when both groups do less than their best. What might otherwise be a zero-sum
game can thus become a negative-sum game. (Sowell (2004), p. 14)
This line of critique is also addressed in Fryer and Loury (2005a), where it is coined
‘Myth No. 3’ because “confident a priori assertions about how affirmative action
affects incentives are unfounded. Indeed, economic theory provides little guidance”
(ibid., p. ). The simple contest game in the style of Tullock (1980), introduced in the
next section, is an attempt to fill this gap in theoretical analysis by addressing the
question whether the criticized trade-off does exist in this kind of stylized model. An
affirmative answer to this question would then imply that optimizing players reduce
their respective effort levels if they face affirmative action policies which creates the
mentioned trade-off.
The implementation of affirmative action is modeled as a biased contest rule3 where
weak contestants are favored because ethical perception interprets their weakness as
being the consequence of past discrimination. The alternative perception, i. e. holding
the contestants ethically responsible for their heterogeneity, requires instead the imple-
mentation of an unbiased contest rule. Both policies are defined formally as restrictions
on the contest rule which imply different incentives for the individuals depending on
the implemented policy option. The key question is therefore how individuals react to
the distortion of incentives that is induced by the two policies.
There exists a limited number of articles with a similar focus. Fu (2006) models
college admission as a two-player all-pay auction under complete information and
shows that favoring the discriminated player to some extent induces the maximal
expected academic effort (interpreted as the expected test score) by both candidates.
A similar conclusion is derived in Schotter and Weigelt (1992) that analyze, also
experimentally, a two-player tournament with unobservable effort. However, none of
the models mentioned so far specifies the normative objective of affirmative action, i.e.
in these papers affirmative action is considered simply as a deviation from an unbiased
‘equal treatment’-policy. This is a crucial difference to the contest model presented
below because here the normative objective of affirmative action is explicitly defined
and integrated into the model.4 Kranich (1994) formalizes a similar idea for a two-
3The underlying game theoretic model is an asymmetric contest game with n heterogeneous players.
Asymmetric contest games are applied in different frameworks, for example, to analyze legal
presumption in trials; see Bernardo, Talley, and Welch (2000), with the interpretation of prior
probabilities; see Corchon (2000), or in a two-stage rent-seeking contest; see Leininger (1993).
4In Fryer and Loury (2005b) a model with incomplete information is introduced where a continuum