Abstract
Sadness has an appraisal theme of loss and helplessness and prior research has shown that sadness leads to increased (vs neutral or happy) hedonic food consumption (e.g. M&M’s). In this research, we test the robustness of sadness’ effect on consumption and further argue that it is the innate helplessness associated with sadness (situational helplessness) that links sadness and hedonic consumption. Specifically, we find that an opportunity to make a choice (increased sense of control) attenuates sadness’ effect on hedonic consumption (study 1) and that making a choice attenuates the helplessness experienced by sad individuals at that moment (general helplessness), thereby reducing hedonic consumption (study 2). Moreover, we show that this effect extends to choice for both self and others (study 2). Implications for research and practice are discussed.
JEL Classification:
1. Introduction
There is no mystery so great as misery.
Imagine that Pat has just received some upsetting news. Pat is feeling sad and down. Sometime later, he is looking for a snack and sees a packet of butter popcorn and opens it. Now imagine that just before he gets ready to eat, he gets a call from a friend who tells him that he is buying a new car and cannot decide which of the two brands in his shortlist he should finally consider. The friend asks Pat to choose the two car brands for him, from the shortlist. A couple of important questions emerge. Will the amount of butter popcorn that Pat eats be influenced because he is sad? Importantly, will the fact that he made a choice before consuming the popcorn influence how much he eats? Existing research (Garg et al., 2007; Kemp et al., 2014; Tice et al., 2001) has already answered the first question and established that incidental sadness increases consumption of tasty, unhealthful, hedonic food products. Our research focuses on the second question and suggests that the opportunity to make a choice will in fact influence Pat’s consumption. In the process, our results highlight the mechanism driving the sadness–hedonic consumption relationship.
While research has established the persistence of emotion effects in affecting behavior across related (e.g. Ashkanasy, 2002; Luce, 1998) and unrelated tasks (for reviews, see Forgas, 1995; Loewenstein and Lerner, 2003), it is only in the last decade or so that the focus has shifted from the simple, dichotomous, valence-based approach (positive vs negative) to the examination of specific, discrete emotions (Lerner and Keltner, 2000, 2001). This has enabled differences to surface even across emotions of the same valence such as anger, sadness, and fear (Garg et al., 2005b, 2017; Lerner et al., 2004). Among the negative emotions, sadness has been studied extensively, and this is understandable given that it is one of the most often experienced emotions. However, the examination of sadness on subsequent consumption behaviors has primarily adopted a valence-based lens in the existing research. This is in contrast to existing research which has demonstrated that relative to other emotions, including other negative as well as positive emotions, sadness has unique influences (Garg et al., 2007; Garg and Lerner, 2013; Lerner et al., 2004). Specifically, sadness leads to positive consumption behaviors such as increased valuation (vs disgust) of new products (Lerner et al., 2004) and increased hedonic consumption (vs happiness, Garg et al., 2007).
Each emotion is associated with a core theme; sadness has one of loss and helplessness (Keltner et al., 1993; Lazarus, 1991). This theme in turn triggers implicit goals in line with the unique appraisal of the emotion. For sadness, the implicit goals evoked are as follows: changing one’s circumstances (Lerner et al., 2004) and acquiring or substituting a reward (Raghunathan and Pham, 1999). Behavioral evidence supporting action tendencies arising from these compensatory consumption goals has been evinced across different domains (e.g. Chien-Huang and Hung-Chou, 2012) such as food consumption, where sad individuals consumed significantly more of a hedonic food product (e.g. butter popcorn and M&M’s candies) compared to those who were happy (Garg et al., 2007) and new product acquisition (e.g. a water bottle), where sad individuals were willing to pay significantly more than those in a neutral state (Lerner et al., 2004). In addition, sad individuals choose the new object more often than do neutral individuals when presented with a choice between a new object and money (Cryder et al., 2008).
Given sadness’ theme of loss and helplessness, two distinct pathways suggest themselves in an effort to mitigate sadness’ effect: one, to attenuate the sense of “loss” associated with sadness via compensatory consumption and second, to decouple the sense of “helplessness” associated with sadness (labeled as “situational helplessness”). However, a recent paper finds that compensating sadness for its sense of loss is not sufficient and that sadness’ effect replicates even when individuals have an opportunity to engage in hedonic consumption (Garg et al., 2018). The current research thus seeks to elucidate the other potential mechanism—helplessness—that might be driving the sadness–consumption relationship and establish the theory-driven mediators/moderators that can explain and attenuate the influence of sadness on consumption behaviors. Specifically, we argue that the theme of helplessness associated with sadness is responsible for its effect on consumption. Furthermore, we find that this effect is mitigated by giving people a choice that allows them to exert some level of control and thereby alleviate their sense of felt helplessness at the current moment (labeled as “general helplessness”).
