When Choosing Implies Losing: Does Flipping a Coin Increase Forfeiture Thoughts?

Coin flips are a common type of random device that can aid decision making. Other random decision-making aids are, for example, rolling a die, spinning a roulette or lottery wheel, using a counting-out rhyme, or picking petals from a flower. The need for random suggestions to help decision making has been recognized by the tech industry and a variety of apps are available for this purpose (e.g., “Decision Roulette” or “Tiny Decisions”). These devices allow a clear suggestion to be obtained, which, despite being random, may influence the decision-making process.

A different group of decision-making strategies that individuals can turn to is asking friends, family, or an expert. Similar to using a randomizer, receiving advice from a person may result in a clear suggestion. Different from a randomizer, however, other people might have an agenda when giving advice. While the advice may be well-intended and helpful, it is not random and recipients could wonder about potential motives. The social context may further change the likelihood that individuals follow the advice, for example, if they feel obliged towards the advice-giver.

Prior research has studied coin flips in a variety of settings, including fairness-related decisions (Keren & Teigen, 2010), personal decisions (Levitt, 2020), prosocial requests (Lin & Reich, 2018), or moral dilemmas (Gordon-Hecker & Olivola, 2019). Choosing randomized outcomes can signal indecisiveness or perceived indifference (Dwenger et al., 2019), yet lotteries can also be effectively used to allocate tasks, resources, or burdens (Elster, 1987), especially when fairness is important (Keren & Teigen, 2010). This reliance on randomizers is not confined to limited circumstances. Indeed, in one study, 46% of participants sampled from the general public had used a coin flip to make a decision before, suggesting that almost every second participant had first-hand experience (Jaffé et al., 2020).

Oftentimes, individuals flip the coin with the intention to follow its outcome, therefore using the coin as a “decider.” This allows for clear-cut decisions, and individuals may reduce perceived responsibility for the outcome (Steffel et al., 2016). In contrast to the decider function, recent research has investigated consequences of flipping a coin as a “catalyst,” meaning that the coin flip does not determine the outcome, but affords a different perspective, for instance, by triggering stronger affective reactions (Jaffé et al., 2019) or lowering the need for further information before making a final decision (Douneva et al., 2019). A recent survey showed that 82% of participants who had flipped a coin prior to deciding also indicated that they had also used a coin flip and experienced feelings; moreover, 65% of participants indicated that they had used a coin flip and then suddenly knew what they really wanted (Jaffé et al., 2020). These numbers illustrate that individuals are well acquainted with the catalyst phenomenon and apply this strategy in everyday life.

In this manuscript, we hypothesize that flipping a coin may shift individuals’ perspective from an acquisition to a forfeiture scenario. This hypothesis is based on a combination of different streams of research, including the notion that a coin flip may render a decision as if decided and hence more concrete. We complement these theoretical considerations with an example outlined in Table 1.

In a situation where individuals need to choose between options, which are not yet owned (acquisition scenario; see Dhar & Wertenbroch, 2000), individuals may ask themselves which of the options they would prefer. While in acquisition situations thoughts revolve around obtaining something, we suggest that coin flips and other random devices may trigger the realization that individuals will also forgo something. By highlighting that choosing one option means giving up the other, the coin flip may move individuals from an acquisition perspective to a perspective that includes forfeiture thoughts. This change in perspective may be particularly salient given that people usually do not consider what other options they will lose or give up when considering a particular option—a phenomenon known as opportunity cost neglect (Frederick et al., 2009).

One reason why this change in perspective may occur, is that receiving a clear suggestion renders the potential ownership of one option less hypothetical. This reduction in psychological distance likely triggers a more concrete mindset (Liberman & Trope, 2008; Trope & Liberman, 2010), bringing details into focus. A more concrete mindset is associated with prevention (Lee et al., 2010), disadvantages (Eyal et al., 2004), and loss frames (White et al., 2011), all of which are in the conceptual vicinity of forfeiting.

A more concrete mindset and focus on details may further result in a closer consideration of the options available. This closer consideration “may induce attachment to the options” (Carmon et al., 2003, p. 15), which could increase attachment and feelings of prefactual ownership. Prefactual ownership, again, could make the decision scenario more similar to a forfeiture scenario, where individuals have to decide which of two previously received options they want to give up (Dhar & Wertenbroch, 2000). In such a situation, opportunity costs could become more salient, and as a result, individuals who flip a coin might have more forfeiture thoughts compared to individuals that do not flip a coin.

These different streams of prior research converge in the prediction that coin flips and other random devices may generally increase forfeiture thoughts. We complement this main effect hypothesis with the following more specific considerations:

Differential effects when a coin suggestion is congruent versus incongruent with preferences. In situations where initial preferences exist, the outcome of a coin flip may be congruent or incongruent with these preferences. The incongruent case appears to be particularly interesting: If a preference for option A is opposed by a suggestion for option B, a conflict arises and forfeiture thoughts may be particularly strong. Receiving an incongruent suggestion could feel unsettling for individuals, as it poses the immediate question of whether they should reject the suggestion, which could then increase deliberation about the choice options (Sokolova & Krishna, 2016). An incongruent suggestion might further symbolize conflict, highlight losses, and trigger the need to check which option would be associated with smaller losses, again resulting in more forfeiture thoughts.

In contrast to the incongruent case, a congruent suggestion could make individuals feel validated and confident in regards to making the decision. We hence assume that more forfeiture thoughts arise when the coin flip’s suggestion is incongruent (vs. congruent) with individuals’ initial preferences.

