Invisible Borders — Persisting Scars

“Of all the manifestations of inequality and oppression under apartheid, none was as stark or potentially as enduring, as the territorial separation of people along racial lines”

— Edward Patrick Lahiff (1997: 10)

The Ten Bantustans

Bantustans as “Nation States”

Transkei

Bophuthatswana

Venda

Ciskei

KwaZulu

Lebowa

Gazankulu

QwaQwa

KwaNdebele

KwaZulu: The Largest Homeland

47% lived in bantustans in 1991; 29.5% still do today

Population density (UNPD 2022). Yellow borders: all bantustans. Blue border: KwaZulu.

Internal Migration

Change in Grade 2 learner share, 2001–2021. Red = out-migration. Green = in-migration. Blue borders: former bantustans.

The Institutional Archaeology of a School System

Every school in KwaZulu-Natal, coloured by the apartheid-era department that created it. Red = KwaZulu Bantu Education.

This Paper

Having been educated in a former bantustan in 2021 causes a 10% reduction in school completion — and this gap is persistent, not converging.

Three contributions:

Causal identification of persistence — using a geographic regression discontinuity design along the former KwaZulu border, comparing schools mere kilometres apart on either side of a line that no longer formally exists.
The persistence finding — the scar is not healing. Thirty years of democracy have not closed the gap that apartheid engineered.
Mechanisms — exploiting variation in exposure to Bantu Education across age cohorts and differences in contemporary traditional authority regimes to isolate how persistence transmits.

Why Do These Borders Persist?

Theoretical Framework

Two Literatures, One Border

Institutional Persistence (Economics)

Acemoglu, Johnson & Robinson (2001); Dell (2010); Dell & Olken (2020).

If the Peruvian Mita, abolished in 1812, still causes lower consumption today — what of institutions abolished only 30 years ago, many of which were never actually dismantled?

Decentralised Despotism (Political Theory)

Mamdani (1996); Chatterjee (1993); Mbembe (2001).

The colonial state ruled the countryside through a “clenched fist” — fusing judicial, legislative, and executive power in the chief. The post-colonial state inherited this architecture. South Africa is no exception.

This paper bridges both: causal measurement of institutional persistence, explained through the theoretical architecture of decentralised despotism.

“The authority of the chief thus fused in a single person all moments of power: judicial, legislative, executive, and administrative. This authority was like a clenched fist… a regime that breathed life into a whole range of compulsions: forced labour, forced crops, forced sales, forced contributions, and forced removals.”

— Mahmood Mamdani (1996: 23)

What Was Never Dismantled

The 1994 transition abolished the formal bantustans. It did not abolish the governance architecture that made them function.

Inherited intact:

Native Law → Chapter 12 of the democratic Constitution
The Chieftaincy — strengthened by the Traditional and Khoi-San Leadership Act (2019)
Communal land tenure — no private title in former homelands
The Ingonyama Trust — ~30% of KwaZulu-Natal held under sole trusteeship of the Zulu monarch
Spatial inequality — no systematic place-based reparations programme

The distinction that matters:

Apartheid redress has primarily focused on demographic representation (BEE, employment equity).

Meanwhile, urban bias—common across much of Africa—has further marginalised these rural peripheries.

This misses the spatial dimension of apartheid harm: 29.5% of South Africans living in places that were designed to produce disadvantage — where the institutional design has not been undone.

Measuring What Has Been Theorised

Wolpe (1972): The homelands subsidised white capital by forcing reserves to bear the costs of social reproduction. When subsistence collapsed, the state shifted from exploiting pre-capitalist economies to enforcing subsistence through coercion.

Bundy (1979): The rise and fall of the South African peasantry was not a natural process — it was produced by land dispossession and fiscal extraction.

Murray (1981): The migrant labour system destroyed family life — 80% of men aged 25–50 absent from the homelands by 1990. Today, 65.6% of children in former homelands are raised by one or fewer parents.

This paper: These mechanisms are not only theoretically coherent — they are empirically measurable, they are causally identifiable, and their effects are growing.

The Design of Dispossession

Making land insufficient:

1913 Native Land Act — 7% of the country for Black residence
1936 Native Trust & Land Act — expanded to 13.7%, where it remained
Sharecropping abolished; subsistence destroyed
Hut taxes (1857) forced market participation, creating bureaucratic legibility (Scott 1998)

Making labour cheap:

Families in the reserves provided housing, childcare, subsistence — allowing capitalists to pay below the real cost of labour reproduction (Wolpe 1972)
By 1990: remittances = ¼ of migrant income, but 45–60% of homeland GNP

Making control total:

Influx control — preventing Black urbanisation
Communal tenure — preventing both commercial farming and a landless proletariat
Forced removals — 3.5 million people displaced (1950–1980)
Denaturalisation — the 1970 Act stripped SA citizenship entirely

“Too poor to support independence. Too distant to threaten white cities. Just close enough to supply labour.”

