Derivatives in Economics: Marginal Cost, Marginal Revenue and Profit Maximisation
Here is the secret that makes the second year of economics much easier than the first: every time a textbook says “marginal”, it means “derivative”. Marginal cost is the derivative of total cost. Marginal utility is the derivative of utility. Once you see that, half of microeconomics collapses into one operation.
Economics students are often taught marginal cost as a table. Produce 10 units, total cost £100. Produce 11 units, total cost £108. So the marginal cost of the 11th unit is £8. Fine. Intuitive. And completely unable to answer the question the exam actually asks: at what output is profit maximised?
Tables can only tell you about the jumps you happened to tabulate. If the profit-maximising output is 10.4 units, a table stepping in whole numbers will never find it. You need something that works at every point, not just the ones in the rows.
That something is the derivative. And the moment you connect the word “marginal” to the operation “differentiate”, a large amount of economics stops being vocabulary and starts being arithmetic.
What a derivative actually measures
Forget the formal limit definition for a moment. A derivative answers one question: if I nudge the input by a tiny amount, how much does the output move?
That is exactly what “marginal” means in economics. Marginal cost asks: if I produce one more unit — really, an infinitesimally small bit more — how much does total cost rise? Marginal revenue asks the same about revenue.
The table approach measures this over a discrete jump: ΔTC/ΔQ. The derivative measures it at a single point: dTC/dQ. As the jump shrinks toward zero, the first becomes the second.
MC = dTC/dQ MR = dTR/dQ MPL = dQ/dL
Every “marginal” in economics is a derivative of the corresponding “total” with respect to the relevant quantity.
The translation table. Total cost → marginal cost. Total revenue → marginal revenue. Total utility → marginal utility. Total product → marginal product. In every single case the relationship is the same: the marginal function is the derivative of the total function. There are no exceptions to learn.
The rules you actually need
Economics uses a surprisingly small slice of calculus. These five rules will carry you through almost every AP, A-Level and first-year undergraduate question.
| Rule | Statement | Example |
|---|---|---|
| Power rule | d(xn)/dx = nxn−1 | d(3Q³)/dQ = 9Q² |
| Constant rule | d(c)/dx = 0 | d(500)/dQ = 0 |
| Sum rule | Differentiate term by term | d(Q² + 5Q)/dQ = 2Q + 5 |
| Product rule | d(uv) = u’v + uv’ | Used for TR = P(Q)·Q |
| Chain rule | d(f(g(x))) = f'(g)·g'(x) | d(2Q+1)⁵/dQ = 10(2Q+1)⁴ |
The power rule does most of the work. The constant rule is why fixed costs never appear in marginal cost — they differentiate to zero. That single fact explains the shutdown rule, sunk cost reasoning, and why an airline will sell a last-minute seat for £30 that “cost” £180 to provide.
From total cost to marginal cost
Suppose a firm’s total cost function is:
TC = 2Q³ − 15Q² + 60Q + 200
Differentiate term by term. The 2Q³ gives 6Q². The −15Q² gives −30Q. The 60Q gives 60. The 200 — the fixed cost — gives zero.
MC = dTC/dQ = 6Q² − 30Q + 60
Notice what just happened. The £200 of fixed cost has vanished entirely. It is still being paid — it is still in TC — but it does not affect the decision about how much to produce, because producing one more unit does not change it. Calculus enforces the sunk cost principle automatically, without you having to remember it.
We can also extract average cost and check a famous relationship:
AC = TC/Q = 2Q² − 15Q + 60 + 200/Q
Note that fixed cost does survive in average cost, as the 200/Q term. That term shrinks as Q rises — which is precisely the “spreading the overhead” effect that drives economies of scale.
Why MC cuts AC at its minimum
Every textbook draws MC intersecting AC at the bottom of the AC curve, and most students memorise it without knowing why. Calculus makes it obvious.
AC is at a minimum when dAC/dQ = 0. Since AC = TC/Q, apply the quotient rule:
dAC/dQ = [Q·(dTC/dQ) − TC] / Q² = (MC − AC)/Q
Set that to zero. Since Q > 0, the only way the expression equals zero is if MC = AC. So AC is stationary exactly where MC crosses it. That is not a drawing convention — it is a mathematical necessity.
