Constant-mean model

For $R_{i\tau}=\mu_i+\varepsilon_{i\tau}$ estimated on $L_1$ days: (a) derive $\operatorname{Var}(CAR_i)=\sigma^2(L_2+L_2^2/L_1)$; (b) show that the event-window ARs are equicorrelated with correlation $1/(L_1+1)$; (c) compute the variance inflation from estimation error for $L_1=240$ and $L_2=21$, a one-month window. Is it negligible?