Although the psychosocial profile of the homeless population has been investigated extensively, there are not many studies comparing the profile of homeless people to that of the general population and of vulnerable people (e.g. the poor) who are not homeless (Koegel and Burnam, 1988). One of the problems with that type of study is that the concept of vulnerability to homelessness is not easy to define. At first, research centreed almost exclusively on mental health disorders as risk factors (Fisher and Breaky, 1991). Further studies have identified a number of additional variables involved in the risk of becoming homeless such as, for example, the lack of robust family or social networks (Unger et al., 1998), the lack of schooling and job training (Brooks and Buckner, 1996 ; Stronge, 1993), employment difficulties (Epel, Bandura and Zimbardo, 1999), health problems (Ropers and Boyer, 1987), and stressful life events (Muñoz et al., 1999). It is also well known that socio-economic variables, especially poverty and difficulties to gain access to the housing market, may influence the onset of homelessness.

In order to determine which variables are most important and their exact role in homelessness, some researchers have begun to introduce control groups that are similar to the homeless group in some specific aspects. The majority of the studies focusing on those groups have analysed single women with and without children (Metraux and Culhane, 1999), the chronic mentally ill, people with substance-abuse disorders (Clark, Teague and Henry, 1999), and the elderly in high-risk situations (Keigher, Berman and Greenblatt, 1991). Although these populations provide an interesting contrast, they may differ from the homeless with respect to important key variables such as their socio-economic level.

The main objective of the investigation was to compare two vulnerable groups (i.e., homeless people and users of services for the homeless) with available data (on socio-demographic, employment, educational and health variables) from a representative sample of the Spanish general population (Juárez, 1994). Comparisons of data from homeless groups and the general population are relatively rare. Furthermore, it is also necessary to assess the specificity of the findings by including other groups of disadvantaged people. Thus, as in other comparative studies of vulnerable populations, we used a control group of people who have a regular place to sleep but who use services for the homeless (i.e., soup kitchens and public baths). The second control group made up of vulnerable housed people was included as a means of analysing factors that, besides economic difficulties, may differentiate the homeless population.

Subjects were defined as homeless if they were older than 18 years and if, on the day of the interview, they had spent the previous night in one of the following places: the street, a shelter for the homeless, or some other place unfit for living (i.e., the subway or an abandoned building). The ideal sample size, 300 persons, was determined by using the generic formula for random sampling of finite populations.

Following the strategy of previous studies of a similar nature (e.g. Wright and Devine, 1992), the strategy used to ensure a representative sample of the true homeless population in the city required making an exhaustive list of all the services specifically aiming to assist the homeless population (i.e., shelters, soup kitchens, clothing banks and public baths) in Madrid and then spending one night (S-night) interviewing the homeless on the streets in order to establish the frequency of their service usage in the previous four weeks. In our S-night study, most of the homeless people (97%) interviewed reported having used either a shelter or a soup-kitchen service within the previous four weeks. In line with these results, the final sample was selected exclusively from shelters and soup kitchens, and visited daily for four weeks. This method ensured that the sample would be representative with respect to the day of the week, the type of centre where the contact was made, and the size of each centre. Of the estimated sample, 54% of the interviews were to be conducted in shelters and 46% in soup kitchens. The final sample consisted of 289 interviews — 54.3% from shelters and 45.7% from soup kitchens. A weighting procedure was included: first, the data were weighted for the inverse of the estimated probability of selecting each sampled individual in each centre, and the probability of duplication (determined by the subject’s frequency of use of other centres or resources during the sampling period) ; second, the data were weighted to correct the difference between the ideal and empirical sample sizes. Thirty per cent of the potential subjects refused to be interviewed. Other investigators using this strategy in various European cities have found similar refusal rates (Marpsat and Firdion, 1998).

To ensure comparability, the variables selected for the study were those that were relevant to the condition of homelessness, such as socio-demographic characteristics, educational level, employment status and health status, in the FOESSA report (Juárez, 1994). Participants were given an informed-consent form that explained the need for the interview and the information the subject would be asked to provide, as well as the time commitment (approximately one hour) necessary for its administration.