The intuition is the one everyone already has about test scores. If your next exam scores above your current average, your average rises. Below it, your average falls. Your average is flat only when the new score exactly equals it. MC is the “next score”; AC is the running average.
Marginal revenue and why it is not price
For a firm in perfect competition, price is fixed at P regardless of output. So TR = P·Q, and differentiating with P constant gives MR = P. Price and marginal revenue are the same thing.
For any firm with market power, they are not — and this is where students lose marks. If the firm faces a downward-sloping demand curve, selling more requires cutting the price on every unit, not just the last one.
Say inverse demand is P = 100 − 2Q. Then:
TR = P·Q = (100 − 2Q)Q = 100Q − 2Q²
MR = dTR/dQ = 100 − 4Q
Compare the two. Demand has intercept 100 and slope −2. MR has the same intercept 100 but slope −4 — exactly twice as steep.
A shortcut worth memorising. For any linear inverse demand P = a − bQ, marginal revenue is MR = a − 2bQ. Same intercept, double the slope. It falls out of the product rule every time, and it saves you a derivation under exam pressure — though you should still show the working if the question asks you to derive it.
Common error — assuming MR = P for a monopolist. This is probably the single most frequent calculus mistake in microeconomics papers. MR = P only under perfect competition, where the firm is a price taker. Any downward-sloping demand curve means MR < P at every positive output, because the price cut applies to all units sold, not just the marginal one.
Profit maximisation: where it all comes together
Profit is revenue minus cost: π = TR − TC. To maximise it, differentiate and set to zero.
dπ/dQ = dTR/dQ − dTC/dQ = MR − MC = 0
Therefore MR = MC — the first-order condition for profit maximisation.
That is the whole derivation. The most famous rule in microeconomics is one line of calculus. MR = MC is not a separate law that firms obey; it is simply what “the top of the profit hill is flat” looks like when you write it algebraically.
The condition everyone forgets
Setting the first derivative to zero finds a stationary point. It does not tell you whether you have found a maximum, a minimum, or a point of inflection. For that you need the second derivative.
d²π/dQ² < 0 → maximum | d²π/dQ² > 0 → minimum
Economically, d²π/dQ² < 0 means the slope of MR is less than the slope of MC — MC must be cutting MR from below. If MC cuts MR from above, you have found the point of minimum profit, which is a real solution to MR = MC and a completely wrong answer to the question.
Common error — stopping at the first-order condition. A cubic cost function typically produces two solutions to MR = MC: one is the profit maximum, the other the profit minimum. Examiners set these deliberately. Always compute the second derivative, state its sign, and say what it tells you. It is usually an explicit mark.
Full worked example
A monopolist faces inverse demand P = 120 − 3Q and has total cost TC = Q³ − 6Q² + 30Q + 50. Find the profit-maximising output, price and profit.
Step 1 — Build TR and derive MR.
TR = (120 − 3Q)Q = 120Q − 3Q²
MR = 120 − 6Q (note: same intercept, double slope — as predicted)
TR = (120 − 3Q)Q = 120Q − 3Q²
MR = 120 − 6Q (note: same intercept, double slope — as predicted)
Step 2 — Derive MC.
MC = dTC/dQ = 3Q² − 12Q + 30
MC = dTC/dQ = 3Q² − 12Q + 30
Step 3 — Set MR = MC.
120 − 6Q = 3Q² − 12Q + 30
0 = 3Q² − 6Q − 90
0 = Q² − 2Q − 30
120 − 6Q = 3Q² − 12Q + 30
0 = 3Q² − 6Q − 90
0 = Q² − 2Q − 30
Step 4 — Solve.
Q = [2 ± √(4 + 120)] / 2 = [2 ± √124] / 2 = [2 ± 11.136] / 2
Q = 6.57 or Q = −4.57. Reject the negative root — output cannot be negative.
