Why Homogeneity Is a Useful Test
Returns to scale asks what happens to output when all inputs are multiplied by the same positive factor. Homogeneity gives a precise way to answer: if output is multiplied by a fixed power of that factor, the power identifies the production function’s returns to scale. This tutorial focuses on that test and on Euler’s theorem, which provides a derivative-based check. The material is aligned with the SOA VEE Economics (Validation by Educational Experience) syllabus.
For production, the standard classification compares the degree with one. If \(r=1\), multiplying all inputs by \(t\) multiplies output by \(t\), giving constant returns to scale. If \(r>1\), output grows by a greater proportion than the common input scale factor, giving increasing returns to scale. If \(r<1\), output grows by a smaller proportion, giving decreasing returns to scale. This comparison describes scaling all inputs together; it is not a claim about changing only one input.
Finding the Degree by Scaling
A direct test is often the safest method: substitute \(tx_1,\ldots,tx_n\) into the function, simplify, and see whether the result is \(t^rF(x_1,\ldots,x_n)\) for one fixed exponent \(r\). The identity must hold for every admissible input vector and every \(t>0\). Checking a single input point is not enough to establish homogeneity.
For a Cobb-Douglas function \(F(K,L)=A K^\alpha L^\beta\), where \(A>0\) and \(K,L>0\), scaling both inputs gives \(F(tK,tL)=A(tK)^\alpha(tL)^\beta=t^{\alpha+\beta}F(K,L)\). Thus the degree is \(\alpha+\beta\). For a sum of terms, the same shortcut works only if every term has the same total degree: otherwise the powers of \(t\) generally cannot be factored into one common power.
Worked Example: A Cobb-Douglas Function with Decreasing Returns
Suppose a hypothetical workshop’s production function is \(F(K,L)=2K^{1/3}L^{1/3}\), where \(K\) is the capital input and \(L\) is the labor input. Find its degree, classify its returns to scale, and check the classification by calculating output before and after scaling both inputs by 8 from \(K=L=8\).
Method. First scale both inputs algebraically. The exponents add to \(1/3+1/3=2/3\), so \(F(tK,tL)=2(tK)^{1/3}(tL)^{1/3}=t^{2/3}F(K,L)\). The degree is \(2/3\), which is less than one. The function therefore has decreasing returns to scale.
Numerical check. At \(K=L=8\), each cube root is 2, so \(F(8,8)=2(2)(2)=8\). Scaling each input by 8 gives \(K=L=64\). Each cube root is then 4, so \(F(64,64)=2(4)(4)=32\). The output ratio is \(32/8=4\), while the input scale factor is 8. The prediction also gives \(8^{2/3}=4\), matching the computed output ratio.
Check and common mistake. Output and inputs both increased, but output increased by a factor of 4 rather than 8; this confirms decreasing returns. A common error is to compare the change in output with the change in just one input. Returns to scale requires scaling both inputs by the same factor.
When a Sum Is Homogeneous
Each product of powers \(K^\alpha L^\beta\) scales by \(t^{\alpha+\beta}\). Therefore, a sum of such terms is homogeneous when all terms have the same sum of exponents. The coefficients can differ: a constant coefficient does not change a term’s degree. If the terms have different degrees, the function generally has no single homogeneity degree.
Worked Example: A Sum of Terms with the Same Degree
Consider \(F(K,L)=K^{1/2}L^{1/2}+K^{1/4}L^{3/4}\), for positive inputs. Test for homogeneity, classify the returns to scale, and verify the result by scaling the inputs from \((1,1)\) to \((9,9)\).
Method. The first term has degree \(1/2+1/2=1\), and the second has degree \(1/4+3/4=1\). Scaling gives \(F(tK,tL)=tK^{1/2}L^{1/2}+tK^{1/4}L^{3/4}=tF(K,L)\). The function is homogeneous of degree 1 and has constant returns to scale.
Numerical check. At \((1,1)\), both terms equal 1, so output is 2. At \((9,9)\), the first term is \(\sqrt{9\cdot9}=9\). The second is \(9^{1/4}9^{3/4}=9^{(1/4+3/4)}=9\). Output is \(9+9=18\), and the output ratio is \(18/2=9\), exactly the input scale factor.
Check and common mistake. Both terms scale by the same factor, so they can be added after scaling. It would be incorrect to conclude that a sum is homogeneous merely because one of its terms is homogeneous; every term must have the same degree for this test to establish a common degree.
Euler’s Theorem and Its Justification
For differentiable homogeneous production functions, the degree can also be recovered from the function and its partial derivatives. This is useful as a consistency check, but it requires both homogeneity and differentiability at the input vector being considered. The scaling definition itself does not require differentiability.
The identity follows directly from the scaling property. Fix an input vector \(x=(x_1,\ldots,x_n)\) and define \(g(t)=F(tx_1,\ldots,tx_n)\). Homogeneity says \(g(t)=t^rF(x)\). Differentiate both sides with respect to \(t\), then set \(t=1\). By the chain rule, the derivative of the left side is \(\sum_j x_j(\partial F/\partial x_j)(x)\); the derivative of the right side is \(rF(x)\). Equating them gives Euler’s identity.
