Thesis (Ph. D.)--University of Chicago, Graduate School of Business, March 2001. / Includes bibliographical references. Also available on the Internet.

Modelling and forecasting the stock market remains a challenge because of the high volatilities in individual stock prices and the market itself. Hence, this topic has received much attention in the literature since forecast errors represent the systematic risk faced by investors. Therefore, the ability to reliably forecast the future values of the shares would provide essential help in reducing that risk to those investors. The main aim of this research is to develop and calibrate a framework that can be used to model the daily share prices of the companies from the banking sector and hence produce informative and reliable one step-ahead forecasts using an adaptive BPNN. To this end, a novel forecasting algorithm is proposed. This algorithm proposes six steps that, when followed, possibly will lead to obtaining superior forecasting models for the share prices from the banking sector. In addition, novel technical indicators, and further information reflecting market knowledge were developed in this research so as to improve the modelling and forecasting share prices for the banking sector, alongside a novel application of the correctly identified turning points which provided an accurate assessment of the performance of the forecasting models. Furthermore, a selection of a set of inputs that are salient to financial data was identified. The research was to inform and improve share price forecasts of the banking sector. The historic open share prices for HSBC, Lloyds TSB, RBS and Barclays were used as case studies and the results give evidence to conclude that useable forecasting models can be obtained by employing the developed framework to the share prices from the banking sector in terms of the correctly identified turning points and the direction of the shares which are achieved more than 70% of the time. The empirical results show that using the market knowledge as input generally improved the modelling and forecasting of the share prices from the banking sector.

This PhD dissertation research primarily aims to empirically investigate three major, yet distinct and stand alone, financial topics using data from Turkey. These are (i) analysis of effects of financial statements on stock returns and their conditional volatility using an event study methodology; (ii) analysis of relationship between credit rating and firm-specific characteristics in a cross-section framework; and (iii) analysis of credit ratings and stock market performance using both firm-specific characteristics and selected macroeconomic variables in a panel data framework. My contribution remains as a pioneering effort to analyse these issues at least in the context of Turkey. The first empirical work, based on event study methodology employing a modified asymmetric time-varying volatility model, concludes that there are significant return and volatility effects stemming from announcement of audited semi-annual financial statements. This conclusion is also valid for the pre-announcements and postannouncement dates. It should be noted that event study methodology is a test of the informational efficiency subject to the joint hypothesis problem. My results can be considered as evidence against the informational efficiency of the Turkish stock market. The second area of research is on the relationships between corporate credit ratings and firm-specific characteristics. This investigation is based on a cross-sectional regression analysis. It concluded that size, sector, systemic risk and volatility are important factors on credit ratings and stock returns. The third empirical work employs panel data methodology to analyse the relationship between corporate credit rating and stock market performance indicators. It is found that both firm-specific characteristics and macroeconomic variables have significant effects on stock returns.

Market timing on the Johannesburg Stock Exchange has been the subject of numerous researches in South Africa, using different market conditions. This research aims to explore the effectiveness of a strategy of market-timing against a buy and hold strategy, at a reasonable level of market timing ability Seven equally weighted share portfolios were created from 56 different companies listed on the Johannesburg Stock Exchange. These companies’ share movements have a positive or negative correlation to the exchange rate movements. From the seven portfolios, three equally weighted group portfolios were created; one portfolio was containing shares with a positive exposure to exchange rate changes and a second portfolio was containing shares with a negative exposure. The research is exploratory in nature and observed that, the levels of predictive ability reported for Proposition 2 is generally higher than the findings presented in Proposition 1. This is ascribed to the higher index used to determine the various levels of return. The levels of predictive ability are consistent with results of other market timing studies, using different market conditions. / Dissertation (MBA)--University of Pretoria, 2010. / Gordon Institute of Business Science (GIBS) / unrestricted

The backdrop to this study is derived from an international paper which analysed the characteristics and performance of concept stocks across different North American stock exchanges. The empirical findings indicated that concept stocks underperform the market as well as control stocks. Concept stocks were also found to exhibit abnormally high levels of research and development spending. This study replicates the North American research in the South African setting by analysing the characteristics and performance of concept stocks on the Johannesburg Securities Exchange. The method used analyses a number of the characteristics of the stocks based on financial indices. Their performance is measured using the buy-and-hold abnormal returns method. The results indicated that the concept stocks outperformed their control stocks on certain accounting characteristics, but underperformed on others. The returns generated by the concept stocks significantly underperformed the control stocks, but only in Years 3, 4 and 5. The concept stocks are therefore only accurately valued over the long term. / Dissertation (MBA)--University of Pretoria, 2010. / Gordon Institute of Business Science (GIBS) / unrestricted

