ECO331: Midterm Questions
Experiment Design: Internal & External Validity
1. Experiment designs that vary levels of the focus variable only across subjects are called
between-subject designs, and those who use several different levels for each subject are
called within-subject designs. Between subjects – each subject receives one treatment.
Within subjects – one subject receives all treatments.
E.g. test levels of OLD vs NEW soles – you can give some boys old material, and others new
material to test the differences. However, to account for the differences in lifestyles between
subjects, you could conduct a within-subject design by placing one shoe with OLD and the
other with NEW. This helps to eliminate noise, or the differences between people.
Problems with within-subject experiments – learning over many trials, order effects, easier
identification of experimenter demand. Solutions: Randomly assigned orders, crossovers
(ABA), dual trial.
2. In GNR, what experimenter demand effects might we be worried about? Consider both
strong and weak.
5. Niederle and Vesterlund found that although there are no gender differences in
performance, more men choose tournament rather than women. This shows that men are
overconfident and embrace competition, while women shy away from it. Gender differences
in top level management may be partially explained by preferences for competition.
Different from GNR, it looks at self selection, and who chooses to compete.
6. They used a within-subject design, because each subject receives all the different scenarios
and treatments. No, we shouldn’t be concerned, because there were only four trials, and the
relative results were not shown to the subjects – so regardless of whether they knew
experimenter demand, they would choose based on their best interests.
7. Perhaps instead of just a small sample of students, if they did similar tests for athletes (who
are by nature more competitive), or run tests for professionals in the field, they would see
different results. If they did this experiment in other parts of the world, perhaps the results
would differ as well. (External Validity)
People and Probabilities
6. Insensitivity to predictability is when we don’t pay enough attention to random factors, and
predict future performance based on one instance. This describes the bias in which people feel
comfortable making intuitive predictions based on insufficient information. In the Barberis