In this paper we examine two different methodes for dealing with heterogeity in an unbalanced panel. Focus is on testing efficiency on the fixed and random coefficient specification, and comparing returns to scale, technical change and price elaticities between the models. The empirical results, based on data from Norwegian paper and pulp industry, indicates fixed coefficient specification to be more efficient. However, we cannot in our study se any significant differences in our estimated results, i.e. elasticities. But if we don't take heterogeneity into account, the estimates may not be consistent.