Time-Series Properties of Financial Series and Implications for Modeling

This paper investigates the time-series properties of a wide range of corporate financial and accounting series. Unit root tests developed by Dickey and Fuller (1979) are applied to these series. The results support the hypothesis that most of these series contain both permanent (random walk) and transitory components. The results show that most financial series are dominated by a random walk component. However, for some series, such as net sales, net income, earnings per share, and returns on investments, there is a relatively significant stationary component, which suggests the presence of successful smoothing for these series. We show that smoothing may reduce volatility of financial series but it cannot produce a deterministic growth trend. Implications of nonstationarity for financial modeling are explored.