The Veblen Effect Is Mostly Folklore
The rising demand curve for Veblen goods is not Veblen's, is not in his book, and was drawn in 1950 by an economist who said only a segment of it could slope upward. The one shop experiment where unit sales rose with price peaked in the middle of the range, and the best field experiment on status goods found that cutting the price sold more.
Answer to the question
Are Veblen goods real?
The status premium is real and has been measured in the field, where 21.0 per cent of bank customers took a platinum card against 13.7 per cent for identical benefits at the same price. The upward-sloping demand curve attached to Veblen goods is a separate and weaker claim. Leibenstein coined it in 1950 in a paper with no data, and predicted a backward S in which only a segment rises.
Key takeaways
- An unbranded walkie-talkie in a randomised shop experiment sold 47 units at the lowest of three prices, 74 in the middle and 36 at the top.
- Leibenstein, who defined the Veblen effect in 1950, predicted a backward S bounded by the income constraint above and satiety below, and his paper contains no data.
- He credited the underlying observation not to Veblen but to John Rae, writing before 1834, and the phrase Veblen goods appears nowhere in Veblen's book.
- In the best field experiment on status goods, 21.0 per cent took a platinum card against 13.7 per cent for identical benefits, and a 25 per cent discount raised take-up.
Forty-seven units, then seventy-four, then thirty-six. That is one unbranded walkie-talkie sold at three rising prices across eighteen stores of an American toy chain, the three numbers coming from a Wharton trade piece of 2003 and not the published paper. They are the closest thing to transaction evidence anybody has produced for Veblen goods, the things people are said to buy because they cost a lot, and the researchers thought status had nothing to do with it. Read the first two and the folklore looks vindicated. Read the third and it falls over: at the top price sales dropped below where they started. That is the backward S the economist who named the Veblen effect predicted, not the straight climb.
Take the strong version seriously first. Some goods are bought for their price, so raising the price raises the quantity sold and the demand curve for Veblen goods slopes up. That is a claim with a shape, and a shape can be looked for.
So where did the shape come from?
1899: the book that draws no curve
Not from Thorstein Veblen. The Theory of the Leisure Class contains no price schedule, no quantity and no demand analysis, and the phrase Veblen goods appears nowhere in it, which we established from the primary text. The nearest thing to the modern claim is in Chapter Six, about the judgement of beauty:
> The marks of expensiveness come to be accepted as beautiful features of the expensive articles.
He argues that people who start by disapproving of an inexpensive way of living end by thinking cheap things dishonourable, without noticing. That is a claim about taste, which the modern term converts into a claim about a market.
Nor is the book evidence. Across fourteen chapters there is no footnote, no table, no figure and no dataset, and Veblen twice concedes his support is thin. This publication cited him at second hand for a year and learned what he wrote only on reading him in full, in a digital text faulty enough that quotations need checking against print.
So who drew the curve?
1950: Leibenstein names it and hedges it
Leibenstein did, in the Quarterly Journal of Economics, and the definition is his:
> By the Veblen effect we refer to the phenomenon of conspicuous consumption; to the extent to which the demand for a consumers' good is increased because it bears a higher rather than a lower price.
He separated that from the snob effect, which in his words is a function of the consumption of others while his Veblen effect is a function of price, and introduced the construct the idea runs on: the real price you paid, against the conspicuous price, his term for the price other people think you paid.
The paper contains no data. Its method is a thought experiment with imaginary questionnaires, then diagrams.
The sentence the field dropped is worse for the folklore. The paper rules out a curve that slopes upward throughout. It will normally be shaped, he writes, like a backward S, for two stated reasons: a price so high that nobody buys at it, owing to the income constraint, and some point of satiety. The straight rising line drawn for Veblen goods is not in the paper it comes from.
The straight rising line drawn for Veblen goods is not in the paper it comes from.
He also gives the idea away, writing that Veblen was neither the discoverer nor the first to elaborate conspicuous consumption, and naming John Rae, who wrote before 1834. We have read neither Rae nor Norine Foley's 1893 fashion article, also cited.
Could the formal economics supply the missing curve?
