Greed is the insatiable desire for more. In three preregistered high-powered experiments, we examined the role of dispositional greed in inaction inertia, the phenomenon that people less likely act on a discount after missing a more attractive one. In Study 1 with a within-subjects design, we found strong evidence that higher greed weakens inaction inertia such that greedy people always want more and are less influenced by missed discounts. Studies 2 and 3 failed to replicate this moderation, both in between- and within-subjects designs. An integrative data analysis suggests that the relationship between greed and inaction inertia is more complex, with non-linearity and facet-specific effects. Overall, our results indicate that people dispositionally differ in how missed discounts influence future purchase decisions as a function of greed, but that these differences are more intricate than what current theorizing on greed predicts.
People like discounts and marketers know that. But when people miss discounts (i.e., when they do not take advantage of them), this might have detrimental consequences. Missing a discount could lead to inaction inertia: the phenomenon that “bypassing an initial action opportunity decreases the likelihood that subsequent similar action opportunities will be taken” (Tykocinski et al., 1995; p. 793).
For instance, imagine that last week you wanted to buy a ski pass at a reduced price, $40 instead of the regular $100. You somehow forgot about it and missed the opportunity. The ski pass is now offered for $90 instead of $100. Will you buy it now? In Tykocinski et al.’s experiments, people were less likely to buy the $90 ski pass after missing the $40 one (i.e., when there was a large value difference between the two offers) than after missing an $80 discount (i.e., small value difference). Thus, people who did not make use of the earlier, greater, discount (inaction) were more likely to refrain from purchasing the product later (inertia).
Ample research has examined inaction inertia across different contexts (for reviews, see Chen et al., 2021; Van Putten et al., 2013), among others, decisions about going on a vacation (Zeelenberg et al., 2006), and pension plans (Krijnen et al., 2019). Outside the everyday consumer context, inaction inertia has been found to influence negotiations (Terris et al., 2020; Terris & Tykocinski, 2016) and the willingness to accept job offers (Foster & Diab, 2017).
In this paper, we address whether all people are equally likely to fall prey to the inaction inertia effect. Only a few studies have investigated personality differences in inaction inertia. These studies found that people who tend to focus on the present (those with an assessment or state orientation) show more inaction inertia (Mathmann et al., 2017; Van Putten et al., 2009). This is consistent with experimental findings that inaction inertia is attenuated when people mentally decouple the past from the present (Van Putten et al., 2013). Yet, we believe that other personality characteristics could also influence the susceptibility to inaction inertia.
Dispositional Greed
In the present research, we focus on a crucial personality trait in social behavior: greed. Greed is excessive, as greedy people want more than they need – as much as possible and as good as possible of any desirable thing. Greed is not restricted to money. It includes non-monetary goods and intangible experiences, such as power, status, and the number of sexual partners (Hoyer et al., 2024).
In addition to wanting more, greed is also associated with clinging to one’s possessions. Seuntjens et al. (2015) found that “stinginess,” and “not being generous” were central elements of the concept of greed. Hence, they conclude that greed is not only about acquiring more than a person currently has but also about keeping what the person already has – so that acquiring would lead to increased possession in total. Thus, an aversion to losing seems to be a part of greed (cf. Krekels & Pandelaere, 2015).
Greed and Inaction Inertia
We decided to examine the role of dispositional greed in inaction inertia for two main reasons. First, it may help us to understand who is affected most by missing discounts and, thus, most likely to suffer from inaction inertia. Second, it may also teach us more about what greed is and what greed does.
We assumed that greedy people generally produce a weakened inaction inertia effect, defined as the difference in purchase likelihood between a small (e.g., 80$ vs. 90$) - and a large (e.g., 40$ vs. 90$) value difference. However, our reading of the literature led us to propose two competing hypotheses about why greedy people should produce a weaker inaction inertia. These competing predictions should manifest in the direction of the main effect of greed on purchase likelihood (see Figure 1). Depending on which core element of greed (Seuntjens, Zeelenberg, Breugelmans, et al., 2015) one focuses on, increased greed may either lead to an increased purchase likelihood (H3a) or to reduced purchase likelihood (H3b).
The first hypothesis (H3a), the “greedy people always want more” hypothesis, builds on the finding that greed is mostly excessive acquisitiveness. It assumes that greedy people are always attracted to discounts, as it provides an opportunity to get something for a good price. Previously missed discounts are irrelevant, as greedy people focus on acquiring material and immaterial things (Hoyer et al., 2024; Seuntjens, Zeelenberg, Breugelmans, et al., 2015; Seuntjens, Zeelenberg, van de Ven, et al., 2015). Thus, in the context of inaction inertia, for greedy people, we could expect an overall higher likelihood to act on current discounts and a weaker inaction inertia effect (i.e., a reduced difference between the small and the large value difference condition due to an overall higher purchase likelihood).
The competing hypothesis (H3b), “greedy people hate losing”, assumes that greedy people are stingy, value money more, and show more loss aversion (Krekels & Pandelaere, 2015; Seuntjens, Zeelenberg, Breugelmans, et al., 2015). This leads us to expect that missing a discount hurts more for greedy people. After all, it is not possible to acquire the best possible offer anymore. The result may be an overall lower likelihood to act on the current discount and a weaker inaction inertia effect (i.e., a reduced difference between the small and large value difference condition, due to an overall lower purchase likelihood). This hypothesis is consistent with Mussel and Hewig (2016), who found that greedy individuals show more negative affect in response to losses, and with Tykocinski and Pittman (1998), who argued that inaction inertia occurs because people aim to distance themselves from missed discounts that are associated with negative affect.
