UC Santa Cruz

Virtual economies are growing as internet technology continues to advance. In Aggregate Dynamics in a Large Virtual Economy: Prices and Real Activity in Team Fortress 2, we analyze a large and complete set of transaction data from the Team Fortress 2 (TF2) virtual economy, which was designed to allow for decentralized barter as the sole exchange institution. A small subset of goods emerges endogenously to act as media of exchange. Taking one of these money goods as numeraire, we generate daily prices for thousands of goods. We then generate macroeconomic indicators, including nominal growth and inflation. We find evidence of a particular sort of nominal rigidity related to the circulation of multiple types of currency goods, and also find some localized asset price bubbles associated with announcements by the game designers.

Continuing work with this complete set of transactions in Emergence of Networks and Market Institutions in a Large Virtual Economy, we construct trader and goods networks, and track them over time using metrics such as node strength, assortativity, betweenness and closeness. The trading platform was designed to make barter exchange as attractive as possible; money was not part of the design and all players were created equal. Yet, within weeks, several specific goods emerged as media of exchange, and various specialized traders appeared that facilitated exchange. Eventually trade was predominantly money-mediated and market-makers played a major role. Our results illustrate how network analysis can capture the spontaneous emergence of economic institutions by applying common conceptions of centrality to goods and traders.

The closing work is Minimum effort coordination in continuous time - An experimental analysis with changing payoff structures. The minimum effort game is among the most studied coordination games because there exists no social dilemma yet coordination failure has been commonly observed. We extend the minimum effort game to continuous to test whether that is sufficient to induce high levels of coordination. We find that continuous time is not always enough to induce efficient coordination in spite of decreased signaling costs and extensive information about the decisions of other players. We finally extend our model to examine the effects of ``gradualism" on coordination and find that they are still less efficient compared to treatments with a mild penalty parameter.