Database Reference
In-Depth Information
We distinguish here two classes of controlled experiments, namely laboratory
experiments and natural experiments.
(Levitt & List, 2007) review five classes of games used in the economic literature
to measure social preferences through
laboratory experiments
, including fairness,
trust, and conditional reciprocity. The reviewed games include dictator and ultima-
tum games, public goods games, trust and gift exchange games. As an example,
(Fershtman & Gneezy, 2001) adopt trust games, dictator and ultimatum games to
test for ethnic discrimination. The trust game assumes a “player A”, who is given a
fixed amount of money and asked to transfer a certain amount to “player B”. The
transferred amount is triplicated. Then, “player B” can choose to transfer any part of
the received amount back to “player A”. Players A and B are randomly paired from
students of different ethnicity. The lower average amount of money transferred to
players of a specific ethnicity, compared to others ethnicities, is considered evi-
dence of discrimination. Recent controlled experiments can be found in the context
of sports card market (J. List, 2004), employment (Feltovich & Papageorgiou, 2004;
Falk et al., 2008) and wages differentials (Guth et al., 2010; Dickinson & Oaxaca,
2009), beauty and speech differences (Andreoni & Petrie, 2008; Rodin & Ozcan,
2011). Moreover, gender (Slonim & Guillen, 2010), racial (Castillo & Petrie, 2010)
and district-based (Falk & Zehnder, 2007) differences have been studied in the con-
text of in-group discrimination and favoritism.
Natural experiments
occur in real life (yet, controlled) situations. The experi-
menter only observes the behavior of participants, who typically are not aware of
the experiment. Television game shows are a typical example, where discriminatory
choices of participants can be studied in a controlled environment. Discrimination
analysis has been reported in (Antonovics et al., 2005, 2009; Bagues & Villadoniga,
2008; Levitt, 2004), with data gathered from the
We a k e s t L i n k
game show, in (Lee,
2009) with data from
American Idol
TV contest show, and in (J. A. List, 2006) with
data from
Friend or Foe?
. Sources of favoritism to attractive people by analysing
data from a TV game show based on the prisoner's dilemma are studied in (Belot
et al., 2008). In addition to the criticism of external validity, natural experiments
have also the problem that not all factors are under control, e.g., the selection of
participants to a TV game show.
6.5
Profiling Perspective
Profiles consists of patterns, rules, or any other form of knowledge that can be used
to screen people when searching for those with a certain behavior. They occur in
many context, from criminal investigation to marketing, from genetic screening to
web site personalization, from fraud prevention to location-based services. Profiling
is the process of extracting profiles, either by manually eliciting them from domain
experts or by automatically inferring them from historical data using increasingly
sophisticated machine learning and data mining techniques. The process of profiling
also concerns the application of profiles to screen individuals, e.g., as in the case of
credit risk scoring and in the identification of security risks - which are covered in
Search WWH ::
Custom Search