This study addresses the influence of income smoothing on earnings persistence in the Brazilian market. We argue that accounting choices made by firms in order to report stability can affect the informational level of performance reported to the market, and thus affect its usefulness, particularly for company valuations and fund allocating decisions. The sample was set based on open companies listed on the BM&FBOVESPA between 2004 and 2013. The firms in the sample were separated into two groups with regards to the presence of income smoothing, based on the EM1 and EM2 metrics from Leuz, Nanda, and Wysocki (2003). We then inserted a multiplicative dummy variable into the basic persistence model, to assess the effect of smoothing on persistence. The results show increased transience in the time series, caused by the presence of income smoothing, which denotes that reported stability, conveyed via a series of smooth earnings, decreased the sustainability of reported performance, captured by persistence of the series. Additional controls show a negative influence of smoothing on persistence for small firms and when incurring losses, and a positive one for operations in sectors with accounting regulation. We conclude that, at least for the firms in the sample, reporting of constant performance resulted in a loss in the sustainability of income as a form of information on which investors, fund providers, and administrators base their decisions.
earnings persistence; income smoothing; time-series properties of earnings; quality of accounting information
Lucro Operacional – (c) EM1 alto e EM2 baixo, extremos sem suavização por postura conservadora 
Lucro Líquido Fonte: Elaborado pelos autores. 
Lucro Operacional – (b) EM1 e EM2 altos, extremos sem suavização 
Lucro Operacional – (a) EM1 e EM2 baixos, extremos com suavização 
Fluxo de Caixa Operacional
e 
3. Residuals: (a) not normal (Shapiro-Wilk with approximation from Royston, p-value <0.001), (b) heteroskedastic (Wald modified, p-value <0.001) and (c) without autocorrelation (Cumby-Huizinga with robustness to heteroskedasticity, p-value = 0.4657). 4. Estimations robust to heteroskedasticity. 5. Additional robustness tests: (a) Control by year: no significant year. Without impact for the estimated parameters. (b) Adjustment by TA at the beginning of the period: Same persistence behaviors, but with small losses in the statistical significance of the estimated parameters - Coefficients (t): [A] 
2. [B] Smoothing model, as in Eq. 5: 
