In-class exercise
Task 1: Sampling from a population (45min)¶
Consider Section II.A of the paper.
Explain what “representative in terms of x, y, z” implies for the sample being representative of the U.S. population (that is, the subset working full time in a paid job) and what it does not imply.
Describe a sampling procedure that would help in terms of generating a representative sample along more dimensions. You do not need to worry about whether this is feasible to do for a private company.
Give an example of a study where a sample that is chosen to be representative along a few observables might lead to wrong conclusions.
Do you think that the sample is an issue for this paper? Why or why not?
Task 2-1: Survey Design (60min)¶
Describe the survey design in Roth and Wohlfahrt. What is the key idea underlying the design? Are the authors able to establish causality?
Briefly summarize their key results.
Task 2-2: Control of confounders (60min)¶
Assess the design of Roth and Wohlfahrt with respect to
anchoring,
experimenter demand,
and attention
effects.
Briefly check their results — do you expect any to be driven by any of these effects?