The Current Research
We designed three studies that stayed close to how inaction inertia is typically studied (cf. Tykocinski et al., 1995, Exp. 3-6 and most subsequent research). That is, comparing responses between two conditions (large vs. small value differences). Participants read inaction inertia scenarios and were asked about the likelihood of acting on a discount after having missed a more attractive discount. The current offer either represented a large or a small difference in comparison to the missed offer (e.g., regular price $100, current offer $90, missed offer $40 [large value difference] vs. $80 [small value difference]). Additionally, we used multiple scenarios to avoid idiosyncratic effects exclusive to one particular decision problem (“stimulus sampling”, cf. Wagenaar et al., 1988; Wells & Windschitl, 1999) and thus increase the evidential value of our research.
We also examined valuation as a mediator (cf. Arkes et al., 2002; Van Putten et al., 2013; Hypothesis 2), to test how greed affects the process behind inaction inertia (Hypothesis 4). When the missed opportunity is perceived as an informative reference point to understand the actual value of a product, then any inferior discount will lead to feelings of being overcharged. Because we assume that the effect of greed on inaction inertia comes from the desire the offer elicits, we hypothesize that greed has the same effect on valuation as on purchase likelihood. We thus predict that the hypothesized effect of greed is mediated by the offer’s valuation, such that greed moderates the impact of value difference on valuation (Moderated Mediation Hypothesis). Furthermore, we assumed that the effect of greed on purchase likelihood (independent of the value difference) is also due to more/less valuation of an offer. Overall, this led to the following hypotheses:
H1: People are less likely to act on an offer if the value difference between the missed offer and the current offer is large (vs. small) [Inaction Inertia].
H2: Inaction inertia is mediated by the current offer’s valuation [Mediation Hypothesis I].
H3: High dispositional greed (vs. low dispositional greed) will reduce the inaction inertia effect … [Moderation Hypothesis]
-H3a: … in such a way that greedy people (vs. non-greedy) will overall display increased purchase likelihood [“greedy people always want more” hypothesis]
-H3b: … in such a way that greedy people (vs. non-greedy) will overall display reduced purchase likelihood [“greedy people hate losing” hypothesis]
H4: The effect in H3 is mediated by the current offer’s valuation such that greed moderates the impact of value difference on valuation [Moderated Mediation Hypothesis]
H5: Dispositional greed influences purchase likelihood through overall levels of valuation [Mediation Hypothesis II]
Study 1
Method
Transparency and Openness
This study was pre-registered via https://aspredicted.org/4396-2p4x.pdf. Data, code, and materials can be found on: https://researchbox.org/374. The scenario descriptions and questionnaire items are in the Supplementary Materials. We report all manipulations, measures, and exclusions in the following studies. Data were analyzed using R (R Core Team, 2025).
Design and Procedure
This experiment had a single within-subjects factor value difference (small vs. large) and the between-subjects continuous variable greed. Half of the participants first responded to six inaction inertia scenarios, followed by the dispositional greed assessment. For the other half, the order was reversed. For each participant, three scenarios were randomly chosen to be presented in the small-difference format and three in the large-difference format. Finally, participants provided demographic information and were debriefed.
Power Analysis & Sample
Sample size was determined before any data collection. A power analysis in G*Power 3.1.9.4 (Faul et al., 2009), for the mixed ANOVA design with two groups and two repeated measures (i.e., two levels of the within-subjects factor), indicated that 510 participants were necessary to detect a small effect (f = 0.08; d = 0.16) with 95% power.1 We considered this a conservative approximation for our analysis, in which personality was a continuous (versus categorical) between-subjects predictor and each value difference level was measured three times (versus once). We decided to oversample to 550 in case of dropouts due to the pre-registered exclusion criteria (i.e., being younger than 18, less than intermediate English proficiency, technical problems, or response times below one minute). None had to be excluded.
Participants were recruited on Prolific Academic in the summer of 2021 for a 5-minute (median duration = 04:58) study for £0.63 (~$0.89) compensation. As all scenarios referred to US$ as currency, we recruited 551 English native speakers from the US (360 female, 186 male, 5 diverse; Mage 29.30, SD = 9.55, range = 18-74).
Materials
Dispositional Greed Scale (DGS). We measured individual differences in greed with the 7-item Dispositional Greed Scale (Seuntjens, Zeelenberg, van de Ven, et al., 2015). The DGS is validated and available in eight different languages (Zeelenberg & Weller, 2025).This scale consists of statements like “As soon as I have acquired something, I start to think about the next thing I want.” Participants indicated agreement with statements (1 = Strongly disagree; 5 = Strongly agree; M = 2.86, SD = 0.79). The internal consistency was acceptable (α = .78).
