In August 2026, Nepal’s 20 commercial banks reported their best year in a while. Combined net profit for FY 2082/83 came in at roughly NPR 69.78 arba, up 32 percent from the year before, according to ShareSansar’s tally of the banks’ fourth-quarter reports. Headlines called it a recovery. Yet three months earlier, Nepal Rastra Bank (NRB) had reported that the banking sector’s non-performing loan (NPL) ratio had climbed to 5.60 percent at the end of the third quarter, and was still rising.
Bad loans up. Profits up sharply. Both can be true at once, and the reason is almost always the same line in the income statement: loan loss provisioning. Nepal banks set aside money against loans that may not be repaid, and that set-aside is the single biggest swing factor between a bad year and a great one. It is also the line most retail investors skip.
Here is the position this piece takes. On NEPSE, a bank’s headline profit tells you less than its provisioning does. If you learn to read one line, read this one. It tells you how much pain the bank has already recognized, how much it is still deferring, and whether this year’s dividend is being paid out of real earnings or out of a reserve being let go.
What loan loss provisioning actually is
When a bank lends NPR 1 crore, it books the loan as an asset worth NPR 1 crore. But not every borrower pays. A provision is the bank’s admission, in advance, that some of that asset is worth less than its face value. The bank charges the provision as an expense in the income statement, which reduces profit, and parks the same amount as an allowance against the loan book on the balance sheet, which reduces the net value of loans.
Two things follow. First, provisioning is not cash leaving the bank. Nobody is paid. It is an accounting estimate of a future loss. Second, because it is an estimate, it can move in both directions. When a bad loan is recovered or upgraded, the provision is reversed, and that reversal flows back into profit as income. That second point is where most of the interesting stories on NEPSE come from.
How loan loss provisioning works at Nepal banks: the NRB buckets
NRB does not leave the estimate entirely to bank management. Its Unified Directives sort every loan into categories based mainly on how long payments are overdue, and attach a minimum provision to each. The commonly cited structure, drawn from summaries of NRB’s Unified Directives 2081, looks like this:
- Pass: paying on time or only briefly overdue. Minimum provision of 1 percent.
- Watchlist: overdue by up to three months, or showing other early warning signs. Minimum provision of 5 percent.
- Substandard: overdue by three to six months. Minimum provision of 25 percent.
- Doubtful: overdue by six months to a year. Minimum provision of 50 percent.
- Loss: overdue by more than a year. Provision of 100 percent.
Pass and watchlist together make up “performing” loans. Substandard, doubtful and loss are the non-performing loans that feed the NPL ratio you see quoted in every quarterly report.
The pass-loan rate is worth a closer look, because NRB has been cutting it. In its Monetary Policy for 2025/26, NRB stated that the loan loss provision on pass loans “has been gradually reduced from 1.2 percent to 1.0 percent.” The last step, from 1.1 percent to 1 percent, came in the mid-term review of the 2024/25 policy in February 2025, as reported by NewsPolar at the time. The stated aim was to free up lending capacity in a slow economy.
That policy choice has a direct earnings effect that rarely gets mentioned. Pass loans are the overwhelming majority of any bank’s book. When the required provision on them falls by a fifth of a percentage point, banks need less in their allowance, and the excess can be released. That release is profit that did not come from lending a single extra rupee.
A worked example: one borrower, three bad years
Numbers make this concrete. Take a mid-sized hypothetical bank with an annual net profit of about NPR 300 crore. One of its corporate borrowers, a hotel developer, has an outstanding loan of NPR 100 crore. The figures below are illustrative, not drawn from any real bank.
Year one, everything is fine. The loan is pass. The bank holds a 1 percent provision: NPR 1 crore.
Year two, the hotel misses payments for four months. The loan drops to substandard. Required provision jumps to 25 percent, or NPR 25 crore. The bank already holds NPR 1 crore, so it must charge an extra NPR 24 crore to its income statement. One borrower has just taken 8 percent off the bank’s profit before tax effects.
Year three, the hotel is more than a year overdue. The loan is now loss. Required provision is 100 percent, NPR 100 crore. Another NPR 75 crore is charged. One loan has now consumed a quarter of a year’s profit.
Now scale that up. A bank with five or six loans like this in the same year, which is not unusual when a whole sector like hospitality, construction or cement slows down, can watch its profit evaporate even while its interest income holds steady. That is the mechanism behind several of the weakest results of the last two years.