1996 to 2005: models, then a laboratory
It supplies conditions instead. Bagwell and Bernheim proved in 1996 that Veblen effects cannot ordinarily arise when preferences satisfy a single-crossing property, and may emerge when it fails. The standard assumption has to break before the effect can appear.
Two later models get the positive slope and disagree about why. Amaldoss and Jain derive it in 2005 from a desire for uniqueness, and report a laboratory test supporting it, though the abstract gives no sample size. Corneo and Jeanne derive it from conformism, in a 1994 Bonn working paper. We never got past its abstract, which is why nothing here is credited to the published 1997 article. Two models, opposite motives, one shared shape, and the shape cannot test either.
The applied work on Veblen goods splits the same way. Fassnacht and Dahm ran eight experiments in 2021 and found the effect exists, their outcome measure being willingness to buy, with nothing bought. A 2026 preprint by Müller, not peer-reviewed, reads Swiss watch and Ferrari data and finds it barely exists.
What happens when somebody charges real money?
Since 2005: real prices and real purchases
Gaur and Fisher published the one field experiment we found in which real unit sales rose with price. The same products were given different prices across panels of stores, with the panels assembled by a clustering method so that each stood for the chain rather than by a random draw. Their own sentence: the experiment yielded the unexpected result that demand increases with price in some cases.
Two of the three products sloped downward. The third is the walkie-talkie. Put to merchandise managers, the explanation came back as a complex unbranded item whose quality a shopper cannot judge, so [price is read as the indicator](https://thewhyweproject.com/articles/why-the-dearer-wine-tastes-better), plus a gift price point, twenty dollars beating fifteen.
Status, then, still needs its own test.
It has one. Bursztyn and colleagues worked with a large Indonesian bank, pre-registered the design, and telephoned 835 customers with a paid credit card upgrade. One group was offered every instrumental benefit of the platinum card on a plain card, the other the platinum card itself. At the same price, 13.7 per cent took the benefits and 21.0 per cent the platinum card.
Then the sentence this turns on. The authors state plainly that they did not use randomised price variation, and report what they did instead: a 25 per cent discount offered to those who had declined raised take-up of the benefits upgrade to an estimated 17.4 per cent. Cutting the price sold more.
Cutting the price sold more.
Their second experiment comes closest to the folklore. Of 180 platinum holders sampled, 93 were reached and offered an identical diamond card for 50,000 rupiah more than theirs. Told the bank had lowered the income threshold so that lower-income customers qualified, take-up went from 21.6 to 40.5 per cent. But the manipulation was [information about who else could have the card](https://thewhyweproject.com/articles/which-kind-of-scarcity-works-on-you), not its price, and on 93 people the gap fell short of conventional statistical significance.
Both readings are strong. The status case: a field experiment with real money found a third of the demand for [a prestigious product attaching to the prestige alone](https://thewhyweproject.com/articles/what-the-luxury-label-actually-buys), strengthening when exclusivity was threatened, with a theorem showing it can hold in equilibrium. The quality case: Rao and Monroe's meta-analysis gathered 36 studies and drew 54 price to perceived-quality relationships out of them, and across those 54 the association was moderate and replicated.
We take the quality reading for the slope and the status reading for the premium, because that is what the designs can carry. Rao and Monroe are careful about their limits: laboratory studies only, mostly cheap and frequently bought goods, and whether the association grows for expensive purchases is, they say, not documented.
Two studies would settle it, and neither has been run so far as we can find:
- A randomised price experiment on a conspicuous good, measuring what people actually bought.
- A measurement of the conspicuous price as Leibenstein defined it, the price you think other people think you paid. We searched for that construct and found no study of it, which is narrower than saying visible spending has gone unstudied.
Both are doable. Until then the honest position on Veblen goods is that the premium is measured and the slope is only drawn.
One demonstration of upward-sloping demand does exist, and it is clean. Jensen and Miller gave vouchers cutting the price of the staple grain to 1,300 poor urban households in two Chinese provinces. In Hunan, lowering the price of rice led households to buy less rice, and removing the subsidy reversed it. That is a real-income effect near subsistence, Giffen behaviour, with no status in it, and the one proven upward-sloping curve in economics belongs to people too poor to eat anything else.
