| This dissertation develops and tests empirically a model of asset returns that captures the dramatic fluctuations, time-varying volatility, and correlations across seemingly disparate assets that financial data often exhibit. The model treats individual securities as options on underlying assets. The key assumption is that investors receive information of varying quality regarding the actual values of those underlying assets. As a result, rational investors who seek to price a particular security must solve a signal-extraction problem to determine their best estimate of the underlying asset value.; When information quality is relatively poor so that the signal-to-noise ratio is low, the investors' estimates will diverge from the actual value of the underlying asset. Conversely, when the quality of information is relatively high, investors can determine the value of the underlying asset with much greater accuracy. Given the variation in information quality, the result of the signal-extraction process is that the investors' estimates of the underlying asset value can diverge from the actual underlying asset value. When the arrival of high-quality information leads the investors to revise quickly their estimates of the underlying asset value, prices can jump dramatically.; The model may shed light on unusual patterns in asset returns and the causes of sudden asset-price changes (e.g. the 1987 stock market crash) where analysts have had difficulty uncovering corresponding movements in fundamentals. Simulations demonstrate that the model produces simulated data that are completely consistent with results. Further, the empirical tests presented in the dissertation demonstrate empirical that the predictions of the model are consistent with financial market data. |
