How big an emergency fund actually needs to be

"Three to six months" is a slogan, not a number — the figure a household actually needs falls out of its essential monthly expenses, how long gaps really last, and how many incomes stand behind it.

Updated September 2026

Three to six months of what, exactly?

The familiar advice leaves out the two things that decide the answer: what the fund has to replace each month, and how long it has to keep doing it. Get them wrong and a fund can look generous while covering half the runway the household thinks it has.

The fund has to replace essential expenses, not income. A household earning $94,000 gross does not need to replace $94,000 a year. During a gap, saving stops, restaurant spending shrinks, and retirement contributions pause. What does not stop is rent, groceries, power, transportation to interviews, and childcare.

The line between essential and discretionary

The federal Consumer Expenditure Survey shows where money goes before anyone starts trimming. For 2024 — the most recent year published, released December 19, 2025 — average annual spending was $78,535 per consumer unit. The full category breakdown of that total, and what it implies for a needs-versus-wants split, is worked through in the 50/30/20 budget rule guide; what matters here is one line inside it. Of the $10,169 that went to food, $6,224 was groceries and $3,945 food away from home — close to two-fifths of the food budget in a largely pausable category. The release publishes those dollar amounts but not their shares of food, so that fraction is arithmetic on published figures, not a BLS statistic.

Applied to one household spending $5,600 a month — every dollar figure below is illustrative, not a survey average:

Essentials come to $4,240 a month, about 76 percent of total spending and about 71 percent of this household's roughly $6,000 of monthly take-home pay — well above the 50 percent needs share the 50/30/20 budget rule guide screens on, which is exactly the structural-cost signal that guide reads a high needs share as, rather than a discipline problem. That is roughly $978 a week. That weekly burn rate is the number worth writing down, because unemployment duration is published in weeks. Working through a version of this table in the budget calculator, then feeding the essentials total into the emergency fund calculator, produces a defensible figure instead of a slogan.

How long gaps actually last

BLS publishes two duration measures, and the distance between them is the whole argument about three months versus six. For September 2026 (published October 2, 2026), the median duration of unemployment was 11.5 weeks and the mean 24.8 weeks. The same release reported a 4.2 percent unemployment rate, 7.1 million unemployed, and 1.9 million unemployed for 27 weeks or longer — 27.1 percent of all unemployed people, up from 23.6 percent a year earlier.

Measured against a $978 weekly essential burn:

The two ends of the familiar rule happen to bracket those measures: 11.5 weeks is about two and a half months, 24.8 weeks a little under six. That is a coincidence, not the origin of the rule — no published source traces "three to six months" to these statistics. Two mismatches matter. The median row is not three months of essentials: three months here is $12,720, nearly $1,500 above the 11.5-week figure. And six months, $25,440, buys about 26 weeks — just short of the 27-week line defining long-term unemployment, so the 27.1 percent already past that line are by definition in spells six months would not have covered. How much of that tail to fund for depends on how specialized the role is, how concentrated local employers are, and how far essentials could be cut again.

Netting unemployment insurance against the burn rate

Those rows assume no money arrives at all. Unemployment insurance is run by each state within federal guidelines, so the amount and the duration both depend on the state law the claim is filed under. As the Department of Labor describes the general pattern, benefits are a percentage of a worker's earnings over a recent 52-week period, subject to a maximum amount set by the state, and benefits can be paid for a maximum of 26 weeks in most states. Additional weeks are available in some periods of high unemployment. Because the percentage, the cap and the number of weeks are all set state by state, the only reliable figure for any one household is the one published by its own state agency.

The arithmetic consequence is large. Illustratively, a $500 weekly benefit — within the range of state maximums in those tables — leaves $478 a week of the $978 burn to find from savings, so 26 weeks of that gap costs about $12,400 rather than the $25,440 the same 26 weeks would cost with nothing arriving. A household expecting benefits can subtract its own expected weekly amount from the burn rate before reading the table above; one that does not expect to qualify should read it as written.

