Endowment Effect
The endowment effect is Richard Thaler's 1980 observation that people demand more to give up a thing than they would have paid to acquire it. Kahneman, Knetsch and Thaler demonstrated it with coffee mugs in 1990, where sellers wanted roughly twice what buyers would pay for the same mug handed out minutes earlier. Charles Plott and Kathryn Zeiler argue the gap is produced by the experiment rather than by preferences, and that dispute has never been settled. This page is graded mixed for that reason.
What it is
Ask someone what they would pay for a coffee mug. Then give them the mug and ask what they would sell it for. The second number is bigger, and it is bigger by a lot.
Thaler named this in "Toward a positive theory of consumer choice", in the first issue of the Journal of Economic Behavior and Organization in 1980. His definition is narrower than the one that later escaped into general circulation, and worth recovering. He called the endowment effect "the underweighting of opportunity costs". Money leaving your pocket is coded as a loss. Money you fail to receive is coded as a gain you did not get. Prospect theory weights the first more heavily, so anything already in your possession is defended more fiercely than the identical thing would be pursued. Ownership was the consequence of the mechanism, not its premise.
The version most people know, that merely holding a thing makes you value it more, arrives ten years later.
Kahneman, Knetsch and Thaler built it in the Journal of Political Economy in 1990. Half a room is given a mug, the other half is not, and the room then trades. If ownership does not move valuations, about half the mugs should change hands, because the mugs were handed out at random and the two halves of the room have the same distribution of taste. In their first experiment the ratio of actual to predicted trading volume was 0.20 for mugs and 0.41 for pens. Across the incentivised consumption-goods markets it averaged 0.31, and median selling prices ran at more than double median buying prices.
The control is the part that makes the paper hard to dismiss. In the same sessions, with the same people and identical trading rules, they ran markets in induced-value tokens, where each participant was simply told what a token was worth to them. Those markets hit the predicted volume exactly. Whatever was suppressing trade in mugs was not transaction costs, was not confusion about the rules, and was not inexperience.
They then added a third role. A seller holds a mug and names a price. A buyer holds cash and names a price. A chooser holds nothing and is asked, for each price, whether they would rather have the mug or the money. The chooser and the seller face economically identical terms. They differ only in who is holding the object. Median valuations came out at $7.12 for sellers against $3.12 for choosers, and in a repeat with 117 students at the University of British Columbia, $7.00 against $3.50. Income effects cannot explain a gap between two positions that are the same position.
That is the case for. The case against is procedural, and it is not a fringe objection.
Plott and Zeiler argued in the American Economic Review in 2005 that the gap is an artefact of how experimenters ask the question. Subjects, on their account, misunderstand the elicitation mechanism, and the standard designs do not control for that misunderstanding. Their own procedure trained participants extensively on a modified Becker-DeGroot-Marschak mechanism, guaranteed anonymity, and ran practice rounds before eliciting any valuation. Under those conditions they report no gap at all. In 2007 they applied the same argument to the exchange design and found the asymmetry disappeared once the procedural signals were removed.
Andrea Isoni, Graham Loomes and Robert Sugden replied in 2011, in the same journal, that Plott and Zeiler's own unpublished data from the same study showed a significant and persistent gap when the identical procedures were applied to lotteries rather than mugs, and reported new data confirming it. Plott and Zeiler replied in the same issue. Nobody has conceded, and anyone who tells you the endowment effect is either established or debunked is picking a side in an argument that is still running.
Two later results move the ground under both camps. Tuba Tunçel and James Hammitt's 2014 meta-analysis in the Journal of Environmental Economics and Management finds the disparity is not solely an artefact of hypothetical transactions or weak method, that it is smaller for ordinary private goods than for public and non-market goods, that it shrinks where subjects have real market experience with the good, and that, contrary to the earlier Horowitz and McConnell review, it is significantly smaller under incentive-compatible elicitation. Dietmar Fehr and Dorothea Kübler took the exchange design out of the laboratory entirely and into the German Socio-Economic Panel, giving respondents either a towel or a notebook as a thank-you and offering a swap at the end of the interview. The trading rate was 32 per cent against the 50 per cent that no endowment effect predicts. Their second finding is the one to sit with: the trading behaviour had no relationship with the respondents' measured loss aversion.
So the phenomenon appears outside a psychology laboratory, its size depends heavily on the good and on the procedure, and the loss-aversion explanation of it is in worse condition than the phenomenon it explains.