Idiosyncratic volatility: an analysis of aggregate and Individual effects

• Autor(a) principal: Carolina Magda da Silva Roma
• Data de Publicação: 2017
• Tipo de documento: Tese
• Idioma: por
• Título da fonte: Repositório Institucional da UFMG
• Texto Completo: http://hdl.handle.net/1843/33869
• topic: CSV; EGARCH; Expected Return; Mercado; Administração

This work is focused on the study of idiosyncratic risk at the aggregate and firm-level and its effect on returns. For that analysis we use a model-free measure, namely the cross-sectional variance (CSV), easily obtained at any frequency, as proposed in Garcia, Mantilla-García, and Martellini (2014). This measure was employed in the U.S. and Brazilian stock markets using equal- and value-weighted market returns, at daily and monthly frequencies. For U.S. stocks, first the main findings in the aforementioned authors’ work were replicated and their sample was extended until December 2014. The results relative to these estimations imply that aggregate idiosyncratic volatility is still positive and significant in predicting equal- and value-weighted market returns. In a second moment, credit ratings of firms rated by Standard & Poor’s were used to construct a CSV measure according to the firm’s credit quality into investment grade, non-investment grade and all rated firms. The analysis relying on CSV by levels of ratings pointed out to the relevance of monthly value-weighted measures in predicting market returns in the whole sample. Using Brazilian stocks from January 2000 to June 2016, the behavior of CSV was investigated as a proxy to idiosyncratic variance over time and its predictive power to forecast market returns throughout the whole timespan, subperiods, up and down markets, expansion and recession periods as well testing its power with the inclusion of market variance, investor sentiment, aggregate dividend yield and expected and unexpected measures of illiquidity. It was also analyzed whether that measure can predict risk factors omitted from the CAPM model as presented in Carhart (1997) and Fama and French (2015) five-factor model. Overall, these findings suggest that the firm-specific risk aggregated over the stocks in the sample is not able to robustly predict market returns or risk factors. Using credit ratings of Brazilian firms showed to be very limited due to the number of rated companies, therefore a new CSV based on ratings or the stock’s market capitalization (size) were constructed; however, these approaches also do not corroborate the idea that aggregate idiosyncratic variance is useful in predicting market returns. After that, attention was focused on understanding the cross-sectional effects of expected idiosyncratic volatility and expected return using an EGARCH model developed by Nelson (1991) and a Skewed – Generalized Error Distribution (Skew-GED). Portfolios were formed based only on expected idiosyncratic risk and controlling for other characteristics which have been shown to affect the stock return, specifically size, book-to-market, momentum and reversal return applying Ang et al.’s (2006) methodology to account for these variables in a double sorting procedure. The abnormal returns generated by these portfolios, in general, are not statistically significant when controlling for other characteristics. Fama and MacBeth (1973) cross-sectional regressions to evaluate the relation between expected returns and expected idiosyncratic volatility controlling for portfolio beta, market capitalization, book-to-market, turnover, momentum, coefficient of variation of turnover, and lagged return were run. Using the three different models consistently suggest that expected idiosyncratic volatility is not related to expected returns when the forward-looking return observation is not included in the estimations supporting Fink, Fink and He (2012) results.