WealthPast logo WealthPast
People
Blog
Calculators
Proverbs & SayingsOn This DayHolidaysAboutContact

Leonid Kantorovich’s Linear Programming: Scarcity, Shadow Prices and the Cost of Constraints

Leonid Kantorovich’s linear programming explained through scarce inputs, feasible production, shadow prices and why an optimized plan depends on its objective and data.

Leonid Kantorovich’s linear programming turned a basic financial problem into a formal calculation: how should limited labor, materials and equipment be allocated among competing uses? The method can identify a best feasible plan under stated assumptions. It cannot decide society’s values, repair bad data or guarantee that a mathematical optimum is fair.

Written by Ibraham

Leonid Kantorovich in 1975 — historical photograph by Andrey Bogdanov
Leonid Kantorovich in 1975 — historical photograph by Andrey Bogdanov. Credit: Андрей Богданов (Andrei-bogdanoff yandex.ru). Wikimedia Commons source · CC BY 3.0. Existing source crop where applicable; no AI alteration.

A production problem became an economic method

NobelPrize.org records that Kantorovich was born in St. Petersburg on January 19, 1912 and became a professor at Leningrad University in 1934. The Royal Swedish Academy notes that his 1939 work addressed efficient resource use inside enterprises. He shared the 1975 economics prize with Tjalling Koopmans for contributions to the theory of optimum allocation of resources.

The work concerned the use of national and enterprise resources; it did not make those resources Kantorovich’s personal assets. No factory output, accounting price or modeled surplus should be relabeled as his wealth.

Optimization begins with a feasible set

Imagine a fictional workshop that makes tables and chairs. A table contributes $40 after directly identified variable costs and requires four labor hours plus three units of wood. A chair contributes $25 and requires two labor hours plus two units of wood. The workshop has 40 labor hours and 30 units of wood.

The manager cannot simply choose the product with the larger dollar contribution. Ten tables use all labor and 30 units of wood, contributing $400. Fifteen chairs use all wood but only 30 labor hours, contributing $375. A mixed plan may perform differently. Linear programming compares all feasible combinations rather than relying on one ratio.

Wassily Leontief’s input–output lesson, scheduled before this article, maps interindustry requirements. Kantorovich’s allocation problem asks how to choose among feasible activities when inputs are scarce.

A shadow price values one more unit inside the model

A binding constraint prevents the objective from improving. Its shadow price measures how much the optimized objective would change if that constraint were relaxed by one unit, within the range where the current solution remains valid. If one extra labor hour raises the model’s contribution by $7, then $7 is an internal marginal value—not automatically the market wage or a universal social value.

The distinction matters financially. A company may rationally pay up to the modeled marginal value for short-term capacity, but overtime fatigue, quality failures and contract rules may be missing. Shadow prices can reveal opportunity cost; they cannot make omitted consequences disappear. They must also be recalculated when demand, technology or available capacity changes, because yesterday’s binding constraint may be slack today.

The objective function carries a value judgment

Maximizing contribution margin, output, employment or reliability can produce different plans from the same physical constraints. A hospital that maximizes the number of treatments may choose differently from one minimizing expected life-years lost. A public planner may also impose equity, resilience or regional constraints that a narrow profit model excludes.

Jan Tinbergen’s targets-and-instruments lesson explains why a policy needs enough independent tools for its goals. Trygve Haavelmo’s identification lesson, also scheduled earlier, shows why estimated relationships need more than correlation before they become model inputs.

Optimization is a question generator

Before trusting a result, test whether coefficients change at larger scale, whether inputs can really be divided, and whether uncertain demand makes the plan fragile. Run alternative scenarios instead of reporting one exact answer as destiny. The sourced Leonid Kantorovich profile records his verified chronology and the limits of personal-wealth evidence.

Sources and Method

Checked November 12, 2026. The workshop, contribution and shadow-price figures are original teaching examples, not historical enterprise data or investment advice.

Share with