We provide strong support for the underappreciated expected earnings hypothesis of negative correlation between aggregate stock returns and earnings (Sadka and Sadka (2009); Choi, Kalay, and Sadka (2016)). For the 1970 to 2000 period studied by Kothari, Lewellen, and Warner (2006), our powerful modeling strategy incorporating macroeconomic information reveals that aggregate returns are significantly negatively correlated with expected aggregate earnings changes but uncorrelated with unexpected aggregate earnings changes. However, this negative correlation changes after 2000, perhaps from heightened volatility or accounting changes. We also show that underlying macroeconomic information explains the power of aggregate earnings to predict future GDP growth.