Towards a demographic profile of the street homeless in South Africa

Abstract

This paper provides a demographic analysis of the Human Sciences Research Council homelessness survey, and starts to develop a profile of the adult and child street homeless populations in the northern part of South Africa. Attempting to sketch the population dynamics of homelessness, the paper reviews research methodology, looks at the age, gender and employment status of the respondents, and identifies the foreign-born and citizen population elements. The data suggest that cross-border migrants represent a significant share of the street homeless, and particularly of street children. In addition, it appears that the child and adult populations are largely separate, with older street children leaving the streets to be replaced by in-migrating rural-born homeless adults.

Keywords:

• street homeless
• homelessness
• demography
• unemployment
• immigrants
• migrants

1. Introduction

This article looks at the population aspects of a multidisciplinary study of homelessness carried out by the Human Sciences Research Council (HSRC) between 2005 and 2008, and examines the relation between demographic factors and the phenomenon of homelessness. It considers briefly what the literature has to say, reviews research methods and begins to profile the street homeless population, using the HSRC 2007 survey data. The aim is to discover where the homeless come from and why the sad problem of people who have no housing or shelter persists in South Africa.

The aim of the demographic component of the study was to gain a clear picture of the underlying determinants of homelessness and the mechanisms by which the street population continues to draw in new members and replace itself as its existing members die or move into permanent homes. This could help us find better ways to prevent homelessness, if the research can identify the point in individuals' migration histories where the path forked and they took the route that led to a homeless state.

2. Literature review for the demographic profiling

Homelessness as a South African phenomenon is often attributed to poverty: that is, the high prevalence of poverty in this country is taken as the causal factor in determining the emergence and persistence of homeless people on the streets (Olufemi, 2001). However, little if any attention has been paid so far to the role of population mobility in the origins of the homeless populations in this country, although there are indications that some of these are relatively mobile, and that certain movements of homeless people may be characteristic in some South African cities (cf. Cross et al., 1998). However, since there are perhaps 10 million very poor people in South Africa, and of these it has been suggested that fewer than 50 000 are homeless (data from Olufemi, 2001), it becomes clear that in considering the origins and determinants of homelessness we need to take on board other factors and processes that lead to people being excluded, in addition to economic factors and poverty as such.

To date, there has been little research dealing with homelessness among adults in South Africa (Olufemi, 2000), although some studies of street children are available (Stavrou, 2001; Jackson, 2002). Olufemi's work from the University of the Witwatersrand (Wits) Department of Town and Regional Planning has been central to what is known of the South African metro homeless, and particularly those in Gauteng. For instance, Olufemi found that women were generally less likely to be homeless than men, but also that homeless women were more often subject to mental illness than men, and were also less likely to be living on the streets due to alcohol or substance abuse than were men (Olufemi, 2002). A study of homelessness was also commissioned by the Gauteng Economic and Social Development Units, but it appears to have yielded few results.

It is widely thought that migration of the poor from South Africa's rural areas is the most likely source of the metro homeless population. Press reports confirm the annual migration of Gauteng's metro homeless to the milder climate of Durban, which causes the population of Durban's main shelter to double during the winter season (Kirk, 2001).

3. Research questions

The key research questions addressed here are:

• Where are the homeless located on the urban–rural continuum?

• What are the demographic dynamics of the homeless populations?

• What proportion of the homeless in the study area has foreign origins?

4. Methodology

The obstacles confronting any study that tries to address the homelessness question in quantitative terms are formidable: probably no other kind of demographic population is as difficult to survey, and the problems involved are frequently discussed in the homelessness literature. A well-known commentator notes that:

One of the most difficult tasks in the study of homelessness is counting homeless individuals accurately … The problem with a census is that it is traditionally accomplished on the basis of domicile, or a dwelling, so that to count the undomiciled requires a different methodology. Also, the homeless themselves are a changing group of people … so that the problems of obtaining an accurate count of ‘the homeless’ are enormous. (Glasser, 1994:110; original italics)

Glasser continues:

Many homeless people, as a safety precaution, hide themselves from view. Others move around throughout the night, when street counts are typically performed … In spite of all the challenges, there is much to be learned … (1994:123)

Both qualitative and quantitative work was required to create a research design for the HSRC homelessness project that could tackle these obstacles. The parameters and dynamics of South Africa's homelessness problem were not well known when the HSRC undertook its large-scale homelessness project, and the institutional actors were not always clearly identified. In addition, it was clear from the literature that to undertake any quantitative survey would be extremely challenging. The first phase of the study began with networking, followed by qualitative fieldwork. After a stakeholder consultation workshop, extensive pilot work and further consultation with selected stakeholders, the study embarked on its second phase – a survey using a quantitative questionnaire.

