A production function maps the relationship between a firm's inputs and the maximum output it can produce. Every cost curve and supply decision you'll encounter later in the course traces back to the structure of production, so getting this right now pays off throughout the rest of intermediate micro.

Mathematical representation and components

The general form of a production function is:

$Q = f(L, K, M)$

where $Q$ is output, $L$ is labor, $K$ is capital, and $M$ is raw materials. The function tells you the maximum output achievable for any given combination of inputs, which means it assumes technical efficiency (no waste).

Inputs fall into two categories depending on the time horizon:

Fixed inputs can't be adjusted in the short run (e.g., factory size, heavy machinery)
Variable inputs can be changed freely (e.g., hours of labor, quantities of raw materials)

Returns to scale

Returns to scale describe what happens to output when you scale all inputs by the same proportion. Suppose you double every input:

Constant returns to scale (CRS): Output exactly doubles.
Increasing returns to scale (IRS): Output more than doubles. This often arises from specialization and division of labor that become possible at larger scales.
Decreasing returns to scale (DRS): Output less than doubles. This typically stems from coordination problems: as organizations grow very large, management layers multiply and communication slows down.

More formally, for a scaling factor $t > 1$ :

CRS: $f(tL, tK) = t \cdot f(L, K)$
IRS: $f(tL, tK) > t \cdot f(L, K)$
DRS: $f(tL, tK) < t \cdot f(L, K)$

A function that satisfies CRS is called homogeneous of degree 1. More generally, if $f(tL, tK) = t^k \cdot f(L, K)$ , the function is homogeneous of degree $k$ . When $k > 1$ you have IRS; when $k < 1$ you have DRS.

Don't confuse returns to scale (a long-run concept where all inputs change proportionally) with diminishing marginal returns (a short-run concept where one input changes while others stay fixed). These are completely independent properties. A production function can exhibit increasing returns to scale and diminishing marginal product of labor at the same time. This distinction is one of the most common mistakes on exams.

Short-run vs. long-run production

The short run and long run aren't defined by calendar time. They're defined by whether any inputs are fixed.

Characteristics and flexibility

In the short run, at least one input is fixed. Typically that's capital: you can't build a new factory overnight. The firm adjusts output by changing variable inputs like labor.
In the long run, all inputs are variable. The firm can resize its plant, adopt entirely new technologies, or restructure operations.

The actual calendar length of the "short run" varies by industry. For a restaurant adjusting staff levels, it might be days. For a power utility building a new plant, the short run could last years.

Short-run decisions: hiring or laying off workers, purchasing more raw materials, adding overtime shifts.

Long-run decisions: building new facilities, investing in automation, entering new markets.

Mathematical representation and components, Labor Productivity and Economic Growth | OpenStax Macroeconomics 2e

The envelope relationship

The long-run production (and cost) function envelopes all possible short-run functions. Each short-run function corresponds to a particular level of fixed capital. The long-run function traces out the best outcome you could achieve if you were free to choose the optimal level of capital for each output level.

Concretely, for every output level $Q$ , the long-run cost is the minimum across all possible short-run costs:

$LRC(Q) = \min_{K} SRC(Q, K)$

This is the envelope theorem at work, and it becomes especially important when you study long-run cost curves. The key implication: the long-run cost of producing any quantity is always weakly less than the short-run cost, because the long run gives you strictly more flexibility.

Inputs and outputs in production

Productivity measures

Three measures describe how productive a variable input (usually labor) is in the short run:

Total Product (TP): The total output produced at a given level of the variable input.
Average Product (AP): Output per unit of the variable input. $AP_L = \frac{TP}{L}$
Marginal Product (MP): The additional output from one more unit of the variable input. $MP_L = \frac{\Delta TP}{\Delta L}$ , or in continuous form, $MP_L = \frac{\partial f}{\partial L}$

The relationship between MP and AP follows a clean rule:

When $MP > AP$ , average product is rising (the marginal unit pulls the average up).
When $MP < AP$ , average product is falling (the marginal unit drags the average down).
When $MP = AP$ , average product is at its maximum.

