Avaliação de empresas: análise da confiabilidade dos modelos de estimação do valor intrínseco das ações negociadas na BOVESPA no período 1995-2007

Abstract

Investment valuation has become increasingly relevant in business decisions inside and outside organizations, as in the buying and selling of securities. Previous research has been devoted to measuring and comparing the performance and reliability of stock intrinsic value estimation models, which appraise the present value of expected future worth after considering the organizations economic and financial fundamentals. Several elements seem to have contributed to the upsurge of Brazilian stock market efficiency and to the improvement of intrinsic value estimation models reliability after the implementation of an economic stabilization program in 1994: globalization, financial markets internationalization, technological advances in communication and data transmission, accounting standards harmonization and the substantial increase in trading volume and in the number of investors at São Paulos stock exchange (BOVESPA). Conversely, economic crises and euphoric episodes in the stock market may have negatively influenced the models reliability. This dissertation aimed to measure the reliability of valuation models in assessing the intrinsic value of stocks traded in BOVESPA and analyze how this reliability evolved throughout the 19952007 period. To this end, a quantitative and descriptive inquiry was conducted in which the models reliability was measured by their accuracy, bias and explanatory power. The models used in the research were the Discounted Cash Flow to Equity model, the Residual Income Valuation model and the Abnormal Earnings Growth model, as described by Ohlson and Juettner-Nauroth. The samples comprised the stocks in the BOVESPA index (Ibovespa) in the January-April portfolio of the year subsequent to each valuation date. Considerable change has been perceived in the constitution of the portfolio over the years of the study. The alternative use of ex ante and ex post growth and expected return rates, the conditional analysis regarding size, market-to-book ratio, stock liquidity and capital structure, as well asthe complementary analysis of the models have proved relevant, adding to the understanding of the evolution of the models reliability. Although the coefficient of determination has frequently been used as a measure of the models explanatory power in previous research, the results suggest caution, showing the occurrence of spurious relationships between the variables due to scale discrepancy and to the violation of econometric assumptions of the OLS regressions, particularly the ones pertaining to homoskedasticity and linearity. Nevertheless, the results also showed a decrease in the number of violations through the years. The models accuracy and bias analyses pointed to the increase of their reliability, especially after 2003. The models ability to identify periods of extreme market behavior, as suggested in previous research with reference to the American market, has also been verified. However, the results do not imply that the models reliability has increased sufficiently so that the inclusion of other variables may be dismissed, there being strong signs that these may be relevant in establishing market price. The results have also signaled that markets, efficient or not, are subject to anomalies and investors irrationality. Nonetheless, they appear to have always adjusted to intrinsic values, even though the exact time of the adjustment does not seem to have been foreseeable. Finally, comparing the models reliability, the Discounted Cash Flow to Equity model was less reliable than the others throughout the period of the study.