Economic Perspectives on Homelessness

Economic Perspectives on Homelessness

Economic Perspectives on Homelessness B O’Flaherty, Columbia University, New York, NY, USA ª 2012 Elsevier Ltd. All rights reserved. Glossary Adverse...

107KB Sizes 1 Downloads 127 Views

Economic Perspectives on Homelessness B O’Flaherty, Columbia University, New York, NY, USA ª 2012 Elsevier Ltd. All rights reserved.

Glossary Adverse selection A situation that exists when one party’s willingness to engage in a transaction is information to the other party that the first party is more likely to have unobserved characteristics that make the transaction disadvantageous to the second party. For example, willingness to buy individual health insurance conveys information about the likelihood of being in poor health. The classic statement of adverse selection is due to Groucho Marx: ‘I would not join any club that was willing to have me as a member.’ Filtering Models of the housing market in which houses age and are passed down to different groups. Homeless In official US usage, a person is homeless tonight if he or she is sleeping in a shelter or in a place not designed for human accommodation, such as a transportation terminal or a lean-to. HUD In the United States, the Federal Department of Housing and Urban Development. Microdata Data in which the units of observation are individuals or households.

Introduction The economics profession came to the study of home­ lessness with one great advantage: it had been thinking about unemployment for many decades. The Great Depression and its attendant explosion of unemployment are the central drama of the twentieth century for most Western economists, just as the World Wars are the central dramas for students of diplomacy. For economists, therefore, it has been natural to think about the lack of a place to live the same way we think about the lack of a place to work. This predisposition has spared economics from most of the discussions in other disciplines about structural versus individual explanations of homelessness. Unemployed people have worse human capital characteristics than employed people, but changes in macroeconomic condi­ tions, not in human capital characteristics, explain most of the fluctuations in unemployment. Thus economists did not find debates about structural or individual explanations of homelessness new or enlightening. The analogy has also meant that economists have concentrated on aggregates

HOME/HOMELESSNESS

Moral hazard A situation that exists when contracts or programme provisions create incentives for one party to take unobservable or unverifiable actions that are detrimental to the interest of the other party or of the programme operator. For example, banks that are assured of a government bailout are more likely to make risky loans. Ordinary least squares (OLS) A statistical procedure in which a regression equation is fit to minimise the unweighted sum of squared residuals. Point-in-time (PIT) count The number of people who are homeless on a particular night. This is distinguished from counts that ask how many people have experienced at least one night of homelessness in a longer time period, such as a year. Rent control Legal limitations on rent increases between leases. Shelter A facility in which people who would otherwise be homeless live, without tenancy rights or an expectation of a long stay. Often, no rent is charged.

and on explanations, rather than the nitty-gritty of helping individual people, or even evaluating the efficacy of spe­ cific programmes. But that is changing. This article reviews the work that economists have done on modern homelessness – the wave of home­ lessness that began in the United States around 1980. It will concentrate on the United States, because almost all economic work on homelessness is about the United States (we exclude work on informal settlements like favelas in developing countries). The approach is broadly chronological but does follow themes as they develop. The first section is about how housing policies affect homelessness. The first papers about homelessness addressed this question, as has some recent work. However, long periods have passed with very little work on this topic. The second section is about what deter­ mines the volume of homelessness and why homelessness rose in the 1980s. The third section concentrates on policies explicitly targeted to homeless people and people at risk of becoming homeless. This area has burgeoned recently.

37

38

Economic Perspectives on Homelessness

How Do Housing Policies Affect Homelessness? Rent Control The earliest writings by economists were triggered by a series of popular articles starting in 1987 that claimed that rent control caused homelessness. The responses used the same data set that the popular articles did, a cross section of ‘expert estimates’ compiled by HUD in 1984, but ran ordinary least squares (OLS) regressions with a few obvious controls aside from rent control. The result vanished with this simple step. The rent control issue was revisited almost a decade later with better data sets and more sophisticated econo­ metrics – but still with a cross section of cities. Two papers (Grimes and Chressanthis, 1997; Troutman et al., 1999) found that rent control increases homelessness in a statistically significant way, but one concluded that the effect was not economically significant. Both papers used 1990 census data and tried to account for endogeneity. Neither, however, had a convincing approach to this problem and neither tried to account for the effect of rent control on rents. A final paper (Early and Olsen, 1999) used microdata and modelled the effects of rent control on rents and on vacancies. It found that rent control may reduce home­ lessness but the effect is not statistically significant. However, using microdata biases down the estimated size of the effects of aggregate variables; thus, the conclu­ sion appears to be that rent control has little effect on homelessness.

