It is well-known to people working in homeless serving sectors that the causes of homelessness are many and complex. Data used by homeless serving sectors, data such as the number of people making use of homeless shelters, is used by the sector and by governments to measure progress and to evaluate policy changes. But the complexity of homelessness means these data need to be carefully interpreted. The data are the product of many data generating processes some of which are offsetting and some of which are additive at different points in time. Decomposing a time series into data generated by different types of homelessness is important for a proper understanding and evaluation of policy interventions. In this paper we introduce and apply to the analysis of homeless shelter data a method of time series decomposition frequently used in the study of the business cycle, namely, the Hodrick-Prescott (HP) filter. We chose to focus on the HP filter because it can be easily used by non-experts in an Excel workbook and because it has been found to produce data that analysts recognize as “reasonable” estimates of data describing slowly changing trends and more frequently changing cycles. We provide interpretations of these trends and cycles – what we refer to as structural and frictional homelessness – that we believe correspond with the homeless serving sector’s understanding of the sources of homelessness.
Ronald Kneebone (Fri,) studied this question.