Behavioral Econs 101: The endowment effect and loss aversion Part 2 – Experimental economics in action
In this fourth installment of our Behavioral Econs 101 series(for parts 1, 2 and 3, see here, here and here) we continue our discussion of the endowment effect. In Part 3, we explained how Prospect Theory helps us understand why losses feel so much more painful than an equivalent gain feels good. Prospect Theory offers one way to better understand why we are reluctant to give up something we already own, even if we had obtained the item for free or at a heavily discounted price. Equipped with insight into how a difference between the willingness to pay(WTP) and the willingness to accept(WTA) transaction prices can arise, we now turn our attention towards examining the actual laboratory experiments that were designed to first study the endowment effect. Along the way, we’ll also(like any good economist should) discuss some of the criticisms that have been leveled at this approach to studying the endowment effect.
by HK Lim
Knetsch and Sinden’s experiment: $2 or lottery tickets?
In Part 3 in our Behavioral Econs 101 series, we explained the phenomenon where people ascribe more value to the items they already own via the endowment effect. Jack Knetsch and John Sinden[1] designed and conducted a terrifically simple experiment with the aim of measuring the endowment effect. In their experiment, half of the subjects were selected at random and given $2 while the other half were given lottery tickets. In each experiment, the winner of the lottery would receive his or her choice between $50 in cash or $70 in vouchers that were redeemable at a local bookstore. After allowing for an interval during which the subjects were given another activity, each group was then given a choice. Those who had not received a lottery ticket were told they could now purchase one for $2, whereas those who were already holders of lottery tickets were now told that they could sell theirs for $2.
What you should notice here(and this is really the key insight) is that both groups are really being asked the same question, that is: “Which would you prefer, the lottery ticket or $2?” Now before reading on, take a quick guess about how standard economic theory would answer this question?
What you should notice here(and this is really the key insight) is that both groups are really being asked the same question, that is: “Which would you prefer, the lottery ticket or $2?”
What was your guess? Well, according to standard economic theory, there really should not be any difference whether a subject first received a lottery ticket or $2(those familiar with the Coase Theorem would understand the deeper implications of this point). If the lottery ticket is valued at more than $2 by a participant, he/she should ultimately end up with one, otherwise if the ticket is valued at less than $2, the subject should end up with $2 instead. So what was the outcome of the experiment? Well, the experiment’s results handily rejected this prediction! Those who started with a lottery ticket overwhelmingly opted to keep it (83%), while only a minority of those who started out with the money opted to buy a ticket (38%). So, people were more likely to keep what they started with than they were to trade it for something else. This was completely consistent with the predictions of the endowment effect(and relatedly, loss aversion but we need to be careful here). Note that this remained true even when the initial allocations were made at random. This was as clear a result as could be expected and flat out contradicted conventional economic thinking.
Among the criticisms of this original experiment[2], the foremost objection was that the experiment’s participants might have been inadvertently confused by the one-off nature of the experiment. The reasoning was that the results would be very different had participants been allowed the opportunity to learn from participating in repeated iterations of this same experiment. The second objection relates to the criticism that the participants were immersed in an artificial and contrived “exchange environment” and this observed endowment effect would dissipate in an actual market context. In this latter case, it was argued that fluctuations in buyer’s and seller’s trading prices would serve to “discipline” participants into behavior that was more consistent with standard economic theory. The third criticism was that the stakes were low in this experiment, and if the financial stakes were much higher, this effect would be greatly weakened.
Of college mugs and tokens
To address these criticisms, Daniel Kahneman and Richard Thaler worked with Jack Knetsch[3] to design a new watertight experimental setup to test for the endowment effect. The key feature of the endowment effect that they set out to demonstrate was the prediction that it would reduce the volume of exchange occurring in the market.
The key feature of the endowment effect that they set out to demonstrate was the prediction that it would reduce the volume of exchange occurring in the market.
Compared to Knetsch and Sinden’s original experiment, they made two significant changes. First, they included an actual market as part of the experiment’s design and second, they directly incorporated Vernon Smith’s concept of induced value as an explicit experimental device. With the induced value methodology, participants buy and sell tokens that have no value outside of the laboratory setting. Participants are separately assigned a unique personal value for their tokens that can be redeemed only at the end of the experiment. This allowed for a gradated range of values for each token to be assigned to participants. The induced value experimental device was utilized because it was not expected that anyone would exhibit an endowment effect in the use of a token any more than they did when using a $10 bill for example. Using tokens in this first stage of the experiment served as a control for the later versions of the experiment.
