American Journal of Community Psychology

, Volume 37, Issue 1, pp 47–61

Homelessness in the United States: Assessing Changes in Prevalence and Public Opinion, 1993–2001

Authors

  • Carolyn J. Tompsett
    • Research Group on Homelessness and Poverty, Department of PsychologyWayne State University
    • Research Group on Homelessness and Poverty, Department of PsychologyWayne State University
  • Melissa Guzicki
    • Research Group on Homelessness and Poverty, Department of PsychologyWayne State University
  • Manuel Manrique
    • Research Group on Homelessness and Poverty, Department of PsychologyWayne State University
  • Jigna Zatakia
    • Research Group on Homelessness and Poverty, Department of PsychologyWayne State University
Original Article

DOI: 10.1007/s10464-005-9007-2

Cite this article as:
Tompsett, C.J., Toro, P.A., Guzicki, M. et al. Am J Community Psychol (2006) 37: 47. doi:10.1007/s10464-005-9007-2

A national survey was administered in 1993–1994 (N = 360) and repeated in 2001 (N = 435) to assess the prevalence of homelessness as well as attitudes, opinions and knowledge regarding homelessness. No significant changes in prevalence were found, despite a strong US economy during most of the 7–8 year period. Respondents in 2001 had less stereotyped views of homeless people and were more supportive of services, but came to see homelessness as a less serious problem that was less often due to economic factors. This “mixed” set of findings may reflect both beliefs on the benefits of a good economy and an increased awareness of the complexity of homelessness. Across the surveys, younger, female, liberal, and less wealthy respondents demonstrated more sympathetic attitudes towards homeless people.

KEY WORDS

public opinionhomelessnesschangeprevalence

How serious of a problem is homelessness? This question may be easy to answer for those who advocate on behalf of homeless individuals and families—so long as any one person is homeless, the problem is too great. However, when new policy or services aimed at addressing the issue of homelessness are initiated, data on prevalence is often used to justify the need for these programs (Burt, Aron, Lee, & Valente, 2001). Public opinion data may also be another tool for developing new policy, as it can help convince policy-makers. Data on the relative stability of the prevalence of homelessness, and/or public opinion regarding homelessness, could effect the strength of the impact of these two sources of information on the issue. It is worthwhile to examine the possible existence of change, as well as possible determinants of change, in prevalence and public opinion regarding homelessness in the interest of gauging support for future initiatives directed at alleviating homelessness.

Prevalence

Citing statistics on prevalence can lend weight to the arguments of advocates for homeless individuals and families. Examining changes in prevalence rates could be useful for determining the effects of existing social programs, refining theories on the causes of homelessness, and supporting movements for new programs. Unfortunately, measuring the prevalence of homelessness can be problematic, with different methodologies yielding widely varying estimates, particularly when conflicting definitions of homelessness are considered. It has become widely recognized that point prevalence estimates tend to provide underestimates as they undercount people who are homeless for only brief periods of time, and the common method of surveying services misses homeless people not using services (Burt et al., 2001; Link et al., 1994; Toro & Warren, 1999). Telephone surveys conducted with the general population also yield underestimates, as the currently homeless and others without telephones are not surveyed. However, this methodology has the advantage of reaching people who had once been homeless, even if only very briefly, and it can estimate lifetime prevalence (Link et al., 1994; Toro & McDonell, 1992). Recent estimates have identified rates of literal homelessness ranging from an annual prevalence of 1% when surveys of service providers are consulted (Burt et al., 2001), to 3.1% over 5 years and 7.4% for lifetime prevalence when telephone surveys with the general population are used (Link et al., 1994). While these estimates clearly establish homelessness as a troubling problem that affects many Americans, the variety of methods used hampers efforts at tracking changes in rates over time. The current study attempts to address this problem by using the same telephone survey to collect prevalence data at two time points.

Public Opinion

While early research focused on examining the personal traits of homeless people to identify possible explanations for their homelessness, more recently it has been recognized that structural factors (e.g., availability of low-income housing or social programs) play as great a role, if not a greater role, in determining pathways from poverty to homelessness (Blasi, 1990; Shinn, 1992; Zlotnick, Robertson, & Lahiff, 1999). To address these structural causes, social policies may become increasingly vital. There is evidence that public opinion is linked to changes in social policy (Jason & Rose, 1984; Monroe, 1983; Page & Shapiro, 1989).

Media Reflecting Public Opinion

When policymakers do not have relevant public opinion data readily available, they may rely on the media to provide this information for them. While media attention may respond to public interest in a subject, often the relationship between the media and public opinion works in the other direction or is entirely absent (Cook et al., 1983; Mutz & Soss, 1997). One example of an inaccurate media portrayal of public opinion regarding homelessness is the concept of “compassion fatigue.” In the late 1980s and early 1990s, media attention to homelessness dropped dramatically. News stories suggested that Americans had once been sympathetic to the plight of homeless people but that, after years of over-attention, they were now tired of reading and hearing about the issue (Buck, Toro, & Ramos, 2004; Hombs, 2001). Studies of actual public opinion in the United States have shown this media presentation of a “compassion fatigued” public to be inaccurate (Link et al., 1995; Toro & McDonell, 1992). It appears that portrayals of the issue of homelessness may reflect the motivation of the media to publicize “new news” rather than the “old news” of ongoing social concerns (Buck et al., 2004).

Influences on Public Opinion

The media's coverage of a social issue may shape public opinion by framing the issue in a certain light, and by creating an illusion of popular consensus that leads individuals to reassess their personal views (Behr & Iyengar, 1985; Mutz & Soss, 1997; Qualter, 1989). Evidence suggests that public opinion regarding value priorities and policy preferences tend to change gradually over time as the public slowly re-evaluates values in light of environmental (societal) changes (Behr & Iyengar, 1985; Rokeach & Ball-Rokeach, 1989). As the public evaluates their values concerning a given social issue such as homelessness, media coverage of other issues may indirectly lead to changes in public opinion.

One social factor that may be particularly influential is the perception of change in the economy (Durr, 1993; Elliot et al., 1997). During times of increasing economic stability, the public may be less supportive of new programs to alleviate homelessness, or their view of the relative importance of the issue may shift as other social issues increase in urgency. As respondents’ level of personal income appears to be an inconsistent predictor of their attitude towards the poor or homeless (Feagin, 1975; Toro & McDonell, 1992), the influence of economic trends on public opinion may be more a function of perceptions of available resources for assisting the poor or homeless, rather than the indirect effects of respondents’ personal economic comfort. While public opinion trends appear gradual overall, in times of crisis opinion may shift more abruptly (Page & Shapiro, 1982). However, in these cases the abrupt shift in opinion is usually related to the crisis itself, with opinion on unrelated issues unaffected.