4.1 HSRC qualitative field study

To discover the role that migration plays in the social and economic processes of homelessness, a series of 94 qualitative interviews on demographics and migration were carried out in Gauteng, Mpumalanga and Limpopo, along the migration corridor stretching from South Africa's eastern border to metro Gauteng, and these interviews were interpreted together with the qualitative cases data-set developed for the life histories research (Morrow, this issue). Responding to the interest of the national Department of Social Development, the demographic qualitative interviewing emphasised the rural sector, where we had reason to believe many of the street homeless population originated, but where homelessness had not been studied in South Africa. On the basis of the findings of the qualitative demographic interviews, along with the qualitative findings of the other HSRC teams who were part of the homelessness study, the parameters were established for the questionnaire survey sample that would cover both the rural and the metro urban sectors.

4.2 HSRC quantitative questionnaire survey

The survey attempted as far as possible to put numbers to the groupings and processes identified by the qualitative work. Some basic orientation was provided by the homelessness data collected by the national Census in 2001, although the data could not be used to produce a sampling frame due to the lapse of time.

It was clear from the homelessness methodology literature, and from the difficulties that the national Census faced, that carrying out a quantitative study of homelessness is among the most difficult survey tasks known. As Glasser (1994) notes, the street homeless have no fixed regular address and move around constantly in unpredictable ways, and are also inclined to hide from interviews; accordingly, it is difficult or impossible to have certainty ahead of time, when constructing the survey sample, about where homeless respondents will be found on the day of the survey when the interview teams are sent out. As a result, surveys of the homeless rarely have a reliable spatial sampling framework or a population structure to work from, and it is extremely difficult to be sure that all the homeless individuals in a given survey district have been identified and reached for interviews, without any respondents being counted twice.

While all possible precautions were taken for the 2007 homelessness survey, without a reliable baseline Census data-set to establish the true overall structure of the population it was not possible to design the samples to reflect the actual street homeless population structure and population distribution. Under these conditions the sampling assumptions were to some extent a best guess, and therefore it is not possible to guarantee the accuracy of the survey results in detail. Rather, the results reflect broad trends in the South African homelessness population as they are reflected in the northern part of the country.

4.3 Structure of the samples

The layout of the survey samples, which was developed to reflect the priorities and needs of HSRC's research partners and the conditions on the ground as reflected by the qualitative results, was not straightforward. The qualitative studies gave good indications of the presence and distribution of both urban and rural homeless people, making it clear that the homeless were not entirely a metro urban population, and that the rural homeless population was of a significant size. This work – together with the preliminary local survey carried out around the HSRC building in Pretoria by Aliber et al. (2004) before the main study began, and a 100-case metro pilot study conducted by two HSRC researchers, Kholadi Tlabela and Zakes Langa, early in the HSRC project – threw some light on the difficulties of identifying the areas in which homeless people could be found so that the field data could be gathered. Both these studies helped to contextualise and structure the survey sample, and to set the questions for the formal sample pilot study by the HSRC that preceded and laid the groundwork for the 1245-case HSRC homelessness survey itself. Without this preliminary information, the survey itself would not have been feasible.

However, the papers that came out of these studies were based mainly on qualitative material, and did not provide any hard numerical data that could be used to structure the statistical sampling and direct the field teams in detail across the entire sampling area. In addition, there was little available official national data based on either large surveys or the national Census that could provide a background population structure or indicate in which precise localities the South African homeless population was concentrated at the time of the main survey. The 2001 Census data-set was shown by the survey pilot study to be out of date, as the homeless in 2007 were no longer concentrating in the Census enumeration areas identified as homeless-frequented localities in 2001. This lack of valid and up-to-date statistical information about the size and whereabouts of the national homeless population made ordinary sampling procedures impossible, and places limitations on the detailed representivity of the final survey coverage.

