Accounting earnings properties and determinants of earnings response coefficient in Brazil

• Autor(a) principal: Pimentel, Renê Coppe
• Data de Publicação: 2009
• Tipo de documento: Tese
• Idioma: eng
• Título da fonte: Biblioteca Digital de Teses e Dissertações da USP
• Texto Completo: http://www.teses.usp.br/teses/disponiveis/12/12136/tde-30012010-095104/
• topic: Accounting earnings; Capital markets; Contabilidade financeira; Corporate finance; Finanças das empresas; Financial accounting; Lucro contábil; Mercado de capitais

A fundamental issue at the interface of economics, finance, and accounting involves the relation between a firm\'s reported earnings and its stock returns. The lack of research in this field using Brazilian data and the limitations of previous research in terms of time-series data (small length available) motivates the present research. In addition, the practical justification of this research is that time-series properties of accounting earnings and the determinants of Earnings Response Coefficient (ERC) have a direct application in earnings forecasting and the valuation process. Based on this, the general objectives of this dissertation are to analyse the earnings time-series properties and to find the economic determinants of ERC in Brazil. Consequently, this dissertation is divided into three main sections/studies: (1) An analysis of the time-series properties of accounting earnings and the long-term relationship among price, return and earnings; (2) An analysis of the relevance and significance of ERC for individual companies and pooled data; and, (3) Elucidation of the economic determinants of ERC in Brazil. In order to achieve these objectives, quarterly and annual data were gathered and analysed. The quarterly sample is composed by 71 firms with quarterly data from the first quarter of 1995 until first quarter of 2009 (57 time-observations), and the annual sample is composed by 61 firms and annual observations from 1995 to 2008 (14 time-observations). Two measures of accounting earnings (SEPS and UNEPS) and two measures of stock returns (RET and ARET) were used. Additionally, proxies of systematic risk (BETA), expected economic growth opportunity (GRO), leverage (LEV), risk-free interest rate (INTER) and size (SIZE) were used as measures of the economic determinant of ERC. In each study, the two different measures of earnings and returns resulted in a combination of four functional models (regressions), in an annual and a quarterly basis. These models were estimated into firm-specific level and pooled data by using different methods (OLS and GLS); these varieties of designs, periodicity and estimations provide a robust analysis. The results of the first study show that earnings present, for most firms, stationarity series and seasonal fluctuation. The evidence also suggests that the accounting earnings in Brazil follow an auto-regressive model AR(1). Test results indicate long-term relationships between earnings and prices/returns, although, it is not possible to robustly infer about the Granger causality direction since a general behaviour was not identified. The second study indicates that for annual and quarterly firm-specific regressions between earnings and stock returns, only a few companies presented a significant relationship. However, the annual pooled analysis presents positive and significant coefficients, and contemporaneous observations (at t level) seem to fit better in the models than the lagged variable of return. Cross-sectional weight in the panel aggregates some refinement to the models in terms of significance and explanatory power. In the quarterly pooled regressions, coefficients with statistical significances were found; nevertheless, these regressions report an extremely low or nonexistent explanatory power, suggesting a slight relationship between the variables. The results of the third study show that systematic risk, interest rates and size significantly explain cross-sections and intertemporal variations of ERC according to previous hypothesis. On the other hand, differently from what has been hypothesized, expected economic growth and leverage do not significant explain cross-section variations of ERC in Brazil. Since the interest rate level in Brazil is higher than those in developed countries and given that interest rate levels affect both earnings and discount rate, the regressions presented different signals according to the proxy for return used. Finally, it is possible to conclude that, by including the significant factors noted above, the empirical specification of the earnings-returns relation is significantly improved, however, given some contrasting results presented here, this dissertation advocates for further research in this field.