Of everything in this series, this is the effect most likely to be costing a Canadian business money right now, and the one least likely to have been noticed, because the loss arrives as an absence rather than as a decision.
Key Takeaway
Money illusion is "people's inclination to think of money without taking inflation sufficiently into account, i.e., in nominal terms rather than in real terms"[1]. The original findings were successfully replicated with N = 604[1]. And the second-order result matters more than the first: many people "do not only seem to suffer from money illusion; they also expect that other people's preferences and decisions are affected by money illusion"[2].
Our Grades For These Claims
Applying the scheme from the first article in this series.
Grade A for the questionnaire findings. A 1997 paper in a leading economics journal, successfully replicated in 2020 with a larger sample, and reproduced in a different country with nearly identical results.
Grade C for the interpretation as money illusion specifically, because a 2019 paper argues the results reflect something else entirely, and we could not obtain its full argument.
Grade C for aggregate market effects, where one paper's results were argued to be an experimental artefact and two other studies found weaker or absent effects.
Our position: the pattern is solid and its explanation is contested, which is a distinction most treatments of this topic skip entirely.
A Note On Method
Everything here is verified to August 2026.
We did not obtain the 1997 paper. Every description of its problems reaches us through papers replicating, criticising or citing it.
Our best sources are a replication paper[1], the authors' own reply to a critique[2], a hosted copy of a 2001 experimental paper[3], and an academic working paper reproducing the study in another country[4].
We found a conflict in the paper's title, appearing as both Money Illusion and On Money Illusion, and a conflict on the date of the founding reference.
All arithmetic is ours.
This article discusses behavioural economics research. It is not pricing, employment, economic or investment advice, and the inflation rates used are invented for illustration rather than forecasts.
The Definition
The concept, stated by a paper that replicated the original.
Shafir, Diamond and Tversky "described money illusion as people's inclination to think of money without taking inflation sufficiently into account, i.e., in nominal terms rather than in real terms"[1].
The paper is Money Illusion, Quarterly Journal of Economics, 112(2), 341–374, 1997[5].
Two sourcing notes, ours.
One source gives the title as On Money Illusion[2] where the others give Money Illusion. We use the shorter form, which is far better attested, and flag the other. This is the thirteenth bibliographic inconsistency recorded in this series.
One source also lists the authors in the order Shafir, Tversky and Diamond[2] rather than Shafir, Diamond and Tversky. We use the better-attested order.
Fisher, A Century Ago
The origin, which is older than most of what this series covers.
Reference lists identify Fisher, I. (1928), The Money Illusion, Toronto: Longmans[6].
We did not obtain it and report the citation only.
One observation, ours. The concept is nearly a century old and was named by an economist rather than a psychologist. That matters for how the dispute later ran: economics has a standing argument that individually costly errors get competed away, and this article's later sections are largely about that argument.
The Equivalence
The arithmetic behind this article's title. Ours, and we checked it before using it.
A 5 percent raise during 12 percent inflation leaves someone 6.25 percent worse off in real terms.
A 7 percent pay cut with no inflation leaves them 7.00 percent worse off.
The two differ by 0.75 percentage points. Precisely, a 5 percent raise at 12 percent inflation is identical to a 6.25 percent cut at zero inflation.
Two observations.
One of those is experienced as a raise and the other as a betrayal, and they are the same event.
And the asymmetry is not symmetric in the useful direction: the version that feels acceptable is the one where the employer did nothing wrong and the employee is worse off anyway, which means nobody has to have a difficult conversation for the loss to occur.
It Replicated
The evidence, which is unusually strong for this series.
A replication reports: "We successfully replicated Problems 1 to 4 of Shafir, Diamond, and Tversky's study (1997) on money illusion (MTurk; N = 604)."[1]
The design detail: 604 participants, being 288 males, 315 females, 1 other, mean age 40.09 with a standard deviation of 11.48; the required sample had been estimated at 166; and "sensitivity analyses indicated that this sample had approximately 99.99% power to detect all the original effects with a two-sided α = 0.05."[7]
One result, quoted: on Problem 2, "Carl was selected as having had the best deal by a majority of participants (421/604, 70%); a majority indicated Ben as the one having made the second best deal (481/604, 80%); and a majority indicated Adam as the one having made the worst deal (456/604, 76%). We concluded that this replication was successful."[7]
Two observations, ours.
The sample was 3.6 times the size required, and the power analysis was reported. That is the practice this series has spent thirty-eight articles asking for.
