Participants are divided into equal numbers of "workers" and "employers," with half of the workers being "green" and half "purple." Each worker sees a random cost of investing in human capital and then decides whether to incur this cost and invest. Workers are paired with employers, who can see the worker's color, but not the cost or investment decision. The employer gives the worker a test, with a good test result being more likely if the worker invested. Seeing the worker's color (purple or green) and the test result, the employer must decide whether to hire the worker or not. This experiment implements the Coate/Loury model of statistical discrimination, which may have asymmetric equilibria in which the employers favor one color of worker, expecting them to invest more often. In addition, there may be symmetric equilibria with no color effects, which are also common in class experiments.

If it develops, statistical discrimination is based on a self-confirming cycle of low employer expectations and low worker aspirations for the less favored group (Holt, Markets, Games, and Strategic Behavior, 2006, Addison-Wesley, Chapter 32).