Where an emergency fund should actually sit

The short version

The arrangement that passes all three tests below is a split: a small tranche in a savings account, the bulk in a sweep deposit or a scheme that lends only for very short periods, and part of it at a bank other than the one your salary lands in.

How much should be in it at all is a different question, answered in full in the emergency fund guide. This article assumes you have a rupee figure already and takes only the next decision: where those rupees go.

That decision gets written up almost everywhere as a yield comparison, which is the wrong axis entirely. The money's job is to be intact, present and obtainable on a day you are not thinking clearly. Yield is the fourth thing it has to do, and the gap between the good answers is smaller than most people assume — the last section prices it.

Three tests, and they are ordered

Not a list of considerations to balance. An ordered list, where a later test is only worth applying to options that already passed the earlier ones.

One — can you have it within a day or two, on a day of your choosing, without asking anyone's permission. Not “can it be liquidated”. Money you can get on the fourth working day is not an emergency fund; it is a short-term investment that will be described as one right up until the fourth working day arrives.

Two — can it be worth less than you put in, ever, over any window you might actually hold it. Not “is it likely to fall”. The fund exists because shocks arrive at bad moments, so the relevant question is what the balance does in exactly the conditions that produce the shock. Anything whose value is set by a market rather than by a contract fails this test to some degree, and the degree is what distinguishes the options.

Three — can you reach it while distracted and frightened. This is the test nobody writes down and the one that fails most often in practice. A person dealing with a hospital admission or a termination letter has roughly the executive function of someone with a bad fever. Anything requiring a forgotten password, a co-signatory, a physical branch visit, or a decision about which of five holdings to break, will be attempted badly or not at all.

Return is fourth. It is not zero — over a decade, money parked in something losing ground to inflation is a real cost, and pretending otherwise is dishonest. But it is fourth, and the ordering is the whole argument of this article.

The realistic options, scored on those tests

Four places in India where this money credibly goes. Each passes some tests and fails others, and the failures are more informative than the passes.

Two rows deserve unpacking, because both are routinely described in terms that are more definite than the underlying reality.

A sweep-in or auto-sweep arrangement links a term deposit to a savings account so that balance above a threshold converts to deposit, and converts back the moment a withdrawal needs it. It is a bank product, not a regulated standard — the sweep threshold, the deposit tenure, the order in which deposits break, and what a broken deposit actually earns are all set by the bank and differ across banks. The concept is worth understanding through the deposit arithmetic and the account type it attaches to; the terms are worth reading on the specific account.

Liquid and overnight schemes are the row where honesty matters most. Their unit price is set by what the underlying paper is worth, so it can fall — small, rare, and not impossible, which is why they fail test two where a deposit passes it. They keep that risk small by lending for very short periods, so a rate move has almost no time to act on the holding. What they hold and how the two differ is a separate question with its own answer; the relevant point here is that buying into these categories has an earlier daily cut-off than other schemes — 1:30 pm IST — and that redemption timing, which is the property this article cares about, varies enough that it is worth confirming on the scheme document rather than assuming next-day.

What does not appear in the table: equity of any kind, gold, property, and anything with a lock-in. Those fail test two, test one, or both, and the emergency fund guide covers why.

The trap: your fund and your salary at the same bank

One class of emergency takes out the fund and the account it is reachable through at the same moment, and it is the one every checklist leaves off.

Ask what set of events would make you need the fund urgently. Job loss, a medical event, an accident, a family emergency — and one more that gets left off every list: the bank itself being the problem. A co-operative bank placed under RBI directions that cap withdrawals. An account frozen on a court or tax attachment. A block after suspected fraud on the account. A KYC lapse that locks the login. A multi-day technical outage.

In every one of those, an emergency fund held at that bank is not slow. It is gone for the duration, and the duration is not yours to set.

Now follow the chain one step further, which is where it gets genuinely counterintuitive. A liquid fund is not a bank deposit and is not affected by a bank failing. But a mutual fund redemption pays out to the bank account registered against the folio. If that registered account is at the bank under directions, the redemption completes perfectly and the money credits into an account you cannot withdraw from. The holding was diversified away from the bank. The payout rail was not, and the rail is what you were actually relying on.

Deposit insurance does not solve this either, and it is worth being precise about why. The cover is capped at ₹5 lakh per depositor per bank, principal and interest together — the mechanics, including how joint holdings and different capacities are treated, are set out separately and are less generous than most summaries suggest. But the ceiling is the second problem. The first is timing: where RBI places a bank under All Inclusive Directions, the interim insured payout runs to 90 days, and the liquidation route has its own separate timetable. Insurance answers “will I get it back”. It does not answer “can I pay the hospital on Tuesday”, and the emergency fund exists to answer the second question.

So the structural fix is a second bank — ideally a different kind of institution, on the reasoning that banks of the same type face the same pressures — holding a tranche large enough to cover a few weeks. It costs a little admin, and it is the one arrangement above in which the fund survives the scenario that made you need it. The DICGC per-bank ceiling gives a second, independent reason to split, which is why the deposit insurance article is worth reading alongside this one.

The test nobody runs

Test three — reachable while distracted — is the only one you can actually verify in advance, and almost nobody does.

