Previous research on building pre-tender cost estimating stresses the importance of giving accurate estimates and minimizing estimating errors. Cost models, especially those mathematical models using mean square error or the like for model training and validation, often treat positive errors (overestimates) and negative errors (underestimates) of equal magnitude the same with an implicit assumption that the regret or disutility of positive errors (overestimates) is equal to that of negative errors (underestimates). A survey was conducted in Hong Kong to study estimating practice and in particular, the attitude of clients and estimators towards estimating errors. This involved the use of regression analysis to model the relative disutility of underestimates (in terms of overestimates) for four different building types. Both clients and estimators are found to be risk averse - tolerating overestimates more than underestimates - and, arguably, clients are satisfied with overestimates. However, they have contrasting views on the desired characteristics of estimates with clients considering the ability to identify cost sensitive elements to be more important than accuracy. In this regard, the formalization of value analysis as part of the cost advice function under typical cost consulting agreements should be the best form of improvement.