When \(F>0\), dividing by output gives \(\big(\sum_j x_j F_{x_j}\big)/F=r\). This ratio is the elasticity of output with respect to a common proportional scaling of all inputs, evaluated at that input vector. For a homogeneous function it equals the same degree everywhere it is defined. For a function that is not homogeneous, the ratio may still be calculated locally, but it does not by itself establish a global returns-to-scale classification.
Different-Degree Terms: Test Before Classifying
A function containing terms of different degrees cannot generally be assigned the degree of one selected term. The scaling definition tests the whole function. A useful diagnostic is to compare output ratios at different scale factors from the same base input vector: if the function were homogeneous of degree \(r\), each ratio would equal the corresponding scale factor raised to \(r\).
Worked Example: A Mixed-Degree Function and a Local Euler Ratio
Let \(F(K,L)=K^{1/2}L^{1/2}+K^{1/4}L^{1/4}\), with positive inputs. Starting at \((1,1)\), calculate output ratios when both inputs are scaled by 4 and by \(1/4\). Then use partial derivatives at \((4,4)\) to calculate the Euler ratio. Decide whether the function has one global homogeneity degree.
Scaling test. At \((1,1)\), the first term is 1 and the second term is 1, so \(F(1,1)=2\). At \((4,4)\), the first term is \(\sqrt{16}=4\), and the second is \(16^{1/4}=2\). Therefore \(F(4,4)=6\), and the output ratio for scale factor 4 is \(6/2=3\).
At \((1/4,1/4)\), the first term is \(\sqrt{1/16}=1/4\). The second term is \((1/16)^{1/4}=1/2\). Thus \(F(1/4,1/4)=3/4\), and the output ratio is \((3/4)/2=3/8\). If the function had degree \(r\), these ratios would be \(4^r\) and \((1/4)^r\), whose product must equal 1. The actual product is \(3(3/8)=9/8\ne1\), so no one degree satisfies the scaling definition.
Euler ratio at \((4,4)\). For the first term, each partial derivative is \(1/2\) at \((4,4)\). For the second term, the derivative with respect to \(K\) is \((1/4)K^{-3/4}L^{1/4}\), which equals \((1/4)(4^{-1/2})=1/8\) at \((4,4)\); by symmetry the derivative with respect to \(L\) is also \(1/8\). Hence \(F_K=F_L=1/2+1/8=5/8\). Since \(F(4,4)=6\), $$ \frac{K F_K+L F_L}{F} =\frac{4(5/8)+4(5/8)}{6} =\frac{5}{6}. $$
Check and common mistake. The Euler ratio is a local value here, not a global degree: the function failed the scaling test. A tempting but incorrect calculation is to take the first term’s degree, 1, as the degree of the whole sum. That would predict \(F(4,4)=4F(1,1)=8\), whereas direct calculation gives 6. The two terms have degrees 1 and \(1/2\), so they do not scale together.
A Reliable Testing Procedure
The following sequence keeps the definition, classification, and derivative check distinct. In particular, Euler’s theorem is a consequence of homogeneity; it should not be used to assume homogeneity.
Replace each input \(x_j\) by \(tx_j\), using one common \(t>0\).
Check whether the result is \(t^rF(x_1,\ldots,x_n)\) with one exponent \(r\) that works for all inputs and scale factors.
If the function is homogeneous, compare its degree \(r\) with 1: greater means increasing, equal means constant, and less means decreasing returns to scale.
For a differentiable homogeneous function, verify \(\sum_j x_j F_{x_j}=rF\) at a convenient interior input vector.
The distinction matters in production analysis because a function’s algebra can look familiar while its terms scale differently. Identifying homogeneity correctly tells you whether a single scale response applies across input combinations. It also prevents confusing an overall proportional change in all inputs with a change in just one input. Euler’s theorem adds a compact check, but only after the scaling property has been established.
Check Your Understanding
For each problem, test the stated scaling carefully; use the definition before relying on a shortcut.
- A production function is \(F(K,L)=5K^{1/5}L^{1/2}\). Find its degree and classify returns to scale. What output multiplier results when both inputs are tripled? (Answer: degree \(7/10\); decreasing returns to scale; output multiplier \(3^{7/10}\).)
- For \(F(K,L)=K^{2/5}L^{3/5}\), determine the degree and the output multiplier when both inputs are scaled by 6. State the returns classification. (Answer: degree 1; multiplier 6; constant returns to scale.)
- Let \(F(K,L)=K^{1/2}L^{1/2}+2K^{1/3}L^{2/3}\). Find the degree, then calculate output at \((1,1)\) and at \((4,4)\) to check the classification. (Answer: degree 1; outputs 3 and 12; constant returns to scale.)
- For \(F(K,L)=K^{1/2}L^{1/2}+K^{1/2}\), calculate \(F(1,1)\), \(F(4,4)\), and \(F(1/4,1/4)\). Does one homogeneity degree fit both scale tests from \((1,1)\)? (Answer: outputs 2, 6, and \(3/4\); no, the function is not homogeneous of one degree.)
- For \(F(K,L)=K^{1/2}L^{1/2}\), calculate \(F\), \(F_K\), and \(F_L\) at \((4,9)\), then verify Euler’s identity and state the degree. (Answer: \(F=6\), \(F_K=3/4\), \(F_L=1/3\); \(4(3/4)+9(1/3)=6=F\); degree 1.)