And the replication authors chose Problems 1 to 4 only, excluding 5 to 7 because those combined money illusion with other phenomena, being mental accounting and fairness concerns[7]. That is a deliberate narrowing, honestly declared, and it means the replication tested the cleanest version rather than the whole paper.
And Across Cultures
Two further replications, in different countries and different decades.
China. A study replicating the landmark survey reports: "We find that money illusion is prevalent in China as well. Respondents in the Chinese sample are often somewhat more likely to base decisions on the real monetary value of economic transactions compared to respondents in the U.S. sample." And: "Respondents in the Chinese sample are often somewhat less prone to money illusion than respondents in the United States."[4]
Brazil. A pre-registered study of 372 Brazilian participants adapting the four conditions reports: "The results found were very similar to the original findings: depending on the terms used (real, nominal, or neutral framing), participants showed varying inclinations towards opting for economically advantageous opportunities."[8]
Two observations, ours.
Three independent replications in three countries, one of them pre-registered, all successful. On the grading scheme this series uses, that is as good as questionnaire evidence gets.
And the Chinese result is interesting for being directional rather than binary: the effect was present but somewhat smaller. A finding that varies in magnitude across populations while persisting is more credible than one that appears identically everywhere.
The Second-Order Finding
The result we think matters most, and it is rarely quoted.
Summarising the original, a later paper states: "Their results suggest that people's preferences as well as their perceptions of the constraints are affected by nominal values. Moreover, many people do not only seem to suffer from money illusion; they also expect that other people's preferences and decisions are affected by money illusion."[2]
Three observations, ours.
That second sentence describes a belief about other people, not a mistake about arithmetic. People expect nominal framing to move others.
Which means someone could understand the real arithmetic perfectly and still act on nominal terms, because they expect their counterparty to respond to nominal terms. That is not an error; it is a correct prediction about a counterparty.
And it is what makes the phenomenon survive sophistication. A firm that knows a flat fee is a real cut may still hold it flat because it expects the client to react to the nominal number, which is the expectation this finding says people hold.
The Strongest Challenge
The objection, stated in the form its own proponents give it.
A discussion paper sets out the case against relevance: "A powerful intuitive argument supports the view that money illusion is largely irrelevant for economics, however: the illusion has detrimental effects on peoples' economic well-being and they thus have a strong incentive to make illusion free decisions. Therefore, people will ultimately make illusion free decisions, implying that money illusion has little or no impact on aggregate outcomes, at least in the long run."[6]
Two observations, ours.
That is a serious argument and it is the standard economic response to any documented individual error: costly mistakes get competed away.
And note that this series has encountered it before in a different form. The eighteenth article recorded a dispute over whether the endowment effect survives market experience. The general question, whether laboratory errors persist in markets, is one of the field's permanent fault lines.
From Questionnaires To Markets
The methodological limit, stated by the people who went and tested it.
Fehr and Tyran write of the original: "While these authors asked subjects hypothetical questions we directly observe subjects' behavior in our experiments. In our view the study of Shafir et al. neatly shows that questionnaires can be a very useful instrument to examine the nature of money illusion at the individual level. It is, however, also clear that it is impossible to examine aggregate effects of individual interactions and adjustment processes with this method."[3]
Three observations, ours.
That is a model of how to criticise a paper. It grants what the method establishes, states precisely what it cannot establish, and then does the study that addresses the gap.
The limit is real and general. A questionnaire can show that individuals respond to nominal framing. It cannot show what happens when many such individuals interact, adjust, and learn.
And it is the same distinction the twenty-seventh article drew between a stated preference and payroll data. Hypothetical response and observed behaviour are different evidence, and the gap between them is where most disputes in this series live.
Nominal Inertia
What the market experiments were looking for.
Fehr and Tyran published Does Money Illusion Matter? in the American Economic Review, 91(5), 1239–1262, in December 2001[3][5].
Their later work extends the argument. A discussion paper states: "It is the purpose of this paper to show that this argument can be seriously misleading because it neglects the strategic repercussions of money illusion. We show experimentally that even if learning in the context of an individual optimization problem does remove individuals' money illusion, there can be" consequences at the aggregate level, with our source truncating[6].
Two observations, ours.
The argument is strategic rather than psychological. Even individuals who have learned to think in real terms may act on nominal terms if they expect others to, which is the second-order finding above arriving in a market setting.
And that would mean the competing-away argument fails not because people cannot learn, but because learning is not enough when your best move depends on what everyone else does.
A Conflict In The Literature
A disagreement we found while sourcing, reported because it cuts against the market evidence.
