It has been widely considered that when faced with natural hazard risks in the future, people adopt precautionary measures in order to alleviate the impact of a hazardous event. This study used the data from the Household Survey of Post-Morakot Social Impact and Recovery-Wave 1 implemented in June, 2010 in Taiwan via face to face interview with the representatives of the households that were forced to relocate after the typhoon. The raw data contains 1658 observations, representing 1658 households. Based on a two-stage approach, we, in the first stage, investigate the determinants of three types of households’ risk perceptions for the future, respectively, with the following explanatory variables: one’s experience with disaster with damage incurred, one’s trust in the authorities as well as local communities regarding their capacity of emergency response, one’s socio-demographic backgrounds and one’s residential areas. In the second stage, we assess the power of previously investigated risk perceptions, as well as of other factors, in explaining households’ intention to adopt measures for preparedness, for mitigation and for recovery, respectively. Subject to the categorical characteristics of dependent variables, an ordered probit model was employed. Our estimation results confirm the association between precautionary behaviour taken before Morakot and households risk perceptions after typhoon Morakot. However, it is inconclusive regarding whether or not former actions could reduce risk perceptions in a later stage. A negative correlation between trust in the central government and perceived impact of property loss is observed which suggests that the central government’s emergency response takes effect on reducing the damage on property. In comparison, households with higher trust in local communities tend to have lower perceived likelihood of a typhoon disaster. The corresponding predicted values confirmed that majority of the sampled households have high or very high risk perceptions. Households’ socio-demographic background and the locations of their residency are related to risk perceptions, primarily perceived probability and impact on property. The extent to which one trusts the central government and local communities in their disaster response capacity explains his/her intention to take precautionary actions. These results imply that households’ trust in the central government indicates their dependency on the government and hence results in weaker intention to take self-protect actions. Furthermore, households who trust in local communities do not consider that it is the communities’ responsibility to take protection measures and being part of the communities, they recognise the necessity of precautionary actions at the community level. Furthermore, we found that excessive compensation is correlated with weaker intention to purchase property insurance but a similar effect is not observed with respect to the intention to take up insurance against personal accidents. This can be explained by the fact that the authorities offered to the households affected by typhoon Morakot a financial compensation package that had extensive coverage on flooded or damaged housing and this can disincentivize households to purchase property insurance. Finally, the results confirm that certain predictors have both direct and indirect effects on the intentions to take certain precautionary measures, when indirect effects are mediated by risk perceptions. These predictors include trust in the central government for preparedness and mitigation, trust in local communicates for preparedness, mitigation and recovery, age for preparedness measures, income and illiteracy for mitigation measures and ethnicity for recovery measures. Moreover, the direct and indirect effects in some cases can counteract with each other. Thus if risk communication is to be sought in order to promote households’ precautionary behaviour against typhoon hazards, not only the information about the possibility and potential impact associated with a hazard but also households’ attitudes and socio-demographic factors ought to be taken into account in the development of communication strategies.