A Raffle Game


The endowment effect (EE) is the finding that when given an item, people have a tendency to value it more than an equally-desirable item that they do not own. The phenomenon was originally thought to be caused by loss aversion as a reference-point effect; however, recent work has argued against this and has suggested that it is instead caused by ownership per se.

Research Questions / Hypotheses

The present study attempted to resolve this controversy by testing the prediction that if the EE is just a reference-point effect then it should be possible to elicit a comparable effect without an actual endowment, just by providing a salient reference point. Specifically, we sought to explore whether the EE can a) be replicated in a purely online setting; and b) be elicited without actual endowment, so is better conceptualised as a reference-point effect. It was hypothesised that for both our endowment and no-endowment conditions, when presented with a choice between two raffle tickets, most people would choose the ticket corresponding to their initial raffle as opposed to choosing the ticket that would enter them into the alternative raffle. If the EE is truly a reference effect, there should be no significant difference in choices between these two conditions. In addition, we predicted that participant’s valuation of the ticket associated with the initial raffle would be greater than participant’s valuation of the ticket associated with the alternative raffle in both endowment and no-endowment conditions.


201 REP participants were recruited for this study. Two participants were excluded due to providing an incomplete response, and selecting both tickets as preferred when requested to select only one ticket (making N = 199). Age, gender and ethnicity were not recorded.


Materials: The experiment was administered online via Qualtrics. Participants could complete the online survey on any personal computing device, including desktop computers, laptops, tablets and smartphones    Procedure: People were randomly assigned to one of two conditions, endowed or non-endowed. Participants in the endowed condition were initially presented with either a purple or a green ticket, which entered them in the raffle for the correspondingly coloured hamper. Both hampers were visible on participants’ screens. They were informed that ‘this is your ticket. We are giving it to you. Because it is [purple/green], it enters you in the raffle to win the [purple/green] hamper of food. Both the purple and green hampers are valued at $50. All tickets have an equal chance of being drawn in the raffle. If you win, you will be contacted via your UniMelb student email and receive the corresponding hamper of goods.’ Participants were then asked to look carefully at both hampers, and informed they were currently entered in the draw for the purple/green hamper, but if they preferred, they could be entered into the draw for the green/purple hamper instead. They were then presented with both tickets, and asked to select which one they would like to enter the draw with.    After choosing the ticket to enter the hamper draw with, participants’ valuation of this ticket was measured by giving them a short list of prices ranging from $0.50 up to $10 in intervals of $0.50. For each price participants were asked to indicate whether they would exchange their chosen ticket and switch to the other ticket (and therefore hamper raffle) if they were paid the respective amount. This measured the explicit value of the chosen ticket in comparison to the other ticket. Participants were informed that at the end of the experiment, for two randomly selected people one of the listed prices would be drawn and enacted. Depending on their response to this price they either had to enter the hamper raffle associated with their chosen ticket, or received the respective amount of money and had to enter the other raffle with the other, non-chosen ticket. This well-established price elicitation method determines the smallest amount of money necessary to make a participant switch from the chosen ticket to the other ticket. This measure is conceptually similar to the classic willingness-to-accept-measures. After completing the pricelist, participants were debriefed and granted with REP credits.    The procedure for non-endowed condition was identical to that of the endowed condition, with the exception that participants were not initially presented with a raffle ticket. Instead, they were simply told ‘you are currently entered into the draw for the [purple/green] hamper.’ Again, participants were asked to look carefully at both hampers, presented with both tickets, and asked to indicate which ticket they would like to enter the raffle with. The explicit valuation task, debriefing and credit allocation followed.    Order effects of both tickets and hampers were accounted for by counterbalancing the order in which the coloured tickets and hampers were presented on screen.