Think of it like a batting average: if today's game (marginal) is better than your season average, your average goes up. You can prove this formally by differentiating $AP_L = \frac{f(L,K)}{L}$ with respect to $L$ and noting the sign depends on whether $MP_L$ exceeds $AP_L$ .

The three stages of short-run production

Stage	TP	AP	MP	Firm behavior
I	Increasing at an increasing rate	Rising	Positive and above AP	Firm is underusing its fixed input; should keep adding labor
II	Increasing at a decreasing rate	Falling	Positive but below AP	Rational operating range; diminishing returns set in
III	Decreasing	Falling	Negative	Too much variable input; adding labor actually reduces output

A rational firm always operates in Stage II. In Stage I, the fixed input is underutilized, so you can always do better by hiring more labor. In Stage III, marginal product is negative, meaning you're literally reducing output by adding workers. Stage II is where the firm faces the real tradeoff: more output is possible, but each additional unit of labor contributes less than the last.

Note that the boundary between Stage I and Stage II occurs where $AP_L$ is maximized (i.e., where $MP_L = AP_L$ ). The boundary between Stage II and Stage III occurs where $MP_L = 0$ .

Mathematical representation and components, Evolution of the Marketing Orientation | Boundless Marketing

Input optimization: isoquants and isocosts

Isoquants are curves showing all combinations of two inputs (say $L$ and $K$ ) that produce the same level of output. They work like indifference curves but for production. Key properties: they slope downward, they can't cross, and they're convex to the origin (reflecting a diminishing MRTS).

Isocost lines show all input combinations a firm can afford for a given total cost $C$ :

$C = wL + rK$

where $w$ is the wage rate and $r$ is the rental rate of capital. The slope of the isocost line is $-\frac{w}{r}$ .

The firm minimizes cost by choosing the input combination where the isoquant is tangent to the isocost line. At that point:

$MRTS_{L,K} = \frac{MP_L}{MP_K} = \frac{w}{r}$

The marginal rate of technical substitution (MRTS) measures how much capital you can give up when you add one unit of labor, holding output constant. It's the absolute value of the slope of the isoquant. Cost minimization requires that the rate at which you can substitute inputs in production equals the rate at which market prices let you substitute them.

An equivalent way to state this condition: at the optimum, the marginal product per dollar spent is equal across all inputs.

$\frac{MP_L}{w} = \frac{MP_K}{r}$

If this didn't hold, you could reallocate spending from the input with a lower bang-per-buck to the one with a higher bang-per-buck and produce the same output for less cost.

Technology's role in production

Technological progress and productivity

Technological improvement shifts the production function upward: the firm gets more output from the same inputs. Two broad categories:

Process innovations make existing production methods more efficient (e.g., a faster assembly technique).
Product innovations create new or improved goods (e.g., a lighter, stronger building material).

Total factor productivity (TFP) captures the portion of output growth that can't be explained by simply using more labor or capital. In a Cobb-Douglas framework $Q = A \cdot L^\alpha K^\beta$ , the parameter $A$ represents TFP. When $A$ rises, the firm is genuinely getting better at turning inputs into output, not just throwing more resources at the problem.

Technology can also change the substitutability between inputs. Automation, for instance, raises the marginal product of capital relative to labor, which shifts the cost-minimizing input mix toward more capital and less labor.

Innovation types and industry impacts

Incremental innovations improve existing products or processes gradually (annual software updates, small manufacturing refinements).
Disruptive innovations fundamentally reshape markets or create entirely new ones (smartphones replacing standalone cameras, GPS devices, and music players).

Technology also affects how long the "short run" lasts in practice. In cloud computing, you can scale server capacity in minutes, making the short run very short. In heavy industry, retooling a factory takes years. Learning-by-doing, where workers become more productive through accumulated experience, is an additional channel through which production functions shift upward over time without any change in measured input quantities.