We adapt the classical SUR procedure to a minimum cross entropy approach to estimate linear systems of equations where the errors across equations are correlated. We conclude that our entropy-based approach may provide a reasonable substitute for SUR in cases where classical methods may not be applied due to shortages of data.
