This document proposes using a family of nested power EWMA estimators based on an exponential power distribution to estimate VaR. It summarizes previous research showing that asset returns are often fat-tailed rather than normally distributed. The power EWMA model nests standard EWMA, robust EWMA, and power EWMA estimators. The document empirically evaluates the conservativeness, accuracy, and efficiency of these models when estimating the VaR of several stock market indices, using measures that address each criterion. The results show that power EWMA and related estimators offer superior coverage of extreme risks compared to standard EWMA, and accurate VaR estimation.