A 2019 experimental paper on money illusion, financial literacy and numeracy states: "By contrast with previous results, for instance, Fehr and Tyran (2001) find nominal inertia is higher in nominal" conditions, with our source truncating mid-sentence[9].
The same paper reports of its own results: "our evidence does not support this explanation. Indeed, highly financially literate participants do not make more errors in noncongruent than in congruent choices, but they are sensitive to computational difficulties."[9]
Three observations, ours.
The phrase by contrast with previous results signals a disagreement with the 2001 experiments, and our source truncates before we can see what the contrast is. We report the existence of the disagreement and not its content.
The second quotation points at a different explanation entirely: computational difficulty rather than a preference for nominal framing. If people fail these problems because the arithmetic is hard, that is a different phenomenon wearing the same name.
And financial literacy not helping is the uncomfortable part for anyone in this profession. On this result, knowing more finance did not reduce the errors; sensitivity to computational load did.
Where That Leaves It
Our summary of a genuinely mixed position.
Four statements we think the evidence supports.
The questionnaire finding is solid. Three independent replications in three countries, one pre-registered, one with 3.6 times the required sample and reported power.
The interpretation is contested. At least one paper points at computational difficulty rather than nominal preference, and we could not read its full argument.
The aggregate question is separate and harder. The people who ran the market experiments said so themselves, and a later paper disagrees with their results in a way we could not resolve.
And the second-order finding survives all of it, because expecting others to respond to nominal terms is a belief about people rather than a claim about markets.
The Fee Held Flat
The application, and the reason we think this is the costliest item in the series. The inflation figures below are invented for arithmetic and are not Canadian CPI.
A fee held flat is a price cut equal to cumulative inflation, taken without anyone deciding to take it.
On invented annual rates of 3.4, 6.8, 3.9, 2.4 and 2.0 percent across five years, a fee unchanged throughout has fallen about 16.6 percent in real terms.
Three observations, ours.
Nobody signed off on a 17 percent price reduction. It arrived as an absence.
It is invisible in the accounts, because revenue per client is flat rather than falling, which is exactly the nominal presentation the literature says people respond to.
And a firm applying money illusion to itself would notice the loss and a firm not applying it would not, which makes this the rare case in this series where the effect is being suffered by the reader rather than exploited by them.
The Catch-Up Problem
An arithmetic trap in the correction itself. Ours, and it is the part most likely to be got wrong.
A fee that has lost 16.6 percent of its real value requires an increase of 19.8 percent to restore it, not 16.6.
Two observations.
The two numbers are different because they are percentages of different bases. Recovering a fall of x percent requires a rise of x divided by one minus x.
And using the smaller number leaves you permanently short. That is a nominal-versus-real confusion committed by the person actively trying to correct for one, which we find the most instructive thing in this article.
Why Firms Do Not Cut Pay
The classic application, stated carefully because we did not obtain the evidence for it.
The literature connects money illusion to sticky prices. The original authors are recorded as distinguishing phenomena in the real economy that suggest money illusion, of which "one is that prices are sticky" and "a second is that indexing does not occur in contracts and laws in times of relatively low inflation, as theory would pr"edict, with our source truncating[4].
We did not obtain the original paper and report only this characterisation of its argument.
Two observations, ours.
The indexing point is the sharper of the two. If people reasoned in real terms, contracts would routinely be indexed, and mostly they are not.
And for a professional firm this is directly checkable: how many of your engagement letters contain an indexation clause? If the answer is none, the firm has been repricing by negotiation instead, which is more expensive and less reliable.
The Honest Use Of This
The line we would draw, in the manner of the twenty-eighth article. Ours.
Three positions.
Using this to notice your own losses is straightforwardly good. A firm that recognises a flat fee as a real cut is correcting an error, not exploiting one.
Using it to present a real cut as a nominal rise is the reverse. An increase below inflation described as an increase is accurate and incomplete, and the twenty-eighth article's test applies: would the person, shown both descriptions side by side, feel they had been told the same thing?
And the honest middle is what we would actually recommend: say the real number. Telling a client that a fee is rising four percent against inflation of three is a different conversation from telling them it is rising four percent, and it is the conversation that survives their doing the arithmetic later.
What To Do
Price the absence, not just the decision. A fee held flat is a cut equal to cumulative inflation, and nobody signed off on it.
Get the catch-up arithmetic right. Recovering a 16.6 percent real fall needs a 19.8 percent increase, and using the smaller figure leaves you permanently short.
Check your engagement letters for indexation. The literature treats the absence of indexing as evidence of the phenomenon, and repricing by negotiation instead is more expensive.
