Essays on Forecasting in Finance
Type of DegreePhD Dissertation
MetadataShow full item record
This dissertation is composed of three essays related to forecasting in various financial markets. The first essay reviews over 60 studies published from 2008 to 2021 that investigate the causes of the mortgage crisis leading to the Great Recession. Because the mortgage crisis led to significant changes in finance, economics, law, education, and other areas, the causes and potential solutions to it have been studied extensively. Other literature reviews since 2008 that focus on mortgage crisis research have highlighted subprime mortgages defaults and determinants of mortgage defaults. This literature review contributes to the mortgage crisis research by reviewing several other areas related to the causes and resolution of the mortgage crisis, including foreclosure laws, securitization, and strategic default. This review is particularly timely, considering the amount of upheaval in real estate and mortgage markets caused by coronavirus disease 2019 (COVID-19). Some economists are concerned that COVID-19 could trigger another mortgage crisis with significant economic spillover. Although such a crisis is difficult to forecast, this literature review provides a much-needed compilation of work spanning multiple fields that could help avert or ameliorate another crisis. Despite the various methodological approaches used and differing data sources, many studies come to similar conclusions on the effects of laws, securitization, borrower characteristics, and neighborhood characteristics on default and foreclosure. Many of the studies on foreclosure laws find that laws designed to help prevent foreclosure (recourse versus non-recourse and judicial versus non-judicial) only extended the foreclosure timeline and did not actually prevent foreclosure. Furthermore, most of the research related to securitization of loans found that securitized loans had a higher rate of foreclosure compared to portfolio loans. Within the arena of strategic default, studies find that emotions and morality play a significant role, but education, information, and lender relationship are also important factors. The second essay examines the forecasting ability of banks in response to changes in local housing prices. Using Freddie Mac housing prices, banking data from the Federal Deposit Insurance Corporation, and mortgage rates from RateWatch, I measure the impact of changes in housing prices on changes in mortgage rates. RateWatch provides branch level rates, which in turn allows a much finer degree of granularity regarding local mortgage rates than previous studies. Thus, by using the RateWatch data, this study contributes to both the real estate literature and the banking literature by tying mortgage rate-setting decisions to local house price movements, increasing our understanding of rate-setting decisions and the effects of centralized rate-setting. The major finding of this essay is that during the incredible housing price expansion from the late-1990s to the mid-2000s, local lenders adjusted mortgage rates faster than non-local lenders in response to changes in house prices. However, in the housing price crash following the runup, there was no difference in how local and non-local banks adjusted rates. The final essay develops a technical analysis-based model for forecasting in the stock market. Technical analysis is widely used in the practice of managing investment portfolios, but approaches used to study technical analysis in most academic research do not mimic actual usage within the industry. This study fills a significant gap by replicating more accurately the application of technical analysis in portfolio management through the use of a unique probit model. The probit model uses variables derived from five common technical analysis indicators to forecast one month in advance. Using S&P values from 1871 to 2019, probit market-timing model outperforms the standard buy-and-hold portfolio in out-of-sample tests. Out-performance is particularly notable in periods of high volatility, while during calm, steady uptrends, the model tends to match market performance.