Q* = 6.57 units
Q = [2 ± √(4 + 120)] / 2 = [2 ± √124] / 2 = [2 ± 11.136] / 2
Q = 6.57 or Q = −4.57. Reject the negative root — output cannot be negative.
Q* = 6.57 units
Step 5 — Check the second-order condition.
π = TR − TC = (120Q − 3Q²) − (Q³ − 6Q² + 30Q + 50) = −Q³ + 3Q² + 90Q − 50
dπ/dQ = −3Q² + 6Q + 90
d²π/dQ² = −6Q + 6. At Q = 6.57: −6(6.57) + 6 = −33.4 < 0 ✓ Maximum confirmed.
π = TR − TC = (120Q − 3Q²) − (Q³ − 6Q² + 30Q + 50) = −Q³ + 3Q² + 90Q − 50
dπ/dQ = −3Q² + 6Q + 90
d²π/dQ² = −6Q + 6. At Q = 6.57: −6(6.57) + 6 = −33.4 < 0 ✓ Maximum confirmed.
Step 6 — Find price and profit.
P = 120 − 3(6.57) = £100.29
π = −(6.57)³ + 3(6.57)² + 90(6.57) − 50 = −283.6 + 129.5 + 591.3 − 50 = £387.20
P = 120 − 3(6.57) = £100.29
π = −(6.57)³ + 3(6.57)² + 90(6.57) − 50 = −283.6 + 129.5 + 591.3 − 50 = £387.20
Notice that P = £100.29 while MC at that output is 3(6.57)² − 12(6.57) + 30 = £80.65. Price exceeds marginal cost by roughly £20 — the monopoly markup, sitting there in the numbers.
Elasticity as a derivative
Elasticity is usually taught as a ratio of percentage changes, which is fine for a two-point calculation. But point elasticity is a derivative in disguise:
PED = (dQ/dP) × (P/Q)
The derivative dQ/dP captures the slope; the P/Q term converts it into a unit-free percentage measure.
This matters because it connects to marginal revenue through one of the most useful relationships in the subject:
MR = P(1 + 1/PED)
Read what that implies. If demand is elastic (PED = −2), MR = P(1 − 0.5) = 0.5P — positive, so raising output raises revenue. If demand is unit elastic (PED = −1), MR = 0 — revenue is at its maximum. If demand is inelastic (PED = −0.5), MR = P(1 − 2) = −P — negative, so selling more actually reduces revenue.
This yields a result students find surprising and examiners love: a profit-maximising monopolist never operates on the inelastic portion of its demand curve. Why? Because MC is always positive, so MR = MC requires MR > 0, which requires elastic demand. The firm would always do better by raising price and selling less.
Case Study — Airline Seat Pricing and the Vanishing Fixed Cost
A Boeing 737 flying London to Edinburgh costs roughly £11,000 to operate — crew, fuel, landing fees, aircraft lease, maintenance. With 180 seats, the “average cost per passenger” is about £61. So no rational airline should sell a seat below £61. Except they do, constantly, and it is the profit-maximising thing to do.
The reason is the constant rule. Once the flight is scheduled and the aircraft is going regardless, nearly all of that £11,000 is fixed with respect to the passenger count. The marginal cost of one additional passenger is: a little extra fuel for their weight (roughly £2), a meal or snack if provided, and the credit card processing fee. Call it £8.
So MC ≈ £8, not £61. Any seat sold above £8 adds to profit, right up until the plane is full. This is why the last-minute seat is £30 while the average cost is £61 — and why the airline is behaving perfectly rationally, not dumping.
Two things follow. First, this is exactly why airlines invest so heavily in yield management: with MC near zero and capacity fixed, the entire profit problem becomes an exercise in price discrimination — charging each segment close to its willingness to pay. Second, it explains the industry’s notorious fragility. When MC is far below AC, competition drives prices toward MC and no one covers their fixed costs. The airline industry has destroyed more capital than almost any other, and the shape of its cost function is why.
Research Spotlight — Do Firms Actually Set MR = MC?
The uncomfortable finding: mostly, no — and economists have known this since 1939.