So the next time the dear thing looks better, ask whether you are reading the price as information about the object or as information about yourself, because on everything measured so far the evidence is much stronger on the first.
Common questions
Are Veblen goods real?
The status premium is real and measured in the field: at the same price, 21.0 per cent of bank customers took a platinum card against 13.7 per cent for identical benefits without it. The upward-sloping curve attached to Veblen goods is a separate claim. Our verdict on the slope leans on Rao and Monroe's 36 studies, where price moved perceived quality moderately, and we found no randomised price test on a conspicuous good measuring real purchases.
Did Veblen invent Veblen goods?
No. The Theory of the Leisure Class carries no price schedule, no quantity and no demand analysis, and the phrase is nowhere in it. The effect was defined by Leibenstein in 1950, who credited the observation to John Rae, writing before 1834, whom we have not read.
Does a higher price ever raise unit sales?
Once, in a shop experiment with real transactions: an unbranded walkie-talkie sold 47 units at the lowest of three prices, 74 in the middle and 36 at the top. The researchers put that middle peak down to quality inference and a gift price point.
Sources
Leibenstein, H. (1950). 'Bandwagon, Snob, and Veblen Effects in the Theory of Consumers' Demand.' Quarterly Journal of Economics, 64(2), 183-207. doi:10.2307/1882692. [FULL TEXT READ at source, from the scan hosted at gwern.net; the article signs off 'H. Leibenstein. Princeton University', so the forename should be verified separately before it is printed]: the origin of the effect and the most valuable source in this file. The definition in his words: 'By the Veblen effect we refer to the phenomenon of conspicuous consumption; to the extent to which the demand for a consumers' good is increased because it bears a higher rather than a lower price.' He separates it sharply from the snob effect, 'the former is a function of the consumption of others, the latter is a function of price', and introduces the distinction between the real price, what the consumer paid, and the conspicuous price, 'the price other people think the consumer paid for the commodity', refined in a footnote to 'the price that the consumer thinks other people think he paid'. FOUR THINGS THE TEXTBOOK VERSION LOSES. First, the paper contains no data of any kind: the method is a gedankenexperiment with imaginary questionnaires followed by diagrams, and he states that for his purposes he can 'quite legitimately abstract from the psychological and sociological elements'. Second, and most dropped, the paper sets aside the curve can slope upward throughout: it will normally be shaped 'like a backward S' because 'there must be a price so high that no units of the commodity will be purchased at that price owing to the income constraint' and because 'there must be some point of satiety for the good', and his conclusion says only that 'some portions of it may even be positively inclined'. Third, he credits the idea away from Veblen in print, writing that the treatment 'was made famous by Veblen although Veblen, contrary to the notions of many, was neither the discoverer nor the first to elaborate upon the theory of conspicuous consumption', naming John Rae 'writing before 1834' as having 'quite an extensive treatment of conspicuous consumption, fashions, and related matters pretty much along Veblenian lines', with Rae tracing the notion to Horace and the keeping-up-with-the-Joneses idea to Alexander Pope, and pointing also to Norine Foley's article 'Fashion', Economic Journal, 1893. Neither Rae nor Foley has been read by the vault and both must be cited only as Leibenstein's report. Fourth, he rates his own contribution as akin to 'the purchase of highly speculative stocks rather than the purchase of government bonds'.
Bagwell, L. S., & Bernheim, B. D. (1996). 'Veblen Effects in a Theory of Conspicuous Consumption.' American Economic Review, 86(3), 349-373. [ABSTRACT READ IN FULL at Bernheim's own Stanford publication page, which prints it entire; full text not retrieved, the JSTOR copy being paywalled. Author order confirmed there and at the Crossref reference record of Amaldoss and Jain 2005: Bagwell first, Bernheim second]: a theorem, not a finding, and the sharpest statement available of how hard the effect is to obtain even on paper. In the authors' words: 'While Veblen effects cannot ordinarily arise when preferences satisfy a single-crossing property, they may emerge when this property fails.' Where it does fail, 'budget brands are priced at marginal cost, while luxury brands, though not intrinsically superior, are sold at higher prices to consumers seeking to advertise wealth', and luxury brands 'earn strictly positive profits under conditions that would, with standard formulations of preferences, yield marginal-cost pricing'. The standard assumption rules the effect out and the assumption has to be broken to let it in, which the article should state plainly.
