This thesis focuses on the treatment of unobserved heterogeneity and unobservables in economic models. More precisely we centre our attention to the effect of these two issues on the field production economics. In this framework it is hardly ever assumed that all individuals included in the studied sample use the same technology. However, if some individuals use different technologies estimating a common technology might yield biased parameter estimates. To control for differences in technological characteristics among individuals, two-stage models have been often used. However, several authors have recently advocated the implementation of one-stage models such as the so-called latent class models. Another source of potential bias arises when the researcher try to model two different economic phenomena which have similar effects on available data. The motivation for this thesis arises from the fact that the two phenomena presented above are not observable to the researcher, and they have to be identified from the data using specific empirical models. Thus, in this thesis we try to shed light on how alternative proposed models are able to deal with these issues. The thesis is composed by three essays and a review of the literature. The first essay compares the results of a latent class model with the use of a priori information to split the sample. The second one analyses the productivity differences between extensive and intensive dairy farms. The last one analyze under which conditions technological catch-up can be disentangled from technical change, in stochastic frontier models. We study this issue using Monte Carlo techniques.