Why the right number differs by household

One earner versus two

BLS family data for 2025 annual averages, released April 23, 2026, found both spouses employed in 49.1 percent of married-couple families and only one spouse in 23.4 percent; among married-couple families with children, both parents were employed in 66.3 percent.

Suppose the illustrative household has two earners taking home $3,100 and $2,900. If the first job goes, the remaining $2,900 covers 68 percent of the $4,240 of essentials, leaving a $1,340 monthly shortfall — so a $12,720 fund covers it for about nine and a half months. A single-earner household with identical essentials burns the same $12,720 in three months flat.

The caveat is correlation. If both earners share an employer, an industry, or a one-employer town, the two incomes are not independent protections and the dual-income discount largely disappears.

Salaried versus commission, self-employed, or variable income

For irregular earners the fund smooths recurring dips rather than bridging one cliff. The Federal Reserve's 2025 Survey of Household Economics and Decisionmaking, published May 13, 2026 and fielded in October 2025 among nearly 13,000 adults, found 30 percent of adults had income that varied at least occasionally — similar to 2024, and up from 28 percent in 2023 — while 11 percent struggled to pay bills because of it. Among the self-employed, 58 percent said income varied month to month and 22 percent struggled with bills, against 28 percent and 10 percent among people working for someone else.

A practical method: pull 24 months of net income, find the worst three consecutive months, and fund that gap. If the worst quarter averaged $3,100 a month against $4,240 of essentials, the shortfall is $1,140 a month, and covering all three months costs about $3,420 on top of the job-loss buffer. The horizon comes from the method — three months, because that is what was measured — not from any published figure.

Dependents

Dependents do not add a flat percentage; they move the essentials line itself. Childcare alone is $880 a month in the table above, and it is among the hardest lines to pause, because places are usually lost rather than suspended. Sizing off essentials makes that increase flow through automatically — which is why essentials, and not income, are the right base.

Where the money sits — and what it costs to wait

What makes this fund different from every other savings target is timing, not return. The money has to be available in days, because the rent falls due before the next paycheck arrives, and that constraint is set by the calendar rather than by anything the balance earns while it waits. We do not recommend particular places to hold it.

What that waiting costs in purchasing power is set out in the guide to what inflation does to money sitting still. The consequence specific to this page is narrower, and it is about maintenance rather than erosion: the target is denominated in last year's essentials, so it goes stale on its own without anyone spending a dollar. A fund that covered six months of groceries, rent and fuel on the day it was sized covers slightly less of the same basket a year later, which makes re-sizing part of keeping the fund rather than an optional refinement. Re-running it annually, and whenever rent changes, a dependent arrives, or the income mix shifts, keeps it honest. The savings goal calculator turns the gap into a monthly contribution, and the fund belongs on the balance sheet in the net worth calculator rather than being counted twice as both a buffer and an investment.

The mistake: sizing against income

The most common error is multiplying income rather than expenses. Three months of this household's roughly $6,000 monthly take-home pay is $18,000 — 1.4 times the $12,720 that three months of essentials requires, and 1.8 times if the multiple is taken off the $7,800 of gross monthly pay instead, as it sometimes is. The overshoot has a cost: the target feels unreachable, so the fund never gets finished, and an unfinished fund is the one that fails. It fails the other way too — where essentials have crept close to take-home pay, "three months of income" and three months of essentials are nearly identical, so income-based sizing leaves no margin and quietly assumes spending can be cut when it cannot.

