Introduction
Public discussion about creditworthiness in South Africa often assumes a single assessment: the borrower is either suitable for credit or they are not. In the South African regulatory terms, the creditworthiness rests on two distinct assessments with different purposes, beneficiaries and legal consequences: credit risk assessment and affordability assessment. Although these assessments may share certain elements and may increasingly be performed at the same time through automated decisioning, they are not interchangeable. The distinction matters for law, supervision, remedies and policy as lending becomes more digital and data driven.
Two assessments, two purposes
Credit risk assessment and affordability assessment answer different questions and serve different regulatory objectives.
Credit risk assessment assesses whether the lender is likely to be repaid. It is prudential in nature and protects lenders and the financial system from losses. In the banking sector, credit risk is integrated into capital and solvency requirements through frameworks influenced by Basel, which translate Probability of Default, Loss Given Default and Exposure at Default into risk-weighted assets for capital adequacy.
Affordability assessment assesses whether the consumer can repay credit instalments without becoming over-indebted. It is a conduct and consumer protection requirement under the National Credit Act (sections 81 and 82 read together with Regulation 23A). Affordability assessment protects consumer welfare and dignity and provides the basis for remedies such as debt review and reckless lending findings.
The two assessments differ not only in purpose, but also in beneficiary. Credit risk assessment protects lenders and the financial system; affordability assessment protects consumers. A borrower may therefore be a good credit risk and still be legally unsuitable for credit on affordability grounds. This duality is a foundational feature of South African credit regulation.
Prudential and Conduct regulation in South Africa
South Africa’s Twin Peaks model reinforces the institutional distinction between the two assessments. The Prudential Authority supervises solvency, capital adequacy and risk management in banks, insurers and designated financial conglomerates. Non-bank credit providers, including retailers, micro-lenders and digital lenders, are not prudentially regulated on a standalone basis, even though they form part of the financial services sector. (Certain non-bank credit providers may be prudentially captured at the group level where they form part of a banking or insurance conglomerate). The National Credit Regulator supervises the credit market under the National Credit Act, while the Financial Sector Conduct Authority supervises market conduct, consumer protection and fairness. This configuration makes the legal separation between credit risk and affordability visible in supervisory mandates.
Within the Twin Peaks model, combined automated credit risk and affordability assessments create supervisory challenges. A singular credit decision may implicate prudential considerations (credit risk) and conduct considerations (affordability), but the outcome does not indicate which assessment failed. Without clarity, it becomes difficult for supervisors to understand market behaviour or emerging risks.
Data, modelling and behavioural credit risk
Credit risk assessment and affordability assessment differ in their relationship to data variables and prediction. Affordability assessment is grounded in verified financial facts: income, living expenses, existing debt obligations and disposable income. These inputs speak to current capacity.
Credit risk relies increasingly on behavioural and predictive variables. Traditional credit bureau data remains central, but digital credit models incorporate additional variables such as employment category, transactional data and consumption patterns. Machine-learning methods allow these variables to be aggregated to estimate default probability and support risk-based pricing. These models are commercially efficient and improve financial inclusion for consumers with limited credit histories, but they also raise explainability and fairness concerns because many inputs function as socio-economic indicators. For example, a credit risk model may use data such as geolocation or employment type, which can indirectly lead to bias. Affordability does not raise equivalent concerns because it is based on direct financial verification. When credit risk and affordability assessments are combined into a single automated decision, it becomes difficult to determine whether a decline was based on affordability or influenced by indirect biases in credit risk modelling.
Some behavioural or transactional variables may overlap with affordability inputs. For example, transactional data may help verify income or expenditure patterns. However, overlap does not eliminate conceptual distinction. Predictive variables improve estimation of default probability; affordability requires evidence of financial capacity. One assessment is probability-based; the other is factual.
The distinction is material: affordability assessment is statutory and evidence-based and credit risk is predictive and model-based. The two assessments may use overlapping information, but they process it for different purposes and under different regulatory frameworks. For example, the National Credit Act requires that debt repayment history be assessed through a consumer credit information file obtained from a credit bureau. Credit bureau data can therefore support both assessments, but in different ways. For credit risk, credit scores, behavioural variables and repayment patterns inform probability of default modelling. For affordability, credit bureau data identifies existing obligations and past repayment behaviour for the purpose of verifying a consumer’s financial means, prospects and obligations.
Within a credit bureau report, certain elements are relevant to both assessments, such as existing obligations and repayment behaviour. Other elements are relevant only to credit risk. Credit scores, for example, estimate the probability of default and are used for pricing and portfolio management; they are not relevant for affordability, even though they are typically contained in the same credit bureau report. The presence of shared inputs does not collapse the two assessments.
Digital lending and combined decisioning
Digital credit markets increasingly execute credit risk and affordability assessments in a single automated workflow. Operationally, this is efficient. Automated underwriting reduces manual steps, increases speed and allows lenders to serve thin-file and digitally active consumers at scale. However, when both assessments are performed together and the system produces a single outcome, the legal distinction between the two assessments becomes difficult to observe from the outside. A singular decision does not indicate whether a credit decline arose from credit risk, from affordability, or from both. The distinction remains relevant in law, even when concealed in practice.
The issue is not the simultaneousness of assessment, which is commercially rational, but the transparency of the basis of decision. Digital credit systems have made the act of credit decisioning more efficient, but less transparent.
Machine learning underwriting introduces additional complexity. Conduct regulators increasingly consider questions of algorithmic fairness and explainability. Prudential supervisors continue to focus on risk, capital and model validation. The two sets of concerns converge operationally in digital credit markets but remain distinct in supervisory mandates.
