Many studies document loss aversion in the housing market, where expected losses lead to higher sales prices. However, exposure to expected losses may correlate with unobservables that influence housing prices. Under the assumption that multiple psychological biases appear together, we estimate loss aversion by identifying sellers who appear psychologically biased by exhibiting focal point or round number bias in their choice of mortgage amount at purchase. Using both difference-in-differences and regression discontinuity approaches, we find evidence of loss aversion on sales prices based on a stronger correlation between loss and sales price for our subsample of sellers who exhibited round number bias, but our estimated effects are substantially smaller than the results that arise from directly estimating the effects of expected loss on the sales price. In addition, lumpy sellers are less likely to sell relative to the control group. We show that expected loss correlates strongly with predetermined mortgage, housing unit, and census tract attributes, but the interaction between a round mortgage amount and expected loss exhibits far fewer failures of balance. Further, the magnitude of the sample-wide relationship between expected loss and sales price is eroded substantially by the inclusion of balancing test controls, as well as by the inclusion of a running variable for the mortgage amount, while the magnitude of the relative estimates for the round mortgage amount subsample is quite stable. Evidence from an earlier experiement showing a positive relationship between reporting round numbers and loss aversioin provide supports for our identification strategy.