This paper analyses the differences between forward (risk-neutral expectation) and realized variance. Four different assets are examined: S&P 500, FTSE 100, eurodollar and short sterling futures. Model-free measures of risk-neutral forward variance are estimated using the full cross section of option prices of a given maturity. Intra-day futures data are used in the estimation of realized variance. A mean squared error criterion is devised to inform the choice of optimal intra-day frequency. Market microstructure issues are considered for both equity and interest rate futures markets. The bias and efficiency of forward variance as a predictor of realized variance is reexamined.