Adoção das IFRS e a relevância da informação contábil utilizando regressão quantílica

• Ano de defesa: 2017
• Autor(a) principal: Santos Neto, Magno dos
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal do Espírito Santo; BR; Mestrado em Ciências Contábeis; Centro de Ciências Jurídicas e Econômicas; UFES; Programa de Pós-Graduação em Ciências Contábeis
• Palavras-chave em Português: Relevância da informação; Regressão quantílica; International Financial Reporting Standards (IFRS); Contabilidade - Normas; Divulgação de informações contábeis; Análise de regressão; Ciências Contábeis; 657
• Link de acesso: http://repositorio.ufes.br/handle/10/8874

The objective of this research is to analyze how convergence to international standards (IFRS) in Brazil affects the relevance of the accounting information, evaluated through the model of Collins, Maydew and Weiss (1997). Two samples were constructed, with and without treatment of discrepant values and exclusion of negative, non-probabilistic equity listed on the São Paulo Stock Exchange. The samples were formed by annual data from 2005 to 2014. The methods adopted were MQO, usually used in the estimation of the parameters of the linear value relevance models, and quantile regression, considering the distributions of the variables analyzed, the presence of discrepant values and the sample size. The results of the quantile regressions indicate that: (i) the adoption of IFRS in Brazil impacted the associative capacity of accounting profit in relation to the share price; (ii) the adoption of IFRS in Brazil impacted the associative capacity of shareholders’ equity in relation to the share price; (iii) and it is possible to affirm that the relevance of the adoption of IFRS varies according to the percentile of company stock prices. Such findings suggest that treatments for discrepant values and the exclusion of net worth do not totally eliminate the effects of discrepant values combined with the lack of normality of the variables influence the estimates of linear models. The quantile regression was more efficient and less likely to estimate errors than the traditional estimation methods (OLS and MV).