To accommodate a rural homelessness sample and relate it to the core metro urban sample in Johannesburg and Pretoria, the sampling approach adopted was to look at the whole extent of the migration corridor stretching from South Africa's northeastern border with Swaziland, westward across Mpumalanga and southern Limpopo, into Gauteng. This selection emerged from the qualitative research together with the life histories recorded: it was based on the recognition that Gauteng is the magnet for the entire northern homeless population, with an attraction that goes beyond South Africa's borders. Rural migration is also a key issue – the metro homeless population originates largely in the rural areas.

To approach the homeless in individual localities across these broad areas, the study team, the research partner organisations, the study's methods consultant and the HSRC Socio-Economic Surveys section spent many months exploring and preparing for what would clearly be one of the HSRC's most difficult survey exercises. With the assistance of cooperating civil society bodies, the project team was able to develop a complex but largely effective survey methodology tailored to the needs of the homeless population in South Africa. Although it is not possible to guarantee detailed accuracy of the results, the survey does broadly reflect trends in the South African homeless population and will help to give scholars and government workers a general overview of South African homelessness.

Separate surveys were conducted among adults aged 18 or older and children aged 12–17. Totally different sampling approaches were adopted for the two surveys, mainly because of the different spatial distributions of the two groups. Furthermore, the children's sample was restricted to Gauteng because the expected small concentrations elsewhere in the study area did not warrant targeting children specifically, while the adult sample included Mpumalanga and Limpopo. This difference should be borne in mind when interpreting the results of the two surveys.

Accordingly, the adult sample included 940 adults, with 678 inside Gauteng, mainly in the metro (374) and sub-metro (304) sectors (see Table 1). In Mpumalanga and Limpopo, feeder areas for the metro homeless population concentrated in Gauteng, and another 125 adults were interviewed in the secondary cities, 82 in the small towns, and 55 in the rural areas comprising farms and former homelands. The overall sample was 79 per cent metro and 21 per cent non-metro. This distribution is in line with the way the sample was planned, making use of the qualitative findings. However, without reliable numbers for the parent population, there is no way to be certain that it accurately reflects the actual urban–rural balance of the national homeless population. Although the decision on allocating the sample's urban and rural coverage benefited from both the qualitative research and the audit pilot results, it does not in itself represent a statistical finding.

Table 1: Structure of the homelessness survey sample

Sex	Number of cases by sample category										Total of all respondents
	Metro: Gauteng and North West						Non-metro: Mpumalanga and Limpopo
	Metro		Sub-metro		Rural		Secondary cities	Small towns		Rural
	Adults	Children	Adults	Children	Adults	Children	Adults	Adults	Children	Adults
Male	306	211	271	50	–	20	122	62	3	53	1098
Female	68	19	33	2	–	–	3	20	–	2	147
Total	374	230	304	52	–	20	125	82	3	55	1245
Metro = metropolitan municipalities (Ekurhuleni, Johannesburg and Tshwane); sub-metro = non-metropolitan municipalities in Gauteng (e.g. Mogale City); rural = non-urban areas outside built-up cities or towns; secondary cities = large towns not classified as metropolitan areas (e.g. Mbombela (Nelspruit) in Mpumalanga); small towns = named urban centres not classified as cities, sub-metros or metros (e.g. Mdutjana in Mpumalanga).

Of all the respondents, both adults and children, 1098 in the total sample were male and only 147 female. There were 814 (87 per cent) adult men and 126 (13 per cent) adult women. This breakdown of the homeless population is in line with Olufemi's 2001 Johannesburg findings, which emphasised the relatively small numbers of women in the street homeless population at all levels, as well as the difficulty of making contact to interview homeless women, who often spend as much time as possible off the streets out of concern for personal safety. In spite of the best efforts of the field team to find more homeless women living on the streets in the sample areas selected, the final sample stood at a mere one-eighth (13 per cent) of adult women.

Twenty per cent of these homeless adult women were interviewed in the non-metro sector, outside Gauteng province. The equivalent figure for men was 22 per cent outside Gauteng, aligning with the overall non-metro share of the total sample. By comparison, the rural sample element for Gauteng itself is very small in respect of the overall sample, as Gauteng is an overwhelmingly urbanised province, and the homeless tend to congregate in the built-up areas.

