We add to the debate about whether models of earnings dynamics should allow for unobservable heterogeneity in expected earnings growth rates by examining implications for a statistic originally proposed to estimate the variance of persistent earnings shocks. While that statistic is unbiased under a common specification, we derive biases that would arise under alternative models and use them to draw inferences about their empirical relevance and to estimate key parameters. Most results cast doubt on substantial heterogeneity in growth rates, though some leave room for a modest role. Estimates of shocks' variance and persistence are more robust.