The literature documents a significantly negative average variance swap payoff (VSP) for the S&P 500 index but generally not for the constituent stocks. We show that this result is affected by biases arising from (i) an intraday momentum effect and (ii) the use of an incoherent measure of return variation. Accounting for these issues, we find stronger evidence of a significant average VSP both at the index level and also for equities. We decompose the index variance risk premium (VRP) into factors related to the VRP of equities and the correlation risk premium (CRP) and assess their predictive power for aggregate stock returns.