Seen, Not Sheltered: Homelessness and the Limits of Being Counted

A Photograph That Did Not Need Updating

There is an image that keeps recurring in South Africa's public memory of its own homelessness crisis: people sleeping directly outside the gates of Parliament, in Cape Town, in the exact precinct where the laws governing their situation are written. Journalist Athi Mtongana captured one version of it in 2019, describing simply what was already visible to anyone walking past, that homelessness continues to grow across the country in full view of the institutions meant to answer for it. Seven years and one national census later, that image has not required updating. Only the number attached to it has become more precise.

The Contradiction That Demands an Answer

In June 2024, the United States Supreme Court ruled, in City of Grants Pass v. Johnson, that a municipality may enforce anti-camping ordinances against homeless residents even where no shelter bed exists for them to occupy instead. That same year, United Nations human rights experts issued a direct rebuke of the pattern the ruling endorsed, urging governments to urgently scrap unfair laws criminalising homelessness and poverty (OHCHR, 2024). Neither statement described a new phenomenon; both named, in the language of the present, an arrangement South Africa had already perfected considerably earlier, for reasons that predate either statement by roughly two centuries.

For most of the democratic era, South Africa's homelessness crisis existed largely as an argument about missing data. That excuse closed in February 2025, when Statistics South Africa released its Profile of Homeless Persons in South Africa, 2022, the country's first dedicated national homelessness count, recording 55,719 homeless people nationally, more than four times the 13,135 counted in 1996 (Stats SA, 2025). The report's own causal breakdown reads as an indictment of policy neglect in its own right: unemployment was cited as the primary reason for homelessness by 54% of respondents, more than five times the next most common cause, drug and alcohol use, at 11.7%; lack of income accounted for a further 8.4%, family conflict for 8% (Stats SA, 2025). The population counted was disproportionately male (70.1%) and disproportionately black (76.7%), concentrated most heavily in Tshwane (18.1% of the national total), followed by Johannesburg (15.6%), Cape Town (11.9%), and eThekwini (10.1%) (Stats SA, 2025).

None of this should have surprised anyone who had read the preceding decade's unemployment figures. Cross, Seager, Erasmus, Ward and O'Donovan (2010), in the foundational academic review of the subject, had already identified unemployment and the erosion of the informal kinship safety net as the dominant drivers of street homelessness fifteen years before the state produced a matching number, describing South Africa's street homeless in language considerably more urgent than any government report has since managed: the proverbial skeletons at the feast, the excluded poorest who enter unobserved and stand by gaunt and starved, terrifying to the invited guests but deprived of any capacity to join the party (Cross et al., 2010; Cross & Seager, 2010; Naidoo, 2010).

This is not a country without evidence. It is a country that took nearly three decades to formally admit what its own researchers had been telling it since at least 2010.

Even the Number Is Contested

Before this analysis leans further on the 2022 figure, the figure itself deserves scrutiny, because the research gap here is not incidental; it is structural. U-Turn Homeless Ministries, one of the largest service providers operating in Cape Town, has publicly disputed the Stats SA count as an undercount, estimating the true figure for Cape Town alone at closer to 20,000, well above the roughly 6,700 the national census would imply for that metro on its 11.9% share (U-Turn Homeless Ministries, cited in AllAfrica, 2024; Stats SA, 2025). Earlier estimates from The Borgen Project (2024), compiled before the census existed, had put the national figure as high as 200,000; a number the 2022 count does not so much correct as radically undercut. Some of this gap is methodological, and defensible: point in time street counts systematically miss people in transitional housing, couch-surfing, or moving between provinces, a limitation the census methodology does not fully resolve (Naidoo, 2010; Du Toit, 2010). Some of it, though, is simply institutional distrust, earned over decades of undercounting, that a single well-designed survey cannot immediately dissolve. The state now has a number. It has not yet earned full confidence in it.

