Colin Camerer, Caltech
Robin Hogarth, Univ Chicago

The effect of economic incentives in experiments: The labor theory of
cognition 


We review a couple dozen studies which have varied the amount of
financial incentive subjects have for superior performance (from none,
to some, to a lot) in a variety of economics experiments.  Tasks
include judgment, choice under risk, bargaining, games, coalition
formation, and markets.  Our general conclusions are that: (i) Paying
subjects some money reduces the frequency of outliers and thoughtless
responses.  For some tasks (e.g. speculative bubbles in markets, or
"weak-link" stag hunt coordination games) which are sensitive to
outliers, paying ssubjects is therefore important. (ii) When responses
have "social presentation" value-- e.g., subjects want to appear
altruistic, risk-taking, or charitable-- then paying subjects
something will reduce the frequency of the presentations they make
when unpaid.  (iii) In a wide variety of tasks, "floor" and "ceiling"
effects imply that whether subjects are paid, and how much, make
little difference for responses.  That is, some tasks are so easy that
self-selected volunteers will do them well for no pay (e.g. choice
under uncertainty); and others are so hard that no amount of pay in an
experimental timeframe will produce superior performance

Generally, we think progress can be made by sketching a "labor theory
of cognition" (Smith and Wlaker, 1993 Econ Inquiry) in which subject
thinking is treated as a laborious activity which responds to
incentive.  We add to the Smith-Walker framework the idea that
subjects need "cognitive capital" to perform well, and while added
incentive draws out more labor, it does not draw out more capital in
the short-run.  Hence the effect of incentive is limited (and can be
negative) by the amount of cognitive capital (which depends on subject
skill and education, the nature of the task, etc.)