Salience in strategic choices

• Main Author: Arjona, David Rojo
• Published: University of East Anglia 2011
• Subjects: 330.01
• Online Access: http://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.554300

Schelling proposes salience as a solution to the problem of multiplicity of equilibria and focal point as equilibrium concept. Salience in strategic situations refers to those choices with psychological appeal. Given the experimental evidence, theories ignoring salience might suffer from a bias of omitted variables. This requires correction, because salience relates to one of the central topics in economics - bargaining. This thesis examines and tests two candidate theories to explain behaviour in games with salience: team-reasoning and, especially, level-k. In particular, this thesis offers two methods to produce experimental, falsifiable tests. The first method pursues the independent identification of the concept behind salience - what is a focal point? (chapter 2) - which helps to address a possible puzzle in the empirical literature: apparent differences between coordination games. In particular, two games are studied: open sets (Mehta et al. 1994) - e.g. "Choose an animal" - and closed sets (Bardsley et al. 2010) - e.g. "Choose one of the following animals: dog, cat, lion, tiger, monkey". This identification allows the best-rule hypothesis of team-reasoning to be tested (chapter 3). The second method uses the fact that level 0 is the same across different games with the same non- strategic features and, therefore, identifiable through pure coordination games. Clear predictions about what higher levels do can then be drawn, and tested, in the rest of the games (chapter 4). The main results are as follows. First, focal points are related with prototypicality and typicality. There are differences between sets; and coordination is higher in open sets because concepts explaining coordination are more correlated. Second, although level-k finds support in both open and closed sets, team-reasoning can now be discarded in closed sets. And, more generally, doubts about level-k are cast, because unexplained and systematic deviations are found.