that switching 100 subsidies to extremely poor recipients from recipients who are not so poor would reduce PIT homelessness by about 7.7%. In the early 2000s, HUD conducted a controlled experiment where some randomly selected welfare families received housing choice vouchers (often called section 8 vouchers) and some did not. Housing choice vouchers essentially eliminated homelessness among recipients (Mills et al., 2006). The experience of the con­ trol group leads to the inference that for every 100 welfare families receiving vouchers, PIT homelessness falls by about 3.5 families, if the additional vouchers have no further housing market effects. Finally, some economists (Mansur et al., 2002) have used simulations, rather than trying to find new estimates of structural parameters. One paper simulates the effects of different housing subsidy policies in the four largest California metropolitan areas. The most interesting pol­ icy is a universal housing voucher for poor renters that is considerably less generous than housing choice vouchers. The policy causes major reductions in homelessness. However, the reduction in PIT homelessness for each 100 households assisted is about the same as that estimated in the rest of the literature: 2.2 per hundred in the San Francisco metropolitan area, for instance. The similarity between results in the subsection should not be surprising. On any night, the proportion of people in practically any conventionally defined group of poor people is pretty small (provided the definition does not refer directly to housing consumption). Providing housing subsidies to any such group will cause at most a commen­ surate reduction in PIT homelessness.

Housing Subsidies Compared with rent control, surprisingly few papers examine the effect of low-income housing subsidy pro­ grammes on homelessness. The first paper (Early, 1999) to do so appeared in 1999. It used microdata to compare the characteristics of homeless people with those living in subsidised housing. The two sets of people have distinctly different characteristics. This paper concludes that 3.8–5% of subsidised households would be homeless if they were not receiving subsidies. If marginal subsidised households are like average subsidised households, expanding subsidies by 100 would reduce point-in-time (PIT) homelessness by 3.8–5%. A later paper (Early and Olsen, 2002) looks at a cross section of metropolitan areas and regresses PIT home­ lessness on the availability of housing subsidies, the extent to which subsidies are targeted to extremely poor people, and a variety of control variables. The paper does not test for the endogeneity of subsidies. The number of housing subsidies has no effect on homelessness, but better target­ ing reduces homelessness. The paper’s coefficients imply

Regulation A recent paper (Raphael, 2010) looks at the effect of housing market regulation on homelessness. Housing market economists have repeatedly shown that more stringent regulation is associated with higher housing prices, especially for poor people, and the work we discuss below repeatedly shows that higher rents, especially for poor people, are associated with more homelessness. The paper collapses the steps and shows that more stringent regulation is associated with more homelessness. The measure of regulatory stringency in this paper primarily gauges how hard it is to construct new housing for middle-income people. No empirical work has been done yet on the effect of regulations that focus on housing for low-income people – for instance, regulations on minimum room size, parking requirements, maintenance codes, window protection, sharing by unrelated persons, boarding homes, and lodging houses.

Economic Perspectives on Homelessness

What Determines How Many People Are Homeless? Economists have been trying to understand what deter­ mines the number of homeless people almost as long as they have been trying to assess the effects of housing policies, because evaluating policies is hard unless you know what conditions to control for, and designing poli­ cies is hard unless you know what causes the problem you are trying to alleviate. This work has been both theore­ tical and empirical.

Theoretical Only one complete theoretical model of housing and homelessness (O’Flaherty, 1996) has been published. It is a ‘filtering model’: there are many (a continuum) qualities of housing, and better housing gets worse with age, if it is not maintained. Housing is built for rich and middle-income people, then (possibly) filters down to poorer people, and may eventually be aban­ doned. Rates of construction, maintenance, filtering, and abandonment are endogenous. The lowest possible qualities of housing may not be available on the market because of maintenance costs, regulation, or the value of vacant land. People are homeless if they are better off spending whatever income they have on nonhousing goods, rather than on the lowest quality of housing on the market. Increases in homelessness can then occur either because of changes in the housing market that raise the price of the lowest available quality or because of changes in the income distribution that raise the number of people with income below the threshold for purchasing housing. The model illuminates what happened in the housing markets between 1970 and 1990 to explain why homelessness rose. Rents for low-quality housing rose dra­ matically, and the rate of abandonment fell. The quantity of low-quality housing also probably fell. The most plausible explanation for these trends is rising income inequality, particularly a shrinking middle class. If the primary reason why homelessness rose, for instance, were greater poverty or more substance abuse or more mentally ill people outside of institutions, the housing market would not have reacted like this: in all cases we would have seen more low-quality housing, not less, and except in the case of more mentally ill people, lower rents for that housing. Rising income inequality, on the other hand, produces the phenomena we observed. The intuition is that houses for low-income people are hand-me-downs from middleincome people. A relative reduction in the size of the middle-income population tightens the supply of housing for low-income people. Tighter supply manifests itself in