The overall sample total for street children was 305, made up of 284 (93 per cent) male youth from the streets and 21 (7 per cent) female street youth. Since the street children's sample was almost entirely obtained inside Gauteng, this aspect of the sample distribution should not be taken to indicate that there were or are no street children in the rural sector: the qualitative results clearly indicate that this is not the case.

4.3.1 Use of data from the national Census

The data in Table 2 make clear how difficult it is to collect reliable data on the number of homeless persons in any area. Although unemployment has climbed in the interval, there is no clearly attributable reason why the 2001 figures should be about 10 times higher than those for 1996. In view of the possibility of measurement errors, differences in area definition, weighting differences and methodological changes introduced into the Census in the same period, the figures from the two censuses shown in Table 2 were considered inappropriate for estimating absolute numbers of the homeless or any certain trend.

Table 2: Numbers and proportions of homeless persons in the study area from the censuses of 1996 and 2001

Study area component	1996		2001
	Number	Percentage	Number	Percentage
Metro	483	0.03	5538	0.20
Non-metro	199	0.06	1295	0.38
Total	682	0.02	6833	0.22

5. Results

5.1 A demographic profile of the homelessness based on the HSRC survey sample

This section describes the basic demographic characteristics of the sample, in relation to some of the social and economic background factors.

5.1.1 Age and sex

The age–sex distribution is shown in Figure 1. The homeless population is clearly dominated by males. It should therefore be no surprise that only about one-eighth (12 per cent) of the 1244 respondents in the sample with known age and sex were female. Among the children (12–17 years old) in the sample, girls represent a mere 7 per cent, while the proportion of women in the adult sample is slightly higher at 13 per cent. The low proportion of adult women in the sample was despite specific sampling attempts to maximise their inclusion.

Figure 1: Age–sex structure of the homeless respondents in the sample

It is important to view the sample for children separately from and parallel to the sample for the adults, since the two were surveyed independently. The ages of the children should therefore be compared only with the ages of other children in the sample, and not with the ages of the adults. Similarly, the numbers of children should not be compared with the numbers of adults; in other words, Figure 1 does not reflect the share of children in the total street population.

In the sample population of street children most were in the older categories, showing a skew toward ages 16 and 17 years for both boys and girls that is not reflected in Figure 1. In the adult sample, the bulk of the street population was male and between the ages of 25 and 44, with a steep peak in the age 25–34 interval. There were few adult women in the sample, and the age distribution was flat by comparison, with a very low peak stretching across the years 18–44.

On the whole, the street homeless population is young, made up mainly of older children and working-age adults, and is predominantly male. From other sources with extended experience working with the street homeless population (see Co-ordinated Action with Street People [CASP], 2000), it is clear that in fact the population of street adults is notably larger than the population of street children. However, there were few elderly people of either sex, particularly elderly women, still on the streets.

Table 3 presents additional characteristics of the homeless adult population in the three provinces surveyed. The samples for Mpumalanga and Limpopo were substantially smaller than the sample for Gauteng. Levels of education were surprisingly similar in all three samples, clustering closely around seven years, with Mpumalanga's average very slightly higher than those of Gauteng and Limpopo. The Gauteng population seems to have no clear advantage over the more rural provinces in terms of realised education, and it also appears clear that the street homeless are not generally people who have no education. However, their educational levels are not competitive enough to enable them to enter the job market, which tends to require 10 or 12 years of education (Wittenberg, 1999).

Table 3: Profile characteristics, average levels for homeless adults by province

Province	Education level (mean years)	Mean percentage reporting employed	Mean percentage disability grant	Mean lifetime number of moves^a	Number of cases
Gauteng	7.04	27	1	1.82	683
Mpumalanga	7.30	40	0	1.54	118
Limpopo	6.99	18	2	1.50	142
Total	7.06	27	1	1.73	939
^aA ‘move’ in this article means a change in place of residence from one town or city suburb to another.

A fairly high share of the three populations combined (27 per cent) reported being employed, or working for pay; however, it is not clear how often the jobs referred to were regular or permanent employment. The CASP (2000) report from Cape Town points out that local businesses often benefit from the availability of street people to provide casual labour on a piecework basis.

By comparison, the share of the street adult population reporting that they received government disability grants was strikingly low at 1 per cent overall. Not one street homeless respondent in Limpopo said they received a disability grant. In the light of Seager and Tamasane's findings (this issue) that the street homeless are frequently subject to accidents and exposed to the risk of violence and often injured or disabled, this low figure appears to point to a shortfall in street homeless people's access to government support grants to which they are legally entitled.