I. How the “market” works
To better understand Kahneman, Thaler and Knetsch’s setup, let us briefly describe how their “market” works(See [2] for more details). Starting with 12 participants, we randomly assign to them induced token values from 25 cents to $5.75, in increments of 50 cents. Six tokens are then randomly distributed among the participants and a market is implemented by asking participants who had received a token to answer a series of questions as follows:
At a price of $6.00 I will sell_____ I will not sell_____
At a price of $5.50 I will sell_____ I will not sell_____
At a price of $5.00 I will sell_____ I will not sell_____
etc
The lowest price at which a participant is prepared to part with their token is referred to as their reservation price. For example, someone with an induced token valuation of $5.25 would be willing to part with a token at a price of $5.50 but not for $5, hence his/her reservation price would be $5.50. In the same manner, potential buyers of tokens would answer a similar series of questions but this time asking about their willingness to buy a token(over the exact same range of prices). With this information from participants, the supply and demand curves for this market can be constructed and a market-clearing price can be determined.
II. According to economic theory
According to (neoclassical) economic theory, if the market functions well, the six participants valuing the tokens most should end up owning all six tokens. In the actual implementations of this experiment, six participants were chosen at random(three buyers and three sellers) after the induced market was constructed(and demand and supply curves identified) and were paid according to the preferences that they had stated on their forms as well as the market-clearing price for that particular run of the experiment. In each of these experiments, the market-clearing price was determined by supply and demand and the number of transactions(volume) was found to be within a unit of that predicted by economic theory. In this version, transaction volumes were found to be within a unit of 3 transactions which is half the number of initial tokens allocated, which is exactly as predicted. This suggested that participants understood the task well and the market mechanism did not impose high transaction costs that impeded exchange so far as this experimental setup was concerned.
III. Enter the college mugs
The next stage of the experiment involved using college coffee mugs(purchased at $6 from the Cornell bookstore). With 44 participants and 22 mugs, the mugs were again randomly allocated among the participants, and all participants(mug owners and non-mug owners) were allowed to examine the mugs for a short while before the actual exchange portion of the experiment began. In this stage of the experiment, four induced markets were conducted with participants listing their willingness to buy or sell the coffee mugs depending whether they were or were not in possession of one. In each trial, the buyers and sellers of mugs always remained the same(to allow for participant learning). Participants were told that only one of these four trials would be binding and the trades from that trial would be executed. Participants were however not told ahead of the trials which actual trial would count.
The same experimental setup was repeated for boxed college ballpoint pens, this time with the sticker price of $3.98 clearly visible. Again, four trials were conducted with again one to count, similar to the mugs scenario, but this time with the buyers of mugs switching to become potential sellers of pens.
IV. Theory versus experiment
Economic theory predicts that for either of these markets, since transactions costs are insignificant and income effects are trivial, when the market clears, the mugs or the pens should typically end up in the hands of the participants that value them the most. Since the mugs and pens were allocated at random, on average, the transaction volume should be close to half the number of mugs or pens available, that is 11 transactions for 22 mugs(or pens). In the four experimental trials for the mugs, the transactions numbered 4, 1, 2 and 2 and in the markets for pens, the average transactions numbered 4 or 5, all well below the predicted 11 transactions! When the trial data was analyzed, the low volume of trade was related to the differing reservation prices of buyers and sellers. The median mug owner was unwilling to sell a mug for less than $5.25 whereas the median buyer was unwilling to pay more than $2.25-$2.75. The market clearing price thus varied between $4.25 and $4.75. The ratio of selling to buying prices was close to 2 and this was also true for the markets for pens(Recall how this gels with the predictions of Prospect Theory and loss aversion from Part 3). Even with repeated experiments, these observations held true with median selling prices approximately twice buying prices and trading volume about half that predicted by economic theory. This was the endowment effect at work.
V. Crossing your t’s – Are we observing a reluctance to sell or to buy?
For completeness, it is also necessary to examine if lower trading volumes might have arisen more due to a reluctance to sell or to buy. To examine this point, another series of experiments was conducted, this time with participants from Simon Fraser University(SFU). Here, 77 students were randomly assigned to three different groups. In the first of these groups, called the sellers, participants were given SFU coffee mugs and asked if they would sell the coffee mugs at various prices between $0.25 to $9.25. The second group, the buyers, were asked if they were willing to buy a mug for the same set of prices. The third group, the choosers, were not given mugs but were asked whether they would prefer to receive a mug or cash for the same set of prices. Notice that the sellers and choosers are effectively in identical situations, having to decide if they would prefer to own a mug or instead receive money.
Surprisingly, the choosers behaved more like buyers than sellers. With the median reservation prices observed to be $7.12 for sellers, $3.12 for choosers and $2.87 for buyers! Thus, it was concluded that the low trading volume was mainly related to an owner’s reluctance to part with his/her endowment(the mug) rather than a buyers’ unwillingness to part with their cash. This cleverly designed experiment also had another important advantage in eliminating the possibility of a dominant income effect playing a role in the outcomes of the previous version of the experiment since the sellers and choosers were effectively in the same situation.
