Most of us, at some point in our lives, find ourselves looking back and wondering what might have happened if we had chosen differently – if we had taken the other job, bought the other house, stayed, or left?
Usually, the answers matter only to us.
For the Public Investment Corporation (PIC), the stakes are considerably higher. It manages money for clients including the Government Employees Pension Fund (GEPF), which has more than 1.2 million active members and about 565 000 pensioners and other beneficiaries.
So, when researchers ask, “What if the money had been invested differently?”, they are not indulging in the kind of personal hindsight most of us occasionally do. They are applying the question to billions of rand in pension savings.
That is what the Bureau for Economic Research (BER) has done with the Isibaya portfolio, which is the PIC’s unlisted, developmental investment division.
The BER, established in 1944, is one of South Africa’s oldest economic research institutes. It is part of Stellenbosch University’s Faculty of Economic and Management Sciences and produces economic research, forecasts, and business surveys.
For this analysis, its researchers worked through 11 390 Isibaya transactions between 1 March 2006 and 30 September 2025 and applied the same amounts on the same dates to the FTSE/JSE All Bond Index (ALBI).
Isibaya’s recorded closing value was R65.4 billion. Under the bond-market counterfactual, the researchers calculate a theoretical closing value of R135.1bn.
The difference is R69.7bn.
It sounds like a loss. To be fair, it isn’t.
The BER calls it an opportunity cost: what the money might have been worth under a different investment route. It is not an accounting loss, and the comparison does not suggest that the GEPF should simply have put its Isibaya allocation into bonds.
There are some important limits to the comparison. The ALBI calculation does not include fees, taxes, transaction costs, or the potential market impact of moving such large amounts of money into bonds. It also cannot capture the developmental objectives of Isibaya. The BER is therefore comparing the financial outcome with a liquid bond-market alternative, not asking whether bonds would have delivered the same developmental benefits.
But why compare Isibaya with bonds in the first place?
The answer lies in what Isibaya was created to do.
It was never supposed to be just another investment portfolio
The GEPF invests in conventional assets such as equities and fixed income, but it also put part of its assets into a developmental investment mandate.
That mandate began in the 1990s at 3.5% of assets for black economic empowerment transactions, later rising to 5%. By the time of the Mpati Commission, the mandate provided for 10% for private equity and impact investing.
The idea was simple: earn a financial return while also contributing to transformation, economic growth, and job creation.
That meant investing in businesses, infrastructure, housing, and financial services, much of it outside the listed market. But there is a trade-off. Unlike a listed share or bond, an unlisted investment has no continuously quoted market price and may be harder to sell when trouble starts.
And then came the scrutiny
In October 2018, President Cyril Ramaphosa appointed the Judicial Commission of Inquiry into Allegations of Impropriety at the PIC, chaired by Justice Lex Mpati. The Commission examined investment decisions, governance failures and possible political interference, including transactions involving Independent Media, Lancaster/Steinhoff, Ayo, Tosaco, Ascendis, and VBS.
One recurring issue was cumulative exposure. In the Maponya Matome Investment Holdings case, the Commission found several PIC investments involving the same individual and no formal limit on the number of investments with a single counterparty or total exposure. It calculated exposure at R1.85bn, rising to R2.2bn including accrued interest, and questioned the PIC’s assessment of cumulative risk. It recommended that the GEPF’s R2bn approval threshold take cumulative investments into account.
In the VBS matter, the Commission found no impropriety by the PIC board in the decision to invest, although it found evidence of ineffective governance involving two PIC executive directors.
The BER is not revisiting those findings. Instead, it asked what Isibaya produced financially over almost 20 years, and what those same cash flows might have produced in the listed bond market.
Where the money worked – and where it did not
The BER calculates that Isibaya generated an annualised return of 4.25% between March 2006 and September 2025.
Average inflation over the period was 5.53%, putting the inflation-adjusted return at approximately -0.93%. The same cash flows invested in the ALBI would have produced a 9.02% annualised return.
The 4.25% is an average, though. The investments behind it had very different outcomes.
Across the investment families constructed from the transaction data, the BER calculates R64.1bn in gross nominal gains and R29.35bn in gross nominal losses. The five largest loss-making families accounted for 62.5% of losses, against 43.5% of gains from the five biggest winners.
At the top of the loss table is AfriSam, a cement producer.
The PIC invested R26bn in the AfriSam investment family. R16.4bn was returned, and nothing remained at the end of the period, leaving a nominal loss of R9.6bn.
AfriSam’s contribution to the BER’s ALBI gap is R42.3bn, reflecting what the same cash flows might have earned in the bond market through to September 2025.
The investment dates to 2008, when the PIC provided R6bn in bridge finance as part of a debt refinancing. The PIC later reported that the highly leveraged structure came under severe pressure as growth slowed; weaker cement demand, oversupply, competition, and rising electricity and input costs squeezed margins.
