where ˆ L = L(ˆ θ) is the log-likelihood of the model in Equation (1) and ˆ L0 = L(ˆ θ0) is the log-likelihood of the alternative constant model setting B equal to a matrix of zeros, and ˆ θ and ˆ θ0 are the maximum-likelihood estimates for the two models. The moment persistence measure LR∗V AR(1) might be interpreted similarly to the R2 in a simple linear regression: the more variation a model can explain, relative to a simple mean model benchmark, the better the predictability. In the case of Equation (2) a better predictability implies higher LR∗V AR(1) and, hence, higher persistence.