Returns to Scale Versus Diminishing Marginal Returns

Two Different Questions About Production

A production function describes the maximum output that can be produced from specified quantities of inputs, given the production method and relevant conditions. Two useful questions about that function may sound similar but require different comparisons. Returns to scale asks what happens when all inputs change together. Diminishing marginal returns asks what happens to the marginal product of one input as that input changes while other inputs remain fixed.

The distinction matters because a production technology can have diminishing marginal returns to labor and, at the same time, increasing returns to scale. There is no contradiction: the first statement holds capital fixed, while the second changes labor and capital together. To classify a production function correctly, always identify which inputs are changing.

Returns to Scale: Change All Inputs Together

For a production function homogeneous of degree \(r\), scaling every input by a factor \(t>0\) scales output by \(t^r\). If \(r=1\), the technology has constant returns to scale; if \(r>1\), it has increasing returns to scale; and if \(r<1\), it has decreasing returns to scale. For example, doubling both labor and capital means \(t=2\), so output changes by the factor \(2^r\).

This classification concerns a proportional change in the entire input bundle. It does not describe what happens when only labor changes. For a production function that is not homogeneous, the same comparison can still be made at particular input levels, but a single degree \(r\) may not characterize the technology at every scale.

Diminishing Marginal Returns: Change One Input

The marginal product of labor, written \(MP_L\), is the additional output associated with a small increase in labor, holding capital and other inputs fixed. If the production function is differentiable, \(MP_L\) is the partial derivative of output with respect to labor. Diminishing marginal returns to labor occur when this marginal product falls as labor rises with capital fixed. For a twice-differentiable function, a standard condition is that the second partial derivative with respect to labor is negative over the range being studied.

The derivative gives a local marginal product. With discrete input units or a production table, marginal product can instead be measured by the change in output divided by the change in labor, keeping other inputs fixed. A declining marginal product means later additions to labor contribute less output per unit than earlier additions over the range examined. It does not necessarily mean that total output is falling; total output can continue to rise while rising at a decreasing rate.

Why the Tests Are Different

For a homogeneous function \(f\) of degree \(r\), the scale test follows directly from the definition of homogeneity: multiplying all inputs by \(t\) multiplies output by \(t^r\). The marginal-product test, by contrast, differentiates with respect to one input while leaving the other input unchanged. These operations are not interchangeable.

Consider the Cobb–Douglas form \(f(L,K)=A L^\alpha K^\beta\), where \(A>0\), \(L>0\), \(K>0\), and the exponents are constant. This function has degree \(\alpha+\beta\), because multiplying both inputs by \(t\) multiplies output by \(t^\alpha t^\beta=t^{\alpha+\beta}\). Its labor marginal product and second derivative are:

When \(0<\alpha<1\), the second derivative is negative for positive \(A,L,K\), so labor has diminishing marginal returns with capital held fixed. The returns-to-scale classification, however, depends on \(\alpha+\beta\), not on \(\alpha\) alone. The two tests therefore use different information about the same function.

Common Exam Trap and Practical Interpretation

A frequent error is to describe an input change without specifying what happens to the other inputs. For example, the statement “labor rises and output rises” is not enough to establish either returns to scale or diminishing marginal returns. Both concepts require a controlled comparison: all inputs change together for returns to scale; only the input of interest changes for its marginal returns.

Wrong calculation: Suppose a candidate observes that output rises from 12 to 18 when labor rises from 4 to 9 and capital remains at 9. They calculate \(18/12=1.5\) and conclude that the production function has increasing returns to scale. This is not a returns-to-scale test, because capital was not scaled along with labor.

Correction: For the function \(q=2\sqrt{LK}\), the observed fixed-capital comparison concerns labor’s marginal product, not returns to scale. To test scale, scale both inputs by the same factor: \(q(tL,tK)=tq(L,K)\), which establishes constant returns to scale. The mistake is not the arithmetic \(18/12=1.5\); it is assigning that one-input comparison the wrong economic meaning.

These distinctions help an analyst describe a production technology precisely. A firm considering a larger operation needs an all-input comparison to assess how output responds when its input bundle expands. A manager deciding whether to add labor to an existing facility needs to consider labor’s marginal product with the facility’s other inputs held fixed. These are distinct production questions, and neither answer should be substituted for the other.