NPS moves the risk rather than removing it

What a defined contribution actually defines

NPS declares no interest rate. What is fixed is what goes in and how it is invested. What comes out is whatever those investments are worth on the day you stop — and EPF and PPF work the other way round.

That difference has a name, and the name arrives last here because the mechanism is what matters. An old-style pension promised an income, calculated from your salary and years of service. Whatever it cost to fund that promise was somebody else's problem — the employer's, or the state's. If markets disappointed, the shortfall was theirs to make up.

NPS promises nothing about the income. It promises that your contribution is invested in the allocation you chose and that you own the result. The institution has stopped carrying the outcome; you carry it now. A scheme built the first way is called defined benefit; one built the second way is defined contribution.

This is not a criticism of the design. A defined-benefit promise has one characteristic failure: someone must fund a guarantee whose true cost is not knowable on the day it is made, and the shortfall surfaces decades later in somebody else's budget. A scheme that cannot fail that way has a real advantage. The trade is plain enough: the scheme becomes solvent by construction, and the variability lands on the individual. The three phases of a retirement plan read differently once you know which of the two you are in.

Inside the account there are units, not a balance

Money paid into NPS does not sit as a balance that later gets credited with interest. It buys units of a pension fund at a price struck from what that fund holds — the same per-unit valuation mechanism used by mutual funds, the NAV.

Two things follow, and both are worth being exact about. Your unit count rises only when you contribute. The value of each unit moves with the securities the fund owns, every day, in both directions. Nothing is added to your account for the passage of time, because nobody has promised a rate that time could accrue.

You choose how the money is spread across asset classes — equity, corporate debt and government bonds among them — within limits the scheme sets, and that choice is the single largest thing under your control. It decides how much of your corpus is exposed to the equity market and how much to interest rates. It does not decide the outcome; it decides which uncertainty you are holding.

So the honest description of an NPS statement is a valuation, not a balance. Read on a good day it flatters, read on a bad one it alarms, and neither reading tells you anything about the number that will be there in twenty years.

EPF, PPF and NPS: the same goal, three places to put the risk

Set the three side by side and the useful comparison is not which returns more. It is which variable each one holds still, and which it lets move.

The current figures make the contrast concrete. EPF's rate for the year is 8.25% for FY 2025-26 — and the route that number travels is part of the point. The EPFO's Central Board of Trustees recommends a rate; the Central Government ratifies it; EPFO then directs that it be credited to accounts. Three stages, months apart, and this year's rate has completed all three. Coverage that reports a recommendation as though the money were already in the account is describing stage one as stage three.

PPF pays 7.1%, reset by notification each quarter. That is a different mechanism again — not a board and a ratification, but a periodic administrative decision that can change the rate on money already deposited.

One acronym to keep separate while reading any of this. GPF is a fourth scheme, for government employees, and it is neither PPF nor EPF despite the family resemblance in the name. Its rate is not stated here, and a rate quoted for one of these three under the label of another is the commonest factual error in Indian coverage of them — the acronyms are not interchangeable.

The conversion nobody rehearses

At exit, a portion of the corpus does not come to you as money. It is applied to an annuity — a contract that pays a stated income for as long as you live.

What that income will be is not known in advance, and it is not a function of your corpus alone. An annuity is priced from prevailing long-term interest rates and from mortality assumptions at the moment the contract is written, and the rate is then set for the life of the contract. The same corpus, converted in two different rate environments, buys two different incomes for life. What an annuity is and why its headline return looks poor is a subject of its own; what matters here is when the price is struck.

The arithmetic is worth doing once, on figures chosen because they divide easily and not because anyone is quoting them. A corpus of ₹50 lakh converted at an annuity rate of 6% pays ₹3 lakh a year for life. The same ₹50 lakh converted at 5% pays ₹2.5 lakh — a sixth less, every year, for as long as the contract runs, and nothing about the saver changed.

And the price is struck once. Consider the asymmetry against everything else in the same plan. Contributing monthly for thirty years spreads your purchase price across hundreds of dates, which is the entire reason staggered investing is discussed at all. The conversion reverses that discipline in a single transaction: one date sets the income for the remainder of a life.

One further feature of the conversion, mentioned because it is the other half of the same trade: a level annuity income is fixed in rupee terms, so inflation works on it for as long as it is paid. Options exist that start lower and rise, and the comparison between them belongs with the annuity itself rather than here.

The one place the two risks partly cancel — and where they do not

The accumulation risk and the conversion risk are not independent, and the direction of the relationship is worth knowing because it is counterintuitive and entirely mechanical.

Take the bond side of the corpus first. When long-term interest rates fall, the market price of bonds already issued rises, because their older, higher coupons become more valuable — the mechanism is worked through in duration and interest-rate risk. So a fall in rates lifts the value of the debt portion of your corpus. That same fall in rates is what makes an annuity more expensive per rupee of income. The two move in opposite directions: a larger corpus buying less income per rupee, and the reverse when rates rise.