Differential effects depending on the coin’s suggestion. While the coin may generally increase forfeiture thoughts (compared to a control group), it is interesting to speculate about differential effects as a function of the coin’s suggestion. On the one hand, the specific suggestion could result in more forfeiture thoughts for the not suggested option, because this is the option that individuals would forfeit when simply doing what the coin suggests. On the other hand, once individuals are in a forfeiture mode and have adopted a more concrete mindset, forfeiture or loss-related thoughts may also pertain to the option suggested by the coin. This would result in more forfeiture thoughts regarding both the not suggested and the suggested option. This alternative prediction is consistent with Kahneman and Tversky’s work (1984) on losses looming larger than gains, and a loss framework making it more difficult for individuals to give up either one of the choice options (Park et al., 2000). Put differently: Once potential losses (forfeiting something) have entered the picture, individuals may mentally simulate ownership for the suggested and the not suggested option, and accordingly simulate forfeiting both options, thus increasing forfeiture thoughts irrespective of the coin’s suggestion. Given that alternative predictions are viable with respect to the effects of coin suggestion (including the null hypothesis of no difference), we test the notion of coin suggestion in an exploratory manner only.

Potential impact of the coin on acquisition thoughts. It is interesting to speculate that using a coin flip may not only result in more forfeiture thoughts, but also impact acquisition thoughts. On the one hand, it seems likely that an increase in forfeiture thoughts goes hand in hand with a decrease in acquisition thoughts. However, although a coin flip might change the decision setup to resemble a forfeiture scenario, differences to a full forfeiture decision setup remain. In the classical forfeiture decision, both options are owned and one has to be given up (see Dhar & Wertenbroch, 2000). In the coin flip condition, only one option is suggested, and individuals decide whether they want to give up and forfeit this option in order to acquire the other one. This could result in a situation where acquisition thoughts are present and eventually strengthened, too, because forfeiting one option also goes hand in hand with acquiring the other.

H1: Flipping a coin (compared to not flipping a coin; control group) increases forfeiture thoughts generally, that is, for both choice options.

H2: In situations where individuals receive a coin flip suggestion that is incongruent (vs. congruent) with their initial preference, individuals report more forfeiture thoughts.

H3: In situations where individuals receive a coin flip suggestion that is congruent (vs. incongruent) with their initial preference, individuals feel validated and therefore more confident that they will make a good decision.

H4: Individuals report more forfeiture thoughts for the option they forfeit when following the coin’s suggestion (i.e., the not suggested option) compared to the option suggested by the coin.

H5: We speculate that flipping a coin (compared to not flipping a coin; control group) could also impact acquisition thoughts for the choice options. Given that one may speculate about both an increase and a decrease of acquisition thoughts, we examine acquisition thoughts in an exploratory fashion.

In this registered report, we present two laboratory studies (total N = 310), which we complement with a third preregistered study to replicate and extend the previous results. Testing Hypothesis 2 (comparing congruency vs. incongruency in the coin-condition) required a smaller study design (two groups) whereas testing Hypothesis 1 (comparing a control to a coin-congruent to a coin-incongruent condition) requires a larger study design (three groups). For reasons of feasibility, we started with the smaller design and tested Hypothesis 2 first, that is, focused on the effects of a congruent versus incongruent coin flip outcome on forfeiture thoughts in Study 1. Next, we implemented the larger study design in Study 2 that included a control group and tests both Hypotheses 1 and 2 (and informs on Hypothesis 3). Both Studies 1 and 2 provide information in regards to Hypothesis 4.

In Studies 1 and 2, participants were asked to choose between a small package of Smarties and a Snickers bar,¹ from now on referred to as snack A and snack B, respectively. Asking participants about a consumption-related decision with snacks allowed us to create a realistic decision scenario in the lab, because participants were able to factually receive the option they chose. Moreover, consistent with prior research suggesting that people are less reluctant to flip coins for less (vs. more) important decisions (Keren & Teigen, 2010), we reasoned that participants would be less sceptical about a coin flip when we use snacks compared to more important decisions or dilemmas with negative outcomes only.

Before choosing (but after indicating their preference), participants in the coin condition were shown a coin flip and its outcome suggesting one of the snacks. In all studies, participants were randomly assigned to flip a coin or not.² Although individuals report hesitation in regards to flipping a coin, especially for personal decisions (and might not choose to do so in real life, see Elster, 1987), previous research did not reveal systematic differences in participants’ affective experiences when looking at the coin outcome depending on their willingness to flip it in the first place (Jaffé et al., 2020; Studies 2a and 2b). Throughout all studies, the coin flip was virtual and integrated into the study to ensure that all participants would receive a random suggestion that could be automatically recorded and would not be subject to potential reporting biases (cf. research on asking participants to flip a coin and report the outcome to detect deceptive behavior, e.g., Abeler et al., 2014; Bryan et al., 2013; Schindler & Pfattheicher, 2017).

After the virtual coin flip, participants indicated the extent to which they were thinking about forfeiting either snack. We assessed forfeiture thoughts before participants made their decision to ensure that forfeiture thoughts were not overpowered or obscured by post-decisional cognitions (e.g., dissonance reduction, Festinger, 1962; Shultz & Lepper, 1996; or regret aversion, Zeelenberg et al., 1996). The data, code, and materials³ can be accessed via the following repository: https://osf.io/hr3b9/.

