The persistence of earnings per share

The persistence of innovations to accounting earnings per share, EPS, has important implications for equity valuation, yet it remains a largely neglected subject. This paper employs various empirical tests in order to measure the persistence of shocks to EPS for the S&P 500 index. Within the I(0)/I(1) paradigm the empirical evidence rejects the I(1) specification, supporting instead a trend-stationary representation. When fractional orders of integration are considered, the results indicate that the detrended series is long memory (d > 0) and mean reverting (d < 1). The responses decay slowly to zero, albeit 50 quarters after an initial shock the responses remain significantly different from zero. Likewise, the variance ratio evidence suggests that the effect of a shock persists over time spans characteristic of the business cycle. © 2007 Springer Science+Business Media, LLC.