Hall and Hitch (1939), interviewing 38 British firms at Oxford, found that almost none calculated marginal cost or marginal revenue. Instead they used full-cost pricing: estimate average cost at normal output, add a conventional markup, charge that. The finding was so awkward that it triggered a decade of argument about whether the marginalist theory of the firm was empirically bankrupt.
Blinder, Canetti, Lebow and Rudd (1998) revisited the question with a far larger and more rigorous study — 200 structured interviews with US firms representing a broad slice of GDP. Their results largely confirmed Hall and Hitch. Most firms reported that marginal cost was roughly constant, not rising; many could not readily distinguish marginal from average cost; and price changes were infrequent, with a median of just 1.4 per year.
Machlup (1946) offered the standard defence, and it remains the best one: firms need not consciously compute derivatives any more than a cyclist consciously solves the differential equations governing balance. Competition selects for firms whose pricing approximates the optimum, whatever heuristic produced it. The model predicts behaviour; it does not claim to describe managers’ mental processes.
Why this matters for you: in an evaluation question, “MR = MC assumes firms know their cost and demand functions, which survey evidence from Hall and Hitch through Blinder et al. suggests they typically do not — though Machlup’s as-if defence means the model may still predict well” is a genuinely strong point. It shows you know the model’s limits without abandoning it.
The wider pattern
Once you internalise marginal-equals-derivative, a large amount of economics unifies:
| Concept | Derivative | Optimisation rule |
|---|---|---|
| Marginal cost | dTC/dQ | MR = MC for max profit |
| Marginal revenue | dTR/dQ | MR = 0 for max revenue |
| Marginal utility | dU/dX | MUx/Px = MUy/Py |
| Marginal product | dQ/dL | MRPL = wage for hiring |
| Marginal propensity to consume | dC/dY | Multiplier = 1/(1 − MPC) |
Every row is the same operation applied to a different total. And every optimisation rule is the same move: differentiate, set to zero, check the second derivative.
The natural next step is constrained optimisation — maximising utility subject to a budget, or minimising cost subject to an output target. That requires Lagrange multipliers, which extend exactly this logic to problems with a constraint attached.
AP & Cambridge A-Level Exam Technique
1. Write TR = P × Q explicitly before differentiating. Students routinely differentiate the demand curve instead of the revenue function and get MR wrong from the first line. Substitute the inverse demand into P·Q, expand, then differentiate.
2. Always check the second-order condition. State d²π/dQ², evaluate it at your Q*, give the sign, and say what it means. On cubic cost functions this is frequently a standalone mark and is the most commonly dropped one.
3. Reject negative roots explicitly. Quadratics give two answers. Write “Q = −4.57 is rejected as output cannot be negative.” Do not silently discard it — the examiner cannot award a mark for reasoning you did not show.
4. Watch fixed costs. They vanish from MC (constant rule) but survive in AC as the FC/Q term. A question that gives you a fixed cost is usually testing whether you know it drops out.
5. Substitute back into the right function. To find price, put Q* into the demand equation, not the MR equation. Putting Q* into MR gives you the marginal revenue at the optimum, which is not the price and is rarely what was asked.
6. Use MR = a − 2bQ as a check, not a substitute. If a question says “derive marginal revenue”, show the product-rule working. Use the shortcut to verify your answer, or when the derivation is not itself being assessed.
7. Interpret, do not just calculate. After finding Q* and P*, add a sentence: “Price exceeds MC by £20, indicating market power and an allocative inefficiency.” Numerical answers alone rarely reach the top band.
Practice Questions
Question 1 — Deriving marginal functions (5 marks)
A firm has TC = 4Q³ − 20Q² + 90Q + 400 and faces inverse demand P = 200 − 5Q. (a) Derive MC. (b) Derive MR. (c) State the marginal cost of production when Q = 0 and explain your answer.
(a) MC = dTC/dQ = 12Q² − 40Q + 90 [2]
The fixed cost of 400 differentiates to zero.
The fixed cost of 400 differentiates to zero.