In Study 1 we tested whether receiving an incongruent versus congruent suggestion results in more forfeiture thoughts (H2). We furthermore assessed whether individuals report more forfeiture thoughts for the not suggested versus the suggested option (see H4).

Participants and design. One hundred fifty-one individuals participated in this study, which was conducted in university buildings using tablet computers. From this sample, we excluded three participants because of their self-indicated participation in previous studies using a similar setup, one indicating explicit reasons why they wanted their data to be excluded, and two self-identifying as not having completed the study carefully (a value of ≤ 4 on a scale from 1 to 9). This resulted in a sample size of 145 participants (84 female, 58 male, 3 no information; M_age = 22.26, SD_age = 4.08). A sensitivity analysis conducted with G*Power (Faul et al., 2007) suggests that this sample is sufficiently large to detect effect sizes of f ≥ .12 with standard criteria (α = .05, power = .80, correlation among repeated measures 0.5). Participants received the chosen snack as compensation.

Participants were randomly assigned to one of two between-subjects conditions: For half of the participants, the coin suggested choosing snack A, for the other half snack B. We determined whether participants had received a suggestion that was congruent versus incongruent with their initial preference and used this (in)congruency as a between-subjects independent variable. Coin suggestion (suggested versus not suggested option) served as a within-subjects independent variable. Forfeiture thoughts served as the dependent variable.

Materials and procedure. Participants were welcomed to a study on decision making. After providing informed consent, participants learned that the study focused on snacks and that we were interested in their preference between two snacks: a package of Smarties and a Snickers bar. Participants indicated their preference by clicking on one of the two options and were then introduced to the coin flip. We explained that a coin flip can sometimes help in making decisions and that we would like them to try flipping a coin in this study. Participants could try the coin flip several times to ensure that they would believe its outcome to be random. The two snack options were then displayed again, the coin was virtually flipped, and it recommended one option by appearing above snack A or B. On the next page, participants were then asked about their forfeiture thoughts with respect to the two options: “To what extent did you think about forfeiting the option suggested by the coin [placeholder for either snack A or B, depending on condition]?” and “To what extent did you think about forfeiting the option not suggested by the coin [placeholder for the other snack]?” Answers were recorded on a Likert scale (ranging from 1 = not at all to 7 = very much). Demographics were assessed and participants were thanked for participating. Before leaving the room, participants picked one of the two snacks as compensation for their participation (this final choice was not recorded, because it was not the focus of this study).

Overall preferences. Seventy-five participants (52%) preferred snack A, while seventy participants (48%) preferred snack B. Cross-tabulated with the coin’s (random) suggestion, this resulted in a cell distribution of n = 34 (n = 29) preferring snack A (B) and receiving a congruent suggestion, and n = 41 (n = 41) preferring snack A (B) and receiving an incongruent suggestion.

Analyses. To test Hypothesis 2, we computed a mixed 2 x 2 ANOVA with (in)congruency as a between factor and coin suggestion as a within factor. Forfeiture thoughts served as the dependent variable. The main effect of coin suggestion, F(1, 143) = 0.01, p = .916, η²_p = .00, and the (in)congruency x coin suggestion interaction, F(1, 143) = 0.00, p = .950, η²_p = .00, were not significant. However, there was a significant effect of (in)congruency, F(1, 143) = 29.16, p < .001, η²_p = .17: Participants reported more forfeiture thoughts when the coin suggestion was incongruent (vs. congruent) with their initial preference, M_incongruent = 3.77, SD = 1.18, M_congruent = 2.66, SD = 1.30. To illustrate that this difference was significant for both levels of coin suggestion, we conducted separate one-way ANOVAs. For both the suggested and not-suggested option, receiving an incongruent compared to a congruent suggestion resulted in more forfeiture thoughts: Suggested option: M_incongruent = 3.77, SD = 2.13 vs. M_congruent = 2.63, SD = 1.93, F(1, 143) = 10.92, p = .001, η² = .07; not suggested option: M_incongruent = 3.78, SD = 2.15 vs. M_congruent = 2.68, SD = 2.12, F(1, 143) = 9.42, p = .003, η² = .06.

Study 1 investigated whether congruency versus incongruency between initial preference and suggestion impacts forfeiture thoughts. Consistent with Hypothesis 2, the coin flip elicited more forfeiture thoughts for both snack options when the coin’s suggestion was incongruent versus congruent with individuals’ preferences.

We tested the additional assumption that participants might report more forfeiture thoughts for the not suggested versus the suggested option (see H4), but did not find support (the main effect for coin suggestion was not significant). Note that this null effect may be interpreted as being consistent with findings suggesting that a loss framework makes it more difficult for individuals to give up either one of the choice options (Park et al., 2000). Individuals may mentally simulate forfeiting both the not suggested and the suggested option, thus increasing forfeiture thoughts overall. But note that no strong conclusion can be drawn from this non-significant result.

Study 1 remains silent on whether participants in the congruent and/or incongruent condition experience more forfeiture thoughts compared to a situation without suggestion. To test whether the coin flip generally increases forfeiture thoughts (H1), a control group without a coin flip is required and was included in Study 2.

Study 1 focused on thoughts and did not assess final choices. However, to check whether choice behavior might be affected, too, we assess choices in Study 2. Furthermore, we investigate whether congruency between initial preferences and suggestion may result in feeling validated (see Hypothesis 3), while incongruency could lead participants to doubt their initial preferences.

