Essays on contagion of risk.

摘要

In this dissertation, I study asset pricing and contagion. In my first and third essays, I study contagion of interbank and sovereign credit risk. In my second essay, I study how contagious sentiment is associated with asset prices.;In my first essay, I study interbank credit risk contagion by estimating the joint and conditional counterparty risk of financial institutions. Using a unique feature of credit default swap (CDS) data, I present a new approach for valuing CDS spreads that enables identification of the joint and conditional default probabilities allowing for time-variant recovery rates. I use CDS spreads and option implied default probabilities to disentangle time-variant joint default probabilities and time-variant recovery rates. I show that when the market is in distress, counterparty risk estimate assuming a fixed recovery rate underestimates the true joint default probability. I measure interbank counterparty risk of CDS dealers from 2007 to 2010 and find that the fixed recovery rate model underestimates expected counterparty risk by approximately 21%. Finally, I show that the conditional default probability of a bank conditional on default of its counterparties is correlated with existing systemic risk measure.;My second essay examines how social network information associates with asset prices. Sentiment is contagious as it can spread through a social network, and the speed of contagion is dependent on the number of followers the sender has. After collecting more than 2 million Twitter tweets on S&P; 500 firms, we compare firm level sentiment with stock returns. We find that firm level sentiment from tweets is significantly related to the firm's stock returns, and that the number of followers moderates this relationship.;My third essay proposes a method to refine the joint default probability estimates of sovereigns, by estimating the covariance of marginal default probabilities and allowing for time-varying recovery rates. I show that a relationship exists between the joint default probability and the covariance of marginal default probabilities, and that estimating the covariance from covariance of Credit Default Swap (CDS) spreads assuming a fixed recovery rate is biased. I use autocovariance of CDS spreads to estimate covariance of marginal probabilities.