Generally an unbiased statistic is preferred over a biased statistic. This is because there is a long run tendency of the biased statistic to under/over estimate the true value of the population parameter. Unbiasedness does not guarantee that an estimator will be close to the population parameter.
