Thesis advisor: Zhijie Xiao / This dissertation consists of three independent studies in Financial Economics. The first chapter focuses on the predictive power of the implied correlation index on the future S&P 500 Index returns. The second chapter investigates a nonlinear contemporary relationship between stock returns and oil price changes. The last chapter discusses the relationship between impact trading costs and a number of market factors that affect the costs. In the first chapter, I investigate the predictive power of the implied correlation index on the future S&P 500 Index returns. This new index was launched by Chicago Board Option Exchange following 2007-2008 financial crisis. As it is derived from the S&P 500 Index option price and the option prices of the largest 50 S&P Index stocks, it is widely regarded by market participants as a gauge of average expected future stock return correlation. Because of its role in measuring systematic risks, any changes in this index may provide useful information about the future market movements. Motivated by this index's forward-looking characteristic, I propose a linear regression where the future S&P 500 Index multi-period return is regressed on a number of controls such as the current period changes of S&P index and the implied correlation index etc. I use weekly data and three different sample splits for in-sample estimation and out-of-sample performance evaluation. I find that the implied correlation index is informative for the period 2007-2009 in predicting the S&P 500 Index returns of 28 to 39 weeks. My model consistently outperforms the random walk model using the Superior Predictive Ability test. This implied correlation index is also useful in predicting the S&P index future multi-week returns for the period 2009-2011 and a longer time span from 2007-2011. I also do a test for the Efficient Market Hypothesis by incorporating the implied volatility index in the regression. There is no evidence supporting the view that the market is efficient for those time periods. In the second chapter, I estimate a nonlinear contemporary relationship between stock returns and oil price shocks. Previous studies on this issue suffer a number of limitations. For example, they do not control the factors potentially driving the economy and the oil market simultaneously. Although, Kilian and Park (2009) does a good job in identifying the relationship by distinguishing different oil market shocks, they use a linear regression framework and do not address the contemporary relationship. Considering the different impacts of different-size oil shocks on stock returns, I propose a two-step estimation procedure for identifying their relationship. In the first step, I follow Kilian and Park's methodology, i.e. a structural vector autoregression, to estimate the demand-specific oil shocks. During the second step, I use a nonparametric quantile regression to estimate the relationship between stock returns and the estimated exogenous oil price shocks. This way, I can control for the factors that simultaneously drive the economy and the oil market and am able to identify a nonlinear relationship of stock returns with oil shocks at the same time. The result shows that different-size oil price changes do have quite different impacts on stock returns. I also find an asymmetric effect of large oil shocks on large stock returns. Specifically, the positive impact of the large negative oil shocks on stock returns is much bigger than the negative impact of the large positive oil shocks on stock returns. I carry out a robust check by running regressions for a number of different model setups and the result persists. I also compare my model with Kilian and Park' SVAR model and it turns out that my model is a big improvement on their model in explaining the stock return variations. The third and last chapter focuses on impact trading cost and its relationship with several market factors. In this chapter, I focus on one of financial market microstructure issues, the immediate impact trading cost for major NASDAQ stocks. The immediate impact cost is the extra cost that market traders pay when they execute a large volume transaction without delay during the time when the market is less liquid. Because the market depth is defined to be the market's ability to sustain relatively large market orders without impacting the price of the security, this cost is closely linked to the trading volume. When trading volume becomes large, market liquidity gets worse and therefore the relationship between immediate impact cost and trading volume is virtually nonlinear. People trading in the market are interested in this relationship because they hope to figure out the best strategies in the situation where they want to execute a large volume order when the market is not deep. Another measure of market depth or liquidity people often use is market spread. Because it is the compensation for market makers' willingness to hold an imbalanced portfolio when the market is not liquid, it is regarded as another important factor linked to the impact cost. In this chapter, I use a nonparametric model to estimate the unknown relationship between immediate impact cost and market factors such as trading volume, market spread etc. for the major NASDAQ stocks. The result shows that, for many stock transactions, there is a certain volume threshold of trading volume beyond which impact costs increase dramatically. I find that for 99% of trading, immediate execution is optimal. I also identify a negative relationship between the occurrence likelihood of a large trading cost and the stock market cap. / Thesis (PhD) — Boston College, 2012. / Submitted to: Boston College. Graduate School of Arts and Sciences. / Discipline: Economics.

<p>The purpose of this study is to examine what affects the changes in credit spreads. A</p><p>regression model was performed where the explanatory variables were; volatility,</p><p>SP&500 index, interest-rate level the slope of yield curve and the dependent</p><p>variable was credit spread for each of CSUSDA, CSUSDBBB, and CSUSDB. We</p><p>found a positive correlation between these independent variables (Volatility, S&P</p><p>500index) and a negative correlation between interest-rate level and credit spreads.</p><p>These results were consistent with our hypothesis. However, the link between the</p><p>slope of yield curve and credit spreads was positive and that was inconsistent with</p><p>our hypothesis and some previous studies. The conclusion of this paper was a</p><p>change in credit spread is related to the variables that we used in our model. And</p><p>these variables explained about 50 per cent of this change.</p>

The purpose of this study is to examine what affects the changes in credit spreads. A regression model was performed where the explanatory variables were; volatility, SP&amp;500 index, interest-rate level the slope of yield curve and the dependent variable was credit spread for each of CSUSDA, CSUSDBBB, and CSUSDB. We found a positive correlation between these independent variables (Volatility, S&amp;P 500index) and a negative correlation between interest-rate level and credit spreads. These results were consistent with our hypothesis. However, the link between the slope of yield curve and credit spreads was positive and that was inconsistent with our hypothesis and some previous studies. The conclusion of this paper was a change in credit spread is related to the variables that we used in our model. And these variables explained about 50 per cent of this change.