Inaction Inertia Scenarios. We used a variety of scenarios from previous research. Our selection was based on the replicability of the effect, guided by Chen et al. (2021). The scenarios included: from Tykocinski et al. (1995) the ski pass (p. 795; see also the introduction section of this paper), the car (p. 795), and the fitness club scenario (p. 796), from Zeelenberg et al. (2006) the couch scenario (p. 93) and an adapted version of the city trip scenario (p. 94), and from Krijnen et al. (2019, p. p. 55) the retirement saving scenario.
Each scenario was presented on a single page, with the dependent measures below. Purchase likelihood was assessed via “How likely is it that you will buy [a short description of the current offer]?” (cf. Tykocinski et al., 1995), valuation via “Forgetting for a moment the initial offer that was available, how valuable would you rate the current offer now?” (Both: 0 = not at all, 10 = extremely) (cf. Arkes et al., 2002, p. p. 380).
Results
Analytical Procedure
Due to the nested structure of our data (six responses nested in 551 participants), we conducted a multilevel moderated mediation analysis in the R-package lavaan (Rosseel, 2012). Random intercepts were modeled at the participant level. Greed, purchase likelihood, and valuation were z-standardized at the grand mean. Deviating from the preregistration, value difference was coded with 0.5 (small) and -0.5 (large) instead of 1/-1. Note that the different coding does not change the results of the significance tests. Due to this coding strategy, a positive effect of value difference thus corresponded to the standardized mean difference (Cohen’s d) of the inaction inertia effect. With this coding, the Greed x Value Difference interaction also corresponded to the difference in the inaction inertia effect for 1SD increase in greed. The correlations between the variables can be found in the Supplementary Materials, Table S5a.
Pre-registered Analyses
Purchase Likelihood. The main results are displayed in the left panel of Figure 2. Participants reported higher purchase likelihood if the value difference was small (vs. large), b = 0.70, CI95% [0.64, 0.76], z = 23.23, p < .001, thus replicating the inaction inertia effect2, consistent with H1. In line with H3, this effect was significantly smaller at higher levels of greed, b = -0.15, CI95% [-0.21, -0.08], z = -4.60, p < .001. As expected from H3a and in contrast to H3b, greedy individuals reported overall higher levels of purchase likelihood, b = 0.10, CI95% [0.05, 0.16], z = 3.93, p < .0013.
Valuation. To further test the mediating role of valuation in inaction inertia, we first repeated the same analyses on the mediator valuation as the dependent variable (right panel Figure 2). Participants reported higher valuation if the value difference was small (vs. large), b = 0.61, CI95% [0.55, 0.67], z = 19.99, p < .001. Again, this effect was significantly smaller at higher levels of greed, b = -0.12, CI95% [-0.18, -0.06], z = -3.85, p < .001. Also, greedy individuals reported overall higher levels of valuation, b = 0.10, CI95% [0.05, 0.16], z = 3.62, p < .001.
Note. Greed was z-standardized. Shaded areas represent 95% confidence intervals.
Note. Greed was z-standardized. Shaded areas represent 95% confidence intervals.
Moderated Mediation Analysis4. As the results for valuation were very similar to those for purchase likelihood (Figure 2), we next conducted the preregistered moderated mediation analysis based on the procedure of Hayes (2015). The analysis is similar to model 8 in PROCESS (Hayes, 2022), but adapted to different software and a multilevel data structure. Note that our research design and a mediation analysis cannot test a causal effect of valuation because the relationship between valuation and purchase likelihood is entirely correlative (Fiedler et al., 2011) although the effect of the value difference on purchase likelihood and valuation are induced experimentally.
In line with H2, the inaction inertia effect was mediated by valuation, with a significant indirect effect, b = 0.39, CI95% [0.35, 0.44], z = 16.63, p < .001. This mediation was only partial, as the direct effect was still significant, b = 0.31, CI95% [0.25, 0.36], z = 11.54, p < .001.
In line with H4, there was a significant index of moderated mediation (Hayes, 2015), b = -0.08, CI95% [-0.12, -0.04], z = -3.85, p < .001: A small value difference increases valuation, but the increase is less pronounced for individuals with high greed, which in turn statistically explains the effects found on purchase likelihood. Notably, the direct effect was also moderated by greed, b = -0.07, CI95% [-0.11, -0.02], z = -3.07, p = .002, with a stronger direct effect for people who were low (vs. high) in greed (see Figure 3). Last, the main effect of greed on purchase likelihood as postulated in H3a was fully mediated by valuation with a significant indirect effect, b = 0.07, CI95% [0.03, 0.11], z = 3.60, p < .001, and an insignificant direct effect, b = 0.03, CI95% [-0.00, 0.07], z = 1.84, p = .067, thus supporting mediation hypothesis H5. To summarize, we find strong support for all hypotheses except H3b (and the data strongly support the competing H3a). The results of all analyses are summarized in Figure 3.
Note. Value difference was coded as -0.5 (large) and 0.5 (small). Greed, purchase likelihood, and valuation were standardized at the grand mean. Regression weights in brackets reflect the direct effects after controlling for valuation. The indirect effect was stronger for individuals low (-1 SD) in greed, b = 0.47, CI95% [0.41, 0.54], z = 14.63, p < .001, than for individuals high (+1 SD) in greed, b = 0.32, CI95% [0.26, 0.38], z = 10.44, p < .001. The direct effect was also stronger for individuals low (-1 SD) in greed, b = 0.37, CI95% [0.30, 0.44], z = 10.64, p < .001, than for individuals high (+1 SD) in greed, b = 0.24, CI95% [0.17, 0.31], z = 7.14, p < .001. We also tested whether greed moderated the relationship between valuation and purchase likelihood (b-path). This was not the case, b = 0.00, p = .765. * = p < .05, ** = p < .01, *** = p < .001.