What these numbers cannot tell you

• Duration statistics describe who is unemployed now, not how long a search takes. BLS states that these measures reflect still-in-progress spells rather than completed ones, and cautions against reading them as how long it takes to find a job. Two effects pull in opposite directions: truncation at the survey date understates spells still running, while a point-in-time count over-represents long spells. Neither 11.5 nor 24.8 weeks forecasts a search length.
• The median describes the shorter half. Half the people unemployed in September 2026 had been looking for 11.5 weeks or less at that point, and the mean is more than double that because a long tail of continuing spells exists. Planning to the median plans for the shortest half of spells in progress.
• The $400 question measures method, not balance. The widely quoted 63 percent of adults who said they would cover a hypothetical $400 emergency expense exclusively with cash, savings, or a card paid off at the next statement — unchanged from each of the previous three years, and down from a high of 68 percent in 2021 — describes how people would pay, not what they hold. The more comparable figure is the three-month one: 55 percent had savings covering three months of expenses in 2025, unchanged from 2024 and down from a 59 percent high in 2021, while 30 percent could not cover three months by any means. The share who do hold three months rises with income — 21 percent of adults with family income under $25,000, 39 and 55 percent in the middle bands, 75 percent at $100,000 or more.
• Job loss is not the only call on the fund. A transmission, an out-of-pocket medical bill, and a flight home for a funeral can land in the same month as a layoff. Sizing purely to unemployment duration assumes the shocks queue politely.
• One data caveat. The 2025 family-employment figures rest on 11-month averages that exclude October, one instance of a gap running through several 2025 BLS annual series; the salary to hourly guide quotes what BLS says it does to year-over-year comparisons.

None of this is advice about any particular situation. It is the arithmetic and the published evidence behind a number usually quoted without either.

Common questions

How often should the target be recalculated?

At least annually, and whenever a fixed essential line moves — rent, childcare rates, medical costs, or the number of earners behind the household. The fund is denominated in a basket that keeps moving: the CPI rose 3.4 percent in the 12 months to August 2026, so a target set a year earlier already buys less than it did. Once the essentials table exists, recalculating is quick, because only the changed lines need updating.

What does a part-built fund actually buy?

Proportional time, which is not the same as nothing. At the illustrative $978 weekly essential burn, $5,000 covers about five weeks of full essentials, and considerably longer if any income or benefits arrive during the gap. That is why the first few thousand dollars does the heaviest work: it converts an immediate crisis into a decision with some time attached. A half-built fund is a working fund, not a failed one.

Does unemployment insurance change the target?

It changes the burn rate rather than the logic. Benefits replace part of prior wages for a limited number of weeks, and the weekly amount, the number of weeks and the eligibility rules are all set by each state. The practical step is to find the state's own figures and subtract the expected weekly benefit from the weekly essential burn before sizing the fund. A household that may not qualify — and eligibility is genuinely state-specific — should size as though nothing arrives.

Is childcare really an essential that cannot be paused?

Usually, for two reasons that have nothing to do with arithmetic. A place given up is typically not held open, so pausing can mean re-entering a waitlist at whatever the new rate turns out to be. And the job search itself needs coverage, because interviews, assessments and travel have to happen somewhere. That is why it sits on the essential side of the illustrative table at $880 a month despite being one of the largest single lines on it.

How should a variable-income household use its worst quarter?

Treat it as a second, smaller buffer with a different job. The job-loss fund bridges a cliff; the smoothing buffer covers recurring dips. Measuring the worst three consecutive months out of the last 24 gives a shortfall that can be funded directly — $1,140 a month in the illustrative case, or about $3,420 for the full quarter. Whether the two buffers sit in one balance or two is bookkeeping; what matters is not counting the same dollars for both jobs.

Where should the fund be held?

We do not name places to put money. The trade-off worth reasoning about is that the fund has to be reachable in days, which rules out anything carrying a lock-up or a settlement delay, while anything that simply sits loses purchasing power at roughly the rate of inflation. Those two requirements genuinely conflict, and how a household resolves them depends on how fast it might need the money and how large the fund has grown relative to its near-term risks.