Table 1 presents the sex and age composition of the three samples, showing significant differences between the homeless, the domiciled service-users, and the general population samples.

Table 1
Sex and age distribution of the three samples (in%)

These structural differences were controlled for by weighting the data on the homeless group and the service-user group using the age-sex structure of the general population. The results for these two groups are thus not the actual values observed, but they allow comparisons to be made between the three groups. The weighted percentages were then compared by means of a chi-square analysis. In order to determine the level of statistical significance, Bonferroni’s correction was applied.

Table 2
Marital status and living circumstances (in%)

The percentage of people from each group, given an age-sex structure identical to that of the general population, who live with their nuclear families or with part of their nuclear families (the “spouse and children”, “parents”, “only spouse” or “only children” categories added together) reaches 23.2% for the homeless group, 31.3% for the service-users group, and 83.4% for the general population. Only 10.6% of the general population live alone without children, whereas this is the case for 43.3% of the service-users group and for 68.3% of the homeless group. On the other hand, the service-users live up to five times more often with friends than the homeless people. This evidence seems to support the protective role of social relations in situations of vulnerability.

It appears that half of the homeless live on their own, which could reveal greater difficulties in developing and maintaining strong, stable family and social ties. According to certain authors, the lack of social support, revealed in the inability to establish personal or social relationships, is one of the most relevant factors causing or maintaining psychological dysfunction (Sarason, Sarason and Pierce, 1990). Marital and family relationships are the strongest ties people have within their social network. Relationship difficulties can lead vulnerable people (especially women, adolescents, the elderly, and others who lack sufficient resources) to become domiciled service-users rather than homeless.

The results of the current study concerning social support are similar to those found in previous investigations. For instance, Shinn, Knickman and Weitzman (1991) studied single mothers and discovered that those in domiciled service-user situations had more solid and stronger social networks than those who were in shelters. They concluded that social relationships may be a protective factor against severe socio-economic difficulties. In addition, studies in various cities of the United States have found that the homeless have much lower social contact rates when compared to poor domiciled service-user families (e.g. Sosin, Schwingen and Yamaguchi, 1993).

Table 3
Level of education (in %)

Regarding educational levels, the adjusted results of this study show that approximately 67% of the homeless group had completed at least part of their secondary education and almost 10% were found to have received a technical or similar diploma, or a university degree. In the domiciled service-users group, 75% had completed part of their secondary education, and 28% had received a technical or university degree. These high percentages could be due to the presence of the Latin American and eastern European immigrants who, after having studied in their native countries, became homeless or “at-risk” in Madrid. Whatever the case, these results do not differ much from those found by Padgett et al. (1995) in New York or by Ropers and Boyer (1987) in Los Angeles, where over half of the homeless were found to be high school graduates. For Spain, Cabrera (1998) reported 12% of his sample of homeless people to be university graduates, a percentage very similar to that of the general population, which was 11.2% in 1991.

Thus, although education has been deemed important to people’s personal and social development, the results of this study do not provide strong evidence to support the inference that this variable is a determining factor of homelessness or even of domiciled service-use. It seems that the educational system in Spain works well to ensure a basic level of mandatory education that does not vary much across groups.

Table 4
Current employment status and employment status one year ago (in %)

Only a small percentage (13.6%, adjusted for an age-sex structure identical to that of the general population) of the homeless group reported doing paid work (even one hour a day) the week before the interview, whereas this percentage reached 23.2% and 37.7% respectively in the domiciled service-users group and in the general population.

With respect to those who were unemployed and looking for work, the percentage was similar for the homeless and domiciled service-users, both currently and one year ago. The increase in the proportion of people who were currently unemployed but not looking for a job in the intervening year was higher for the homeless than for the other two groups, a likely sign of discouragement and hopelessness. In addition, the only statistically significant within-group difference in this variable was found in the homeless sample because of the increase in comparison to the previous year.