Note. Value difference was coded as -0.5 (large) and 0.5 (small). Greed, purchase likelihood, and valuation were standardized at the grand mean. Regression weights in brackets reflect the direct effects after controlling for valuation. The indirect effect was stronger for individuals low (-1 SD) in greed, b = 0.47, CI95% [0.41, 0.54], z = 14.63, p < .001, than for individuals high (+1 SD) in greed, b = 0.32, CI95% [0.26, 0.38], z = 10.44, p < .001. The direct effect was also stronger for individuals low (-1 SD) in greed, b = 0.37, CI95% [0.30, 0.44], z = 10.64, p < .001, than for individuals high (+1 SD) in greed, b = 0.24, CI95% [0.17, 0.31], z = 7.14, p < .001. We also tested whether greed moderated the relationship between valuation and purchase likelihood (b-path). This was not the case, b = 0.00, p = .765. * = p < .05, ** = p < .01, *** = p < .001.
Discussion
Study 1 showed a reliable inaction inertia effect. It also showed that dispositional greed moderated inaction inertia. Greedy people were less susceptible to the effect - they were overall more likely to act on a current offer, supporting the “greedy people want more” hypothesis. In addition, although we cannot infer causality, the valuation of the current offer mediated the effect of the missed discount on the likelihood of acting on the current discount, replicating Arkes et al. (2002) and Zeelenberg et al. (2006). This mediation was also moderated by dispositional greed. The greedier a person was, the higher they valued the current offer, and the less they were influenced in their valuation by missing a previous offer.
However, Study 1 was limited by the multiple-scenario within-design, which might trigger contrast effects between different scenarios. Also, we collected data exclusively on US participants. To tackle both points, we ran a second study with a single-scenario between-participants design in a new population.
Study 2
In Study 2, we ran a single-scenario between-participants design in a new population to increase generalizability (https://aspredicted.org/dh4v-dbgq.pdf).
Method
This experiment had two between-subjects conditions (value difference: small vs. large). The procedure was the same as in Study 1, except that we collected data on a single scenario (the fitness club scenario; Tykocinski et al., 1995, p. p. 796), without assessing valuation. This scenario does not contain monetary values and can thus be easily used in countries with different currencies.
For an interaction in a 2 x 2 between ANOVA design, a sample size of 1053 was necessary to detect a small effect (f = 0.10; d = 0.20) with 90% power (α = 0.05).5 We decided to oversample to 1100 participants. We once more recruited on Prolific Academic in the fall of 2023. We collected data from native English speakers from the UK (584 female, 514 male, 2 not disclosed; Mage =43.41, SD = 13.83) to increase generalizability given the US participants of Study 1. The study’s median duration was 01:19, with £0.15 (~$0.19) compensation. The Prolific sample was restricted to participants indicating English as their first language, having an approval rate of 95-100, and not having participated in the previous study. No further exclusions were made.
Results
We calculated a linear regression analysis in which z-standardized purchase likelihood was predicted by the interaction between z-standardized greed and value difference (coded once more as 0.5 = small and -0.5 = large). Once more, participants reported higher purchase likelihood if the value difference was small (vs. large), b = 0.80, CI95% [0.69, 0.90], t = 14.46, p < .001, replicating the inaction inertia effect (Figure 4). Higher levels of greed were associated with increased purchase likelihood, b = 0.12, CI95% [0.05, 0.18], t = 3.61, p < .001. However, contrary to Study 1 and H3, the interaction was not significant, b = -0.04, CI95% [-0.17, 0.09], t = -0.63, p = .528.
Note. Greed was z-standardized. Shaded areas represent 95% confidence intervals.
Note. Greed was z-standardized. Shaded areas represent 95% confidence intervals.
Discussion
Study 2 replicated the inaction inertia effect in a between-subjects design. But this time, dispositional greed did not moderate the effect. Greedy people were more likely to act on a current offer, independently from the size of the missed previous offer. As this between-design study showed effects contrary to Study 1, we ran a third study to systematically test whether the replication failure was due to the specific scenario, and whether there were systematic differences in how greed moderates inaction inertia in within-subjects and between-subjects designs.
Study 3
To make sense of the results of the previous studies, Study 3 replicated the within-subjects procedure of Study 1 and the between-subjects procedure of Study 2 but with a different scenario (https://aspredicted.org/nrsx-23v7.pdf).
Method
Study 3 contained both a within-subjects and a between-subjects design, referred to as Study 3a and 3b. Study 3a used the same within-participants design as Study 1 but with four instead of six scenarios (city trip, couch, retirement saving, fitness club; with two randomly assigned scenarios for each within-subjects condition). Study 3b followed a single-scenario between-participants design like Study 2 with only the car scenario.
