Experimental and Behavioral Economics
Experimental Design
Sebastian Strasser
University of Munich
June 1, 2011
Purpose of Experiments
Main Objective: Causal inferences from observed behavior
I Random assignment of subjects (remember the urn!)
I Maximum of control over (unobserved) independent variables
Methodological cornerstones
I Salience and incentivized actions
I Replicability of findings
I No deception of subjects
The most important design questions - I
I One-Shot vs. Repetition
I Issues: Different set of theoretical equilibria, learning, boredom, stability of results, etc.
I Matching Protocol
I Partner Matching
I Identifiability vs. Non-Identifiability
I Stranger Matching
I Perfect Stranger Matching
I Within-subject design vs. Between-subject design
I Issues: Learning, heterogeneity, obviousness of research question
The most important design questions - II
I The role of information (who knows what at which point in time)
I Presentation of Instructions, framing, wording
I Order effects
I Control for (un-)observables (risk, social preferences, time, experience, beliefs, knowledge, socio-demographics, etc.)
I Confounds (experimenter demand effect, anonymity, etc.)
I Calibration of earnings
I Factorial designs (n ×m)
Some more specialized design issues
I Strategy vector method vs. direct choices
I Communication
I Group decisions (voting, etc.)
I Implementation of (potential) losses
I Control questions, quizzes, trial periods
I Real effort tasks
I Questionnaires
I Classroom experiments
I Pen-and-Paper vs. computerized (see zTree later)
The concept of independent observations
Essential to the statistical analysis
I Set of persons that inno wayinteracts with a different set of persons before or during the experiment in terms of ...
I ... actions
I ... information
I ... beliefs