Experimenter Demand Effect in the Lab

Using four standard results from the experimental economics literature, we test their sensitivity to experimenter demand effects. Following de Quidt et. al 2018, we test if and how demand effects affects the standard predictions of probability weighting , hyperbolic discounting , endowment effect and charitable giving in a dictator game. Specifically, we ask if using positive and negative experimenter demand effects appropriately can reverse the standard predictions observed in the literature.

2020-12-04

2021-01-31

sooner payment, WTP for lottery, WTA for lottery, giving to charity

Hyperbolic discounting : amount chosen to be received as the sooner payment out of $10 endowment
Probability weighting : WTP for two lottery, one having a 10% chance at winning $10 and another having 90% chance of winning 10$
Endowment effect : WTA and WTP for each of the lotteries
Charitable giving : amount allocated from the endowment of $10 to send to charity

We use a within-subject design in an individual decision-making study. Each participant is asked to make eight decisions, for each they begin with a $10 endowment. They are told they would be paid for only one of them.

For probability weighting, the participants are endowed with $10 to purchase a lottery ticket once for a 10% lottery of winning $10 and another for a 90% lottery of winning $10, while a random price for the lottery is determined between $0.01 and $10. The participant is asked to report their highest willingness to pay (WTP) for each of the lotteries. Using BDM, if the lottery price is greater than WTP, the participant do not buy the lottery and earn the $10.00.
If the lottery price is less than or equal to WTP, the participant buy the lottery and earn $10.00 minus the lottery price, plus the outcome of the lottery.

For endowment effect (in addition to WTP elicitation mentioned above), the participants are also asked to report their lowest WTA for both types of lotteries, and additionally they also receive $10 endowment for these decisions with a lottery price determined randomly. If the lottery price is less than the reported WTA, the participant does not sell the lottery and earn $10.00 plus the outcome of the lottery. If the lottery price is greater than or equal to their WTA, they sell the lottery and earn $10.00 plus the lottery price.

For hyperbolic discounting, the participants are asked to make two decisions, each asking them to decide how they would like to divide an endowment of $10, between a sooner and later payment, with the later payment. For one decision, individuals are asked to choose between today and a week from today, while in the other they are asked to choose between tomorrow and a week from tomorrow. For both the decisions, the participant earns an additional 20% interest on the amount they decide to receive later.

For charitable giving, we use a dictator game with the respondent framed as a charity. The participants are asked to decide how much of a $10 endowment, they would like to donate to the charity. For one decision, the charity receives the amount they donate while in another the donations are matched, i.e. the charity receives twice the amount of donation.

We have three treatments, in which we vary the kind of experimenter demand, following de Quidt et. al (2018). In a positive (negative) treatment, participants are urged to choose higher (lower) numbers, while in the control treatment nothing additional is mentioned.

We use a within-subject design in an individual decision-making study. Each participant is asked to make eight decisions, for each they begin with a $10 endowment. They are told they would be paid for only one of them.

For probability weighting, the participants are endowed with $10 to purchase a lottery ticket once for a 10% lottery of winning $10 and another for a 90% lottery of winning $10, while a random price for the lottery is determined between $0.01 and $10. The participant is asked to report their highest willingness to pay (WTP) for each of the lotteries. Using BDM, if the lottery price is greater than WTP, the participant do not buy the lottery and earn the $10.00.
If the lottery price is less than or equal to WTP, the participant buy the lottery and earn $10.00 minus the lottery price, plus the outcome of the lottery.

For endowment effect (in addition to WTP elicitation mentioned above), the participants are also asked to report their lowest WTA for both types of lotteries, and additionally they also receive $10 endowment for these decisions with a lottery price determined randomly. If the lottery price is less than the reported WTA, the participant does not sell the lottery and earn $10.00 plus the outcome of the lottery. If the lottery price is greater than or equal to their WTA, they sell the lottery and earn $10.00 plus the lottery price.

For hyperbolic discounting, the participants are asked to make two decisions, each asking them to decide how they would like to divide an endowment of $10, between a sooner and later payment, with the later payment. For one decision, individuals are asked to choose between today and a week from today, while in the other they are asked to choose between tomorrow and a week from tomorrow. For both the decisions, the participant earns an additional 20% interest on the amount they decide to receive later.

For charitable giving, we use a dictator game with the respondent framed as a charity. The participants are asked to decide how much of a $10 endowment, they would like to donate to the charity. For one decision, the charity receives the amount they donate while in another the donations are matched, i.e. the charity receives twice the amount of donation.

We have three treatments, in which we vary the kind of experimenter demand, following de Quidt et. al (2018). In a positive (negative) treatment, participants are urged to choose higher (lower) numbers, while in the control treatment nothing additional is mentioned. Thus a positive (negative) demand effect urges one over (under) weight the lottery chances, be more (less) patient in hyperbolic discounting decisions and finally more (less) generous in dictator games.

computer program

individual

No

800

800

100 by each treatment arm

We use the effect sizes suggested by de Quidt et al. (2018) We use effect sizes and variability in their data to calculate the sample size needed to generate >90% power across all of our hypothesis.