Kuhn-Tucker Conditions for Consumer Choice

Why Boundary Choices Need More Than Tangency

The interior tangency rule compares marginal utility per dollar across goods when the consumer purchases positive quantities of both. But a consumer may choose zero units of a good. At that boundary, the usual equality need not hold: the relevant test is whether a small affordable move into the zero-consumption good would improve utility. Kuhn-Tucker conditions formalize that test.

The conditions combine the budget constraint, nonnegative consumption, and complementary slackness. In particular, they distinguish a good whose quantity is positive from one whose quantity is zero. This tutorial develops the conditions for two goods, derives the boundary inequality, and applies the method to candidate consumer choices.

Model and assumptions: A consumer chooses quantities \(x\) and \(y\) to maximize a utility function \(U(x,y)\), subject to \(p_xx+p_yy\leq m\), \(x\geq0\), and \(y\geq0\). Prices \(p_x,p_y\) and income \(m\) are positive. Assume \(U\) is differentiable at the candidate bundle. The Kuhn-Tucker conditions are necessary at a local optimum when a constraint qualification holds; with positive income, a strictly feasible bundle with positive quantities exists. The conditions identify candidates and do not, by themselves, establish that every candidate is a global maximum.

Writing the Kuhn-Tucker Conditions

Write each constraint in “greater than or equal to zero” form: \(m-p_xx-p_yy\geq0\) for the budget, and \(x\geq0\), \(y\geq0\) for nonnegative consumption. Assign nonnegative multipliers \(\lambda\), \(\alpha\), and \(\beta\) to these three constraints, respectively. The Lagrangian is

\mathcal{L} =U(x,y)+\lambda(m-p_xx-p_yy)+\alpha x+\beta y.

The first-order stationarity conditions set the derivatives of the Lagrangian with respect to each choice variable equal to zero. Alongside them are feasibility, nonnegative multipliers, and complementary slackness:

\begin{aligned} U_x(x,y)-\lambda p_x+\alpha&=0,\\ U_y(x,y)-\lambda p_y+\beta&=0,\\ m-p_xx-p_yy&\geq0,\qquad x\geq0,\qquad y\geq0,\\ \lambda&\geq0,\qquad \alpha\geq0,\qquad \beta\geq0,\\ \lambda(m-p_xx-p_yy)&=0,\qquad \alpha x=0,\qquad \beta y=0. \end{aligned}

Here \(U_x\) and \(U_y\) are the marginal utilities of the two goods. Complementary slackness says that a constraint can have a positive multiplier only when it is binding. For example, if \(x>0\), then \(\alpha x=0\) forces \(\alpha=0\). If \(x=0\), the multiplier \(\alpha\) may be positive, recording that the nonnegativity constraint is active.

Interpretation: If \(x>0\), stationarity gives \(U_x=\lambda p_x\). If \(x=0\), stationarity and \(\alpha\geq0\) give \(U_x\leq\lambda p_x\). Likewise, a positive \(y\) satisfies \(U_y=\lambda p_y\), while \(y=0\) requires \(U_y\leq\lambda p_y\). Thus, a consumed good satisfies equality in marginal utility per dollar; a zero-consumption good satisfies an inequality.

When the budget binds, \(\lambda\) is the common marginal utility per dollar for goods consumed at positive quantities. If only one good is consumed, its marginal utility per dollar sets that level, while the zero-quantity good must not offer a greater marginal utility per dollar at the boundary. These statements concern derivatives at the candidate bundle.

Why the Boundary Condition Is an Inequality

Suppose \(x=0\), \(y>0\), and the budget binds. A small increase in \(x\) must be financed by reducing \(y\). For a small change \(\varepsilon>0\), the budget-preserving change is \(dx=\varepsilon\) and \(dy=-(p_x/p_y)\varepsilon\). The first-order change in utility is therefore

dU =U_x\varepsilon-U_y\frac{p_x}{p_y}\varepsilon =\varepsilon\left(U_x-\frac{p_x}{p_y}U_y\right).

