Quattrone and Tversky: Causal versus diagnostic contingencies

Quattrone and Tversky. 1984. Causal versus diagnostic contingencies: On self-deception and on the voter's illusion. Journal of Personality and Social Psychology 46 (2): 237-248.

In Brief

People sometimes act as though acts that are merely 'correlated' with a particular outcome actually 'cause' that outcome. This may explain why people bother to vote: They know that their decision to turn out is correlated with the decision by people like them, so they turn out thinking that this will cause people like them to turn out as well.

The authors report two experiments to support this claim. The first, a pseudo-medical experiment, establishes the logic and serves as an analogy; the second demonstrates that this logic might apply to voting.

See also Quattrone and Tversky (1988).

Experiment 1: Cold tolerance and longevity

Volunteers are led to believe that they are participating in a study of heart type. People with a Type 1 heart are likely to have more illnesses and die younger than people with a Type 2 heart. Subjects do a pain tolerance test (submerging their arm in ice cold water for as long as they can), then exercise vigorously on an exercise bike. Half are told that people with a Type 2 heart will have more cold tolerance after exercising; half are told the opposite. The pain/cold tolerance test is repeated, and almost all participants behave as though they have a Type 2 heart (either lengthening or shortening the amount of time they are willing to keep their arm in the water).

Subjects confused a 'diagnostic' effect with a 'causal' one. Holding their arm for more (or less) time in the water is merely diagnostic of whether you have a healthy heart; it does not cause a change in your heart's type. Yet people (perhaps subconsciously) behave as if they can actually change their heart's type.

Experiment 2: The voter's illusion

A similar logic works in voting. If I am part of a class of voters (e.g. young, married, white, student), then my voting behavior probably correlates with the voting behavior of others like me--as do my political views. Thus, my decision to vote is 'diagnostic' of whether people like me will vote. It is not 'causal,' only 'diagnostic' (i.e. correlated). Yet I will behave as my decision to vote does have a causal effect on voting--I will turn out, expecting my decision to increase the turnout rates of my class. This is the "voter's illusion."

The authors use an experimental (hypothetical) survey to test this. Respondents are told to imagine that they belong to Party A in a fictional two-party state. Each party has 4 million supporters, and there are 4 million independents. One group of respondents is told that political analysts all agree that the independents will decide the election (though it's not clear in which direction); all the partisans will turn out at equal rates, so the decision will be swung by the center. Another group is told that the independents are evenly divided, so the election will be decided by whichever party gets more supporters to the polls. Respondents are then asked whether they would vote, and whether they think their decision to vote would affect other Party A supporters. Sure enough, those told that partisan mobilization will decide the election are more likely to want to vote and to think that their vote will have a causal effect on others.

Comments and Criticism

The pseudo-medical study presents the logic far more persuasively than the voting survey does. The authors need to find a better test of the "voter's illusion" than a hypothetical survey. I have no reason to believe that people will behave the same in real life as they claim they would on a hypothetical questionnaire.