Corneo, G., & Jeanne, O. (1997). 'Conspicuous consumption, snobbism and conformism.' Journal of Public Economics, 66(1), 55-71. doi:10.1016/S0047-2727(97)00016-9. [THE PUBLISHED ARTICLE WAS NOT READ and nothing may be attributed to it. What was read in full at source is the abstract of its direct predecessor, University of Bonn SFB 303 Discussion Paper A-461, December 1994, titled 'Conspicuous Consumption and the Existence of Upward Sloping Demand Curves', at the Bonn abstract page, which gives author order as Corneo first, Jeanne second]: the working paper models conspicuous consumption as a device that signals the consumer's social status, status being some function of the individual's rank in the wealth hierarchy, and says the approach 'makes it possible to distinguish between snob and conformist consumer behavior'. THE COUNTERINTUITIVE RESULT THE ARTICLE NEEDS: 'It is shown that when behavior is conformist, the market demand curve for the good can exhibit a positive slope.' Not when buyers are snobs. When they are conformists, which is the opposite of the exclusivity story the trade tells. Keywords listed include 'law of demand'. The 1997 version's abstract as relayed by RePEc also notes unconventional implications for luxury taxation and the voluntary provision of public goods, and that is a secondary relay rather than a read.
Amaldoss, W., & Jain, S. (2005). 'Pricing of Conspicuous Goods: A Competitive Analysis of Social Effects.' Journal of Marketing Research, 42(1), 30-42. doi:10.1509/jmkr.42.1.30.56883. [ABSTRACT READ IN FULL from the publisher-deposited abstract at the Crossref item record; full text not retrieved. Author order confirmed at Crossref: Amaldoss first, Fuqua School of Business, Duke University; Jain second, Robert H. Smith School of Business, University of Maryland]: a duopoly model plus a laboratory test. The modelling results: in a duopoly the authors 'identify the conditions under which the desire for uniqueness can increase demand among some consumers as the price of a product increases'; the desire for uniqueness leads to higher prices and firm profits while a desire for conformity leads to lower prices and profits; and, strikingly, 'consumers purchase high-quality products not because of their desire for uniqueness but despite it'. THE EMPIRICAL SENTENCE, AND IT IS ONE SENTENCE: 'In a laboratory test, the authors find support for the claim that demand for a product among consumers who desire uniqueness may increase as its price increases.' That is the clearest demonstration in the formal literature that the effect can be produced at all. It is a laboratory test, the abstract gives no sample size and no design, none could be verified on this pass, and the article may therefore report that a laboratory test supported it and may not attach a number to it. NOTE THE UNRESOLVED CONTRADICTION: Amaldoss and Jain get the positive slope from uniqueness, Corneo and Jeanne get it from conformism, and two models predicting the same slope from opposite motives is a sign the prediction is not pinning down the mechanism.