Demographic Predictors

Along with an examination of change in public opinion over time, information on demographic predictors of public opinion may prove useful in identifying sources of support for public policy. Toro and McDonell (1992) found that younger and female respondents were more generally sympathetic regarding the homeless population, and were more aware of related structural factors. Examining effects of education, Phelan, Link, Stueve, and Moore (1995) found that better educated respondents expressed greater tolerance for homeless people, but less support for economic aid for them, while Lee, Jones, and Lewis (1990) noted that increased education predicted less support for a personal deficiency model of homelessness. Studies have also examined the effects of political party allegiance and income on attitudes towards homelessness, with varying results (Furnham, 1996; Toro & McDonell, 1992).

The Current Study

As community psychologists, we have a clear commitment to producing research that is directly applicable to social concerns. Data on the state of public opinion on homelessness, as well as the stability of this public opinion over time, may guide advocates and policymakers as they gauge support for new programs. As the media provides an unreliable portrayal of public interest, rigorous study of public opinion appears crucial to provide accurate evidence of public support. The current study assessed public opinion on homelessness over an 8-year period between 1993–1994 and 2001, examining changes over time as well as demographic predictors of public opinion. Between these time points, the economy in the United States was on a steady rise, with unemployment decreasing and positive media reports emphasizing the prosperity enjoyed by all Americans (Cooper & Madigan, 1999). In 2001, the economy was just beginning the slump that would eventually become a longer-lasting recession. However, at this time, it was not yet widely recognized that such a recession had begun (Pellegrini, 2001).

The current study measured prevalence at both time points using a telephone survey methodology, in order to shed light on the possible effects of a healthy economy on rates of homelessness in the general population. The public's perception of the seriousness of homelessness as a problem may also be affected by the positive economic trends reported in the media, along with other attitudes and attributions for homelessness and perceptions of the homeless population itself. In addition, this study looked for possible effects of the events of September 11 within the 2001 sample, as it was considered possible that such a dramatic event, along with its overwhelming media coverage, might impact the public's views of unrelated social issues such as homelessness. Finally, demographic predictors of public opinion regarding homelessness were examined, and change in these predictors over time was also assessed.

METHOD

Participants

Participating households were selected using random digit dialed telephone numbers generated by Survey Sampling, Inc., weighted by population density to be nationally representative. The random digit dialing method ensures that unlisted numbers are sampled, but approximately 6.2% of households were not included in the 1993–1994 sample, and 5.1% in the 2001 sample, because they did not have a telephone (Federal Communications Commission, 2001). An additional 0.5–2.0% were excluded from each sample because they had only a mobile phone (Telephone Communication, Survey Sampling, Inc., May 20, 2002). An adult within each household was randomly selected using a variation of Frey's (1989) method, as used in other recent surveys on homelessness (Link et al., 1994, 1995; Toro & McDonell, 1992); 13% of households reached in 2001, and approximately 11–15% in 1993–1994, were determined to be eligible and agreed to participate. While these response rates may appear low, they are similar to those obtained by pollsters calling respondents at random in recent years (Council for Marketing and Opinion Research, 2003). There is some research, including studies using a very similar survey, that suggests that higher response rates obtained through persistent callbacks and other incentives may not significantly impact results as compared with less time-intensive administrations of the same survey (Langer, 2003; Link et al., 1994, 1995).1 All interviews were conducted anonymously. The final samples included 360 respondents in 1993–1994 (female = 235, male = 125) and 435 respondents in 2001 (female = 260, male = 175).

1

Measure

The survey instrument was adapted from those of Toro and McDonell (1992) and Link et al. (1994, 1995). A total of 159 items were used in both surveys. These items were designed to assess respondents’ attitudes, knowledge, and opinions regarding homelessness, as well as policy-related beliefs, their personal experiences with homelessness and homeless people, and demographic characteristics of the respondents. Respondents’ personal experiences with homelessness were assessed by querying whether they had ever considered themselves homeless, following up with further questions if they responded affirmatively. Several items addressing respondents’ attitudes towards homelessness were forced-choice questions (e.g., “Which of the following do you think should be most responsible for helping the homeless:” 1 =: government; 2 =: churches and charities; 3 =: the homeless themselves), with the remainder using four-point Likert scales (e.g., “How much does a shortage of affordable housing contribute to homelessness?” 1 =: a lot, 2 =: some; 3 =: a little; 4 =: not at all). An even number of response choices was used in Likert scales to prevent “neutral” responses, and many items were reverse coded to avoid response bias. Continuous measures addressing knowledge of characteristics of homeless people were also included, asking the respondent to estimate the percentage of homeless persons sharing a given characteristic (e.g., “How many homeless people out of 100 are male?”). The perceived importance of homelessness relative to other issues was assessed by asking if the respondent believed a given issue (e.g., “strengthening the national defense”) was more, equally, or less important than reducing homelessness.2 Test-retest reliability assessed by Toro and McDonell (1992) indicated over 80% reliability for all items included in the current study.

2

Three factor analyses using orthogonal varimax rotation were used to reduce 69 items to eight factors: General Compassion, Limit Public Rights, Trustworthy, Social Isolation, Street People, Housing/Services Needed, Economic Factors as Cause, and Personal Factors as Cause (see Tables IIII). Because analyses conducted separately for each sample resulted in similar factor structures, pooled analyses were used here for the construction of factor scores (N = 360 + 435 = 795). Three of the resulting factors are comparable to those found by Toro and McDonell (1992), with the remaining five representing additional nuances of attitudes. Exact factor scoring was used to construct the variables used in all subsequent analyses.
Table I.

Pooled General Attitude Factors

Description of item

General compassiona

Limit public rightsa

Feel compassionate when think about homeless

  .74

 

Angry that so many are homeless in a rich country

  .62

 

Willing to have shelters, housing in neighborhood

  .57

 

Would pay higher taxes to help the homeless

  .50

 

Would be careful not to touch a homeless person

−.45

 

Homeless people make money illegally

−.46

 

Less compassion for the homeless than before

−.64

 

Homeless should have the: Right to beg

 

−.76

 Right to set up a tent in public places

 

−.76

 Right to sleep in public

 

−.70

Percent of total variance

20.3 

18.0   

Note. Loadings over .40 included in the table; loadings greater than .50 are in boldface. N = 792.