(b) TR = (200 − 5Q)Q = 200Q − 5Q² [1]
MR = dTR/dQ = 200 − 10Q [1]
Consistent with the a − 2bQ shortcut: a = 200, b = 5, so MR = 200 − 10Q. ✓
(c) At Q = 0, MC = 12(0) − 40(0) + 90 = £90. [1]
This is the cost of producing the very first unit. It is not zero, and it is entirely independent of the £400 fixed cost — fixed costs are incurred whether or not any output is produced, so they do not enter the marginal cost of the first unit or any other.
Question 2 — Profit maximisation (8 marks)
A monopolist faces P = 90 − 2Q with TC = Q² + 10Q + 100. Find the profit-maximising output, price and profit. Verify the second-order condition.
Step 1: TR = (90 − 2Q)Q = 90Q − 2Q² → MR = 90 − 4Q [2]
Step 2: MC = dTC/dQ = 2Q + 10 [1]
Step 3: Set MR = MC: 90 − 4Q = 2Q + 10 → 80 = 6Q → Q* = 13.33 units [2]
Step 4 — SOC: π = 90Q − 2Q² − Q² − 10Q − 100 = −3Q² + 80Q − 100
dπ/dQ = −6Q + 80; d²π/dQ² = −6 < 0 ✓ maximum confirmed [1]
Here the second derivative is constant and negative, so profit is concave everywhere and the single stationary point must be the maximum.
Step 5: P = 90 − 2(13.33) = £63.33 [1]
π = −3(13.33)² + 80(13.33) − 100 = −533.1 + 1066.7 − 100 = £433.33 [1]
Question 3 — MC, AC and elasticity (7 marks)
(a) Prove using calculus that MC = AC at the minimum point of the AC curve. (b) A monopolist claims it maximises profit at an output where PED = −0.7. Explain why this cannot be correct.
(a) AC = TC/Q. Differentiate using the quotient rule: [1]
dAC/dQ = [Q·(dTC/dQ) − TC·1] / Q² [1]
Since dTC/dQ = MC and TC/Q = AC, this simplifies to (MC − AC)/Q [1]
At the minimum of AC, dAC/dQ = 0. Since Q > 0, the numerator must be zero, so MC = AC. [1]
dAC/dQ = [Q·(dTC/dQ) − TC·1] / Q² [1]
Since dTC/dQ = MC and TC/Q = AC, this simplifies to (MC − AC)/Q [1]
At the minimum of AC, dAC/dQ = 0. Since Q > 0, the numerator must be zero, so MC = AC. [1]
(b) Use MR = P(1 + 1/PED). [1]
At PED = −0.7: MR = P(1 + 1/(−0.7)) = P(1 − 1.429) = −0.429P, which is negative. [1]
Profit maximisation requires MR = MC. But marginal cost cannot be negative — producing more cannot reduce total cost. So MR = MC is impossible where MR < 0. [1]
The firm would raise profit by cutting output and raising price: revenue would rise (since demand is inelastic) and costs would fall. A profit-maximising monopolist therefore always operates where demand is elastic (PED < −1).
Question 4 — Application and evaluation (10 marks)
A rail operator’s marginal cost of carrying one additional passenger on an already-scheduled train is close to zero, while its average cost per passenger is £45. (a) Explain using calculus why MC and AC diverge so sharply. (b) Evaluate whether the operator should sell off-peak tickets at £10.
