What Portfolio at Risk Measures and 3 Ways Microfinance Lenders Misread It
Bob Schulte
May 20, 2026
8 mins read
Portfolio at Risk (PAR) measures the outstanding balance of all loans with at least one overdue installment, divided by the total gross loan portfolio balance. It is a balance-weighted exposure ratio, not a count of late loans.
That distinction matters because three common calculation errors make PAR look better than the portfolio actually is. Lenders count late loans rather than summing their balances, and write-offs are excluded from the denominator, compressing the ratio without any real recovery. Rescheduled loans remain in the numerator because they are now current under the revised terms.
Each error is silent. None of them triggers an alert in a spreadsheet or a basic LMS (loan management software).
In the microfinance portfolios I have reviewed, PAR figures in board decks and donor reports are most often wrong in execution. The formula is right, but the inputs are not. That is what this article addresses.
PAR is a balance-weighted ratio. The numerator is the total outstanding balance of every loan with at least one installment overdue beyond a defined threshold.
That threshold is the DPD (Days Past Due) marker. Every loan that crosses it is included in the numerator at its full outstanding balance, not just the overdue installment amount.
Three thresholds matter operationally:
PAR more than 30 is not the same as a 30-day delinquency rate. A delinquency rate typically measures only the amount of overdue installments. PAR more than 30 measures the full outstanding balance of every loan with one installment past 30 days. On a portfolio with large loans in arrears, the difference is significant.
A single large delinquent loan carries more weight in PAR than five small current ones. That is why using the loan count as a proxy produces a misleading number. It hides the size concentration inside the ratio.
The most common PAR error is using the number of loans with overdue payments as the numerator instead of their outstanding balance.
It happens because it is easier. Counting flagged accounts in a spreadsheet takes seconds. Summing their outstanding balances takes longer, especially without an automated system that aggregates balances by DPD bucket.
The problem is that the two numbers tell different stories on the same portfolio.
Here is a simple example:
Your portfolio: 100 loans, total outstanding balance $500,000. 10 loans are 30+ days overdue.
Count-based PAR > 30: 10 ÷ 100 = 10%
Now, say 5 of those 10 loans are large – together they carry $80,000 in outstanding balance.
Balance-based PAR > 30: $80,000 ÷ $500,000 = 16%
Same portfolio. Same loans. Six percentage points apart.
That gap is the difference between a board that thinks portfolio stress is moderate and one that understands it is elevated.
As of Q3 2024, Aeris Insight tracked 61 US CDFI Small Business and Micro Lenders. Eleven had more than 30-day delinquencies above 10% – up from five to six CDFIs at FYE 2022 and FYE 2023. That rising concentration matters most when delinquency clusters in larger loans. Count-based reporting hides that clustering entirely.
One caveat: count-based figures are not wrong in every context. Some funders request them. But they are not PAR by the CGAP definition. If your institution runs both, label them separately. Calling a count-based figure PAR in a donor report is a categorization error, not just a calculation preference.
Run the balance-based figure alongside the count-based one every reporting cycle. If they diverge by more than 3–4 percentage points, delinquency is concentrating in your larger loans.
When loans are written off, they are removed from the gross loan portfolio. The denominator gets smaller. If the numerator shrinks at the same rate, PAR holds steady. But when write-off activity is heavy and concentrated in at-risk loans, the denominator can compress faster than the numerator, and PAR falls without any real improvement in collections.
The San Francisco Fed’s CDFI financial data primer documents this directly: portfolio quality ratios at US CDFIs can be misleading in periods of high write-offs or loan sales because changing portfolio ratios do not reflect a decline in the total number of dollars at risk.
The mechanism is the same whether the trigger is a write-off cycle or a loan sale. The denominator shrinks. The ratio improves. The underlying exposure does not change.
Compare your PAR more than 30 trend month-over-month against your write-off volume over the same period. If PAR falls as write-offs rise, the denominator is driving the improvement, not collections.
OFN’s FY2020 membership data puts the sector-wide net charge-off rate at 0.48% – well below the threshold at which write-off activity meaningfully compresses PAR denominators.
