This study constructs a quantitative stock selection model across multiple sectors with the application of the Bayesian method. It employees factors from the Taiwan stock market which could explain stock returns. Under this structure, each sector that has different significant factors is allowed to be imported into sub models. The factors are calculated into alpha scores and used to do stock selection. Therefore, the demonstration of both intra and inter-sector alpha scores into sector-specific integration alpha scores is an important concept in this study. Furthermore, an enhanced index fund is built based on the model and related to the benchmark to illustrate the power of this model. Once the contents of a portfolio are decided, this model could provide stock selection criterion based on the predictive power of stock return. Finally, the results demonstrate that this model is practical and flexible for local stock portfolio analysis.

Quantitative analysis is one branch of portfolio management. The advantages of quantitative analysis are fast and objective. It has developed significantly in recent years because of the improvements in computer technology. This thesis applies the structure of a multi-factor model (MFM) to undertake quantitative analysis. Singapore has one of the most prosperous financial markets in Southeast Asia. The Singapore Stock Exchange (SGX) and Financial Times and the London Stock Exchange (FTSE) are now in cooperation, which has added vitality to this market. It has great influence in global financial markets, and this is why we select its security market to be our target in MFM. The model refers the multi-factor processes of Jeng and Tsai (2011) . For backtesting, we adopt an enhanced strategy as testimony. We transmit information from the MFM to the enhanced strategy. Then we create the stock weightings to constitute the enhanced portfolio. This model includes 68 significant descriptors, 14 composite factors and 7 industry factors. The Singapore MFM shows 43% adjusted R-Square in the sample period. The enhanced portfolio we suggested has an information ratio of 76.80% with a tracking error of 4.02% and 1.53% for monthly turnover rate.

This study constructs a cross-market risk model based upon local multi-factor risk models of Taiwan, China and Singapore equity markets. This model allows each local market to adopt different local factors rather than force all local markets to use one parsimonious set of factors. We employ the world, country, industry, and global risk factors to build a structural model which could explain the relationship between local factors across market by further decomposing local factor returns. Therefore, this model could provide both in-depth and broad coverage analysis of international equity portfolios. Furthermore, we build a simple portfolio and its corresponding benchmark to illustrate the usage of our model. Once the contents of a portfolio are decided, this model could provide not only the risk estimation and decomposition in advance but also the performance attribution compared with the benchmark after the portfolio is realized. The analytical viewpoint could also easily change with different numeraire perspectives. The result demonstrates that this model is practical and flexible for international equity portfolio analysis.

This Thesis attempts to construct a Taiwan equity multi-factor model using fundamental cross-sectional approach step by step. It is found that the model involves 28 explanatory factors (including 20 industry factors) and its explanatory power is 58.6% on average. The results of the estimations can be considered very satisfactory. Moreover, based on MFM, this study simulates applications of equity market neutral strategies through quantitative techniques over the period Jan.2003 ¡V Dec.2003. The results verified that the three major characteristics of equity market neutral portfolio performance are: 1) providing absolute return; 2) lack of correlation to the equity benchmark; and 3) low volatility due to hedged portfolio structures.

The estimation results of Brazilian risk premia are not robust in the literature. For instance, among the 133 market risk premium estimates reported on the literature, 41 are positives, 18 are negatives and the remainder are not significant. In this study, we investigate the grounds for this lack of consensus. First of all, we analyze the sensitivity of the US risk premia estimation to two relevant constraints present in the Brazilian market: the small number of assets (137 eligible stocks) and the short time-series sample available for estimation (14 years). We conclude that the second constrain, small T, has greater impact on the results. Following, we evaluate the two potential causes of problems for the risk premia estimation with small T: i) small sample bias on betas; ii) divergence between ex-post and ex-ante risk premia. Through Monte Carlo simulations, we conclude that for the T available for Brazil, the betas estimates are no longer a problem. However, it is necessary to wait until 2041 to be able to estimate ex-ante risk premia with Brazilian data. / Os resultados das estimações de prêmios de risco brasileiros não são robustos na literatura. Por exemplo, dentre 133 estimativas de prêmio de risco de mercado documentadas, 41 são positivas, 18 negativas e o restante não é significante. No presente trabalho, investigamos os motivos da falta de consenso. Primeiramente, analisamos a sensibilidade da estimação dos prêmios de risco norte-americanos a duas restrições presentes no mercado brasileiro: o baixo número de ativos (137 ações elegíveis) e a pequena quantidade de meses disponíveis para estimação (14 anos). Concluímos que a segunda restrição, T pequeno, tem maior impacto sobre os resultados. Em seguida, avaliamos as duas potenciais causas de problemas para a estimação de prêmios de risco em amostras com T pequeno: i) viés de pequenas amostras nas estimativas dos betas; e ii) divergência entre prêmio de risco ex-post e ex-ante. Através de exercícios de Monte Carlo, concluímos que para o T disponível no Brasil, a estimativa dos betas já não é mais um problema. No entanto, ainda precisamos esperar até 2041 para conseguirmos estimar corretamente os prêmios ex-ante com os dados brasileiros.