Gaur, V., & Fisher, M. L. (2005). 'In-Store Experiments to Determine the Impact of Price on Sales.' Production and Operations Management, 14(4), 377-387. doi:10.1111/j.1937-5956.2005.tb00227.x. [ABSTRACT READ IN FULL from the publisher-deposited abstract at the Crossref item record; full text not retrieved. Author order confirmed at Crossref: Gaur first, then at the Stern School of Business, New York University; Fisher second, at the Wharton School]: a genuine randomised field experiment with real transactions, and the only one found in which unit sales rose with price. The same product was assigned different price points across store panels and the resulting real sales used to estimate a demand curve, with panels formed by a variant of the k-median problem to control for differences between stores and to be representative of the whole chain. The authors' own sentence: 'Our experiment yielded the unexpected result that demand increases with price in some cases. We present likely reasons for this finding from our discussions with retail managers.' STUDY DETAIL FROM A SECONDARY SOURCE, AND THE ARTICLE MUST SAY SO: Knowledge at Wharton, 'The Price is Right, or Is It? Determining the Impact of Price on Sales', 5 November 2003, the Wharton School's own publication, read in full. It reports 18 stores of the Zany Brainy toy chain, nearly one third of its 53 stores, over six weeks, three products at three price points each: a family game centre at 19.99, 24.99 and 29.99 United States dollars, a Phonics traveller at 24.99, 29.99 and 34.99, and an unbranded headset walkie-talkie at 14.99, 19.99 and 24.99. Store managers were not told and labels did not show the original list price so nothing read as a markdown. Two products sloped downward. The walkie-talkie sold 47 units at 14.99, 74 units at 19.99 and 36 units at 24.99, so demand peaked in the middle of the range and fell at the top, which is Leibenstein's backward S rather than the straight upward line of the folklore. THE AUTHORS' OWN EXPLANATION IS NOT STATUS: per the Wharton account they put it to merchandise managers and came away with price-quality inference, the walkie-talkie being a complex unbranded electronic item whose quality shoppers cannot judge, and a price-point convention, 19.99 being a more popular gift price than 14.99. Nobody in this experiment was buying a toy walkie-talkie to be seen with it. One further line from the Wharton piece, usable as an anecdote and nothing more: Fisher reports that 'one retailer I worked with found that half the time the products it tested sold better at the higher price, an illogical result that led the company to conclude that the test was invalid', an unnamed retailer with no data attached.
Bursztyn, L., Ferman, B., Fiorin, S., Kanz, M., & Rao, G. (2018). 'Status Goods: Experimental Evidence from Platinum Credit Cards.' The Quarterly Journal of Economics, 133(3), 1561-1595. doi:10.1093/qje/qjx048. [FULL WORKING TEXT READ at source, the September 2017 author copy at home.uchicago.edu. Author order confirmed at the Crossref item record for the QJE version: Bursztyn first, University of Chicago and NBER; Ferman, Sao Paulo School of Economics FGV; Fiorin, then UCLA Anderson; Kanz, World Bank; Rao, Harvard and NBER. The first experiment was pre-registered at the American Economic Association's Social Science Registry, RCT ID AEARCTR-0000828]: the strongest field evidence that exists on status goods, run with a large Indonesian bank on real customers at real prices, and the source that decides the article. EXPERIMENT 1, n = 835: paid credit card upgrades were offered by telephone, the control group receiving every instrumental benefit of the platinum card as an upgrade on a nondescript card and the treatment group the platinum card itself. At the same price, take-up was 13.7 per cent for the benefits and 21.0 per cent for the platinum card, a gap of 7.3 percentage points, p = 0.029, so roughly a third of the demand was for the status and nothing else. Transaction data showed platinum cards were more likely to be used in social contexts. THE FINDING THAT SETTLES THE DEMAND QUESTION: the authors state plainly that 'We did not use randomized price variation', and report that a 25 per cent discount offered to customers who had declined raised demand for the benefits upgrade from 13.7 per cent to an estimated 17.4 per cent. In the best field experiment on status goods, cutting the price raised take-up and demand sloped downward. They use that to size the status premium: the average consumer values the status aspect of the card at roughly what a 25 per cent discount is worth. EXPERIMENT 2, the positional-externality test and the closest thing to a field Veblen effect anyone has: 180 platinum holders with incomes of at least 500 million rupiah were sampled and 93 were reached. All were offered a functionally identical 'diamond' card at an annual fee of 650,000 rupiah, 50,000 more than the platinum card they already held, differing only 'in a new and different design and color'. The treatment group was additionally told the bank had cut the platinum income threshold so that lower-income customers now qualified. Take-up rose from 21.6 per cent to 40.5 per cent, p = 0.069. HOLD THAT AT THE RIGHT ANGLE: people agreed to pay more for no extra function and twice as many agreed when exclusivity was underlined, but the manipulation was information about who else could have it, not the price. The p-value of 0.069 is above the conventional threshold and the final sample is 93 people, and both facts belong in the prose.
