Allen–Uzawa elasticity of substitution (AES)

Overview

The Allen–Uzawa elasticity of substitution (AES) is a way to quantify how easily one input can be substituted for another in a cost-minimizing production setting with three or more inputs. It generalizes the familiar two-input “Hicks elasticity of substitution.” Intuitively, a higher AES between two inputs means that when the price of one input changes, the firm can more readily adjust toward the other input without sharply increasing cost. AES matters for understanding how input price shocks (wages, energy prices, interest rates) reconfigure optimal input mixes, how quickly costs rise under new regulations (e.g., carbon prices), and how technology choices shape flexibility.

Origins and Credit

The measure originated with R. G. D. Allen (1938), who defined a “partial” elasticity of substitution for multi-input production functions. Hirofumi Uzawa (1962) later reformulated it using the cost function—clarifying invariance properties and placing it squarely in duality theory—leading to the widely used “Allen–Uzawa” label. In the special case of two inputs, AES reduces to the Hicks elasticity of substitution.

Core Idea and Mechanics

Consider a firm minimizing cost to produce a fixed output level. Let C(w, y) be the minimum cost of producing output y at input prices w. The AES between inputs i and j is defined (locally) as:

σ_ij^AU = [ C · (∂²C/∂w_i∂w_j) ] / [ (∂C/∂w_i) · (∂C/∂w_j) ].

By Shephard’s lemma, ∂C/∂w_i equals the firm’s cost-minimizing demand for input i, and ∂²C/∂w_i∂w_j is the compensated cross-price effect (how the demand for i changes when the price of j changes, holding output fixed and always re-optimizing). The AES is thus a normalized curvature measure of the cost function:

• Sign: If σ_ij^AU > 0, inputs i and j are (net) substitutes—a higher price of j increases the demand for i. If σ_ij^AU < 0, they are (net) complements.
• Magnitude: Larger absolute values mean stronger substitution or complementarity. Near zero indicates little scope to adjust one input in response to the other’s price.
• Symmetry: σ_ij^AU = σ_ji^AU.

In the two-input case, this definition coincides with the Hicks elasticity of substitution and can be interpreted as the percentage change in the input ratio needed to offset a one-percent change in the marginal rate of technical substitution along an isoquant. With three or more inputs, AES remains a well-defined, symmetric pairwise measure but does not, by itself, summarize all the directional substitution effects induced by a specific price change.

Key Assumptions and Conditions

• Cost minimization for a given output, with a twice-differentiable, well-behaved technology (so a smooth cost function exists).
• Interior solution: all relevant inputs are used in positive amounts at the comparison point.
• Local, small changes: AES is a local elasticity; large shocks may move the firm to different regions or corners.
• Stable prices of other inputs and fixed output during the comparison (compensated setting).

Implications

• Input price pass-through to mix and cost: When energy and capital have a high AES, an energy price increase can be partially offset by capital-deepening (e.g., more efficient equipment), making total cost less sensitive.
• Technology choice and flexibility: Processes with higher AES among key inputs offer more room to adapt to future price or policy shocks—valuable under uncertainty.
• Policy evaluation: Carbon pricing, wage floors, or tariffs will reallocate cost shares depending on AES. Low AES between regulated and alternative inputs implies steeper cost increases for a given policy.
• Estimation and planning: Empirically, AES is often recovered from flexible functional forms (e.g., translog cost functions). Managers can use estimated AES to stress test input strategies.

Example in Practice

Suppose a chemical manufacturer uses three inputs intensively: energy (E), capital (K), and labor (L). An econometric study using a translog cost function yields:

• σ_EK^AU ≈ 0.9 (energy–capital are fairly substitutable),
• σ_EL^AU ≈ 0.2 (energy–labor are weak substitutes),
• σ_KL^AU ≈ −0.1 (capital–labor are mild complements).

Now consider a 25% increase in energy prices. The high σ_EK^AU signals that the plant can economically respond by investing in energy-efficient capital (heat recovery, better drives), cushioning cost growth. The low σ_EL^AU warns that staffing levels will not adjust much via energy–labor substitution. Since K and L are slightly complementary (negative AES), expanding capital to save energy will tend to pull labor up, not down. The plant’s planning team therefore prioritizes capital retrofits and schedules additional cross-training rather than expecting labor cuts to offset energy costs.

Limitations and Common Misunderstandings

• Interpretation in multi-input settings: Unlike the two-input Hicks elasticity, AES in higher dimensions is a symmetric, pairwise curvature measure—useful but not a complete description of how input ratios change when a particular price moves.
• Directional substitution: For analyzing “how much i replaces j when w_j rises,” many economists prefer the Morishima elasticity of substitution (MES), which is directional and more closely tied to observable demand responses.
• Local and interior: AES is defined at interior points and for small changes. Near corners (e.g., zero usage of an input) or with discrete technology choices, it can be undefined or misleading.
• Estimation sensitivity: Numerical values depend on the functional form (CES vs. translog), data quality, and identification. Treat estimates as approximations within a model, not structural constants.
• Sign and concavity: The cost function is concave in prices, so own-price second derivatives are nonpositive, but cross-partials (and thus AES signs) can be positive or negative; do not assume all inputs are substitutes.

Related Concepts

• Hicks elasticity of substitution
• Morishima elasticity of substitution
• Translog cost function
• CES production function
• Shephard’s lemma
• Slutsky (substitution) matrix