Baker and Wurgler (2002) proposed market timing theory and indicated the observed capital structures are the outcomes that managers timed the equity market and took advantages of timing when information asymmetry is low and stock price is high. But many scholars argue that Baker and Wurgler¡¦s timing proxy is noisy, this study attempts to use the concept of Huang and Ritter (2009) to test market timing effect on capital structure more directly by cost of equity. The cost of equity in this study is estimated by Fama and French three factors model with five-year rolling regression which is different from Huang and Ritter (2009). The empirical results show that publicly traded firms in Taiwan Stock Exchange from 1996 to 2007 tend to issue debt when the cost of equity is high and issue equity when the cost of equity is low which means the timing of financing behavior exists but it has no long-lasting effect on capital structure. Indicating that the observed capital structures of publicly traded firms in Taiwan Stock Exchange aren¡¦t the outcomes that managers timed the equity market which is not identical to the perspectives of Baker and Wurgler (2002) and the speed of adjustment of capital structure of publicly traded firms in Taiwan Stock Exchange is very fast.

The traditional fund performance-evaluating measures rely on historical returns; however, this return-based performance measures are demonstrated with less precision but more biases than the holding-based measures by many studies. Therefore, the paper uses both return-based measure and holding-based measure by Cohen et al (2005). The equity fund samples begin from January 2004 to December 2008 with monthly returns and seasonally holdings of equity funds. The purpose of the study is to compare the predicting ability and information-containing ability between the four models: CAPM alpha model, Fama & French alpha model, CAPM holding based alpha model, and Fama & French holding based alpha model. At the end, the study analyzes the stock selection ability of Taiwan equity funds with the model which has the best predicting and information-containing power. The result shows that the best predicting power models are Fama & French model and CAPM model. Also, the best information-containing models for predicting future returns are Fama & French model and CAPM model. Thus, the study uses both Fama & French model and CAPM model to analyze the stock-picking ability of Taiwan equity funds. And we find that the funds have no stock selecting ability under Fama & French alpha model, but have the contrary results under CAPM alpha model. However, considering the number of the factors and the explanation of the two models, we conclude this paper with Fama & French model which shows Taiwan equity funds having no stock selection ability.

Submitted by Luiz Fernando Alves Jr (luizfalvesjr@yahoo.com.br) on 2011-09-14T17:50:22Z No. of bitstreams: 1 Dissertacao Luiz Alves Jr.pdf: 640831 bytes, checksum: 63a6f4da7dee84ca741e9567c860b77e (MD5) / Approved for entry into archive by Suzinei Teles Garcia Garcia (suzinei.garcia@fgv.br) on 2011-09-14T19:26:29Z (GMT) No. of bitstreams: 1 Dissertacao Luiz Alves Jr.pdf: 640831 bytes, checksum: 63a6f4da7dee84ca741e9567c860b77e (MD5) / Approved for entry into archive by Suzinei Teles Garcia Garcia (suzinei.garcia@fgv.br) on 2011-09-14T19:26:45Z (GMT) No. of bitstreams: 1 Dissertacao Luiz Alves Jr.pdf: 640831 bytes, checksum: 63a6f4da7dee84ca741e9567c860b77e (MD5) / Made available in DSpace on 2011-09-14T19:27:22Z (GMT). No. of bitstreams: 1 Dissertacao Luiz Alves Jr.pdf: 640831 bytes, checksum: 63a6f4da7dee84ca741e9567c860b77e (MD5) Previous issue date: 2011-08-14 / Fama & French (1993) Three Factors Model is an extension of the Sharpe-Lintner & Black (1972) asset-pricing model, the CAPM. In this model, the market value and ratio price to book value of the firms are incorporated as explainable variables to the CAPM, adding to the market-risk factor β of the model. The Three Factors Model was originally developed to the American stock market and then successfully rolled-over to several different countries markets. Some studies have been done to the Brazilian case however the lack of data has compromised the results. The aim of this work is to test the Fama & French (1993) model to the Brazilian stock market using a new methodology to build the portfolios. The innovation in this work is the usage of a moving data-base that incorporates new stocks to the portfolios as they debut in the stock market and reaches the liquidity parameters for the model. In the previews works, the databases were made of fixed sets of stocks. This methodology aims to mitigate the lack