Crucially, the participants of Study 3a responded to five consecutive inaction inertia scenarios – first the car scenario for the between-subjects manipulation, then the four within-subjects scenarios. Once we had collected the required sample size for the within-subjects design of Study 3a, the four within-subjects scenarios were removed from the study so that further participants only responded to the single between-participants scenario of Study 3b. Therefore, the participants from Study 3a also provided data for Study 3b, but not vice versa. This design allowed a (nearly) direct replication of Study 1 and a conceptual replication of Study 2 with the required sample sizes while keeping the study costs low.
We collected data on Prolific Academic in the fall/winter of 2023. The recommended sample sizes were the same as in Study 1 (N = 550) and Study 2 (N = 1100). We recruited native English speakers from the US. Study 3a counted 545 participants (319 female, 223 male, 3 not disclosed; Mage = 45.48, SD = 14.26). The study’s median duration was 03:29, with £0.45 (~$0.55) compensation. Study 3b counted 1101 participants (645 female, 445 male, 11 not disclosed; Mage = 43.65, SD = 14.35). For half of the participants, those who only provided data for Study 3b (vs. 3a and 3b combined), the study’s median duration was 01:32, with £0.15 (~$0.19) compensation. The Prolific sample was restricted to participants indicating English as their first language, and not having participated in any of the previous studies. No further exclusions were made. Demographic data was collected separately from Prolific.
Results
The correlations between all variables can be found in the Supplementary Materials, Table S5a and S5b. See Figure 5 for a visualization of the effect of greed and value difference on purchase likelihood and valuation.
Study 3a
Purchase Likelihood. We replicated the inaction inertia effect6, as participants who were in the small (vs. large) value difference condition reported higher purchase likelihood, b = 0.71, CI95% [0.64, 0.79], z = 19.28, p < .001. In contrast to Study 1, this effect did not depend on the level of greed, b = -0.02, CI95% [-0.10, 0.05], z = -0.59, p = .557, but greedy people did report higher purchase likelihood, b = 0.07, CI95% [0.02, 0.12], z = 2.75, p = .006 (see Figure 5).
Valuation. Participants reported higher valuation if the value difference was small (vs. large), b = 0.63, CI95% [0.56, 0.70], z = 17.02, p < .001, but this effect did not depend on the level of greed, b = -0.02, CI95% [-0.10, -0.06], z = -0.57, p = .572. Greedy people generally reported higher valuation, b = 0.10, CI95% [0.05, 0.16], z = 3.64, p < .001.
Moderated Mediation Analysis. As in Study 1, we conducted a multilevel moderated mediation analysis but did not find a significant index of moderated mediation (Hayes, 2015), b = -0.02, CI95% [-0.08, 0.04], z = -0.57, p = .572 (see Figure 6).
Note. Greed was z-standardized. Shaded areas represent 95% confidence intervals.
Note. Greed was z-standardized. Shaded areas represent 95% confidence intervals.
Note. Value difference was coded as -0.5 (large) and 0.5 (small). Greed, purchase likelihood, and valuation were standardized at the grand mean. Regression weights in brackets reflect the direct effects after controlling for valuation. Inaction inertia was partially mediated by valuation, bindirect = 0.48, CI95% [0.43, 0.54], z = 15.95, p < .001, bdirect = 0.23, CI95% [0.18, 0.28], z = 8.67, p < .001, but not moderated by greed, b = -0.01, CI95% [-0.05, 0.04], z = -0.26, p = .797. Similar to Study 1, the main effect of greed on purchase likelihood was fully mediated by valuation, bindirect = 0.08, CI95% [0.04, 0.12], z = 3.53, p < .001, bdirect = -0.01, CI95% [-0.04, 0.02], z = -0.56, p = .579.
* = p < .05, ** = p < .01, *** = p < .001.
Note. Value difference was coded as -0.5 (large) and 0.5 (small). Greed, purchase likelihood, and valuation were standardized at the grand mean. Regression weights in brackets reflect the direct effects after controlling for valuation. Inaction inertia was partially mediated by valuation, bindirect = 0.48, CI95% [0.43, 0.54], z = 15.95, p < .001, bdirect = 0.23, CI95% [0.18, 0.28], z = 8.67, p < .001, but not moderated by greed, b = -0.01, CI95% [-0.05, 0.04], z = -0.26, p = .797. Similar to Study 1, the main effect of greed on purchase likelihood was fully mediated by valuation, bindirect = 0.08, CI95% [0.04, 0.12], z = 3.53, p < .001, bdirect = -0.01, CI95% [-0.04, 0.02], z = -0.56, p = .579.
* = p < .05, ** = p < .01, *** = p < .001.
Study 3b
The results were similar to Study 2. Participants reported higher purchase likelihood if the value difference was small (vs. large), b = 1.38, CI95% [1.08, 1.69], t = 8.89, p < .001. Higher levels of greed predicted increased purchase likelihood, b = 0.29, CI95% [0.13, 0.44], t = 3.67, p < .001. Value difference and greed did not interact, b = -0.07, CI95% [-0.38, 0.24], t = -0.45, p = .652.
