Why Minimum Efficient Scale Matters
The long-run average-cost curve describes the lowest average cost at which a firm can produce each output when it can adjust its inputs. Minimum efficient scale identifies the smallest output associated with the lowest average cost. It is a useful benchmark for comparing efficient production scale with the amount a firm might need to produce; it does not, by itself, determine the output a firm will choose.
The phrase “the smallest output” requires care. A cost curve may approach its lowest level without reaching it, or it may attain that level at many outputs without a smallest one. In either case, simply pointing to a low or flat part of the curve does not establish an exact minimum efficient scale. The feasible output set and the set of outputs attaining the minimum must be checked.
This definition allows for several shapes. If average cost has one unique lowest point at a feasible output, that output is MES. If the lowest cost is attained throughout a closed interval whose left endpoint is feasible, MES is the left endpoint. But a set of minimizing outputs can lack a least element: for example, if it is \((0,4]\), every member has a smaller member in that same set. In that case, the minimum cost is attained, but exact MES is not defined.
Locating a Smooth Minimum
For a long-run total-cost function \(C(q)\), average cost is \(AC_{LR}(q)=C(q)/q\) for positive output. Marginal cost is the derivative \(MC_{LR}(q)=C'(q)\), where the derivative exists. For a differentiable cost function, the relationship between marginal and average cost helps identify a smooth interior minimum.
Since \(q>0\), the sign of the derivative of average cost is the sign of \(MC_{LR}(q)-AC_{LR}(q)\). Thus, where marginal cost is below average cost, average cost is falling; where marginal cost is above average cost, average cost is rising. If a smooth interior minimum occurs, the derivative of average cost is zero there, so marginal cost equals average cost. That equality is a useful condition, not a complete test for a global minimum: endpoints, nondifferentiable points, and other stationary points must also be considered.
Worked Example: A Unique Interior Minimum and a Near-Minimum Threshold
A hypothetical producer can choose any positive output \(q\), measured in hundreds of units per month. Its long-run average cost is \(AC_{LR}(q)=100+(q-8)^2\) dollars per unit. Find exact MES. Then find the smallest output whose average cost is no more than \(4\) dollars per unit above the minimum.
Method. First minimize the average-cost function over the feasible set \(q>0\); then apply the stated threshold to define near-minimum output. The threshold is part of this question’s definition of “near,” not a universal convention.
Step 1: Find the global minimum. The squared term is nonnegative, so \(AC_{LR}(q)=100+(q-8)^2\ge 100\). Equality occurs when \((q-8)^2=0\), or \(q=8\). Since \(8\) is feasible, the minimum is attained there, and it is the unique minimizing output. Therefore exact MES is \(8\) hundreds of units per month, or \(800\) units per month.
Step 2: Check the interior condition. Differentiating gives \(dAC_{LR}/dq=2(q-8)\), which is zero at \(q=8\). The implied total cost is \(C(q)=q[100+(q-8)^2]\). Its marginal cost at \(q=8\) is \(C'(8)=100\), and average cost there is also \(100\), as the interior condition predicts.
Step 3: Apply the near-minimum threshold. The minimum average cost is \(100\), so the maximum allowed average cost is \(100+4=104\). Solve \(100+(q-8)^2\le104\), which gives \((q-8)^2\le4\), or \(6\le q\le10\). The least feasible output in this interval is \(q=6\), or \(600\) units per month. At that output, average cost is \(100+(6-8)^2=104\), confirming that the threshold is met exactly.
Common mistake avoided: Choosing \(q=8\) as the answer to both questions would confuse exact MES with the smallest output inside the specified near-minimum band. The exact minimum occurs at \(8\), but the stated tolerance permits an output of \(6\).
Exact Minimum, Infimum, and Near-Minimum Scale
A curve can have a lower bound that it never reaches. In that situation, there is no minimum average-cost value over feasible outputs, even if the curve gets arbitrarily close to one. The lower bound is called an infimum. This distinction matters because exact MES is defined using outputs that attain the minimum, not outputs that merely approach an unattained lower bound.
“Near-minimum scale” can still be useful, but it must be defined using a tolerance. For an absolute tolerance \(\varepsilon>0\), one possible definition is the least feasible output satisfying \(AC_{LR}(q)\le m+\varepsilon\), where \(m\) is the minimum average cost if attained, or the infimum if it is not. The near-minimum scale exists only if this set of outputs has a least element. Another analysis might choose a percentage tolerance instead; the chosen rule should always be stated.