At a local optimum, this feasible move cannot raise utility to first order, so \(U_x-(p_x/p_y)U_y\leq0\). Since \(y>0\), its stationarity condition gives \(U_y=\lambda p_y\). Substitution yields \(U_x\leq\lambda p_x\), the Kuhn-Tucker inequality for the zero-quantity good. Equality is possible at a threshold where the consumer is just indifferent, to first order, about introducing that good.

Identify the active quantities.
For each good, decide whether the proposed choice is positive or zero. A positive quantity forces its nonnegativity multiplier to zero.

Apply stationarity.
Use marginal-utility equalities for positive quantities and marginal-utility inequalities for zero quantities.

Check feasibility and complementary slackness.
Verify the budget, the signs of all multipliers, and that each multiplier times its constraint slack is zero.

Worked Applications

Worked Example: Testing a Zero-Consumption Candidate

A consumer has \(U(x,y)=0.5\ln(1+x)+4\ln(1+y)\), prices \(p_x=2\) and \(p_y=1\), and income \(m=8\). Test the proposed boundary bundle \((x,y)=(0,8)\) against the Kuhn-Tucker conditions.

Method. Since the proposed bundle spends \(2(0)+1(8)=8\), the budget binds. Since \(y>0\), its nonnegativity multiplier is \(\beta=0\). The marginal utility of \(y\) is \(U_y=4/(1+y)\), so at \(y=8\), \(U_y=4/9\). Stationarity for \(y\) gives \(\lambda=U_y/p_y=(4/9)/1=4/9\).

At the boundary \(x=0\), the marginal utility is \(U_x=0.5/(1+x)=0.5\). The required boundary inequality is \(U_x\leq\lambda p_x\). Its right-hand side is \((4/9)(2)=8/9\), and \(0.5\leq8/9\), so the inequality holds. The multiplier for the nonnegativity constraint on \(x\) is \(\alpha=\lambda p_x-U_x=8/9-1/2=16/18-9/18=7/18\geq0\).

The budget multiplier is positive, the budget slack is zero, and \(\alpha x=(7/18)(0)=0\), while \(\beta y=0(8)=0\). Thus the proposed boundary bundle satisfies the Kuhn-Tucker conditions. This test does not replace any separate check that the candidate is a global maximum. The common mistake is to demand \(U_x/p_x=U_y/p_y\) even though \(x=0\); the correct test is the inequality \(U_x/p_x\leq\lambda\).

Worked Example: Solving an Interior Candidate

Let \(U(x,y)=\sqrt{x}+\sqrt{y}\), with prices \(p_x=1\), \(p_y=2\), and income \(m=6\). Find the interior Kuhn-Tucker candidate.

For an interior candidate, \(x>0\) and \(y>0\), so \(\alpha=\beta=0\). The stationarity equations are \(1/(2\sqrt{x})=\lambda\) and \(1/(2\sqrt{y})=2\lambda\). Equating the two expressions for \(\lambda\) gives \(1/(2\sqrt{x})=1/(4\sqrt{y})\), hence \(\sqrt{x}=2\sqrt{y}\) and \(x=4y\).

The interior budget binds, so \(x+2y=6\). Substitute \(x=4y\): \(4y+2y=6\), \(6y=6\), and \(y=1\). Therefore \(x=4\). The multiplier is \(\lambda=1/(2\sqrt{4})=1/4\). The other stationarity equation gives \(1/(2\sqrt{1})=1/2=2(1/4)\), as required.

The candidate is feasible because \(4+2(1)=6\), and both quantities are positive, so \(\alpha=\beta=0\) satisfies complementary slackness. This method uses equal marginal utility per dollar only after confirming both quantities are positive. Applying a boundary inequality to this interior candidate would be the wrong case.

Worked Example: Finding the Income Threshold for a Boundary

Consider \(U(x,y)=\ln(1+x)+2\ln(1+y)\), with prices \(p_x=3\), \(p_y=1\), and income \(m\). Test the candidate \(x=0,\ y=m\), and find the condition on income under which it satisfies the boundary Kuhn-Tucker inequality.

The candidate spends \(3(0)+m=m\), so the budget binds. Since \(y=m>0\), stationarity for \(y\) gives \(\lambda=U_y/p_y=[2/(1+m)]/1=2/(1+m)\). At \(x=0\), \(U_x=1/(1+0)=1\). For zero consumption of \(x\), the condition is \(U_x\leq\lambda p_x\), or \(1\leq[2/(1+m)]3=6/(1+m)\).

