We use a framed field experiment considering hypothetical stocking rate decisions made by grazing enterprise managers and estimate non‐linear multinomial logit models for a range of nested non‐expected utility and expected utility models. The risk and decision‐bias parameters for five models estimated for individual responses are shown to be significantly related to land condition but in ways which suggest behavioural aspects of decision making are critical in understanding land management and stocking rate decisions. Our results show that individual heterogeneity in decision making amongst farming groups is likely to be a significant source of variation in farming intensity and technology adoption decisions. This heterogeneity does not appear to be a reflection of socio‐demographic characteristics. Furthermore, decision functions appear to be biased toward selection of simpler representative functions (e.g. Expected Utility) for sample averages. This suggests that experimental findings that Expected Utility is representative for actual decisions may be due to sample averaging rather than reflect actual behaviour.