Worked Example: Falling Average Cost That Never Reaches Its Infimum
A hypothetical service provider has feasible output \(q>0\), measured in thousands of transactions per month, and long-run average cost \(AC_{LR}(q)=30+18/q\) dollars per transaction. Determine whether exact MES exists. Then find the smallest output with average cost at most \(1\) dollar per transaction above the infimum.
Method. Separate the question of whether a minimum is attained from the question of whether an explicitly defined near-minimum output exists. This avoids treating an asymptotic limit as an achieved cost.
Step 1: Identify the lower bound. Since \(18/q>0\) for every feasible \(q>0\), \(AC_{LR}(q)>30\) for every feasible output. As \(q\) increases, \(18/q\) approaches zero, so average cost approaches \(30\). Therefore the infimum is \(30\), but no feasible output attains it. There is no minimum average-cost value, and exact MES is not defined.
Step 2: Apply the near-minimum tolerance. The specified rule allows average cost up to \(30+1=31\). Solve \(30+18/q\le31\). Subtracting \(30\) gives \(18/q\le1\). Because \(q>0\), this is equivalent to \(q\ge18\). Thus the smallest qualifying output is \(18\) thousand transactions per month.
Step 3: Verify the boundary. At \(q=18\), average cost is \(30+18/18=31\), which meets the threshold exactly. For any smaller positive \(q\), \(18/q>1\), so average cost exceeds \(31\).
Common mistake avoided: Reporting a finite output as exact MES because the cost curve is “close to” \(30\) would confuse an unattained infimum with an attained minimum. The output \(18\) is a tolerance-based near-minimum result, not exact MES.
Plateaus, Feasibility, and the Meaning of “Smallest”
If average cost is constant at its minimum over an interval, the least minimizing output depends on the interval’s feasible boundary. A left endpoint that is included in the feasible set and in the minimizing set is the least minimizer. If the endpoint is excluded, it cannot be selected merely because it is the interval’s boundary. There may be no least minimizing output.
Worked Example: A Minimum-Cost Plateau with No Least Output
A hypothetical production technology permits every positive output \(q\), with average cost in dollars per unit given by \(AC_{LR}(q)=50\) for \(0<q\le4\), and \(AC_{LR}(q)=50+(q-4)^2\) for \(q>4\). Determine whether exact MES exists. Then, using an absolute tolerance of \(2\) dollars per unit above minimum cost, determine whether a smallest near-minimum output exists.
Method. Find the entire minimizing set, not just its interval boundary, and test whether that set contains a least feasible member. Apply the same test to the tolerance set.
Step 1: Find the minimum and its minimizing set. For \(0<q\le4\), average cost is \(50\). For \(q>4\), the squared term is positive, so average cost exceeds \(50\). The minimum is therefore \(50\), and the set of outputs attaining it is \((0,4]\).
Step 2: Test for a least minimizer. The endpoint \(0\) is not feasible. Moreover, for any proposed minimizing output \(q\in(0,4]\), the output \(q/2\) is also feasible, is still in \((0,4]\), and is smaller than \(q\). Thus no member of the minimizing set is least. The minimum is attained, but exact MES is not defined.
Step 3: Find the near-minimum set. The tolerance rule is \(AC_{LR}(q)\le50+2=52\). Every \(q\in(0,4]\) satisfies it. For \(q>4\), solve \(50+(q-4)^2\le52\), giving \((q-4)^2\le2\), and hence \(4<q\le4+\sqrt{2}\). Combining both ranges, the near-minimum set is \((0,4+\sqrt{2}]\).
Step 4: Check whether that set has a least output. For any output \(q\) in this set, \(q/2\) is positive, smaller, and also in the set. Therefore the tolerance set has no least member either; no smallest near-minimum output exists under this rule. Its infimum is \(0\), but \(0\) is infeasible.
Common mistake avoided: Calling \(0\) the MES or near-minimum output would select an infeasible endpoint. Calling \(4\) the MES would ignore the smaller feasible outputs that attain the same minimum cost.
Discrete Output Choices and Interpretation
Not every firm can choose every positive output. A production process may require whole batches, fixed machine configurations, or other restrictions, so the feasible set can be discrete. In that case, compare average cost only at feasible outputs. Interpolating between listed choices can produce an output that the technology does not permit.