What the Shelters Cannot Absorb

Whichever figure is closer to correct, the practical shortfall is not in dispute. Of the 55,719 people the state has counted, only around 20% were found to be in a shelter on the night of enumeration (Stats SA, 2025; EWN, 2026). The City of Cape Town's own facilities offer a fraction of the beds its service providers say the metro needs, and shelters including The Haven reported running at 85% capacity even before winter demand peaked, with waiting lists standard practice rather than exception (AllAfrica, 2024). Shelters in and around Johannesburg reported turning people away entirely, citing capacity constraints and funding shortfalls that have not moved in step with the population the census confirmed (EWN, 2026).

This is the arithmetic underneath the policy failure: even a state fully committed to sheltering everyone it has counted does not currently have anywhere near enough beds to do so.

The Colonial Throughline

None of this fragmentation is incidental. It is, in a specific and well-documented sense, inherited.

Magnus Killander, writing in the South African Journal on Human Rights (2019), traces South Africa's vagrancy legislation back to 1809, through the Cape colonies, the Boer republics, the Union, and the apartheid government, arguing that these laws were adopted, consistently across each successive regime, with the explicit aim of subjugating the indigenous population, and functioned throughout as an instrument of social control over the poor, applied with deliberately racist implementation (Killander, 2019; Holness, 2020). The Cape's Vagrancy Act of 1879, amended in 1889 and folded into the Vagrancy and Squatting Act of 1895, alongside Natal's own 1869 Vagrancy Act, established the legal architecture criminalising the mere presence of a poor person in public space more than a century before any current municipal by-law was drafted (Killander, 2019; Mashau, 2022).

Killander's central finding is that this architecture did not end with apartheid. It persists today, structurally intact, in the municipal by-laws of Johannesburg, Tshwane, eThekwini, and Cape Town, each empowering officials to fine, arrest, or remove people for sleeping, washing, begging, or sheltering in public space (Killander, 2019; Holness, 2020). Holness (2020), examining eThekwini's by-laws independently in Law, Democracy and Development, reached the identical conclusion: the colonial project appears to be perpetuated in by-laws that criminalise poverty. What changed between the nineteenth century and now is largely the vocabulary. Vagrancy became public nuisance. Removal became reintegration. The underlying mechanism, making a poor person's presence in public space the offence, rather than the conditions that produced it, has simply been carried forward and relabelled (Killander, 2019; Holness, 2020).

The Cape Town Cases

Nowhere is that continuity easier to trace in real time than in a single, still unresolved legal contest brought by the Ndifuna Ukwazi Law Centre.

In 2021, eleven homeless applicants, since referred to in the litigation as the Gelderbloem applicants, challenged the City of Cape Town's Streets, Public Places, and Noise Nuisances By-law (2007) and its Integrated Waste Management By-law (2009) as unconstitutional and discriminatory. The applicants won: the court found the City was misusing its own by-laws to criminalise the poor, a ruling consistent with the constitutional concerns Killander (2019) and Holness (2020) had each already raised independently about this legal architecture. One applicant had been fined R300 for sleeping rough in District Six. The same month the ruling came down, the City amended the by-law, not to remove its punitive core, but to permit police to arrest homeless people who declined an offer of shelter; precisely the pattern of legal defeat followed by administrative reformulation that Holness (2020) had already documented in eThekwini.

The applicants returned to court the following year, this time also challenging the City's practice of confiscating belongings during so-called clean-up operations, after officials cleared people from the Castle of Good Hope and Culemborg without prior notice. By 2024, the City proposed a further amendment; Ndifuna Ukwazi noted that while the revised draft removed the explicit arrest power, it continued to criminalise anyone who declined "alternative shelter," a category the organisation's clients argue is frequently inadequate or unavailable at the precise moment it is offered, an argument the shelter capacity figures above make difficult to dismiss as rhetorical.

A related, procedurally distinct matter reached the Constitutional Court in December 2024: Charnell Commando and Others v City of Cape Town, concerning the Bromwell Street residents facing eviction into homelessness, and whether the City bore a constitutional obligation to secure emergency accommodation before that eviction proceeded. The mechanism differs, private eviction rather than direct by-law enforcement, but the underlying question is the same one running through Gelderbloem and the wider academic literature alike: what, precisely, does the state owe a person who is about to become, or already is, homeless (Killander, 2019; Holness, 2020; Naidoo, 2010)?