Run the film backward and you get the opposite. Suppose the hotel restructures, pays down arrears, and the loan is eventually upgraded. Some or all of that NPR 100 crore provision is reversed, and it lands in the income statement as a write-back. The bank’s profit jumps, not because it earned more from customers, but because it stopped expecting a loss it had already booked.
Why provisioning is where rising NPLs actually bite
Most coverage treats the NPL ratio as the headline number. It is not. The NPL ratio tells you what share of the loan book is in trouble. Provisioning tells you what that trouble is costing shareholders, and when.
Consider the third-quarter FY 2082/83 data NRB released, as reported by Nepal News and Khabarhub on May 17, 2026. NIC Asia Bank had the highest NPL ratio among commercial banks at 8.85 percent, with NPR 22.32 arba in loan loss provisions. Nepal Investment Mega Bank had an NPL ratio of 8.41 percent and the largest provision pile in the sector at NPR 28.45 arba. Kumari Bank held NPR 28.20 arba against an NPL ratio of 6.94 percent. At the other end, Everest Bank reported an NPL of just 0.61 percent and Standard Chartered Bank Nepal 1.81 percent.
Those provision figures are cumulative balances, years of charges stacked up. They are the cushion. What matters for this year’s profit is the change in that cushion. A bank whose NPLs are high but already fully provided may have its worst behind it. A bank whose NPLs are rising faster than its provisions is pushing the cost into future years.
This is also why the gap between NPLs and dividend capacity matters. ShareSansar’s analysis of FY 2082/83 results shows that four banks reported negative distributable profit: NIC Asia, Himalayan, Nepal Investment Mega and Prabhu. Distributable profit is what is left after statutory and regulatory reserves are carved out of net profit, and provisioning-related reserves are a large part of that carve-out. Our explainer on distributable profit versus net profit walks through the full calculation.
The NFRS 9 twist: two provisions, one bank
Since FY 2081/82, there has been a second layer. Under NRB’s NFRS 9 Expected Credit Loss Related Guidelines, 2024, Nepali banks moved to full implementation of the expected credit loss (ECL) model, the international accounting approach that asks a bank to estimate future losses using probability of default and loss given default, not just how many days a loan is overdue.
So now every bank effectively computes two numbers: the provision required under NFRS 9 ECL, and the minimum provision required under NRB’s directives. NRB’s Procedure for Approval of Financial Statement Publication and Dividend Distribution, 2082, issued in August 2025, settles which one wins. If the NFRS provision is lower than NRB’s directive-based requirement, the shortfall must be transferred to a regulatory reserve. The same procedure requires capitalized interest and grace-period interest to be parked in that reserve too.
In plain terms, the bank books whichever is higher. If the accounting model says the loss is smaller than NRB’s rulebook demands, the difference does not hit the income statement, but it does get locked away from shareholders in the regulatory reserve. It cannot be paid out as a dividend.
This creates a trap for investors who look only at net profit. A bank can report healthy profit under NFRS while a growing chunk of it is being diverted into the regulatory reserve. EPS looks fine. Distributable profit per share does not. If you have ever wondered why a “profitable” bank declares a token dividend, this is often the answer.
Write-backs: when a profit jump is not a profit jump
Here is the uncomfortable part of the FY 2082/83 recovery story. A 32 percent jump in sector profit happened in a year when, by NRB’s own third-quarter data, the NPL ratio kept rising. Some of the biggest percentage gains came from banks that had suffered the heaviest provisioning hits the year before. Himalayan Bank’s profit rose more than tenfold from a very low base, and NIC Asia swung from a loss of NPR 3.87 arba to a small profit of NPR 0.18 arba, according to ShareSansar and Ratopati.
Some of that is genuine. Banks that front-loaded provisions in FY 2081/82 had less left to charge in FY 2082/83, and lower deposit costs helped margins. But part of a recovery like this typically comes from three provisioning-related sources that do not repeat:
Lower fresh charges. If the worst loans were already provided at 100 percent last year, they cost nothing this year even if they are never recovered.
Write-backs. Recoveries and upgrades reverse old provisions into income.
Rule changes. The cut in the pass-loan rate to 1 percent mechanically lowers required provisions on the healthiest part of the book. NRB’s own Monetary Policy for 2025/26 also tightened one rule in the other direction: a non-performing loan that becomes regular for six months must now first move to watchlist for six months before it can return to pass. That slows how quickly upgrades can generate write-backs.
This is opinion, but an informed one: a bank-wide earnings recovery built partly on reduced provisioning in a year of rising bad loans is a recovery to be skeptical of. It does not mean the profit is fake. It means it is lower quality, and harder to repeat.
