In-Group Favoritism: Why a Coin Flip Is Enough to Pick a Side

In-group favoritism is what people do with a boundary once they are on one side of it. The boundary can be deep, a nationality or a faith held from childhood, or it can be a coin flip performed a minute ago, and the behavior that follows is the same in kind: members of the in-group are rated more warmly, trusted more readily and given more of whatever is being divided. Henri Tajfel's minimal group experiments of 1971 established the second case. Before them, discrimination was explained by conflict over resources or a history of hostility; after them, categorization alone was enough.

Tajfel and John Turner's social identity theory, set out at the end of the 1970s, gave the mechanism. Part of a person's sense of who they are comes from the groups they belong to, and people want that part to compare well. A group with no history offers nothing to compare except the next division of points, so the division becomes the comparison. That is why the second 1971 experiment found that the size of the gap pulled harder than the size of the gain: the gap is what the comparison runs on.

Marilynn Brewer argued in 1999 that in-group favoritism is mostly in-group love rather than out-group hate: people extend trust, cooperation and generosity preferentially to their own, and the out-group suffers by omission more than by attack. In minimal groups there is nothing to hate, and the favoritism appears anyway; hostility needs more, usually threat or competition.

The effect is robust and moderate. Daniel Balliet, Junhui Wu and Carsten De Dreu reviewed over 200 studies of cooperation in 2014 and found that people cooperate more with in-group than with out-group members across experimental games, by a moderate margin; the effect was the same size for real groups as for groups created in the laboratory, and larger when people knew that the partner knew their group. The bias is strong enough to matter in hiring, grading, lending and refereeing, and weak enough that a person who knows about it can usually override it on a given decision.

Tajfel, Michael Billig, Robert Bundy and Claude Flament published the minimal group experiments in the first volume of the European Journal of Social Psychology in 1971; Tajfel had described the same experiments for a general audience in Scientific American the year before. The participants were boys aged 14 and 15 from one Bristol school, 64 in the first experiment and 48 in the second.

The first experiment began with a task of estimating the number of dots flashed on a screen. The boys were then told they had been grouped as overestimators or underestimators, or as more and less accurate; in fact the grouping was random. Each boy then went alone to a cubicle with a booklet of matrices: rows of 14 boxes, each holding a number for one person and a number for another, the people identified only by code number and group, the points worth a tenth of a penny each. Some matrices divided points between two members of the boy's own group, some between two members of the other group, and some between one of each. In the intergroup choices a large majority of the boys gave more to their own group's member; the mean choices were significantly above rank 7.5, the point of maximum fairness on a 14-box matrix, while choices between two members of the same group clustered around that point.

The second experiment replaced the dots with paintings. The boys viewed slides of works by Klee and Kandinsky, shown without signatures, expressed preferences, and were told they were in the Klee group or the Kandinsky group; again the assignment was random. New matrices were built so that three motives pulled in different directions: maximum joint profit, the most points for both people together; maximum in-group profit, the most for the in-group member regardless of the other; and maximum difference, the largest gap in the in-group member's favor. Maximum joint profit had little effect on the choices. Maximum in-group profit mattered. The clearest effect came from maximum difference: the boys sacrificed points for their own group's member in order to widen the gap, and their choices between two in-group members were closer to joint profit than their choices between two out-group members, even though generosity to two outsiders cost the in-group nothing.

Tajfel called the result gratuitous discrimination: nothing in the setup gave the boys a reason to favor one group, and the categories were, in his word, flimsy. The effect has been reproduced many times with adults and other arbitrary labels.

The pattern appears wherever a boundary is visible and something is being divided or judged. The rows are illustrations, not findings.

In-group favoritism is fast and does not feel like favoritism from the inside, so the counters are procedures that remove the group label from the decision or widen the group.

• Blind the decision where you can. Names, schools and team colors off the page until the judgment is made. Blind review and blind grading exist because this bias is reliable.
• Fix the rule before you know the groups. An allocation formula, a scoring rubric or a selection criterion written in advance cannot be bent toward the in-group after the fact.
• Ask the swap question. Would you make the same call if the two people exchanged groups? If the answer changes, the group is doing the deciding.
• Notice the matrix in front of you. Many real decisions are the first kind of choice in the demo: a step that gives everyone more is available, and the only reason not to take it is that the other group would gain faster.
• Widen the circle on purpose. A shared goal that puts both groups in one larger category reduces the pull; the boundary that matters is the one that is salient at the moment of the decision.

In-group favoritism is a bias in how groups are treated, and it is worth separating from biases in what groups are believed to be like.

• It is not out-group homogeneity bias, the belief that members of another group are more alike than members of your own. That is a judgment about variety; favoritism is a judgment about worth and a decision about rewards, and Park and Rothbart found the homogeneity effect without any sign that it depended on preferring one's own group.
• It is not stereotyping, which is the content of beliefs about a group. Minimal groups have no content to stereotype, and the favoritism appears anyway; stereotypes are one route to it in groups with a history, not its source.
• It is not prejudice, in the psychological sense of a negative attitude toward a group and its members. Brewer's distinction applies: minimal-group favoritism is mostly in-group love, and out-group hostility requires conditions, such as threat or competition, that the matrices do not supply.