Financial data are typically not identically, independently and normally distributed (iid-normal). Yet, standard tests of asset-pricing models are based on this assumption, and we have little information on how sensitive the tests are to violations of iid-normality. Recent evidence suggests that test outcomes may be sensitive to these violations. In this paper, we use Australian data to compare the standard test results with those that do not require iid-normality: the GMM-J test and bootstrap-based tests. We find that different tests produce differences in prob values at least as large as those in US studies but that test outcomes are generally robust.