HOMELESS IN AMERICA, H OMELESS IN C ALIFORNIAJohn M. Quigley, Steven R aphael, and Eugene Smolensky*Abstract—It is generally believed that the increased incidence of home-lessness in the United States has arisen from broad societal factors, suchas changes in the institutionalizatio n of the mentally ill, increases in drugaddiction and alcohol usage, and so forth. This paper presents a compre-hensive test of the alternate hypothesis that variations in homelessnessarise from changed circumstances in the housing market and in the incomedistribution . We assemble essentiall y all the systematic information avail-able on homelessness in U.S. urban areas: census counts, shelter bedcounts, records of transfer payments, and administrative agency estimates.We estimate similar statistical models using four different samples of dataon the incidence of homelessness , de ned accordin g to very differentcriteria. Our result s suggest that simple economic principles governing theavailability and pricing of housing and the growth in demand for thelowest-qualit y housing explain a large portion of the variation in home-lessness among U.S. metropolitan housing markets. Furthermore, rathermodest improvements in the affordability of renta l housing or its avail-ability can substantiall y reduce the incidence of homelessnes s in theUnited States.I . Introduc tionTHE visibility of street beggars and those sleeping inpublic places increased substantially about two decadesago, a nd the homeless became a substantive political issueat approximately the time of the inauguration of RonaldRea gan in 1981. The  rst of the authoritative counts of thehomeless appeared shortly thereaf ter (Hombs & Snyder,1982), followed by compilations of expert opinion by gov-er nment agencies (U.S. Department of Housing and UrbanDe velopment, 1984), and by estimates pr oduced by univer-sity- based scholars (such as Rossi, 1989), nonpro t researchce nters (Burt & Cohen, 1989), and by th e U.S. Bureau of theCensus ( Taeuber, 1990). T hese various esti mates differsubstantia lly in methodology and de nition, and their inter-pre tation is subject to political manipulatio n as well aslegit imate statistical controversy ( Jencks, 1994; O’Flaherty,1996; Cordray & Pion, 1997). Nevertheless, these estimatesdo reveal a simple fact: the incidence of homelessnessincr eased substantially during the 1980s and has yet todecl ine.N otwithstanding the debate s surrounding enumeration,identif ying the time trend has been considerably m oresucce ssful than uncovering the under lying causes of home-lessne ss and apportioning blame. The lis t of usual suspectsincludes the deinstitutionalization of the mentally ill, thecr ack epidemic of the mid-1980s, and the relatively highcost of low-quality housing. Several prominent social sci-entists (in particular, Jencks (1994)) have downplayed therole of housing affordability, placing greater emphasis ondeinstit utionalization and the ravaging consequences of in-cr eased drug usage.1H owever, there are reasons to question these conven-tional explanations. The introduction of crack cocaine sub-stanti ally reduc ed the cost of getting high, thus inducingoffsetting income and price effects on housing consumption.M oreover, the onset of the crack epidemic is often dated tothe mid-1980s, nearly  ve years after noticeable increases inhomelessness (Reuter, MacCoun, & Murphy, 1990). Thedecl ine in mental hospital populations has been largelyoffset by increases in the numbers of the mentally ill whoare con  ned in other institutional se ttings. Hence, the num-ber of mentally ill who are “institutionalized,” broadl yspeaki ng, may not exhibit much of a trend. O’Flaherty(1996) refocuses the deba te on housing costs, offering am odel of urban housing markets that, when combined withthe well-documented increase in income inequality duringthe 1980s (Reed et al., 1996), points to an increase in theincide nce of homelessness.T he arguments regarding the relative importance of thevar ious determinants of homelessness a re indirect. Theyre ly on the established fact that homelessness has incr eased,and they make indire ct infere nces from trends in potentialca usal factors. In this paper, we analyze directly the deter-m inants of homelessness using essentially all the availablesystem atic survey information about variations in the inci-dence of homelessness across U.S. housing markets. Weanal yze two national cross-sectional data sets in which theunit of observation is the city or the metropolitan area andtwo county-level data sets for a single large state, Califor-nia. Three of these data sets correspond closely wit h whatone might label complete counts of the “of cial homeless.”The fourth, for California, measures homelessness amongAF DC-eligible families. This latter data set is particularlyusef ul in that it provides observations on county-levelca seloads over an eight-year period; we are thus able to usestanda rd panel techniques to address the unobserved heter-ogeneity which is not easily captured in simpler statisticalm odels.2Received for publicatio n August 6, 1999. Revision accepted for publi-cation March 30, 2000.* University of California, Berkeley.A previous version of this paper was presented at the Annual Meetingsof the American Economic Association, New York, January 1999, thePublic Policy Institute of California, and elsewhere. We are grateful toJohn Landis , Suzanne O’Keefe, and Michael Smith-Heimer for assistancein obtaining primary data, and to Erin Mansur and Larry Rosenthal forresearch assistance . We are als o grateful to David Levine, Harriet New-burger, Brendan O’Flaherty, and two anonymous referees for extensivecomments. This work is supported by the Public Policy Institute ofCalifornia and by the Berkeley Program on Housing and Urban Policy.The raw data used in this study may be downloaded at http:// urbanpolic y.berkeley.edu.1A similar emphasis—and the rejection of housing market explana-tions— are even more apparent in the European literature on homeless-ness. See Fitzpatrick (1998) for a survey.2Another homelessness measure for U.S. cities is a 1984 survey “basedon the opinions of local of cials involved with the problem of homeless-ness.” (U.S. Department of Housing and Urban Development, 1984). Thissurvey has been heavily criticized on methodological grounds (Rossi,1989; Early & Olsen, 1998) and has formed the basis fo r politicallyinspired