This paper concerns itself with the following problem: Suppose the true production function is of the CES type with constant returns to scalle. If we fit an unrestricted Cobb-Douglas production function instead, what is the nature of the bias in the estimate of the returns to scale parameter?