Why Short-Run Production Has Three Stages
In the short run, at least one input is fixed while a variable input changes. A production schedule can therefore show how output responds as more units of the variable input are added to the fixed capacity. Average product and marginal product summarize two different aspects of that response: output per unit of input, and the additional output from an extra unit. Together, they divide the production range into three useful stages.
The stages are not three different production technologies. They are ranges along a production relationship, classified by how average and marginal product compare and whether marginal product is positive or negative. The short-run setting matters: the classification describes changes in a variable input while other relevant inputs are held fixed.
Stage I: average product is rising; in the standard smooth case, marginal product is above average product. Stage I ends where average product reaches its maximum and marginal product equals it.
Stage II: marginal product is positive but below average product. Total product is still rising, while average product is falling. Stage II ends where marginal product reaches zero and total product is at its maximum.
Stage III: marginal product is negative. Adding the variable input reduces total output.
At \(L=0\), average product \(Q(L)/L\) is not defined. The stage description therefore applies to positive input quantities. In a discrete schedule, marginal product is usually calculated as the change in output divided by the change in input. The exact boundary may fall between observed rows, so a table may identify a transition only approximately.
How Average and Marginal Product Locate the Stages
Average product is an average across all units of the variable input already used. Marginal product is the contribution associated with an additional unit. If that additional contribution exceeds the current average, it pulls the average upward; if it is below the average, it pulls the average downward. This familiar average-and-increment logic explains why comparing \(MP\) with \(AP\) locates the first two stages.
The equality follows by differentiating \(Q(L)/L\) with the quotient rule, then substituting \(Q(L)/L=AP(L)\) and \(dQ(L)/dL=MP(L)\). Since \(L>0\), the denominator is positive. Thus, \(AP\) rises when \(MP>AP\), is locally at a turning point when \(MP=AP\), and falls when \(MP<AP\). At an interior maximum of a differentiable average-product curve, its derivative is zero, so \(MP=AP\).
This equality is a boundary condition, not a claim that marginal product always equals average product at every input level. Nor does the equation alone say whether marginal product itself is increasing or decreasing. In the familiar three-stage pattern, marginal product falls after its peak, crosses average product at average product’s maximum, remains positive through Stage II, and reaches zero at the maximum of total product.
Worked Example: Classifying a Discrete Production Schedule
A workshop keeps its floor space and equipment fixed while varying the number of assistants. Its hypothetical output schedule is below. Calculate average product and marginal product at each positive input level, then identify the stages represented.
| Assistants \(L\) | Total output \(Q\) | Average product \(Q/L\) | Marginal product \(Q(L)-Q(L-1)\) |
|---|---|---|---|
| 0 | 0 | Not defined | — |
| 1 | 8 | 8 | 8 |
| 2 | 18 | 9 | 10 |
| 3 | 30 | 10 | 12 |
| 4 | 40 | 10 | 10 |
| 5 | 48 | 9.6 | 8 |
| 6 | 54 | 9 | 6 |
| 7 | 56 | 8 | 2 |
| 8 | 55 | 6.875 | −1 |
Method and calculations. Each row adds one assistant, so marginal product is the difference between consecutive total-output values. For example, at \(L=3\), \(MP=30-18=12\), while \(AP=30/3=10\). At \(L=5\), \(MP=48-40=8\), while \(AP=48/5=9.6\). At \(L=8\), \(MP=55-56=-1\), while \(AP=55/8=6.875\).
Stage I. Average product rises from \(8\) at \(L=1\) to \(9\) at \(L=2\), then to \(10\) at \(L=3\). At \(L=4\), it remains at \(10\). The schedule therefore shows the maximum average product at both \(L=3\) and \(L=4\); the fourth assistant adds \(10\), equal to the average product at \(L=3\), so the average stays level.
Stage II. After this maximum, average product falls, but marginal product remains positive: it is \(8\), \(6\), and \(2\) for the fifth, sixth, and seventh assistants. Output continues to increase, reaching \(56\) at \(L=7\).
Stage III. The eighth assistant has marginal product \(-1\), since output falls from \(56\) to \(55\). That negative increment places the move from seven to eight assistants in Stage III. The table checks out: output rises through \(L=7\), where the last positive marginal product occurs, and then falls.
Common mistake avoided: Average product at \(L=5\) is \(9.6\), but that is not the fifth assistant’s marginal product. The correct marginal product is \(48-40=8\). Confusing the total output per assistant with the output added by the most recent assistant can lead to the wrong stage classification.
Stage Boundaries in a Smooth Production Function
The same comparisons can be made using a differentiable production function. Consider the hypothetical short-run function \(Q(L)=16L+4L^2-L^3\) for nonnegative \(L\), with all other inputs fixed. Its average product is defined for \(L>0\), and its marginal product is the derivative of total output.
To find the maximum of average product, differentiate it: \(dAP/dL=4-2L\). Setting this equal to zero gives \(4-2L=0\), so \(L=2\). For \(0<L<2\), the derivative is positive and average product rises. At \(L=2\), \(AP=16+8-4=20\), and \(MP=16+16-12=20\). This confirms the Stage I boundary condition \(MP=AP\).
