Forecasting Corporate Bankruptcy: Applying Feature Selection Techniques to the Pre- and Post-Global Financial Crisis Environments
By Parker Levi
I investigate the use of feature selection techniques to forecast corporate bankruptcy in the years before, during and after the global financial crisis. Feature selection is the process of selecting a subset of relevant features for use in model construction. While other empirical bankruptcy studies apply similar techniques, I focus specifically on the effect of the 2007-2009 global financial crisis. I conclude that the set of bankruptcy predictors shifts from accounting variables before the financial crisis to market variables during and after the financial crisis for one-year-ahead forecasts. These findings provide insight into the development of stricter lending standards in the financial markets that occurred as a result of the crisis. My analysis applies the Least Absolute Shrinkage and Selection Operator (LASSO) method as a variable selection technique and Principal Components Analysis (PCA) as a dimensionality reduction technique. In comparing each of these methods, I conclude that LASSO outperforms PCA in terms of prediction accuracy and offers more interpretable results.
View Data (Email for Access)
Advisors: Professor Andrew Patton, Professor Michelle Connolly | JEL Codes: G1, G01, G33
By Christopher G. MacGibbon
This thesis develops a new Multi-Horizon Moment Conditions test for evaluating multi-horizon forecast optimality. The test is based on the variances, covariances and autocovariances of optimal forecast errors that should have a non-zero relationship for multi-horizon forecasts. A simulation study is conducted to determine the test’s size and power properties. Also, the effects of combining the Multi-Horizon Moment Conditions test and the well-known Mincer-Zarnowitz and zero autocorrelation tests into one forecast optimality test are examined. Lastly, an empirical study evaluating forecast optimality for four multi-horizon forecasts made by the Survey of Professional Forecasters is included.
Advisors: Andrew Patton, Grace Kim and Kent Kimbrough | JEL Codes: G1, G17, G00
By Daniel Roeder
Portfolio Optimization is a common financial econometric application that draws on various types of statistical methods. The goal of portfolio optimization is to determine the ideal allocation of assets to a given set of possible investments. Many optimization models use classical statistical methods, which do not fully account for estimation risk in historical returns or the stochastic nature of future returns. By using a fully Bayesian analysis, however, this analysis is able to account for these aspects and also incorporate a complete information set as a basis for the investment decision. The information set is made up of the market equilibrium, an investor/expert’s personal views, and the historical data on the assets in question. All of these inputs are quantified and Bayesian methods are used to combine them into a succinct portfolio optimization model. For the empirical analysis, the model is tested using monthly return data on stock indices from Australia, Canada, France, Germany, Japan, the U.K.
and the U.S.
Advisor: Andrew Patton | JEL Codes: C1, C11, C58, G11 | Tagged: Bayesian Analysis Global Markets Mean-Variance Portfolio Optimization
By Yuci Zhang
In nance, the beta of an investment is a measure of the risk arising from exposure to general market movements as opposed to idiosyncratic factors. Therefore, reliable estimates of stock portfolio betas are essential for many areas in modern nance, including asset pricing, performance evaluation, and risk management. In this paper, we investigate Static and Dynamic Conditional Correlation (DCC) models for estimating betas by testing them in two asset pricing context, the Capital Asset Pricing Model (CAPM) and Fama-French Three Factor Model. Model precision is evaluated by utilizing the betas to predict out-of-sample portfolio returns within the aforementioned asset-pricing framework. Our findings indicate that DCC-GARCH does consistently have an advantage over the Static model, although with a few exceptions in certain scenarios.
Advisor: Andrew Patton, Michelle Connolly | JEL Codes: C32, C51, G1, G12, G17 | Tagged: Beta, Asset Pricing, Dynamic Correlation, Equity, U.S. Markets