Category: Effects
Type: A judgment of a money amount by the currency units rather than by what those units buy
Origin: Irving Fisher, The Money Illusion (Adelphi Company, New York, 1928); the survey problems are Eldar Shafir, Peter Diamond, and Amos Tversky, Quarterly Journal of Economics (May 1997); the laboratory prices are Ernst Fehr and Jean-Robert Tyran, IEW Working Paper No. 45 (revised May 2000)
Also known as: A bias toward the nominal evaluation
Type: A judgment of a money amount by the currency units rather than by what those units buy
Origin: Irving Fisher, The Money Illusion (Adelphi Company, New York, 1928); the survey problems are Eldar Shafir, Peter Diamond, and Amos Tversky, Quarterly Journal of Economics (May 1997); the laboratory prices are Ernst Fehr and Jean-Robert Tyran, IEW Working Paper No. 45 (revised May 2000)
Also known as: A bias toward the nominal evaluation
Quick Answer — Money Illusion is a tendency to judge a sum by the number of currency units rather than by what those units buy. Irving Fisher gave the pattern its name in 1928. In one question from a 1997 survey, 44 of 69 people called Barbara happier. Her raise was 5 percent during 4 percent inflation. Ann’s raise was 2 percent, and prices had not risen. A later pricing game at the University of Zürich is a different object. It is a money-supply cut inside a laboratory, and it is not this raise.
What is Money Illusion?
Money Illusion is a tendency to judge a sum by the number of currency units rather than by what those units buy.Failure to perceive that the dollar, or any other unit of money, expands or shrinks in value.Peter Howitt and Don Patinkin quote that sentence from page 4 of Irving Fisher’s 1928 book. The number on a wage, a price, or a contract is the nominal amount. What that number buys is the real amount. The two come apart when the price level moves. A 5 percent raise during 4 percent inflation adds more currency units than a 2 percent raise with no inflation, and it adds less purchasing power. A currency unit is a step whose length changes. Counting more steps does not say the trip was longer if each step got shorter. Eldar Shafir, Peter Diamond, and Amos Tversky, writing in 1997, describe the pattern as a tendency to think in nominal rather than real monetary values. Their proposal is that people keep both representations, and that the nominal one biases the judgment. The bias is easier to see when the question is about happiness or a job than when it asks who is better off in economic terms. Framing Effect is the neighboring fact that the wording of a question changes the choice. Mental Accounting splits money by the label on its source or its use. Denomination Effect is about the face value of the note in hand. Neither of those two pages is the price level between a number and a basket.
Money Illusion in 3 Depths
- Beginner: A bigger number on a paycheck can be a smaller basket. Write the basket before you celebrate the number.
- Practitioner: Put last year’s prices beside this year’s wage, rent, and the thing you were about to buy. Compare the baskets, then decide.
- Advanced: The 44 of 69 is one happiness question in a 1997 survey. The Zürich prices are a four-person game after a cut in the money supply. Neither figure is next year’s raise, or a central bank’s forecast.
Origin
The name is Irving Fisher’s. The Money Illusion was published by the Adelphi Company in New York in 1928. Howitt and Patinkin, in The New Palgrave Dictionary of Economics, quote his page 4 for the sentence above, and they treat the term as his. This page did not re-read the 1928 book. Later textbook accounts also place a version of the pattern in John Maynard Keynes’s 1936 discussion of resistance to cuts in money wages. This page does not carry a wage count from that book. The survey problems are Eldar Shafir, Peter Diamond, and Amos Tversky, “Money Illusion,” The Quarterly Journal of Economics, volume 112, number 2 (May 1997), pages 341–374. The issue is marked in memory of Amos Tversky (1937–1996). Their abstract says the term refers to a tendency to think in terms of nominal rather than real monetary values. The counts below are the tables of that study printed by Ignazio Ziano, Jie Li, Shue Man Tsun, Hoi Ching Lei, Anvita Anil Kamath, Bo Ley Cheng, and Gilad Feldman, Journal of Economic Psychology, volume 83 (March 2021), article 102349. The 1997 PDF was not re-read. Ziano and his coauthors say the 1997 respondents were Princeton undergraduates, passengers at Newark airport, and visitors to shopping malls in New Jersey. They also say the 1997 paper reports a sample size for each problem and does not say which place produced which finding. The 150, the 69, and the 139 stay attached to their questions. The laboratory prices are Ernst Fehr and Jean-Robert Tyran, “Does Money Illusion Matter?,” IEW Working Paper No. 45, Institute for Empirical Research in Economics, University of Zurich. The copy read here is the revised version of May 2000, and the text dates that version to December 1999. The same study was published in the American Economic Review, volume 91, number 5 (December 2001), pages 1239–1262. The numbers on this page are the working paper. Fehr’s byline there is the University of Zürich. Tyran’s is the University of St. Gallen.Key Points
Four readings keep each percentage on the question that produced it. A survey rank is not a sale. A contract checked on a page is not a signed order.1
Separate the currency units from the basket
The 1997 earnings problem introduces Ann and Barbara. They graduated from the same college a year apart and took similar jobs with publishing firms. Each started at 600. In Barbara’s first year there was 4 percent inflation, and in her second year she received a 5 percent raise, $1,500. Asked who was doing better in economic terms as they entered that second year, 106 of 150 people said Ann, and 44 said Barbara. That is 71 percent and 29 percent. The question that asks who is better off economically is the question on which a majority used the real comparison.
