There is a version of this experiment I keep returning to, and it happens on a card table in a psychology building in North Carolina in the early 1970s. A woman sits down and is told she will be rating consumer products. In front of her is a glass jar of chocolate chip cookies holding either ten cookies or two, with the same jar, the same cookies, and the same script. Then a second man walks in carrying a jar of his own, apologizes for interrupting, and asks the first experimenter whether he has enough cookies to spare. The two men swap jars and the second man leaves. That swap is the entire point of the study, and it is the reason I still think about a paper published fifty years ago.
🧠 The Jar on the Table
Stephen Worchel at Virginia, with Jerry Lee and Akanbi Adewole at North Carolina, published “Effects of Supply and Demand on Ratings of Object Value” in the Journal of Personality and Social Psychology in 1975. Two experiments, 200 female undergraduates in total. The first one ran 146 students from an introductory psychology class at Chapel Hill, and twelve of them were dropped from the analysis for refusing to eat the cookies or for being suspicious about the setup.
The task was simple. Taste nothing, just look, then rate the cookies on how much you like them, how attractive they seem, and how much you would pay.
The two-cookie jar beat the ten-cookie jar. Same cookie. Same brand, same size, same jar. Only the count differed, and the count was enough to move the price.
Then came the swap, which is where the design gets clever. In one condition the subject’s jar started with ten cookies and ended with two, because the second experimenter either needed them (his own subjects had eaten more than expected) or had taken them by mistake. In another condition the jar started with two and ended with ten. The subject watched the change happen, which means she was not rating scarcity in the abstract. She was rating a supply that had just moved.
Cookies that fell from ten to two were rated more valuable than cookies that had been sitting at two the whole time. F(1,122) = 18.08, p < .001. A shortage you watched arrive is worth more than a shortage that was always there.
And the reason attached to the shortage mattered. Cookies scarce because other people had been eating them beat cookies scarce because an experimenter made a clerical error, F(1,122) = 4.26, p < .05. The number on the table was not the message. The message was that somebody else wanted them.
⚙️ How Scarcity Actually Works
The tempting explanation is that people want what they cannot have, full stop. That is too flat to be useful, because it cannot explain the ordering the study actually produced. Here is the chain as I understand it.
Start with the fact that a brain never prices an object in isolation. It prices the object times a rough coefficient for how obtainable it will be later. This is Brock’s commodity theory from 1968, stated plainly: any commodity is valued to the degree that it is unavailable. The cookie does not change. The multiplier does.
Next, the removal of an option generates its own pressure. Brehm’s reactance work, which the authors cite directly, had already shown that when your freedom to obtain something is eliminated, the thing becomes more attractive. Ten cookies means you could take any of them. Two means the supply is closing. What closes gets wanted harder, and that rebound is not a decision you make. It arrives before the decision.
Then there is the inference. A shrinking supply is read as information about the object rather than about the jar. If other people are taking them, the reasoning goes, they must know something I do not. This is why the demand explanation outperformed the accident explanation. Scarcity is never just a quantity. It is a quantity plus a rumor about other people.
The fourth step is where the arithmetic hides. Ten dropping to two is not experienced as “two.” It is experienced as minus eight, and losses are weighted more heavily than equivalent gains. The reference point is the past quantity, not zero, which is why a jar that was always at two barely registers. No drop, no loss, no bump. This is the same machinery as relative deprivation: people revolt over what they had and lost, not over what they never had.
Put those together and you get a ranking that is not “less is better” but something stranger. Falling supply beats constant scarcity. Constant scarcity beats constant abundance. And constant abundance beats a supply that suddenly increased.
🤔 The Opposite of Scarce
Everyone quotes the first half of this study. Almost nobody quotes the second, which is the part I find genuinely useful: cookies that went from two to ten were rated lower than cookies that had been at ten the whole time.
An increase is not neutral. It is a signal, and the signal reads as nobody wanted these. Which means the real antonym of scarcity is not abundance. It is unwantedness. A product that abruptly becomes easy to get has not merely lost its premium. It has been publicly demoted.
I should also be honest about the size of the effect, because this study gets cited as if it were a magic trick. In the second experiment, Worchel and his coauthors reported that if they left the suspicious subjects in the analysis, the scarcity effects disappeared entirely. Among the subjects who had not caught on, only the willingness-to-pay rating reached significance at p < .05. Liking was marginal at p < .10. Attractiveness was not significant at all, F(1,122) = 1.13.
So the 1975 result is a modest price bump in a small all-female student sample, not a lever that turns gravel into gold. What survived is the direction, not the magnitude. The 2022 meta-analysis by Barton, Zlatevska and Oppewal, pooling 416 effect sizes from 131 studies, found scarcity cues reliably raise purchase intentions, and then found something worth taping to a wall: in advertising, supply-based scarcity produced an effect of 0.69, time-based scarcity 0.26, and demand-based scarcity minus 0.03. Telling people everyone is buying it did essentially nothing. Telling them production is limited did a lot.
🔗 What This Means If You Sell Anything
Anyone building a product whose value rests on being taken seriously should be careful with this lever, because the lever changes what the product is.
A limited number of slots converts a relationship into a transaction with a countdown attached. The first time it works. The second time the customer learns to wait. The third time they conclude the slots refill, and you have built the two-to-ten jar on purpose, in public, at your own expense. Scarcity borrowed against trust has to be repaid.
The honest version is the one with a real constraint behind it: how many deep readings one person can actually do in a week, how many pieces a workshop can physically produce, how many dates a season holds. Real constraint communicates care about quality. Invented constraint communicates that you think the buyer is stupid, and buyers are slow but not that slow.
For anyone doing diligence on a consumer brand, the same distinction is a diagnostic. Separate scarcity that comes from capacity, craft, or calendar from scarcity that resets every Monday. The second kind shows up in the cohort data as a first-purchase spike followed by a flat or falling repurchase rate, because the mechanism trains waiting rather than wanting.
Serialized fiction sits on the good side of this without trying. A chapter a week is a real constraint, and the waiting it creates is rhythm rather than pressure. Rhythm produces anticipation. Pressure produces defense, and defense is expensive to undo.
🎲 The Footnote
Twelve women were excluded from the first experiment for refusing to eat the cookies or for being suspicious, which is my favorite data-cleaning note in social psychology. The cookies were the treatment, and some people simply declined to be treated.
Also, the line everyone remembers from this paper — that a limited edition often costs four times as much as the everyday product — is not a finding. It is a throwaway observation in the opening paragraph, before any data appear. Fifty years later it circulates as the result.