In recent years there has been increasing empirical evidence that appears to support the view that the Fama and French three-factor model is highly effective in capturing the systematic risks associated with equity rates of return. It has equally been recognised that the three-factor model does not have the theoretical sophistication of the Capital Asset Pricing Model (CAPM). This comparison presents a puzzle that hinges on a search for explanations of the sources of the two extra risk factors that are central to the three-factor model. These factors are: first, the size premium (defined as the difference between rates of return on a large size stocks and small size stocks); and, second the value premium (defined as the difference between rates of return on high Book-to-Market stocks and low Book-to-Market stocks). The purpose of this thesis is to offer a careful empirical analysis of the Fama and French three-factor model, which will add to our knowledge about the source of the systematic risks associated with these two factors. The study consists of three sections. In the first section, the three-factor model is tested under the time-varying volatility condition by using Generalized Autoregressive Conditional Heteroscedastic (GARCH) models in two time periods, June 1963-December 1991 and September 1927-December 2005 in the US market. The results indicate that the time-varying volatility does not improve the performance of the three-factor model in explaining the rates of return, but it does enhance the efficiency of the regression model by reducing the value of standard deviation and serial autocorrelation within residuals. In the second and third section, the potential relationship of the value premium with several macroeconomic risk factors, measured as the industrial production, inflation rate, the money supply, and the interest rate, are tested from January 1959 to December 2005 in US market. By using the methodology of the Cointegration test to focus on the long run relationship and conditional volatility by GARCH model to focus on risk relationship, the results suggest that i) the value premium is related to the changes of fundamental risk; ii) there is an asymmetric effect on the price of the value stock and growth stock under different business conditions; iii) and the three risk factors are driven by a similar source of macroeconomic activity change, but the interactive relationship between these three risk factors is essential in explaining the rates of return, thus, they should be used together. Overall, the results in this thesis support the view that the Fama and French three-factor model is a strong model in explaining rates of return, and that the value premium is generated from systemic risk and should be used in the equilibrium asset pricing model. The finding is useful for academics and practitioners alike.

published_or_final_version / Business / Master / Master of Philosophy

No description available.

No description available.

No description available.

Following the important work on unit roots and cointegration which started in the mid-1980s, a great deal of econometric works has been devoted to the study of the subtleties and varieties of near nonstationarity and persistence that characterize so many economic and financial time series. In recent years research activity has gained importance with outstanding contributions made on estimation and testing of a wide variety of long memory processes, together with many interesting and imaginative applications over a wide variety of different fields of economics and finance. For these reasons, this study provides empirical evidence to an aspect of fractional differencing and long memory processes, or the long memory of volatility. Evidence of long memory persistence is explored using stock price indices for eight emerging economies in both Asian and Latin American markets. The concern with the presence of long memory in higher moments of return series was first drawn by Ding, Granger and Engle (1993), using asset returns. Baillie, Bollerslev and Mikkelsen (1996) developed the fractionally integrated GARCH, or FIGARCH, process to represent long memory in volatility. The measure of long-memory persistence in the volatility is employed either using the original rescaled range statistic by Hurst (1951) and its modified version proposed by Lo (1991). Further analysis of the presence of long memory persistence is conducted using autocorrelation analysis. All the findings point in the same direction, that is, the existence of long memory in volatility irrespective of the measure chosen. Estimation of different models of volatility is undertaken beginning with the ARCH specification and until the FIGARCH model. The results show the effects to be higher in Latin American countries than in the Asian ones. This result seems consistent with the degree of intervention in the Latin American markets, known to be much higher. / Other possible explanations for the occurrence of long term persistence are also pursued such as the Regime Switching modelisation proposed first by Hamilton and Susnel (1994) with the SWARCH approach. Results show that this approach can bring another possible explanation for persistence, specially in economies like Brazil that, have very different regimes for the period covered in this study.

This thesis contributes to our knowledge of the behaviour of the Chinese stock market by offering an empirical investigation into issues such as market segmentation, inter‐relationships between Chinese stock markets and inter‐relationships with foreign stock markets. Basic questions which have been typically analysed for developed stock markets are considered in this thesis. These include an analysis of core concepts such as volatility; causal links with economic variables and the reasons why the theoretical stock price may be different from the actual stock price. Methodological methods include; cointegration, generalised autoregressive heteroscedastic modelling (GARCH), vector autoregressive framework modelling (VAR) and panel data analysis. Both daily and monthly observations are used over a time period from 1996 to 2006. The results indicate that there is a rich set of reasons why we may observe phenomena such as a discount on B shares and a relationship between A shares and B shares. The findings also suggest that China is not isolated from the rest of the world and that there is evidence of inter‐relationships with foreign stock markets and that Chinese stock market prices are close to their fundamental value. This is not generally the finding for developed stock markets. Overall, it appears that the methodological approaches usually associated with developed stock markets can serve us well as useful tools in creating a deeper understanding of the underlying fundamentals describing the Chinese stock market.

This thesis examines the spatial distribution of prespawning herring (<I>Clupea harengus, </I>L.) in the northern North Sea and its relationship with various environmental factors. Data were collected from 1992 to 1995 during the ICES coordinated herring acoustic surveys. The spatial distribution and organization of the species was studied by means of geostatistics. A significant interannual variability was observed in the spatial distribution and the horizontal dimensions of the herring structures. It was found that herring had narrowed their spatial occupation and had been progressively confined to distinct areas. These areas were stable throughout the four years, had the highest herring abundances throughout the studied period and appeared to be vacated last as the population declines. These were around the Shetland and Orkney islands and in the south part of the studied area. The confinement of the herring population in a smaller region of the surveyed area was believed to be related to the decreasing stock. Herring tended to aggregate mainly in mesoscale clusters of dense schools having a diameter of 7˜9n. miles and occasionally in longer range spatial patterns associated with, or corresponding to, oceanographic features. The number and dimensions of the structures that the population was aggregated in, had gradually decreased. A method for a robust analysis and optimal mapping of the highly aggregated herring population was introduced by combining the robust estimator of the variogram with the cross-validation technique. The method effectively handled the destabilizing effect that the few high value observations had and revealed the underlying spatial dependence (larger scale structural component) by defining better the correlation ranges of herring.

No description available.

Thesis (PhD)--University of South Australia, 1999