Experimental Design Details
We use the draw-from-a-bag version of the die-roll game.
The task will be explained through the actions of a single subject.
First, the subject enters the lab and sits at a computer terminal. There, he will read the instructions.
When it is his turn, the subject will be prompted to approach the table at the front of the lab. There, he will be randomly assigned a folder.
The subject will then randomly draw one envelope from the folder. The subject will privately open the envelope to reveal a value. This value is the subject's "true message". This is the number that the subject is asked to report for his payment.
Once the subject makes his draw, he will close the envelope and put it back into the folder. He will then go to the "writing station" to write down his report.
Once the subject has written down his report, he will enter a second room of the lab. There, he will submit his report to the second experimenter. The second experimenter pays the subject in cash for his reported value. Then the subject leaves the experiment.
The experimenter observation is implemented with the following design:
Each folder is uniquely numbered 0-99. Each subject is assigned a unique subject ID.
When the subjects are randomly assigned a folder, the experimenter writes down both the subject ID and the folder ID that he drew from.
We manipulate the elements of the folders so that some proportion of the folders allow for the true message to be observed by the experimenter (the proportion that is observable is determined by the treatment).
We construct "observable" folders and "unobservable" folders.
Unobservable folders are constructed as they would be in a traditional draw-from-a-bag version of the experiment. An unobservable folder contains 10 envelopes. Each of those ten envelopes contain a different value from the set {1, 2, 3, 4, 5, 6, 7, 8, 9, 10}.
Observable folders are constructed so that all 10 envelopes contain the same value. For example, an "Observable Folder 1" will contain ten envelopes, where each of those envelopes contain the value 1. If a subject draws from that folder, the experimenter knows with certainty that the subject's true message was 1.
In the 0% treatment, 0% of the folders are observable; in the 20% treatment, 20% of the folders are observable; and so on.
In the instructions, subjects are told the exact distribution over folder distributions.