Aside: Implications for the microeconomic theory of consumer behavior
In microeconomics, it is assumed that indifference curves do not intersect so as to ensure that a well-defined preference ordering exists to facilitate optimizing of choice bundles. This non-intersection criteria is associated with the idea of reversibility of indifference curves, in that if an individual holds x and is indifferent between trading it for y, the reverse should also be true.
Knetsch[4] experimentally demonstrated how this can be violated in the presence of loss aversion. Using pens, depending on whether the participant started out with pens or money, each participant was asked how much he would require to sell a pen or how much money he would pay to for a pen. In this manner, indifference curves in different individuals could be shown to intersect if one started out with pens and the other with money(See Fig.1 below).
That the indifference curves derived above are starkly different is suggestive of the main result. The pens are worth more to those who start out with pens than to those who started with money. While this particular situation examined indifference curves arising from different individuals, it does offer evidence of a contradiction of the assumption of representative agent reversibility in aggregate.
VI. Does the endowment effect enhance the appeal of a gift one owns?
One more remaining question remains to be answered: does the endowment effect enhance the appeal of a gift one already owns? To paraphrase, “do subjects endowed with a gift value it more than those who do not receive it?” This question was examined in an experiment by Loewenstein and Kahneman[5]. In their experiment, half the student participants were given pens while the other half were given tokens redeemable for an unspecified gift. The participants were then tasked with ranking the attractiveness of six gifts as potential prizes for a future experiment. In the final stage, participants were asked to choose between a pen and two chocolate bars. As before, there was a marked endowment effect with participants endowed with a pen choosing it 56% of the time, but only 24% of the other participants chose a pen! On the other hand, when making rankings on the basis of attractiveness, the participants who were endowed with pens did not rank them as more attractive. This implied that the endowment effect worked more by enhancing the pain of giving something up as opposed to enhancing the appeal of the gift.
Conclusions and some brief criticisms
The series of experiments we have described illustrate how the endowment effect can play a role in our willingness to give up something we already own(however recently obtained). This is typically understood to be related to loss aversion(see part 3 of our Behavioral Econs 101 series here) and works more by enhancing(or exacerbating) the pain of giving something up. There have been numerous criticisms of this approach in the literature. Without going into too much detail, let us point the interested readers to three major criticisms. The first is Haneman’s[6] approach that attempts to explain these findings via a neoclassical lens without a need to utilize prospect theory. However, as we described above, Kahneman, Knetsch, and Thaler(1991) find that the endowment effect persists even when one fully controls for wealth effects. The other criticism we highlight is attributable to Shogren, et al.[7] which noted that the experimental technique used by Kahneman, Knetsch and Thaler(1990) to demonstrate the endowment effect in fact created a situation of artificial scarcity. By performing a more robust experiment with the same goods used(chocolate bars and mugs), they then found little evidence of the endowment effect. Thirdly and lastly, we have also been examining if a trading bid-ask book that permits participants to revise their WTP and WTA as the experiment is in progress will weaken the endowment effect. This would be extremely helpful in helping clarify the persistence of the endowment effect and loss aversion in actual market trading environments.
References
1. Knetsch, Jack L., and Sinden, John A. (1984). “Willingness to Pay and Compensation Demanded: Experimental Evidence of Unexpected Disparity in Measures of Value.” Quarterly Journal of Economics 99, no. 3: 507-21.
2. Knez, Peter., Smith, Vernon L., and Williams, Arlington W. (1985). “Individual Rationality, Market Rationality, and Value Estimation”. American Economic Review. 75 (2): 397–402.
3. Kahneman, Daniel, Knetsch, Jack L., Thaler, Richard H. (1990). “Experimental Tests of the Endowment Effect and the Coase Theorem”. Journal of Political Economy. 98 (6): 1325–1348.
4. Knetsch, Jack L. (1990) “Derived Indifference Curves,” working paper, Simon Fraser University.
5. Lowenstein, George, and Kahneman, Daniel. (1991) “Explaining the Endowment Effect,” working paper, Department of Social and Decision Sciences, Carnegie Mellon University.
6. Hanemann, W. Michael (1991). “Willingness To Pay and Willingness To Accept: How Much Can They Differ? Reply”. American Economic Review. 81 (3): 635–647.
7. Shogren, Jason F.; Shin, Seung Y.; Hayes, Dermot J.; Kliebenstein, James B. (1994). “Resolving Differences in Willingness to Pay and Willingness to Accept”. American Economic Review. 84 (1): 255–270.