Worked Example: Tied Minimum Across Feasible Output Options
A hypothetical manufacturer can produce only \(3\), \(5\), \(8\), or \(12\) thousand components per month. Its long-run average costs at those choices are, respectively, \(74\), \(62\), \(62\), and \(65\) dollars per component. Find exact MES and the smallest output whose average cost is within \(5\%\) of the minimum.
Method. Compare the costs over the stated feasible set, retain every output attaining the minimum, and then calculate the percentage threshold. The common error is to report a minimizing output without resolving a tie by the “smallest” condition.
Step 1: Find the minimum. The lowest listed average cost is \(62\) dollars per component. It is attained at \(q=5\) and \(q=8\), both measured in thousands of components per month.
Step 2: Select the least minimizing output. Of the two minimizing outputs, \(5<8\). Therefore exact MES is \(5\) thousand components per month.
Step 3: Apply the percentage tolerance. Five percent of the minimum is \(0.05(62)=3.10\) dollars, so the maximum permitted cost is \(62+3.10=65.10\) dollars per component. The feasible choices with cost at most \(65.10\) are \(5\), \(8\), and \(12\), whose costs are \(62\), \(62\), and \(65\). The output \(3\) is excluded because its cost is \(74\). The least qualifying output is \(5\) thousand components per month.
Step 4: Check the result. The selected output has cost \(62\le65.10\), while the only smaller feasible output, \(3\), has cost \(74>65.10\). This verifies the near-minimum answer as well as the exact MES.
Common mistake avoided: A tie does not make exact MES ambiguous when one minimizing output is smaller than the others. Conversely, checking only the lowest cost value without identifying all outputs attaining it would not establish which output is the least minimizer.
What MES Does and Does Not Tell You
MES summarizes the cost side of a production decision. It identifies the least feasible scale associated with the lowest long-run average cost, when such a least scale exists. If the curve declines over a wide output range, producing substantially below a chosen near-minimum threshold may mean giving up some average-cost savings. If the curve is flat around its minimum, several scales may have very similar average costs, so the precise minimum may be less informative than a clearly stated near-minimum range.
MES is not automatically the firm’s actual output, its profit-maximizing output, or the output that best matches demand. Those decisions involve considerations beyond the average-cost curve. Nor does a low average cost at a large output by itself prove that every smaller output is inefficient: the shape of the curve and the firm’s feasible choices matter. The concept is most useful when the cost curve, feasible output set, and any near-minimum tolerance are all made explicit.
The distinction between a minimum and an infimum, and between a closed and an excluded boundary, is central to careful scale analysis. This treatment of minimum efficient scale is aligned with the SOA VEE Economics (Validation by Educational Experience) syllabus.
Check Your Understanding
For each problem, check both the cost condition and whether the selected output is the least feasible output satisfying it.
- A hypothetical firm can choose any \(q>0\), and \(AC_{LR}(q)=72+(q-6)^2\) dollars per unit. Find exact MES and the smallest output with average cost no more than \(5\) dollars above minimum. (Answer: Exact MES is \(q=6\); the smallest near-minimum output is \(q=6-\sqrt{5}\approx3.764\).)
- A provider has \(AC_{LR}(q)=24+10/q\) for every feasible \(q>0\). Does exact MES exist? Under an absolute tolerance of \(2\) above the infimum, find the smallest qualifying output. (Answer: Exact MES does not exist because the infimum \(24\) is not attained; the smallest qualifying output is \(q=5\).)
- Feasible outputs are \(q\in[2,9]\), and \(AC_{LR}(q)=40\) for \(2\le q\le5\), while \(AC_{LR}(q)=40+(q-5)^2\) for \(5<q\le9\). Find the minimum cost, its minimizing set, and exact MES. (Answer: Minimum cost is \(40\); the minimizing set is \([2,5]\); exact MES is \(q=2\).)
- A hypothetical plant can choose only \(q=4,7,11\), with average costs \(91,84,87\) dollars per unit, respectively. Find exact MES and the smallest output within \(5\%\) of minimum cost. (Answer: Exact MES is \(q=7\); the \(5\%\) ceiling is \(88.20\), so the smallest near-minimum output is \(q=7\).)
- For feasible outputs \(0<q\le6\), average cost is \(55\) dollars per unit. Does exact MES exist? Does a smallest output within \(3\) dollars per unit of minimum cost exist? (Answer: Neither exists: both the minimizing set and the tolerance set are \((0,6]\), which has no least element.)