Search

To link to full-text access for this article, visit this link: http://dx.doi.org/10.1016/j.jedc.2010.06.009 Byline: Abhijit Sengupta, Danica VukadinoviA Greetham Abstract: Brand competition is modelled using an agent based approach in order to examine the long run dynamics of market structure and brand characteristics. A repeated game is designed where myopic firms choose strategies based on beliefs about their rivals and consumers. Consumers are heterogeneous and can observe neighbour behaviour through social networks. Although firms do not observe them, the social networks have a significant impact on the emerging market structure. Presence of networks tends to polarize market share and leads to higher volatility in brands. Yet convergence in brand characteristics usually happens whenever the market reaches a steady state. Scale-free networks accentuate the polarization and volatility more than small world or random networks. Unilateral innovations are less frequent under social networks. Author Affiliation: Unilever Research and Development, Colworth House, Sharnbrook MK44 1LQ, UK Article History: Received 7 April 2009; Accepted 1 June 2010

We propose a solution concept--which we call viable proposals--for games in coalitional form that endogenously determines a coalition structure together with an associated set of payoff configurations as parts of the solution. We show that there is at least one viable proposal for every game; further, whenever a game has a nonempty core, the set of payoff configurations identified by the solution coincides with the core of the game. Finally, given any arbitrary coalition structure and payoff configuration, a viable proposal can be reached in finitely many steps of successive domination even when the core of the game is empty.

It is known that among the diagnosable systems modeled as in PMC model, there exist several subclasses of systems that can be diagnosed more efficiently than the general case. Some of these subclasses are D(n, t, X) systems, self-implicating systems, systems having connection assignment digraph of node connectivity t and so on. It has been shown in the literature that some of these classes are unrelated. An obvious question is: does there exist any class of systems that subsumes all these subclasses? In this paper, we have found an affirmative answer to this question. We present a generalization of the self-implicating structure and show that this generalized class subsumes all these subclasses. We find a characterization of this new class and find an efficient diagnosis algorithm.