms of the contract but also provides new information. The new behaviour produced by contract is a response to the combination of a new payoff structure and the new info
Papers to reference for question on incentives/ incentives and social preferences
-Gneezy & Rustichini (2000): 3 exps
-Ariely Et Al (2009): skill games India
-Fehr & Goette (2007): bikes Zurich
-Falk & Ichino (2006): envelopes
-Mas & Moretti (2009): scanners
-Kube Et Al (2012): library
-Brown Et Al (2004): 7 firms, 10 workers
-Fehr & List (2004): trust game CEOs
-Falk & Kosfield (2006): PA game with restricted option