Participants and design. An a priori power analysis conducted with G*Power (Faul et al., 2007) assuming a medium effect size of f = .25, α = .05, three groups, and a desired power of .80 resulted in a required sample size of 159 participants. To compensate for potential participant dropout, we added 10% to this number while keeping group sizes balanced, therefore aiming for a sample size of 177 participants. We obtained data from 176 participants, but excluded four participants because of their participation in a previous study, two because they indicated explicit reasons as to why they wanted their data to be excluded, and six because they indicated that they had not carefully completed the study (a value of ≤ 4 on a scale from 1 to 9). The resulting sample comprised 165 participants (109 female, 55 male, 1 no information; M_age = 22.80, SD_age = 5.29). Participants received the chosen snack as compensation.

Participants were randomly assigned to one of three between-subjects conditions: a control without any suggestion (n = 57), coin suggests snack A (n = 54), or coin suggests snack B (n = 54). Based on this setup we recoded the factor condition as control versus coin congruent versus coin incongruent. Participants were asked about their forfeiture thoughts regarding both snacks (within-subjects factor) that was coded as snack (snack A vs. snack B) or coin suggestion (suggested vs. not suggested option). Level of forfeiture thoughts served as the main dependent variable of interest.

Materials and procedure. Study 2 included a control group in which participants did not flip a coin and did not receive a suggestion regarding a snack. Instead, control participants were shown four different hour glasses, each turning for 20 seconds, to equalize the study length in all conditions. The crucial difference between conditions therefore lies in receiving a specific suggestion in the coin but not in the control group.

After having indicated their preference by clicking on the respective snack, participants were introduced to the coin (the control group did not receive any further information at this point). Participants could test the coin several times to ensure that they would believe in its random nature. The two snack options were then displayed again, the coin was virtually flipped, and it recommended one option by appearing above snack A or B.

Participants were then asked about their forfeiture thoughts with respect to the two options using the same items as in Study 1, while control participants answered: “To what extent did you think about forfeiting snack A [placeholder]?” and “To what extent did you think about forfeiting snack B [placeholder]?” Answers were recorded on a Likert scale (ranging from 1 = not at all to 7 = very much).

Coin participants then indicated whether the outcome of the coin had made them feel validated and whether it had made them feel unsettled (on scales ranging from 1 = do not agree at all to 7 = agree very much). Finally, all participants were asked to describe their mood on a Likert scale, to probe whether the effect on forfeiture thoughts is idiosyncratic (instead of generalized to other constructs). The mood scale was anchored on 1 = somewhat negative to 7 = somewhat positive. All participants then chose a snack in the study, which they could then pick up on their way out.

Overall preferences. Ninety-one participants (55%) preferred snack A, while seventy-four participants (45%) preferred snack B. This resulted in a cell distribution of n = 25 (n = 20) preferring snack A (B) and receiving a congruent suggestion, and n = 34 (n = 29) preferring snack A (B) and receiving an incongruent suggestion.

Final choices. Looking at choice proportions, 91 (55%) participants chose snack A and 74 (45%) chose snack B. In the coin condition, 58 (54%) participants chose snack A and 50 (46%) snack B, while 33 (58%) participants chose snack A and 24 (42%) snack B in the control condition. Choice proportions did not differ between the coin and control groups (χ²(1) = 0.27, p = .625). When comparing initial preferences with final choices, only six participants chose against their initial preference (five in the coin and one in the control condition, no significant difference between conditions, χ²(1) = 0.88, p = .432).

Comparisons between the three between conditions. To compare conditions regarding forfeiture thoughts, we conducted a 3 x 2 mixed ANOVA with condition (incongruent vs. congruent vs. control) as a between-subjects factor and snack (A vs. B) as a within-subjects factor. Assumptions regarding homogeneity of covariances were violated (Box’s test p = .024), and assumptions regarding homogeneity of the error variances was violated for snack A (Levene’s test p = .041; snack B: p = .571). We therefore report the results from the mixed ANOVA below, but also computed separate analyses for each snack and report the Welch’s ANOVA result for snack A.

The mixed ANOVA shows a significant main effect of condition, F(2, 162) = 8.86, p < .001, η²_p = .10, no effect of snack, F(1, 162) = 1.58, p = .211, η²_p = .01, and no interaction between condition and snack, F(2, 162) = 0.07, p = .935, η²_p = .00. We then used post-hoc Tukey HSD tests to compare the incongruent versus the control condition as well as the congruent versus the control condition. Both comparisons were significant, see Table 2. When computing separate one-way ANOVAs for snacks A and B, the effect of condition remained significant, F_A(2, 102.35) = 5.75, p = .004, and F_B(2, 162) = 4.32, p = .015, η² = .05, yet not all of the post-hoc comparisons reached significance, see Table 2.

We further tested whether conditions differed regarding mood, but homogeneity of the error variances was not given (Levene’s test p = .037). We therefore report the main effect from the Welch’s ANOVA, F(2, 99.96) = 3.79, p = .026. Post-hoc Tukey HSD comparisons show that participants in the coin-incongruent condition do not differ from the control condition, p = .651, and the same is true for coin-congruent versus control participants, p = .198. Participants in the coin-incongruent compared to the coin-congruent condition, however, indicated lower levels of positive mood, p = .028 (see Table 2).