Harry Markowitz work in the 50’s spring-boarded modernportfolio theory. It gives investors quantitative tools to compose and assessasset portfolios in a systematic fashion. The main idea of the Mean-Varianceframework is that composing an optimal portfolio is equivalent to solving aquadratic optimization problem.In this project we employ the Maximally Predictable Portfolio (MPP) frameworkproposed by Lo and MacKinlay, as an alternative to Markowitz’s approach, inorder to construct investment portfolios. One of the benefits of using theformer method is that it accounts for forecasting estimation errors. Ourinvestment strategy is to buy and hold these portfolios during a time periodand assess their performance. We show that it is indeed possible to constructportfolios with high rate of return and coefficient of determination based onhistorical data. However, despite their many promising features, the success ofMPP portfolios is short lived. Based on our assessment we conclude thatinvesting in the stock market solely on the basis of the optimization resultsis not a lucrative strategy / Modern portföljteori har sitt ursprung i Harry Markowitz arbete på 50-talet. Teorin ger investerare kvantitativa verktyg för att sammansätta och utvärdera tillgångsportföljer på ett systematiskt sätt. Huvudsakligen går Markowitz idé ut på att komponera en investeringsportfölj genom att lösa ett kvadratiskt optimeringsproblem. Det här examensprojektet har utgångspunkt i Maximally Predictable Portfolio-ramverket, utvecklat av Lo och MacKinley som ett alternativ till Markowitz problemformulering, i syfte att välja ut investeringsportföljer. En av fördelarna med att använda den förra metoden är att den tar hänsyn till uppskattningsfelen från prognostisering av framtida avkastning. Vår investeringsstrategi är att köpa och behålla dessa portföljer under en tidsperiod och bedöma deras prestanda. Resultaten visar att det mha. MPP-optimering är möjligt att konstruera portföljer med hög avkastning och förklaringsvärde baserat på historisk data. Trots sina många lovande funktioner är framgången med MPP-portföljer kortlivad. Baserat på vår bedömning drar vi slutsatsen att investeringar på aktiemarknaden uteslutande på grundval av optimeringsresultatet inte är en lukrativ strategi.

Day-ahead spot electricity markets are the most transparent spot markets where one can find integrated supply and demand curves of the market players for each settlement period. Since it is an indicator for the market players and regulators, in this thesis we model the spot electricity prices. Logarithmic daily average spot electricity prices are modeled as a summation of a deterministic function and multi-factor stochastic process. Randomness in the spot prices is assumed to be governed by three jump processes and a Brownian motion where two of the jump processes are mean reverting. While the Brownian motion captures daily regular price movements, the pure jump process models price shocks which have long term effects and two Ornstein Uhlenbeck type jump processes with different mean reversion speeds capturing the price shocks that affect the price level for relatively shorter time periods. After removing the seasonality which is modeled as a deterministic function from price observations, an iterative threshold function is used to filter the jumps. The threshold function is constructed on volatility estimation generated by a GARCH(1,1) model. Not only the jumps but also the mean reverting returns following the jumps are filtered. Both of the filtered jump processes and residual Brownian components are estimated separately. The model is applied to Austrian, Italian, Spanish and Turkish electricity markets data and it is found that the weekly forecasts, which are generated by the estimated parameters, turn out to be able to capture the characteristics of the observations. After examining the future contracts written on electricity, we also suggest a decision technique which is built on risk premium theory. With the help of this methodology derivative market players can decide on taking whether a long or a short position for a given contract. After testing our technique, we conclude that the decision rule is promising but needs more empirical research.

本研究的目的為建立一套適用於台灣股市的主動式量化投資策略。本研究利用多因子 Alpha 模型為分析架構，試圖掌握多維度的股價影響因子，以資訊係數（Information Coefficient）、T-test of ICs、成功率（Success rate）以及 Quintile 累積報酬做因子有效性的檢定，篩選出穩定且有效解釋股價報酬的月頻率因子，再組合因子形成Alpha 股票評分，Alpha 可拆解成三部分，包括市場波動度、因子預測下一期報酬的能力以及因子的獲利能力。本論文以此評分做為股票投資權重的依據，建構一個以台灣中型 100 指數為標竿指數的投資組合。實證結果發現，此主動式量化投資策略能夠有效擊敗標竿指數，獲得平均每個月 3.7%的超額報酬。 　　本研究並嘗試以設定原始權重保留率的方法，控制追蹤誤差以降低週轉率與交易成本，實證結果發現，此方法可有效降低追蹤誤差，但隨著保留率提升，資訊比率（Information Ratio）與投資組合的超額報酬將降低。 / The objective of this study is to build an investment process of active quantitative stock selection model. In this study, we use the Alpha Multi-factor model to find a multitude of factors which are significantly relative to the stock return. The tests we conduct to select the factors that end up in the final multi-factor model are monthly Information Coefficient, T-test of ICs, success rate and quintile cumulative return. Then we examine how to optimally combine correlated factors and calculate the Alpha score for each stock for each period. Alpha is Volatility times IC times Score. Volatility is the cross-sectional volatility of the residual return. IC is the predictive power of the model. And Score are the cross-sectional scores for each stock. We utilize a simple method to construct the portfolio that uses the Alpha score to adjust the weight of component stocks in the benchmark. The empirical result reveals that this investment process successfully outperform the Taiwan Mid-Cap 100 Index benchmark. Moreover, this study tries to decrease the turnover rate and transaction costs by controlling the tracking error. We set the original weight retention rate of the benchmark to control the tracking error. The empirical result reveals that the method works. But as the retention rate rises, the Information ratio and the excess return drops.