Qualitative case material and reports from the homeless population in shelters suggest that the level of access to grants for the shelter population is much higher, largely because the shelter staff are able to act as intermediaries and help the homeless approach government offices and register for benefits. The street homeless, as a very poor or destitute population often of rural origin, are often afraid or unwilling to approach government offices or staff on their own; many indicated in interviews that they perceive their unkempt appearance and irregular situation as unsuitable and perhaps likely to expose them to rejection. For homeless street people who do not have anyone to help them fill in forms and approach bureaucratic staff, this diffidence contributes to cutting them off from grant access almost completely.

Lastly, Table 3 shows that the street homeless adult population is highly mobile, with most individuals having been active migrants. The overall average for number of moves from one area (e.g. town or suburb) to another was reported at 1.73, although this may be an underestimate since it is likely that people forget some of their moves. Nearly all of the adult homeless population appeared to have moved at least once, from their original place to the place of the interview, and many had moved more than once.

5.1.2 Family, children and friends

A large majority of adult respondents (70 per cent) had never been married. Almost one-eighth (12 per cent) said they were married at the time of the survey, 5 per cent lived together, 3 per cent were separated, 7 per cent were divorced, and 3 per cent were widowed. Two-thirds (66 per cent) of the adult respondents told the interviewers that they had regular companions or partners. Three-quarters (75 per cent) of these were friends, 16 per cent lovers, and 7 per cent blood relatives or spouses. Among the children in the sample, 78 per cent had regular companions or partners, and 84 per cent of these were friends, 12 per cent were relatives, and 3 per cent were lovers. More than three-fifths (61 per cent) of the adult respondents had children, but only about one-eighth (12 per cent) of those with children had them living with them at the time of the survey. Of those under the age of 18 years, 3 per cent had children of their own – and 29 per cent of these had them living with them.

Just over one-half of the adult respondents (51 per cent) did not have any financial dependants. What is noteworthy, however, is that 37 per cent of the child respondents in the sample said they shared money with or gave money to others. This finding might perhaps help to confirm the proposition that many children on the streets are not entirely homeless and are involved in economic activity on behalf of others. However, it is not clear that the recipients of this kind of contribution or support were family members – it is likely that in many cases they were companions or connections in the street community, or in some cases gang members or older children extorting protection money. For Cape Town, the CASP (2000) report emphasises the role of friendships and alliances among children and adults living on the street as a form of adaptation to street life, and the Gauteng Street Alliance report of 2005 notes the same kind of social mutualism among the homeless. These contributions could of course be a heavy burden for some of these children, and could sometimes compel them to resort to socially unacceptable activities in order to obtain cash to support others.

5.1.3 Urban and rural location and origin

Figure 2 shows that nearly one-half the adult respondents considered their family homes to be in rural areas of South Africa. If the figures are adjusted to leave out the respondents born in other countries and those who could not answer the question, then 58 per cent of street homeless metro adults count their home area as rural. In the non-metropolitan parts of the study area, the corresponding proportion among adults was 82 per cent. Taking into account all of the South African adults who answered this question, the share of rural home communities comes to 64 per cent.

Figure 2: Respondents' locality types at the ‘family home’

For the children's sample, which pertains to Gauteng only, the pattern was significantly different. Excluding the children born outside South Africa and the children who were not sure of their home community, just under two-thirds of the South African children who replied to the question said their family home was in the urban sector. Compared with the adults, the street children in the sample represent a mainly metro urban population, and would not have moved nearly as far as the adults before arriving on the streets at their place of interview.

5.1.4 Citizenship and migration

The role of cross-border migration in creating the present street homeless population in South Africa appears to be very significant. About one-seventh (14 per cent) of the respondents were citizens of other countries. Non-citizens represented 11 per cent of the adult sample, but in the sample of street children the proportion was as high as 23 per cent. Figure 3 shows these distributions. The large proportion of non-citizens among the children in Figure 3 stands out. From Figure 4 it is clear that a majority of the foreign-born children in the sample originated in Zimbabwe. Among adult non-citizen respondents the main countries of origin were Mozambique and Zimbabwe.

