High-impact ?oods have become a virtually annual experience in Malaysia, yet ?ood insurance has remained a grossly neglected part of comprehensive integrated ?ood risk management. Using discriminant analysis, this study seeks to identify the demand-side variables that best predict ?ood insurance purchase and risk aversion between two groups of residential homeowners in three districts of Johor State, Malaysia: those who purchased ?ood insurance and those who did not. Our results revealed an overall 34% purchase rate, with Kota Tinggi district having the highest (44%) and thus the highest degree of ?ood risk aversion. The Wilks’ lambda F test for equality of group means, standardised discriminant function coef?cients, structure correlation, and canonical correlation has clearly shown that there are strong signi?cant attribute differences between the two groups of homeowners, based on the measures of objective ?ood risk exposure, subjective risk perception, and socioeconomic cum demographic variables. However, the measures of subjective risk perception were found to be more predictive of ?ood insurance purchase and ?ood risk aversion.