Table II.

Pooled Social Attitude Factors

Description of Item

Trustworthy

Social isolation

Street people

Homeless are more likely to commit violent crimes

−.62

  

I would not trust someone known to have been homeless

−.62

  

Homeless should not be allowed to gather in public

−.59

  

Most homeless people have family troubles

−.55

  

Even if they seem right, homeless people may be more dangerous

−.54

  

Most homeless people have job skills

  .54

  

Most homeless people could care for a home

  .55

  

The more homeless people in an area, the worse the neighborhood

−.49

  

It is natural to be afraid of a street dweller

−.49

  

Most homeless Respect neighbors property

  .47

  

Homeless are not more dangerous

  .46

  

Homeless people: Have trouble making friends

−.41

  

 Spend waking hours with relatives

 

−.66

 

 Receive unemployment wages as income

 

  .50

 

 Receive income through relatives

 

−.55

 

 Spend waking hours looking work

 

−.59

 

 Sleep at the homes of family members

 

−.65

 .44

 Sleep on the streets

  

 .66

 Spend most waking hours outside

  

 .51

 Sometimes sleep in jail

  

 .48

 Often go without food

  

 .43

 Sleep in abandoned buildings

  

 .56

Percent of total variance

13.2 

9.8

9.3

Note. Loadings over .40 included in the table; loadings greater than .50 are in boldface. aN = 777.

Table III.

Pooled Cause/Solution Belief Factors

 

Housing/services needed

Economic factors as cause

Personal failings of homeless

Would this reduce homelessness: Free alcohol and drug treatment

  .66

  

 More temporary and emergency housing

  .64

  

 Child care programs so homeless mothers can work

  .60

  

 Increasing minimum wage employers must pay

  .55

  

 Federal emergency housing

  .55

  

 Committing the seriously mentally ill

  .54

  

 Providing housing for released mental patients

  .53

  

Should the government: increase minimum wage

  .51

  

 Increase spending drug and alcohol treatment

  .49

  

 Provide tax breaks for low-income developers

  .47

  

Would more low-income housing reduce homelessness

  .48

  

Helping the homeless receive welfare benefits would reduce homelessness

  .51

  .40

 

Would providing rent subsidies reduce homelessness

  .51

  .42

 

Should the government: increase federal rent subsidies

  .50

  .47

 

 Build more affordable housing

  .48

  .51

 

A shortage of government aid contributes to homelessness

 

  .58

 

Unemployment is a major cause of homelessness

 

  .52

 

Should the government spend more on welfare benefits

 

  .52

 

If there were more jobs, fewer would be homeless

 

  .48

 

Job training programs would solve homeless problem

 

  .43

 

The economic system contributes to homelessness

 

  .47

 

Many homeless lost their homes through foreclosure

 

  .46

 

Lack of affordable housing contributes to homelessness

 

  .47

 

Failure of society to provide good schools contributes to homelessness

 

  .42

 

Many homeless remain so by choice

 

−.48

 

Cutting off welfare would reduce homelessness

 

−.41

 

Factors contributing to homelessness: Irresponsible behavior

 

−.46

 

 Mental illness

  

 .62

 Release of mental patients

  

 .64

 Physical illness

  

 .57

 Bad luck

  

 .42

Percent of total variance

14.8 

13.5 

6.3

Note. Loadings over .40 included in the table; loadings greater than .50 are in boldface. N = 780.

Procedure

For both samples, surveys were administered by telephone by trained interviewers associated with a large Midwestern university. Calls were placed at varying times throughout the day, with most taking place in the evening between 5:00 and 9:00 P.M. (respondents’ time). Once respondents were contacted, they were asked to take part in an anonymous survey on housing lasting 15–20 min. This introduction was used to avoid some of the response bias that might arise if respondents were initially informed that the survey focused on homelessness. After a respondent had the opportunity to pose questions regarding the study and consent was obtained, the interview was administered in its entirety. Interviews were conducted from September, 1993 through May, 1994 for the first sample, and from January, 2001 through December, 2001 for the second sample.

Results

The lifetime prevalence of literal homelessness was 8.1% for the 1993–1994 sample and 6.2% for the 2001 sample (χ2(1) = 1.07, p = .30). Lifetime prevalence for any type of homelessness reported (including “precarious housing”) was 11.7% for the 1993–1994 sample and 12.9% for the 2001 sample (χ2(1) = .24, p = .63). Prevalence for literal homelessness experienced within the past 5 years was 3.9% for the 1993–1994 sample and 1.9% for the 2001 sample (χ2(1) = 3.08, p = .08).

Eight multivariate analyses of variance (MANOVAs) were conducted to assess changes across the two surveys. The following dependent variables were considered in each of these MANOVAs: (1) key opinion items including whether federal spending should be increased, seriousness of the problem at the local, city, and national levels, whether homelessness is improving; (2) indicators of influence on views, including how much the respondent reads and is influenced by media reports on homelessness, and frequency of talking about homelessness; (3) exposure and sympathy related items including, willingness to pay higher taxes, willingness to have services for homeless people nearby, frequency of being approached and giving to panhandlers, and ranking of homelessness as important compared with six other issues; (4) all eight attitude factor scores; (5) estimates of personal characteristics of percent male, married, with children, and having frequent contact with relatives; (6) social characteristics of percent eventually finding a permanent home, percent on public assistance, percent without a high school degree, and percent high school graduate; (7) mental health and related estimates, including percent mentally ill, percent mentally retarded, percent alcoholic, percent substance abusing, and percent with a criminal record; and (8) importance of homelessness ranked against specific social issues of strengthening the national defense, cleaning up the environment, improving health care, reducing the national debt, improving public education, and developing child care programs. Significant differences were found in all eight analyses. Several significant univariate differences between samples, presented in Table IV and V, were found for behavior, attitude, and opinion items.3
Table IV.