2
Keep happiness and quitting on their own questions
A second group was asked who was happier. Twenty-five of 69 said Ann, and 44 of 69 said Barbara. That is 36 percent and 64 percent. A third group was told that each woman had a job offer from another firm, and was asked who was more likely to leave. Ninety of 139 said Ann, and 49 said Barbara. That is 65 percent and 35 percent. The three questions were asked of different people. The 150, the 69, and the 139 are not one class, and they are not three answers from the same person. Ziano and his coauthors recalculated the tests against a 50 percent split. Each of the three is marked below the 0.01 level. Their association test across the three questions has a Cramér’s V of 0.26, with an interval from 0.17 to 0.37.
3
Rank the sale by what the money buys
A second problem gave Adam, Ben, and Carl an inheritance of 154,000, which the text calls 23 percent less than he paid, and a 2 percent profit in real terms. Ben sold for 246,000, called 23 percent more than he paid, and a 2 percent loss in real terms. The real order the authors state is Adam, then Ben, then Carl. In the 1997 ranks, Carl was placed first by 207 of 440 people, 47 percent. Ben was placed second by 315 of 427, 74 percent. Adam was placed third by 228 of 426, 54 percent. The three denominators are not the same. The 47 percent is a survey rank. It is not the share of houses sold in a year of inflation.
4
Read the contract in both units
A fourth problem asks the reader to sign, in 1991, as the head of a corporate division in Singapore, for computer systems to be delivered in January 1993. The systems are priced at 1,200. The other sells at the 1993 price, so the real profit stays essentially the same. When the text spelled out the real gains and losses, 38 of 47 chose the contract that keeps real profit steady. That is 81 percent. When the text highlighted the nominal amounts, 29 of 49 chose that same real-steady contract. That is 59 percent. When the text was neutral, 23 of 43 chose it. That is 54 percent. These are three groups. Ziano and his coauthors mark the real-wording test below 0.001, the nominal-wording test at 0.05, and the neutral test at 0.11. Their Cramér’s V across the three wordings is 0.25, with an interval from 0.13 to 0.42.
Applications
Use the basket, the question, and the wording. Keep each count inside the group that produced it.Write the basket beside the raise
Before you treat a percent raise as a gain, list rent, food, and transit at last year’s prices and at this year’s. Ann’s 2 percent with no inflation and Barbara’s 5 percent with 4 percent inflation are the survey pair. The list is the comparison. The 106 of 150 is the share who picked Ann when the question said “economic terms.”
Ask the economic question in its own words
A household or a team can feel pleased by a larger percent and still be able to say who can buy more. The 1997 design asked those as separate questions, of separate people. The 64 percent who called Barbara happier are 44 of 69. They are not a correction of the 106 of 150, and they are not a vote at your workplace.
Rank a sale by the goods the money buys
A price that is up in currency units can be down in goods if the price level rose by more. Adam, Ben, and Carl are a written inheritance of $200,000 and a written sale. Use their real order, Adam then Ben then Carl, as the check. The 207 of 440 who put Carl first are that survey. They are not a closing on your street.
Put both columns on the contract
A fixed money price and a price that moves with the price level are the two Singapore contracts. Write what each pays if prices rise by 10 percent and if they rise by 30 percent. The 38 of 47 is the real-wording cell, and it is 47 people. It is a prompt to fill both columns. It is not the share to expect on the next agreement.
