Why Scale and Unit Cost Are Different Questions
A firm that increases output may find that each unit costs less to produce, costs about the same, or costs more. These patterns matter for production planning: they help describe how the firm’s cost efficiency changes as it operates at different scales. The key measure is long-run average cost, not simply the number of inputs used or the total cost of production.
The long run is the planning horizon in which the firm can adjust all inputs. For each output level, the firm chooses an input combination that produces that output at the lowest possible expenditure, given its technology and input prices. Thus, a comparison of long-run average costs compares efficient production plans at different output levels. It does not describe what happens when output changes while one or more inputs are fixed.
These terms describe a relationship between output and unit cost. They are not labels for whether a firm is large or small in absolute terms. A firm can have economies of scale over one range of output and diseconomies over another. If \(LRAC\) is differentiable, the local versions are identified by the sign of its slope: a negative slope means average cost is falling, a zero slope means it is locally unchanged, and a positive slope means it is rising. The derivative is a local test; for an interval, compare average cost at the relevant output levels.
Classifying a Long-Run Cost Function
The classification can be made directly from a cost function by dividing total cost by output. This step matters: a change in total cost alone does not tell us whether unit cost has risen or fallen. Total cost will generally increase when output increases, even when average cost is decreasing.
Worked Example: Finding the Economies-of-Scale Range
A hypothetical manufacturer has long-run cost function \(C(q)=q^3-12q^2+60q\), for \(q>0\). Find its long-run average cost and determine where economies or diseconomies of scale occur.
Method. Divide the cost function by output, then examine how the resulting average cost changes with \(q\). This avoids the mistake of classifying scale economies from total cost alone.
First calculate average cost:
The completed-square form shows that average cost is lowest at \(q=6\). To classify the ranges, differentiate the average-cost function:
For \(0<q<6\), \(2q-12<0\), so average cost falls as output rises: there are economies of scale. At \(q=6\), the slope is zero and average cost is at its minimum. For \(q>6\), \(2q-12>0\), so average cost rises: there are diseconomies of scale. A zero slope at one output is not a range of constant economies.
Check the function at three output levels. At \(q=2\), \(C(2)=8-48+120=80\), so \(LRAC(2)=80/2=40\). At \(q=6\), \(C(6)=216-432+360=144\), so \(LRAC(6)=144/6=24\). At \(q=10\), \(C(10)=1000-1200+600=400\), so \(LRAC(10)=400/10=40\). These values confirm that average cost falls from 40 to 24 and then rises to 40.
Check. The direct values agree with \(LRAC(q)=(q-6)^2+24\): they give 40, 24, and 40, respectively. The classification is based on average cost, not on the fact that total cost rises from 80 to 144 to 400.
Connection to Returns to Scale
Returns to scale describe how output changes when all inputs are multiplied by the same factor. The earlier material on returns to scale distinguishes increasing, constant, and decreasing returns as properties of a production technology. Under standard assumptions, those properties help explain long-run average-cost patterns: if production has increasing returns to scale and input prices stay constant, a proportional increase in output can be achieved with a less-than-proportional increase in inputs and cost.
The following result makes that connection precise for a homogeneous production function. Suppose production is \(F(x)\), where \(x\) is the vector of inputs, and \(F(tx)=t^rF(x)\) for \(t>0\). The positive number \(r\) is the degree of homogeneity: \(r>1\) means increasing returns to scale, \(r=1\) constant returns, and \(0<r<1\) decreasing returns. Assume input prices are fixed and positive, all inputs can be adjusted, and cost minimization is feasible at the output levels considered.
Why the result holds. Take an input bundle that minimizes cost for output \(q\), with cost \(C(q)\). Multiplying every input by \(\lambda^{1/r}\) produces \(\lambda q\), because homogeneity gives \(F(\lambda^{1/r}x)=\lambda F(x)=\lambda q\). At unchanged input prices, this scaled bundle costs \(\lambda^{1/r}C(q)\). Therefore, the minimum cost at \(\lambda q\) cannot exceed that amount: \(C(\lambda q)\leq\lambda^{1/r}C(q)\).
Now take a cost-minimizing bundle for \(\lambda q\). Scaling all of its inputs by \(\lambda^{-1/r}\) produces \(q\). The cost of that scaled bundle is \(\lambda^{-1/r}C(\lambda q)\), so the minimum cost at \(q\) cannot exceed that amount: \(C(q)\leq\lambda^{-1/r}C(\lambda q)\). Rearranging gives \(C(\lambda q)\geq\lambda^{1/r}C(q)\). The two inequalities establish equality. Dividing by \(\lambda q\) gives the average-cost formula. Its exponent is negative for \(r>1\), zero for \(r=1\), and positive for \(0<r<1\), which proves the classification.
Worked Example: Increasing Returns and Falling Unit Cost
A hypothetical workshop produces output according to \(q=KL\), where \(K\) is machine input and \(L\) is labor input. The technology is homogeneous of degree 2. Machine input costs 4 dollars per unit and labor costs 1 dollar per unit. Find the cost-minimizing inputs and average cost at outputs \(q=1\), \(q=4\), and \(q=16\).
Method. For each required output, minimize expenditure \(4K+L\) subject to \(KL=q\). Since \(L=q/K\), the cost as a function of \(K>0\) is \(4K+q/K\). Its derivative is \(4-q/K^2\), which is zero at \(K=\sqrt{q}/2\). The second derivative is \(2q/K^3>0\), so this point minimizes cost. Substituting into the production constraint gives \(L=2\sqrt{q}\); the minimum cost is \(4\sqrt{q}\).
