gripping the reins: is there an endowment effect for power? foster.pdfgigi foster (unsw) matt nagler...
TRANSCRIPT
Gripping the Reins:
Is there an endowment effect for power?
Gigi Foster (UNSW)
Matt Nagler (CUNY)
June 2019
Sydney Market Design Workshop, UTS
Motivation
“Power tends to corrupt, and absolute power corrupts
absolutely” – Lord Acton
How can economists interrogate the effects of power?
Fehr et al AER 2013 Lure of Authority: Higher effort
provision by principals in principal-agent games, and
evidence that powerful principals (seemingly irrationally)
want to retain their power
Because of “disutility for being overruled”? (p. 1327)
A behavioral economics observation: The endowment
effect for goods
Research question
(it’s in the title) but what does “power” mean?
(from Wikipedia, excluding the physics/electrical definitions):
(A) the ability or capacity to do something or act in a particular
way
(B) the capacity or ability to direct or influence the behavior of
others or the course of events
Power in the real world often brings with it a potential to get
resources; actual resources themselves; and feelings.
How to translate these ideas to the lab?
Approach
Construct power differences between pairs of people via
two standard lab games (DG and PGG)
Validate that a power difference is perceived in each case
Use the multiple price list mechanism to elicit WTP and
WTA for that power
Run a third stage to elicit the endowment effect for a
tangible good (a mug or water bottle)
With the data:
1. Estimate the EE for the two different types of power
2. Compare the estimated EE for power to the EE for the
tangible item
3. Predict the EE for power based on observables
Design sketch
Earn endowment (high or low, depending on performance)
EE setting 1
EE setting 2
EE setting 3
Questionnaire
Payout
Main stages consist of TI, DG, PGG.
one 16-subject and one 32-subject session: ordering is
PGGTIDG
one 24-subject session: DGTIPGG
N=72 total
“Main stages”, each consisting of multiple rounds
(next slide…)
Based on performance in a boring task (“Finding
Letters on Pages,” Azar 2019)
Replenishes the wallet in each round
Main stagesEach main stage (TI, DG, PGG) consists of multiple rounds.
1st 5 rounds: subjects practice, playing two roles simultaneously
• the more and less powerful roles in DG and PGG
• owning vs not owning “tokens” in the TI game
• matched randomly to members of their half of subjects (high or low
endowment)
Then (in DG and PGG) power perception elicitation:
• half: just before assignment to roles (“hypothetical”)
• other half: just after assignment (“non-hypothetical”)
Random assignment to roles (powerful/less powerful, own/don’t own)
Then market-clearing trades of roles
Then play one round, as one (randomly assigned or traded) role or as the
(randomly assigned or traded) owner or potential buyer of a TI, that “counts”
Instructions for responding to the MPL mechanism are presented either before the TI
practice rounds (if TI is the first main stage) or after the DG/PGG practice rounds (if
DG or PGG is the first main stage)
Design sketch: Streamlined
5 practice rounds, playing both roles
Power perceptions elicitation (DG and PGG), before
or after random single role assignment
MPL market-clearing role trades
Play a round that counts, with each subject in one
role
[“Counting” means the round is entered into a group of three,
from which one is chosen to be paid out at the end of the
experiment (experimental dollars converted to AUD at an
exchange rate that gives us a match to the average per-
hour payment level convention at the Bizlab)]
Parametrizing powerIn DG: The dictator is more powerful than the receiver (we use
neutral language to describe the roles… shape names)
In (dyadic) PGG: Create a power difference by giving one party the
role of “civilian” and the other the role of “gendarme.” The gendarme
has the option to pre-commit to punishing or rewarding the civilian
depending on her ex-post observed level of contribution to the public
good.
Marginal per capita return for both players of 0.75. Assume
endowment of 100. The gendarme may commit up to 25 of his
post-contribution ED to reward (add 25 ED) or punish (destroy
75 ED) his partner. Max contribution to the public good is 25
ED.
With these settings, neither player can come away with less than
zero income from a round of the PGG
What’s on the questionnaireAge, gender, country of identifying culture, program of study,
international student status
To what degree are you influenced by your peers?
Position on the political spectrum
Mother’s and father’s education levels
When you were a child, how well off were your parents or
guardians, in a financial sense?
