The marginal product of an input is the additional output produced by using one more unit of that input, holding other inputs constant. The law of diminishing marginal returns states that as more of one input is used (while others remain fixed), the marginal product eventually decreases. This reflects the reality that inputs become less productive when combined with fixed amounts of other inputs.
Calculate marginal products from production data. Plot total product and marginal product curves side-by-side to see the relationship. Identify the point at which marginal product becomes negative (where adding more input actually reduces total output).
Your prerequisite, the production function, established that output depends on inputs in some systematic way: Q = f(K, L). Now we zoom in on the margin. If you add one more worker to a factory, how much extra output do you get? That increment is the marginal product of labor (MP_L) — the change in total output from employing one additional unit of labor, holding capital fixed. It is the partial derivative of the production function with respect to that input, or in discrete terms, the extra output from one more unit.
To build intuition, imagine a pizza kitchen with one oven. The first cook runs the whole operation — prep, bake, and box — and is highly productive. The second cook helps substantially, splitting tasks and increasing throughput. By the fifth cook, they are sharing oven space and stepping around each other. The sixth adds less value than the fifth. This is diminishing marginal returns: as you add more of one input while holding others constant, each successive unit of that input adds less to output than the previous one. The workers haven't changed — the kitchen has become crowded relative to the fixed oven, so additional labor is less productive at the margin.
The relationship between total product and marginal product follows a predictable pattern you can read off a graph. When MP_L is above zero and rising, the total product curve is accelerating upward — each new worker adds more than the last. When MP_L is positive but falling (the onset of diminishing returns), total output is still increasing but at a slower rate — the total product curve is flattening. When MP_L falls to zero, total output reaches its peak — adding another worker neither helps nor hurts. If MP_L goes negative, total product actually falls — the kitchen is so overcrowded that adding labor reduces total output. Graphically, marginal product is simply the slope of the total product curve, so every feature of the MP curve can be read from the curvature of the total product curve.
Diminishing returns is a short-run phenomenon — it applies precisely because at least one input is held fixed. In the long run, you could expand the kitchen, buy another oven, or restructure the whole production process. Returns to scale — what happens when you scale all inputs together by the same proportion — is a distinct concept. A firm can have constant returns to scale in the long run (double all inputs, double output) while still exhibiting diminishing returns to labor in the short run (when capital is fixed). This distinction becomes essential when you study isoquants and factor demand: the curvature of isoquants and the shape of cost curves are both determined by how marginal products behave as input ratios change.