Leontief Input–Output Model

Leontief input–output model (also termed input–output analysis or inter-industry analysis) is a quantitative multisectoral framework in business and economics theory and applied general equilibrium theory that models how the output of each industry in an economy serves both as an intermediate input into other industries and as a final good for household consumption, investment, government purchases, and exports. Formulated by Russian-American economist Wassily Leontief in 1936 and expanded in his 1941 monograph The Structure of American Economy, 1919–1929—work for which he received the 1973 Nobel Memorial Prize in Economic Sciences—the model represents the circular flow of production as a system of linear equations governed by fixed-proportion Leontief production functions. By inverting the matrix of net inter-industry coefficients via the Leontief inverse, economists and national statistical agencies can trace how a unit change in final demand for any single product propagates upstream through every tier of the supply chain.

The Input–Output Transactions Table

Consider a national or global economy partitioned into n distinct production sectors, indexed by i, j ∈ {1, 2, …, n}. Over a given accounting period (typically one calendar year), let:

• x_i ≥ 0 denote the total gross output produced by sector i;
• z_ij ≥ 0 denote the intermediate flow of goods or services produced by upstream sector i and consumed as an intermediate input by downstream sector j; and
• d_i ≥ 0 denote the final demand for sector i's output by autonomous end-users (households C_i, gross capital formation I_i, government spending G_i, and net exports NX_i).

Across each row i of the input–output table, the material balance identity requires that total gross output equal the sum of intermediate deliveries to all n industries plus final demand:

Down each column j, total gross outlay x_j equals the sum of intermediate purchases from all upstream industries ∑_{i = 1}ⁿ z_ij plus primary value added v_j (compensation of employees, return to capital, and net production taxes), linking the input–output table directly to gross domestic product where ∑_{i = 1}ⁿ d_i = ∑_{j = 1}ⁿ v_j = GDP.

Fixed Technical Coefficients and the Balance Equation

Leontief assumed a fixed-proportions production technology (zero elasticity of substitution between intermediate inputs) in which producing one unit of gross output in sector j requires a constant quantity a_ij of input from sector i, known as the technical coefficient (or direct input coefficient):

Assembling the coefficients into the n × n technical coefficient matrix A = [a_ij], the gross output vector x = (x₁, …, x_n)^⊤, and the final demand vector d = (d₁, …, d_n)^⊤ yields the matrix balance equation:

where I is the n × n identity matrix and (I − A) is the Leontief matrix.

Neumann Power-Series Expansion

Whenever the matrix (I − A) is invertible, the unique gross output vector x required to satisfy any exogenous final demand vector d is given by:

where L = [l_ij] ≡ (I − A)⁻¹ is the Leontief inverse matrix (or total requirements matrix). Provided the spectral radius (largest eigenvalue magnitude) of the non-negative matrix A is less than unity—ρ(A) < 1—the Leontief inverse admits a convergent Neumann series expansion:

Each power of A captures a distinct supply-chain tier of production:

1. Zeroth Round (Id): The immediate final demand vector d itself.
2. First Round (Ad): The direct intermediate inputs required by final-goods producers to assemble d (for instance, the steel, glass, and microchips directly purchased by an automobile plant).
3. Second and Higher Rounds (A²d + A³d + ⋯): The indirect intermediate inputs required upstream to manufacture the direct inputs (the iron ore, coal, electricity, and chemical feedstocks needed to produce the steel and microchips, ad infinitum).

The Hawkins–Simon Viability Theorem

Because physical production x cannot be negative, an input–output system A is economically viable (productive) if and only if (I − A)⁻¹ ≥ 0 element-wise, ensuring that x = (I − A)⁻¹d ≥ 0 for every non-negative final demand vector d ≥ 0. David Hawkins and Herbert A. Simon (1949) proved that for A ≥ 0, the following conditions are equivalent:

1. Hawkins–Simon Condition: All n leading principal minors of (I − A) are positive (det(I_k − A_k) > 0 for every k = 1, 2, …, n).
2. Perron–Frobenius Spectral Condition: The dominant Perron–Frobenius eigenvalue (spectral radius) satisfies ρ(A) < 1.
3. Braeuer–Solow Column-Sum Sufficiency: If every column sum of A expressed in common monetary units is less than unity (∑_{i = 1}ⁿ a_ij < 1 for all j, meaning every sector generates positive value added v_j > 0), then ρ(A) < 1 and (I − A)⁻¹ ≥ I ≥ 0.