Rosenberg generalized trust item
Pearlin and Schooler Locus of Control item battery
Rosenberg self-esteem item set
Initial questions
Q1: What is the EE for the tangible item?
Q2: What is the EE for the dictator role in the DG?
Q3: What is the EE for the gendarme role in the
PGG?
Q4: How do the above values compare?
Q5: Did subjects really perceive power differences
across the DG and PGG roles?
Q6: What information from the questionnaire predicts
the elicited values for the tangible item and/or the two
types of power, and the perceptions of power?
Initial results: Power perceptions (adapted
from Anderson et al 2012)
Role/Type Power perception (1 to
7 scale)
Gendarme (n=36) 4.03*Civilian (n=36) 3.51*Dictator (n=36) 4.08*Receiver (n=36) 3.02*
* p-value of one-sided t-test of difference between role
types lies between 0.05 and 0.01
There is a perceived power difference of between 0.5 and 1
points on the 7-point scale of power perceptions.
In the ____ role in the game,
1. I could get the other player to do what I want.
2. My wishes would not carry much weight. (RC)
3. I think I would have a great deal of power.
4. Even if I try, I would not able to get my way.
(RC)
5. If I want to, I would get to make the decisions.
Initial results: The estimation target, High
Endowment TypesRole/Side of Market Elicited value in
experimental dollars
Gendarme WTA (by gendarme) 131.11**
Gendarme WTP (by civilian) 46.67**
Dictator WTA (by dictator) 163.89**Dictator WTP (by receiver) 84.44**Mug WTA (by owner) 104.44*Mug WTP (by non-owner) 69.44*
WTA/WTP results, High Endowment Types (n=36)
* p-value of one-sided t-test of difference between role types lies between 0.05 and 0.01
** p-value of one-sided t-test of difference between role types is smaller than 0.01
There is a strong endowment effect for power and a modest
endowment effect for the tangible item for high-type subjects.
WTA/WTP = 2.8 for the gendarme role power; 1.9 for the
dictator role power; and 1.5 for the mug
Initial results: The estimation target, Low
Endowment TypesRole/Side of Market Elicited value in
experimental dollars
Gendarme WTA (by gendarme) 69.44*Gendarme WTP (by civilian) 46.67*Dictator WTA (by dictator) 81.67*Dictator WTP (by receiver) 50.56*Water bottle WTA (by owner) 50.56Water bottle WTP (by non-owner) 46.11
WTA/WTP results, Low Endowment Types (n=36)
* p-value of one-sided t-test of difference between role types lies between 0.05 and 0.01
There is a modest endowment effect for power and no
significant endowment effect for the tangible item for low-type
subjects.
WTA/WTP = 1.5 for the gendarme role power; 1.6 for the
dictator role power; and virtual unity for the water bottle
Initial results: What predicts rein-gripping?
Predicting WTA for the Gendarme Role in the Public Goods Game
Independent Variables
Full sample
(n=36)
High types
(n=18)
Low types
(n=18)
Constant 59.19 *** 57.03 ** 77.13 *
(18.90) (22.92) (34.69)
Age>22 56.22 *** 86.15 ** 20.05
(17.01) (28.95) (24.04)
Left-leaning politically 66.21 *** 109.55 ** -5.26
(21.30) (33.63) (31.15)
Right-leaning politically 58.48 *** 63.44 * 56.15 *
(17.97) (34.23) (25.82)
Trusting -63.87 *** -77.11 * -65.76 **
(17.17) (35.99) (27.37)
Econ program student 47.14 28.96 123.89 **
(30.03) (48.87) (50.61)
Eng program student 30.62 * 49.80 * 3.61
(15.68) (25.63) (25.49)
High type 19.83
(16.03)
Adjusted R-sq 0.5522 0.5305 0.2899
*** p < .01 ; ** .01 < p < .05 ; * .05 < p < .10. Dummies for session included (not shown).
Initial conclusionsOur power-game manipulations appear to “work”: power
differences between roles are perceived by subjects
We estimate modest to strong endowment effects for the power
residing in both the dictator role and the gendarme role
The tangible item endowment effect is not in evidence for low-
endowment (water bottle) subjects, but modest and significant
for high-endowment (mug) subjects
Rein-gripping is strongly predicted by age, political extremity,
and lack of trust, particularly for high-endowment types