In this paper we address the question "what is the variance of long horizons portfolios?" The work is directly motivated by ``Are Stocks Really Less Volatile in the Long Run'' (2011) by Pastor and Stambaugh. In that work, they use their "Predictive Systems" framework to challenge the conventional wisdom that stocks are less volatile over long horizons when compared to short horizons. Their conclusion is reached by the incorporation of parameter uncertainty and ``imperfect'' predictors. Preserving the economic motivation of their approach, we develop parsimonious alternatives to "Predictive Systems" and show that, when compared to the correct benchmark, stocks can still be appealing for long run portfolios. Central to our results is a careful assessment of the priors investors have about key quantities underlying market behavior. In particular we explore the impact of priors that enforce the correlation between expected and realized returns to be negative in a time-varying volatility context.