of data problem in the Brazilian stock market. The period of analysis is from 2000 to 2011, and the stocks used to build the portfolio are those with reliable data and the ones that present at least one trade per 2 months. The analysis of the Three Factors Model were made using the Black, Jensen & Scholes (1972) linear regression approach, the same applied by Fama & French (1993) in their work. Sixteen portfolios were used as the dependent variables. They were built trough the crossing of 4 groups of stocks organized according to their market value (ME) and their ratio price to book value (ME/BE). Two portfolios were built as the independent variables. They are a set of stocks that mimic the market value risk factor, the SMB portfolio, and the price to book-value risk factor, the HML portfolio. The method used to estimate the parameters of the equation was the Ordinary Least Square. The results found for the Brazilian stock market were very similar to the ones found by Fama & French (1993). The first one was the same empirical contradictions of the CAPM found by Fama & French (1993) for the American market. The Betas from the CAPM had no apparent relation to the expected return of the stocks. Next, the Three Factors Model presented a higher explaining (R²) power to the portfolios returns and was statistically significant to 15 of the 16 tested portfolios. The coefficient of the regressions related to the risk factors SMB and HML presented, in the vast majority, the same signals of the ones found by Fama & French (1993). A small discrepancy was found in some HML coefficients and it was explained by the performance of the Brazilian economy and stock market in the period. At last the Three Factors Model proved itself a much better tool to evaluate the risk factors of Brazilian stocks then the CAPM. / O modelo de três fatores de Fama & French (1993) é uma extensão do modelo de precificação de ativos de Sharpe (1963), Lintner (1965) e Black (1972), o CAPM. Em Fama & French (1993), o valor de mercado e o valor contábil das empresas são adicionados como variáveis explicativas ao fator de risco de mercado β do CAPM. O objetivo deste trabalho é testar o poder explicativo do modelo de três fatores para o mercado acionário brasileiro. A inovação deste trabalho foi a utilização de um universo de ações móvel, no qual os títulos que são lançados na Bovespa no período de análise vão sendo incorporadas à base de dados conforme atingem os requisitos do modelo. Trata-se de uma abordagem inovadora, já que tradicionalmente o universo de ações que compõe a amostra é rígido do início ao fim do período de análise. Esta abordagem foi desenvolvida com o intuito de mitigar o problema de falta de dados do mercado acionário brasileiro. O período de análise foi de 2000 à 2011, e as ações utilizadas foram todas aquelas que possuíam um histórico confiável e apresentaram pelo menos um negócio à cada dois meses. A análise do Modelo de Três Fatores foi realizada utilizando a metodologia de séries temporais de Black, Jensen e Scholes (1972), da mesma forma que Fama & French (1993). Como variável dependente foram utilizadas 16 carteiras, oriundas do cruzamento das ações dividas em 4 percentis iguais pelos seus valores de mercado (ME), e 4 percentis iguais pela razão valor de mercado pelo valor contábil (ME/BE). Como variáveis independentes foram construídas duas séries de retorno que replicam os fatores de risco valor de mercado, SMB, e a razão valor de mercado pelo valor contábil, HML. Estas foram construídas pela diferença dos retornos das ações de maior ME e menor ME; e pela diferença do retorno das de maior ME/BE, pelas de menor ME/BE. O método de estimação dos coeficientes das regressões utilizado foi o dos mínimos quadrados ordinários. Os resultados do Modelo encontrados para a bolsa brasileira foram similares àqueles encontrados por Fama & French (1993). O Modelo apresentou maior poder explicativo para os retornos dos portfolios analisados que o CAPM, e mostrou-se estatisticamente significante para 15 das 16 carteiras. Os coeficientes das regressões relativos aos fatores de risco SMB e HML apresentaram, em sua maioria, os mesmo sinais que os encontrados por Fama & French (1993). Foi encontrada uma discrepância relativa ao sinal do fator HML para as carteiras de maior ME/BE, cuja explicação está atrelada ao momento da economia e mercados no período. Por fim, o Modelo e a discrepância foram reavaliados dividindo-se o período de análise em pré e pós-crise de 2008. O modelo mostrou maior poder explicativo para o período pós-crise que para o pré-crise. A mesma discrepância do sinal de HML foi encontrada no pré-crise, entretanto não foi verificada no pós-crise.