Lifetime and Five-Year Prevalence
of H-omielessness in the
Bruce G.Link,PhD, Ezra Susser, MD, DrPh, Ann Stueve, PhD, Jo Phelan, PhD,
Robert E. Moore, DrPh, and ElmerStruening,PhD
The number of homeless people in
the United States is of vital importanceto
publichealth.1I The conditions ofhomeless-
nessposesevere problemsfor the control
of infectious diseases such as tuberculosis
and the acquired immunodeficiency syn-
drome andputhomeless peopleat risk of
serious adversephysicaland mental health
Althoughthere is consensus that the
number ofhomelesspeople grewdramati-
cally during the 1980s,4'"
debate hasdeveloped concerningjusthow
largethisgrowing populationhas become.
claimed that the number of homeless
peoplein the United States is ashighas 2
to 3 million.9 However, surveysthat tryto
actually count people who are currently
homeless usually produce much smaller
Because some of these surveys12'13
are based on scientific sampling proce-
dures rather than on assessments of local
informants, theywould appeartoprovide
a better indication of the size of the
homeless population than do the esti-
mates of advocates. However, counting
currently homeless people poses
tremelydifficult problems, problemsthat
some believe have led to severe under-
The first problem is finding people
who are currentlyhomeless. Surveys may
miss the so-called hidden homeless, who
sleep in box cars, on the roofs of tene-
ments, incampgrounds,or in other places
that researchers cannot effectivelysearch.
The second problem is that, once
located, respondents may refuse to be
interviewed or deliberately hide the fact
of key informants
in the nation's largest
for homeless people
that they are homneless.810 In one study,
the interviewers identified enough people
they believed were homeless but who
number of homeless street peopleidenti-
The third problemis missing people
whoexperience relativelyshort or intermit-
tent episodes of literal homelessness.
Such people are less likely to be counted
in 1-night or 1-week surveys; however,
they must be explicitly included in re-
search studies if factors that are initial
causes of homelessness are to be differen-
tiated from factors that cause it to persist.
extrapolating from data collected in one
geographical location or from one sub-
group of the population to other geo-
graphical areas or subpopulations. For
example, Burt and Cohen12 surveyed
homelessness in large cities and then
assumed that rural and suburban areas
had rates that were one third the rate in
urban areas to generate a national esti-
Although systematic surveys of cur-
rent prevalence would seem to provide
the most accurate estimates of the size of
the homeless population, the problems
just enumerated pose serious questions
about the accuracy of the numbers the
surveys provide. We used
it when asked to increase the
is the problem of
The authors are with the Division ofEpidemiol-
ogy,School ofPublic Health, Columbia Univer-
sity, New York, NY. Bruce G. Link, Ezra
Susser, and Elmer Struening are also with the
New York State Psychiatric Institute, New
Requests for reprints should be sent to
Bruce G. Link, PhD, Epidemiologyof Mental
Disorders, 100 Haven Ave, Apartment 31D,
New York, NY 10032.
Thispaperwas accepted August 17, 1994.
Editor's Note. See related editorial by
Rosenheck (p1885)in this issue.
American Journal of Public Health
Link et al.
prevalence and annual incidence from
these data. Second, our finding concern-
ing relatively equal prevalence bycommu-
nity size suggests that the size of the
homeless population outside major urban
areas may have been underestimated.
Current prevalence studies have generally
assumed that homelessness is substan-
tially less prevalent (one third as high)
outside urban areas-an assumption that
our results suggest may be unwarranted.
Third, to make projections to annual
prevalence and incidence, these studies
used data on the duration of homeless-
ness among the people they sampled on a
given night or during a givenweek. People
who are chronically homeless are over-
sampled in such "one-shot" assessments.
As a result, estimates of average duration
are biased upwards and estimates of
annual prevalence and incidence are
biased downwards. Under this scenario,
even if point-prevalence rates reported in
past studies are accurate, projections to
the rate at which our society generates
incident cases of homelessness may be
vastly underestimated. Because of these
possibilities we propose that the question
of the number of homeless people in the
United States be reopened. Doing so is
critically important for two reasons.
First, because point-prevalence stud-
ies dramatically underestimate the rate at
which our society is generating homeless-
ness, they minimize the magnitude and
the seriousness of the problem. The place
of homelessness on the national agenda
and the resources available to address the
problem could suffer as a consequence.
Second, although studies of current
prevalence can be important for planning
the delivery of services, they can at the
same time lead to inappropriate conclu-
sions about the initial causes ofhomeless-
ness and the characteristics ofpeople who
become homeless. If surveys of currently
homeless people miss large numbers of
people who become homeless, as our
study suggests they do, it means that such
surveys substantially oversample people
who are chronically homeless. As a conse-
quence, studies of current prevalence
locate and describe-again and again-
those who are most thoroughlybeset with
dramatic personal problems. Bolstered by
seemingly consistent evidence from such
studies, researchers and policy makers
run the risk of substantially overestimat-
ingthe importance ofthepersonaltroubles
of chronically homeless people in under-
standing the causes ofhomelessness.