In addition to physical quantity flows, the Leontief model possesses a dual price system. If v̄_j = v_j / x_j denotes the primary value-added cost (wages plus capital rental plus energy/carbon taxes) per unit of output in sector j, zero economic profit requires that each sector's unit price p_j equal its intermediate input costs plus its unit value added:

Notice that in the Leontief model without joint production, the Non-Substitution Theorem (developed by Paul Samuelson, Nicholas Georgescu-Roegen, and Tjalling Koopmans in 1951) holds: even if each industry has a smooth continuum of alternative production techniques, when there is only one scarce primary factor (labor) and constant returns to scale, the cost-minimizing technical matrix A and equilibrium relative prices p are completely independent of the composition of final demand d.

Beginning in the late 1960s and 1970s (Leontief, 1970), the Leontief framework became the standard engine of Environmentally Extended Input–Output (EEIO) life-cycle assessment and global carbon-footprint accounting (implemented in multi-regional databases such as EXIOBASE, WIOD, and GTAP). Let f = (f₁, …, f_n)^⊤ denote the vector of direct environmental intensities—such as metric tons of CO₂ emitted or cubic meters of freshwater withdrawn per million dollars of gross output in each sector. Then the total embodied environmental footprint E associated with a final consumption bundle d is:

The row vector m^⊤ = f^⊤(I − A)⁻¹ gives the cradle-to-gate emission multipliers, capturing Scope 1 (direct on-site), Scope 2 (purchased electricity and heat, f^⊤A), and full Scope 3 (all upstream tier-2 through tier-∞ supply-chain) emissions. Similarly, in modern macroeconomics, Daron Acemoglu, Vasco Carvalho, Asuman Ozdaglar, and Alireza Tahbaz-Salehi (2012) embedded Cobb–Douglas input–output networks into stochastic general equilibrium models, proving that when the out-degree distribution of the Leontief inverse is heavy-tailed (due to "hub" input suppliers), idiosyncratic microeconomic productivity shocks do not wash out by the law of large numbers and instead generate aggregate business-cycle fluctuations.

What was the Leontief Paradox in international trade?

In 1953, Wassily Leontief applied his 1947 U.S. input–output table to test the Heckscher–Ohlin theorem (see comparative advantage and Stolper–Samuelson theorem). Because the United States was the most capital-abundant nation in the world after World War II, theory predicted U.S. exports would be more capital-intensive than U.S. import-competing goods. Using the Leontief inverse (I − A)⁻¹ to calculate total (direct plus indirect) capital and labor embodied in a million dollars of U.S. exports versus imports, Leontief discovered that U.S. exports were roughly 30% less capital-intensive (and more labor-intensive) than U.S. imports—a puzzle subsequently resolved by accounting for U.S. human capital (skilled labor), R&D intensity, and natural-resource abundance.

Why must the spectral radius of the technical matrix A be less than one?

If the dominant eigenvalue ρ(A) ≥ 1, producing one unit of output across the economy requires at least one unit of that same composite bundle as intermediate input simply to replace what was used up in production, leaving zero net surplus (x − Ax ≤ 0) to satisfy positive final consumption d > 0.

1. Leontief, W. W. (1936). "Quantitative Input and Output Relations in the Economic Systems of the United States." The Review of Economics and Statistics, 18(3), 105–125. https://doi.org/10.2307/1927837
2. Hawkins, D., & Simon, H. A. (1949). "Note: Some Conditions of Macroeconomic Stability." Econometrica, 17(3/4), 245–248. https://doi.org/10.2307/1905526
3. Leontief, W. (1970). "Environmental Repercussions and the Economic Structure: An Input–Output Approach." The Review of Economics and Statistics, 52(3), 262–271. https://doi.org/10.2307/1926294
4. Acemoglu, D., Carvalho, V. M., Ozdaglar, A., & Tahbaz-Salehi, A. (2012). "The Network Origins of Aggregate Fluctuations." Econometrica, 80(5), 1977–2016. https://doi.org/10.3982/ECTA9623
5. Miller, R. E., & Blair, P. D. (2009). Input–Output Analysis: Foundations and Extensions (2nd ed.). Cambridge University Press. https://doi.org/10.1017/CBO9780511626982