Case Study
Ernst Fehr and Jean-Robert Tyran asked whether a small tendency to read nominal payoffs as real ones could slow a price cut that everyone had been told was coming. The setting is a computerized laboratory. The participants in the negative-shock design are 130 undergraduates from different disciplines at the University of Zürich. They were paid a show-up fee of CHF 15, which the paper converts as 28, and that average includes the show-up fee. A session lasted about 90 minutes. Each person is a firm in a group of four and chooses an integer price from 1 to 30. The real payoff depends on one’s own price and on the average price of the others. The best price rises when that average rises, which the authors call strategic complementarity. The real payoffs are the same in every condition. The conditions differ in whether the payoff table is printed in nominal units or in real units, and in whether the other three firms are people or a computer. Twenty-two people faced the real table and a computer. Twenty-four faced the nominal table and a computer. Ten groups of four, 40 people, faced the real table and human opponents. Eleven groups of four, 44 people, faced the nominal table and human opponents. Those four counts are the 130. Before the shock the money supply is 42, and the average equilibrium price is 18. The shock cuts the money supply to 14, one third of 42, and the average equilibrium price falls to 6. Subjects were told the shock was coming, they kept the old tables, and they received the new ones. In the human conditions, the phase before the shock and the phase after it each run for 20 periods. With the real table and a computer, all 22 people adjusted at once to the new equilibrium. The average price sits on 18 before the shock and on 6 after it, except in a few periods. After the second period of the new phase, all of them are on the equilibrium in most periods. With the nominal table and a computer, 80 percent of the 24 adjusted fully at once. In the first period after the shock the average price is 2.0 units above 6. The paper’s sentence is that this is a small amount of money illusion at the individual level. The 80 percent is not a head count. Twenty-four times 0.80 is not an integer the paper prints. With human opponents the nominal table changes the path. In the last period before the shock, 93 percent of the real-table group and 80 percent of the nominal-table group played the equilibrium, and average prices were close to 18. In the first period after the shock, the real-table average is 3.1 units above 6. Thirty-five percent played the new equilibrium, 23 percent were one or two units above it, and 42 percent were more than two units above it. The nominal-table average is 7.1 units above 6. The paper says 11.5 percent played the new equilibrium, and 73 percent were three or more units above it. A regression rejects equality with 6 for two periods on the real table and for twelve periods on the nominal table, at the 5 percent level. The second period prints four averages: 7.0 on the real table with a computer, 7.4 on the nominal table with a computer, 7.7 on the real table with people, and 12.9 on the nominal table with people. The nominal-minus-real gap is 0.4 units with a computer and 5.2 units with people. The paper calls the gap with people 13 times the gap with a computer in that period. During the first ten periods after the shock, 84 percent of the subjects in the real human treatment and 80 percent in the nominal human treatment chose the exact best reply to the price they expected from the others. The paper’s phrase is a percent of subjects across those periods. It is not a head count. The stickiness the authors emphasize is in the expectation. Real income is measured against the equilibrium payoff. In the first period after the shock the average shortfall is 52 percent with the real table and human opponents, and 65 percent with the nominal table and human opponents. Over the first ten periods, groups on the nominal table lose 26 percent of the potential payoff. Groups on the real table lose slightly less than 14 percent. A second design raises the money supply instead of cutting it. It is another 96 undergraduates, 48 with the real table and 48 with the nominal table, 12 groups of four in each. They are not inside the 130, and this design has no computer opponents. Each phase lasts 15 periods. The pre-shock equilibrium prices of the two firm types are 11 and 14, so the group average is 12.5. After the shock those prices are 22 and 28, and the average is 25. In the first period after the shock the nominal-table average is 20.5, and by the fourth period it is almost exactly 25. The real-table average jumps to 22.5, which is 2.5 units below 25. Equality with 25 is rejected for three periods on the nominal table and for two periods on the real table. After the cut, the nominal table had been out for twelve periods. The same nominal framing adjusts faster when the move is a rise in prices. After the experiment, people could say whether they believed others would read a high nominal payoff as a high real payoff. Thirty percent of the subjects in the nominal human treatment agreed strongly or totally, and a further 25 percent agreed weakly. The paper does not print the denominator of those two shares. The lesson is a four-person pricing game in Zürich, a cut from 42 to 14 that moved the equilibrium average from 18 to 6, and a nominal table that left the human groups 7.1 units above 6 in the first new period. It is not Ann’s raise, and it is not a shop’s prices.Boundaries and Failure Modes
