Why Returns to Scale Matter
A producer considering a larger operation needs to know how output responds when the whole production process expands. If labor, equipment, and other inputs are all doubled, does output double, more than double, or less than double? Returns to scale answer this question by comparing a proportional change in every input with the resulting change in output.
This is a property of a production technology: it describes the relationship between input bundles and feasible output. It does not ask what happens when just one input changes while the others stay fixed. In the earlier material on isoquants and input combinations, an input bundle was a particular combination of productive inputs. Here, the key operation is to multiply every component of that bundle by the same positive factor.
The comparison amount is important. When inputs rise by a factor of \(\lambda\), constant returns mean output rises by that same factor—not that output stays constant. Increasing returns mean output rises by a larger factor, while decreasing returns mean it rises by a smaller factor. A classification is global only if the relevant comparison holds for every input bundle and every scale factor greater than one. If it holds only for one comparison, report that result as specific to that bundle and factor.
How to Make the Comparison
A reliable method is to identify the original input bundle and output, find the common input scale factor, calculate output at the scaled bundle, and compare that output with the original output multiplied by the same factor. The calculation must scale all inputs equally.
Write the inputs as \(x_1,\ldots,x_n\) and the corresponding output as \(q=f(x_1,\ldots,x_n)\).
Check that each new input equals the same \(\lambda\) times its original amount, with \(\lambda>1\).
Find output at the scaled bundle and compare it with \(\lambda q\). Greater, equal, or less output indicates increasing, constant, or decreasing returns at that comparison.
Worked Example: Classifying a Pair of Observed Input Bundles
A hypothetical repair facility uses technicians and diagnostic stations. At 2 technicians and 5 stations, it completes 7 repair jobs per shift. At 6 technicians and 15 stations, it completes 21 jobs per shift. What does this pair of observations indicate about returns to scale?
Method. Compare the input bundles component by component. Technicians increase from 2 to 6, a factor of \(6/2=3\). Stations increase from 5 to 15, also a factor of \(15/5=3\). Thus every listed input has been tripled.
The original output is 7 jobs. Under constant returns to scale, tripling all inputs would produce \(3(7)=21\) jobs. The observed output at the expanded bundle is also 21 jobs.
The comparison indicates constant returns to scale for these two bundles. It does not, by itself, establish that the facility has constant returns at every possible input bundle. That broader conclusion would require a production function or enough additional evidence to support it.
Common mistake avoided: Saying that output “increased” is not enough to identify returns to scale. The relevant comparison is 21 jobs against the constant-returns benchmark of 21 jobs, not against the original 7 jobs.
Classifying Production Functions
When a production function is given, substitute the scaled inputs into the function. For an input bundle \((K,L)\), where \(K\) denotes capital and \(L\) denotes labor, proportional scaling by \(\lambda\) means evaluating output at \((\lambda K,\lambda L)\). To establish a global classification, the comparison must hold for all relevant positive input bundles and all \(\lambda>1\), not just for one convenient numerical choice.
For example, consider the fixed-proportions technology \(f(K,L)=\min(K,2L)\). Scaling both inputs by \(\lambda\) gives \(\min(\lambda K,2\lambda L)\). Because multiplying both quantities being compared by the same positive number preserves which is smaller, this output is \(\lambda\min(K,2L)\). The technology therefore has constant returns to scale. This conclusion follows from scaling both inputs together, not from assuming that the input mix stays fixed in a general technology.
Worked Example: Increasing Returns in a Cobb-Douglas Technology
Consider the production function \(f(K,L)=K^{0.75}L^{0.75}\). Classify its returns to scale by comparing output at \((K,L)=(1,1)\) with output when both inputs are multiplied by 4.
Method. First calculate the original output:
The scaled bundle is \((4,4)\). Substitution gives:
The constant-returns benchmark is \(4(1)=4\). Since \(8>4\), output rises by more than the fourfold increase in inputs. The example indicates increasing returns at this bundle and scale factor.
