We examine the factors underlying the presence of earnings announcement premia. We find that the premia persist beyond the sample period examined in prior studies (ending in 1988), although they decline in magnitude after 1988. Further, premia are lower on the expected than the actual earnings announcement dates. We document that increases in voluntary disclosures result in lower premia, despite the increase in return volatility over time. Finally, our evidence suggests that the premia are not completely eliminated because of the costs of arbitrage.