Abstract: As an alternative to the quantile-based QVaR, the risk measure EVaR based on the Expectile model is simpler to calculate and can more accurately reflect the effects of extreme values. In order to make full use of the information contained in mixed frequency data, a MIDAS-Expectile regression model was constructed, and the estimation of the parameters and conditional EVaR were obtained based on the nonlinear asymmetric least squares method. The asymptotic normality of the estimates and coverage test for conditional Expectile were also given. In addition, the likelihood function and information criterion of the Expectile regression model were given from the perspective of maximum likelihood estimation, which could compare and test different models. In order to study the financial risks of cryptocurrencies, in the empirical part, the MIDAS-Expectile regression model was applied to the measurement of cryptocurrency returns risk, and the risk contagion of other tradition financial markets to this emerging financial asset was discussed. The empirical results of the risk of cryptocurrency monthly data indicate that signals from other financial markets will have a significant or positive or negative impact on the risks of the cryptocurrency market, and that the cryptocurrency market is not isolated from traditional financial markets.

Key words: MIDAS-Expectile regression model, EVaR, cryptocurrency, nonlinear asymmetric least squares, maximum likelihood