Figure 3: Citizenship of respondent

Figure 4: Countries of birth among foreign-born respondents

5.1.5 Sleeping locality

Table 4 shows that three-quarters of the children slept on the street (75 per cent), almost one-tenth slept in a park (9 per cent), and 4 per cent slept in a shelter. Table 5 shows that, of the children not sleeping in a permanent home, 27 per cent reportedly did not have a permanent home, 11 per cent were temporarily stranded, and for the majority (61 per cent) their home was too far away to commute. Compared with the children, a notably lower proportion of adults indicated that their homes were too far away to commute (40 per cent); this probably reflects the longer distances and resulting longer durations of absences from home. A slightly higher proportion of the adult sample slept in shelters (10 per cent).

Table 4: Where do you sleep?

Where do you sleep?	Adults		Children		Total
	Number	Percentage	Number	Percentage	Number	Percentage
Street	436	46	230	75	666	53
Shelter	97	10	12	4	109	9
Park	127	13	28	9	155	12
Friends, connections	34	4	2	1	36	3
Other	249	26	33	11	282	23
Total	943	100	305	100	1248	100

Table 5: If not sleeping in a permanent home, what is the situation?

Response	Adults		Children		Total
	Number	Percentage	Number	Percentage	Number	Percentage
I have no permanent home	316	34	82	27	398	32
I am temporarily stranded	208	23	33	11	241	20
My home is too far away to commute	366	40	185	61	551	45
Other reason	33	4	4	1	37	3
Total	923	100	304	100	1227	100
Number of cases missing due to non-response = 21 (20 adults, one child).

It follows that the sample for the HSRC survey appears to be correctly targeted and did successfully access the street homeless fraction within the total population on the Gauteng streets, many of whom may otherwise represent a day-stroller population that is not in fact homeless. Although some of the respondents reported that they slept at shelters, this fraction is small at 10 per cent for adults and 4 per cent for the children. These low proportions may point to an overlap and interchange between the homeless whose main home is the streets and the homeless who spend most or all of their time in formal shelters. If so, this category makes the point that the street and shelter elements of the homeless population are not entirely separate.

However, in addition to the children and adults who reported sleeping on the streets, with friends or in parks or shelters, nearly one-quarter in both samples said they slept in some other locality. The qualitative findings draw attention to such sleeping places as abandoned buildings or municipal facilities that are not closely guarded, but this category could include respondents living in formal housing or in shack areas, and commuting to the street localities where they were interviewed. The results in Table 4 suggest that very few of the survey sample in either group slept in permanent accommodation.

5.1.6 Homelessness status

Table 5 shows that only 27 per cent of the children and 34 per cent of the adults in the sample identified themselves to the interviewers as having no claim to a home at all. Another 23 per cent of adults and 11 per cent of children characterised their predicament as temporary, although without saying how long they thought such a situation might last. Implicit in this reply is the claim that the respondent does have a home, or a home place, but that for the time being they are unable to live there – a period of time that might often stretch into the indefinite future.

Nearly all of the rest, 40 per cent of adults and 61 per cent of the children, did not characterise themselves as being without a home, but said they were separated from any such home by travel distance that prevented active contact. Unpacking this answer is not easy: it acknowledges being out of regular touch with the family home, but leaves open the question of whether the respondent could claim to be a current member of the family household. That is, the respondent refuses to accept the stigma of homelessness, but acknowledges being away from home on at least a semi-permanent basis and leaves open the issue of having lost contact with or formally left the home family. The claim to have a home that they can rarely reach may therefore represent a hopeful wish that at some point in the future it may be possible to return to the family of origin, or at least to the place of origin.

Since most of the children in the sample and some of the adults are originally from localities within Gauteng, the distances involved are not always very great, although for the very poor even a fairly short trip is extremely expensive and may be impractical. Some of the respondents living on the streets may in fact actually be earning income, and may still retain their household-member status while returning home only very rarely, if at all. However, for most of the South African respondents, in light of their other replies, it looks likely that social membership of the family household will have lapsed, or have been cut off by a previous expulsion or by the extended duration of their absence without contact.

In effect, the children and adults who gave this kind of answer may have been putting a more hopeful gloss on their situation as they saw it and as they wished to present it to outsiders, while their actual situation may actually have been as far into real homelessness as that of the respondents who stated flatly that they had no home at all.