Expect the nominal number to be what lands. The second-order finding is that people expect others to respond to nominal terms, which is why this survives people knowing better.
Say the real number anyway. Four percent against inflation of three is a different conversation from four percent, and it is the one that survives the client's later arithmetic.
Do not assume financial literacy protects you. One 2019 paper reports that highly financially literate participants did not make fewer errors, though they were sensitive to computational difficulty.
Treat the aggregate question as open. The market evidence is contested, including by a paper that says its results contrast with the best-known experiment.
Note which side of this you are on. Almost everything in this series describes an effect a firm might exploit. This one describes a loss a firm is probably taking.
The Limits Of This Analysis
Several caveats matter. This article reviews behavioural and experimental economics research. It is not pricing, employment, economic or investment advice. Every inflation rate used is invented for arithmetic illustration; none is Canadian CPI or any other published statistic, and none is a forecast. Everything is verified to August 2026. We did not obtain the 1997 paper, and every description of its contents reaches us through papers replicating, criticising or citing it. We did not obtain Fisher (1928), the 2001 experimental paper beyond its front matter and a methodological passage, the later coordination-failure paper beyond a truncated passage, or the 2019 paper beyond two truncated passages. We report the existence of a disagreement between a 2019 paper and the 2001 experiments without its content, because our source truncates mid-sentence. One source gives the original's title as "On Money Illusion" and reverses two authors' order; we use the better-attested forms and flag the variant. All arithmetic is ours, including the equivalence in the title, the cumulative erosion figures and the catch-up calculation. The application to professional fees, the indexation observation and the honest-use section are our own reasoning, untested. This article was built on a partially drafted earlier version whose citations we re-verified from source before use, and one of its claims, that two sources gave a variant title, we could confirm for only one and corrected accordingly.
Frequently Asked Questions
What is money illusion?
Has it replicated?
Is it disputed?
What is the second-order finding?
What does this cost my firm?
How much do I need to raise it by?
References
- Publisher record for Revisiting "money illusion": Replication and extension of Shafir, Diamond, and Tversky (1997), on Shafir, Diamond and Tversky (1997) having described money illusion as people's inclination to think of money without taking inflation sufficiently into account, in nominal terms rather than in real terms; and on the authors having successfully replicated Problems 1 to 4 of that study on Mechanical Turk with N = 604; together with a related record for a Brazilian replication describing money illusion as leading people to fail to consider the impact of inflation on the real value of money, and noting consequences including reluctance to sell a house or a stock at a nominal loss but a real gain, and the oversized appreciation of a nominal wage raise that is actually a real wage cut in times of high inflation. Note: a publisher record; we obtained the abstract and did not obtain the paper. sciencedirect.com
- Fehr, E., & Tyran, J.-R., Does Money Illusion Matter? Reply, hosted copy, on Shafir, Tversky and Diamond having provided evidence indicating that one or both preconditions for the absence of money illusion are frequently violated; on their results suggesting that people's preferences as well as their perceptions of the constraints are affected by nominal values; on many people not only seeming to suffer from money illusion but also expecting that other people's preferences and decisions are affected by money illusion; and its reference list giving the original as Shafir, E.; Diamond, P.A. and Tversky, A. (1997), On Money Illusion, Quarterly Journal of Economics. Note: a reply paper, obtained in part. This is the only source we located giving the title as "On Money Illusion" and the only one listing the authors as Shafir, Tversky and Diamond; both variants are flagged in the body and the better-attested forms are used. researchgate.net
- Fehr, E., & Tyran, J.-R. (2001). Does Money Illusion Matter? American Economic Review, 91(5), 1239–1262, hosted institutional copy of the published paper together with the authors' working paper version, on the authors observing subjects' behaviour directly in experiments while Shafir and colleagues asked hypothetical questions; on the study of Shafir and colleagues neatly showing that questionnaires can be a very useful instrument to examine the nature of money illusion at the individual level; on it also being clear that it is impossible to examine aggregate effects of individual interactions and adjustment processes with that method; and on the most important advantage of experimental methods for the authors' purposes being the ability to directly observe the evolution of individual and aggregate behavior after a nominal shock. Note: we obtained the paper's front matter and this methodological passage; we did not obtain its results. econ.uzh.ch