— and this was by design, not accident.

From Exploitation to Enforcement

By WWII, subsistence agriculture in the reserves had collapsed. The organic wage subsidy no longer functioned.

“The attempt to retain the structure of the ‘traditional’ societies… for the purposes of reproducing and exercising control over a cheap African industrial labour force… by the enforcement of low levels of subsistence”

— Harold Wolpe (1972)

As rural subsistence collapsed, the bantustans evolved to facilitate absolute spatial control.

The inherent contradiction: Maintaining this control required the systemic educational deprivation of Black South Africans, even as capital increasingly demanded a more educated industrial workforce. Total control came at the cost of education, which came at the cost of higher-skilled labour.

Bantu Education: The Mechanism of Transmission

The Bantu Education Act (1953) restructured Black education to produce compliant labourers.

	Black	White
1953	R17.99	R127.84
1963	R11.56	—
1968	—	R144.57

Per capita Black education spending fell after the Act.

“There is no place for [a Black man] in the European community above the level of certain forms of labour”

— Hendrik Verwoerd

In 2018: 78% of white students received a bachelor’s pass vs. 23% of Black students.

This is the mechanism of intergenerational transmission: parents subjected to Bantu Education pass the human capital deficit to their children. The funding gap compounds across generations.

The Natural Experiment

Identification Strategy

The Border as Natural Experiment

The former KwaZulu border creates a natural experiment: schools that are geographically adjacent — sharing the same climate, terrain, labour market, and ethnic composition — but that fell on different sides of an administrative boundary that determined which education department ran them, which governance regime applied, and which institutional infrastructure was built.

The strategy:

Compare schools within a narrow bandwidth (10km) on either side of the former KwaZulu border
Schools just inside: administered by KwaZulu Bantu Education, governed by the chieftaincy, on communal land
Schools just outside: administered by the Natal Education Department, governed by provincial law, with freehold tenure possible
Same geography. Same people. Different institutions.

A geographic regression discontinuity design, double-differenced against borders that were never homeland borders.

The 10km Bandwidth: Inside vs. Outside

Dark red: KwaZulu core. Light red: 10km inside the border. Medium red: 10km outside the border.

Were the Homelands Inherently Marginal?

A key concern: perhaps the borders were drawn around land that was already poor — and the persistence reflects geography, not institutions.

We test this directly. Using the same RDD:

On average, the former homelands receive better rainfall and have better soils than their provincial averages
But topsoil degradation shows a sharp discontinuity at the border — it is worse just inside
This degradation is a consequence of overcrowding and institutional neglect, not a pre-existing condition
Geological features (elevation, slope) show no discontinuity

The lands were not inherently marginal. They were made marginal.

“It would be difficult or impossible to assign to the natives such locations of an extent sufficient for their support…”

— Earl Grey, 1849

The Scars

Results

Education at the Border

NSC Mathematics

NSC Geography

Orange = outside the former homeland. Teal = inside. The dashed line is the former border.

The Scar Outlived the Curriculum

Students who entered school after 1994 score ~10 points higher on NSC exams than those who entered under apartheid.

But did this improvement close the homeland border gap?

	Est.	SE	p
Border gap change (post − pre reform)	−1.54	0.82	0.058^†

^† p < 0.10. N = 1,585 schools within 4.3 km of the border.

We compare the border gap before 1994 with the border gap after — isolating what the curriculum reform changed
The gap at the border did not close; if anything, schools outside the homeland improved more (−1.54 pp)
The educational scar is structural and institutional, not driven by the apartheid curriculum alone

The left panel shows the temporal RD: students who entered school after the democratic transition score substantially higher. The right panel asks a deeper question. We compare the spatial discontinuity at the KwaZulu border for cohorts educated under Bantu Education (born ≤ 1987) versus cohorts educated after the reform (born ≥ 1988). If Bantu Education drove the border gap, abolishing it should have shrunk the spatial discontinuity. It did not. The −1.54 coefficient suggests that schools just outside the homeland were better positioned to capitalise on the new curriculum. Homeland schools, suffering from decades of structural neglect, could not adapt as quickly. The implication: the persistence of the border effect is not about a racist curriculum — it is about the deep institutional infrastructure that the curriculum sat on top of, and which democracy never dismantled.