Changes in Attitude and Opinion Items

 

1993–4 Survey

2001 Survey

 

Variable

M

SD

M

SD

F

Opinion and behavior itemsa

 Frequency of panhandler requests (1: never; 4: more than 10 times)

2.15

1.11

2.76

 .99

55.00 (1, 667)***

 Frequency of giving to panhandlers (1: never; 4: almost always)

1.64

1.48

2.54

1.10

76.95 (1, 667)***

Attitude factorsb

 Street people

 .21

1.06

−.18

 .91

31.04 (1, 774)***

 Housing/services needed

−.13

1.04

 .11

 .95

11.26 (1, 774)***

 Economic factors

 .17

 .97

−.14

1.01

18.81 (1, 774)***

General opinionsc

 Seriousness in county (1: not at all serious; 4: very serious)

2.88

1.04

2.56

 .96

16.98 (1, 666)***

  In city

3.37

 .84

3.18

 .83

 8.57 (1, 666)***

 Problem of homelessness improving (1: worse; 3: improving)

1.32

 .58

1.61

 .74

31.82 (1,666)***

Influences on opiniond

 Frequency of reading about homeless (1: none; 4: a lot)

2.82

 .95

2.53

 .92

18.75 (1, 783)***

 Importance of media influence on views (1: not important; 3: very important)

2.14

 .70

1.88

 .71

26.44 (1, 783)***

 Frequency of talking about homelessness (1: never; 4: often)

2.53

 .91

2.38

 .86

 5.34 (1, 783)*

Comparison issuese

 Strengthening national defense (1: homelessness more important; 2: equally important; 3: homelessness less important)

1.73

 .79

2.02

 .81

23.66 (1, 737)***

 Cleaning up environment

2.11

 .72

2.00

 .71

 4.72 (1, 737)*

 Reducing national debt

2.06

 .81

1.68

 .77

41.61 (1, 737)***

 Improving public education

2.36

 .66

2.47

 .64

 5.38 (1, 737)*

Note. Here and in Tables VXI, MANOVAs were used as a conservative screen. Post hoc (Student–Newman–Keuls) tests were conducted only when both the MANOVA and associated univariate ANOVA were statistically significant (p < .05).

aMANOVA F = 21.76 (5, 663), p < .001.

bMANOVA F = 9.19 (9, 766), p < .001.

cMANOVA F = 9.65 (5, 662), p < .001.

dMANOVA F = 13.66 (3, 781), p < .001.

eMANOVA F = 17.35 (6, 732), p < .001.

*p < .05. **p < .01. ***p < .001.

Table V.

Changes in Views of Homeless Characteristics

 

1993–1994 Survey

2001 Survey

 

Variable

M

SD

M

SD

F

Family characteristicsa

 Percent male (0–100)

52.18

24.85

64.62

15.47

68.09 (1, 734)***

  With children

41.52

27.17

50.67

24.29

23.29 (1, 734)***

  Regular contact relatives

27.30

22.83

35.39

24.55

21.24 (1, 734)***

Demographic characteristicsb

 Percent find permanent home

29.35

22.12

33.45

22.23

 6.14 (1, 722)*

  Receive public assistance

41.10

28.59

48.32

25.54

12.86 (1, 722)***

  Did not finish high school

56.44

20.53

51.50

18.22

11.75 (1, 722)***

Personal characteristicsc

 Percent mentally ill

29.01

21.98

37.01

22.89

23.31 (1, 738)***

  Mentally retarded

16.63

14.22

20.38

18.34

 9.47 (1, 738)**

  With criminal record

34.09

23.34

42.55

23.22

24.31 (1, 738)***

aMANOVA F = 20.11 (4, 731), p < .001.

bMANOVA F = 9.21 (4, 719)***, p < .001.

cMANOVA F = 8.04 (5, 734), p < .001.

*p < .05. **p < .01. ***p < .001.

3

Eight additional MANOVAs, using the same dependent variable groupings as the main set reported above, assessed for pre- and post-September 11 differences on the dependent variables of interest in the second sample (N = 236 pre-Sept. 11, N = 185 post-Sept. 11). The only significant effects at the multivariate and univariate levels indicated that post-Sept. 11 respondents were more likely to consider strengthening the national defense (univariate F(1, 377) = 62.33, p < .001) and improving public education (F(1, 377) = 4.93, p < .05) as more important social issues than reducing homelessness.

Both samples were pooled and eight multivariate multiple regression analyses were conducted to assess the association of gender, race, age, education, political conservatism, and income level with attitude, opinion, and knowledge factors and items.4 The same dependent variable groupings were used here as in the MANOVAs described above. Each successive predictor variable, entered in the order listed above, controlled for all others previously entered. Interactions between survey year and each of the other independent variables were entered in a final step. Variables showing significant effects in the multivariate analyses were then examined in univariate multiple regression analyses. Hierarchical multiple regression was used to reduce individual sample effects. Simultaneous entry resulted in all results remaining significant, with the exception of the effects of age on General Compassion and estimates of percent of homeless with children, and the effects of education on Social Isolation. These findings should be interpreted with caution.

4

Relatively few interactions between survey year and other predictors were found across the eight analyses, indicating that demographic predictors tend to demonstrate stable patterns, at least over the 8-year period of the present study. One significant interaction for age (β = −.25; F = 11.25 (1, 620), p < .001) found that older people were less likely to support increased federal spending on homelessness in 1993–1994 (F = 21.11(1, 322), p < .001), while age was not predictive in 2001. Two significant interactions for race, on seriousness of homelessness at the national level (F = 7.64(1, 620), p < .01), and perception of whether homelessness is improving (F = 5.12(1, 620), p < .05), were uninterpretable as these effects were nonsignificant within each year. Two significant interactions were found for education, on estimates of percent alcoholic (F = 5.26(1, 722), p < .05) and percent drug abusing (F = 9.03(1, 722), p < .01). For both variables no effect was found in 1993–1994. In 2001, higher education correlated with lower estimates of homeless people being alcoholic (β = −.25; F = 23.70(1, 387), p < .001) and drug abusing (β = −.23; F = 21.63(1, 387), p < .001). Five interactions were found between political orientation and survey year, one of which, estimates of the percent of homeless people who receive public assistance (F = 2.68(3, 705), p < .05), was not interpretable as effects were nonsignificant within each time point. The remaining four interactions included estimates of percent male (F = 3.13(3, 717), p < .05), percent with regular contact with family (F = 3.17(3, 717), p < .05), percent who eventually find a permanent home (F = 3.39(3, 705), p < .05), and percent who abuse drugs (F = 4.91(3, 720), p < .01). All four of these effects were nonsignificant in 2001. However, Student Newman–Keuls post hoc tests revealed that in 1993–1994 the estimates given by the “other” group were significantly lower than the remaining groups on percent male (F = 2.89(3, 329), p < .05), percent with regular family contact (F = 3.65(3, 329), p < .05), and percent who eventually find a permanent home (F = 3.00(3, 324), p < .05). The “other” group and the conservative respondents gave significantly higher estimates than the liberal respondents for percent abusing drugs (F = 7.51(3, 330), p < .001).