- Publisher record and hosted working paper for Are individuals in China prone to money illusion?, on the authors replicating the landmark study of Shafir, Diamond and Tversky (1997) to examine whether individuals in China are prone to money illusion; on the finding that money illusion is prevalent in China as well; on respondents in the Chinese sample often being somewhat more likely to base decisions on the real monetary value of economic transactions compared to respondents in the U.S. sample, and often being somewhat less prone to money illusion; on the survey questions having been translated into simplified Chinese with Chinese names and prices and dates adapted; on the original having been administered at an airport, in two New Jersey shopping malls and among Princeton undergraduates; and on Shafir, Diamond and Tversky distinguishing phenomena in the real economy that suggest money illusion, one being that prices are sticky and a second being that indexing does not occur in contracts and laws in times of relatively low inflation, our source truncating. Note: a publisher record and a hosted working paper; we did not obtain the full study. sciencedirect.com
- Bibliographic database record confirming Shafir, Eldar, Diamond, Peter, & Tversky, Amos (1997), Money Illusion, The Quarterly Journal of Economics, 112(2), 341–374; and Fehr, Ernst, & Tyran, Jean-Robert (2001), Does Money Illusion Matter?, American Economic Review, 91(5), 1239–1262, together with earlier working paper versions from 2000. Note: a bibliographic record, used for citation confirmation. ideas.repec.org
- Discussion paper, Fehr, E., and Tyran, J.-R., Money Illusion and Coordination Failure, on a powerful intuitive argument supporting the view that money illusion is largely irrelevant for economics, namely that the illusion has detrimental effects on people's economic well-being so they have a strong incentive to make illusion free decisions and will ultimately do so, implying little or no impact on aggregate outcomes at least in the long run; on the purpose of the paper being to show that this argument can be seriously misleading because it neglects the strategic repercussions of money illusion; on the authors showing experimentally that even if learning in the context of an individual optimization problem removes individuals' money illusion there can be further consequences, our source truncating; and its reference list identifying Fisher, I. (1928), The Money Illusion, Toronto: Longmans. Note: a discussion paper; we obtained the introduction only, and the passage stating the authors' own result truncates before the result. docs.iza.org
- Repository copies of Revisiting "Money Illusion": Replication and Extension of Shafir et al. (1997), on participants having completed Problems 1 to 4 in random order as part of a larger set of experiments; on those four problems having been chosen because they represented simple money illusion demonstrations in different contexts without the complexity added by combining money illusion with mental accounting and fairness concerns investigated in Problems 5 to 7; on the final sample having been determined by the project budget and being larger than the estimated required sample of 166; on 604 participants having been recruited from Mechanical Turk, being 288 males, 315 females and 1 other, with mean age 40.09 and standard deviation 11.48; on sensitivity analyses indicating approximately 99.99 percent power to detect all the original effects at a two-sided alpha of 0.05; on Shafir and colleagues having argued that while people know the real value of money, the nominal representation is more salient and simpler; and on Problem 2 results in which Carl was selected as having had the best deal by 421 of 604 participants at 70 percent, Ben as second best by 481 of 604 at 80 percent, and Adam as worst by 456 of 604 at 76 percent, with the authors concluding the replication was successful. Note: repository copies of the replication paper; we obtained these passages and not the full paper. researchgate.net
- Publisher record for Replication: The money illusion effect in a Brazilian sample and meta-analyses, on the study aiming to replicate the four conditions outlined in the original research adapted to the Brazilian context, covering earnings, transactions and contracts; on it being a cross-sectional and pre-registered study of 372 Brazilian participants conducted via mobile phone or computer; on the results found being very similar to the original findings, with participants showing varying inclinations towards opting for economically advantageous opportunities depending on whether real, nominal or neutral framing was used; and on the original 1997 article having garnered over 1,242 citations on Google Scholar by 2023. Note: a publisher record; we obtained the abstract and did not obtain the paper or the meta-analyses its title refers to. sciencedirect.com
- Accepted manuscript of a 2019 paper on money illusion, financial literacy and numeracy, on agents having difficulties understanding the impact of price fluctuations on revenues and being confused about the differences between real and nominal returns or values even when they know the inflation rate; on the authors' evidence not supporting a particular explanation, with highly financially literate participants not making more errors in noncongruent than in congruent choices while being sensitive to computational difficulties; and on the authors stating that by contrast with previous results, Fehr and Tyran (2001) find nominal inertia is higher in nominal conditions, our source truncating mid-sentence. Note: an accepted manuscript; we obtained two passages, both truncated. We report the existence of a disagreement with the 2001 experiments without being able to report its content. sciencedirect.com
This article reviews behavioural and experimental economics research and is not pricing, employment, economic or investment advice. Every inflation rate used is invented for arithmetic illustration; none is Canadian CPI or any other published statistic, and none is a forecast. The 1997 paper was not obtained and is described only through papers replicating, criticising or citing it. All arithmetic is the authors' own.