Money Illusion is the wrong name in three retellings. The happiness question is treated as the economic one. The laboratory average is treated as a wage. A person who can define inflation is treated as immune. The first boundary is the questionnaire. Problem 1’s three questions are three groups. The places of recruitment are not tied to a problem, and the paper, as Ziano and his coauthors read it, prints no age and no gender. The contract cells are 47, 49, and 43. The neutral cell’s test sits at 0.11. A shopping problem in the same 1997 paper told readers that, within six months, benefits, salaries, and all prices in the United States had risen by 25 percent, so they earned and spent 25 percent more. A leather armchair they had planned to buy or to sell went from 500, which is the same 25 percent. Ziano and his coauthors report the original buy answers, on 362 responses, as 7 percent more likely to buy, 55 percent the same, and 38 percent less likely. The sell answers, also 362, are 43 percent more likely to sell, 42 percent the same, and 15 percent less likely. They mark those integer counts as recalculated, and the reprint pools a dollar wording with a percent wording. A person who is less willing to buy that chair has not been made poorer by the chair. Salaries moved with the chair. The second boundary is the laboratory. The prices are integers from 1 to 30 in a game of four. The money supply was assigned, and the subjects were paid. The 7.1 units and the 65 percent belong to the first period of the nominal human treatment after the cut from 42 to 14. The positive-shock 96 are a second design, with different equilibrium prices, so that the count of dominated strategies would be roughly comparable. Adding 96 to 130 invents a sample the paper keeps apart. With a computer, the nominal table’s first-period gap is 2.0 units, and the long gap appears when the other firms are people whose best price rises with the average. A market without that complementarity is a different study. In 2014 Fehr and Tyran replied, in the American Economic Review volume 104, number 3, pages 1063–1071, to a 2013 comment by Luba Petersen and Abel Winn. The reply says money illusion still matters for adjustment after a fully anticipated negative shock, and that the comment misreads both data sets. This page does not carry the 2013 counts. A later Fehr and Tyran paper on coordination failure is not this design either. The third boundary is the claim that a correct economic answer, or a stated knowledge of inflation, removes the pattern. On the economic question, 106 of 150 picked Ann. On the happiness question, asked of other people, 44 of 69 picked Barbara. Ziano and his coauthors also ran the four problems on Mechanical Turk in a sample of 604. On the earnings problem their shares were 111 of 200 for Ann on the economic question, 70 of 203 for Ann on happiness, and 138 of 201 for Ann on leaving. Carl was ranked first by 421 of 604, 70 percent. Forty-eight of the 604 said they knew the inflation rate. Knowing it, tracking it, or the estimate itself was not associated with the choices in the tests they emphasize. Two regressions, among many, were significant at 0.03 and 0.024, and the authors say those can be chance. The 2021 shares are a second sample. They are not the 1997 cells, and they are not a curriculum result. Loss Aversion is a greater weight on a loss than on a matched gain. The Zürich reluctance to cut a nominal price sits inside this laboratory. It is not the 1979 gamble counts. Anchoring Effect is a starting number that pulls a later judgment. The pre-shock average of 18 is the equilibrium this design reached. It is not that study.Common Misconceptions
Three retellings travel with the name. The percent on the raise is treated as the living standard. A person who can say what inflation is gets marked immune. A label on a dollar, or a slow sale of a loser, is given this page’s pairs.A larger percent raise is a larger living standard
A larger percent raise is a larger living standard
Ann’s raise is 2 percent, and prices did not rise. Barbara’s raise is 5 percent, and prices rose 4 percent. The currency units favor Barbara. The basket favors Ann. When the question said “economic terms,” 106 of 150 picked Ann. When a different group was asked who was happier, 44 of 69 picked Barbara. The percent on the letter is one column. The basket is the other.
Knowing the inflation rate removes the pattern
Knowing the inflation rate removes the pattern
The economic question and the happiness question already split inside the 1997 design. Ziano and his coauthors asked a further sample whether they knew the inflation rate. Forty-eight of 604 said they did. That report was not associated with the choices in the tests the authors emphasize. A definition of inflation, written in the margin, is not the same act as comparing the two baskets before the choice.
This is a label on a dollar, or a slow sale of a loser
This is a label on a dollar, or a slow sale of a loser
Mental Accounting is a label by source or by use, such as a ticket paid for yesterday. Denomination Effect is the face value of the note. Disposition Effect is a higher rate of selling positions above their purchase price than positions below it. Money Illusion is the price level between the printed number and the basket. The Zürich game shows a slower cut in a nominal price inside that laboratory. It does not move those other pages’ counts onto a wage.
Related Concepts
Money Illusion sits beside the wording of a question, beside a label on a sum of money, beside the face value of a note, beside the extra weight of a loss, and beside a number that sticks.Framing Effect
The same facts, worded differently, draw different choices. The Singapore contract moved from 38 of 47 to 29 of 49 when the wording left the real column and stressed the nominal one.
Mental Accounting
A label by source or by use, judged inside that label. The split on this page is the price level, between the currency units and the basket.
Denomination Effect
Spending that depends on the face value of the note in hand. A 5 percent raise and a 2 percent raise are a different comparison.
Loss Aversion
A greater weight on a loss than on a matched gain. The slower nominal price cut in the Zürich game is this page’s laboratory, not the 1979 gamble counts.
Anchoring Effect
A starting number that pulls a later judgment. The average of 18 is the pre-shock equilibrium of the Zürich design. It is that design’s price, not an anchor experiment.