Abstract:

Stock valuation, the pricing of common stocks issued by corporations, is one of the key issues in finance. It plays a central role in the areas of both investment and corporate decisions. For investors, it is crucial for portfolio selection decisions. For corporations, it may affect many corporate decisions including security issuance and mergers and acquisitions.

Although there has been great interest in stocks from investors, corporate managers and finance researchers, there are not as many stock valuation models that are of practical use. This can be due to the fact that the uncertainties in stocks are more complex to model and thus to estimate compared to other financial products. This dissertation contributes to stock valuation literature by proposing new methodologies and applications to better estimate stock behavior in the future.

The stock valuation literature is divided into two main categories. Some researchers try to predict future returns by using relative value measures. Some others, on the other hand, use fundamental company data and analysts forecast to evaluate a stock. Models or methods from both categories are shown to have predictive power of future returns. However, the main challenge is to detect wrong signals. We introduce statistical learning theory to combine the information gathered from technical and fundamental analysis. Each view has weaknesses under different market conditions, and our methodology successfully filters the misleading signals to yield more robust estimates for future returns.

The valuation of mergers and acquisitions has long been an important issue in the corporate world. We have seemingly entered an age in which we regularly read news of bigger-than-ever-before mergers, so it is crucial to use sophisticated methods or models for merger or acquisition decisions. Investment bankers rarely apply modern financial engineering methods when considering a purchase or merger of two companies. Therefore, using the stock valuation techniques we develop to the corporate arena is a natural extension. We pose the problem of combining two publicly traded enterprises as a multistage stochastic program, which extends to the valuation of Bermudan options due to the similarity of the decision processes in mergers and options.