How to spot provisioning trouble early
You do not need a Bloomberg terminal for this. Every commercial bank publishes a quarterly unaudited report on its website and on NEPSE, and they all follow NRB’s format. Our guide on how to read a bank’s quarterly report covers the full layout. For provisioning specifically, check four things.
1. The impairment charge line. In the statement of profit or loss, find “Impairment charge/(reversal) for loans and other losses.” A positive number is a charge that reduced profit. A negative number, often shown in brackets, is a reversal that added to profit. Compare it with the same quarter last year. If net profit grew sharply and this line flipped from a large charge to a reversal, most of your growth story is in this line.
2. Net interest income against net profit. Net interest income sits above provisioning. If net profit is up 40 percent and net interest income is flat, the gap is almost certainly coming from provisioning or one-off items.
3. NPL against provisions over time. Track the NPL ratio for four quarters, then check whether the provision balance on the balance sheet grew at a similar pace. NPLs rising faster than provisions means the bank is under-reserving relative to its problems. The cost has not gone away. It has been postponed.
4. Distributable profit per share. This is in the “Statement of Distributable Profit or Loss” that NRB requires. If it is far below EPS, regulatory reserves, including the NFRS shortfall, are absorbing earnings. That gap caps the dividend no matter what management says at the AGM.
A fifth, softer signal: watchlist. Watchlist loans are performing loans that are already slipping, and they are where tomorrow’s NPLs come from. A bank with a swelling watchlist and a falling NPL ratio may simply be catching problems one category earlier. Our piece on Nepal’s rising bank NPLs looks at how the sector-wide trend developed.
What provisioning means for bank valuations on NEPSE
Bank shares on NEPSE are usually priced off two numbers: price to earnings and price to book. Provisioning distorts both.
On earnings, the distortion is timing. Provisions bunch up when the economy turns and unwind when it recovers, so a bank’s EPS swings far more than its underlying business does. Buying a bank because its trailing P/E looks low after a write-back year is buying at the top of the provisioning cycle, not the bottom.
On book value, the distortion is adequacy. Book value is only as honest as the loan valuation behind it. If a bank is under-provided, its true book value is lower than reported, and a stock that looks cheap at 1.1 times book is more expensive than it appears. If a bank is conservatively provided, the reverse is true, and future write-backs are a hidden asset. The core banking ratios guide explains how provisioning also feeds into the capital adequacy ratio, which NRB’s 2082 dividend procedure now ties directly to whether a bank can pay a cash dividend at all.
There is also a quiet asymmetry worth naming. Banks that provided heavily and early, taking the pain in FY 2081/82, are the ones with the most potential for write-backs as the economy stabilizes. Banks that provided slowly have the opposite profile: their reported profits may look steadier now, but the bill is still waiting. The market tends to reward the steady-looking bank. That is often backwards.
Fact, estimate and opinion, separated
To be explicit about what this piece rests on.
Fact: NRB’s directives set minimum provisions by loan category. NRB cut the pass-loan provision from 1.2 percent to 1 percent across 2024/25. NFRS 9 ECL became fully effective in FY 2081/82, and NRB’s 2082 procedure requires any shortfall against directive-based provisions to go into the regulatory reserve. The sector NPL ratio was 5.60 percent at the end of the third quarter of FY 2082/83. Sector net profit for FY 2082/83 rose about 32 percent.
Estimate: The worked hotel-loan example is illustrative. How much of the FY 2082/83 profit jump came from lower provisioning versus genuine margin improvement varies by bank and is not published as a sector total.
Opinion: A profit recovery that leans on lower provisioning while bad loans are still rising is lower quality than it looks. Provisioning, not headline profit, is the better guide to a Nepali bank’s real earning power.
The verdict
Loan loss provisioning is the line where a bank’s problems become shareholders’ problems. The NPL ratio is the warning light. Provisioning is the bill. And because the bill can be paid early, paid late, or partly refunded later through write-backs, it is also the line that does the most to make bank earnings on NEPSE look better or worse than the underlying business.
Before you buy a bank share on the back of a strong year, open the latest quarterly report and find the impairment line. Ask whether profit grew because the bank earned more, or because it set aside less. Check whether distributable profit is close to net profit or far below it. If the answer is “set aside less” and “far below,” the strong year is probably borrowed from the next one.
The banks that will pay steady dividends through this cycle are not the ones with the best-looking profit this year. They are the ones that already took their provisioning medicine.
This is analysis, not financial advice.