(In)congruency. Looking at differences within the coin group with a 2 x 2 mixed ANOVA, with (in)congruency as a between-subjects factor and coin suggestion as a within-subjects factor, we descriptively found a tendency for more forfeiture thoughts (across both snacks) when preference and suggestion were incongruent versus congruent, see Table 2. However, the main effect of (in)congruency, F(1, 106) = 1.48, p = .226, η²_p = .01, the main effect of coin suggestion, F(1, 106) = 0.05, p = .825, η²_p = .00, and the interaction between congruency and coin suggestion, F(1, 106) = 0.35, p = .555, η²_p = .00, were not significant. Participants’ mood and feelings of validation were significantly higher in the congruent compared to the incongruent condition (see Table 2), whereas feelings of unsettlement did not differ (see Table 2).

Study 2 investigated how flipping a coin impacts forfeiture thoughts compared to not flipping a coin. Providing support for Hypothesis 1, flipping a coin (regardless of its suggestion) increased forfeiture thoughts compared to a control group (see Table 2). There were no differences between groups regarding mood, demonstrating an idiosyncratic effect of flipping a coin on forfeiture thoughts.

As in Study 1, the coin flip descriptively elicited more forfeiture thoughts when it was incongruent versus congruent with participants’ preferences. Unlike in Study 1, this pattern failed to reach significance. As in Study 1, we further did not find that forfeiture thoughts were stronger for the not suggested versus suggested option. Based on these findings, we dropped Hypothesis 4.

Study 2 further showed that congruency (vs. incongruency) increased feelings of validation, which provides preliminary support for Hypothesis 3. Increased feelings of validation after congruent (vs. incongruent) suggestions have also been found in previous research on catalysts in consumer contexts (Douneva et al., 2019; Study 2), and illustrate one reason for why people might experience coin flips as helpful.

Study 2 supports the hypothesis that using a coin increases forfeiture thoughts compared to a no-coin control group (H1). Studies 1 and 2 provide mixed support for Hypothesis 2 that incongruency (vs. congruency) between preference and suggestion increases forfeiture thoughts. We conducted a preregistered Study 3 that put Hypothesis 2 to another test and extended the previous studies by investigating further consequences of flipping a coin: We hypothesized that a suggestion that is congruent (vs. incongruent) with initial preferences does not only increase feelings of validation (see Study 2), but that these feelings also increase individuals’ confidence in making a good decision (H3). To test the explorative Hypothesis 5, we investigated whether flipping a coin impacts acquisition thoughts by probing for bidirectional differences in acquisition thoughts between control and coin participants. We also exploratorily looked at individuals’ choices and decision satisfaction. The preregistration for Study 3 can be found here: https://osf.io/dybnq.

Participants and design. An a priori power analysis conducted with G*Power (Faul et al., 2007) for mixed ANOVAs focusing on the between-subjects factor assuming a medium effect size of f = .25,⁴ α = .05, two groups (as we compared two conditions each and adding a third condition decreases the sample size per cell), and a desired power of .95 results in a required sample size of 79 participants per cell. Computing the power analysis for a one-tailed t-test based on the same assumptions (d = .50, α = .05, desired power of .95) results in a required sample of 88 participants per cell. To compensate for potential participant dropout, we added 20%⁵ to the required sample size, therefore aiming for a sample size of 106 participants per cell and 318 participants overall. The study was conducted in the lab as the first study after a COVID-19 lockdown, and 320 participants completed the study.⁶ We excluded all participants that indicated that they a) participated in a similar study (n = 7), b) want their data to be excluded (n = 9), and c) did not carefully complete the study (a value of ≤ 4 on a scale from 1 to 9; n = 4).⁷ The resulting sample comprises 300 participants (190 female, 100 male, 10 other/no information, M_age = 23.21, SD = 4.43). Participants received the chosen snack as compensation.

Participants were randomly assigned to one of the following three between-subjects conditions: control without any suggestion (n = 100), coin suggests snack A (n = 100), or coin suggests snack B (n = 100). Taking participants’ preferences into account, we coded the factor condition as control group (no suggestion; n = 87) versus coin-congruent (suggestion in line with initial preference, n = 96) versus coin-incongruent condition (suggestion not in line with initial preference, n = 87). Thirty participants indicated being undecided, which was coded as missing for this factor. Participants were first asked about forfeiture and then about acquisition thoughts regarding snack A and B (within-subjects factor type of thoughts). Snack A and B therefore served as within-subjects factor (labelled as snack when coded as snack A vs. snack B; labelled as coin suggestion when coded as suggested vs. not suggested option). We chose to assess forfeiture and acquisition thoughts within participants to investigate correlations. Forfeiture or acquisition thoughts served as dependent variables in the respective hypotheses tests.

Materials and procedure. Study 3 was similar to Study 2 with the following exceptions: We randomized whether snack A was displayed on the left (and snack B on the right) side of the screen, or vice versa. This display factor was only a control and not a design factor.

We asked for participants’ preference for snack A or B, but participants also had the option to indicate being undecided (scale with options 1 = snack A, 2 = undecided, 3 = snack B). After indicating their initial preference (and flipping the coin in the coin condition), participants were first asked about their forfeiture thoughts and then about their acquisition thoughts regarding the two options. We used the same phrasing of the questions in all conditions: “To what extent did you think about forfeiting / acquiring snack A [placeholder]?” and “To what extent did you think about forfeiting / acquiring snack B [placeholder]?” The order of the questions followed from the presentation order, meaning that we always first asked about the snack presented on the left side of the screen and second about the snack presented on the right side of the screen. Answers were recorded on a Likert scale (ranging from 1 = not at all to 7 = very much).