Sources

• BLS — The Employment Situation, September 2026 (dated archive) — Median duration 11.5 weeks and mean duration 24.8 weeks; 4.2 percent unemployment rate; 7.1 million unemployed; 1.9 million unemployed 27 weeks or longer at 27.1 percent of all unemployed; released October 2, 2026. Re-fetched and all six figures confirmed.
• FRED — Median Duration of Unemployment (UEMPMED) — Supporting series for the 11.5-week September 2026 median. Secondary to the dated BLS release.
• FRED — Average (Mean) Duration of Unemployment (UEMPMEAN) — Supporting series for the 24.8-week September 2026 mean. Secondary to the dated BLS release.
• BLS TED — Long-term unemployed share, September 2025 — Year-earlier comparison: long-term unemployed were 23.6 percent of all unemployed in September 2025.
• BLS — Consumer Expenditures, 2024 (dated archive) — Average annual expenditures $78,535 per CONSUMER UNIT; housing $26,266/33.4 percent, transportation $13,318/17.0 percent, food $10,169/12.9 percent, food at home $6,224, food away from home $3,945, healthcare $6,197/7.9 percent, entertainment $3,609/4.6 percent; released December 19, 2025. Re-fetched: the release uses 'consumer unit' throughout and does NOT publish food-away-from-home as a share of food.
• BLS — Employment Characteristics of Families, 2025 (dated archive) — Both spouses employed in 49.1 percent of married-couple families, one spouse in 23.4 percent, both parents employed in 66.3 percent of married-couple families with children; the 11-month-average caveat from the uncollected October 2025 survey; released April 23, 2026. Re-fetched and confirmed.
• BLS — Consumer Price Index, August 2026 (dated archive) — 12-month CPI-U change of 3.4 percent, shelter 3.0 percent, food 2.7 percent; released September 11, 2026. Re-fetched: the dated archive returns the August 2026 release and carries all three figures.
• BLS — Labor Force Statistics from the CPS: Concepts and Definitions — Duration of unemployment measures reflect still-in-progress spells rather than completed spells, and should not be read as how long it takes to find a job. Replaces the draft's thinner citation to the 'How the Government Measures Unemployment' page.
• Federal Reserve — Economic Well-Being of U.S. Households in 2025: Savings and Investments — 63 percent would cover a $400 expense with cash or its equivalent; 55 percent had three months of expenses saved in 2025, unchanged from 2024 and down from a 59 percent high in 2021; 30 percent could not cover three months by any means; three-month share by family income 21 / 39 / 55 / 75 percent. Re-fetched: the table header is 'Have savings to cover three months of expenses', so the income-band figures are shares who HAVE the savings.
• Federal Reserve — Economic Well-Being of U.S. Households in 2025: Income and Expenses — 30 percent of adults had income that varied at least occasionally — a similar level to 2024 and up from 28 percent in 2023; 11 percent struggled to pay bills because of it; self-employed 58 percent and 22 percent; those working for someone else 28 percent and 10 percent. Re-fetched: page confirms the 'similar to 2024' wording and contains none of the draft's $3,100/$1,140/$5,700 figures.
• Federal Reserve — Economic Well-Being of U.S. Households in 2025: Executive Summary — The 63 percent $400 share was 'unchanged from the previous three years, but down from a high of 68 percent in 2021' — four flat readings, not three.
• Federal Reserve — Survey of Household Economics and Decisionmaking (SHED) overview — 2025 report published May 13, 2026; survey of nearly 13,000 adults fielded October 2025.
• U.S. DOL ETA — Significant Provisions of State Unemployment Insurance Laws, effective July 2026 — State weekly benefit formulas (commonly 1/26 of the average of the two highest quarters, i.e. about half of average weekly wages, with explicit-percentage states at 50-60 percent), state maximum weekly amounts, and the 'Benefit Weeks Payable' column showing most states capped at 26 weeks with several lower and many varying with earnings history and the state unemployment rate. Replaces the draft's citation to the state-laws index page, which does not itself list benefit weeks.
• U.S. Department of Labor — Unemployment Insurance — Each state administers a separate unemployment insurance program and sets its own eligibility guidelines.

Read the other guides, or browse every calculator.