Essays in the economics of networks: Network analysis in experimental games, international migration, and finance

Abstract

It is critical to understand the interaction among agents in the economic system. The decision of an agent depends on the interaction structure. Also, the interaction structure is formed by agents’ decisions in the economic system. Network science provides helpful frameworks for understanding the interaction among economic agents in economic systems.

This dissertation applies network science to three areas: cooperation in the finitely repeated prisoner’s dilemma game, international migration networks, and systemic risk in finance. To understand the cooperation in the finitely repeated prisoner’s dilemma game, I study the effect of networks on the cooperative actions of players using three approaches: the game-theoretical model, the belief-based learning model, and the experimental approach. I find the subgame perfect pairwise-Nash equilibria in which all players defect and are completely connected or isolated in the game-theoretical model. However, I find the reputation-building process to be selected in network formation and cooperation by forming links among cooperators who exclude defectors in the network in the belief-based learning model. In the experimental approach, I find that strong connections among cooperators who exclude defectors promote cooperation, and it supports the results in the belief-based learning model.

I apply network analysis to understand international migration among countries and the difference between migration and tourism. Using various data sets, I analyze the relationship between international migration or tourism network formation and many factors to drive migration or traveling decisions. The explanatory variables are classified into non-network structural variables and network structural variables. Social, economic, language, distance, safety, and climate change factors are non-network structural variables. The community structure of a network and three centralities, such as the in-degree, betweenness, and closeness centralities, are considered network structural variables. International migration network formation shows the characteristics of individuals' decisions as an investment: the desire to live in a more stable and safer society. However, international tourism network formation shows consumption characteristics: the desire to enjoy traveling and leisure. International migration and tourism network formation are associated with network structural variables, showing network effects. International migration due to climate change and the negative effect of climate change on tourism are not significantly observed. However, international migration and tourism to the countries with more climate-related disaster frequencies are observed with high statistical significance.

I introduce a novel systemic risk measure defined by the combined effects of direct and indirect connections among stocks in the US stock network using the degree centrality and community’s influence strength (CIS). I construct stock networks using minimum spanning trees (MSTs) based on the cross-correlation matrix of stock returns. The risk-adjusted volatility, not explained by the six-factor model, including the Fama-French five factors and momentum factor, is explained by the degree centrality and CIS with high statistical significance. I find increases in systemic risk during the recession after the dot-com bubble and the subprime mortgage crisis.