Doctor of Philosophy / The motivation for the research presented in this thesis stems from the recent availability of high frequency limit order book data, relative scarcity of studies employing such data, economic significance of transaction costs management, and a perceived potential of data mining for uncovering patterns and relationships not identified by the traditional top-down modelling approach. We analyse and build computational models for order submissions on the Australian Stock Exchange, an order-driven market with a public electronic limit order book. The focus of the thesis is on the trade implementation problem faced by a trader who wants to transact a buy or sell order of a certain size. We use two approaches to build our models, top-down and bottom-up. The traditional, top-down approach is applied to develop an optimal order submission plan for an order which is too large to be traded immediately without a prohibitive price impact. We present an optimisation framework and some solutions for non-stationary and non-linear price impact and price impact risk. We find that our proposed transaction costs model produces fairly good forecasts of the variance of the execution shortfall. The second, bottom-up, or data mining, approach is employed for trade sign inference, where trade sign is defined as the side which initiates both a trade and the market order that triggered the trade. We are interested in an endogenous component of the order flow, as evidenced by the predictable relationship between trade sign and the variables used to infer it. We want to discover the rules which govern the trade sign, and establish a connection between them and two empirically observed regularities in market order submissions, competition for order execution and transaction cost minimisation. To achieve the above aims we first use exploratory analysis of trade and limit order book data. In particular, we conduct unsupervised clustering with the self-organising map technique. The visualisation of the transformed data reveals that buyer-initiated and seller-initiated trades form two distinct clusters. We then propose a local non-parametric trade sign inference model based on the k-nearest-neighbour classifier. The best k-nearest-neighbour classifier constructed by us requires only three predictor variables and achieves an average out-of-sample accuracy of 71.40% (SD=4.01%)1, across all of the tested stocks. The best set of predictor variables found for the non-parametric model is subsequently used to develop a piecewise linear trade sign model. That model proves superior to the k-nearest-neighbour classifier, and achieves an average out-of-sample classification accuracy of 74.38% (SD=4.25%). The result is statistically significant, after adjusting for multiple comparisons. The overall classification performance of the piecewise linear model indicates a strong dependence between trade sign and the three predictor variables, and provides evidence for the endogenous component in the order flow. Moreover, the rules for trade sign classification derived from the structure of the piecewise linear model reflect the two regularities observed in market order submissions, competition for order execution and transaction cost minimisation, and offer new insights into the relationship between them. The obtained results confirm the applicability and relevance of data mining for the analysis and modelling of stock market order submissions.

The main purpose of this paper is to investigate the effect (if any) of investor sentiment on asset prices. To calibrate the ability of various market sentiment variables in forecasting stock returns, we followed the recursive regression methodology by Pesaran and Timmermann (1995,2000), taking into account the influences of regime switches on trading decisions of investors in real time. Our results suggest that stock returns may be difficult to predict when stock market is relatively unstable and investors are unsure of which forecasting model to be employed for trading strategies. This finding is not consistent with the empirical results of Pesran and Timmermann (1995). We also find that net buy (sell) of investment trusts and security dealers become in a close relation with stock returns after 1998, implying that institutional investors seem to reasonably capture the sentiment of the market and their trading strategies may reflect information asymmetries between managers and investors.

本研究係探討台灣股市流動性的提供者，研究樣本係台灣證券交易所四支流動性大小不同的標的個股，而樣本期間橫跨民國八十五一至五月及民國八十六年全年，分別代表證券市場的盤整期間，大多頭及大多頭。 研究方法係將證券市場上的參與者依其身份別分別自營商、國內籌資投信、國外籌資投信、外資、法人及自營人六種身份，再依績效指標及行為指標去判斷何者較有可能是市場中流動性的提供者。一個好的流動性提供者必須在績效指標方面有穩定的表現及在行為指標方法符合流動性供給者的交易方式。 研究結果發現自營商及法人的獲利狀況十平穩定，不論市場狀況如何，均能有明顯大於零的表現，且其賣單金額比例較為平均，顯示其進出狀況較為對稱，且對獲利進行迴歸時亦顯示其買賣愈均衡時獲利愈大，因此就績效指標及行為指標來看此兩種身份別的市場參與者最符合市場中流動性提供者的特性。 而在投信及外資方面則發現其獲利狀況變動程度相當大，隨著市場狀況的不同而對其有著明顯的影響，其賣單金額比例較大幅度的擺動亦顯示出其較傾向單方向的交易，因此可知其操作決策與獲利狀況會因績效壓力會造成追漲殺跌的交易特性，因此買賣之間比例並不均衡且獲利受市場狀況影響甚大，並不是合適的流動性提供者。自然人在平均獲利上並沒有特別的優勢或劣勢，但其損益狀況並不對稱、使用信用交易的比例甚大，且愈傾向以市價進行委託的自然人其信用交易比例亦愈高，顯示其在擴充槓桿的同時並沒有獲得相對應的報酬。因此儘管自然人提供了市場上大部份的成交量，但其並不是市場中流動性穩定的提供者。