Choices: As participants were asked to choose between the ticket corresponding to the raffle they were initially assigned to (which was the ticket they were initially given in the endowment condition) or the ticket corresponding to the raffle for the other hamper, frequency data of participants’ choices was recorded. Participants were categorised as either keepers, who chose to stick with their initial ticket or switchers, who parted with their initial ticket and opted for the alternative ticket. This frequency data was then subject to a chi-squared analysis, and converted into percentages for further analysis.    In the endowment group, 65 out of 100 (65%) participants were keepers, a proportion that was significantly greater than 50% (p = .002), whereas in the no-endowment group 57 out of 99 (58%) participants were keepers, a proportion that was not significantly different from 50% (p = 0.13). These proportions were not significantly different from each other (X2(1, N = 199) = 1.16, p = .28). Thus, an EE was found only in the endowment condition. The fact that the difference between conditions was not significant suggests that this test is underpowered. We therefore analysed the price valuation data to see if we could obtain clearer findings.    Price Valuation: In addition to choosing a ticket, participants were then asked to provide an explicit valuation for this ticket, specifying the amount of money they would be willing to accept in order to give up their chosen ticket and enter the alternative hamper raffle with the other ticket instead. That is, participants who chose the ticket corresponding to the initial raffle indicated how much they would have to be paid to accept the other ticket instead. Participants who chose the other ticket indicated the amount of money they would need in order to accept the ticket corresponding to the initial raffle. Two sets of analyses were conducted, the first testing whether the mean WTA price was greater than zero within each condition, the second comparing the mean WTA prices between the two conditions.    We coded the valuations using the procedure of Chapman (1998). If participants were keepers, their indicative price was coded as a positive amount; if participants were switchers, their indicative price was coded as negative . Therefore, for each condition (endowed v. non-endowed) a series of values was recorded (such as +0.50, +4.00, –2.50, –7.00; with positive values indicating the participant was a keeper, and negative indicating switchers), providing a continuous measure of preference for the endowed ticket such that higher positive prices indicated a stronger preference for the endowed ticket. Within each condition, participants’ WTA prices were averaged across all participants (ie. averaging over both keepers and switchers) to get a single price value. An independent samples t-test was then conducted, comparing the mean WTA price to zero. Addition, the mean WTA prices for the endowed and non-endowed conditions were compared via a third independent samples t-test.    Independent-samples t-tests were conducted to compare the explicit price valuations for each condition and between conditions. The mean valuation in the endowment condition was $1.48 (SD = $5.09), which was significantly different from zero (t(99) = 2.91, p = .0045, 95%CI[0.47 2.49]). For the no-endowment condition, the mean valuation was $0.06 (SD = $4.91), which was not significantly different from zero (t(98) = 0.12, p = .90, 95%CI[-0.92 1.04]. The valuations in the two conditions were significantly different from each other (t(197) = 2.00, p = .047, 95%CI[0.02 2.81]. So, again, we found an EE only in the endowment condition but now, using this more sensitive analysis, we find a significant difference between conditions, which is what we would expect as we did not find a significant EE in the no-endowment condition.


The present research aimed to extend upon and refine Kogler et al.’s (2013) study in an attempt to clarify the cognitive processes involved in eliciting the EE, and to explore whether the EE may be better conceptualised as a reference-point effect (Pryor et al., 2018). Utilising a newly designed lottery paradigm, the present study was able to elicit the EE in a purely online domain through merely assigning ownership of a proxy good (ie. a ticket). To our knowledge, no such study has successfully stretched the concept of ownership to this extent in an online setting, thus these findings contribute to online EE literature in an exciting and novel way, and carry significant implications for electronic commerce (henceforth referred to as e-commerce).    The prediction that the EE is just a reference-point effect was not supported, however, with no such ‘endowment’ effect observed in the no-endowment condition. That is, we found no significant difference between the amount of people choosing to keep versus switch their initial ticket (and therefore hamper) in the no-endowment condition, suggesting that the EE does indeed require some level of endowment or perceived ownership in order to be elicited.    Collectively, these findings indicate that although some degree of endowment is necessary to generate the effect (at least in an online setting), this degree can be quite small –even the virtual bestowal of a proxy good, representing a mere possibility to own a good is sufficient enough to induce the effect.