39

higher prices and lower quantities – and in some cases, people on the street. Research has also shown directly that the number of substance abusers or of noninstitutionalised prime-age adults with serious mental illness did not rise substan­ tially, if at all, at the time that homelessness rose. Empirical This theory was tested soon after it appeared. Research found that homelessness was more closely associated with inequality than with poverty, as predicted. Aside from theory-driven work, a large number of papers in the 1990s tried to find the determinants of metropolitan-area homelessness. They ran regressions on cross sections of metropolitan area homeless rates (usually either in 1984 or in 1990) and used characteristics of the metropolitan areas as explanatory variables. Most of these studies find that greater homelessness is related to tighter housing markets (higher rents for poor people and lower vacancy rates) and better weather (higher January temperatures and less precipitation). Demographic characteristics are almost uniformly insig­ nificant. The proportion of the population that is male, minority, poor, single, mentally ill, or substance-abusing rarely matters, even though all of these characteristics on an individual level are associated with greater homelessness. A small number of studies have also looked at time series of shelter population in a single city. All these papers find seasonal effects for single adult shelters, but some papers find little else. Papers on the New York City shelters (O’Flaherty and Wu, 2006, 2008) find moderate macroeconomic effects, particularly for families, but little impact of rents. This may be in part a data problem, because only aggregate rents are available in the time series – rents for poor people cannot be broken out – and the rate of rent change in New York City shows very little variation during the periods studied. Demographic factors like prison, jail, and mental hospital populations and releases do not affect single-adult shelter population significantly. Why do demographic characteristics make so little difference to aggregate PIT homelessness? Essentially, because the proportion of any demographic group that is homeless at any time is small. For instance, about 0.7% of poor people in an average metropolitan area were homeless in March 1990, and around 2% of the severely mentally ill people were homeless on an average night in 2000. A simple cross-section model can derive the impli­ cation of these small proportions. Suppose that cities differ in the proportion of the population that is poor (or severely mentally ill) and in the price of housing. People with income below P are poor. People are

40

Economic Perspectives on Homelessness

homeless in city c when their incomes are below H(c), which varies between cities because the price of hous­ ing varies. Variations in the proportion between P and H(c) affect poverty population but not homelessness. Variation in housing prices affects homelessness but not poverty population; so will variation in the propor­ tion of poverty population below H(c). Since almost all the poverty population is between P and H(c), almost none of the variation in poverty population is asso­ ciated with variation in homelessness, but variation in housing prices is associated with variation in homelessness.

How Do Policies Targeted at Homelessness Work? The final type of work that economists have done has looked at the effects of programmes and policies explicitly directed at homeless and near-homeless people.

placement into subsidised housing can be a tool – albeit expensive and imperfect – that administrators can use to reduce shelter population. These and other New York City shelter papers also examine the effect of different kinds of shelters and of shelter capacity. Shelter population seems to be pretty sensitive to what shelter administrators do and somewhat impervious to what goes on in the outside world. This finding, however, may simply reflect the fact that shelter conditions are more accurately measured. A number of empirical studies are also ongoing and should be available in the next several years. These include analyses of homelessness prevention efforts in New York City, Ten-Year-Plans nationally (these are plans that communities adopt at federal urging to ‘end homelessness in ten years’) and federal supportive housing programmes. More studies are likely.

Theoretical Empirical Most of this work has been empirical and has studied the New York City shelter system. This system is large, with over 30 000 people most of the time, and has developed good records. The major controversy that has surrounded this system has been the incentive effect of family placement into subsidised housing. Many families receive subsidised hous­ ing in order to leave the shelters, and sheltered families are more likely to receive housing subsidies than otherwise similar unsheltered families. Critics have charged that these subsidies incentivise families to become homeless, and that the subsidy programme is ultimately self-defeat­ ing. Two surges in shelter population – one in the early 1990s and the other in the early 2000s – are often blamed in the media and even in the minds of some city officials on overly aggressive placement policies. Economists emphasise incentives, and so the claim that incentives affect how people act is not foreign. However, the claim that incentive effects are so strong as to reverse the primary population-reducing effect of placement is very strong and is testable empirically. Two studies (Cragg and O’Flaherty, 1999; O’Flaherty and Wu, 2006), one centred on each incident, estimated the incentive effect. They found that it was present, but not large enough for placements to increase shelter popu­ lation. A hundred placements reduce shelter population by about 30 – not a hundred, as would be the case if there were no incentive effects, not by zero, as would be the case if incentive effects were strong enough to offset placements. Most of the incentive effect operates by delaying exits of already-sheltered families, not by drawing new families into the shelters. Thus

