Note: Below is intended to get some ideas going on good class-room experiments and not intended as a protocol for doing one. I will post at a later time on some existing experiments and also potential on some protocols if I can work some out. Suggestions welcome.
Thinking of the following simple class experiment to illustrate both the efficient market hypothesis and the extent of overconfidence of being able to beat indices by lay people. Basic idea of the set up is below. Everything will be recorded anonymously though students will be able to identify themselves in the results session if they remember the stocks they picked. Could potentially be conducted as a web survey, with the advantage that students would be able to look at the stock index in their own time and not have to remember the stocks.
1. Email the class a link to google finance and some simple instructions for how to look up prices of stocks on the NYSE.
2. Tell them that they will be asked to pick stocks on the NYSE in class as part of a classroom exercise. Tell them to write down the names of 12 stocks they would like to pick. (Hard to think of how to enforce this part without using online).
3. At start of class, hand each student a paper survey asking them to allocate a hypothetical 1000 dollars across 12 stocks in any proportion they wish. Tell them that they will hold this portfolio for one month at which stage it will be evaluated in terms of rate of return (excluding any dividends). Later these can be keyed into separate portfolios on the google finance toolbar.
4. Ask them how confident they are that they will outperform the overall market index for the NYSE
5. Ask them how confident they are that they will outperform the class average return
6. Elicit a few simple financial capability questions (and gender)
7. One month later:
- plot average rates of return across the class
- show correlation with degree of confidence of beating market; beating class average; gender and financial sophistication.
8. Basic prior is that picking 12 random stocks on the index should yield a normal distribution of average returns that are unrelated to any of the personal factors measured.