- Indifference curves and budget constraints
- Consumer equilibrium — maximising utility formally
- Expected utility theory · coming soon
Water is priceless. Why is it so cheap?
You cannot live for a week without water. You can live your whole life without a diamond. Yet a litre of tap water costs almost nothing, and a small diamond costs more than many people earn in a month.
Adam Smith noticed this in 1776, in The Wealth of Nations. Things with the greatest "value in use", he wrote, often have little "value in exchange", and the other way round. He could describe the puzzle. He could not fully explain it. Economists call it the paradox of value, or the water–diamond paradox.
Here is a smaller version of the puzzle. You have not eaten since breakfast, and a plate of identical dumplings arrives. The first one is wonderful. By the eighth, you would rather stop. Only the number you had already eaten changed.
The lecture on the law of demand stated one idea without giving it a name: each extra cup of coffee is worth less to you than the one before. This lecture names that idea, gives it rules, and uses it to solve the water–diamond paradox. The difficulty is not the arithmetic, which is simple. The difficulty is one word, marginal. Once that word is clear, the paradox disappears.
Utility: a name for the benefit of consuming
Economists use the word utility for the satisfaction, or benefit, a person gets from consuming a good or service. To compare choices, it helps to give utility numbers. The unit is an imaginary one called a util.
Nobody can measure utils with a machine, so treat the numbers in this lecture as a model. Using utility numbers like amounts is called cardinal utility. A theory that needs only rankings ("I prefer this to that") is ordinal. The lecture on indifference curves shows that rankings are enough.
Total utility and marginal utility
Two quantities do all the work.
- Total utility (TU) is the whole satisfaction from all the units consumed.
- Marginal utility (MU) is the extra satisfaction from one more unit. "Marginal" means "at the edge": the next unit, not all of them.
When Q goes up one unit at a time, ΔQ = 1, so MU is just the rise in TU. Here is the dumpling meal, as an invented example.
| Dumplings eaten | TU (utils) | MU of that dumpling (utils) |
|---|---|---|
| 1 | 30 | 30 |
| 2 | 54 | 24 |
| 3 | 72 | 18 |
| 4 | 84 | 12 |
| 5 | 90 | 6 |
| 6 | 92 | 2 |
| 7 | 92 | 0 |
| 8 | 88 | −4 |
Each dumpling adds less than the one before. TU keeps rising until the 6th dumpling, stays level at the 7th, and falls at the 8th. TU is simply a running total of the MU column: 30, then 30 + 24 = 54, and so on.
The two graphs tell one story. While the bars are positive, TU climbs. A bar of zero means TU stops climbing. A negative bar means the unit makes the person worse off, so TU falls.
Other things being equal, as a person consumes more units of a good in a given period, the marginal utility of each extra unit eventually falls.
Each phrase in the law does a job.
- Other things being equal (ceteris paribus). Tastes, health and everything else are held fixed.
- In a given period. Eight dumplings in one meal is very different from eight dumplings in a month.
- Eventually. MU can rise for the first few units. The first two lessons of a new language may give you very little, because you still cannot say anything useful. Around the tenth lesson, you start to hold real conversations, so that lesson is worth more. But sooner or later, extra lessons add less and less.
Mohammed's marginal utility from the 3rd, 4th and 5th dumplings is falling. Is he getting less satisfaction from his meal as he eats them?
Show the answer
No. His total satisfaction is still rising: 72, then 84, then 90 utils. What is falling is the extra satisfaction each new dumpling adds: 18, then 12, then 6.
Falling MU means TU is rising more slowly. TU falls only when MU becomes negative. Mixing up "less extra" with "less in total" is the most common mistake in this topic.
Why the next unit is worth less
The law is not just an observation. There is a reason behind it. People use each unit of a good for its most important purpose first.
Think of water. Your first litres each day go to drinking. Without them you would be ill. The next litres go to cooking, then to washing, then to cleaning the floor, and then to the plants. Each extra litre goes to a less important use than the one before. So the marginal utility of water falls as you have more of it. The litres are identical. Their uses are not.
From marginal utility to the demand curve
Now measure utility in money instead of utils: how much would a person pay for one more unit? This is the willingness to pay that you met in the lecture on consumer surplus. Here it is the marginal utility of each extra unit, expressed in dollars.
A sensible buyer keeps buying while the next unit is worth at least its price to her. Use the builder below to see what this rule produces.
