ECON10071 Chap.7 Equality-Constrained Optimisation
Equality-Constrained Optimisation
Equality-Constrained Optimisation restricts choices to a curve or surface. At a regular optimum, the objective and constraint gradients are parallel, and the Lagrangian expresses that tangency with a multiplier. The first-order system includes one equation for every choice and one feasibility equation for every constraint.
These conditions generate candidates; concavity, substitution, candidate comparison or a bound must still establish existence and optimality.
What this chapter covers
- 01
Constraint tangency
- 02
Lagrangian construction
- 03
Multiplier equations
- 04
Feasibility conditions
- 05
Multiple equality constraints
- 06
Existence and classification
Worked example · free
Allocate a fixed total
- 3The Lagrange conditions give 1/x=λ and 2/y=λ, so y=2x.
- 4Feasibility gives x=4 and y=8; strict concavity proves a unique global maximum.
Key terms
- Lagrangian
- An objective combined with equality constraints through multipliers.
- Constraint gradient
- A normal direction to the equality-constraint surface.
- Multiplier
- A variable enforcing a constraint and often measuring a local value effect.
Equality-Constrained Optimisation FAQ
Why is the unconstrained gradient not zero at a constrained optimum?
The objective may still rise in an infeasible direction. At a regular constrained optimum, its gradient is instead parallel to the constraint gradient, so no feasible tangent movement improves the objective.
Can I divide by a choice variable in the first-order conditions?
Only after treating the zero case separately. Dividing immediately can discard boundary or branch candidates. Split the cases, solve each and retain only feasible candidates.
What proves a Lagrange candidate is global?
Strict concavity on a convex feasible set can prove uniqueness, while a complete comparison on a compact set or substitution into a one-variable bounded objective can also establish a global maximum.
Exam move
Write every constraint in zero form and choose a consistent multiplier sign. Differentiate with respect to all choices and multipliers, branch before dividing and reject infeasible candidates. Practise both one- and two-constraint problems. Always add the classification argument, because the multiplier equations alone are only necessary candidate conditions.
Start each problem by separating the objective from the constraint system. Rewrite every equality as a zero expression and choose one Lagrangian sign convention. A different consistent convention can reverse the multiplier sign without changing the optimal choices, so the danger lies in switching conventions halfway through. Differentiate with respect to every choice variable and every multiplier.
Count the resulting equations before solving: one stationarity equation per choice and one feasibility equation per equality. This count catches omitted constraints and unused multipliers. Solve with explicit branching. First-order equations often contain a choice variable multiplied by a bracket. Record the zero-variable branch before dividing by that variable, because a boundary or corner can otherwise disappear.
Substitute candidate relations into the original equality rather than only into a rearranged version. Reject complex or inadmissible solutions using the stated domain. For several constraints, keep a separate multiplier for each and verify that their constraint gradients satisfy the required regularity condition at the candidate.
If a constraint gradient vanishes, use substitution or a direct parameterisation to inspect the feasible directions. Pair every multiplier calculation with an alternative method on selected exercises. Substitute a linear constraint into the objective to reduce the dimension, or parameterise a curve and differentiate the reduced function. Agreement confirms the choice coordinates and often supplies the missing classification.
Complete squares or inspect concavity to decide whether the candidate is a local or global optimum. On an unbounded feasible curve, examine the objective as the parameter grows in both directions; a stationary maximum can coexist with no minimum. Use a final checklist: feasible point, all stationarity equations, multiplier value, objective value and proof of status.
Explain the tangency by saying that objective and constraint gradients are parallel at a regular constrained optimum. Do not treat the multiplier equations as a proof of existence. For logarithmic objectives, preserve positivity throughout; for quadratic objectives, compare the reduced form with the curvature conclusion.
In timed revision, alternate maximum, minimum, compact and unbounded examples so that the classification step cannot be replaced by a memorised closing sentence. Add a sensitivity question to selected problems. Change the right side of the equality slightly, predict the direction in which the optimal choices should move and resolve the system.
Compare the change in the optimised value with the multiplier interpretation under the chosen convention. Treat this as a local comparison and check whether the regularity and curvature conditions remain valid. The exercise connects the geometric tangency, the algebraic multiplier and the economic meaning of relaxing a resource requirement.
When a reduced-form solution is available, differentiate the optimised value with respect to the constraint parameter and compare it with the multiplier under the chosen sign convention. Record any sign reversal caused by the way the constraint entered the Lagrangian.
Working through Equality-Constrained Optimisation in ECON10071? Sia is AskSia’s AI Mathematics tutor — ask any ECON10071 Equality-Constrained Optimisation question and get a clear, step-by-step explanation grounded in how ECON10071 is taught and assessed. Read this chapter free, then take your hardest questions to Sia.