Repository logo

Essays on Social Networks



Change log


To, Yu Yang Tony 


This thesis consists of three essays on the economics of social networks. It broadly deals with understanding the value of social connections on favour exchange and information exchange. Social networks facilitate trust, learning, and communication, all crucial in the modern online environment. Examining the effects of network structure provides new tools and insights on decision-making and behaviour. Chapter 1 develops a model of repeated favour exchange on social networks where individuals choose between allocating the opportunity to the expert (market action) or a friend (favouritism action). Assuming favouring a friend reduces one’s payoff, favouritism cannot be sustained in a stage game. However, by introducing a grim-trigger strategy where a selective group of individuals favour each other, favouritism can be sustained in an infinitely repeated game. In particular, the maximum clique of the network defines favouritism behaviour that is coalition-proof where no group of individuals have incentives to deviate collectively. While aggregate surplus increases with network connectivity, it decreases with the number of favouritism-practising agents. Additionally, favouritism exacerbates payoff inequality that arises from degree inequality: Favouritism players cooperate to extract a large portion of the aggregate surplus at the expense of market players, creating a negative externality on the economy. Chapter 2 conducts an experiment to study the impact of network structure on opinion formation. At the start, subjects observe a private signal and then make a guess. In subsequent periods, subjects observe their neighbours’ guesses before guessing again. Inspired by empirical research, we consider three canonical networks: Erdös-Rényi, Stochastic Block and Royal Family. We find that a society with ‘influencers’ is more likely to arrive at an incorrect consensus and that one with ‘network homophily’ is more likely to persist with diverse beliefs. These aggregate patterns are consistent with individuals following a DeGroot updating rule. Chapter 3 studies incentives for verifying information in social networks. Agents derive value from sharing correct information and suffer a reputational loss from sharing false information. So agents can undertake costly verification prior to sharing information. We show incentives for verification are increasing in degree. This implies that information quality is increasing in average degree and is higher in more egalitarian networks. We then introduce an external agent whose goal is to maximise views through a choice of news source accuracy. We find that denser networks lead to higher accuracy when information accuracy is either expensive or cheap, and sparse networks lead to more accurate information otherwise.





Goyal, Sanjeev


favouritism, information, learning, networks


Doctor of Philosophy (PhD)

Awarding Institution

University of Cambridge