At that rate, write-off-driven PAR distortion is minimal. But individual institutions with higher charge-off rates – as Aeris reports, in the bottom quartile of the BML peer group face meaningful denominator compression risk in any reporting period in which large impaired loans are written off.
The fix is to report PAR alongside write-off volume in the same period. A board report that shows only PAR without write-off activity is structurally incomplete.
Build a two-row summary into every PAR report: PAR more than 30 this period, and gross write-offs this period. If the second row is rising while the first is falling, the ratio is doing accounting work, not signalling portfolio recovery.
Rescheduled loans where the original repayment terms were modified because the borrower could not meet them must stay in the PAR numerator. A loan that is now current under a revised schedule is not a recovered loan. It is a loan that demonstrated repayment difficulty and received new terms.
Removing it from the numerator because it is technically on time understates portfolio stress. It also concentrates the error in exactly the periods when stress is highest: economic downturns, when rescheduling activity spikes.
US CDFI lenders saw this firsthand during 2020 and 2021. The CDFI Fund’s Equitable Recovery Program deployed over $1.73 billion in grants to 604 CDFIs responding to COVID-19-related portfolio stress. [Source]
Across the sector, widespread payment deferrals and loan modifications held reported PAR flat during a period of significant underlying borrower stress. Institutions that excluded deferred and modified loans from their PAR numerators reported stable portfolio quality, while their actual at-risk exposure was rising.
The CGAP/MicroRate consensus standards, which apply globally to microfinance institutions and are referenced by US-based MFIs reporting to mission-aligned investors, are explicit: rescheduled and restructured loans belong in the PAR numerator regardless of their current payment status under revised terms.
This is also the point where PAR and a simple delinquency report diverge most sharply. A delinquency report shows the current payment status. PAR measures credit risk. A loan modified to avoid default has reduced delinquency risk on paper, but carries the same credit risk it did before modification.
Maintain a separate rescheduled loan register. Add the outstanding balances of all rescheduled and deferred loans to your PAR numerator every reporting cycle, regardless of current payment status. If your LMS does not surface modification history alongside payment status, you are working with an incomplete numerator.
The three metrics appear in the same conversations. They measure different things. Using one in place of the other produces a number that looks credible but answers the wrong question.
Here is how they differ:
| Metric | What it measures | Numerator | Denominator | Standard threshold |
|---|---|---|---|---|
| PAR | Outstanding balance at risk of default | Balance of all loans with 1+ overdue installments | Total gross outstanding portfolio | More than 30 days (MFI standard) |
| NPL Ratio | Non-performing loan exposure | Balance of non-performing loans | Total gross loans | More than 90 days (commercial banking standard) |
| Repayment Rate | Cash collection performance | Payments received in the period | Payments due in the period | No universal standard |
1. The most damaging substitution is the repayment rate for PAR. A US microlender can post a 97% repayment rate and a rising PAR of more than 30 simultaneously. That happens when small loans are repaid in full, but a few large loans fall behind. The cash metric looks healthy. The balance-based risk metric does not.
2. NPL ratio carries a different problem. It uses a 90-day threshold from the commercial banking industry. A microfinance portfolio operating without collateral cannot wait 90 days to flag stress. By the time a loan crosses PAR more than 90, the window for intervention has closed.
3. Use PAR more than 30 for weekly portfolio monitoring. Use PAR more than 90 for provisioning decisions. Use the repayment rate to measure your collections team’s performance. Never report one in place of another.
An accurate PAR calculation starts with one data output: outstanding loan balances segmented by days-overdue bucket. Most manual information system setups cannot produce this on demand. They track payment status but do not aggregate outstanding balances by DPD bucket without a custom report.
Bryt’s Aging Reports segment outstanding loan balances by delinquency category. They provide MFI portfolio managers with numerator inputs for PAR > 30, PAR > 60, and PAR > 90 reads, without a custom build or spreadsheet assembly.
See how Bryt’s reporting supports portfolio quality monitoring for microfinance operations.