Afterwards, participants were asked about feelings of validation (see Study 2). Furthermore, we assessed how confident participants were that they will make a good decision (1 = not confident at all to 7 = completely confident). Finally, participants were asked to choose a snack and to rate how satisfied they were with their decision (1 = very dissatisfied to 7 = very satisfied).

All analyses have been preregistered in the here presented order.

When being asked for their initial preference, 118 participants preferred snack A and 152 preferred snack B. Consistent with pre-registration, participants who indicated being undecided were not included in the main analyses (n = 30). Descriptive results for all analyses can be found in Table 3.

Effect of condition on forfeiture thoughts (H1). To test whether flipping a coin compared to not flipping a coin increases forfeiture thoughts regarding snacks A and B, we conducted a 3 x 2 mixed ANOVA with condition as a between-subjects factor (coin-congruent vs. coin-incongruent vs. control) and snack as a within-subjects factor (snack A vs. snack B). Forfeiture thoughts served as the dependent variable. We observed a significant main effect of snack, F(1, 267) = 4.34, p = .038, η²_p = .02, indicating that participants reported more forfeiture thoughts for snack A compared to snack B. Results indicate no significant effect of condition, F(2, 267) = 0.34, p = .713, η²_p = .00, and no significant effect of the interaction between snack and condition, F(2, 267) = 0.81, p = .445, η²_p = .01.

As we observed neither a significant effect of condition nor a significant interaction, we refrained from further analyses and simply investigated whether there were descriptive tendencies showing that participants in the coin-incongruent and the coin-congruent condition reported more forfeiture thoughts than participants in the control condition. This was the case for snack B but not for snack A (see Table 3). Overall, the data is not in support of Hypothesis 1.

As an exploratory addition to the confirmatory data analysis, we also report means and standard deviations of forfeiture thoughts in the different conditions depending on the spatial presentation of the snacks (snack A on the left or right side of the screen) in the Appendix.

Effect of (in)congruency on forfeiture thoughts (H2). Focusing on the coin conditions only, we used a 2 x 2 mixed ANOVA to test the impact of (in)congruency (between-subjects factor: congruent vs. incongruent) and coin suggestion (within-subjects factor: suggested vs. not suggested snack) on forfeiture thoughts (dependent variable). Results indicate neither a significant main effect of (in)congruency, F(1, 181) = 0.11, p = .737, η²_p = .00, nor of coin suggestion, F(1, 181) = 0.27, p = .602, η²_p = .00. However, we find a significant interaction between (in)congruency and coin suggestion, F(1, 181) = 74.68, p < .001, η²_p = .29.

To understand this interaction, we ran separate analyses for the suggested versus the not suggested option. For the suggested snack, forfeiture thoughts were higher for participants that received an incongruent versus congruent suggestion, F(1, 181) = 46.27, p < .001, η²_p = .20. To illustrate, consider the following example with snack B being suggested: When participants preferred snack A and the coin suggested snack B (an incongruent suggestion) versus when participants preferred snack B and the coin suggested snack B (a congruent suggestion), participants reported more forfeiture thoughts for snack B (the suggested snack). Forfeiture thoughts also differed for the not suggested snack, F(1, 181) = 49.69, p < .001, η²_p = .22, but the pattern is reversed: Congruency resulted in more forfeiture thoughts regarding the not suggested snack (see Table 3). To illustrate, consider the following example with snack A being not suggested: When participants preferred snack A and the coin suggested snack B (an incongruent suggestion) versus when participants preferred snack B and the coin suggested snack B (a congruent suggestion), participants reported more forfeiture thoughts for snack A (the not suggested snack). Overall, the results show that coin participants had more forfeiture thoughts regarding the snack they did not prefer and, therefore, do not support Hypothesis 2.

Effect of (in)congruency on validation and confidence (H3). To test whether congruency versus incongruency increases feelings of validation and confidence in the decision, we used one-tailed t-tests with (in)congruency as independent variable and feelings of validation and confidence as respective dependent variables. Results show significantly higher feelings of validation when participants received a congruent versus an incongruent suggestion, t(170.96) = 5.81, p < .001, d = 0.85, see Table 3. This pattern was descriptively replicated when looking at confidence, but differences did not reach significance, t(165.45) = 0.63, p = .265, d = 0.09, see Table 3. Results therefore only provide partial support for Hypothesis 3.

Consistent with pre-registration, we exploratorily investigated effects of receiving a congruent versus an incongruent suggestion versus no suggestion (control group) on confidence with a one-way ANOVA. Results, however, did not show a significant effect of (in)congruency on confidence, F(2, 267) = 0.64, p = .529, η²_p = .01; see also Table 3 for the descriptive results.

Effects of condition on forfeiture versus acquisition thoughts (H5). To investigate whether condition impacts acquisition thoughts, we conducted a 3 x 2 mixed ANOVA with condition as a between-subjects factor (coin-congruent vs. coin-incongruent vs. control) and snack as a within-subjects factor (snack A vs. snack B). Acquisition thoughts serve as the dependent variable. Results show a significant main effect of snack, F(1, 267) = 5.16, p = .024, η²_p = .02, indicating that participants reported more acquisition thoughts regarding snack B compared to snack A (see Table 3). Results indicate neither a main effect of condition, F(2, 267) = 1.97, p = .141, η²_p = .02, nor an interaction effect between condition and snack, F(2, 267) = 0.29, p = .750, η²_p = .00. The descriptive pattern is in line with H5 (see Table 3): Participants in the coin-incongruent and the coin-congruent condition had more acquisition thoughts than participants in the control condition. Overall, however, the differences are small and the data do not support Hypothesis 5.