Main effects for gender are presented in Table VI.2 These effects were all consistent with a general pattern of the female respondents being more concerned about homelessness and having more compassionate views towards homeless people. With the exception of their higher scores on Street People, perhaps reflecting sympathy for people perceived as sleeping in unstable circumstances, women demonstrated less stereotyped views of homeless people. Age effects, presented in Table VII, generally indicated that older respondents tended to be less supportive of the homeless population in general.
Table VI.

Gender Main Effects

Variable

Female mean

Male mean

F

Opinion and behavior itemsa

 Willing to have shelters, housing in neighborhood (composite, 0–6 scale)

 3.86

 3.28

13.25 (1, 653)***

 Frequency of panhandler requests

 2.33

 2.11

 6.88 (1, 614)***

Attitude factorsb

 General compassion

  .21

−.33

54.94 (1, 763)***

 Trustworthy

  .09

−.16

11.29 (1, 763)***

 Street people

  .09

−.16

11.50 (1, 763)***

 Housing/services needed

  .14

−.21

22.19 (1, 763)***

 Economic factors

  .13

−.21

20.43 (1, 763)***

General opinionsc

 Increase federal spending (1: decreased; 2: same; 3: increased)

 2.54

 2.41

 5.44 (1, 651)*

 Seriousness in county

 2.82

 2.51

15.16 (1, 651)***

 In city

 3.33

 3.14

 8.32 (1, 651)**

 Nationwide

 3.69

 3.38

32.65 (1, 651)***

 Homelessness improving

 1.41

 1.60

11.80 (1, 651)***

Influences on opiniond

 Frequency of reading about homeless

 2.74

 2.54

 7.46 (1, 764)**

 Frequency of talking about homeless

 2.56

 2.26

19.14 (1, 764)***

Family characteristics of the homelesse

 Percent male

57.22

61.34

 6.33 (1, 723)*

 Married

31.06

26.02

12.91 (1, 723)***

Demographic characteristicsf

 Percent did not finish high school

52.27

56.74

 8.84 (1, 712)**

Personal characteristicsg

 Percent with criminal record

36.19

42.55

12.44 (1, 727)***

Comparison issuesh

 Strengthening national defense

 1.83

 1.97

 5.07 (1, 728)*

 Cleaning up environment

 2.00

 2.14

 6.81 (1, 728)**

 Reducing national debt

 1.81

 1.94

 4.64 (1, 728)*

aMANOVA F = 4.57 (5, 649), p < .001.

bMANOVA F = 9.13 (8.756), p < .001.

cMANOVA F = 8.53 (5, 647), p < .001.

dMANOVA F = 7.18 (3, 762), p < .001.

eMANOVA F = 9.12 (4, 720), p < .001.

fMANOVA F = 5.13 (4, 709), p < .001.

gMANOVA F = 5.39 (5, 723), p < .001.

hMANOVA F = 3.05 (6, 723), p < .01.

*p < .05. **p < .01. ***p < .001.

Table VII.

Age Main Effects

Variable

ΔR2

β

F

Opinion and behavior itemsh

 Willing to have shelters, housing in neighborhood

 .0496

−.22

37.64 (1, 647)***

 Frequency of panhandler requests

 .0422

−.21

22.66 (1, 647)***

 Willingness to pay higher taxes (any)

 .0399

−.20

28.16 (1, 647)***

 Willingness to pay extra $25 per year in taxes

 .0291

−.17

24.39 (1, 647)***

Attitude factorsb

 General compassion

 .0063

−.08

 4.73 (1, 758)*

 Limit public rights

 .0482

  .22

37.86 (1, 758)***

 Trustworthy

 .0175

−.13

13.79 (1, 758)***

 Social isolation

 .0117

  .11

 9.11 (1, 758)**

 Street people

 .0083

−.09

 6.50 (1, 758)*

 Housing/services needed

 .0315

−.18

25.27 (1, 758)***

 Economic factors as cause

 .0118

−.11

 9.24 (1, 758)**

General opinionsc

 Increase federal spending

 .0177

−.14

 9.71 (1, 644)**

 Seriousness in county

 .0053

−.07

 7.39 (1, 644)**

 In city

 .0120

−.11

11.56 (1, 644)***

 Nationwide

 .0098

−.10

 9.46 (1, 644)**

Family characteristicsd

 Percent with children

 .0058

−.08

 4.25 (1, 724)*

Demographic characteristicse

 Percent find permanent home

 .0139

−.12

 8.77 (1, 710)**

Comparison issuesf

 Strengthening national defense

 .0220

  .15

17.52 (1, 727)***

 Improving health care

 .0110

  .11

 6.83 (1, 727)**

 Reducing national debt

 .0266

  .16

19.39 (1, 727)***

aMANOVA F = 13.88 (5, 640), p < .001.

bMANOVA F = 8.54 (8, 748), p < .001.

cMANOVA F = 4.48 (5, 637), p < .001.

dMANOVA F = 3.98 (4, 716), p < .01.

eMANOVA F = 3.15 (4, 704), p < .05.

fMANOVA F = 6.59 (6, 719), p < .001.

*p < .05. **p < .01. ***p < .001.

The three main ethnic groups (White, African American, and Other) were contrasted using Student–Newman Keuls post-hoc analyses. Results are presented in Table VIII . African American respondents stood out from the other two racial groups on several of the findings, for example African Americans were more likely to support an increase in federal spending, acknowledged greater importance of the media on their views, and endorsed higher levels of Street People and Economic Factors than respondents in the White and Other groups.
Table VIII.

