Random Assignment Lab
Chapter 3. Could chance alone produce the gap?
In the resume study, 392 of 4,870 resumes received a callback. The gap between the two name groups was 3.2 percentage points. Suppose the names made no difference. The 392 callbacks would then fall into the two groups by chance. The tool deals the same 392 callbacks into two groups of 2,435 at random, 1,000 times, and plots the gap each time. The callback totals are real. Source: Bertrand and Mullainathan (2004).
The bars show gaps that chance produces when the name has no effect. The dark line marks the gap the study found. If almost no shuffle reaches the line, chance is a poor explanation for the result. Random assignment is what allows this comparison, because it makes chance the only other explanation.