Three to six months of expenses, in cash, before you invest a dollar.
Ask where the six came from. Nobody knows. It's not derived from anything — not median unemployment duration, not any distribution of emergency costs, not a model. It's a round number that got repeated until it became furniture.
I'm not saying don't hold cash. I hold cash. I'm saying the rule is a placeholder for a calculation you were supposed to run yourself, and almost nobody runs it.
The Rule Ignores Every Variable That Matters
"Six months" is the same answer for:
- A tenured government employee with a working spouse and disability insurance
- A commission-only contractor with no dependents and three clients
- A solo founder with no income and a mortgage
These people have wildly different risk profiles. One of them could arguably hold two months. One should probably hold twelve. Giving all three the same number isn't simplification — it's abdication dressed as advice.
The variables that actually determine the right size:
Income volatility. Salaried with a stable employer is a different animal from commission, contract, or equity-heavy comp. This is the single biggest input and the rule ignores it entirely.
Correlation between your income and your investments. If you work in tech and your portfolio is tech-heavy, your job and your assets crash together. That's the scenario the buffer exists for, and it's the scenario where the buffer is most needed and most likely to be undersized.
Number of independent income sources. Two earners in unrelated fields is close to a buffer on its own. One earner is not.
Access to non-cash liquidity. Roth contributions come out penalty-free. A HELOC exists. A taxable brokerage can be sold in a day. These aren't as good as cash, but pretending they're worth zero massively overstates how much cash you need.
What your actual disasters cost. Not "six months of expenses" in the abstract — the real, specific numbers. Insurance deductibles. Your car's likely failure cost. Your actual health plan's out-of-pocket max. Most "emergencies" are four-figure events, not income replacement events, and those are the ones that actually happen.
Too Much Cash Is A Real, Compounding Cost
The rule is treated as risk-free. It isn't.
$30,000 sitting in cash for twenty years instead of invested is not a neutral decision. At a 7% real return, that's the difference between $30,000 and roughly $116,000 — an $86,000 decision, made by default, usually without anyone framing it as a decision at all.
And people don't hold six months. They hold six months and then keep going, because cash feels safe and investing feels like an action that requires confidence. I've watched people carry two years of expenses in a savings account for a decade while describing themselves as bad with money. They weren't bad with money. They were extremely good at one thing and it cost them a house.
Inflation is the other half. Cash at 2% in a 3% inflation environment loses purchasing power every year with perfect reliability. A "safe" 100% certain small loss is still a loss — it just doesn't feel like one because the number on the statement never goes down.
What The Buffer Is Actually For
Here's the reframe that made it click for me.
The emergency fund's job is not to cover emergencies. Its job is to stop you from having to sell assets at a bad time or take on high-interest debt.
That's it. That's the entire function. It's sequence-of-returns insurance for your personal balance sheet.
Which means the right question isn't "how many months feels safe." It's: what's the largest plausible cash need that would otherwise force me to sell into a crash or hit a credit card at 22%?
For a lot of people that number is much smaller than six months, because job loss is gradual (severance, notice, unemployment insurance) while the genuinely sudden costs — the transmission, the deductible, the flight home — are in the low thousands.
For some people it's much larger. If your income is volatile and correlated with the market, six months is optimistic.
Either way it's a derived number, not a received one.
My Own Number, And Why It Undercuts My Whole Argument
Here's mine, and it's going to complicate everything I just said.
My fixed monthly floor is $2,601 — mortgage $1,005, groceries and household ~$1,000, utilities $270, car insurance $145, internet $133, phone $48. Add $215/month for the Claude Code Max plan, which is a real business expense, and total burn is $2,816/month.
I hold $20,000. That's 7.7 months of full burn, or just under 8 months of the fixed floor.
Now the honest part. I did not derive $20,000 from a risk assessment.
I picked it because it's a round number. Twenty thousand dollars is easy to remember, easy to aim at, and it feels like an achievement when I look at the account. The eight months is an output, not an input. I never sat down with correlation coefficients and unemployment distributions — I picked a number that felt substantial and let the month count fall where it fell.
Which means I did exactly what this post spends a thousand words criticizing: I inherited a number instead of deriving one. Mine was just a rounder, bigger, self-selected number rather than a rounder, smaller, internet-selected one.
Why I'm Not Changing It
Because the psychological argument is real and the spreadsheet argument quietly assumes it isn't.
A buffer you will actually fund and not raid beats an optimal buffer you won't. $20,000 is legible. I can hold it in my head, I know instantly whether I'm above or below it, and hitting it felt like clearing a bar. A derived $17,431.88 has none of those properties. It's more correct and less sticky, and stickiness is what determines whether the money is still there in year three.
So the defensible version of my position isn't "eight months is the right amount of risk cover." It's: the round number bought me compliance, and compliance is worth some inefficiency.
The cost of that choice is real and I can price it — roughly $3,100 more cash than six months of burn would require. At a 7% real return over twenty years, holding that extra sits somewhere around $12,000 of forgone growth. That's what the nice round number costs me. I've looked at the figure and I'm still keeping it, which is a decision rather than an accident, and that's the part I'd defend.
The Part Where The Rule Is Right
I'll concede the important half: for someone with no buffer and no habit, "save six months" is excellent advice. It's directionally correct, easy to remember, and enormously better than zero. The failure mode it prevents — one bad month becoming a credit card balance that takes four years to clear — is the single most common wealth-destroying event there is.
So if you have nothing, go build three months and don't overthink it. The rule is a fine default when the alternative is paralysis.
The myth isn't that you need cash. It's that six is a number rather than a shrug.
Derive It, Don't Inherit It
Write down your floor. Write down your largest plausible single-event cost. Write down how correlated your income is with your investments, and how long you'd actually take to replace it.
You'll get a number. It almost certainly won't be six.
And if you then override it with a rounder one because the round one is the one you'll actually stick to — fine. That's what I did. Just do it knowing what you overrode and what the override costs, instead of inheriting a rule and never checking whether it applies to you at all.
The myth was never that you need cash. It's that somebody else's number could ever be yours by default.
Run it yourself. The rule was never an answer — it was a placeholder for the fact that nobody could size it for you.