We investigate price duration based variance estimators that have long been ignored in the literature. We show i) how price duration based estimators can be used for the estimation and forecasting of the integrated variance of an underlying semi-martingale price process and ii) how they are affected by a) important market microstructure noise components such as the bid/ask spread, irregularly spaced observations in discrete time and discrete price levels, as well as b) price jumps. We develop i) a simple-to-construct non-parametric estimator and ii) a parametric price duration estimator using autoregressive conditional duration specifications. We provide guidance how these estimators can best be implemented in practice by optimally selecting a threshold parameter that defines a price duration event. We provide simulation evidence that price duration estimators give lower RMSE's than competing estimators and forecasting evidence that they extract relevant information from high-frequency data better and produce more accurate forecasts than competing realised volatility and option-implied variance estimators, when considered in isolation or as part of a forecasting combination setting.