Zara's marginal utility for bubble tea, in dollars per cup, is $6, $5, $4, $3, $2 and $1 for her 1st to 6th cup in a week (invented). Change the price. She buys every cup worth at least the price. Each price you visit adds a point to her demand curve below.
Zara buys every cup worth at least the price. Her demand schedule follows straight from her marginal utility.
| Price of a cup ($) | Cups bought per week | Consumer surplus ($) |
|---|---|---|
| 6.50 | 0 | 0 |
| 5.50 | 1 | 0.50 |
| 4.50 | 2 | 2.00 |
| 3.50 | 3 | 4.50 |
| 2.50 | 4 | 8.00 |
| 1.50 | 5 | 12.50 |
As the price falls, more cups pass the test "worth at least the price". So the quantity bought rises. That is the law of demand, and now you can see where it comes from. A demand curve is the marginal utility of each unit, measured in money and read off at each price. The economist Alfred Marshall built demand curves this way in 1890. He assumed that the value of a dollar itself stays about the same while the person buys more of one cheap good.
The paradox of value, solved
Now return to Adam Smith's puzzle. The price of a good depends on the value of the last unit people buy, not on the value of all the units together.
Water is plentiful. People already have enough for drinking, cooking and washing. The last litre a household uses goes to its least important use, perhaps washing a floor. Its marginal utility is tiny, so nobody pays much for it. But the total utility of water is enormous. Without it, we die.
Diamonds are scarce, so the marginal utility of one more is still high. Total utility from all diamonds is far smaller than from all water. But price follows the margin.
That is it. That is the whole idea. Smith compared total value with price. The price is set at the margin. Around 1871–1874, three economists reached this answer separately: William Stanley Jevons in Britain, Carl Menger in Austria and Léon Walras in Lausanne, Switzerland. Historians call it the marginal revolution. Hermann Gossen had stated the law of diminishing marginal utility earlier, in 1854, but few people read his book.
Spending a budget on two goods
Real buyers choose between many goods with limited money. The question becomes: where should the next dollar go?
The answer comes from comparing marginal utility per dollar, MU ÷ P, for each good. A dollar spent on X buys MUX ÷ PX utils at the margin. A dollar spent on Y buys MUY ÷ PY. Spend each dollar where it buys the most.
Suppose a dollar on X buys more than a dollar on Y. Then moving a dollar from Y to X raises total utility. As you buy more X, its MU falls. As you buy less Y, its MU rises. The gap closes. When the two are equal, no move can help. This is the equimarginal principle.
With a fixed budget, total utility is as large as possible when the marginal utility per dollar is the same for every good bought:
Try the principle by hand, with $10.
Good X costs $2 and good Y costs $5 (invented). Buy one unit at a time. Before each choice, compare the utils per dollar of the next X and the next Y in the table.
The table gives each unit's marginal utility and utils per dollar. Spending each dollar where it buys the most:
| Unit | MU of X | Per $ (X, $2) | MU of Y | Per $ (Y, $5) |
|---|---|---|---|---|
| 1st | 20 | 10 | 45 | 9 |
| 2nd | 16 | 8 | 35 | 7 |
| 3rd | 12 | 6 | 30 | 6 |
| 4th | 8 | 4 | 24 | 4.8 |
- 1st X (10 per $) beats 1st Y (9 per $): buy X. $8 left.
- 1st Y (9 per $) beats 2nd X (8 per $): buy Y. $3 left.
- 2nd X (8 per $): buy X. $1 left, too little for anything more.
- Result: 2 X and 1 Y, total utility 36 + 45 = 81 utils. No other affordable bundle gives more.
Notice the end. With 2 X and 1 Y, the last dollar on X bought 8 utils and the last dollar on Y bought 9. They are not equal, because units cannot be split. With goods bought in any amount, you could move money until the ratios matched exactly.
Worked examples
At a fair, Kwame's marginal utility in money for each extra ride is $8, $6, $5, $3 and $1 (invented). A ride costs $4.
Rule. Buy each ride worth at least its price.
Decide. Rides 1, 2 and 3 are worth $8, $6 and $5, all at least $4. Ride 4 is worth $3, less than $4. So he buys 3 rides.
Consumer surplus. (8 − 4) + (6 − 4) + (5 − 4) = $7.
Price cut. If a ride costs $2, ride 4 ($3) also passes, and ride 5 ($1) does not. He buys 4 rides, and his surplus rises to 6 + 4 + 3 + 1 = $14.
