Several restaurant owners consider the same new shopping street, which has custom for a few of them. A team finds a fault late in the evening, and one person staying to repair it would spare everyone the consequences. In the first case too many may act, and in the second too few.
Both games test how a group of people in identical positions coordinate when each must answer yes or no without knowing the answers of the others. The market entry game shows how closely the number of entrants matches what the market can bear. The volunteer's dilemma shows how responsibility thins out as a group grows, so that an action which each member would take alone may not be taken at all.
In the market entry game, analyzed by Selten and Güth (1982), entering pays more than staying out as long as fewer firms enter than the market has room for. The game has many equilibria in pure strategies, in all of which the number of entrants equals the capacity or falls one short of it, and nothing in the game says which firms should be the ones to enter. There is also a symmetric equilibrium in which every firm enters with the same probability. With the payoffs used here that probability is the capacity minus one divided by the group size minus one.
In the volunteer's dilemma of Diekmann (1985), everybody receives a benefit if at least one member pays a cost. No strategy is dominant, and in the symmetric equilibrium each member volunteers with a probability that falls as the group grows, and the probability that nobody volunteers rises with group size toward the ratio of the cost to the benefit. With a cost of 4 and a benefit of 10, a group of four leaves the task undone with a probability of about 0.29, and a very large group with a probability approaching 0.4.
In market entry experiments the number of entrants is close to the capacity from the early rounds onward, without communication. Kahneman (1988), who ran an early version, remarked on how much the result looked like magic to a psychologist, and Erev and Rapoport (1998) took up the word in their title. Sundali, Rapoport and Seale (1995) found that coordination emerged quickly at the level of the group while individuals differed considerably in how they decided. When the payoff after entry depends on relative skill, Camerer and Lovallo (1999) found excess entry among participants who had chosen to take part knowing this, a result they attribute to overconfidence.
For the volunteer's dilemma, the best-known evidence comes from the study of bystanders by Darley and Latané (1968). Of the participants who believed that they alone had heard an emergency, 85 percent reported it promptly, against 62 percent of those who believed that one other person had heard it and 31 percent of those who believed that four others had. In the experiment of Goeree, Holt and Smith (2017), individual rates of volunteering fell with group size as the theory predicts. Outcomes in which nobody volunteered nevertheless became rarer in larger groups, which the theory does not predict. Franzen (1995) examines group size in a related one-shot setting.