Fassnacht, M., & Dahm, J.-M. (2021). 'Growing luxury brands by increasing the price: does the Veblen effect exist?' Luxury Research Journal, 2(1/2), 99-139. doi:10.1504/LRJ.2021.116292. [ABSTRACT READ IN FULL at the Inderscience publisher record; full text not retrieved, the article being paywalled. AUTHOR-ORDER CAUTION, recorded because it is exactly the trap this kind of citation sets: Crossref carries two deposits for this article with the order reversed, DOI 10.1504/lrj.2021.10039169 listing Dahm first and DOI 10.1504/lrj.2021.116292 listing Fassnacht first. The publisher's own record of the published version gives 'Martin Fassnacht; Jil-Marie Dahm', both at WHU Otto Beisheim School of Management, Vallendar. The article follows the publisher record and the vault notes the discrepancy]: the paper 'empirically analyses its existence through eight experiments' across luxury product categories and concludes that 'the Veblen effect exists, yet it varies between hard and soft luxury goods whether intrinsic or extrinsic purchase motivations dominate respectively'. The outcome measure is decisive and is in the keyword list: 'willingness to buy; WTB'. Eight experiments measuring stated willingness to buy, not purchases. The paper's own framing concedes that 'research has neglected the Veblen effect over the past decades'. No sample size, population or design could be verified on this pass, so the article may cite it as eight experiments on willingness to buy finding support, and no further. Accepted 3 October 2020, published online 16 July 2021. Living authors: the criticism is of an outcome measure, never of the people.
Muller, F. (2026). 'Does the Veblen Effect Exist? An Analysis of its Likelihood in Luxury Consumption and Potential Consequences for Pricing Management.' SSRN working paper. doi:10.2139/ssrn.6836182. [ABSTRACT READ IN FULL from the Crossref deposited abstract; full text not retrieved. THIS IS A PREPRINT AND NOT A PEER-REVIEWED ARTICLE and must be labelled one wherever it appears]: the sceptical pole, and the only source found that looks at long-run industry data. In the author's words, the effect, 'if it exists at all, is only temporary and limited in scope as exemplified by the empirical data of the Swiss watch industry and sports car manufacturer Ferrari from 2000 to 2025', and 'the analysis concludes that the effect barely exists, implying a much more complex context for luxury pricing'. It earns a place despite the preprint status because nothing else in the literature attempts the long-run observational test. Living author: the caveat is about peer review, nothing adverse about the man is asserted.
Rao, A. R., & Monroe, K. B. (1989). 'The Effect of Price, Brand Name, and Store Name on Buyers' Perceptions of Product Quality: An Integrative Review.' Journal of Marketing Research, 26(3), 351-357. doi:10.2307/3172907. [FULL TEXT READ at source, the copy hosted by the Carlson School of Management, University of Minnesota. Rao first, then Assistant Professor of Marketing at the University of Minnesota; Monroe second, then the Robert O. Goodykoontz Professor of Marketing at Virginia Polytechnic Institute and State University]: the rival explanation, meta-analysed, and the numbers are specific enough to anchor a paragraph. 36 studies reporting 85 effects, of which 54 are price-perceived quality results, with eta squared as the effect size index throughout. Weighted mean eta squared for price on perceived quality = .12, weighted standard deviation .148, standard error .017, 95 per cent confidence limit plus or minus .033, which the authors call on Cohen's criteria 'moderately large and statistically significant'. Brand name is slightly larger at .142 and significant; store name is .052 and not significant, its confidence interval including zero. THE METHODOLOGICAL FINDINGS ARE THE REAL PRIZE: within-subjects designs produced eta squared of .20 against .10 for between-subjects designs; the strength of the price manipulation was significantly and positively associated with effect size; and price level was not. In their words, 'strong manipulations are more likely to produce large effects, thereby enhancing the chance of a statistically significant effect', and 'weak price manipulations are a possible explanation for previous price-perceived quality studies' failure to find a statistically significant price main effect'. THE LIMIT THE AUTHORS PUT ON THEIR OWN REVIEW, to be quoted rather than paraphrased: 'Only laboratory studies involving one or more of price, brand, and store cues were used in the review.' The products are mostly cheap: prices ran from 11 cents for table salt to 400 dollars for stereo receivers, only four consumer products exceeded 100 dollars and 60 per cent cost less than 30 dollars. Their conclusion: 'the phenomenon has been shown to be present at a moderate level of association for relatively lower priced, frequently purchased goods, but whether the strength of the association increases for higher priced, less frequently purchased goods has not been documented adequately.' For the four results above 100 dollars the weighted mean effect was eta squared = .39, which is large, and four results is four results. One figure from their introduction that is not theirs and must be attributed as their citation of it: they report Tellis and Wernerfelt 1987 as finding the actual correlation between price and product quality across product markets to average r = .27. The vault has not read Tellis and Wernerfelt.
