Experiments in the Classroom: Part I

Last semester, I decided to run a few experiments with the students in the introductory economics class I was teaching. The first had to do with the power of chocolate in improving test scores. This theory arose from my mother-in-law, who always gave her daughter a chocolate on days when she had to take an important test at school.

To begin with, I offered the students two choices of sweets before their in-class test – a chocolate and a non-chocolate option. I asked them to mark their choice on their test paper.

After the in-class test, it transpired that there was no difference between the grades of chocolate-eaters and non-chocolate eaters. But we were worried. What if the choice of chocolate was correlated with something about the student? Perhaps chocolate did help test performance, but this effect was washed out by the propensity of lazy students to eat more chocolate?

To solve the problem, we decided to implement a unique solution: the world’s first (so far as I know) chocolate-test-taking randomised trial. Alternate test papers were marked on the back “chocolate” or “non-chocolate”. Before starting the test, students could grab a sweet. Afterwards, we tabulated the results.

Chocolate group – mean of 12.8
Non-chocolate group – mean of 13.2

There was no statistically significant difference between the groups, indicating that the small difference in means was likely to have just been statistical noise.

Our conclusion: theories about the effect of chocolate on test-taking performance are not supported by the evidence. The only reason to eat chocolate before a test is that you like the taste.

This entry was posted in Economics for Government Course. Bookmark the permalink.

2 Responses to Experiments in the Classroom: Part I

  1. Kevin Cox says:

    Perhaps the effect your mother in law “observes” is because she is doing more than giving chocolat. She is giving encouragement and showing that exam days are special. This in turn means that your wife was better disposed to examinations and so performed better than she would have otherwise. This hypothesis would be difficult to test with a randomised trial but for humans I suspect it is more likely than the effect of chocolat (or money) on the ability to do exams. In other words randomised trials involving human emotions and reasoning are very difficult because the important variables are hard to measure.

  2. Andrew Leigh says:

    Kevin, I think your first 3 sentences may well be right. But the last 2 sentences suggest that you may have missed the point of the randomised trial, which was to test just the chocolate effect – sans love and care.

Comments are closed.