Common Errors in Long-Run Cost and Scale Analysis

Why These Errors Matter

Long-run cost analysis links a firm’s production technology, input prices, and output target. The calculations are often short, but a small conceptual error can reverse the conclusion: a firm may be described as having economies of scale when only one input has diminishing marginal returns, or an interior tangency condition may be imposed when the least-cost choice is at a boundary.

A reliable solution keeps three questions separate: how output changes when all inputs change proportionally; how output changes when one input changes while others are fixed; and how the least-cost input bundle changes as the output target changes. These are related ideas, but none can substitute for the others.

Definition: Returns to scale describe the change in output when all inputs are multiplied by the same positive factor. Marginal returns describe how output changes when one input changes while the other inputs are held fixed. Economies of scale describe how long-run average cost changes as output changes, with technology and input prices held fixed.

Do Not Confuse Returns to Scale with Marginal Returns

For a production function \(q=f(K,L)\), where \(K\) is capital and \(L\) is labor, returns to scale compare \(f(tK,tL)\) with \(f(K,L)\) for \(t>1\). If output rises by more than the factor \(t\), the technology has increasing returns to scale over that comparison; if it rises by exactly \(t\), it has constant returns; and if it rises by less than \(t\), it has decreasing returns.

Diminishing marginal returns to labor instead concern the marginal product of labor, \(MP_L=\partial f/\partial L\), as \(L\) rises with \(K\) fixed. The production function can have diminishing marginal returns to labor and still have increasing, constant, or decreasing returns to scale. The tests change different inputs in different ways.

Worked Example: Two Different Tests on One Technology

A workshop’s production function is \(q=K^{0.4}L^{0.4}\), with positive capital and labor. Classify its returns to scale, and determine whether it has diminishing marginal returns to labor when capital is fixed.

Method. For returns to scale, multiply both inputs by \(t\). The exponents sum to \(0.4+0.4=0.8\), so

f(tK,tL)=(tK)^{0.4}(tL)^{0.4} =t^{0.8}K^{0.4}L^{0.4} =t^{0.8}f(K,L).

Since \(0.8<1\), for \(t>1\) the output multiplier \(t^{0.8}\) is less than \(t\). The technology therefore has decreasing returns to scale.

Now hold \(K\) fixed and differentiate with respect to \(L\):

MP_L=0.4K^{0.4}L^{-0.6}.

As \(L\) increases, \(L^{-0.6}\) decreases, so \(MP_L\) falls. The technology also has diminishing marginal returns to labor. This second conclusion comes from the behavior of \(MP_L\), not from the returns-to-scale test. In another production function, the two classifications need not match.

Check. Doubling both inputs multiplies output by \(2^{0.8}\), approximately \(1.741\), which is less than \(2\). The scale classification is consistent with the exponent-sum test.

A Tangency Is Conditional, Not Universal

In long-run cost minimization, a firm chooses inputs to produce at least a target \(q_0\) at given input prices. Let \(r>0\) be the price of capital and \(w>0\) the price of labor. The cost is \(rK+wL\). When a cost-minimizing solution is an interior point on a smooth isoquant, the least-cost bundle satisfies a tangency condition: the isoquant and isocost have the same slope there.

That condition is not a rule for every technology or every optimum. It does not directly apply at a boundary, where one input may be zero, or where the isoquant has a kink. Nor does a tangency by itself establish a global minimum unless the relevant technology and cost problem support that conclusion. For example, convex isoquants and positive input prices help make an interior tangency a valid cost-minimizing choice; the candidate still needs to be checked against alternatives when the problem calls for it.

Interior tangency condition: If the cost-minimizing bundle uses positive amounts of both inputs, the production function is differentiable there, and the optimum is an interior point, then \(MP_K/MP_L=r/w\). Equivalently, the isoquant slope \(-MP_K/MP_L\) equals the isocost slope \(-r/w\). Boundary solutions and kinked technologies require direct comparison of feasible bundles rather than an imposed tangency.

The isocost slope depends on which input is placed on each axis. With \(K\) on the horizontal axis and \(L\) on the vertical axis, an isocost for total cost \(\bar C\) is \(rK+wL=\bar C\). Solving for \(L\) gives \(L=\bar C/w-(r/w)K\), so its slope is \(-r/w\). Reversing the axes changes the written slope; it does not change the underlying cost trade-off.

Worked Example: A Cost-Minimizing Corner Rather Than a Tangency

A producer has technology \(q=K+2L\). Capital costs \(r=4\) dollars per unit and labor costs \(w=1\) dollar per unit. Find the least-cost bundle that produces at least 12 units. State why an interior tangency calculation is not the right method.