Discussion
Both studies, the within-design Study 3a, and the between-design Study 3b, reliably produced the inaction inertia effect. However, the effect did not depend on the level of dispositional greed. Whereas Study 3b’s results are similar to Study 2 - suggesting that greed does not moderate inaction inertia in between-subjects designs - Study 3a’s results go against Study 1. This time, the inaction inertia effect did once more not depend on greed, although Study 3a was a highly powered direct replication of Study 1 with only minor differences. The central interaction predicted in Hypothesis 3 vanished entirely. Given that the effect had been robust7 in a preregistered study before (i.e., with z = 4.60 and p = .000002), we deemed it unlikely that this had been merely a false positive. Therefore, we conducted an additional integrative data analysis across all studies exploring potential reasons for why the effects did not replicate in Study 3a.
Integrative Data Analysis
In the following, we describe several analyses conducted to explore potential reasons why the effects did not replicate. For parsimony, we only describe the main findings and present more detailed empirical results in the Supplement (see Table S6a to S12b; note “Sample Characteristics” and Figure S1).
Overall Results Across Studies
We first conducted one integrative analysis on all within-subjects data (Study 1 & 3a) and one on all between-subjects data (Study 2 & 3b) where we ran our main preregistered model with study as additional predictor. For the within-subjects data, there was still an overall interaction across both studies, but also a three-way interaction with study, such that the Greed x Value Difference interaction was significantly weaker in Study 3a. For the between-subjects data, there was still no significant interaction. As Study 2 was conducted with UK participants and Study 3b with US participants, and no effect of study on the greed-inaction inertia interaction was found, we ruled out cultural differences as an explanation for now.
Nonlinear Effects
Despite the same sampling criteria on Prolific, the distributions of greed were vastly different between Study 1 and Study 3a (Figure 7). Whereas greed was distributed symmetrically around the scale midpoint in Study 1, Study 3a had many participants with extremely low greed and also a few participants with very high greed.
As a consequence, differences in the results could also arise from non-linear relationships. If the true relationship between greed and inaction inertia is nonlinear, then chances of detecting a linear effect would depend on the distribution of greed in a sample (Bless & Wänke, 2023). For example, if higher greed is associated with weaker inaction inertia effects only for non-extreme values of greed (e.g., between - 1.5 and + 1.5 SD), then this would make it much easier to obtain the expected interaction in Study 1 where greed was distributed mostly in that range.
We therefore added quadratic and cubic trends to the multilevel regressions, visualized in Figure 8. We indeed found evidence for non-linearity in the form of a cubic moderation. Whereas non-extreme greed values between - 1.5 and + 1.5 SD showed the expected interaction pattern with smaller inaction inertia for higher greed, this interaction reversed for extremely high and extremely low greed (see Figure 8). Notably, this pattern emerged for both purchase likelihood and valuation, albeit with slightly different shapes.
Note. Shaded areas represent 95% confidence intervals.
Note. Shaded areas represent 95% confidence intervals.
We also conducted a re-analysis of Study 3a using only participants with greed values between - 1.5 and + 1.5 SD, where the integrative data analysis had shown a linear pattern (see the Supplement, Table S8). This analysis showed all the significant effects we had expected in our hypotheses and that we had found in Study 1, including the moderated mediation (see Figure 9). Overall, this suggests that non-linear relationships could be responsible for why the interaction did not replicate in Study 3a. We did not find any evidence for such non-linear effects or differences in the greed distribution in the between-subjects data.
Note. Value difference was coded as -0.5 (large) and 0.5 (small). Greed, purchase likelihood, and valuation were standardized at the grand mean. Regression weights in brackets reflect the direct effects after controlling for valuation. * = p < .05, ** = p < .01, *** = p < .001.
Note. Value difference was coded as -0.5 (large) and 0.5 (small). Greed, purchase likelihood, and valuation were standardized at the grand mean. Regression weights in brackets reflect the direct effects after controlling for valuation. * = p < .05, ** = p < .01, *** = p < .001.
Differences in Greed
Another explanation for the failed replication might be differences in the psychometric properties of the greed scale. Although both studies showed a similar internal consistency, there might be differences in the factorial structure of the scale. According to the original paper, the scale should be unidimensional (Seuntjens, Zeelenberg, van de Ven, et al., 2015). However, exploratory factor analyses on our data suggested that there were two correlated but distinct factors8. One factor consisted of items 3, 6, and 7, capturing beliefs that more possessions are always desirable (e.g., “One can never have too much money”). The other factor consisted of items 1, 2, 4, and 5, assessing to what extent one is never satisfied with one’s possessions (e.g., “It doesn’t matter how much I have. I’m never completely satisfied”). In reference to Seuntjens et al’s (2015) theoretical differentiation of the core elements of greed, we termed these two factors acquisitiveness and dissatisfaction.
There were notable differences in the distributions of these two facets, most notably for dissatisfaction (Figure 10).
Given that we found non-linear effects for overall greed, we also examined non-linear effects of these two facets. We repeated the same polynomial regression models as reported in the previous section, but with either acquisitiveness or dissatisfaction instead of greed as predictor. We visualize these models in Figure 11. For acquisitiveness, we found the expected linear moderation effect, with higher purchase likelihood and a smaller inaction inertia effect at higher acquisitiveness. For dissatisfaction, however, the moderation was linear only for non-extreme values. For low (< -1 SD) and high (> +1 SD) dissatisfaction, the moderation went in the opposite direction than expected.