To find the point where total output reaches its maximum, set marginal product equal to zero:
The roots are \(L=-4/3\) and \(L=4\). Only \(L=4\) is in the nonnegative input range. Marginal product is positive between \(L=2\) and \(L=4\), and negative for \(L>4\). At the endpoint, \(Q(4)=16(4)+4(4^2)-4^3=64+64-64=64\). The function therefore identifies Stage I up to \(L=2\), Stage II from \(L=2\) to \(L=4\), and Stage III beyond \(L=4\). At \(L=4\), \(MP=0\) and total output is at its maximum.
This example also shows why the stage labels should not be reduced to “marginal product rising” versus “marginal product falling.” The stage boundary at \(L=2\) is located by the maximum of average product, where \(MP=AP\); the boundary at \(L=4\) is where \(MP=0\). The sign and comparison conditions define the stages.
Worked Example: Using Unequal Input Increments
A greenhouse holds its land and lighting fixed. A hypothetical production trial records output at intervals of two workers: \(Q(0)=0\), \(Q(2)=30\), \(Q(4)=56\), \(Q(6)=66\), and \(Q(8)=60\). Calculate the marginal product per worker over each interval and use the observations to classify the ranges.
Method. Because each observation is two workers apart, marginal product per worker is \(\Delta Q/\Delta L\), not the raw difference in output. From zero to two workers, it is \((30-0)/(2-0)=30/2=15\). From two to four, it is \((56-30)/(4-2)=26/2=13\). From four to six, it is \((66-56)/(6-4)=10/2=5\). From six to eight, it is \((60-66)/(8-6)=-6/2=-3\).
Average product at \(L=2\) is \(30/2=15\); at \(L=4\), it is \(56/4=14\); at \(L=6\), it is \(66/6=11\); and at \(L=8\), it is \(60/8=7.5\). The observed average product has fallen after \(L=2\), while marginal product remains positive through the interval from four to six workers. These observations are consistent with Stage II over those positive-output increments. The final interval has negative marginal product, so adding workers from six to eight is in Stage III. Among the recorded input levels, total output peaks at \(L=6\).
The data do not show an increasing-average-product range: average product is already lower at \(L=4\) than at \(L=2\). Therefore, they do not reveal the exact Stage I boundary; it would be unjustified to invent one between unobserved values. The calculations are checked against the output changes: \(30\), \(26\), \(10\), and \(-6\) total units over intervals of two workers give marginal products of \(15\), \(13\), \(5\), and \(-3\) per worker.
Common mistake avoided: Calling the \(26\)-unit output increase from \(L=2\) to \(L=4\) a marginal product of \(26\) ignores that two workers were added. The per-worker marginal product is \(26/2=13\).
What the Stages Say—and What They Do Not Say
The stages provide a compact way to interpret production data. Stage I indicates that average output per unit of the variable input is still increasing. In Stage II, average product has passed its maximum, but extra input continues to add output. In Stage III, extra input reduces total output under the specified short-run conditions.
A declining marginal product is not the same as a negative marginal product. Marginal product can be positive and falling: output still rises, but each additional unit contributes less than the one before. Stage III requires a stronger condition, \(MP<0\). Likewise, a falling average product does not by itself prove that marginal product is negative; it may simply mean that positive marginal product is below the existing average.
The smooth-function boundaries are clear when the relevant maximums are interior and the function is differentiable. With discrete observations, a maximum may lie between recorded input levels, a plateau may occur, and marginal product may not equal average product exactly at an observed row. Use the increments available, state what the data establish, and avoid claiming a precise boundary the schedule cannot support.
The stages describe physical production, not a complete choice of the best input level for a business. Identifying Stage II does not, by itself, determine a profit-maximizing quantity; that decision also requires information about output value and input cost. For production analysis, the essential classification is: rising average product in Stage I, positive marginal product below average product in Stage II, and negative marginal product in Stage III.
Check Your Understanding
Classify each situation using the stated observations. For discrete data, distinguish a measured interval from an exact continuous boundary.
- At a positive input level, \(AP=18\) and \(MP=22\). What happens to average product as input rises locally, and which stage does this indicate? (Answer: Average product rises; Stage I.)
- A smooth production function has \(AP=14\) and \(MP=14\) at \(L=5\), with average product rising before \(L=5\) and falling after it. What boundary is at \(L=5\)? (Answer: The maximum of average product and the Stage I–Stage II boundary.)
- At \(L=7\), total output is 84 units, and at \(L=8\), it is 91 units. Find marginal product over this one-unit increase and classify the increment. (Answer: \(MP=91-84=7\); marginal product is positive, so the increment is not Stage III.)
- A production schedule reports \(Q(3)=45\) and \(Q(5)=69\), with two workers added. Calculate marginal product per worker. If average product at \(L=5\) is 13.8, is average product rising or falling over that addition? (Answer: \(MP=(69-45)/(5-3)=12\); because \(12<13.8\), average product is falling.)
- A differentiable production function has \(MP=0\) at \(L=9\), positive marginal product just below 9, and negative marginal product just above 9. What occurs at \(L=9\), and which stage follows it? (Answer: Total product reaches a local maximum at \(L=9\); Stage III follows.)