Three rounds of litigation. Three rulings or amendments. The same underlying refusal each time, documented in the academic record years before this specific case existed: to treat a homeless person's presence in public space as a housing failure rather than a policing opportunity (Killander, 2019; Holness, 2020; Cross & Seager, 2010).

The Policy That Still Does Not Exist

Here is the finding that should end any debate about whether this is a resourcing problem or a political will problem: as of 2025, South Africa has no national policy framework on homelessness at all. Pophaim (2025), writing in Social Sciences, states this without qualification: homelessness remains a neglected and under-prioritised area of policy intervention in South Africa, and without a national policy framework, it is generally overlooked, resulting in intermittent, fragmented and ineffective responses (Pophaim, 2025). Before the Covid-19 pandemic exposed the consequences of this gap, only three municipalities in the entire country had any homelessness policy at all: Cape Town (approved 2013), Tshwane (2015), and Johannesburg (still in draft as of 2022) (De Beer & Vally, 2021; Muleya & Mlilo, 2023, cited in Pophaim, 2025). The national Department of Social Development confirmed as much directly in 2021, stating plainly that no national department was responsible for tackling homelessness, meaning no dedicated action plan and no ring-fenced budget existed for the population the state would only formally count four years later (Department of Social Development, cited in Daily Maverick, 2021).

There is movement, but it remains pre-legislative. A national diagnostic study was commissioned (Sonko-Najjemba et al., 2022), and a working group led by the Human Sciences Research Council was appointed to develop a Draft Green Paper on Homelessness (HSRC, 2024). A Green Paper is not policy; it is the document that precedes a White Paper, which itself precedes legislation, a process South Africa's own 2024 White Paper for Human Settlements, gazetted in January 2025, took years to complete for the housing sector broadly, and that document does not treat street homelessness as a distinct category requiring its own intervention at all (Republic of South Africa, 2025; Royston, Mahlangu & Mohapi, 2024). Housing policy has a white paper. Homelessness, the more acute condition, has a green one still being drafted.

What X Is Saying, and Why It Constitutes Data

Public discourse on homelessness in South Africa is intermittent. It surfaces around a specific image, a specific number, or a specific court ruling, and recedes until the next one arrives. But when it surfaces, its qualitative texture tells you something the census cannot: not just how many people are homeless, but how the public has been narrating that fact to itself for the better part of a decade, often with more analytical precision than the policy record it is reacting to.

That single rewritten sentence is, in essence, the argument this piece has been building toward, stated in fourteen words rather than several thousand. It is also, unintentionally, an accurate summary of the by-law history Killander (2019) and Holness (2020) document: municipal policy in South Africa has overwhelmingly chosen the first framing over the second, repeatedly, across more than a century of legislative continuity.

This is not a lay paraphrase of the academic literature; it is the academic literature, restated by the litigating organisation itself in the same register Killander (2019) uses in the peer-reviewed original. When an advocacy group's public messaging and a law journal's constitutional analysis converge on identical language independently, the claim has moved well past interpretation into settled fact.

Some of the public record is less an argument than a data point in its own right. Athi Mtongana's 2019 photographs of people sleeping outside Cape Town's parliamentary precinct, captioned only with the observation that homelessness continues to grow across the country, function as a kind of accidental fieldwork: the same institution empowered to legislate a national policy on homelessness has spent years unable, or unwilling, to resolve the presence of homeless people at its own front gate.

The arithmetic in that post holds up against the National Treasury's own published allocations. A commitment of roughly 40,000 social housing units spread across ten years, set against a temporary housing population already exceeding 135,000 and rising annually, is not a rounding error; it is a programme sized for a fraction of the problem it is named after (National Treasury, 2025). The public did the division before this analysis did.