Race Main Effects

Variable

White (N = 629)

African–American M(N = 68)

Other M(N = 98)

MF

Post hoc tests

Increase federal spendinga

 2.45

 2.78

 2.55

 6.70 (2, 649)**

A>O,W

Importance of media influence on viewsb

 1.98

 2.25

 2.10

 4.56 (2, 649)*

A>W

Frequency of panhandler requestsc

 2.83

 2.66

 2.34

 6.61 (2, 613)**

A>W

Social isolationd

 .06

−.25

−.31

 7.35 (2, 762)***

W>A,O

Street peopled

−.03

 .29

−.01

 3.21 (2, 762)*

A>O,W

Economic factors as caused

−.04

 .34

 .01

 4.48 (2, 762)*

A>W,O

Percent of homeless with childrene

45.24

60.08

43.89

 9.99 (2, 722)***

A>W,O

 Mentally Illf

31.71

45.74

35.12

11.74 (2, 726)***

A>O,W

 Mentally retardedf

17.69

27.34

18.84

 9.98 (2, 726)***

A>O,W

 Abusing drugsf

49.84

58.94

51.26

 4.38 (2, 726)*

A>O,W

 With criminal recordf

36.98

49.39

41.65

 8.88 (2, 726)***

A>O,W

aMANOVA F = 2.08 (10, 1290), p < .01.

bMANOVA F = 4.81 (6, 1520), p < .001.

cMANOVA F = 2.28 (10, 1298), p < .01.

dMANOVA F = 2.68 (16, 1508), p < .001.

eMANOVA F = 4.43 (8, 1436), p < .001.

fMANOVA F = 4.74 (10, 1442), p < .001.

*p < .05. **p < .01. ***p < .001.

Effects of education (Table IX) appear mixed with regard to general sympathy towards homeless people. Although respondents with higher education appear to be more personally supportive of homeless people through giving to panhandlers, paying taxes, and holding some less stereotyped views of their characteristics; on a more general level they appear to view homelessness as a personal issue less likely to be improved by structural-level interventions.
Table IX.

Education Main Effects

Variable

ΔR2

β

F

Frequency of reading about homelessnessa

 .0153

  .13

12.07 (1, 752)***

Frequency of panhandler requestsb

 .0099

  .10

11.06 (1, 645)***

Frequency of giving to panhandlersb

 .0153

  .13

10.93 (1, 645)**

Willingness to pay higher taxes (any)b

 .0118

  .11

 9.27 (1, 645)**

Willingness to pay extra $25 per year in taxesb

 .0122

  .11

 6.88 (1, 645)**

Trustworthyc

 .0152

  .12

12.12 (1, 755)***

Social isolationc

 .0117

  .11

 9.18 (1, 755)**

Street peoplec

 .0270

−.17

21.50 (1, 755)***

Housing/services neededc

 .0049

−.07

 3.97 (1, 755)*

Personal failings as causec

 .0066

  .08

 5.12 (1, 755)*

Percent alcoholicd

 .0220

−.15

17.75 (1, 723)***

 Abusing drugsd

 .0175

−.14

13.75 (1, 723)***

 With criminal recordd

 .0051

−.07

 3.96 (1, 723)*

aMANOVA F = 5.44 (3, 746), p < .01.

bMANOVA F = 5.89 (5, 637), p < .001.

cMANOVA F = 7.82 (8, 744), p < .001.

dMANOVA F = 5.61 (5, 715), p < .001.

*p < .05. **p < .01. ***p < .001.

Results for political conservatism indicated that liberal respondents tended to be more sympathetic in general and endorsed less stereotyped characteristics of homeless people (Table X). Effects of income, presented in Table XI, were mixed, and indicated that a refusal to reveal income was as predictive as the individual income groups themselves. It is suspected that many of those who refused to divulge their income may have been wealthier than other respondents or alternatively may have been more suspicious of the interview process.
Table X.

Political Conservatism Main Effects

Variable

Conservative M (N = 291)

Moderate M (N = 196)

Liberal M (N = 189)

Other M (N = 119)

F

Post hoc

Increase federal spendinga

2.34

2.56

2.69

2.41

 8.55 (3, 639)***

L>O,C; M>C

Seriousness in countya

2.50

2.63

2.94

3.01

 7.71 (3, 639)***

O,L>M,C

Importance of media influence on viewsb

2.03

2.09

1.99

1.77

 5.59 (3, 750)**

M,C,L>O

Willing to have shelters, housing in neighborhoodc

3.35

3.70

4.15

3.47

 3.81 (3, 643)*

L>C,O

Frequency of panhandler requestsc

2.28

2.44

2.69

2.36

 3.07(3, 643)*

L>M,O,C

Willingness to pay higher taxes (any)c

1.08

1.33

1.59

0.99

 6.61 (3, 643)***

L>M>O; L>C,O

Willingness to pay extra $25 per year in taxesc

1.84

2.11

2.28

1.91

 7.73 (3, 643)***

L>M>C,O

General compassiond

−.14

 .06

 .33

−.24

 8.51 (3, 753)***

L>M>O; L>C

Limit public rightsd

 .19

−.14

−.17

−.03

 4.67 (3, 753)**

C>O,M,L

Trustworthyd

−.17

 .04

 .26

−.14

 5.01 (3, 753)**

L>M,O,C

Housing/services neededd

−.14

−.01

 .24

−.06

 4.10 (3, 753)**

L>M,O,C

Economic factors as caused

−.25

 .07

 .32

 .04

10.27 (3,753)***

L>M,O>C

Strengthening national defensee

2.05

1.83

1.67

1.90

 6.22 (3, 722)***

C>M,L; C,O>L

Cleaning up environmente

1.98

1.98

2.11

2.29

 5.45 (3, 722)**

O>L,C,M

Reducing national deficite

2.03

1.81

1.72

1.75

 4.28 (3, 722)**

C>M,O,L

aMANOVA F = 3.27 (15, 1740), p < .001.

bMANOVA F = 2.43 (9, 1808), p < .01.

cMANOVA F = 3.06 (15, 1751), p < .001.

dMANOVA F = 3.14 (24, 2150), p < .001.

eMANOVA F = 3.76 (18, 2014), p < .001.

Table XI.

Income Main Effects

Variable

Lower (N = 222)

Middle (N = 252)

Higher (N = 272)

Refused (N = 99)

F

Post hoc

Frequency of talking about homelessnessa

 2.62

 2.43

 2.40

 2.22

 5.04 (3, 750)**

L>R

Willing to have shelters, housing in neighborhoodb

 4.11

 3.68

 3.39

 3.00

 6.52 (3, 643)***

L,M>R; L>H,R

General compassionc

  .17

−.05

  .03

−.28

 5.03 (3, 753)**

L,H,M>R

Social isolationc

−.32

  .14

  .11

  .02

 7.82 (3, 753)***

M,H,R>L

Street peoplec

  .24

  .02

−.26

−.03

 5.03 (3, 753)**

L,M>H; L>R,H

Housing/services neededc

  .16

−.02

−.05

−.26

 2.75 (3, 753)*

L>R

Personal failings as causec

−.01

  .08

  .02

−.30

 3.13 (3, 753)*

M,H,L>R

Percent maled

53.80

59.28

62.99

58.62

 5.87 (3, 718)***

H>L

 Not high school graduatee

56.26

54.16

52.50

49.59

 2.72 (3, 706)*

L>R

 High school graduatee

28.22

30.46

31.46

35.77

 4.43 (3, 706)**

R>H,M,L

aMANOVA F = 2.23 (9, 1801), p < .05.

bMANOVA F = 2.31 (15, 1742), p < .01.

cMANOVA F = 3.33 (24, 2141), p < .001.

dMANOVA F = 2.07 (12, 1871), p < .05.

eMANOVA F = 1.81 (12, 1839), p < .05.