The official sources reviewed by the BER contain no finding of impropriety in the AfriSam investment.
Next comes Bayport/BML, a consumer-lending business operating across African markets. It recorded a R3.82bn nominal loss and a R9.8bn ALBI gap. The exposure included equity and later loans linked to a BEE acquisition structure; Bayport subsequently came under pressure from its capital structure, cash-interest burden, and debt maturities.
Belelani/BVI recorded a R3.17bn nominal loss and a R7.16bn ALBI gap. The investment helped finance Belelani’s acquisition of a 24% stake in Pareto and BVI, linked to shopping-centre assets. The GEPF recorded a R2.33bn impairment in 2020 and a R1.63bn reversal in 2022. The BER says the evidence is more consistent with stress in the acquisition-financing structure than with a collapse in the shopping-centre assets, although the public record does not establish the precise cause or whether the original due diligence was adequate.
The other two largest loss families were Independent Media, the newspaper and publishing group, at R920 million, and Smile Telecoms, a telecommunications company, at R820m. The PIC later reported that assumptions behind Independent Media had not materialised and that its price was high relative to the risks. Smile’s underperformance was attributed to economic conditions, currency depreciation, particularly in Nigeria, and competition, alongside governance weaknesses.
AfriSam, Bayport/BML and Belelani/BVI alone account for R59.3bn, or 85.1%, of the R69.7bn ALBI gap.
Some investments made money, but they earned far less than the bond-market alternative.
Vodacom/uNewco/UCVH, a family of transactions linked to telecommunications and BEE structures, generated a nominal gain of R1.1bn but still contributed R16.6bn to the ALBI gap. Its Extended Internal Rate of Return (XIRR) was 0.92%.
The winners
The largest nominal winner was the MTN BEE investment. It generated a R9.14bn nominal gain, a 21.28% XIRR and R30.1bn in ALBI outperformance and enabled more than 3 200 MTN employees to become shareholders before it was unwound profitably.
Lancaster/Steinhoff BEE generated a R6.35bn nominal gain, a 10.09% XIRR and R1.99bn in ALBI outperformance, supporting Lancaster’s acquisition of a black ownership stake in Steinhoff and intended to increase participation by black suppliers and entrepreneurs.
SA Home Loans, the housing-finance provider, generated a R5.27bn nominal gain and an 8.35% XIRR. The funding supported lending to government employees and the affordable-housing market; the PIC reported more than 6 249 government-employee loans and 4 113 affordable-housing loans in 2018.
IDC01U recorded a R4.55bn nominal gain and a 9% XIRR. Africa Finance Corporation, a development finance institution, recorded a R2.58bn gain, although the BER cautions that the data records only R310 500 of deployment against a R2.4bn closing value.
SA Home Loans still contributed R1.07bn to the ALBI gap. Capitec, only the tenth-largest nominal winner, was among the largest ALBI outperformers because its cash flows came back relatively quickly.
A few investments moved the number a lot
The BER tested what happens when the largest loss-making families are removed.
The full portfolio produces an XIRR of 4.25%; remove AfriSam and it rises to 6.43%; remove AfriSam, Bayport/BML and Belelani/BVI and it rises to 7.72%; remove the five largest loss families and it rises to 8.13%.
The researchers stress that this is a hindsight-based diagnostic, not what the portfolio would necessarily have earned. The excluded capital is not reinvested.
The timing matters too.
Investments first recorded in 2008 generated R9.69bn in gross nominal losses, almost entirely because of AfriSam. The 2015 vintage generated R10.44bn, with Bayport/BML, Belelani/BVI, Smile Telecoms, Allied and Daybreak/Afgri Poultry among the main contributors. Together, the two vintages account for 68.6% of gross nominal family losses.
The more recent investments look different. Those first recorded from October 2015 generated a 7.57% XIRR, while investments first recorded from October 2020 outperformed the ALBI overall. These are not current portfolio returns; they group investments by when they first appear in the transaction records. The five-year result is also affected by incomplete transaction data for Africa Finance Corporation.
What the numbers cannot tell us
The BER’s conclusion is less about choosing between development and financial returns than about whether the two can be achieved together.
The researchers point to investment selection, financing structures and the management of concentrated and related exposures as areas requiring closer attention. They also point to stronger results from some more recent investments, while cautioning that these figures should not be read as current portfolio returns.
The harder question is the developmental side.
The PIC reports jobs, housing, enterprise support and infrastructure outcomes, but the BER could not match those outcomes to the individual Isibaya transactions in its dataset. That means it cannot put a financial value on them or weigh them against the investment results.
And perhaps that is where the “could have, would have, should have” comes back in.
The BER can tell us what happened to the money, and what might have happened under a different investment route.
It cannot put a price on the development those investments were intended to create.