5.1.7 Duration of homelessness

One-seventh (14 per cent) of the adults in the sample said they had been homeless before the age of 18. Table 6 shows the mean number of years that they had been homeless. Adult respondents had on average been homeless for 6.4 years, almost one-fifth of their lives – for a 40 year old this would be nearly eight years. The mean duration of homelessness among children was 2.1 years, and on average this translates into one-eighth of their lives, which for a 17 year old would be just over two years.

Table 6: Mean duration of homelessness (years)

Survey	Sample size (n)	Mean	Standard deviation	Median	Minimum	Maximum
Adults	939	6.4	6.45	4.5	0.12	46.5
Children	305	2.1	2.22	1.5	0.02	17.5
Total	1244	5.3	6.00	3.5	0.02	46.5

6. Toward conclusions: Homeless population characteristics

The discussion in this article of the characteristics and dynamics of the homeless population, based largely on the HSRC survey findings, suggests some possible generalisations about the homeless population in Gauteng, Limpopo and Mpumalanga. For the national level of homelessness, the trends of the 1996 and 2001 Census findings for homelessness obtained from Statistics South Africa data cannot be tracked at all accurately or related to any specific determinants, given significant changes between the two census years in how the numbers were arrived at. However, the steep rise in the numbers of homeless people suggested by the 1996 and 2001 Census estimates may be plausible since the national homeless population was unlikely to have fallen over this period, and some significant rise was likely since the prevailing unemployment rates have put a strain on household incomes. The outlook for a sharp rise in homelessness from 2009 onwards, as the effects of the worldwide economic downturn take hold in South Africa, may be ominous.

6.1 Urban/rural location of the homeless

The results of the 2001 Census of the homeless population, which include both the homeless in shelters and the homeless sleeping on the streets, suggest that the national homeless population was largely in the urban areas, but also that the rural areas had a higher proportion of homeless people than the cities and that the rural homeless were scattered rather than concentrated. Without a reliable census enumeration to give the overall homeless population its true structure and show what overall share is located in the urban and rural sectors, the distribution of the HSRC's survey sample is entirely an artefact of what was assumed when the sample was constructed, and the actual distribution of the homeless population across the urban and rural divide cannot be reliably determined from the survey figures.

However, the Gauteng street homeless population is clearly the largest in the northern provinces, and the street homeless populations observed in non-metro localities during the intensive qualitative phase of the work were very small by comparison with the large and concentrated populations that the Johannesburg pilot reconnaissance reflected in specific metro areas such as Joubert Park. On the basis of the limited and incomplete Census 2001 figures for the homeless, it would appear that the metro urban component of the homeless population may be as high as 81 per cent, leaving a rural homeless population of about 20 per cent of the homeless total.

6.2 Demographic dynamics of the homeless grouping

It is clear that the street homeless population is predominantly male, including both children and adults. Women are comparatively few on the streets, which may be a consequence of the very difficult street living conditions for women noted by Olufemi (2001). Since the children were sampled separately, it is not possible to compare the age structure of the adults with that of the children, but the street children population is reliably reported to be much smaller than that of the street adults (CASP, 2000), and was estimated by both the HSRC and the Gauteng Street Alliance at roughly 3000. The overall national homeless population, including both urban and rural elements, was estimated from the HSRC survey data at perhaps as many as 200 000, although this estimate cannot be confirmed.

The adult male population was strongly concentrated among working-age men with moderate to low levels of education, but the women's population had a flatter age structure that included more youth and elderly, although the elderly were scarce overall on the streets. The street children appear in the sample as predominantly older children, close to the age at which they would no longer be able to use the children's shelter system.

At least one-fifth of the survey sample reported that they had been homeless even as young children under the age of 12. For the rest, homelessness appears to have occurred in late childhood or in adulthood. For the adults, homelessness in most cases had lasted for less than 10 years, although there was a small minority of long-term homeless in the population. The homeless death rate is unknown, as there are no statistics that identify the homeless who die on the streets or in hospitals.

The South African northern population of street people appears from the sample to be comprised mainly of single people (70 per cent), or at least unmarried people who do not have children staying with them. Although nearly two-thirds of the adult sample said they did have children somewhere, only about one in 10 – mainly women – had children living with them in the streets. In the children's sample, fewer than one in 20 reported that they had children of their own.