Submitted by Alexandre Gomes (aludolf@gmail.com) on 2017-01-30T17:16:34Z No. of bitstreams: 1 FUNDOS MULTIMERCADOS BRASILEIROS CRIAM VALOR UMA AVALIACAO DOS ALFAS.pdf: 3747366 bytes, checksum: 91e4b622130ab4ac64d4ffc168ba2458 (MD5) / Approved for entry into archive by Renata de Souza Nascimento (renata.souza@fgv.br) on 2017-01-30T17:18:27Z (GMT) No. of bitstreams: 1 FUNDOS MULTIMERCADOS BRASILEIROS CRIAM VALOR UMA AVALIACAO DOS ALFAS.pdf: 3747366 bytes, checksum: 91e4b622130ab4ac64d4ffc168ba2458 (MD5) / Made available in DSpace on 2017-01-30T17:21:28Z (GMT). No. of bitstreams: 1 FUNDOS MULTIMERCADOS BRASILEIROS CRIAM VALOR UMA AVALIACAO DOS ALFAS.pdf: 3747366 bytes, checksum: 91e4b622130ab4ac64d4ffc168ba2458 (MD5) Previous issue date: 2016-12-16 / This work examines the alpha generation of the Brazilian multimarket fund industry, taking in account fund specific characteristics, fund strategies, investor segment during different economic conditions. The employed dataset represents 1,568 multistrategy funds from 328 different managers within a 10-year timeframe from Dec-2005 to Dec-2015. The proposed model utilizes a stepwise automatic feature selection method, similar to other authors such as Stafylas, Anderson e Uddin (2015), where features are selected from a regressor candidates list that contemplates: equity factors, interest rate factors, credit factors, currency factors and commodities factors. The results found shows that at a 5% level there is positive alpha generation, that is, Brazilian multistrategy funds on average have delivered extraordinary returns on the whole sample and at the more benign market environments. During less benign market environment multistrategy funds does not deliver alpha that is statistically distinguishable from zero and sub-strategy segmentation points to different risk exposures dynamics during different market conditions. / Esse trabalho analisa a geração de alfa da indústria brasileira de fundos multimercado, levando em consideração características específicas, estratégias de investimento e segmento de investidor durante diferentes condições econômicas. A amostra utilizada compreende 1.568 fundos de 328 gestores distintos, observados em granularidade mensal durante período de 10 anos, de dez-2005 a dez-2015. O método empírico utilizado faz uso de processo de seleção automática de regressores via stepwise, modelagem implementada por autores como Stafylas, Anderson e Uddin (2015), selecionados de uma lista prévia de candidatos a regressores amparados na literatura e contemplando fatores de ações, fatores de juros, fatores de risco de crédito, fatores de moedas e fatores de commodities. Os resultados obtidos mostram que a 5% existe geração de alfa positivo na categoria de fundos, isto é, os gestores de fundos multimercado têm capacidade de gerar, na média, retornos anormais na administração dos recursos dos investidores ao longo do tempo, tanto na amostra total quanto nos períodos econômicos mais benignos, contudo não apresentam retornos extraordinários estatisticamente diferentes de zero durante períodos econômicos menos benignos. Além disso, os fundos também apresentam diferenças sensíveis na composição dos retornos quando avaliados por sub-estratégias e em diferentes condições de mercado.

One of banks core businesses today is to, in various ways, lend capital to the market and in return receive interest rate. But giving out credit comes with great risk and, therefore, precautions need to be taken. It is impossible to forecast exactly which obligor (borrower) that will default on its exposure. However, with well functioning risk management, institutions can lower the severity of their loss. In this study, we consider using a multi-factor model to calculate concentration risk for Swedish credit portfolios, which is a type of credit risk that is usually caused by high concentration of credit exposures distributed over few industrial sectors. In its existing form, the multi-factor model uses fixed sector correlations with predetermined sectors as input. Instead, we propose to use a data-driven approach based on data from the Stockholm stock exchange. In a simulation study, we find that the distributions of total credit loss are somewhat different under the original approach than under our proposed approach. This suggests that further research is needed to investigate whether the two approaches are interchangeable.