How to House the Homeless

7. Homelessness as Bad Luck: Implications for Research and Policy

143 Chapter 7 Homelessness as Bad Luck: Implications for Research and Policy BRENDAN O’FLAHERTY Sometimes bad things happen to people. They lose their jobs; companions walk out on them; their health—physical or mental—deteriorates; they get evicted; prices of goods they rely on rise; they lose their benefits. Sometimes they are blameless in these calamities; sometimes they are not. Good things happen to people, too. Stochastic processes are a major consideration in the study of homelessness . Individual narratives of how people become homeless emphasize bad luck, and good luck often figures in how people leave homelessness. More objective studies support this subjective view. Predicting who will become homeless on the basis of observable characteristics is extremely difficult , and even the most thorough studies end up with large errors of both type 1 and type 2. Unobservable characteristics are therefore hugely significant in what leads to homelessness. These characteristics are either persistent or transient. If they were persistent, spells of homelessness would be very long—years or decades long. But most spells of homelessness are measured in days or months, not decades. Homelessness is not an indelible characteristic like a birthmark or a Social Security number. Almost everyone who will be homeless two years from today is housed now, and almost everyone who is homeless today will be housed two years from now. Homeless spells are more like semesters than careers. Some homeless spells are many years long, but these are rare. What is important about these spells is that at their starts they are unpredictable. Bad luck can be decades long as well as days long. Transient unobservable characteristics, then, are responsible for a large proportion of homelessness. Luck is the colloquial term for transient unobservable characteristics (and it is appropriate to use this term even when the probability of becoming homeless is not the same for every person at every time1 ). Looking at homelessness as part of the stochastic variation in people’s lives has many implications; this chapter elucidates a few of them. I show that transitions to homelessness are inherently unpredictable; services cannot be well targeted at people likely to become homeless, no matter how diligently and resourcefully researchers and service providers try. Current housing conditions are the best predictor of future homelessness, although housing market shocks are probably not the major immediate precursor to homelessness. Reducing real income volatility may be more effective in reducing homelessness than raising people’s average level of well-being. Smoothing may work better than uplifting. In paying attention to how individuals experience homelessness, I am not implying that the environment in which they find themselves is unimportant or unchanging. Changes in that environment imply changes in individual probabilities of experiencing homelessness and so changes in the aggregate volume of homelessness. How environment affects aggregate homelessness has been studied extensively (for instance, see Raphael, this volume). The question this chapter addresses is how, in any given environment, homelessness can be prevented or reduced. It is not a question of why so many fires occur, but rather how the fire trucks should be deployed. Both questions are worth studying. I do not examine what sort of shocks poor people face and what sort of shocks precipitate homelessness (see O’Flaherty 2009a). The chapter also examines institutions and programs that might reduce volatility. Reforms in the private housing market do not appear promising, with the possible exception of shared-equity mortgages, partly because housing price shocks are not a major problem. Subsidized housing programs are good at reducing volatility among their recipients but are not designed as insurance programs for the wider community. Aid does not generally flow to those who need it most when they need it most. Shelters, on the other hand, work well as part of the safety net and would work better if they applied some lessons from the theory of optimal unemployment insurance. Post-shelter subsidies should be widespread rather than targeted and most easily available early in shelter spells. The chapter also looks at the nascent field of homelessness prevention. Why Does Risk Matter? Risk matters because homelessness is a transient state for which onset cannot easily be predicted in advance. If some people were fated to be homeless and stay homeless for most of their lives, or if one could anticipate well in advance who would become homeless and when, then analysts 144 How to House the Homeless could ignore risk. Abundant...