The approach to analysis of and interpretation of findings from the Beck Depression Inventory (BDI), a self-report questionnaire, depends on sample characteristics. To extend work using conventional BDI scoring, the BDI's suitability in assessing symptom severity in a homeless and precariously housed sample was examined using Rasch analysis. Participants (n=478) recruited from an impoverished neighbourhood in Vancouver, Canada, completed the BDI. Rasch analysis using the partial credit model was done, and the structural validity, unidimensionality, and reliability of the BDI were studied. A receiver operating characteristic curve determined a Rasch cut-off score consistent with clinical depression, and Rasch scores were correlated with raw scores. Good fit to the Rasch model was observed after rescoring all items and removing Item 19 (Weight Loss), and unidimensionality and reliability were satisfactory. Item 9 (Suicidal Wishes) represented the most severe symptom. Rasch-based scores detected clinical depression with moderate sensitivity and specificity, and were positively correlated with conventional scores. The BDI in a community-based sample of homeless and precariously housed adults satisfied Rasch model expectations in a 20-item format, and is suitable for assessing symptom severity. Future research on depression in similar samples may reveal more information on using specific symptoms to determine clinical significance.