The within-participant assessment of forfeiture and acquisition thoughts further allows to compute correlations between the two variables across participants. Overall, forfeiture and acquisition thoughts were positively and significantly correlated, r = .27, t(268) = 4.66, p < .001. When looking at between-subjects condition separately, correlations were highest in the congruent condition, r = .35, t(94) = 3.62, p < .001, second highest in the control condition, r = .27, t(85) = 2.61, p = .011, and lowest as well as not significant in the incongruent condition, r = .19, t(85) = 1.83, p = .071.

Effects of indecisiveness on forfeiture and acquisition thoughts. We preregistered to separately run analyses for undecided participants in case their number exceeds n = 25⁸ per condition. This was not the case: in total, 30 participants reported being undecided (17 in the coin, 13 in the control group). We therefore only report the descriptive results in Table 3 and refrain from inferential tests.

Adherence to the coin’s suggestion, choice proportions, and satisfaction with choice. To investigate participants’ adherence to the coin flip’s suggestion in an exploratory fashion, we used a χ²-test to examine associations between the suggested snack (snack A vs. snack B) and the chosen snack (snack A vs. snack B). The data, however, do not show a significant association, χ²(1) = 0.72, p = .395, meaning that participants presumably made their choices independently of the coin flip suggestion.

To investigate whether choice proportions differ between coin and control groups, we used a χ²-test to examine associations between condition (coin-congruent vs. coin-incongruent vs. control) and choice (snack A vs. snack B), which was also not significant, χ²(2) = 2.57, p = .277.

We further tested whether satisfaction is positively or negatively correlated with forfeiture thoughts. For participants who indicated a preference at the beginning of the study (n = 270), there is no significant correlation between forfeiture thoughts and satisfaction with the decision, r = .00, t(268) = 0.05, p = .959.⁹

Study 3 was preregistered and designed to replicate previous findings as well as to test additional hypotheses. It builds on results from Studies 1 and 2. Study 2 shows that flipping a coin and receiving both a congruent or an incongruent suggestion resulted in more forfeiture thoughts compared to a control group. Study 3 does not replicate this pattern: In contradiction to Hypothesis 1 and the results from Studies 1 and 2, receiving an incongruent versus congruent suggestion results in similar levels of forfeiture thoughts. Participants reported more forfeiture thoughts for snack A compared to snack B, but this could simply result from pre-study differences in preferences for the snacks. Furthermore, coin participants reported more forfeiture thoughts for the option that they did not prefer (which is, if they stick with their preference, also the one they have to forfeit in the end). Eventually, this could be taken as evidence that coin participants paid close attention to the study instructions and carefully thought about the option they would not get before making their decision. In any case, Hypothesis 2 is not supported by the data.

Study 3 provided partial support for Hypothesis 3: Receiving a congruent (vs. incongruent) suggestion resulted in stronger feelings of validation (replicating Study 2). There was also a tendency for congruent suggestions to increase confidence, but this difference was not significant. In an exploratory fashion, we also investigated acquisition thoughts, but did not find any differences between groups and could therefore not show an impact of the coin toss on acquisition thoughts. Overall, Study 3 findings differ substantially from those observed in Studies 1 and 2; we discuss potential reasons in the General Discussion below.

We also looked at adherence rates between the coin suggestion and participants’ choices. High adherence would indicate that participants may have used the coin flip as a decider, meaning that its outcome would determine their decision. In this study, however, participants were asked to use the coin flip as a catalyst, meaning to consider the suggestion, but to decide on their own. Results suggest that the coin was not used as a decider, but may have acted as a catalyst in participants’ own decision making.

This manuscript investigated whether flipping a coin may afford a change in perspective in a consumer context. More specifically, we hypothesized that flipping a coin and receiving a concrete suggestion would shift individuals’ perspective from an acquisition to a forfeiture scenario and would thereby increase forfeiture thoughts. In three lab studies, participants were asked to make a consumer decision by choosing which one of two snacks they a) prefer and b) want as compensation. Some participants were asked to flip a coin in between these two decisions, which recommended one option over the other. These participants therefore received a suggestion that was congruent or incongruent with their initially stated preference. All participants were then asked to what extent they had thought about forfeiting each snack and proceeded to choose a snack as compensation. This setup allowed us to study whether flipping a coin and receiving an incongruent versus congruent versus no suggestion (in the control group) impacts forfeiture thoughts.

Results across studies are mixed. In Study 1, all participants flipped a coin and results showed that receiving an incongruent (vs. congruent) suggestion resulted in more forfeiture thoughts. In Study 2, this difference emerged descriptively, but did not reach significance. Study 2 further included a control condition without a coin flip. Both the congruent and the incongruent coin conditions resulted in more forfeiture thoughts compared to the control condition, suggesting that flipping a coin shifts participants’ perspective towards a forfeiture scenario. Presumably this is because the coin flip highlights that choosing something means forgoing something else. Results of Study 2 further show that receiving a suggestion that was congruent (vs. incongruent) with participants’ preferences resulted in higher feelings of validation.

Based on Studies 1 and 2, we designed Study 3, but the results strongly differ. Coin and control conditions did not significantly differ regarding forfeiture thoughts (H1). Furthermore, we did not find support that an incongruent (vs. congruent) suggestion results in more forfeiture thoughts (H2). When following up on a suggestion mentioned in the review process to look at acquisition thoughts in an exploratory fashion, we did not observe a significant difference either (H5). Yet, Study 3 provides partial support for Hypothesis 3, showing that receiving a congruent (vs. incongruent) suggestion results in stronger feelings of validation and, in tendency, more confidence.