The same substitution also shows why the classification holds for any positive bundle and any \(\lambda>1\):
Because \(\lambda>1\), \(\lambda^{1.5}>\lambda\). Therefore \(f(\lambda K,\lambda L)>\lambda f(K,L)\) for positive \(K,L\), so this function has increasing returns to scale on that domain.
Common mistake avoided: The exponents sum to \(0.75+0.75=1.5\), but the classification is not based on saying “the sum is greater than one” without explanation. Scaling both inputs produces the factor \(\lambda^{1.5}\), which is then compared directly with the required benchmark \(\lambda\).
Worked Example: Decreasing Returns in an Additive Technology
A hypothetical small-scale producer’s output is described by \(f(K,L)=\sqrt{K}+\sqrt{L}\). At \((K,L)=(4,9)\), compare output with output after both inputs are multiplied by 4.
Method. Calculate the original output:
The scaled input bundle is \((16,36)\), so the new output is:
If the technology had constant returns at this comparison, output would be \(4(5)=20\). Since \(10<20\), this bundle and scale factor show decreasing returns to scale.
In fact, direct substitution gives the result for any positive \(K,L\) and \(\lambda>1\):
For \(\lambda>1\), \(\sqrt{\lambda}<\lambda\). Thus \(f(\lambda K,\lambda L)<\lambda f(K,L)\), establishing decreasing returns for this function. The calculation is checked: the scaled output is 10, which is less than the constant-returns benchmark of 20.
Common mistake avoided: Output doubled from 5 to 10, but inputs quadrupled. Comparing output only with its starting value would miss the relevant benchmark and could lead to the wrong classification.
What the Classification Does—and Does Not—Say
Returns to scale describe the response of output when the entire input bundle is scaled proportionally. They do not describe what happens when only one input changes. For example, doubling labor while leaving capital fixed is not a returns-to-scale comparison, even if output changes substantially.
A production technology may also have different returns to scale in different regions. In that case, a single global label is inappropriate: state the input range or scale comparison for which the result holds. When a problem gives only two bundles, first verify that every input changed by the same factor, and then limit the conclusion to the observed comparison unless more information justifies a broader claim.
A common exam trap is to confuse “output increased” with increasing returns. Suppose all inputs triple and output rises from 8 to 18. Output has increased, but the constant-returns benchmark is \(3(8)=24\). Since \(18<24\), that comparison shows decreasing—not increasing—returns to scale. The correct reference point is always the original output multiplied by the input scale factor.
Check Your Understanding
For each problem, check that the stated comparison scales every input proportionally before classifying the result.
- A workshop produces 9 units at inputs \((K,L)=(3,4)\) and 18 units at \((6,8)\). Identify the scale factor and classify returns for this comparison. (Answer: Inputs double; the constant-returns benchmark is 18, so the comparison shows constant returns to scale.)
- For \(f(K,L)=\sqrt{KL}\), calculate output at \((4,9)\) and at \((12,27)\). Classify returns for this comparison. (Answer: Outputs are 6 and 18; inputs triple and output triples, so constant returns to scale.)
- For \(f(K,L)=K^{0.8}L^{0.5}\), start at \((1,1)\) and double both inputs. Give the scaled output factor and classify the technology. (Answer: Scaled output is \(2^{1.3}\approx2.462\), greater than the constant-returns benchmark of 2; increasing returns to scale.)
- A producer has \(f(K,L)=\sqrt{K}+\sqrt{L}\). At \((9,16)\), both inputs are multiplied by 9. Calculate original and scaled output, then classify. (Answer: Original output is \(3+4=7\); scaled output is \(9+12=21\), less than the benchmark \(9(7)=63\); decreasing returns to scale.)
- A fixed-proportions technology is \(f(K,L)=\min(K,3L)\). Show how output changes when both inputs are multiplied by any \(\lambda>1\), and classify the technology. (Answer: \(f(\lambda K,\lambda L)=\min(\lambda K,3\lambda L)=\lambda\min(K,3L)=\lambda f(K,L)\); constant returns to scale.)