Method. The feasible production constraint is \(K+2L\geq12\), not an equality for every feasible bundle. For instance, \((K,L)=(13,0)\) is feasible and does not satisfy \(K+2L=12\). But any cost-minimizing bundle must produce exactly 12 units: if it produced more, positive input use could be reduced while still meeting the target, lowering cost because both input prices are positive. Thus the minimization can be carried out on \(K+2L=12\).

The cost function is \(C=4K+L\). On the target isoquant, \(L=(12-K)/2\), where \(0\leq K\leq12\). Substitute this expression into cost:

C=4K+\frac{12-K}{2} =6+3.5K.

This cost increases as \(K\) increases. The minimum is therefore at \(K=0\), giving \(L=6\). The minimum cost is \(4(0)+1(6)=6\) dollars. The bundle produces \(0+2(6)=12\) units, as required.

There is no interior tangency: the least-cost bundle is on the boundary \(K=0\). The isoquant slope is \(-1/2\), while the isocost slope, with \(K\) horizontal and \(L\) vertical, is \(-r/w=-4\). They are not equal at the solution, and they do not need to be. Common mistake avoided: setting the slopes equal without first checking whether an interior solution exists.

Deriving and Checking an Interior Solution

For a smooth interior solution, the tangency condition has an economic interpretation: the last dollar spent on either input must contribute the same amount of output at the margin. The marginal output per dollar spent on capital is \(MP_K/r\); for labor it is \(MP_L/w\). If these are unequal, shifting a small amount of expenditure toward the input yielding more marginal output per dollar can increase production without increasing cost. At an interior optimum, therefore, \(MP_K/r=MP_L/w\), which rearranges to \(MP_K/MP_L=r/w\).

Worked Example: An Interior Tangency and a Global Cost Check

A firm produces \(q=\sqrt{KL}\), with \(r=4\) dollars per unit of capital and \(w=1\) dollar per unit of labor. Find the least-cost bundle for a target of 20 units, and verify that the tangency candidate is a minimum.

Method. The production constraint at a minimum is \(KL=20^2=400\). The marginal products are \(MP_K=\frac{1}{2}\sqrt{L/K}\) and \(MP_L=\frac{1}{2}\sqrt{K/L}\). Consequently, \(MP_K/MP_L=L/K\). At an interior tangency,

\frac{L}{K}=\frac{r}{w}=\frac{4}{1}=4, \qquad L=4K.

Substitute into the production constraint: \(K(4K)=400\), so \(4K^2=400\), \(K^2=100\), and \(K=10\) (inputs are nonnegative). Hence \(L=4(10)=40\). The resulting output is \(\sqrt{10(40)}=\sqrt{400}=20\). Cost is \(4(10)+1(40)=80\) dollars.

To check the candidate globally along the target isoquant, write \(L=400/K\), for \(K>0\). Cost becomes \(C(K)=4K+400/K\). Its derivative is \(C'(K)=4-400/K^2\), which is zero at \(K=10\); its second derivative is \(C''(K)=800/K^3>0\) for every \(K>0\). Thus cost is strictly convex along the isoquant and the candidate is the minimum. The isocost slope is \(-4\); the isoquant slope is \(-L/K=-40/10=-4\), confirming the tangency. Common mistake avoided: treating the tangency equation alone as a complete global-minimum check.

Output Growth Alone Does Not Show Economies of Scale

Economies of scale concern long-run average cost, not output growth in isolation. Long-run average cost at output \(q>0\) is \(C(q)/q\), where \(C(q)\) is minimum long-run cost at that output. With technology, input prices, and other relevant conditions unchanged, falling average cost as \(q\) rises indicates economies of scale; rising average cost indicates diseconomies; and constant average cost indicates neither over that range.

A reported rise in output does not reveal how much input use or cost changed, whether the technology changed, or what average cost did. A firm can produce more output while its average cost rises, falls, or stays constant. The comparison needs cost information (or enough technology and input-price information to derive cost), not output figures alone.

Derivation for a homogeneous production function: Suppose the production function is homogeneous of degree \(\rho>0\), so \(f(tK,tL)=t^\rho f(K,L)\), and input prices are fixed and positive. Under unrestricted long-run cost minimization, scaling a least-cost bundle that produces one unit by \(q^{1/\rho}\) produces \(q\) units. Conversely, scaling any bundle producing \(q\) down by \(q^{1/\rho}\) produces one unit. These two directions imply \(C(q)=C(1)q^{1/\rho}\). Thus average cost is \(C(q)/q=C(1)q^{1/\rho-1}\): it falls if \(\rho>1\), is constant if \(\rho=1\), and rises if \(0<\rho<1\). This relationship depends on the stated homogeneous technology and fixed input prices; it should not be inferred from output growth alone.