Recently, there are more and more literatures discuss on the issues of investment strategies of leveraged ETFs. In our works, we concentrate our issues on optimal leverage of ETF of S&P 500 index. Based on ARMA-GARCH model¡¦s assumption, we find out that the forecasting optimal leverage can be shown in a formula which contains return and characteristic function. In this paper, we use MA(1)-GARCH(1,1) to forecast volatility based on 1008 rolling window to forecast one day ahead¡¦s volatility; and our estimation time is start from 1954 to March 2011. In this paper, we present four dynamic leverage models (Normal, Student T, VG, and Best model¡¦s leverage) to find out the payoffs under these models. In our model, the forecasting accuracy is just about 55% which is slightly higher than SPX raise probability. But during long-term compound effect, the dynamic leverage models can out-perform than constant leverage. There may exist some important factors in these results, one of them is the crash forecasting ability. During 1980 to 2011 SPX has 14 big crashes and these models can effectively avoid 10 big crashes. In short-term investment horizon none of these five models are always outperform than others but in long-term investment horizon the strategy of best model¡¦s leverage can always earn money when investment horizon is 2400 days.

While the much more use of leverage could be effective in generating alpha o investment, the Kelly strategy is an attractive approach to capital creation and growth. It is originated from the Kelly criterion dubbed ¡§ fortunes formula ¡§ which maximizes the long run growth rate of wealth. There is a tradeoff of rate of return versus risk/volatility as a asymptotic function solution of leverage or position size determined by the application of EGARCH model in the different residual assumptions given by the Normal, Generalized Hyperbolic, and the Generalized Error distributions. No matter there is any timing ability in any strategy, risk management is much more important especially with many repeated trading. We present the performance and risk control of the leveraged ETFs tracked the S&P 500 index in the past ten years using optimal leverage strategy derived by the full Kelly and fraction Kelly criterion.

本研究嘗試整合類神經網路與法則基礎（rule-based）系統技術，以建立S&P 500指數期貨的交易策略。本研究不同於先前研究之處有下列二方面：一、本研究採用法則基礎系統的方式提供神經網路的訓練範例；二、本研究以理解神經網路（Reasoning Neural Networks）取代後向傳導網路（Back propagation networks）以解決局部最小值與隱藏結點數未知的困境，而實證結果也顯示理解神經網路之表現優於後向傳導網路。首先，由期貨的日價格資料計算出十種技術分析指標值，用這些指標值來表示期貨市場內的各種可能狀況（case）。接著，我們提出FFM（Futures Forecast Model）與EFFM（Extended Futures Forecast Model）來處理市場的各種狀況，預測出隔日的期貨價格改變方向。以法則基礎方法所建立的FFM是用來處理明顯的狀況（obvious cases），並且提供類神經網路好的訓練範例。而EFFM包括四個理解神經網路系統與一個決策機置（voting mechanism），它被用來處理那些不明顯的狀況（non-obvious cases）。從實證模擬的結果顯示，在預測市場時FFM與EFFM有良好的合作 關係。因此，我們以FFM與EFFM為基礎建立一個整合的期貨交易系統（Integrated Futures Trading System，IFTS），並將它用於S&P 500 指數期貨市場作模擬交易，結果我們發現在1988到1993年的測試期間，IFTS 的投資報酬率高於買入持有投資策略。 / This research adopts a hybrid approach to implementing the trading strategies in the S&P 500 index futures market. The hybrid approach integrates both the rule-based systems technique and the neural networks technique. Our methodology is different from previous studies in two aspects. First, we employ Reasoning Neural Networks (RN) instead of back propagation networks to resolve the undesired predicaments of local minimum and the unknown of the number of hidden nodes. Second, the rule-based systems approach is applied to provide neural networks with good training examples. We, first, categorize the daily conditions of the futures market into a variety of cases through processing futures historical data. Then, the dual-forecast models, FFM (futures forecast model) and EFFM (extended futures forecast model), are proposed to predict the direction of daily price changes. The rule-based model, FFM, is designed to deal with the obvious cases and to provide the neural network-based model, EFFM, with good training examples. Meanwhile, EFFM, which consists of four RNs and a voting mechanism, is designed to handle the non-obvious cases. The simulation results show that the cooperation of FFM and EFFM does a good job in predicting the direction of daily price change of S&P 500 index futures. Based on FFM and EFFM, the integrated futures trading system (IFTS) is developed and employed to trade the S&P 500 index futures contracts. The results show that IFTS outperforms the passive buy-and-hold investment strategy over the six-year testing period from 1988 to 1993.