NEW UPDATE: Completed trials may now upload and register supplementary documents (e.g. null results reports, populated pre-analysis plans, or post-trial results reports) in the Post Trial section under Reports, Papers, & Other Materials.
Correspondence Bias and Memory
Last registered on December 28, 2019


Trial Information
General Information
Correspondence Bias and Memory
Initial registration date
October 10, 2019
Last updated
December 28, 2019 5:04 PM EST
Primary Investigator
Humboldt University of Berlin, WZB Berlin Social Science Center
Other Primary Investigator(s)
PI Affiliation
Carnegie Mellon University
PI Affiliation
University of Pittsburgh
Additional Trial Information
Start date
End date
Secondary IDs
This project aims to understand correspondence bias and its determinants, especially its interactions with imperfect memory. Correspondence bias is the tendency to draw inferences about a person's enduring characteristics from behaviors that can be entirely explained by the incentives/environment that induce them. In our first study, we identify the bias through a clean experimental design that deals with three potential confounders. Subjects in our experiment play a dictator game as the dictator in the first stage. In stage 2, they are randomly matched into a group of 4 to play a benign game and a malign game. Benign game promotes cooperation, while the malign game motivates them to harm each other. In stage 3, we elicit their benign premium, which is their willingness to pay, as the receiver in the dictator game, to choose the one they play the benign game with as their dictator. Furthermore, our design enables us to investigate whether there are ways to potentially alleviate correspondence bias or even eliminate it. The second study concerns the interactions between imperfect memory and correspondence bias. The central hypothesis is actions are easier to recall than the environment in which actions take place. Consequently, correspondence bias becomes more severe when people have to rely on their long-term memory to make a decision as it becomes even harder to attribute behaviors to the environment instead of a person's enduring characteristics. We test this hypothesis by comparing benign premium, our measure for correspondence bias, when there is no delay between obtaining information in stage 2 and making partner choice in stage 3 and when there is a 1-hour delay between the two.
External Link(s)
Registration Citation
Han, Yi, Yiming Liu and George Loewenstein. 2019. "Correspondence Bias and Memory." AEA RCT Registry. December 28. https://doi.org/10.1257/rct.3819-1.4000000000000001.
Experimental Details
Intervention Start Date
Intervention End Date
Primary Outcomes
Primary Outcomes (end points)
Benign premium: the amount of money one is willing to pay to be matched with the benign player
Primary Outcomes (explanation)
We construct our primary outcome as follows. In stage 3, subjects are asked to choose between two players they have played with in stage 2 to play a dictator game as the powerless receiver. We elicit their willingness to pay to be matched with the benign game player who they played the benign game with. The more they are willing to pay to be matched with her, the stronger their correspondence bias.
Secondary Outcomes
Secondary Outcomes (end points)
Secondary Outcomes (explanation)
Experimental Design
Experimental Design
We have two treatments in our experiment. The first one is the instant choice condition. This treatment has four major stages. At the first stage, all subjects make a decision as the dictator in the dictator game. At the second stage, they are randomly matched into groups of four to play the benign game and the malign game. The benign game encourages cooperation, while the malign game motivate subjects to harm each other. At the end of the second stage, subjects are able to see their outcomes in the benign game and the malign game and their partners' choices. Third, they are asked, as the receiver, to choose a dictator between a benign game partner and a malign game partner to play the first stage dictator game. We predict that there exists a benign premium: subjects are willing to pay to be matched with the benign game partner. Lastly, they are asked to recall the key information generated in the experiment: who they play with, which game they play, the payoff they get and what actions their partners choose.

The second treatment is the delayed choice condition. The setup of this treatment is the same with the instant choice condition with one exception: there is a 1-hour delay at the end of the second stage. After looking at the results of the benign game and the malign game, we ask subjects to take a break for 1 hour before they can proceed to stage 3. During that hour they are free to choose any activity they want. After the 1-hour delay ends, they have 10 minutes to log back into the study and finish stage 3. If they come back more than 10 minutes late, then they can only get half of their experimental payment excluding the show-up fee.

The central assumption is people can recall the actions of others but not the environment in which those actions are taken. In the delayed choice condition, subjects take a 1 hour break between stage 2 and stage 3. So to utilize the information generated in stage 2 when making the stage 3 decision, subjects need to reply more on their memory in the delayed choice condition. The hypothesis is the benign premium, our measure for correspondence bias, is larger in the delayed choice condition than in the instant choice condition. As we also ask subjects to recall the game they play and others' actions in stage 2, we can directly test whether the 1-hour delay indeed affects people's recall of previous events or not.

To deal with the potential selection issue, we also add a 1-hour delay in the instant choice condition. After subjects finish the recall questions, they enter a 1-hour waiting period. The rule of the delay is the same with the delayed choice condition. After the delay ends, they can see their payoffs. In both the instant and delayed decision condition, we inform the subjects in the recruiting advertisement that the experiment will last around 1 hour and 20 minutes but the actual decision time is around 20 minutes. Subjects are required to have at least 90 minutes of free time at the time of signing up in both conditions.
Experimental Design Details
Randomization Method
simple randomization with a pseudo-random random number generator.
Randomization Unit
Individual-level randomization
Was the treatment clustered?
Experiment Characteristics
Sample size: planned number of clusters
8-9 sessions
Sample size: planned number of observations
150-180 subjects in each treatment, 300-360 in total; 40 subjects per session
Sample size (or number of clusters) by treatment arms
150-180 subjects in each treatment arm
Minimum detectable effect size for main outcomes (accounting for sample design and clustering)
IRB Name
Human Judgement and Decision-Making
IRB Approval Date
IRB Approval Number
Analysis Plan

There are documents in this trial unavailable to the public. Use the button below to request access to this information.

Request Information
Post Trial Information
Study Withdrawal
Is the intervention completed?
Intervention Completion Date
December 04, 2019, 12:00 AM +00:00
Is data collection complete?
Data Collection Completion Date
December 04, 2019, 12:00 AM +00:00
Final Sample Size: Number of Clusters (Unit of Randomization)
Was attrition correlated with treatment status?
Final Sample Size: Total Number of Observations
Final Sample Size (or Number of Clusters) by Treatment Arms
167 control (instant decision), 180 treatment (delayed decision)
Data Publication
Data Publication
Is public data available?
Program Files
Program Files
Reports, Papers & Other Materials
Relevant Paper(s)