The advance in technologies such as massive storage devices and high speed internet has led to an enormous increase in the volume of available documents in electronic form. These documents represent information in a complex and rich manner that cannot be analysed using conventional statistical data mining methods. Consequently, text mining is developed as a growing new technology for discovering knowledge from textual data and managing textual information. Processing and analysing textual information can potentially obtain valuable and important information, yet these tasks also requires enormous amount of computational resources due to the sheer size of the data available. Therefore, it is important to enhance the existing methodologies to achieve better scalability, efficiency and accuracy. / The emerging Grid technology shows promising results in solving the problem of scalability by splitting the works from text clustering algorithms into a number of jobs, each to be executed separately and simultaneously on different computing resources. That allows for a substantial decrease in the processing time and maintaining the similar level of quality at the same time. / To improve the quality of the text clustering results, a new document encoding method is introduced that takes into consideration of the semantic similarities of the words. In this way, documents that are similar in content will be more likely to be group together. / One of the ultimate goals of text mining is to help us to gain insights to the problem and to assist in the decision making process together with other source of information. Hence we tested the effectiveness of incorporating text mining method in the context of stock market prediction. This is achieved by integrating the outcomes obtained from text mining with the ones from data mining, which results in a more accurate forecast than using any single method.

<p>The Chinese stock market is a remarkable emerging market, the two stock markets Shanghai and Shenzhen Stock Exchanges were both established in 1990, and since then they have been playing a very important role in Chinese economy. More and more attention is focused on the emerging Chinese market, and investors have been trying to find the opportunity to achieve abnormal returns through the Chinese stock market. We name this phenomenon market efficiency anomaly, one pattern of which is seasonality effect. In our study, we would like to choose the seasonality effect as the approach.</p><p>This study focuses on Shanghai Stock Exchange Composite Index, and we settle two research questions: Does seasonality effect exist in Chinese Stock exchange? Is the seasonality effect persistent over times?We try to test the seasonality in Chinese stock market by day of the week effect, January effect and semi-month effect. Deductive approach and quantitative research method are used in this thesis. To analyze seasonality effect, the data has been collected from Shanghai Stock Exchange Index and has been tested in four periods: 1992-1996,1997-2001, 2002-2006 and the whole period 1992-2006. Null hypothesis and T-test with α=0.05 is used to test the seasonality effect. The results show that seasonal anomalies like Day of the week effect, positive March effect, and negative July effect exist in the Chinese stock market, while semi-month effect does not occur significantly; but the existing seasonal effect is not persistent over times. The above indicates that the Chinese stock market is not fully efficient yet. Investors may have opportunities to make use of the seasonal anomalies to earn abnormal return.</p>

<p>This thesis consists of an introductory chapter and four self-contained essays on the value relevance of financial statement information.</p><p>Essay 1: The purpose of this essay is to examine relevance of environmental information from an investor’s perspective. The study proposes that the market value of companies will reflect both financial and environmental performance. The theoretical foundation of the study is the accounting based valuation theory outlined by Ohlson (1995). This study provides new insights into how environmental performance is reflected in the market value of Swedish companies listed on the Swedish Stock market.</p><p>Essay 2: In financial accounting research, much effort has been devoted to study the relation between accounting earnings and stock prices. The primary purpose of the second essay is to investigate the effect of alternative return-earnings model specifications to the estimated returns-earnings relation, that is, the earnings response coefficients. The returns-earnings models investigated include the traditional earnings levels and changes, and models including analysts’ earnings forecasts based on Ohlson’s (1995) extended residual income model.</p><p>Essay 3: Fundamental analysis research that focuses on the use of accounting information to estimate equity value, has surfaced as a central theme in market based accounting research of the 1990s (Lee, 1999). The purpose of third essay is to compare two different approaches to valuation based on the theory presented in Ohlson (1995) in terms of explanatory and predictive power of the value estimates. Both approaches are implemented with and without the use of analysts forecasts. </p><p>Essay 4: In this essay data from the Swedish stock market is used to investigate the profitability of two different types of investment strategies based on fundamental-to-value ratios and past insider trading activity. The purpose of the research is to explore four related research questions: (i) Do accounting based trading strategies generate abnormal returns on the Swedish stock market?; (ii) Do trading strategies based on insider trading behaviour generate abnormal returns on the Swedish stock market?; (iii) Do insiders who buy stocks tend to favour value stocks and do insiders who sell stocks tend to dispose growth stocks?; and (iv) Are insiders able to discriminate between temporary high/low fundamentals and temporary low/high prices when buying/selling value stocks and growth stocks? </p>

<p>This paper is an empirical study on Harry Markowitz work on Modern Portfolio Theory. The model introduced by him assumes the normality of assets’ return. We examined the OMX Large Cap List1 by mathematical and statistical methods for normality of assets’ returns. We studied the effect of the parameters, Skewness and Kurtosis for different time series data. We tried to figure it out which data series is better to construct a portfolio and how these extra parameters can make us better informed in our investments.</p>

