Counting the Unsheltered Homeless in Edmonton Using Plant-Capture Studies from Toronto

Counting the Unsheltered Homeless in Edmonton Using Plant-Capture Studies from Toronto

Homeless counts are common in many cities and are regarded as an important basis for civic planning. In Edmonton, they are conducted every two years and include a street count of unsheltered homeless individuals. This is a point-in-time count over the course of a single day in October. Volunteer enumerators are instructed to approach individuals on predetermined walking routes throughout the city where the homeless are known to congregate. The total count is then reported as an estimate of the size of the unsheltered homeless population. However, interval estimates (e.g. Bayesian 95% credible intervals) are not available for any of the homeless counts. Thus we cannot judge the accuracy of the estimation. The reason is that there are no data available in order to evaluate completeness of the sampling frame and estimate the probability of selection in order to calculate interval estimates. In this article, we propose a Bayesian solution that combines the homeless count in Edmonton with plant-captures studies of homelessness from Toronto. We model the probabilites of selection in Edmonton and Toronto as exchangeable in a hierarchical Bayesian model. They are treated as a random sample from a Beta distribution with unknown hyperparameters. Markov chain Monte Carlo is used to sample from the posterior distribution of model parameters. This allows us to estimate the size of the unsheltered homeless population in Edmonton during each enumeration year. We provide guidelines for applying the method to assess the accuracy of homeless counts in other cities.

ORGANIZATION: Faculty of Health Sciences, Simon Fraser University, Canada
PUBLICATION DATE: 2012
LOCATION: Edmonton, AB, Canada