Check. A lower price gives a larger quantity, as the law of demand says. ✓
Use the explorer's goods: X costs $2, Y costs $5, and the budget is $10. Total utility is TU(X) + TU(Y), with TU(X) = 20, 36, 48, 56 and TU(Y) = 45, 80, 110, 134 for 1 to 4 units.
List the affordable bundles that have no room for one more unit from the table. Any other bundle could simply add a unit and do better.
| Bundle (X, Y) | Cost ($) | Total utility |
|---|---|---|
| (0, 2) | 10 | 80 |
| (2, 1) | 9 | 36 + 45 = 81 |
| (4, 0) | 8 | 56 |
(4, 0) leaves $2, but the table stops at 4 units of X, so a 5th X cannot be judged.
Answer. (2, 1) gives 81 utils, the most. It even leaves $1 unspent and still beats (0, 2).
Check with the margin. At (2, 1), the last X bought 16 ÷ 2 = 8 utils per dollar and the last Y bought 45 ÷ 5 = 9. The next X would give 12 ÷ 2 = 6 per dollar, and it costs $2, which you no longer have. ✓
Each month, Ananya spends $200 on food and $50 on books. For her, the last dollar on food adds 3 utils and the last dollar on books adds 8 utils.
Compare at the margin. 8 is more than 3. Moving one dollar from food to books gains about 8 utils and loses about 3, a net gain of about 5.
What happens next. As she buys more books, the MU of books falls. As she buys less food, the MU of food rises. She should keep moving money until the last dollar on each gives the same.
"But food matters more!" In total, it does. But her last food dollar perhaps buys a second snack. This is the water–diamond idea again, inside one person's budget.
A real case: is an extra dollar worth less to a richer person?
The argument. Many arguments for taxes that take a larger share from higher incomes, and for cash transfers to poorer people, rest on one claim. Money itself has diminishing marginal utility. If that is true, $1000 taken from a rich person costs them little, and $1000 given to a poor person helps a lot.
The prediction. In 1738, the mathematician Daniel Bernoulli suggested a precise form. The extra satisfaction from a small gain is inversely proportional to the wealth a person already has. That makes utility rise with the logarithm of wealth. In this model, every doubling of income adds the same amount of utility. So $10 000 more matters as much to someone on $10 000 as $160 000 more does to someone on $160 000.
What happened. Psychologists have tested this using people's own reports of their well-being.
- Daniel Kahneman and Angus Deaton (2010, PNAS) analysed more than 450 000 survey responses from US residents in 2008–2009. People's ratings of their lives rose steadily with the logarithm of income. Their day-to-day emotions also rose with log income, but showed no further progress beyond about $75 000 a year.
- Matthew Killingsworth (2021, PNAS) used 1 725 994 real-time reports from 33 391 employed US adults on their phones. Both measures rose in a straight line with log income, with no plateau.
- In 2023 (PNAS), Killingsworth, Kahneman and Barbara Mellers re-analysed the data together. The flattening appeared only for the least happy group. For most people, happiness kept rising with log income, and for the happiest group it rose even faster.
Verdict. The evidence supports diminishing marginal utility of income in dollar terms. Well-being rises roughly with the logarithm of income, so each extra dollar adds less as income grows. But three cautions keep the policy argument from being automatic. First, a well-being survey is not a direct measure of utility. Second, comparing utility between two people needs a value judgement, a point the economist Lionel Robbins pressed in 1932. Third, as the lecture on taxes showed, taxes also change incentives to work and earn. The evidence supports the claim behind the argument, while the policy decision needs more than this one fact.
Falling MU means total utility rises more slowly. TU falls only when MU is negative. Always say which one you mean: total or marginal.
A buyer compares marginal utility per dollar, not marginal utility alone. In Worked Example 2, the first Y gives 45 utils and the first X only 20, yet the first dollar is better spent on X: 10 utils per dollar against 9.
Utils are a model of one person's choices. "The poor gain 50 utils and the rich lose 5" assumes utility can be compared between people. That comparison is a value judgement, so say so when you use it in an evaluation.
Work each one on paper first. Open a hint only when you are stuck.
-
Lucía's total utility from slices of pizza at one meal is 40, 70, 90, 100, 104, 104 and 98 utils for 1 to 7 slices. Find the marginal utility of each slice. At what point is her total utility largest, and which slice is the first that she would rather not eat?