The majority of the disadvantaged subjects, especially the homeless, reported their current situation as significantly worse than their situation a year ago. They reported that they were working less, looking for work less, and were less motivated. As the domiciled service-users become more discouraged and stop looking for work, their profile becomes more and more similar to that of the homeless group.

Table 5
Perceived overall health status (in %)

Table 6
Perceived overall health status (in %)

As seen in Table 5, the homeless perceived their overall health as much worse than did the service-users and the general population. Individuals from the homeless group (HG) tended to provide more extreme answers on the health scale (considering an age and sex structure similar to that of the general population). Table 5 shows that the perceived overall health status is the worst in the homeless group, slightly better in the service-users group, and clearly much better in the general population.

Regarding specific health problems and health habits (see Table 6), it becomes clear that the disadvantaged groups have higher percentages of subjects reporting poorer health on almost all specific concerns (i.e., dental problems, sleeping problems, back pain, or aching bones). Smaller percentages of the homeless group reported using external aids to treat their health problems (glasses, contact lenses and dentures, for example), which seems to indicate that this group has less access to these types of services.

As in previous studies (De la Rochère, 2003), in the evaluation of specific health concerns, the homeless appear to be similar to the domiciled service-users group except in their access to certain services and external aids (as shown by the percentage who use glasses or contact lenses and dentures). Dental and ocular services are not typically covered as part of Spain’s socialized health-care plan, and this means that most citizens must obtain private health insurance to help pay for medical assistance and follow-up in these specific areas. The results can be interpreted as a function of economic means: fewer homeless have access to the types of services and aids that require payment. The homeless group showed the highest frequency of smoking, a result that is congruent with other studies conducted with homeless people (Connor et al., 2002). In regard to the frequency of alcohol consumption, our data show small differences among groups in daily consumption. Although most of the research in this field has shown disproportionate alcohol consumption in the homeless population compared to the normal population (e.g. Koegel and Burnam, 1988 ; Muñoz et al., 2002), it is important to take into account that the item included in our study measured frequency, not the amount of alcohol consumed each day.

The differences discovered between the homeless and domiciled service-users group in the education variable contrast with the results of previous studies. Shinn et al. (1991) found that a larger percentage of homeless women had completed some studies after high school, compared to poor at-risk mothers. In Spain, Cabrera (1998) reported a larger percentage of subjects with secondary studies in the homeless group, but not in the at-risk sample. In this sense, the role of education in the exclusion process of the homeless deserves more attention and may require other types of more qualitative tactics that would allow researchers to better identify the relations between the different variables involved (e.g. life course analysis).

In regard to the employment situation, it is interesting to note that unemployed domiciled poor people show the same eagerness to look for work as homeless people. Yet the adjusted proportion of the unemployed not actively looking for work is higher among the homeless. This differential pattern may provide an interesting insight into the dynamics of various vulnerable populations, and service-providers should be aware of these differences in order to implement well-differentiated intervention programmes for users of services (i.e., homeless people and poor domiciled people) who may have different needs and be at different stages of a common pathway of discouragement (e.g. actively looking for a job or not) and social exclusion. Future studies might specifically address these issues using longitudinal methods.

Therefore, it seems that there are clear differences between the groups. The group of users of services for the homeless has not received much attention as a differentiated group but, according to these results, it deserves to become a target of specific interventions. Although similar to the homeless group in a number of characteristics, these people still maintain more social ties and more social support and have a better but apparently more fragile employment situation. The general picture emerging from this study is that the domiciled people using services for the homeless seem to be at a mid-point (between the general population and the homeless group) and, as a group, may be moving toward worse conditions.

Lastly, we should mention that this study has some limitations. First, it is a cross-sectional study and, therefore, no causal pathways can be established. Second, it is based on self-report measures that, although valid, may not gather relevant information in specific domains (e.g. health status). Nevertheless, this study reflects a further step in trying to characterize specific vulnerable populations that, as we have shown, may have similarities but, at the same time, relevant differences.