When economists study how firms make production decisions, they often need to understand the complex relationships between different inputs like labor, capital, energy, and materials. One powerful tool for this analysis is the estimation of interrelated factor demand systems using a technique called Seemingly Unrelated Regression Estimation, or SURE. This approach becomes especially valuable when combined with flexible production models like the translog cost function.
Table of Contents
- From production functions to cost functions: The dual perspective
- The translog cost function system: Flexibility meets structure
- Why SURE matters for factor demand estimation
- Handling restrictions and computing elasticities: Making the model work
- Allen elasticities of substitution: Understanding input relationships
- Real-world applications and empirical insights
From production functions to cost functions: The dual perspective
Imagine a manufacturing firm deciding how much labor to hire, how much equipment to purchase, and how much energy to consume. Traditional production theory approaches this from two angles: the production function, which describes how inputs combine to create output, and the cost function, which represents the minimum cost of producing a given level of output at specific input prices.
This duality between production and cost functions is fundamental to understanding factor demand. Shephard’s Lemma provides the mathematical bridge between these perspectives, stating that the derivative of a cost function with respect to an input price equals the firm’s conditional demand for that input. In simpler terms, if we know how a firm’s costs respond to changes in wage rates or equipment prices, we can directly determine how much of each input the firm will use.
The beauty of this approach lies in its practicality. Rather than trying to estimate a complex production function directly, researchers can work with the cost function and derive all the information they need about input demands. This dual approach has become particularly influential in applied econometrics because the compensated demand functions can be computed directly from cost minimization problems.
The translog cost function system: Flexibility meets structure
Among various functional forms used to model production relationships, the translog (transcendental logarithmic) specification stands out for its flexibility. Unlike simpler forms like the Cobb-Douglas function, which assumes constant elasticity of substitution between inputs, the translog functional form can capture many attributes of a cost function implied by economic theory without imposing restrictive assumptions.
Consider a textile manufacturer using five inputs: capital, labor, energy, materials, and services. The translog cost function expresses the logarithm of total cost as a quadratic function of the logarithms of input prices and output. When we apply Shephard’s Lemma to this function, we obtain a system of cost share equations-each showing what proportion of total cost goes to each input.
These cost share equations form the heart of the SURE framework. Because all cost shares must sum to one (every dollar spent goes somewhere), the equations are inherently connected. A shock that increases the labor share must decrease shares going to other inputs. This interdependence means that estimating these equations jointly, rather than separately, captures important cross-equation correlations and produces more efficient estimates.
Why SURE matters for factor demand estimation
The SURE methodology recognizes that while each cost share equation has its own disturbances, these disturbances are correlated across equations. When energy prices spike unexpectedly, for instance, it doesn’t just affect the energy cost share-it ripples through all input decisions. Estimating the system using iterated seemingly unrelated regression accounts for these contemporaneous correlations, producing parameter estimates that are more reliable than those obtained from equation-by-equation estimation.
Handling restrictions and computing elasticities: Making the model work
In practice, implementing a translog cost share system requires careful attention to several technical considerations. One critical issue is singularity. Since cost shares sum to one, including all equations in estimation would create perfect multicollinearity-the equations would be linearly dependent. The standard solution is to drop one equation arbitrarily and estimate the remaining ones. The parameters for the dropped equation can be recovered afterward through the homogeneity and symmetry restrictions imposed by economic theory.
These theoretical restrictions are not arbitrary. Homogeneity restrictions ensure that if all input prices double, the optimal input mix remains unchanged (only the cost level doubles). Symmetry restrictions come from Young’s theorem on the equality of cross-partial derivatives and ensure that the effect of a change in labor prices on capital demand equals the effect of a change in capital prices on labor demand. Imposing these restrictions shrinks the number of parameters considerably, which is particularly useful when working with limited data.
Allen elasticities of substitution: Understanding input relationships
Once the translog system is estimated, researchers can calculate Allen partial elasticities of substitution, which measure how easily firms can substitute one input for another in response to price changes. These elasticities indicate whether inputs are substitutes or complements in production-positive values suggest substitutes (when one becomes expensive, firms use more of the other), while negative values indicate complements (inputs used together).
For example, if the Allen elasticity between labor and services is high and positive, rising wages might lead firms to substitute toward purchased services like outsourcing or automation software. The Allen elasticity of substitution, also known as the partial elasticity of substitution, is perhaps the most popular measure in general applications, though economists debate its interpretation when dealing with more than two inputs.
These elasticities have profound implications for policy analysis. When governments consider energy taxes or minimum wage policies, understanding whether labor and energy are substitutes or complements helps predict how firms will adjust their production processes. Will higher energy costs lead to more labor-intensive production methods, or will they complement automation that uses more capital and less of both labor and energy?
Real-world applications and empirical insights
Applied studies using the translog SURE framework have revealed fascinating insights about industrial production. Research on the U.S. textile industry, for instance, found that most input pairs are substitutes, with particularly strong substitution between labor and services. This finding helps explain how the industry has responded to international competition and rising labor costs through greater use of outsourced services and automation.
The framework also allows researchers to test important economic hypotheses. Do industries exhibit constant returns to scale (proportional increases in all inputs lead to proportional output increases)? Are production technologies homothetic (the input mix doesn’t depend on output level)? These questions can be formally tested within the translog SURE framework using likelihood ratio tests.
What do you think? How might estimates of input substitutability change during major economic disruptions like the oil embargo of the 1970s or recent energy price volatility? Could understanding these elasticities help firms better prepare for future resource price shocks?
Leave a Reply