These results reopen the question of
the scope of homelessness in the the
United States. Our study provides an
assessment ofthe numbers ofpeople who
experience homelessness that conforms to
standards of social science research, but
also (as described above) circumvents
many of the biases that critics of past
prevalence studies have identified. The
fact that our rates are much higher than
the rates provided by studies of current
prevalence supports the critics of those
studies and speaks to the need to reassess
two conclusions that have been drawn
from them. First, the magnitude of the
problem of homelessness
much greater than current prevalence
studies indicate. Second, because those
studies overrepresent chronic, long-term
homeless people, they distort our image
ofwho becomes homeless and mistakenly
overemphasize the importance ofpersonal
deficits as causes ofhomelessness. 0
This research was supported by National
Institute ofMental Health grant MH46101.
We thank Micki Bresnahan, Sharon
Schwartz, and Arthur Whaley for their com-
ments and their assistance. We also thank the
staff of the Center for Survey Research for
their assistance in drafting the survey instru-
ment and for conducting the telephone inter-
1. Susser E, Moore R, Link B. Risk factors
for homelessness. Epidemiol Rev. 1994;15:
2. Brickner PW, Scallan BC. Homeless per-
sons and health care. Ann Intem Med.
3. Gelberg L, Linn LS. Assessing the physical
health ofhomeless adults.JAMA. 1989;262:
4. Susser E, Valencia E, Conover S. Preva-
lence of HIV infection among psychiatric
patients in a New YorkCity men's shelter.
AmJPublic Health. 1993;83:568-570.
5. Struening E, Padgett D. Physical health
status, substance use and abuse, and
mental disorders among homeless adults. J
Soc Issues. Winter 1990;46:65-81.
6. Burt M. Over the Edge:
Homelessness in the 1980's. New York,NY:
The Russell Sage Foundation; 1992.
7. Jahiel R. The size of the homeless popula-
tion. In: Jahiel R, ed. Homelessness: A
The Growth of
Prevention-Oriented Approach. Baltimore,
Md: Johns Hopkins University Press; 1992:
8. Rossi P. Down and Out in America: The
Origins ofHomelessness. Chicago, Ill: Uni-
versity ofChicago Press; 1989.
9. Hombs ME, Snyder M. Homelessness in
Amenca: A Forced March to Nowhere.
Washington, DC: Community for Creative
10. Taeuber CM, Siegal PM. Counting the
nation's homeless population in the 1990
census. In: Taeuber CM, ed. Enumerating
HomelessPersons: MethodsandData Needs.
Washington, DC: Bureau of the Census;
11. Wright J, Devine JA. Counting the home-
less: the Census Bureau's "S-Night" in five
cities. Eval Rev. 1992;16:355-364.
12. Burt M, Cohen B. America's Homeless:
Nwnbers, Characteristics, andPrograms That
Serve Them. Washington, DC: The Urban
Institute Press; 1989.
13. Rossi PH, Wright J, Fisher G. The urban
homeless: estimating composition and size.
14. Applebaum R. Counting the homeless. In:
Momeni J, ed. Homelessness in the United
States, Vol II: Data and Issues. New York,
NY: Greenwood Press; 1987:1-16.
15. Marcuse P. Homelessness and housing
policy. In: Caton C, ed. Homelessness in
America. New York, NY: Oxford Univer-
sity Press; 1990:138-157.
16. Wakesberg J. Sampling methods for ran-
dom digit dialing.JAm StatAssoc. 1978;73:
17. Kish L. Survey Sampling. New York, NY:
John Wiley and Sons; 1965.
18. Hildiroglou MA. SUPERCARP.
Iowa: Iowa State University; 1980.
19. Shlay AB, Rossi PH. Social science re-
search and contemporary studies of home-
lessness. Annu Rev Sociol. 1992;18:129-
20. Kleinbaum DG, Kupper LL, Morgenstem
H. Epidemiologic Research: Principles and
Quantitative Methods. Belmont, Calif: Life-
time Learning Publications; 1982.
21. Rossi P, Fisher G, Willis G. 77Te Condition
ofthe Homeless ofChicago. A Report Based
on Surveys Conducted in 1985 and 1986.
Amherst, Mass: Social and Demographic
Research Institute and NORC, A Social
Science Research Center; 1986.
22. Rosenheck R, Fontana A. A causal model
of homelessness among male veterans of
the Vietnam generation. Am J Psychiaty.
23. Davis J, Smith TW. The General Social
Survey: Cumulative Codebook and Data
File. Chicago, Ill: National Opinion Re-
search Center and University of Chicago;
24. Culhane D, Dejowski E, Ibanez J, Needham
rates in Philadelphia and New York City:
the implications of turnover for sheltered
population counts. Washington, DC: Of-
fice of Housing Research; 1993. Fannie
Mae working paper.
I. Public shelter admission
1912 American Journal ofPublic Health
December 1994,Vol.84,No. 12