A topic in the Open Knowledge Graph — a free, open map of 15,290 topics and the order to learn them in.
Duality in producer theory establishes equivalence between profit maximization and sequential problems of cost minimization (for given output) plus revenue maximization. The cost function C(w,y) and profit function π(p,w) contain all technological information. Shephard's lemma derives factor demands from the cost function through differentiation.
In producer theory, there are two natural ways to think about a firm's problem. The primal approach starts from the production function — the technology that maps inputs (labor, capital) into output — and asks: given input prices and output price, what combination of inputs maximizes profit? The dual approach flips the question: given that you want to produce a specific amount of output, what is the cheapest way to do it? Duality theory proves these two perspectives contain exactly the same information. Everything you can learn about a firm's technology from its production function, you can also extract from its cost function, and vice versa.
The cost minimization problem is solved using the tools of constrained optimization you already know. You minimize total input cost w₁x₁ + w₂x₂ subject to the constraint that f(x₁, x₂) ≥ y, where w is the vector of input prices, x is inputs, and y is the target output level. Setting up the Lagrangian and applying the first-order conditions yields the conditional factor demands x*(w, y) — the cost-minimizing input quantities as functions of input prices and output. Substituting these back gives the cost function C(w, y) = w · x*(w, y), which tells you the minimum cost of producing any output level at any set of input prices.
The remarkable result is Shephard's lemma: the partial derivative of the cost function with respect to an input price equals the conditional factor demand for that input. That is, ∂C(w, y)/∂wᵢ = xᵢ*(w, y). This means you do not need to re-solve the optimization problem to find factor demands — you can simply differentiate the cost function. This is extraordinarily powerful in applied work because cost functions are often easier to estimate empirically than production functions. If you can estimate how costs respond to input price changes, you automatically know the firm's input demands.
The cost function also has elegant mathematical properties that mirror the structure of the underlying technology. It is concave and homogeneous of degree one in input prices (doubling all input prices exactly doubles costs), non-decreasing in output, and non-decreasing in input prices. These properties are not assumptions — they are consequences of cost minimization. The profit function π(p, w) works analogously for the full profit-maximization problem: it is convex in prices, and Hotelling's lemma says its derivative with respect to output price gives supply, while its derivatives with respect to input prices give (negative) unconditional factor demands. Together, these duality results mean that a researcher who observes only market data on prices, costs, and quantities can recover the firm's entire technological structure without ever directly observing the production function.