That is a partial offset, and the word doing the work is partial. It is not a hedge anyone has sized, and this article is not going to size it either. The durations involved do not match, mortality assumptions move for reasons of their own, and only the share of the corpus that is actually converted ever meets an annuity price at all.

The equity side has no such relationship at all. A rise in the equity market lifts the corpus and does nothing whatever to the price of an annuity — so that portion of the corpus arrives at the conversion date carrying its full, unoffset exposure. Which is why the composition of the corpus in the years approaching exit is a different question from its composition at thirty-five, and why preservation is treated as a phase of its own rather than a milder version of accumulation.

Read as a claim, this one is mechanical about the signs and a judgement about the magnitude. The signs are not debatable. How much they offset is, and anyone who tells you the two cancel out has quietly upgraded a direction into a quantity.

The number that looks like a rate and is not one

The specific error this topic produces, over and over: putting an NPS scheme's reported return next to EPF's credited rate and treating the larger one as better.

They are not the same kind of number. EPF's is a rate that was decided and then paid — it describes money that is in the account. An NPS return figure describes what a portfolio happened to do over a stated past period. One is an outcome that has been fixed by a process; the other is a measurement of what already happened, carrying no commitment about what happens next. Past performance does not indicate future results, and in this comparison that line is the substance rather than a formality.

The second half of the mistake runs the other way, and it is the more comfortable one. A declared rate removes the uncertainty about the number. It does not remove the uncertainty about what the number buys — a fixed rate against an inflation path nobody has fixed is still an uncertain real outcome, and the certainty on offer is narrower than it feels.

The comparison that survives both problems is structural rather than numerical. Ask which variable each scheme holds still, who absorbs the year it goes wrong, and at what moments you are exposed. The table above answers those three for all four rows. A single return figure answers none of them.

Which parts of NPS are rules — and why they are not in this article

Everything above is mechanism. It follows from how the scheme is built and it will still be true after the next amendment. A second category of fact about NPS behaves completely differently, and mixing the two is how stale figures get a second life.

These are the rules, and this article deliberately states none of them:

• How much of the corpus must be applied to an annuity at exit.
• How many account types the scheme has, and what each permits.
• The ages, lock-ins and conditions attached to any withdrawal.
• Every tax relief, its section, its ceiling and the regime it belongs to.

All four are set by statute and regulation. All four have moved before, and a reader who memorises a number from an article dated two years ago is carrying a figure the article had no way to update. Check these at the source — the regulator's own site and the current Act — rather than in anyone's explainer, including this one.

The tax point does need its mechanism, because the mechanism is stable even where the figures are not. Relief on a retirement contribution can take two structurally different forms: a deduction that reduces the income you are taxed on, or an arrangement where money routed by an employer is not treated as your income in the first place. Those are different instruments with different ceilings and different conditions, and they are frequently described as one.

Which of them is available to a given person depends on a question most articles skip entirely. Under the Income-tax Act 2025 the new regime is the default (s.202), and it does not carry the general deduction at s.123 — the section formerly numbered 80C — at all. So "NPS saves tax" is not a complete sentence until it names a regime; the choice between them is the subject of the regime comparison. Income for the financial year to March 2026 is still assessed under the repealed 1961 Act, so both numberings are live during the transition.

What can be tested, and what cannot

The accumulation side can be examined against evidence. The conversion side cannot. Of the two uncertainties in this article, exactly one has a record standing behind it.

Start with the testable one. How much a corpus outcome depends on the start date, how deep the falls along the way were, and how wide the spread of outcomes was across every historical starting point are all questions you ask of a price history for a comparable allocation. The answer is a distribution rather than a number, and the useful part of it is the worst window, not the average — the same discipline that governs withdrawal rates.

The conversion side is not testable in the same way. There is no historical record of the rate that will prevail on your exit date, and no amount of studying past annuity pricing produces one. That asymmetry is itself the finding: the part of the plan people research hardest is the part where evidence exists, and the part that fixes their income for life is the part where it does not.

FNOTrader's Mutual Funds app runs on the full AMFI NAV history — around 34 million NAV rows — and reports rolling-return distributions across every available start date, alongside maximum drawdown, for any scheme and period. That covers the first of the two questions above for an equivalent allocation. Historical outcomes describe what happened, not what will happen.

FNOTrader is not a SEBI-registered investment adviser and none of this is retirement advice. What mix of these schemes suits a particular person turns on facts an article does not have — starting with how large a corpus the plan needs and how much of the eventual income has to be certain rather than merely likely.

Common questions

Continue reading

More in Retirement Planning · App: Mutual Funds · Definitions: glossary · Free tools: calculators · All: every article