However, it is clear that the street homeless population is not entirely isolated in the sense that is widely believed. The street constituency may lack strong connections with the housed population in many cases, but it does tend to have strong links with others on the streets. Two-thirds of the adults and more than three-quarters of the children said they had regular friends or partners they associated with, although fewer than one in 10 in either group said they were in a sexual relationship with a partner. It appears that few children are born on the streets or to street homeless women, although it is likely that this does occur in a small number of cases.

This precarious street population may often include people attempting to support children and families, as well as themselves, by using the limited resources of the streets, but at the same time it is possible that most of the meagre resources the South African street homeless can mobilise in this way go to supporting themselves and their immediate networks of friends and associates. Nearly one-half of the adult respondents said they did have financial dependents, although it is not very clear what these links mean; many may not mean they are supporting absent families, and may instead mean network support or generational solidarity on the streets, or, depending on how the question is understood, even funds lost to regular extortion by individuals or gangs.

6.3 Cross-border migrants in the street homeless population

More than one in 10 adult street people, and nearly one-quarter of the street children, were foreign-born at the time of the survey in 2007. This is a remarkably high share, and is likely to be a fairly recent development, increasing as the economic crisis in Zimbabwe has worsened. From what is known of the current in-migration into South Africa, it looks likely that the present high levels of foreign-born presence represent a final effort by many families in the home country to place at least one migrant worker in the South African job market. As of 2009, this migration phenomenon is becoming visible: current press reports indicate more and more children from Zimbabwe crossing the border into South Africa alone.

If this is the case, then many or most of the cross-border street people on the streets at the time of survey may not be individuals who have fallen out of society and become stranded and isolated; instead, many of them, the children especially, may be migrant workers who have moved into the street economy of the homeless since they are unable to enter the formal job market because of their age or lack of documentation and may lack contacts for the informal economy. For any such purposive street migrants, street livelihoods may be a viable substitute for jobs since very small earnings can go further outside South Africa than inside the relatively expensive South African economy.

It is not clear what proportion of the street population of cross-border migrants have deliberately or reluctantly taken up a street livelihood as a last resort to contribute to the support of their families in Africa to the north, and what proportion are stranded individuals who have failed to find a foothold in the urban job market and largely lost touch with their families. Whichever category they belong to, the foreign-born component of the street homeless population in South Africa is likely to be the most dynamic, and perhaps volatile and rapidly changeable, component of the homeless population mix.

6.4 Origins of the street homeless population

There is perhaps one particularly salient conclusion to be drawn from the results of the demographic profiling exercise: the street homeless adults and children are not part of the same population. The adult homeless population is mainly of rural origin, and has migrated into the cities where it is now predominantly found, whereas the population of street children is primarily urban-born, and has not moved long distances to its street location. In other words, the homeless adult population does not exist mainly because street children have remained homeless on the streets as they aged into adulthood. Rather, the street children for the most part appear to leave the streets when they reach adulthood, and the adult street people mostly arrive later in their lives from the rural sector.

If this is the case, then the next question is where the fully grown street children go when they leave the streets. The available data are not much help in answering this question, as they do not cover the formerly homeless. It is possible to infer that many or most of the children follow the usual route for displaced township residents and move into the shack population. It is also possible that some find places in adult homeless shelters, or – if as adults they are able to find employment – that some are able to obtain rental accommodation or subsidy housing and rejoin the housed population. Others, especially if they are able to earn an income, may find places with relatives who have township houses or other housing. Not all the homeless youth do leave the streets, however. Some become homeless adults, and others die as a result of accidents, violence or untreated HIV (Seager & Tamasane, this issue). Others graduate from occasional petty crime into full-time membership of criminal gangs, and gain access to shelter through this connection.

For the adults, it appears that many or most exit the street homeless population after some five or 10 years; again, the exit mechanism is not clear, but it is possible that many become more or less permanent shelter residents, while others die quite young. While it would be reassuring to believe that at least some of the street homeless population successfully rejoin their families, there is little indication that this is happening, and it may be realistic to assume that any interventions to help the street homeless constituency will have to take measures to help them in place on the streets.

Acknowledgements

This work forms part of the Human Sciences Research Council's 2005–2008 study of homelessness. Funding from the National Department of Social Development, the Human Sciences Research Council, the Gauteng Department of Social Development and the Swiss Agency for Development and Cooperation is gratefully acknowledged.