Portfolio at risk (PAR)

Portfolio at risk is the percentage of your outstanding loan book that sits with borrowers who have a payment overdue. It is the single most watched number in lending, because it answers the question every lender, funder, and regulator asks first: how much of the money you are owed is showing signs it may not come back?

The important part is what goes into the numerator. PAR does not count the missed instalment. It counts the entire outstanding balance of every loan that has any payment late. A borrower who owes 10,000 and has missed one instalment of 500 adds 10,000 to your portfolio at risk, not 500.

That is deliberate. Once a borrower misses one payment, the whole loan is in question, not just the part they skipped.

The PAR formula

PAR (N days) = Outstanding balance of loans with a payment over N days late
               ─────────────────────────────────────────────────────────────
                            Total outstanding loan portfolio

Both figures are measured on the same date. The denominator is the gross outstanding portfolio — all principal still owed to you, performing and non-performing together.

Worked example. Your book has 2,000,000 outstanding across 400 loans. Thirty-one loans have a payment more than 30 days late, and those thirty-one loans have a combined outstanding balance of 180,000.

PAR 30 = 180,000 ÷ 2,000,000 = 9%

Why the "N days" matters

PAR is always quoted with a threshold: PAR 1, PAR 14, PAR 30, PAR 90. The threshold sets how late a payment must be before the loan counts as at risk.

The convention almost everyone uses is cumulative and nested. PAR 30 means every loan with a payment 30 or more days late — which includes the loans that are 60, 90, and 200 days late. So:

• PAR 1 ⊇ PAR 30 ⊇ PAR 90
• PAR 1 will always be the largest number, PAR 90 the smallest
• The figures are not meant to be added together

This is the definition used by CGAP, by MIX Market before it closed, and by most central banks and development funders. If you send a PAR 30 figure to an investor, this is what they will assume you mean.

PAR is not the same as aging buckets

Aging buckets — 1–30 days, 31–60, 61–90, 90+ — are mutually exclusive. Each loan appears in exactly one bucket, and the buckets add up to your total delinquent portfolio.

PAR thresholds overlap. Each one contains the ones above it.

Both views are useful and most lenders need both. Aging buckets tell you where the delinquency is concentrated and how it is moving month to month. PAR tells you the headline exposure at a given depth. The mistake is mixing them in one table without labelling which is which, because a branch manager reading "PAR 30: 9%" next to "PAR 60: 6%" will either double-count a defaulting borrower or assume the numbers should sum. They should not.

If you report both, label them plainly. "Payments overdue by more than 30 days" for PAR. "Loans 31 to 60 days late" for buckets.

What good and bad look like

Benchmarks are always quoted against PAR 30. Rough industry bands:

PAR 30ReadingUnder 5%Healthy. Normal for a well-run book.5% – 10%Watch closely. Something in collections or credit assessment is slipping.Over 10%Serious. Expect funder and regulator questions.

Treat these as a starting point, not a rule. Your baseline depends heavily on what you lend:

• Payroll-deducted lending should run very low PAR — often under 2% — because repayment is taken at source. A payroll book at 6% has a real problem.
• Unsecured micro-lending to traders carries structurally higher delinquency, and a disciplined lender may run at 8% with strong recovery rates and still be profitable.
• Group lending with joint liability usually sits lower than individual lending in the same market, because the group absorbs short-term shortfalls before they reach you.
• Seasonal books — agriculture especially — will spike predictably before harvest. Compare against the same month last year, not last month.

Set your own thresholds against your own history and your cost of funds. A rising trend on a low number matters more than a flat high one.

Where PAR gets quietly wrong

PAR is easy to calculate and easy to distort. Most of the distortion is not fraud — it is bookkeeping that drifted.

Written-off loans left on the books. If you write a loan off but never remove it from the outstanding portfolio, it inflates both numerator and denominator and makes your PAR look artificially stable. If you remove it correctly, PAR rises, because your denominator shrank while the healthy loans stayed. That is the honest number. Lenders who avoid writing off because it hurts the ratio are lying to themselves first.

Restructured loans reset to zero. Rescheduling a delinquent loan and calling it current is the oldest way to make PAR look better. A restructured loan that was 90 days late is not a performing loan on day one of its new schedule. Track restructured loans as a separate line and report them alongside PAR, not inside it.

Arrears dated from the wrong day. If a payment was due on the 5th, received on the 20th, but posted on the 25th, the loan's days-late figure depends entirely on which of those dates your system uses. Use the due date and the actual payment date. Posting date is an admin event, not a borrower event.

Partial payments misapplied. If a borrower pays part of an instalment and your system applies it in the wrong order, the loan can appear current when principal is still short — or appear late when it is not. This is why the allocation waterfall matters: penalties, then fees, then interest, then principal, applied the same way every time.

Prepayments hiding arrears. A borrower who pays two instalments early and then misses three may still show a positive balance against schedule in a naive calculation. PAR should test against the instalment due, not the cumulative amount paid.

Different denominators between reports. Gross portfolio, net of provisions, principal only, principal plus interest receivable — each gives a different PAR. Pick one, write it down, and use it everywhere. Most lenders should use gross outstanding principal.

PAR is a symptom, not a diagnosis

A PAR figure tells you something is wrong. It does not tell you what. Read it alongside:

• Collection rate — what percentage of what fell due this month was actually collected. Falling collection rate is the leading indicator; PAR is the lagging one.
• Aging buckets — is the delinquency fresh (1–30 days, recoverable with a phone call) or old (90+, likely gone)? Two lenders with identical PAR 30 can be in completely different positions.
• Roll rates — what share of the 1–30 bucket moves into 31–60 next month. This is where you see collections working or failing.
• Loan loss provisioning — PAR tells you what is at risk; provisioning tells you what you have already set aside against it.
• PAR by officer, branch, and product — a 9% book-level figure often hides one branch at 3% and another at 22%. The average is the least useful cut of the data.

Why most small lenders can't see their PAR

The formula is arithmetic. The problem is the data.

To calculate PAR correctly you need, for every loan, on any given date: the outstanding balance, the full schedule of instalments due, every payment received with its actual date, and a consistent rule for how those payments were applied. On a spreadsheet with a few hundred loans across three branches, that is a week of work — and by the time it's finished, the number is a week old.

So it doesn't get done. Instead the lender knows roughly who is behind because a loan officer keeps a list, and the real position surfaces only when a funder asks for it or a branch is audited. By then the fresh delinquency that could have been recovered with a phone call has aged into the 90-day bucket.

PAR is not valuable as a month-end report. It is valuable on a Tuesday morning, when the loans that went 5 days late over the weekend can still be saved.