(a) [4 marks]
Write total cost as TC = F + V(Q), where F is fixed (track access, rolling stock lease, crew, fuel for the scheduled service) and V(Q) is the variable component. [1]
MC = dTC/dQ = dV/dQ. By the constant rule, F differentiates to zero and disappears entirely. [1]
For a scheduled train, an additional passenger consumes almost no extra resources — a negligible amount of fuel, no extra crew — so dV/dQ ≈ 0, giving MC ≈ 0. [1]
AC = TC/Q = F/Q + V(Q)/Q. The F/Q term does not vanish, and with F large it dominates, producing AC = £45. The divergence is entirely the fixed cost, which appears in AC but is mathematically absent from MC. [1]
Write total cost as TC = F + V(Q), where F is fixed (track access, rolling stock lease, crew, fuel for the scheduled service) and V(Q) is the variable component. [1]
MC = dTC/dQ = dV/dQ. By the constant rule, F differentiates to zero and disappears entirely. [1]
For a scheduled train, an additional passenger consumes almost no extra resources — a negligible amount of fuel, no extra crew — so dV/dQ ≈ 0, giving MC ≈ 0. [1]
AC = TC/Q = F/Q + V(Q)/Q. The F/Q term does not vanish, and with F large it dominates, producing AC = £45. The divergence is entirely the fixed cost, which appears in AC but is mathematically absent from MC. [1]
(b) Evaluation — credit for developed points both ways: [6 marks]
Arguments for selling at £10:
• Since MC ≈ 0, any fare above marginal cost adds to profit. £10 > £0, so each off-peak sale increases contribution toward the fixed costs. Refusing the sale leaves money on the table and the seat empty.
• The comparison against AC of £45 is the classic sunk cost fallacy. Once the train is scheduled, the fixed cost is unavoidable and therefore irrelevant to the marginal decision.
• Off-peak capacity is perishable — an unsold seat on a departed train has zero salvage value. This is the same logic as airline yield management.
• Price discrimination is the point: charging £10 off-peak and £45 peak captures consumer surplus from price-sensitive travellers without cannibalising full-fare demand, provided the segments can be separated.
Arguments against:
• MC ≈ 0 only holds within capacity. If off-peak demand grows enough to require additional carriages or services, MC jumps discontinuously and the £10 fare no longer covers it.
• Revenue dilution: if peak travellers can shift to off-peak, the operator loses £35 per switching passenger. The elasticity and separability of the two segments determine whether the policy gains or loses overall.
• In the long run all costs are variable. A firm that consistently prices at short-run MC never recovers F and eventually exits — the structural problem that makes rail and airline economics so fragile.
• Regulatory and equity constraints may limit differential pricing.
Conclusion: selling at £10 is correct as a short-run marginal decision on spare capacity, and the £45 average cost is the wrong comparator. But the policy is only sustainable if peak fares and other revenue cover F, and if the two markets can genuinely be separated. The calculus tells you the seat is worth selling; it does not tell you the business is viable.
Summary
Marginal means derivative. Marginal cost is dTC/dQ, marginal revenue is dTR/dQ, and profit is maximised where their difference is zero — which is to say where MR = MC. That famous rule is not an assumption bolted onto the theory; it is what setting a derivative to zero looks like when the function is profit.
Two things separate a good answer from a full-marks one. First, the second-order condition: MR = MC finds a flat spot, and only d²π/dQ² tells you whether you are standing on a summit or in a valley. Second, the constant rule, which quietly deletes fixed costs from every marginal calculation — and in doing so explains airline pricing, sunk costs, and why average cost is the wrong number for almost every decision a firm actually faces.
And the evidence says firms do not consciously do any of this. Hall and Hitch found it in 1939, Blinder confirmed it in 1998, and Machlup’s answer still holds: the cyclist does not solve the equations either.
References
- Blinder, A.S., Canetti, E., Lebow, D. and Rudd, J. (1998) Asking About Prices: A New Approach to Understanding Price Stickiness. New York: Russell Sage Foundation.
- Chiang, A.C. and Wainwright, K. (2005) Fundamental Methods of Mathematical Economics. 4th edn. New York: McGraw-Hill.
- Hall, R.L. and Hitch, C.J. (1939) ‘Price theory and business behaviour’, Oxford Economic Papers, 2, pp. 12–45.
- Machlup, F. (1946) ‘Marginal analysis and empirical research’, American Economic Review, 36(4), pp. 519–554.
- Robinson, J. (1933) The Economics of Imperfect Competition. London: Macmillan. (Origin of the Amoroso–Robinson relation MR = P(1 + 1/ε).)
- Varian, H.R. (2014) Intermediate Microeconomics: A Modern Approach. 9th edn. New York: W.W. Norton.