Not every public voice reaches for policy language. Some reach for the older, blunter register of corruption.

This framing sits alongside a housing delivery record that Treasury's own Human Settlements vote acknowledges is under sustained pressure, with informal settlement upgrading grants to provinces falling at a significant average annual rate over the current medium-term expenditure framework even as underlying need has grown (National Treasury, 2025). Corruption and underfunding are not competing explanations here; they are compounding ones.

A more troubling thread runs through the discourse too, one this platform's methodology requires naming rather than filtering out.

The Stats SA (2025) data only partially supports this framing. Drug and alcohol use was cited as the primary reason for homelessness by 11.7% of respondents; a real and non-trivial factor, but well behind unemployment's 54%, and comparable in scale to lack of income (8.4%) and family conflict (8%). The public's dominant explanatory frame for homelessness and the state's own measured one are not the same story; the gap between them is itself a policy communication failure worth naming.

Thapelo Phetlo's case, reported six years before Stats SA (2025) confirmed unemployment as the leading national cause at 54%, is what that statistic looks like from inside a single, named life. The census gave the country a percentage. Journalism had already given it a person.

This is, in essence, a lay restatement of Killander's (2019) entire thesis, arrived at independently, a year before most South Africans would have had reason to read a law journal article on municipal by-laws. The convergence between a reply-thread post and a peer-reviewed constitutional analysis, on this exact causal claim, is not a coincidence worth dismissing. It is what happens when lived experience and academic research describe the same structure from opposite ends.

The Gap Named

Place the census, the litigation, and the absent national framework beside one another, and the shape of the failure sharpens past argument. South Africa now knows, with more precision than at any point in its democratic history, how many homeless people it has, where they are concentrated, and, by its own data, overwhelmingly why (Stats SA, 2025). It knows its shelters can absorb perhaps one in five of the population it has counted (Stats SA, 2025; EWN, 2026). It has, in Pophaim's (2025) own words, no national policy framework governing any of it. And in the one domain where the state has moved with real institutional speed, repeatedly, across four Cape Town legal defeats in three years, that speed has been spent narrowing the definition of an acceptable municipal by-law, not building the shelters, the transitional housing, or the employment pathway its own statistics say the crisis requires (Killander, 2019; Holness, 2020; Stats SA, 2025).

What Needs to Change

Three things, none of them beyond the state's demonstrated administrative capacity when it chooses to exercise it. First, the HSRC's Draft Green Paper (2024) needs to move to a binding national policy on a fixed timeline, not the indefinite consultative drift that has characterised this process since the Department of Social Development first admitted, in 2021, that no such framework existed (Department of Social Development, cited in Daily Maverick, 2021; Sonko-Najjemba et al., 2022). Second, that framework needs to explicitly prohibit the criminalisation of sleeping, sheltering, or existing in public space, closing the loophole Cape Town alone has redrafted three times since 2021, rather than leaving the matter to be re-litigated indefinitely, city by city (Killander, 2019; Holness, 2020). Third, shelter capacity needs to be funded to match the census finding, not the pre-census assumption: if the state's own count implies a substantial shortfall in beds relative to documented need, the National Treasury's Human Settlements and Social Development allocations need to reflect that arithmetic directly, rather than the current pattern of city by city, NGO-dependent capacity building (National Treasury, 2025; AllAfrica, 2024; EWN, 2026).

A Final Word: On Being Seen

"We are invisible on the street," one man told a gathering of homeless South Africans and the researchers documenting their testimony in 2024. "Most people pretend they don't see us. Others treat us all like criminals" (cited in De Beer, AllAfrica, 2024). No government official attended that gathering. If you have ever stepped around someone sleeping in a doorway and felt the specific, fleeting discomfort of being briefly implicated in something you did not personally cause, you have already done more than the state's own conference attendance record managed that year. The state has, at considerable administrative cost, finally learned how to count this man. It has not yet learned how to see him the way he is asking to be seen: not as a nuisance, not as a case, but as someone it owes something to, simply by virtue of having been found.