The failures are mundane and they are specific. A netbanking login not used in eight months, now needing a password reset that sends an OTP to a phone in a hospital bag. A fund folio whose registered bank account was closed two years ago, so the redemption bounces. A deposit in a joint account with “jointly” operating instructions, needing a signature from someone who is the reason for the emergency. A transfer that hits a per-transaction cap on the channel you tried, at 23:00 IST, when the payment has to be made now — the caps differ by channel and by bank, and finding that out during an emergency is the wrong time.

All of these are discoverable in twenty minutes on an ordinary Tuesday. The check that finds them is a small live withdrawal, once a year — move ₹1,000 out of the fund into your spending account by the exact route you would use for real, and note how long it took and what it asked for. A drill that has never run is an assumption.

Two things to fix while you are in there. Confirm the nomination on each account and folio, since a fund that requires a succession process is not available to the family it was meant to protect. And write the whole arrangement down somewhere a spouse or parent can find it — which bank, which scheme, which login — because an emergency fund only one person can locate has a single point of failure that is not financial.

The friction that helps, and the friction that bites

Separating the fund from your spending account is genuinely useful: money you can see is money you will eventually spend, and a fund sitting in the account your card draws on is not an emergency fund so much as a large balance. Distance is doing real work there.

But distance has a documented failure mode that catches people who follow that advice well. An account with no customer-induced transaction for two years is classified inoperative. The classification places duties on the bank rather than penalties on you, but what it means at the counter is a reactivation step — fresh paperwork, at exactly the moment you wanted the money. After ten years an unclaimed balance leaves the bank's own books entirely; the mechanics, and how to claim it back, are their own subject.

The repair costs nothing: one small transfer a year, in or out, scheduled on a date you will not forget. The annual drill in the previous section is already that transaction, which is a pleasant coincidence — one habit satisfies both.

Worth separating from all of this: predictable annual bills — premiums, school fees, festival spending — are not emergencies and do not belong in this pot. They belong in a sinking fund saved for on a schedule. Mixing them is the most common way the fund is empty on the day it is needed.

What the return difference is actually worth

On a fund of ₹3 lakh, one percentage point of annual yield is ₹3,000 a year, or ₹250 a month. That is the whole size of the fourth test, and it is worth computing rather than waving away, because “return doesn't matter here” is usually asserted rather than shown.

Whatever gap you can see between the rates printed on your own savings account and your own deposit or scheme, multiply it by ₹3,000 and you have the annual cost of choosing the slower option. It is a real number and you can compute it exactly, which is better than arguing about it.

Then price the other side. A ₹2 lakh hospital admission on a Sunday, met on a credit card because the money was a working day away, costs whatever that card charges from the day of the transaction until it is cleared — and a cash withdrawal on a card conventionally attracts a fee and no interest-free period at all, so the meter starts immediately. The rate is on the statement. Put it next to the ₹250.

That comparison is the reframe worth carrying away. The yield decision is a known, small, continuous cost. The availability decision is a rare, large, discontinuous one — it costs nothing at all in most years and a great deal in the year it binds. Optimising the first at any expense to the second is trading a certain small gain for an uncertain large loss, which is the shape of the trade nobody would take if it were described that way.

Two honest qualifications, because this cuts both ways. If the fund is large — say a year of expenses for someone self-employed, where a long income gap is the realistic scenario — the yield gap stops being ₹250 a month and becomes worth structuring for, with a small instantly-available tranche and the bulk somewhere better paying. And a fund losing to inflation for a decade is a real erosion, not a rounding error. Neither qualification changes the ordering. Both change how much effort the fourth test deserves once the first three are satisfied.

Checking a parking option on its own record

Everything above is decidable from documents you already have: the account terms, the scheme information document, and your own statements. The one claim that is not — how a particular scheme has actually behaved, as opposed to how its category is described — is a data question rather than an argument.

FNOTrader's Mutual Funds app runs on AMFI's full record of daily per-unit prices — the net asset value, or NAV — around 34 million rows of it, so a candidate scheme's worst single-day fall and its deepest drawdown are inspectable rather than assumed. For money whose entire job is being intact on the day it is needed, the drawdown record is the relevant column and the return column is not.

Historical figures describe what happened over the period stated. They are not a forecast, and past performance does not indicate future results. This describes how the decision is structured; FNOTrader is not a SEBI-registered investment adviser and does not recommend schemes.

Six ways the parking decision goes wrong

1. Optimising the yield first — picking the highest-paying option that is “basically liquid”, then discovering what “basically” meant. The tests are ordered for a reason.
2. All of it at the salary bank — the single arrangement that fails in the specific scenario where the bank is what went wrong, and the one almost everybody defaults into.
3. Diversifying the holding but not the payout account — a scheme at a different fund house whose redemption still credits the one bank you cannot withdraw from.
4. A non-callable deposit — bought for the marginally better rate, unbreakable before maturity, and therefore not an emergency fund at all whatever it is labelled.
5. Leaving the account untouched for years — the distance that stops you spending it also drifts the account towards inoperative status and a reactivation queue.
6. Never testing the route — the password, the mandate, the joint operating instruction and the channel cap are all fine in theory and all discovered on the worst possible day.

Only the first and the fourth are selection errors — a wrong product. The other four are arrangement errors, and nothing about them shows up in a yield table. That is the pattern worth noticing: this decision is more often lost on plumbing than on the product. A related structural point sits in medical emergency planning, where the fund and the insurance policy have to work in the same hour and usually have not been tested together.

Common questions

Continue reading

More in Budgeting & Saving · App: Mutual Funds · Definitions: glossary · Free tools: calculators · All: every article