Just as consumer theory built the demand curve from first principles, producer theory builds the cost side of every profit-maximization problem in the rest of this module. Start from a total cost function, TC(Q), splitting into total fixed cost (TFC) (costs that do not vary with output at all, sunk in the short run) and total variable cost (TVC(Q)), which rises with output because of diminishing marginal returns to a fixed input (a fixed factory, more and more workers): each additional worker eventually adds less extra output than the last, so TVC rises at an increasing rate past some point. Four per-unit measures follow directly: AFC = TFC/Q (always falling, since a fixed cost is spread over more units), AVC = TVC/Q, ATC = TFC/Q + AVC (the sum of the first two), and marginal cost, MC = d(TC)/dQ, the cost of producing one more unit, the single most important number in this entire module, because every profit-maximizing decision from here to the end of oligopoly is a comparison against MC.
Advertisement
How do marginal cost and average cost curves relate?
A cubic total-cost function of the classic textbook shape, TC(Q) = TFC + aQ − bQ² + cQ³, produces the U-shaped AVC and MC curves every principles course draws: AVC falls at first (spreading a rising-but-still-modest variable cost over more units, with early efficiency gains) and then rises (diminishing returns take over), and MC (steeper than AVC, because it reflects the cost of the next unit alone, not an average) crosses AVC from below exactly at AVC's minimum. This is not a coincidence of the specific numbers; it is a calculus identity: AVC's minimum occurs where d(AVC)/dQ = 0, and at exactly that quantity, MC = AVC, because the average of a series can only be falling when the newest addition is below it, and rising when the newest addition is above it, so the average is momentarily flat (its minimum) exactly where the marginal value crosses it. The identical logic gives MC = ATC at ATC's minimum, and because AFC is always positive and always falling, ATC's minimum sits at a larger quantity than AVC's minimum (AFC's continued decline keeps pulling ATC down for a while even after AVC has turned up). Worked Example 4 proves both crossings numerically for a specific firm; Practice P7 asks you to prove them again for a fresh one.
What is the shutdown rule?
These curves hand you the competitive firm's short-run operating rule directly, and it is the single most commonly botched distinction in an introductory course: the shutdown rule (a short-run decision) says produce nothing this period if price falls below AVC's minimum, below that floor, every unit sold loses money even before touching fixed costs, so producing at all adds to the loss beyond the fixed cost you are stuck with regardless. Between AVC's minimum and ATC's minimum, the firm should keep operating in the short run (revenue covers all variable cost and part of the fixed cost, so the loss is smaller than shutting down entirely) but is not covering its full economic cost, if the price stays there, the correct long-run decision is to exit the industry once fixed commitments can be unwound, because the capital tied up here is earning less than its opportunity cost elsewhere. Above ATC's minimum, the firm earns a genuine economic profit and should not merely continue but is exactly the kind of entrant a rational competitor is watching, a fact perfect competition's long-run competitive equilibrium turns into a precise prediction. Note, too, how this cost machinery underwrites one of Phase 4's own moat sources without re-deriving it: M4.04 §6 names efficient scale as a moat source precisely where a market is small relative to the minimum efficient scale (the quantity at which a firm's long-run average cost curve bottoms out) so that a second full-scale entrant would crash the price for everyone; this module gives you the actual cost-curve mathematics that claim rests on, and perfect competition extends the same minimization to the long-run, multi-firm case.
{
"skill": "Scaffolding and cross-checking a payoff-matrix or reaction-function solver for an oligopoly problem",
"use": "Ask an AI assistant to draft Python that solves a stated Cournot or Bertrand problem (reaction functions, Nash equilibrium quantities/prices) or finds the equilibrium/equilibria of a small payoff matrix -- then verify every output against your own hand-solved reaction-function algebra before trusting it.",
"tools": ["Chat assistants -- Claude, ChatGPT, Gemini (drafting and explaining)", "A Python/Jupyter or Pyodide notebook with scipy.optimize or a simple brute-force best-response check"],
"prompt": "Write a Python function cournot_n(n, a, b, c) that returns each firm's equilibrium quantity, total quantity, market price, and per-firm profit for n symmetric Cournot firms facing inverse demand P = a - b*Q and constant marginal cost c, using the closed-form qi = (a-c)/((n+1)*b). Verify it against n=1 (should match the direct monopoly first-order condition) and confirm price falls monotonically toward c as n grows.",
"verify": "Before trusting the function's output for any new (a,b,c,n), re-derive the reaction function qi = (a-c-qj)/2 by hand for n=2 and confirm the code's duopoly answer matches your own algebra to the decimal, and separately confirm the n=1 case exactly reproduces the standalone monopoly MR=MC solution -- a self-consistency check that catches most sign and off-by-one errors in an AI-drafted formula (Guardrail 1-5, AI0.01).",
"diy": "You must already be able to derive a single firm's reaction function from its own first-order condition by hand -- that is exactly what section 7 below builds, and it is the one step no tool should be allowed to hand you fully formed.",
"market": "IN"
}