Resourcing at the Border

Class Size

Teacher Cost-to-Country (CPI-adjusted)

While disparities are visible in school inputs like class size and teacher pay, the mechanism of transmission isn’t sensitive to the institutional and funding gaps we record. The scar runs deeper than contemporary resourcing.

The Border, Visible from Space

1993 — Night light luminosity at the KwaZulu border, one year before democracy

2023 — The same border, thirty years later

The Night Light Gap, Across All Homelands

Former Homeland	Absolute	Proportional (log)	Interpretation
Ciskei	−5.2^***	+0.05	Fell further behind
Bophuthatswana	−5.1^**	+0.40^***	Grew faster; still behind
Venda	−3.5^***	+0.13	Fell further behind
Lebowa	−1.2	−0.17^**	Remained proportionally behind
Transkei	−0.1	−0.25	No significant change
KwaNdebele	+1.3	−0.33^*	Remained proportionally behind
Gazankulu	+1.6^*	+0.10	Some catch-up in levels
KwaZulu	+2.2^***	+0.31^***	Caught up
Kangwane	+3.7^*	+0.24^*	Caught up

Robust RD estimates (rdrobust). Absolute = raw luminosity difference. Proportional = log(2021+1) − log(1993+1), ≈ rate of growth. QwaQwa excluded. ^*p < 0.05, ^**p < 0.01, ^***p < 0.001.

This table is deliberately designed to provoke a double-take. The two columns answer different questions. The absolute column asks: “Did the raw luminosity gap widen?” The proportional (log) column asks: “Did homeland areas grow faster or slower from their starting point?” The answers frequently disagree — and the disagreement is the point. Take Bophuthatswana: in absolute terms, the gap widened dramatically (−5.1 luminosity units). But in proportional terms, the homeland areas actually grew significantly faster (+0.40 log points) — they tripled from a very low base while adjacent areas added more in levels. This is classic conditional convergence: the homeland is “catching up” in growth rates, but from so far behind that the absolute gap is still widening. The only homelands where both measures agree on convergence are KwaZulu and Kangwane — these show genuine catch-up in both levels and rates. For the audience: “Whether you see convergence or divergence depends entirely on how you measure it. And that is itself a finding about how deep the original scar was.”

The Gap Persists

The human capital scar is not converging towards zero.

Over the twenty-six-year period since the end of apartheid, the gap at the former homeland border has endured.

This is not temporary.

This is stubborn persistence.

Thirty years of democracy have not undone the spatial engineering of apartheid. The design is working as intended — even without the designer.

Two Channels of Persistence

1. Intergenerational Transmission

Parents subjected to Bantu Education pass the human capital deficit to their children.

Exploiting discontinuous exposure by age cohort
Parents educated under the harshest deprivation → children with lowest outcomes
The funding gap compounds across generations

2. Institutional Continuity

The governance architecture of the bantustans was never dismantled.

The chieftaincy system persists (and was strengthened in 2019)
Communal land tenure prevents capital formation
The Ingonyama Trust controls ~30% of KZN
Mamdani’s “decentralised despotism” — constitutionalised, not abolished

We exploit variation in both channels to separate the mechanisms. Persistence is not a single process — it is the interaction of inherited human capital deficits and unreformed institutional structures.

The Persistence of Spatial Disadvantage

The spatial dimension of apartheid harm in the former homelands has not been adequately addressed by post-apartheid policy.

Redress has not been broad-based; it has primarily focused on demographic representation, missing the specific geographic scars of the bantustans
Urban bias, a preponderant trend across the continent, persists — the rural poor remain politically and economically peripheral
Key institutions — communal tenure, the chieftaincy, the Ingonyama Trust — were never dismantled
The 29.5% of South Africans in the former homelands are among the most disadvantaged people in the country, living in places designed to produce disadvantage

These findings demand a policy response that considers the spatial legacy of apartheid. The borders of the bantustans remain the most powerful predictor of educational disadvantage in South Africa.

“The most profound challenges to South Africa’s development can be found in its rural hinterlands. These areas, systematically deprived of the most basic resources under apartheid, continue to lag behind in the post-apartheid era.”

— Nelson Mandela, Emerging Voices (2005)

This paper shows that Mandela’s observation is not merely descriptive — it is causal, it is measurable, and it is stubbornly persistent.

petercourtney.co.za

Peter Courtney · RESEP · Stellenbosch University & Tinbergen Institute