All three studies were similar in their design and used identical snacks as stimuli and compensation. Nevertheless, differences occurred, which may potentially account for the observed inconsistencies in findings. We address some of these differences and speculate how future research may better test the impact of random decision-making aids on forfeiture thoughts.

One difference between Studies 1 and 2 versus Study 3 is timing. Studies 1 and 2 were conducted before the COVID-19 pandemic, whereas Study 3 was the first study run in our lab once the university reopened in September 2021. One may speculate that the requirement to wear masks and to disinfect hands frequently may have raised participants’ health concerns and made them feel differently about consuming food items in public. Although this speculation has been derived post-hoc, it may be possible that such changes in cultural practices as a function of the pandemic account for the findings’ heterogeneity.

Study 3 further differed from Studies 1 and 2 in the way we assessed initial preferences: In Study 3, participants could indicate that they were undecided (which about 30 of them did) or whether they had a preference for either of the two snacks. In contrast, participants in Studies 1 and 2 indicated right at the study’s beginning which snack they preferred, without being able to voice indifference. What appears like small differences may have strong psychological impact, as indicating a clear preference for a snack right away may have set more readily the ground for the processes hypothesized here. Indeed, a change in perspective may be more likely induced by a coin flip when a preference has previously been stated, because a clearly stated preference makes the suggestion feel right or wrong and increases awareness that choosing one option means forfeiting the other. In contrast, if a preference has not been stated as clearly or feels more flexible (as might have been the case in Study 3 because of the “undecided”-option), participants may actively search for a strategy to make up their mind. Here, using a coin versus not using a coin might result in similar behaviours: All participants may try to look at the decision from different perspectives and thereby may be less susceptible to the potential that a coin flip has to offer. This could explain why we do not find differences between the coin and control condition in Study 3. Future studies could vary the assessment of preferences by using different scales or implement a different approach to assess initial preferences in the first place, for instance asking for preference ratings for many different snacks and then choosing two that have been rated similarly but slightly differently (see, e.g., Jaffé & Greifeneder, 2021; Krajbich et al., 2010), or by assessing preferences in a separate session. These approaches could also result in more decision difficulty, eventually mimicking a situation where people would be more likely to use a decision aid in the first place.

It appears also instructive to compare the present studies’ setup to those of previous research to better understand the result pattern and develop ideas for future work. Previous research manipulated the decision scenario to create a forfeiture or acquisition setup. In particular, Dhar and Wertenbroch (2000) gave participants two options that participants thought they could keep, but were then asked to forfeit one of the two, resulting in a forfeiture scenario. In the acquisition scenario, participants were told that they could choose between two options. Preference or choice of the hedonic versus utilitarian option served as dependent variables. Across their studies, the authors showed that a forfeiture scenario (vs. an acquisition scenario) resulted in a preference of hedonic (vs. utilitarian) choice options. In our project, we did not manipulate but measured forfeiture thoughts, thereby investigating forfeiture thoughts in a new setup using self-report. Future research may explore alternative setups to either create a forfeiture setting or to assess forfeiture thoughts in a different way. As alternatives to self-reports, for example, researchers could rely on behavioural measures, reaction times, or eye tracking. For eye tracking, the relation between gaze times and forfeiture thoughts would first need to be explored. Participants could attend to the option they want to acquire (resulting in shorter gaze times for the option they want to forfeit) or focus on the option that they want to forfeit to simulate how this decision would feel (resulting in longer gaze times for the option they want to forfeit). If associations between gaze time and forfeiture thoughts prove stable, this measure could provide further insight into the relation between flipping a coin and forfeiture thoughts.

So far, we have speculated about methodological reasons for the heterogeneity in the observed findings. However, it should also be mentioned that the differences could be due to mere chance, such that flipping a coin (compared to a control group) has no significant impact on forfeiture thoughts. A shift in perspective regarding forfeiture thoughts may therefore not be the (primary) reason why people find random decision-making aids, such as coin flips, helpful. Eventually, the aid proves useful through different routes not investigated here.

We investigated whether flipping a coin impacts forfeiture thoughts when making consumer decisions. We conducted three lab studies in which participants were asked to choose between two snacks and to flip a coin. Receiving a suggestion that was congruent versus incongruent with participants’ initial preferences resulted in stronger feelings of validation. However, the pattern regarding forfeiture thoughts was not conclusive, so that the hypotheses that a coin flip may impact forfeiture thoughts were not supported. We discuss potential reasons for the heterogeneity in the observed findings and suggest pathways for future research to improve the investigation of how random decision-making aids may impact individuals’ judgment and decision making.

Contributed to conception and design: MJ, MD, RG Contributed to acquisition of data: MJ, RG Contributed to analysis and interpretation of data: MJ, RG Drafted the article: MJ Revised the article: MJ, MD, RG Approved the submitted version for publication: MJ, MD, RG

We thank all research assistants involved in data collection. We thank Helena Brunt and Sonja Borner for their help in designing the studies. We thank Matthias Stutz for organizing the data collection and Caroline Tremble for proofreading the manuscript.

This research was supported by grant #100019_159445 from the Swiss National Science Foundation.

None of the authors report a conflict of interest.

The data, code, and materials can be accessed via the following repository: https://osf.io/hr3b9/.