A ranking is hard to calculate with. A utility function turns it into a number, and a number can be differentiated. This lesson says exactly what that number does and does not mean, and builds the one slope the whole subject runs on: the marginal rate of substitution.
A utility function u(x, y) represents her preferences if it gives a bigger number to the bundle she likes better: u(A) > u(B) exactly when A ≻ B, and u(A) = u(B) exactly when A ∼ B.
It is a scoring scheme and nothing more. If two bundles score 12 and 6, the only fact is that the first is better. It does not mean she likes it twice as much. Every ranking from Lesson 1 that is complete, transitive and continuous can be scored this way. (Lexicographic preferences cannot: Lesson 3.)
Many different functions score the same ranking. If u represents her preferences and f is any increasing function, then f(u) ranks the bundles in the same order, so it represents the same preferences. Doubling, squaring (for positive u), taking a log (for positive u) and adding a constant are all examples. Such a change is called a monotonic transformation.
Modern theory is therefore ordinal: only the order of the scores means anything. An older cardinal view treats utility as measurable, so that differences and ratios of utility have meaning. The ordinal view asks less, and it is enough for everything in demand theory.
The test for whether a function is a valid replacement is whether it preserves the order of every pair of bundles. Not every function of x and y does:
| If u = xy represents her, then | Same preferences? | Why |
|---|---|---|
| x²y², or xy + (xy)³, or √(xy) | Yes | An increasing function of xy (JAM 2023: u + u³ and u^(1/n)) |
| −xy | No | Decreasing: it reverses every ranking |
| xy + x + y | No | Adds a term that depends on the bundle, not just on u, so two bundles with equal xy can now differ (JAM 2023) |
| √(x + y) against 2x + 3y | Yes, both perfect substitutes | √(x+y) is an increasing function of x + y, and 2x + 3y has the same straight-line curves (JNU 2019) |
| min{x, y} + max{x, y} | Equals x + y | The smaller plus the larger is the sum (JNU 2012) |
The indifference curves, the MRS, the ranking of every pair, and the best affordable bundle all stay the same. The utility numbers, the marginal utilities and the size of utility gaps all change. So any claim that depends on those cannot be a claim about her tastes.
Her tastes are u = xy. Choose a transformation and move bundle A. The curves do not move at all. The right-hand column of the table is the label she would read off the new function; the last column says whether it matches the old one. Watch the marginal utility of x: it can rise, fall or stay flat as you change the labelling, while the MRS stays put.
Her tastes are u = xy. Pick a transformation and move bundle A. The curves are drawn once, because they do not depend on how you label them; watch which rows of the table change.
| What you read off | u = xy | u² | Same? |
|---|---|---|---|
| Utility of A | 8 | 64 | no |
| Utility of B | 10 | 100 | no |
| Which is preferred | B | B | yes |
| MRS at A | 2 | 2 | yes |
| Marginal utility of x at A | 4 | 64 | no |
| Utility gap, A minus B | −2 | −36 | no |
With u = xy, the marginal utility of x stays the same when x goes up by one. With u² it rises. Diminishing marginal utility is a property of the labelling, not of her tastes.
Move a little along an indifference curve, changing x by dx and y by dy. Utility does not change, so the gain from x and the loss from y cancel. The marginal utility of a good is the extra utility from one more unit of it, written MUx = ∂u/∂x, so
It is unchanged by any monotonic transformation, because f′(u) multiplies both MUx and MUy and cancels. That is why the MRS, unlike marginal utility, is a fact about tastes. Some standard values:
| Utility function | MRS = MUx/MUy |
|---|---|
| (perfect substitutes; CUET 2024 asks 2x + 3y) | , the same everywhere |
| (quasi-linear) | , depends on x only |
| (CES) |
Read the wording carefully. “An MRS of 2 eggs for 1 sandwich” (Hyderabad 2010) says she will give up 2 eggs for 1 sandwich. So sandwiches are worth twice as much at the margin: MUsandwich/MUegg = 2, and the ratio the question wants, MUegg/MUsandwich, is ½. At any point on an indifference curve, the absolute value of its slope is the MRS (Hyderabad 2018).
Ordinary tastes show a diminishing MRS: the more x she holds relative to y, the less y she will give up for another unit of x. As she moves down a curve, the MRS falls, so the curve gets flatter. A curve that gets flatter as x rises is convex to the origin (bowed in towards it). Convexity is the formal name for a taste for variety: she prefers a balanced mixture of two equally good bundles to either extreme (Hyderabad 2012: she gives up successively less of one good for each extra unit of the other).
Convexity can fail. “I would rather spend all my time in the country or all in the town than divide myself between the two” prefers the extremes, which is the opposite of convex (Hyderabad 2010). A utility function such as √(x² + y²) has curves that are quarter-circles around the origin: the midpoint of two equally good bundles is worse, and the curves are concave to the origin (Hyderabad 2021).
The two are different. The JNU 2019 table lists bundles worth 1, 4 and 16 utils. They fit u = x²y², so the marginal utility of x, 2xy², rises with x. Yet the MRS, y/x, falls as x rises along a curve, and the curves are convex. Convexity needs a diminishing MRS only.
The older, cardinal approach asks the same question with money. The extra utility from a rupee spent on x is MUx/Px. If it is larger than MUy/Py, a rupee moved from y to x raises her utility. She stops moving rupees when the two are equal. That is the equi-marginal principle:
Rearranged, it says MUx/MUy = Px/Py: the MRS equals the price ratio. So the cardinal rule and the indifference-curve rule are the same condition (Lesson 5 draws it). Questions use it as a ratio: if MUx = 50, MUy = 100 and Py = Rs 20, then 100/20 = 5 per rupee, so Px = 50/5 = Rs 10 (CUET 2019).
Definitions and one-step calculations: the shape of a curve, a budget line, an optimum for Cobb–Douglas tastes. Most CUET, UGC NET, Hyderabad and early JNU questions sit here.
Statements to judge, matching lists, and demands to compute with unusual tastes, up to JAM, GATE and JNU level.
This topic’s hardest questions, from the DSE entrance, are worked through in Lesson 17: Orderings, representation and revealed choice.
These are questions from the TestStop bank. A free account answers 10 a day, each with its worked solution; All-Access opens every set with no daily limit.