Theoretical work has also examined shelters and pro­ grammes to help homeless people, particularly the timing of placement into subsidised housing. This is a problem that combines moral hazard and adverse selection. Moral hazard argues for early placement: delayed placement gives families an incentive to stay in shelters longer. Adverse selection argues for late placement: why ‘waste’ valuable resources on families who are going to leave the shelter on their own pretty soon and do not really need the help? Similar problems have been studied in the unemployment insurance literature. The solution is an incentive-compatible menu of ‘contracts’: families likely to leave the system soon opt for an early placement programme with no hope of late placement, and families likely not to leave opt for a programme with a significant prospect of late placement. Centralised systems of shelters would have great diffi­ culty implementing this solution, but decentralised systems would not, and in fact, some localities (e.g., Hennepin County, Minnesota) come close to it. The other problem to receive theoretical attention is homelessness prevention. Researchers in other disciplines have found that predicting who will become homeless is very hard but assumed that this problem could be con­ quered if they could get better data and use it more cleverly. Economists have applied consumption theory and argued that prediction is in many cases inherently impossible. People become homeless as a result of unfor­ tunate surprises, and surprises by definition cannot be predicted. Understanding exactly what can be predicted is a necessary step in any workable plan for homelessness prevention.

Economic Perspectives on Homelessness

Conclusion Homelessness is the most visible manifestation of extreme poverty in developed countries. It is hard to maintain that economics is doing its job unless it can explain the misery that everyday working people in those countries see around them and what could be done about it. See also: Cost Analyses of Homelessness: Limits and Opportunities; Economics of Social Housing; Ethnographies of Home and Homelessness; Health, WellBeing and Vulnerable Populations; Homeless People: African Americans in the United States; Homelessness: Prevention in the United States; Mental Health and Homelessness; Policies to Address Homelessness: Housing First Approaches; Policy Instruments that Support Housing Supply: Social Housing; Shelter and Development; Social Psychological Perspectives on Homelessness.

41

Grimes P and Chressanthis G (1997) Assessing the effects of rent control on homelessness. Journal of Urban Economics 41: 23–37. Mansur E, Quigley J, Raphael S, and Smolensky E (2002) Examining policies to reduce homelessness using a general equilibrium model of the housing market. Journal of Urban Economics 52: 316–340. Mills G, Abt Associates, Gubits D, et al. (2006) Effects of Housing Vouchers on Welfare Families. Washington: HUD Office of Policy Development and Research. O’Flaherty B (1996) Making Room: The Economics of Homelessness. Cambridge, MA: Harvard University Press. O’Flaherty B and Wu T (2006) Fewer subsidized exits and a recession: How New York City’s family homeless shelter population became immense. Journal of Housing Economics 15: 99–125. O’Flaherty B and Wu T (2008) Homeless shelters for single adults: Why does their population change? Social Service Review 82: 511–550. Raphael S (2010) Housing market regulation and homelessness. In: Ellen I and O’Flaherty B (eds.) How to House the Homeless. New York: Russell Sage Foundation. Troutman WH, Jackson JD, and Ekelund RB, Jr. (1999) Public policy, perverse incentives, and the homeless problem. Public Choice 98: 195–212.

Further Reading References Cragg M and O’Flaherty B (1999) Do homeless shelter conditions determine shelter population: The case of the Dinkins deluge. Journal of Urban Economics 46: 377–415. Early D (1999) A microeconomic analysis of homelessness: An empirical investigation using choice-based sampling. Journal of Housing Economics 8: 312–327. Early D and Olsen E (1999) Rent control and homelessness. Regional Science and Urban Economics 28: 797–816. Early D and Olsen E (2002) Subsidized housing, emergency shelters, and homelessness: An empirical investigation using data from the 1990 census. Advances in Economic Analysis and Policy 2.

Ellen I and O’Flaherty B (eds.) (2010) How to House the Homeless. New York: Russell Sage. Frank RG and Glied SA (2006) Better but Not Well: Mental Health Policy in the United States Since 1950. Baltimore: Johns Hopkins University Press. O’Flaherty B (2004) Wrong person and wrong place: For homelessness, the conjunction is what matters. Journal of Housing Economics 13: 1–15. Quigley J, Raphael S, and Smolensky E (2001) The Links between Income Inequality, Housing Markets, and Homelessness in California. San Francisco: Public Policy Institute of California. Shinn MB, Weitzmann BC, Stojanovic D, et al. (1998) Predictors of homelessness among families in New York City: From shelter request to housing stability. American Journal of Public Health 88: 1651–1657.