The object of this dissertation is to investigate some of the more recent concepts and methods of technical forecasting of market trends, business cycles and individual security cycles and to indicate the sources from which such data may be obtained. This investigation indicates that technical forecasting is a useful technique for the security analyst which may be used to advantage. No attempt has been made to prove that technical methods of forecasting are infallible or that these methods may be considered superior to individual security analysis but rather that they should be used by security analysts in conjunction with other techniques. The problem is divided into three phases. The first phase, presented in Chapters I to IX, discusses technical market trend forecasting and the results are compared with actual market performances for 1959. The second phase, Chapters X and XI, considers the shortcomings of the earnings and valuation approaches of security analysis and contains a discussion of business cycle forecasting. The third phase, Chapters XII and XIII, illustrates a modified technical method of evaluating common stocks. / Business, Sauder School of / Graduate

The main purpose of this study was to test the worth of using methods of timing investment decisions in the stock market. The writer investigated the use of economic and technical indicators in forecasting the most advantageous times for investing and disinvesting in the stock market. Recognition was given to the importance of fundamental analysis in the choice of stocks, and the "balance of the appraisal was devoted to the timing decision, or, "when to "buy?". A null hypothesis was formed to provide the basis for a test on the timing decision. The hypothesis was tested by the use of a model, consisting of economic and technical indicators, and criteria that are developed for the performance of the model. The statistical method in this appraisal comprises of the formation of indexes for forecasting the investment decisions. Some of the leading economic indicators that were developed by the National Bureau of Economic Research, and the Index of Consumer Sentiment of the University of Michigan were formed into a diffusion index, which was tested for the purpose of assigning a weight to its performance. A group of eight currently used technical indicators were then individually tested for their effectiveness in a market forecast. Of these, six were found suitable, and were then incorporated into a composite index. The composite index was then tested for the purpose of assigning a weight to its performance. On the basis of their weighting, the diffusion index and the composite index were then incorporated into the model. By means of tests, suitable criteria were developed for the performance of the model. The model was then used to test the null hypothesis that was formed for this appraisal. The results indicated that there was a significant difference between a buy and hold investment decision, and one that was timed to the indications of the chosen economic and technical indicators. / Business, Sauder School of / Finance, Division of / Graduate

The purpose of the research undertaken in this thesis is twofold: a) to test the relationship between a security analyst's perception of risk based upon financial statement data and overall market return and b) to determine the relationship between the practitioners concept of risk and risk as outlined in the literature. The main data sources for the thesis were the Financial Post computer tape from which "accounting" measures of risk were derived and stock exchange price quotations from which "economic" or "traditional" risk measures were determined. "Accounting" measures of risk considered included the coefficient of variation, standard deviation and mean-absolute deviation of the earnings stream variables, net operating income, net income and net income plus depreciation. The "traditional" or "economic" measures computed were the standard deviation of return and the beta coefficient or volatility index. Arguments were then presented for the relevance of each measure in describing stock market risk. To determine any relationship among various risk measures, a correlation and sectoral analysis was undertaken. The correlation analysis indicated a significant relationship existed among certain "accounting" and "economic" risk measures and in general, this relationship was supported by the sectoral analyses. To indicate the relationship among the risk measures and overall return, a graphical analysis was undertaken. Mixed results were obtained in this analysis, with certain measures of risk displaying a more significant risk/return relationship than did others. Thus, it appears that there does exist some degree of association between "accounting" and "traditional" measures of risk as indicated by the analyses undertaken in this thesis. What the literature is measuring as risk could possibly then be a reflection of what the security analyst views as stock market risk. However, there may be other factors which play an important role in the practitioners formation of risk estimates, factors which are, as of yet, non-quantifiable. / Business, Sauder School of / Graduate

This study attempts to determine the usefulness of Sharpe's Beta Coefficient in explaining the relationship between selected indices and individual securities. Basically, this involved doing a correlation-regression analysis on the returns of randomly selected securities against those of specific market indices. The returns for both variables were calculated traditionally, that is, by taking the price differential between the closing price at the end of the previous and present quarter and adding the quarterly dividend (where applicable) and dividing the total by the initial price. This was performed for six test periods. Generally, the tests yielded negative results. The amount of explained variation in individual security returns by the Beta Coefficient is negligible. This study concludes by providing some explanations and suggesting modifications. / Business, Sauder School of / Graduate

The purpose of this thesis was to investigate the underlying process generating stock prices in the Canadian stock market. The hypothesis that daily price changes of stocks listed on the Toronto Stock Exchange are independent could not be rejected. The distribution of daily stock price changes were found to be significantly non-normal. These results led to the conclusion that the price-generating process in the Canadian stock market can be represented, in the short-run, by a random walk model in which price changes are independently drawn from a non-normal distribution which is possibly a stable Paretian. / Business, Sauder School of / Graduate