This paper reassesses evidence of anticipatory income smoothing reported in DeFond and Park (1997) in light of knowledge about the shortcomings in the model used to identify the managed component of reported earnings. Our tests demonstrate that the empirical results reported in DeFond and Park are implicitly guaranteed by the method used to classify firms based on the incentive to manage or not manage earnings. Results identical to those reported in DeFond and Park (1997) and interpreted in that paper as supportive of anticipatory income smoothing are obtained by simple random assignments of firms into groups where earnings management incentives are hypothesized to exist. The present paper adopts an alternative design, based on using analysts' forecasts as proxies for un-managed earnings (and forecast errors as proxies for discretionary accruals) which avoids the tautological nature of the research design in DeFond and Park. Our results indicate that patterns in measured discretionary accruals and relative earnings performance are consistent with the theory that managers smooth earnings based on both current-year results and expected next-year results. The prevalence and magnitude of this smoothing behavior, however, is considerably lower than inferred in DeFond and Park (1997). Key words: Earnings management; Smoothing