DISCUSSION

The Prevalence of Homelessness Over Time

In spite of an increase in shelter beds and suggestions that homelessness may itself be increasing (Burt et al., 2001), the present study did not find statistically significant differences in the prevalence of lifetime precarious housing, lifetime literal homelessness, or 5-year prevalence of literal homelessness between 1993 and 1994 (with rates of 11.7, 8.1, and 3.9%, respectively) and 2001 (with rates of 12.9, 6.2, and 1.9%). The earlier national study by Link et al. (1994), completed in 1990, also found rates quite similar to those we obtained 3–4 and 11 years later (14.0, 7.4, and 3.1%, respectively, for the three rates presented above). Although the almost two-percentage point decrease in both lifetime and 5-year prevalence hints at a decrease in the rate of homelessness during this period, the lack of statistical significance makes it difficult to make such a conclusion. Furthermore, the rate of precarious housing actually increased slightly (by more than 1%) over the 8-year period studied. While it can be problematic to interpret the lack of significant differences across the surveys (one cannot “prove” the null hypothesis; Grimm, 1993), the large samples involved here suggest that homelessness, despite the strong economy during the 1990s, did not clearly improve during this period.

There are many possible reasons for what appears to be a constant rate of homelessness over the 1990s into the 2000s. While the percentage of individuals living below the official poverty line dropped from 14.5% in 1994 to 11.7% in 2001, the percentage living in “extreme poverty” (below 50% of the poverty line) showed a relatively smaller reduction, from 5.9% in 1994 to 4.8% in 2001 (Current Population Survey, 1994, 2001). It appears that the economic recovery of the 1990s may have had little impact on the extremely poor, from whose ranks homeless individuals and families generally come. This limited economic improvement for the very poor was, perhaps, due to the welfare reforms and other cuts in social and health benefits during this period that may have negated any positive impact of a strong economy (Edelman, 1997). Another possibility is that the growth of services for homeless people during this period may have identified more of those who were homeless in 2001 as compared to 1993–1994.

Public Opinion on Homelessness Over Time

Respondents reported significantly higher frequencies in 2001 both of being approached by and giving to panhandlers, suggesting that while the prevalence of homelessness may have remained stable, panhandlers have become more persistent and/or visible to the general public. Despite this greater contact, respondents also reported seeing homelessness as a less serious problem both in their local county and nearest city. These opinions could be the result of perceptions of a good economy, leading respondents to assume that decreases in unemployment and a strong economy resulted in an improvement in the circumstances for homeless people. Alternatively, they might reflect the influence of reduced media coverage on the topic, as indicated by respondents’ reporting decreases in reading about the homeless and in perceived influence of the media. Despite these changes, respondents did not evidence a change in general compassion towards homeless people, in keeping with prior studies finding no evidence for “compassion fatigue” in the general public (Link et al., 1995).

Attitude changes across the two surveys reflect an increased recognition of diversity in the homeless population and an increase in support for interventions aimed at structural rather than personal factors. However, 2001 respondents seemed less willing to attribute homelessness to economic factors during a time of economic prosperity. This balancing of seemingly conflicting ideologies by supporting structural interventions while minimizing the causal role of structural factors is supported by prior research indicating that Americans tend to espouse complex attributions for poverty, which can appear to be self-contradictory but actually reflect a weighing of different aspects of the same issue; one example would be the finding that many Americans simultaneously believe that existing training programs guarantee anyone a job and that people are poor due to lower-quality education, which appears to be self-contradictory but may in fact represent an awareness of obstacles for the poor in accessing existing training programs (Nilson, 1981). In addition, although the sheer number of media reports decreased during this time, they also became more varied and may have introduced the public to more sophisticated ways of thinking about homelessness (Buck et al., 2004).

Changes in views of the characteristics of homeless people may also reflect an increasingly complex view of the homeless population. Respondents in the 2001 sample believed that a greater percentage of homeless people received public assistance, eventually found a permanent home, and had finished high school. These changed estimates may also reflect a more optimistic view of the availability of effective assistance for homeless individuals. However, respondents may increasingly be holding the belief that homeless people are in some way personally flawed, as evidenced by increased estimates of mental illness and mental retardation. This may be indirectly linked to a positive view of the economy in the sense that respondents may assume that if people are unable to maintain a job and living space in such a positive economy, they must have some inherent personal weakness. It is, of course, possible that observed changes to some extent reflect actual changes in characteristics of the homeless population rather than simply changed perceptions. Unfortunately it was not possible to examine changes in accuracy of views of the homeless because of the lack of availability of sound data on the actual characteristics of the homeless population over the 8 years studied.

The relative importance of homelessness was contrasted with several other well-known issues. At the second time point, homelessness had increased in importance relative to cleaning up the environment and reducing the national debt, but decreased in importance relative to improving public education and strengthening the national defense. These findings suggest that rankings of social issues are subject to current events, such as a strong economy or a catastrophe such as September 11. For example, when the economy is perceived as strong, the reduction of national debt may be seen as less pressing. In addition, we found that respondents within the 2001 sample interviewed after September 11 were more likely to support national defense as more important than homelessness, as compared to those respondents interviewed before September 11. Note that the lack of change at that time on other social issues supports the view that public opinion changes only very gradually and generally does not respond dramatically to current events (Behr & Iyengar, 1985; Rokeach & Ball-Rokeach, 1989).

Predictors of Attitudes on Homelessness

Demographic predictors of attitudes appeared relatively stable across time points. In no case were there contradictory (and significant) findings across the two surveys. This pattern suggests that gradual changes over time in overall attitudes are not likely to extend to changes in the way demographics predict attitudes.