<p>Forecasting the stock market is a complex task, partly because of the random walk behavior of the stock price series. The task is further complicated by the noise, outliers and missing values that are common in financial time series. Despite of this, the subject receives a fair amount of attention, which probably can be attributed to the potential rewards that follows from being able to forecast the stock market.</p><p>Since artificial neural networks are capable of exploiting non-linear relations in the data, they are suitable to use when forecasting the stock market. In addition to this, they are able to outperform the classic autoregressive linear models.</p><p>The objective of this thesis is to investigate if the stock market can be forecasted, using the so called error correction neural network. This is accomplished through the development of a method aimed at finding the optimum forecast model.</p><p>The results of this thesis indicates that the developed method can be applied successfully when forecasting the stock market. Of the five stocks that were forecasted in this thesis using forecast models based on the developed method, all generated positive returns. This suggests that the stock market can be forecasted using neural networks.</p>

<p>Bakgrund: 4 av 5 svenskar äger aktier i någon form och det är många som är intresserade av att maximera sin avkastning. Det har lett till att det skrivs mycket i media om olika sätt att få avkastning högre än marknaden. Om det skulle vara möjligt att nå överavkastning är det en indikation på att marknaden inte är effektiv.</p><p>Syfte: Syftet med studien är att undersöka om det historiskt har gått att nå en överavkastning genom tillämpa momentum- eller motsatstrategin på den svenska aktiemarknaden. Syftet är även att testa om den svenska aktiemarknaden har varit effektiv i svag form.</p><p>Genomförande: Momentumstrategin testades genom att portföljer formades med de tio aktier som har haft högst relativ prisstyrka de senaste sex månaderna. Portföljernas marknadsjusterade avkastning har sedan följts i sex månader. Motsatsstrategin testades genom att vinnarportföljer formades med de tio aktier som har haft högst avkastning de senaste tre åren och förlorarportföljer formades för de med lägst avkastning. Portföljernas marknadsjusterade avkastningar under de kommande tre åren jämfördes sedan med varandra.</p><p>Slutsats: Vi har kommit fram till att det har gått att nå en överavkastning med hjälp av momentumstrategin och att den har fungerat bäst i perioder av stabila uppgångar. Det har däremot inte gått att få överavkastning med hjälp av motsatsstrategin. Vi menar att resultatet beror på att marknaden underreagerar. Därmed kan vi säga att den svenska aktiemarknaden under den undersöka tidsperioden inte har varit effektiv i svag form.</p> / <p>Background: 4 out of 5 Swedes own stocks in some form and many people are interested in maximising their profits. This has led to a lot of publicity in ways to get profits higher than the market. The eventual possibility to receive abnormal returns indicates that the market is inefficient.</p><p>Purpose: The purpose of the study is to investigate if it historically has been possible to receive abnormal returns by implementing momentum- and contrarian strategies on the Swedish stock market. The purpose is also to test if the Swedish stock market has been efficient in weak form.</p><p>Implementation: The momentum strategy was tested by forming portfolios consisting of the ten stocks with the highest relative price strength over the last six months. The portfolio abnormal returns were then followed for the following six months. The contrarian strategy was tested by forming winner portfolios consisting of the ten stocks with the highest abnormal returns over the last three years. Loser portfolios were formed of the ten stocks with the lowest abnormal returns over the last three years. The portfolios abnormal returns were then compared to each other for the following three years.</p><p>Conclusion: We have reached the conclusion that it has been possible to receive abnormal returns by using the momentum strategy. It has been the most successful in periods of steady raises. The contrarian strategy has not generated any excess returns. We believe that reason for the results is that the market under reacts. Therefore we can state that the Swedish stock market not has been efficient in weak form during the examined period.</p>

<p>I denna uppsats<strong> </strong>undersöker jag huruvida CAPM eller APT modellerna ger resultat på den svenska aktiemarknaden mellan 1998 och 2007. Jag undersöker om någon av dessa modeller passar in i den svenska aktiemarknaden och hurdan är resultatet. Det som gör uppsatsen intressant är att jag använder båda modellerna i en och samma studie och jämför resultatet av uppskattningar av modellerna. Som markandsindex har jag använt OMXS30 index och som makroekonomiska variabler i APT-modellen har jag använt inflation, oljepris, industriproduktionsindex och ränta. Resultatet av studien är att med de makroekonomiska variablerna, som jag har använt, ger både CAPM och APT likvärdiga resultat. Slutsatsen av min studie är att APT med de rätta variablerna är en bättre modell att skatta priset på aktier än CAPM.</p> / <p>In this essay, I examine whether CAPM or APT models give results on the Swedish stock market between 1998 and 2007. I examine if either CAMP or APT or both of these models fits in on the Swedish stock market and what the result is. What makes this essay interesting is that I use both models in one and the same study and compare the result of my estimates with these two models. As market index, I have used the OMXS30 index and as macroeconomic variables in APT model, I have used inflation, oil price, industry production index and interest. The result of the study is that with the macroeconomic variables, that I have used with APT and CAPM, gives CAPM and APT equivalent results. The conclusion of the study is that APT, with the correct variables, is better model for estimating the stock prices than CAPM.</p>