A rather loosely defined expression which in general means, or ought to mean, a bias which is due to the nature of the situation and cannot, for example, be removed by increasing the sample size or choosing a different type of estimator.