Because \(1+m>0\), multiply both sides by \(1+m\) to obtain \(1+m\leq6\), hence \(m\leq5\). For \(m<5\), the inequality is strict; at \(m=5\), it holds with equality. For \(m>5\), the proposed boundary candidate fails this necessary condition. A common error is to reverse the inequality and conclude that a zero-consumption good must have higher marginal utility per dollar; the feasible-move derivation shows why its marginal utility per dollar must instead be no greater.

Applying the Conditions Carefully

Kuhn-Tucker conditions are most useful when treated as a matched system, rather than as a single formula. A proposed bundle can fail because it violates the budget, because a multiplier has the wrong sign, because stationarity fails, or because complementary slackness is inconsistent with whether a constraint binds. A check of only the first-order equalities is incomplete.

Status of the good	Quantity condition	Marginal-utility condition
Positive consumption of \(x\)	\(x>0\), so \(\alpha=0\)	\(U_x=\lambda p_x\)
Zero consumption of \(x\)	\(x=0\), with \(\alpha\geq0\)	\(U_x\leq\lambda p_x\)
Positive consumption of \(y\)	\(y>0\), so \(\beta=0\)	\(U_y=\lambda p_y\)
Zero consumption of \(y\)	\(y=0\), with \(\beta\geq0\)	\(U_y\leq\lambda p_y\)

The same logic applies if \(y=0\) and \(x>0\): the positive \(x\) good determines \(\lambda=U_x/p_x\), and the boundary test is \(U_y\leq\lambda p_y\). If both goods are positive, both marginal utilities per dollar equal \(\lambda\). At a boundary, do not impose equality for the zero-quantity good unless the inequality happens to hold with equality.

The Kuhn-Tucker framework is aligned with the SOA VEE Economics (Validation by Educational Experience) syllabus because consumer choice must allow for nonnegative quantities, not merely interior bundles. It makes explicit the economic meaning of a corner: at the chosen bundle, shifting expenditure toward a currently unconsumed good does not provide a first-order utility gain. A candidate that violates the corresponding inequality cannot be an optimum, even if it satisfies the budget constraint.

Key takeaway: For each good, positive consumption implies \(U_i=\lambda p_i\); zero consumption implies \(U_i\leq\lambda p_i\). Combine these conditions with feasibility, nonnegative multipliers, and complementary slackness before treating a bundle as a Kuhn-Tucker candidate.

Check Your Understanding

For each problem, apply the relevant Kuhn-Tucker conditions and check the arithmetic.

A consumer has \(U(x,y)=3\ln(1+x)+\ln(1+y)\), prices \(p_x=2,\ p_y=1\), and income \(m=4\). Test the proposed bundle \((0,4)\): calculate \(\lambda\) from the positive good and decide whether the boundary inequality for \(x\) holds. (Answer: \(\lambda=1/5\); \(U_x=3>\lambda p_x=2/5\), so the candidate fails.)
For \(U(x,y)=\ln(1+x)+3\ln(1+y)\), \(p_x=2,\ p_y=1,\ m=5\), test \((0,5)\), including the nonnegativity multiplier for \(x\). (Answer: \(\lambda=1/2\), \(\alpha=0\), and the boundary inequality holds with equality.)
For \(U(x,y)=\sqrt{x}+\sqrt{y}\), \(p_x=1,\ p_y=2,\ m=6\), solve the interior stationarity and budget equations. (Answer: \(x=4,\ y=1,\ \lambda=1/4\).)
For \(U(x,y)=\ln(1+x)+2\ln(1+y)\), \(p_x=3,\ p_y=1\), and income \(m>0\), test the candidate \((0,m)\) and state the range of income for which its boundary inequality holds. (Answer: it holds when \(m\leq5\), with equality at \(m=5\).)
In the same utility model as question 4, take \(m=8\). Solve the interior conditions and report the bundle and \(\lambda\). (Answer: \(x=1/3,\ y=7,\ \lambda=1/4\); the budget cost is \(3(1/3)+7=8\).)

What You'll Learn