Hint 1
Each MU is the difference between neighbouring totals. The first one is 40 − 0.
Hint 2
Total utility stops rising where MU reaches zero.
Full answer
MU: 40, 30, 20, 10, 4, 0, −6.
Total utility is largest, 104 utils, at 5 and 6 slices. The 6th slice adds nothing (MU = 0). The 7th slice is the first she would rather not eat: its MU is −6, and it lowers her total to 98.
Check: the MUs add up to 98, the total for 7 slices. ✓
-
Using Worked Example 1, make Kwame's demand schedule for prices of $9, $7, $5.50, $4, $2 and $0.50. Explain in one sentence why it slopes down.
Hint 1
At each price, count the rides worth at least that price: $8, $6, $5, $3, $1.
Hint 2
At $5.50, rides worth $8 and $6 pass, and the ride worth $5 does not.
Full answer
$9: 0 rides. $7: 1. $5.50: 2. $4: 3. $2: 4. $0.50: 5.
It slopes down because each extra ride is worth less to him than the one before, so only a lower price makes the next ride worth buying.
Check: the quantity never falls as the price falls. ✓
-
Wei has $15. Good X costs $3 and good Y costs $2. The marginal utilities of the 1st to 5th units are 30, 24, 18, 12, 6 for X and 20, 16, 12, 8, 4 for Y. Find the bundle that gives the most total utility, and check it with the equimarginal principle.
Hint 1
Work out MU ÷ P for every unit: X gives 10, 8, 6, 4, 2 per dollar.
Hint 2
Y gives 10, 8, 6, 4, 2 per dollar too. Buy units in order of utils per dollar until the $15 is spent.
Full answer
In order of utils per dollar: 1st X and 1st Y (10 each, $5 in total), then 2nd X and 2nd Y (8 each, $5 more), then 3rd X and 3rd Y (6 each, $5 more). That spends exactly $15 on 3 X and 3 Y.
Total utility: (30 + 24 + 18) + (20 + 16 + 12) = 72 + 48 = 120 utils.
Check: the last X gives 18 ÷ 3 = 6 per dollar, and the last Y gives 12 ÷ 2 = 6 per dollar. They are equal, and the budget is all spent. ✓ The next best bundle, 2 X and 4 Y, gives only 110.
-
Suppose utility equals the natural logarithm of income, as Bernoulli suggested. Compare the gain in utility from an extra $1000 for a person earning $10 000 and for a person earning $100 000. Then explain why moving $1000 from the second person to the first would raise total utility in this model, and give one reason a government might still hesitate.
Hint 1
The gain is ln(new income) − ln(old income) = ln(new ÷ old).
Hint 2
For the first person, the ratio is 11 000 ÷ 10 000 = 1.1. For the second, it is 1.01.
Full answer
First person: ln 1.1 ≈ 0.0953. Second person: ln 1.01 ≈ 0.00995. The same $1000 is worth about 9.6 times as much to the poorer person.
Moving $1000 from the second to the first: the first gains about 0.0953. The second loses ln(100 000 ÷ 99 000) ≈ 0.0101. Total utility rises by about 0.085.
Reasons to hesitate: adding utilities across two people is a value judgement, not a measurement. A well-being survey is not the same as utility. And higher taxes may change how much people choose to work and earn.
Check: a 10% rise beats a 1% rise, as the log model says it must. ✓
The word "marginal" is one of the most powerful in economics, and this lecture is where it starts. Firms decide output by comparing the cost and revenue of the next unit. Workers weigh the next hour of work against the next hour of rest. Governments ask what the next dollar of spending achieves. The water–diamond paradox taught economists to look at the edge, not the total.
Bernoulli's logarithm returns in the lectures on choice under uncertainty. If each extra dollar matters less, losing $1000 hurts more than winning $1000 helps, so most people dislike risk.
- Total utility is the whole satisfaction from all units consumed, and marginal utility is the extra satisfaction from one more unit.
- The law of diminishing marginal utility says that, other things being equal, the marginal utility of each extra unit eventually falls, because each unit goes to a less important use.
- Total utility rises while marginal utility is positive, is largest where marginal utility reaches zero, and falls when marginal utility is negative.
- A buyer keeps buying while the next unit is worth at least its price, so a demand curve is marginal utility measured in money.
- Price depends on marginal utility, not total utility, which solves the water–diamond paradox.
- With a fixed budget, utility is largest when the marginal utility per dollar is equal across goods, which is the equimarginal principle.