Note. Shaded areas represent 95% confidence intervals.
Note. Shaded areas represent 95% confidence intervals.
General Discussion
In this research, we examined the relationship between dispositional greed and the susceptibility to inaction inertia in three well-powered, pre-registered studies. The inaction inertia effect was robust in each study. Our two within-participant studies also showed that greed moderated the inaction inertia effect, at least for non-extreme levels of greed (Study 1, Integrative Data Analysis); Greedy people were less susceptible to the effect. They were overall more likely to act on a current offer, supporting the “greedy people want more” hypothesis. In addition, although we cannot infer causality, the valuation of the current offer mediated the effect of the missed discount on the likelihood of acting on the current discount, replicating Arkes et al. (2002) and Zeelenberg et al. (2006). This mediation was also moderated by greed. The greedier a person was, the higher they valued the current offer, and the less they were influenced in their valuation by missing a previous offer. However, the moderating effects did not manifest for more extreme levels of greed (Study 3a) and in between-participant studies (Study 2, Study 3b), and it further depended on the specific facet of greed.
Theoretical and Practical Implications
The first central implication of our research is that people dispositionally differ in inaction inertia as a function of greed, at least, if faced with multiple decision situations. Research on personality differences in inaction inertia has been scarce (for exceptions, see Mathmann et al., 2017; Van Putten et al., 2009), perhaps due to the larger sample sizes required (Schönbrodt & Perugini, 2013). Our findings not only suggest dispositional differences in inaction inertia but also that these relate to one of its underlying psychological processes. That is, (moderately) greedy people are less likely to devalue a product after having missed a previous discount, which partially mediates their stronger interest in taking the current offer. Yet, our findings also showcase the robustness of inaction inertia: Across all levels of greed, the inaction inertia effect was present. Thus, inaction inertia seems to be a general phenomenon influencing different individuals, even those rather greedy.
Second, our results give important insights into greed. Greed was consistently associated with higher purchase likelihood, both in within-subjects and between-subjects studies. Thus, greedy people always want more – apparently more than they hate losing. Generally, the hurt of having missed a discount does not offset the valuation of a current discount. As academic research on greed is a surprisingly new research field that has emerged only recently, our findings offer an important insight into what greed is and how it may influence decisions. Thus, our results suggest that dispositional greed may also relate to other sequential choice phenomena. The sunk cost effect, for example, refers to the finding that previous costs and investments strongly impact people’s current choices despite being irrational (Arkes & Blumer, 1985; Thaler, 1980). Based on the current results, one could speculate that greedy people are less likely to honor sunk costs, as they focus more on the future than on the past. Thus, greedy people should be more willing to stop ongoing projects and look for better opportunities in a sunk-cost situation.
The most intriguing and unexpected discovery in our studies is, however, that there are non-linear moderation effects of greed that also differ between different facets of greed. Whereas the original studies on the DGS found a unidimensional solution to be best (Seuntjens, Zeelenberg, van de Ven, et al., 2015, p. 921), Seuntjens et al. (2015) actually proposed a theoretical concept of greed building on two core dimensions, the desire to acquire more (so, acquisitiveness) and the dissatisfaction of never having enough (so, dissatisfaction). Not only did we find these two core dimensions in our data, but also different relationships with inaction inertia. Specifically, we observed linear effects for the facet of acquisition, but non-linear effects for the facet of dissatisfaction. We reason that this dissociation may be attributed to the operationalization of inaction inertia, which favors linear effects of acquisitiveness. The more people desire to acquire more, the more they want to purchase things, irrespective of previous offers. However, concerning the dissatisfaction of never having enough, the scenarios did not specify which possessions a participant already had - so participants may have independently filled in the lack of information. Participants who were somewhat dissatisfied tended to want more, showing a linear moderating effect. When dissatisfaction was high, participants were notably stingy, whereas those with very low dissatisfaction seemed satisfied with any offer. This could be because people on the extreme ends of dissatisfaction might actively perceive money differently (owning vs. using money; compare Lea & Webley, 2006). Additionally, the inaction inertia scenarios lacked the certainty of complete immersion, potentially constraining affective reactions (e.g., never being completely satisfied), while more rational arguments (e.g., always wanting more) persisted. Consequently, we advocate for future research to investigate the dissociation between acquisitiveness and dissatisfaction further. This could be achieved by modifying scenarios to be trade-based rather than purchase-based, incorporating personal possessions.
On a meta-scientific level, our work gives novel insights into why findings in psychology may not replicate. In previous decades, various reasons for replication failures in psychology have been suggested, ranging from insufficient methodological rigor in the original studies (e.g., Simmons et al., 2011) to changes in the construct validity of the study materials (e.g., Stroebe & Strack, 2014). In contrast, our research emphasizes another argument that has been less prominent in the debate on replication failures - non-linear relationships (Bless & Wänke, 2023). If two constructs X and Y actually have a non-linear relationship, then the chances of detecting a linear effect substantially depend on the levels of X represented in a study. In our case, the original (Study 1) and the replication study (Study 3) differed substantially in the distribution of greed, leading to vastly different results when searching for a mere linear relationship between greed (X) and the inaction inertia effect (Y). As a consequence, an (almost) exact replication study failed to replicate the original effect, despite the original study showing robust evidence and both the original and the replication study being sufficiently powered and adhering to the preregistered analysis. Only when restricting the analysis of the replication study to the range of greed from the original study, the same linear effect was obtained. Our research suggests that psychological research should consider non-linearity more often as a reason for replication failures.