Demographic predictors pooled across both time points showed that female and younger respondents considered homelessness a more serious problem, were most sympathetic towards homeless people, were more likely to attribute their plight to economic factors and recommended more structural approaches to alleviating the problem. The “gender gap” in voting has been well documented, with women tending to vote Democratic versus Republican and supporting more liberal causes than men (Manza & Brooks, 1998). Our results are quite consistent with such findings. Toro and McDonell (1992) suggested that, in addition to their tendency to be more politically conservative, older adults in the United States may hold a more pessimistic view towards homeless people as a result of their personal experiences following the Depression. As they and/or others close to them had been able to cope with the hardships of this period, perhaps they believe that homeless people today must “choose” their lifestyles.

Findings on political conservatism were generally consistent with expectations. Thus, those reporting to be more liberal were more likely to support increased taxes and federal spending for homeless initiatives, showed more general compassion, and were more attuned to structural factors relevant to homelessness. Consistent with usual political platforms, politically conservative respondents were more likely to consider strengthening the national defense and reducing the national deficit as more pressing issues than homelessness. It is possible that the larger number of findings on political orientation, as compared to some earlier surveys (e.g., Toro & McDonell, 1992), reflects a growing polarization between political parties on a variety of social welfare issues, including homelessness (DiMaggio, Evans, & Bryson, 1996).

There were a number of other significant predictors of attitudes towards homeless people, although the findings obtained were not as clear-cut as for gender, age, political orientation, and income. African Americans, as compared to Whites or Others, demonstrated a generally more socially conscious and liberal orientation to the issue of homelessness, endorsing higher levels of Economic Factors as Cause, and being more likely to support increased federal spending to assist homeless people. Simultaneously, African American respondents endorsed more stereotyped views of homeless individuals themselves, and were more likely to view homeless people as having children and being mentally ill, mentally retarded, and substance abusers.

Effects found for education appear in conflict with those of Lee, Jones, and Lewis (1990) and Phelan et al. (1995). It appears possible that respondents at higher levels of education may have expected that homeless individuals shared the same opportunities for advancement that they had, decreasing their awareness of possible structural or societal influences that may have contributed to an individual's becoming homeless. A higher level of education was associated with a stronger belief in personal causes for homelessness and a weaker support for structural interventions to help homeless people. However, a higher level of education was also associated with a greater willingness to personally give to panhandlers and pay higher taxes, and fewer stereotyped views of the homeless as street people, alcoholic, or criminal.

Limitations and Strengths of this Study

The present study has a number of limitations that should be acknowledged. As noted earlier, it is likely that prevalence estimates are underestimates, due to the exclusion of individuals who are currently homeless and others who cannot afford a telephone. The study contrasted two large samples, maximizing the chance of finding statistically significant findings. Although MANOVAs were used to screen out findings non-significant at the multivariate level, many of the remaining effects, though significant, were small in magnitude. It should also be pointed out that the data in the present study came solely from the self-reports of respondents during a brief telephone interview. With regard to generalizations about changes in attitudes and values regarding homelessness, although this study uncovered changes in public opinion across time, it did not attempt to isolate the sources of these changes. As a result, although inferences about the influence of different factors over time may be posited, no causal pathways can be definitively asserted in this strictly correlational survey study.

The study also had its strengths. It assessed the prevalence of and public opinion on homelessness through two surveys of large, nationally representative samples. This approach can help understand factors that cause social change and cause public opinion to shift over time.

Conclusion

Overall, these results indicate that while prevalence of homelessness may be more stable than previously thought, public attitudes towards the social issue of homelessness does change gradually over time. Dramatic but not directly related current events such as September 11 appear to only have a narrow impact on perceptions of homelessness (e.g., as it ranks against the suddenly more pressing issue of national security). This pattern of gradual change in public opinion is consistent with findings for other social issues, such as inflation, unemployment, and desegregation (Behr & Iyengar, 1985; Rokeach & Ball-Rokeach, 1989). The public appears to hold increasingly complex views of the homeless population and factors contributing to homelessness. Advocates may take heart that the general public is moving away from old stereotypes about homeless people and may be increasingly willing to support new policy initiatives directed at increasing availability of low-income housing or essential services for homeless individuals and families. However, these findings are confined to a specific time period, and changes in the future may not continue in this direction. As new policy is proposed, it may benefit advocates to conduct further public opinion surveys to gauge the level of public support, as results from older surveys may not be as applicable in different economic times. In addition, attention to demographic differences in opinion may be instrumental in targeting new policy efforts. An older white male policy-maker who is wealthy and votes conservatively may represent one of the least sympathetic demographics with regards to the issue of homelessness. Arguments to persons from such demographic groups may “fall on deaf ears.” However, our data indicate that many if not most other constituencies are likely to be swayed into action by the high prevalence of homelessness that our nation has been tolerating and by the public opinion data that show that most other Americans remain concerned about homelessness and may be supportive of policy changes and structurally targeted interventions.

Footnotes
1

A MANOVA was conducted to examine sample differences on all available demographic variables. The 2001 sample demonstrated significantly higher levels of education (F = 13.52 (1, 760), p < .001), income (F = 12.03 (1, 760), p < .001), and liberal and independent political affiliation (F = 4.30 (1, 760), p < .05). While the education and income findings (not controlled for inflation) may reflect actual changes in the general population over this time period, (U.S. Census Bureau, 1994, 2000, 2002), the finding on political affiliation may simply reflect random sample fluctuation.

 
2

The survey instrument used and details on the factor analyses and many other aspects of this study are available from the authors.

 
3

To simplify presentation, only significant differences are presented in Tables IVXI. Full findings, including those that were non-significant, are available from the authors.

 
4

Gender was used as a dichotomy, with 20 respondents missing gender not included in analyses. Race was divided into categories of White and African American, with all other categories including missing values grouped into a third “Other” category. Education was operationalized as a continuum, ranging from 8th grade or less to graduate work. Political conservatism was divided into four categories encompassing “conservative,” “moderate,” “liberal,” and “other.” Income was also operationalized into three income categories based on the distribution of the samples between $7,500 to annual income over $50,000 (lower = under $25,000, middle = $25,000–$50,000, higher = over $50,000), plus a fourth category for those who refused to disclose their income.

 

ACKNOWLEDGMENTS

Funding for this research came from Wayne State University. Further information on this and related studies by our Research Group can be obtained from Dr. Toro and/or on our website at: http://sun.science.wayne.edu/∼ptoro/.

Copyright information

© Springer Science+Business Media, Inc. 2006