The Balance-Sheet Logic of a Long-Maturity Loan
Colombia’s provisioning reform changes the cost a financial institution must recognize when it originates a covered consumer loan with a long repayment horizon. The key question is how that cost is built: which risks enter the expected-loss calculation, why the 72- and 108-month thresholds matter, and how a lender could respond without automatically charging the borrower more. Tracing that chain makes the later evidence easier to read, because it shows which outcomes should move first and which are choices made after the rule has done its accounting work.
Expected loss is a forward-looking estimate
A provision is best understood as a recognition of possible loss, not as a payment made after loss has already occurred. Colombia’s framework estimates expected loss from three basic ingredients: the probability of default (PD), the exposure at default (EAD), and the loss given default (LGD). In compact form, the paper expresses the calculation as:
Expected loss = PD × EAD × LGD × MA × K
The first three terms answer different questions. PD asks how likely the borrower is to default within the relevant 12-month horizon. EAD asks how much the institution would have exposed when default occurs, including the outstanding principal and other amounts identified by the model. LGD asks what share of that exposure would not be recovered after considering the borrower’s repayment and the value or quality of available collateral.
These distinctions matter because a loan can become more costly to carry through more than one channel. A riskier borrower can raise PD. A larger balance can raise EAD even if the borrower’s risk category is unchanged. Weak or absent collateral can raise LGD because less of the exposure is expected to be recovered. The provisioning system combines those dimensions rather than treating “credit risk” as a single score.
The terms also describe different points in the loss process. PD is about whether a default event may occur; EAD is about how much is at stake when it occurs; LGD is about how severe the loss would be after recovery. A lender can therefore face a high expected loss because a borrower is more likely to default, because the balance is large, because recovery is poor, or because several of these conditions coincide. The formula is useful precisely because it keeps those channels visible.
Imagine two new loans with similar borrower ratings and the same principal. If one is secured by stronger collateral, its LGD may be lower. If the other runs much longer, the maturity terms can raise its expected loss even when the starting borrower information looks similar. The comparison is not a forecast that the longer loan will default; it is a recognition that the institution must carry that exposure through more uncertain future conditions. That recognition is what a forward-looking provision is designed to capture.
Maturity adds a different kind of uncertainty. A loan that runs for many years gives more time for household income, employment, interest rates, collateral values, and the wider economy to change. The SFC’s framework already included a maturity adjustment, or MA, that raises expected loss for repayment horizons beyond 72 months. The 2023 reform added K to capture the additional risk associated with higher leverage at longer maturities. For the covered tiers, K is 1 up to and including 72 months, 1.1 above 72 and up to 108 months, and 1.4 above 108 months.
The multiplication is conceptually important. The K factor does not replace the borrower’s PD, the size of the exposure, or the expected recovery. It scales the loss estimate after those other risks have entered the calculation. Two loans could therefore have similar borrower risk and collateral characteristics but different provisioning costs because one keeps the institution exposed for longer. Conversely, a long loan with a small balance or strong recovery prospects may still have a different absolute provision from a large, weakly secured loan. The rule adds a maturity-linked layer; it does not make maturity the only determinant of risk.
MA and K also play different roles in the policy story. The maturity adjustment already made repayment horizon part of expected loss beyond 72 months. K is the reform’s additional intervention for the leverage and maturity risk of new covered loans. It therefore creates a change at specific boundaries rather than simply continuing a smooth relationship between term and provisions. A loan just below a threshold and a loan just above it may be similar in many respects, but the regulatory calculation treats the latter as belonging to a more expensive risk tier.
That boundary is economically useful but not magical. A borrower does not become suddenly riskier on the day a contract crosses from 72 to 73 months. The threshold is a regulatory decision that approximates a change in exposure and creates a clear rule for institutions. The study can use that decision to compare loans near the treatment categories, but it still has to account for the fact that borrowers and lenders may choose different maturities for reasons unrelated to the rule.
This is why higher provisioning can strengthen resilience without producing an immediate change in the loan contract. The provision records an expense and increases the balance-sheet resources set aside against potential loss. It does not transfer cash from the borrower to the regulator, and it does not mechanically alter the contractual interest rate. The institution initially bears the accounting cost. Whether that cost later reaches the borrower depends on the lender’s margins, capital position, funding conditions, competitive position, and willingness to keep originating the affected loans.
The distinction between expense and cash flow should not be mistaken for harmlessness. A provision can reduce reported profitability and change how attractive a loan is relative to other uses of the institution’s balance sheet. If the cost is small relative to the expected return, a lender may absorb it. If the cost is material for a narrow, risky segment, the lender may protect profitability or capital by changing the amount it disburses, requiring more collateral, charging a different price, or preferring a shorter term. The rule changes the incentive at the institution; the institution chooses the transmission path.
