The study looked at how stock returns vary over time, finding that this variation is more common when certain conditions are met. By using a specific statistical model, the researchers were able to estimate stock betas more accurately, improving the efficiency of predicting stock returns.