At \(q=1\), \(K=1/2\) and \(L=2\). Cost is \(4(1/2)+2=4\), so average cost is \(4/1=4\).
At \(q=4\), \(K=1\) and \(L=4\). Cost is \(4(1)+4=8\), so average cost is \(8/4=2\).
At \(q=16\), \(K=2\) and \(L=8\). Cost is \(4(2)+8=16\), so average cost is \(16/16=1\). The computed average costs fall from 4 to 2 to 1 as output increases.
Check. The cost function \(C(q)=4\sqrt{q}\) implies \(LRAC(q)=4/\sqrt{q}\), which gives 4, 2, and 1 at the three stated outputs. This agrees with the direct calculations and with the increasing returns to scale of the degree-2 technology.
Why Average Cost Can Change in Practice
Economies of scale can arise when a larger operation makes production more efficient. For example, specialized equipment may be used more fully, workers may focus on tasks suited to their skills, or a fixed organizational capability may support more units of output. These are possible explanations for falling long-run average cost, not separate definitions of economies of scale. The definition is always about the direction of change in \(LRAC\).
At larger scales, coordination can become more difficult. Communication across teams, additional management layers, or the complexity of scheduling and quality control may raise the cost per unit. Such pressures can contribute to diseconomies of scale. Constant economies describes a range in which unit cost does not change as output rises; it does not mean that the firm has no costs or that total cost is constant.
The production-function result relies on important assumptions. In particular, it holds with fixed input prices and a homogeneous technology. If input prices change with the scale of the firm, the cost consequences of expanding production may differ from the returns-to-scale prediction. Also, production may have different returns to scale over different ranges, or its scale advantages may be offset by coordination costs. In those cases, classify the observed cost pattern from long-run average cost rather than inferring it from a production label alone.
Worked Example: Reading Cost Data Without Overclaiming
A hypothetical firm estimates its minimum long-run costs at three output levels: \(C(100)=\$1{,}000\), \(C(200)=\$1{,}800\), and \(C(300)=\$3{,}000\). Calculate average cost at each output and describe the scale pattern supported by these observations.
Method. Divide each total cost by its corresponding output, keeping the units consistent. Do not compare total costs alone or claim a precise pattern between observations that have not been supplied.
At \(q=100\), \(LRAC(100)=\$1{,}000/100=\$10\) per unit. At \(q=200\), \(LRAC(200)=\$1{,}800/200=\$9\) per unit. At \(q=300\), \(LRAC(300)=\$3{,}000/300=\$10\) per unit.
Between the observed outputs of 100 and 200, average cost falls from \$10 to \$9 per unit, so the observations show economies of scale over that comparison. Between 200 and 300, it rises from \$9 to \$10 per unit, showing diseconomies over that comparison. The data do not establish exactly how average cost behaves at every intermediate output; that would require more information about the cost function.
Check. The cost changes are consistent with the calculated unit costs: total cost rises at each step, while average cost first falls and then rises. The conclusion concerns the observed comparisons, not an unobserved continuous curve.
Distinguishing Scale Economies from Nearby Ideas
A common conceptual error is to treat economies of scale as another name for diminishing marginal returns. They concern different comparisons. Returns to scale ask what happens to output when all inputs change proportionately. Diminishing marginal returns concern the additional output from increasing one input while at least one other input is held fixed. Economies of scale, by contrast, ask whether minimum long-run cost per unit changes as output changes.
The distinction matters because the long run allows all inputs to adjust, while diminishing marginal returns is a fixed-input comparison. A technology can have increasing returns to scale under the relevant production conditions, yet a firm’s actual average-cost pattern may not match the simple prediction if input prices change or coordination costs become important. Always identify the measure and assumptions before applying a label.
This treatment is aligned with the SOA VEE Economics (Validation by Educational Experience) syllabus.
Check Your Understanding
For each problem, identify the relevant average-cost comparison and state the scale pattern. Treat all cost data as hypothetical.
- A long-run cost function is \(C(q)=3q+q^2/4\). Calculate average cost at \(q=4\) and \(q=12\), then classify the change as output rises between those levels. (Answer: \(LRAC(4)=\$4\), \(LRAC(12)=\$6\); diseconomies of scale over the comparison.)
- A technology is \(q=\sqrt{KL}\), with both inputs costing 2 dollars per unit. For output \(q=3\), the cost-minimizing choice is \(K=L=3\). Calculate total and average cost, then state the scale pattern implied by this technology with constant input prices. (Answer: total cost is \$12 and average cost is \$4 per unit; constant economies of scale.)
- A firm’s minimum long-run costs are \(C(50)=\$400\), \(C(100)=\$700\), and \(C(200)=\$1{,}600\). Calculate average cost at each output and classify each observed interval. (Answer: average costs are \$8, \$7, and \$8 per unit; economies of scale from 50 to 100, then diseconomies from 100 to 200.)
- A production function is homogeneous of degree 2. Output is multiplied by 8, input prices remain fixed, and the original average cost is \(A\). Find the input-scaling factor and the new average cost in terms of \(A\). (Answer: inputs scale by \(\sqrt{8}\); new average cost is \(A/\sqrt{8}\), so the technology implies economies of scale.)
- A differentiable long-run average-cost function is \(LRAC(q)=(q-5)^2+9\), for \(q>0\). Determine whether average cost falls or rises over \(q=3\) to \(q=5\), and over \(q=5\) to \(q=8\). (Answer: it falls from 13 to 9 between 3 and 5, then rises from 9 to 18 between 5 and 8; economies of scale on the first interval and diseconomies on the second.)