Worked Example: The Same Output Growth Can Accompany Different Cost Patterns

Two hypothetical firms each increase output from 100 to 150 units. Firm A has long-run cost \(C_A(q)=4q\). Firm B has long-run cost \(C_B(q)=q+q^2/100\). Calculate each firm’s average cost at both output levels and decide whether output growth alone identifies scale economies.

Method. For Firm A, at \(q=100\), cost is \(4(100)=400\) and average cost is \(400/100=4\) dollars per unit. At \(q=150\), cost is \(4(150)=600\) and average cost is \(600/150=4\) dollars per unit. Average cost is unchanged, so Firm A has constant average cost over this comparison.

For Firm B, at \(q=100\), cost is \(100+100^2/100=100+100=200\), giving average cost \(200/100=2\) dollars per unit. At \(q=150\), cost is \(150+150^2/100=150+225=375\), giving average cost \(375/150=2.5\) dollars per unit. Average cost rises, so Firm B has diseconomies of scale over this comparison.

Both firms’ output rises by 50%, but their average-cost patterns differ. The output increase by itself cannot establish economies of scale. Common mistake avoided: interpreting “more output” as evidence that production has become more efficient without comparing long-run average cost.

Common Errors to Catch Before Submitting a Solution

Changing one input and calling it a scale test. Returns to scale require multiplying all inputs by the same factor. Changing labor alone with capital fixed tests marginal returns to labor.
Assuming every least-cost bundle is a tangency. Check for a boundary solution or a kink. For fixed-proportions technology, for example, an optimum can occur at the kink rather than at a smooth tangency.
Using the wrong isocost slope. With \(K\) horizontal and \(L\) vertical, the slope is \(-r/w\). State the axes and derive the equation before substituting a slope.
Calling every feasible bundle an equality bundle. The production requirement is generally \(f(K,L)\geq q_0\). With positive input prices and a cost-minimizing choice, output will be exactly the target when excess output can be reduced without violating feasibility; that does not make the equality true of all feasible bundles.
Inferring scale economies from an output increase. Calculate or compare long-run average cost at different output levels, holding technology and input prices fixed.

Identify the comparison.
For returns to scale, scale all inputs together. For marginal returns, change one input while holding the others fixed. For economies of scale, compare long-run average cost across output levels.

Check the shape of the solution.
Before using tangency, establish that the candidate is interior and the production function is smooth there. Check boundaries and kinks separately.

Verify axes and units.
Write the isocost equation, derive its slope using the specified axes, and keep input prices and output units consistent.

Use the right evidence for the conclusion.
Output growth alone is not a cost result. Use a long-run cost function or sufficient information about technology and input prices to obtain average cost.

Key takeaway: Returns to scale, diminishing marginal returns, and economies of scale answer different questions. Correct analysis begins by identifying what is changing, then checking whether the relevant optimum is interior, and finally using the appropriate production or cost comparison.

Check Your Understanding

For each problem, distinguish the relevant concepts and show the key calculation before checking the answer.

A production function is \(q=K^{0.3}L^{0.4}\). State its returns to scale. With \(K\) fixed, does its marginal product of labor rise or fall as \(L\) increases? (Answer: Decreasing returns to scale, since \(0.3+0.4=0.7<1\); diminishing marginal returns to labor, since \(MP_L=0.4K^{0.3}L^{-0.6}\) falls with \(L\).)
A producer has \(q=3K+L\), with \(r=6\) and \(w=1\), and must produce at least 18 units. Find the least-cost bundle and explain whether an interior tangency applies. (Answer: \(K=0,L=18\), with cost \(18\); it is a boundary solution, so use direct cost comparison, not an interior tangency.)
A firm has \(q=K^{0.25}L^{0.75}\), input prices \(r=3,w=1\), and a target \(q=3\sqrt{3}\). At an interior tangency, find \(K,L\), and total cost. (Answer: \(MP_K/MP_L=L/(3K)=r/w=3\), so \(L=9K\); substitution gives \(q=3\sqrt{3}K\), hence \(K=1,L=9\), and cost \(3(1)+1(9)=12\).)
A long-run cost function is \(C(q)=5q+q^2/50\). Find average cost at \(q=50\) and \(q=100\), and classify the change over this range. (Answer: \(C(50)=300\), so average cost is \(6\); \(C(100)=700\), so average cost is \(7\). Average cost rises, indicating diseconomies of scale over this range.)
A firm’s output rises from 80 to 100 while both capital and labor rise by 25%, with technology unchanged. What does this comparison indicate about returns to scale over those bundles? Does it by itself establish economies of scale? (Answer: Constant returns to scale over the comparison, because inputs and output each rise by a factor of \(1.25\); it does not establish economies of scale without long-run cost or average-cost information.)

What You'll Learn