
Mathematics for Economics — Constrained Optimization and the Lagrange Multiplier
Mathematics for Economics — With Constraints, Derivatives Alone Are Not Enough
Without constraints, is the first-order condition. With a budget or resource constraint, one must move along its boundary. The Lagrangian is a technique that treats the constraint like a price. Nash equilibria and mixed strategies, where several people choose at once, belong to game theory in microeconomics. This chapter deals only with the extrema of a single decision-maker.
1. The multiplier is the value of one unit of the constraint to the objective
s.t. . .
At , . With a budget of 13 won, and , an increase of 3.125 that matches the approximation . This is , the marginal utility per won of consumer theory.
2. With inequalities and non-negativity, corners can be the solution
KKT adds complementary slackness for non-binding constraints to the equality Lagrangian. If a good’s marginal utility per won is lower than another’s, consumption of that good stops at zero. “Always an interior tangency” is a special solution that assumes positive consumption.
| Tool | Constraint | What the solution tells you |
|---|---|---|
| First-order conditions | None | Candidate stationary points |
| Lagrangian | Equality | Extrema on the boundary and λ |
| KKT | Inequality, non-negativity | Which constraints bind |
The second-order condition checks whether the Hessian is negative definite (for a maximum). Matching only the first order and reading a saddle point as maximum profit picks output the wrong way round. It is the same as the distinction between average and margin in chapter 1.
Once time is added, the same first-order conditions become differential equations. The stability of their steady state is covered in chapter 4.
Check your understanding
If the budget rises 50% from 12 won to 18 won, the linear approximation suggests that utility rises by . In fact and , an increase of 22.5, and the new multiplier is . With this utility function λ grows with income, so the linear approximation holds only for small changes. 50% is not small.
References
- Alpha Chiang and Kevin Wainwright, Fundamental Methods of Mathematical Economics, ch. 12
- Carl Simon and Lawrence Blume, Mathematics for Economists, ch. 18–19
- Harold Kuhn and Albert Tucker, “Nonlinear Programming,” Proceedings of the Second Berkeley Symposium on Mathematical Statistics and Probability (1951)
Oiyo
Editorial DeskThe OIYO editorial desk researches money, law, lifestyle, and self-understanding topics against primary sources and public statistics. Every piece carries source notes and is reviewed on a regular cycle for accuracy and usefulness.