In many statutory valuation jurisdictions, the penalty for over-valuation is greater than the penalty for under-valuation. This follows the principle that tax payers should not knowingly be over-charged when applying ad valorum taxes.In NSW, the main standard for valuation accuracy is framed in terms of the mean value price ratio (MVP). The prior standard for the MVP has been a range of 85% to 100%. That is, a reasonable level of under-valuation is tolerated but not so any level of over-valuation (MVP above 100%). At present, this standard is being reviewed with a suggested range of 90% to 100%.Clearly, a standard for the accuracy of valuations is impacted on by the variation in valuations. This raises a number of considerations. What should be the target level of valuation accuracy in an asymmetric penalty environment? What is an appropriate standard for valuation accuracy and how is impacted by different levels of valuation variation (as will occur for different classes of property)?These issues will be considered in the paper. Penalised least squares and Bayesian approaches will be considered.These ideas may also be applicable in other property applications, such as for investment allocation decisions, where under-performance may be penalised much higher than over-performance.