Survey-elicited ex ante outputs are used to develop an empirical representation of an Arrow–Debreu–Savage state-contingent technology in an activity-analysis framework. An empirical test of output-cubicality is developed for that framework. We apply those tools to assess production characteristics of a sample of Catalan farmers specialized in arable crops. Results suggest that imposing nonsubstitutability between ex ante outputs results in no significant loss of information. Even though the technology appears to be output cubical, efficiency measurements based on ex post output observations do not appear to adequately represent the stochastic production environment and apparently yield downward biased technical efficiency measures.