Jensen, R. T., & Miller, N. H. (2008). 'Giffen Behavior and Subsistence Consumption.' American Economic Review, 98(4), 1553-1577. [FULL TEXT READ at source, the published version hosted at Occidental College]: in this file for one reason, as the standard against which the status literature should be measured. The abstract's first sentence: 'This paper provides the first real-world evidence of Giffen behavior, i.e., upward sloping demand.' Design: a randomised field experiment in which 1,300 poor urban households, 650 in each of two Chinese provinces, covering 3,661 individuals, were assigned to a control group or to one of three treatment groups receiving printed vouchers cutting the price of their dietary staple by 0.10, 0.20 or 0.30 yuan for five months, against pre-intervention prices of about 1.2 yuan per jin for rice in Hunan and 1.04 for wheat flour in Gansu. Vouchers covered 750g per person per day, about twice average consumption, were redeemable only at participating local grain shops, and could not be resold. Consumption surveys were gathered before, during and after. Result: strong evidence of Giffen behaviour for rice in Hunan, where 'lowering the price of rice via the experimental subsidy caused households to reduce their demand for rice, and removing the subsidy had the opposite effect', robust across specifications, and weaker evidence for wheat in Gansu. The elasticity depends nonlinearly on the severity of poverty, so the very poorest do not show the effect and the group in the middle does. Two sentences the article should carry: 'The fact that there has, to date, been no convincing evidence of Giffen behavior stands as a minor embarrassment to economists', and the methodological conclusion that transfers directly, 'the search should not be for a Giffen good, but for Giffen behavior', because a subset of consumers may slope upward while the market does not. CRITICAL CONSTRAINT: this is a real-income effect among households near subsistence. It is not a Veblen effect, it is not a status effect, and presenting it as one would be the worst error available in this piece.
Veblen, T. (1899). The Theory of the Leisure Class: An Economic Study of Institutions. [Read in full by the vault on 2026-09-20 from Project Gutenberg ebook 833, all fourteen chapters, 661,874 characters of which roughly 642,500 are Veblen; logged at Books/The Theory of the Leisure Class/The Theory of the Leisure Class.md with the session note at Books/The Theory of the Leisure Class/Notes/2026-09-20.md. ON THIS PASS the Gutenberg plain text truncated at 136,785 characters, cutting off inside Chapter Four, exactly the failure the session note documented, so Chapter Six could not be re-read at source today and all Chapter Six quotation comes from the vault's own full read as transcribed at Concepts/Pecuniary Canons of Taste.md. en.wikisource.org returned an empty body and is recorded as not read]: unpaginated, cited by chapter. The book is the seed and the source of the vocabulary. It is not a citation source for any demand claim, because it contains none, and the whole point of using it here is that negative. Its evidentiary base, per the vault's audit: no footnote, no table, no figure, no dataset and no date of any study across fourteen chapters, roughly one named source in the whole book quoted for a proposition, and two explicit admissions by Veblen that his evidence is drawn from psychology rather than ethnology and that 'little of a tangible character can therefore be offered in support of the view here taken'. The race and inherited-temperament material, including the Chapter Six lawn argument, is in the hub's quarantine table and is excluded from this article entirely.