The version most people reach for: marginal cost must always be below average total cost, and the two curves eventually meet when output gets large enough. That's wrong on both counts. Marginal cost crosses average total cost at exactly one point — the minimum of the ATC curve — and it does so because of a mathematical relationship between any marginal quantity and its corresponding average.
Why the mistake is the natural reading
When you see the standard U-shaped ATC curve and the MC curve rising steeply beside it, the picture looks like MC is "chasing" ATC from below. That framing makes intuitive sense: if each new unit costs more to produce, shouldn't the running average always stay ahead?
The trap is conflating two different types of numbers. Average total cost tells you the cost-per-unit across all units produced so far. Marginal cost tells you the cost of producing the next unit only. These are not the same quantity — they don't move together, and there's no rule that one must stay below the other across the whole output range.
Before MC crosses ATC, marginal cost is below average total cost. That means each new unit costs less than the current average, so adding it pulls the average down. After MC crosses ATC, marginal cost is above average total cost. Each new unit costs more than the current average, so adding it pushes the average up. The crossover — the point where MC equals ATC — is precisely where the average stops falling and starts rising. That's the definition of a minimum.
This logic applies to any average/marginal pair. It has nothing special to do with production costs; it's a general arithmetic fact.
The correct mechanism
Define average total cost at output $q$ as:
$$ATC(q) = \frac{TC(q)}{q}$$
For ATC to be at a minimum, its derivative with respect to $q$ must equal zero:
$$\frac{d(ATC)}{dq} = \frac{MC(q) \cdot q - TC(q)}{q^2} = 0$$
This requires:
$$MC(q) \cdot q = TC(q)$$
Dividing both sides by $q$:
$$MC(q) = \frac{TC(q)}{q} = ATC(q)$$
So at the minimum of ATC, MC must equal ATC — not by coincidence but by definition. The minimum cannot occur anywhere else.
A useful way to keep this straight: MC is the derivative (instantaneous rate of change) of total cost; ATC is the ratio (running average) of total cost to output. Whenever the marginal value is below the average, the average falls. Whenever the marginal is above the average, the average rises. Equality means the average is momentarily flat — a minimum (on the way down) or a maximum (on the way up; not relevant here because ATC is U-shaped, not inverted-U).
Worked example
Suppose a firm has the following total cost schedule:
| Output (q) | Total Cost ($) | ATC ($/unit) | MC ($/unit) |
|---|---|---|---|
| 1 | 12 | 12.00 | — |
| 2 | 20 | 10.00 | 8 |
| 3 | 27 | 9.00 | 7 |
| 4 | 36 | 9.00 | 9 |
| 5 | 47 | 9.40 | 11 |
| 6 | 60 | 10.00 | 13 |
At $q = 3$: MC = 7, ATC = 9.00. MC is below ATC — each new unit costs less than the current average, so ATC fell to reach 9.00 (down from 10.00 at $q = 2$). In this discrete schedule, ATC ties at 9.00 across $q = 3$ and $q = 4$, which is the expected shape of any integer cost table at its minimum.
At $q = 4$: MC = 9, ATC = 9.00. MC exactly equals ATC. The average is flat — this is the minimum.
At $q = 5$: MC = 11, ATC = 9.40. MC is now above ATC, so the average is rising.
The crossing happens exactly at $q = 4$, the minimum of ATC. Notice that MC doesn't need to be below ATC everywhere before the crossing — it just needs to be below ATC for ATC to still be declining. Once MC rises above ATC, ATC climbs with it.
This also shows why the statement "MC is always below ATC" is false: at $q = 5$ and $q = 6$, MC (11, 13) is well above ATC (9.40, 10.00).
How to internalize it
Think of a class grade average. If your score on the next exam is below your current average, your average drops. If it's above your current average, your average rises. The moment your next score equals your current average, the average is momentarily unchanged — that's its turning point. MC and ATC work exactly the same way, with "next unit cost" in place of "next exam score."
Two quick checks to run when you see a cost-curve diagram:
- Where MC is below ATC, ATC must be sloping downward. If the diagram shows ATC flat or rising in that region, something is wrong.
- The MC curve intersects ATC from below. If the diagram shows MC cutting through ATC from above (i.e., MC was higher, then becomes lower), that's not a standard competitive-firm cost structure.
The same marginal-pulls-the-average logic governs average variable cost, not just ATC — MC also crosses AVC at its minimum for identical reasons. The two minimums occur at different output levels because AVC excludes fixed costs.
If you're also working through the idea that sunk (fixed) costs don't affect marginal decisions, notice the connection: fixed costs raise ATC above AVC at every output level, but they contribute nothing to MC (since MC is the cost of one more unit, and fixed costs don't change with output). That's why the MC curve passes through both the AVC minimum and the ATC minimum — it's the same curve, crossing two different averages.
Check yourself
A firm is producing 50 units. At that output, marginal cost is $18 and average total cost is $22. What happens to average total cost if the firm increases output to 51 units?
A. ATC rises, because marginal cost is positive. B. ATC falls, because marginal cost is below average total cost. C. ATC stays the same, because marginal cost hasn't reached its minimum yet. D. ATC rises, because the firm is past the minimum-cost output level.
Correct answer: B.
Marginal cost ($18) is below average total cost ($22), so the 51st unit costs less than the current per-unit average. Adding it pulls the average down. ATC falls.
Close the gap
The MC/ATC relationship is the kind of thing that reads clearly on a diagram and then evaporates the moment you're asked to reason through a new scenario. The gap isn't the definition — it's applying the marginal-pulls-the-average logic under time pressure, when a question frames the curve in an unfamiliar way.
A tutor that walks you through cost-curve problems in real time can catch the moment you conflate "marginal is below average" with "marginal is small" — the same session where it happens, before it calcifies into a habit.