Limitations
Despite several strengths of this research – such as large and heterogeneous sample sizes, different scenarios, and preregistrations – some limitations compromise our findings. Most obviously, we did not always find a consistent linear moderating effect of greed, limiting our findings to within-participants designs, and making us explore the data beyond our preregistered intent. Given that we detected no order effects for our within-participant design, we can only assume that the within-participant studies (vs. the between-participant studies) showed an effect of greed due to a contrast effect, that is, an increased focus on the numerical values. Alternative explanations for the null-findings in the preregistered linear analyses could be contextual specificity, or measurement sensitivity: Trait greed may manifest differently in different decision contexts, with our inaction inertia scenarios not having matched situations in which trait greed typically influences decision making (cf. Mussel et al., 2014). Or, the instruments used to assess trait greed and inaction inertia may not have been sensitive enough to detect subtle interactions between these constructs.
Second, our research design only offers correlative evidence (e.g., Fiedler et al., 2011) for both the effect of greed on inaction inertia as well as the role of valuation as mediator. Regarding the former, we are unaware of any convincing method to explicitly manipulate greed. Regarding the latter, we aimed to replicate previous research on valuation as a mediator (Arkes et al., 2002, Exp. 3; Zeelenberg et al., 2006, Exp. 4 & 5).
Third, we relied on non-consequential decisions in our study, possibly limiting the generalizability and external validity of our findings. However, it was most important to us to stay close to the original inaction inertia scenarios and not invite additional confounds. Additionally, we believe that using multiple scenarios over multiple studies will somewhat mitigate this point of criticism. Until now, the majority of inaction inertia research is based on hypothetical scenario questions. Recently, however, Shani et al. (2023) found clear evidence for the effect in a real-life setting.
Conclusion
Our research was guided by the idea that greed, an important personality trait, may influence people’s susceptibility to inaction inertia. The literature led us to propose two competing hypotheses about the role of greed. Multiple studies showed that given multiple purchase situations, greedy people want it all and are less impacted by missing discounts, although the latter finding is restricted in its generalizability to non-extreme levels of greed in within-participant scenarios. Nevertheless, our studies reveal several new insights into both inaction inertia and greed, such as a non-linear relation between greed and inaction inertia, and evidence that greed may be better understood as compromising two distinct dimensions (vs. being unidimensional). Based on this, we also highlight promising directions for future research, such as examining the differing effects of greed’s distinct dimensions.
Author Contributions
Contributed to conception and design: VR, MI, MZ
Contributed to acquisition of data: VR, MI, MZ
Contributed to analysis and interpretation of data: VR, MI
Drafted and/or revised the article: VR, MI, MZ
Approved the submitted version for publication: VR, MI, MZ
Acknowledgements
The authors have no acknowledgements to make.
Funding Information
This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors. The publication of this article was funded by the University of Mannheim.
Ethics Statement
This research was approved by the TSB-ethics review board of Tilburg University (RP275). All studies were performed in accordance with the American Psychological Association guidelines and regulations. Informed consent was obtained from participants prior to data collection.
Competing Interests
We have no known conflict of interest or affiliations to disclose.
Supplemental Material
All supplemental material can be found on this paper’s project page on ResearchBox: https://researchbox.org/374
Data Accessibility Statement
Data supporting this study are privately available from researchbox.org at https://researchbox.org/374. The study materials supporting this study are included within the supplementary materials.
Footnotes
Test family: F tests; Statistical test: ANOVA: Repeated measures, within-between interaction; Effect size f: 0.08; α err prob: .05; Power (1-β err prob): .95; Number of groups: 2; Number of measurements: 2; Corr among rep measures: .5; Nonsphericity correction: 1.
We also computed separate t-tests for each scenario, confirming that the scenarios showed the inaction inertia effect, all t’s > 4.97, all p’s < .001 (see Supplementary Materials, Table S1).
The zero-order correlations on person-level were r = .18 for greed and purchase likelihood and r = .17 for greed and valuation.
Robustness Checks (e.g., replications with different r packages) can be found in the Supplementary Materials.
Test family: F tests; Statistical test: ANOVA: Fixed effects, special, main effects and interactions; Effect size f: 0.10; α err prob: .05; Power (1-β err prob): .90; Numerator df: 1; Number of groups: 4
Once more, we computed separate t-tests for each scenario, showing a robust inaction inertia effect, all t’s > 8.82, all p’s < .001 (see Supplementary Materials, Table S2).
We also conducted Bayesian analyses for the correlation between the inaction inertia effect and greed in the two studies. For Study 1, there was strong evidence in favor of the expected correlation, BF10 > 1000. For Study 3a, there was moderate evidence against the expected correlation, BF10 = 0.12.
In Study 1, parallel analysis suggested three instead of two factors. However, only one item loaded on the third factor, and therefore we also extracted two factors to have a more parsimonious model. The correlation between the two subscales was r = .59.