“Never Spend More on a Repair Than the Car Is Worth” — Check the Arithmetic

You have an estimate in your hand for $3,200. Somebody looks it up and the car is worth $2,400. The rule arrives within about a minute: never spend more on a repair than the car is worth.

It is the most repeated sentence in this decision and it compares two figures that answer different questions. What a car is worth is the price of selling it. What a repair costsis the price of keeping it. A car's market value is set by what a stranger would pay to take on its remaining life — which is exactly the thing you are about to buy back, cheaper, by fixing it.

The ratio between those two numbers tells you nothing about which path is cheaper. Here is what does.

The one figure that rests on no forecast

A repair bill is not a cost. It is a purchase of miles, and the price is per mile:

repair cost per mile = the estimate ÷ miles it buys before the next decision

$3,200 buying 3 more years at 12,000 miles a year

  $3,200 ÷ 36,000 miles = $0.089 a mile

Nine cents a mile. The estimate is a written number and your annual mileage is a known number, so that division is exact — it is the only figure in this whole decision that carries no guess at all except the one in the denominator.

Now price the alternative the same way. A $28,000 replacement losing 20% of its value a year, over that same three years:

$28,000 × 0.8³ = $14,336 left at three years
loss in value  = $13,664

$13,664 ÷ 36,000 miles = $0.380 a mile      before a drop of fuel goes in

Thirty-eight cents against nine. Depreciation alone on the newer car is more than four times the repair, per mile, and depreciation never sends an invoice. That asymmetry is why the rule fails so reliably: it weighs a visible bill against an invisible one.

The whole comparison, both sides

Per-mile figures are a way in, not the answer. The honest version prices both paths over the same window and charges each for what it spends, crediting each for what it still owns at the end. Same three years, same 36,000 miles, gasoline at $3.19:

KEEP IT                                  REPLACE IT
  repair              $ 3,200              price               $28,000
  fuel @ 24 MPG       $ 4,785              old car sold as-is  −$ 2,400
  upkeep to come      $ 2,400              fuel @ 34 MPG       $ 3,378
  insurance           $ 3,900              upkeep              $ 1,500
  registration        $   450              insurance           $ 5,700
                                           registration        $ 1,050
  worth at the end    −$   900             worth at the end    −$14,336
  ─────────────────────────────            ─────────────────────────────
  net over 3 years    $13,835              net over 3 years    $22,892

                                           keeping it wins by  $ 9,057

The replacement is genuinely better on fuel — $1,407 cheaper over three years — and it loses anyway, by seven times that. Note also that the repair, the thing the entire argument was about, is under a quarter of the cost of keeping the car. Fuel is larger.

Watch the $2,400. It appears once, as cash coming inon the replace side, and only because that cash arrives only if you actually sell the car. Counting it again on the keep side — as value you “lose” by not selling — is the commonest way this comparison is got wrong, and it always tilts the answer toward buying.

What the rule is actually fumbling for

The rule is not useless. It is a bad proxy for a real question it never asks out loud: how much longer will this car last? That question lives in the denominator, and it is the only place the answer is genuinely sensitive:

the same $3,200 estimate

  buys 36,000 miles    $0.089 a mile     obviously worth doing
  buys  8,000 miles    $0.400 a mile     now it is a real argument
  buys  3,000 miles    $1.067 a mile     you are buying miles at a dollar each

On a car near the end, a big bill genuinely is money thrown away — not because it exceeds the car's value, but because it buys so few miles. The rule stumbles onto the right answer in that case for the wrong reason, which is why it survives.

You cannot know the remaining life, but you can bound it better than you think. What failed, and is the rest of the car in the same condition? A water pump on an otherwise sound car at 130,000 miles is a different proposition from a water pump on a car that also needs struts, tires and an exhaust. Ask the mechanic what else they saw while it was on the lift — the single most useful question in this decision, and it costs nothing.

Then test whether your guess is even carrying the decision. Run the numbers at half your estimate of the remaining life and at double it. If keeping the car still wins at half, stop refining the guess; it is not what the answer turns on. If the answer flips between the two, the guess isthe decision, and what you need is a second opinion on the car's condition rather than a more precise calculator.

Why there is no threshold table here

The federal government answers “when should a vehicle be replaced?” with exactly such a table. Under 41 CFR § 102-34.270, sedans and station wagons hit their minimum replacement standard at “3 years or 60,000 miles”, whichever comes first. It is a floor for a fleet, and it exists because nobody can inspect several hundred thousand vehicles one at a time.

You can inspect yours. You have a written estimate, you know what else the mechanic found, and you know your own mileage — strictly more information than any age-and-mileage cutoff can use. That is the whole reason this is a calculator and not a rule.

It is also worth knowing how ordinary this position is. S&P Global Mobility put the average age of the 289 million light vehicles on US roads at 12.8 years in 2025, and passenger cars at 14.5 — an eighth consecutive annual rise. The average car on the road is well past the age at which this question starts getting asked, and most of the people asking it are keeping the car.

Two things the arithmetic cannot price

Risk. Every figure above is a middle expectation, and it says nothing about the spread around it. An older car has a wider spread: usually it costs what you entered, and occasionally it strands you two states from home on a holiday weekend. A newer one has a narrower spread at a higher average. What that tail is worth depends on things no calculator can ask — a second car in the driveway, a job you can miss a morning of, a repair fund.

The useful way to read a result is as the price of that insurance. If keeping the car is $9,000 cheaper over three years, that is what you are being paid to carry the risk, and you can decide whether it is enough. If the two paths land within a few hundred dollars, the arithmetic has told you it is not the deciding factor — reliability is, and you should decide on that.

Whether the repair is a good repair. This is a cost comparison and nothing else. It does not know whether the estimate is fair, whether the part will hold, or whether the car should be on the road. A mechanic answers those, and a second written estimate is the cheapest thing you can buy in this entire process.

Running it on your own car

Five things you can get before you decide, none of them expensive: the written estimate, what else the mechanic saw, what the car would fetch as-is today, your real annual mileage, and an insurance quote on whatever you would replace it with.

Then the Repair or Replace Calculatorprices both paths over the same window, prints the repair's cost per mile, and shows how much the answer moves if your guess about the remaining life is wrong by half. It prints the value ratio too — labelled, because someone will quote it at you before this is over, and it is better seen than argued with.

Two neighbours to this question. If what you are really weighing is a thirsty old car against an efficient new one, that is a break-even: Does a Hybrid or EV Pay for Itself? works through it. And if you keep the car, the figure worth knowing next is what it costs you all-in — True Cost per Mile puts the repair beside depreciation, fuel, insurance, registration and interest, where it usually turns out to be smaller than it felt in the shop.