How big should an emergency fund be? Three months is the floor, not the answer

"Three to six months of expenses." Every article says it, and it is the only piece of personal-finance advice that gives a hospital consultant on a fixed salary and a seasonal roofer exactly the same answer.

That cannot be right, and the reason it is not right has a name and a formula. This guide gives you both, plus the honest picture of where people actually stand.

What people actually have

46% of US adults have at least three months of expenses saved. 30% have some savings but less than three months. 24% have nothing set aside at all. In the same survey, only 47% said they could cover a $1,000 emergency from savings.

The Federal Reserve's own survey tells a similar story at a smaller amount: in its 2025 data, 63% of adults said they would cover a surprise $400 expense entirely with cash, savings, or a credit card paid off at the next statement — and 12% said they could not pay it by any means at all.

Two things follow from those numbers, and they point in opposite directions. Most people are not close to three months, so a bigger target can look absurd. But the households with the least stable income are exactly the ones for whom three months is furthest from enough. Averaging over that tension is how "three to six" became the standard answer — and how it ended up being wrong for almost everyone individually.

Why one number cannot fit everyone

An emergency fund does one job: it pays your fixed costs during the months your income does not.

So the size you need depends on two things — how much you spend, which the standard rule does account for, and how much your income varies, which it does not. A salary that lands on the same day for the same amount has near-zero variance; the fund only has to survive a job loss. An income that swings between $640 and $5,120 in the same year is exposed to something else entirely: not one dramatic event, but an ordinary bad quarter.

The measure of that swing is the coefficient of variation:

CV = STDEV(last 12 months of net income) / AVERAGE(last 12 months of net income)

It is a ratio, so it works at any income level — a $2,000-a-month freelancer and a $20,000-a-month agency owner can be compared directly. In words:

CV	Shape of your income
under 0.20	Steady. Close to a salary.
0.20 – 0.45	Bumpy. Normal for freelance work.
0.45 – 0.75	Volatile. The buffer has to do real work.
over 0.75	Feast and famine. Non-negotiable.

The rule

months_needed = MIN(12, MAX(3, ROUNDUP(3 + 12 × CV, 0)))
buffer_target = months_needed × monthly_committed_costs

Three parts, each doing a job:

• The 3 is the base — the cover you need even if your income never varies, because the risk of losing it entirely does not depend on how steady it is.
• The 12 × CV is the part everyone else leaves out. Each 0.1 of volatility adds roughly 1.2 months, because a more dispersed income produces both deeper troughs and longer runs of them; the shortfalls compound rather than alternate.
• The cap at 12 stops the rule running away. Past a year of cover you are holding a large amount of idle cash, and the right response to volatility above that point is to change the income, not to keep hoarding against it.

Note what the chart shows: at any real level of variability the answer is already above three months. Three is the floor of this curve, and only an income with literally no variance sits on it. If your income moves at all, the standard advice was under-specified for you from the start.

Worked example

Take a freelancer with $3,502.50 of committed monthly costs — the self-paycheck they draw plus their fixed bills — and twelve months of net income that produced a CV of 0.53.

months = ROUNDUP(3 + 12 × 0.5273) = ROUNDUP(9.33) = 10
target = 10 × $3,502.50 = $35,025

Against the standard advice of three to six months — $10,508 to $21,015 — the measured rule asks for $35,025. That is not the rule being pessimistic. That freelancer had a month that brought in $640 against $3,502.50 of costs, a hole of $2,862.50 in a single month, and it was not their only bad one. Six months of cover would have been a real risk of running out.

What counts as "committed monthly costs"

Not your average spending. The number you need is what you cannot stop paying inside a month:

• Include: rent or mortgage, utilities, insurance, minimum debt payments, groceries, transport to work, childcare, phone, and any subscription you would actually keep during a bad quarter.
• Exclude: discretionary spending you would genuinely cut, savings contributions other than the buffer itself, and anything you could pause with a phone call.

Be honest in both directions. Padding the number makes the target unreachable and you will abandon it; stripping it to the bone makes the fund fail at the moment you need it, which is the only moment it matters.

Getting there when the target looks impossible

$35,025 is a number that makes people close the tab. Three things make it tractable:

1. Count in months, not dollars. "$8,000 saved" is not information. "2.3 months of runway" is, and it goes up in a way you can feel. Runway is just balance ÷ committed_costs.
2. The first month of cover is worth more than the tenth. Going from 0 to 1 month removes the entire category of emergencies that currently go on a credit card at 22%. Going from 9 to 10 does much less. Front-load the effort and let the later months take years if they need to.
3. Fund it from surplus, not from discipline. If your income is variable, the good months are where the buffer comes from. Draw a fixed self-paycheck, leave the rest, and the fund fills itself without any month feeling like a sacrifice — the percentile method sets that fixed amount.

Recheck it quarterly

CV is not a constant. Losing a large client, adding a retainer, or moving from project work to contract work all change the shape of your income, and the target should move with it. Recompute the standard deviation and average on a rolling twelve months every quarter; if the buffer target went up, that is the mechanism telling you something real about your risk, not a mistake in the arithmetic.

The short version

Three to six months is a placeholder for a calculation nobody makes you do. Measure your own volatility, add 1.2 months of cover for every 0.1 of it, cap it at a year, and count what you have in months rather than dollars. The number you get will be specific to you, and you will be able to say exactly why it is what it is.

---

Sources: Bankrate Annual Emergency Savings Report (fieldwork 2–8 December 2025, n = 2,564 US adults); Federal Reserve, Report on the Economic Well-Being of U.S. Households in 2025.