With digital lending, explainability becomes increasingly relevant for understanding the basis of credit decisions. Conduct regulators require explainability to enforce statutory affordability requirements, for bias detection and for activities around consumer education. Prudential supervisors require explainability to validate models and assess capital implications. Consumers require explainability to understand the basis for the decisions and to meaningfully exercise their rights under the National Credit Act.
Consumers generally receive a single decline outcome. Without further detail, consumer understanding defaults to the credit score frame. Improving one’s credit score has become synonymous with improving access to credit, even though affordability is not determined by credit score and cannot be improved through credit score optimisation alone. This reflects the opaqueness of automated decisioning.
The basis for a decline matters. Affordability failures lead to remedies such as debt review and may support allegations of reckless lending. Credit risk failures raise different questions, including pricing and fairness. When the basis for the decline for credit is unclear, outcomes are difficult to interpret legally and practically.
For regulators and policymakers, transparency matters for market diagnostics. Conduct regulators require visibility to assess compliance with affordability requirements and consumer protection. Prudential regulators require visibility to understand risk dynamics. When underwriting decisions collapse the two assessments into a single outcome, supervisory data loses informational content. A market may appear exclusionary for credit risk reasons when the actual constraint is affordability or structural income patterns.
Explainability and transparency within underwriting systems therefore becomes increasingly important. Without explainability and transparency, prudential, conduct and consumer protection objectives become harder to align in practice.
Litigation and remedies under the National Credit Act
Most remedial construction of the National Credit Act is built around affordability assessment. Reckless lending and debt review rely on evidence that the consumer could not afford repayments at the time of contracting. Disputes about reckless lending therefore turn on verified financial information rather than default probability or credit bureau scores.
Attempts to treat sophisticated credit risk models as substitutes for affordability assessment have been resisted by the National Credit Regulator. The premise that predictive modelling “surpasses” affordability because it incorporates more data variables is conceptually flawed. Predicting default does not establish that a consumer can repay instalments without becoming over-indebted. Where parties conflate the two assessments, they blur statutory consumer protection requirements (affordability) with a commercial consideration (credit risk). Evidence relevant to lender risk may not satisfy the National Credit Act’s duty to verify a consumer’s financial means, prospects and obligations.
A consumer may fail credit risk and pass affordability. In such cases, lenders typically respond through risk-based pricing, smaller limits or outright declines, subject to National Credit Act rate caps and internal risk appetite. Credit risk can be priced; affordability cannot because hardship cannot be priced away. Conversely, a consumer may pass credit risk and fail affordability. In such cases, the National Credit Act prohibits the transaction irrespective of pricing, because affordability concerns consumer hardship rather than lender exposure. There are also cases of dual failure, where a consumer is both over-indebted and a poor credit risk. In such cases both assessments point to a decline.
Supplementation within the credit risk and affordability assessments is not inherently impossible. The practical question is therefore about supplementation, not substitution. Predictive credit risk indicators may add useful information to affordability assessment, provided they do not displace the requirement to verify financial means, prospects and obligations. Supplementation would need to occur in settings that are auditable and supervised, so that predictive information supports rather than replaces direct financial verification.
As credit underwriting becomes more digital and data-driven, the distinction between credit risk and affordability becomes more important to maintain. Future reforms are likely to focus on how both assessments can operate in parallel, and use some shared data, without weakening either prudential or consumer protection objectives.
Financial inclusion and market access
Financial inclusion is not only about expanding access to credit (or other financial products or services); it is also about understanding why access is denied. Financial exclusion may stem from credit risk or affordability. Each cause points to a different policy response. If the reason for non-access is unclear, policy interventions can easily miss the mark.
Alternative data for credit risk and affordability
The distinction between credit risk and affordability also matters for how alternative data is used in digital underwriting. Not all alternative data serve the same purpose. Some variables speak clearly to credit risk: for example, behavioural patterns, transactional histories, employment stability or sectoral volatility. Other alternative data points speak more directly to affordability, such as verified income flows, expenditure patterns or recurring obligations. As more alternative data is incorporated into credit decisioning, it is important not to treat all alternative data as if it serves a single assessment.
Recognising these functional differences allows for a more structured classification of alternative data. Such classification, effectively a taxonomy for alternative data, would clarify which information may legitimately support credit risk and which may support affordability. A practical taxonomy could also make it easier to align alternative data with the appropriate regulatory framework: prudential for credit risk; conduct and consumer protection for affordability.
In this way, the distinction between the two assessments does not hold innovation back; it structures innovation. It enables the responsible use of alternative data in digital credit markets by making the use, supervision and oversight clearer.
Conclusion
Credit risk and affordability are distinct legal assessments in South African credit regulation. Their purposes differ, their beneficiaries differ and the consequences of failure differ. Automation and digital credit markets have not eliminated the distinction but have made it less visible. When decisioning collapses into a single outcome, it becomes difficult to determine whether a credit decline arises from lender risk, consumer hardship or structural income volatility.
Transparency in the distinction between the two assessments matters for financial inclusion. If policymakers cannot distinguish between credit risk exclusion and affordability exclusion, policy interventions risk being misdirected. Transparency in the distinction also matters for consumers. Without understanding the basis for a credit decline, consumers cannot meaningfully exercise rights under the National Credit Act. Without clarity, consumers cannot begin to improve their standing because they do not know which assessment is constraining their access to credit.
The regulatory challenge for South Africa is therefore not to merge the two assessments, but to preserve their separation in a digital environment. As digital lending expands and predictive models mature, maintaining clarity between prudential and conduct objectives becomes more- not less- important for consumer protection, market integrity and financial inclusion.