From a maturity threshold to a lender decision
External Circular 026 made that transmission path discrete. It did not impose one continuous surcharge on every consumer loan. It introduced two maturity bands for new covered loans: above 72 months and up to 108 months, and above 108 months. This design recognizes that a nine-year loan is not merely a six-year loan with a few extra payments. It leaves the lender exposed to a longer sequence of possible economic conditions and, under the rule, carries a larger additional factor.
The rule’s scope is as important as its thresholds. Credit cards and revolving credit were excluded, as were payroll loans to pensioners. The empirical study also removes records that cannot be reliably identified within the covered consumer-loan categories. That means the policy’s marginal cost applies to a defined set of new originations rather than to every household debt outstanding in Colombia. A result found in this segment should not be generalized to revolving balances, existing loans, or products that never faced the additional factor.
For a covered new loan, the sequence is straightforward. First, the institution identifies the loan’s risk inputs and maturity. The expected-loss model combines PD, EAD, LGD, the existing maturity adjustment, and the new K factor where applicable. The resulting provision raises the cost of carrying that exposure and, if the provision coverage ratio rises, leaves the institution better prepared for potential deterioration. Only after that accounting effect does the institution decide how to manage the commercial consequences.
There are several possible responses, and they do not carry the same meaning. Raising the interest rate attempts to pass the cost through price. Reducing the loan amount lowers the institution’s exposure. Lowering the loan-to-value ratio requires more borrower equity or collateral relative to the principal. Shortening the maturity reduces the time over which the institution is exposed, although it may increase the scheduled payment. Tightening approval or redirecting borrowers toward another product changes who receives the loan without necessarily changing the terms of the loans that remain in the data. Absorbing the cost leaves the observed contract broadly unchanged but may reduce the lender’s margin.
These choices redistribute risk rather than simply making it disappear. A higher rate may preserve the lender’s return but increase the payment burden for a borrower whose income is already uncertain. A smaller principal may protect the lender while leaving the borrower unable to finance the intended purchase. A lower LTV shifts more loss-absorbing capacity to the borrower through collateral or equity. A shorter maturity can reduce the lender’s time exposure but raise the borrower’s periodic payment. The same provisioning rule can therefore look prudent from the institution’s balance sheet and restrictive from the perspective of a household that needs a long repayment schedule.
The choices also have different visibility in administrative data. A change in loan amount, interest rate, or LTV appears in the terms of an originated loan. A decision not to approve an application, to offer a different product, or to move a borrower to another lender may not appear as a change in any contract that is observed. That is why the study’s measured margins are important but incomplete: they show how the terms of new loans moved, not every way in which credit access could have changed.
The diagram separates the rule’s trigger from the lender’s decision. The maturity threshold triggers a change in the expected-loss calculation; it does not determine which commercial response must follow. This is also why provision coverage is a distinct outcome from interest rates or loan size. Coverage can rise even when the lender absorbs the expense. Conversely, a lender can reduce loan size or LTV as a defensive response, but that does not necessarily tell us whether the institution’s overall buffer improved by the same amount.
The mechanism also explains why the longest loans are the natural place to look for an effect. The additional K factor is larger above 108 months, so the marginal cost is strongest where the repayment horizon is longest. The 72-month tier still receives a higher factor than shorter loans, but the incentive to adjust should be more pronounced where the rule adds 40 percent rather than 10 percent to the existing expected-loss calculation, all else equal. “All else equal” is essential here: the longest loans may serve different borrowers, use different collateral, or be originated by different institutions. Those pre-existing differences must be handled before the policy effect can be read from the data.
Institutional capacity matters for the same reason. Two lenders facing the same regulatory factor may have different margins, capital buffers, funding costs, risk models, or market positions. One may treat the additional provision as a manageable cost of retaining a profitable customer. Another may see it as a reason to reduce exposure to long-maturity loans. The formula creates the common shock; it does not imply a common response.
This is also why the policy can have an uneven incidence without an obvious system-wide break. A large institution may spread a modest additional provision across a broad portfolio or accept lower margin on a strategic product. A smaller institution with fewer ways to diversify may respond at the specific maturity tier where the cost is highest. The result can be a stable average rate and loan amount alongside a real tightening for customers whose lender has less capacity to absorb the shock.
The mechanism therefore predicts a direction of pressure, not a single unavoidable outcome. Higher K raises the cost of long-maturity exposure. Whether that pressure appears as higher price, smaller principal, lower LTV, shorter term, fewer approvals, or no visible contract change depends on the institution’s choices and on competition. The empirical work must observe those margins separately rather than infer one from another.
This gives the study a precise empirical task. It must examine whether provision coverage rises, then test whether rates, loan amounts, and LTV change at the maturity thresholds. It must also ask whether those effects differ across institutions. A comparison of raw averages would not be enough, because long loans could have different terms even without the reform. The next question is therefore not just what the rule predicts, but how the study separates that prediction from the ordinary selection of borrowers, products, and lenders into different maturity ranges.