Our research makes two main contributions. First, it establishes a key mediator underlying the sadness–consumption relationship. Specifically, study 1 clarifies that because sadness is characterized by a sense of loss and helplessness, offering a choice and the associated control to sad individuals delinks the sadness and hedonic consumption relationship. This effect is not obtained for sad individuals who are simply given a gift (vs a choice of a gift). It also rules out an alternative explanation that this influence of choice might be driven by the nature of the gift (hedonic or utilitarian) to show that the effect of choice is robust and independent of the nature of the gift.
Second, our findings further explicate the nature of the mediation and lead to a deeper understanding of sadness in particular and incidental emotion effects in general. Study 2 not only establishes that offering choice reduces felt helplessness at the current time (general helplessness) which delinks sadness and consumption but also examines whether the locus of control (self vs others) in choice matters in attenuating sadness’ effects. Prior research has shown that sad individuals do not recognize that they tend to consume more of hedonic food products (Garg et al., 2007) or choose to pay more for a new product (Cryder et al., 2008). Consistent with these findings, study 2 establishes that the sanguine effects of choice are not limited to choices only relevant to the self, further attesting to the robustness of these effects.
Overall, our research provides insights into the robust effect of sadness on judgments and decisions along with its mediating and moderating factors. In the next section, we summarize the relevant research on sadness and consumption and develop our hypotheses. We then present results from two studies and a pilot study. Finally, we close with a discussion of the implications of our findings, potential limitations, and avenues for future research.
2. Hypothesis development
2.1. Sadness and consumption
Sadness is characterized by a theme of loss and helplessness which in turn can evoke a goal of reward replacement (Smith and Lazarus, 1993). Prior research has revealed that people prefer “comfort foods” or drinks, such as ice cream or hot tea, when sad or depressed (Mick and DeMoss, 1990). Such hedonic products are generally understood to be goods that “provide more experiential consumption, fun, pleasure, and excitement” than more utilitarian products (Dhar and Wertenbroch, 2000: 60) and are associated to a greater extent with the idea of indulgent consumption (Martins et al., 2015; Shao and Shao, 2011). Specifically, while happy individuals display greater control over when consuming a hedonic, rewarding food, sad individuals tend to over-consume (Garg et al., 2007; Tice et al., 2001). However, this effect is not obtained for a healthful, less hedonically rewarding food, such as raisins. Garg et al. (2007) report that happy individuals ate more raisins (perceived to be healthful) than did sad individuals (study 2), while Lerner et al. (2004) find that sadness carries over to influence financial decisions even where real money is at stake. Specifically, in a demonstration of the “misery-is-not-miserly effect” (Cryder et al., 2008), sad individuals were willing to give up more money to acquire a commodity than were neutral individuals.
Interestingly, recent research suggests that while sadness might be associated with both a sense of loss and helplessness, compensating for loss does not seem to alleviate sadness’ tendency to over-consume/spend (Garg et al., 2018; Garg and Lerner, 2013). Specifically, Garg et al. (2018) report that even when participants were offered an opportunity to engage in hedonic consumption in another domain, it did not attenuate sadness’ effect on consumption later on: an effect that was robust across different types of intervening rewarding opportunities. Given the need for reward associated with sadness and its intractability in the face of alternative hedonic, rewarding opportunities, we argue that sadness’ effect on consumption will be attenuated by providing individuals with an opportunity to exercise a choice and thereby control. A theoretically rich line of research has established that individuals prefer perceived control and choice (vs no-choice), even when the choices are trivial (Averill, 1973; Iyengar and Lepper, 2000; Langer, 1975) . This is because choice relates to an increase in self-determination and sense of control (e.g. (Deci and Ryan, 1985; Langer, 1975). Thus, choice and the associated control should delink the sadness–helplessness connection and thereby attenuate its impact on consumption, highlighting the need to understand the process driving the association between sadness and compensatory consumption.
Garg and Lerner (2013) report that this might indeed be the case and that offering a choice to participants decouples the link between sadness and consumption. While their findings are novel and interesting, two key unanswered questions need to be addressed. First, in their study, participants in the choice (“high control”) condition were offered the choice of a gift after the emotion induction, whereas participants in the endowment (“baseline”) condition were given the gift before the emotion induction. Although this design implies that their baseline condition was particularly suited to replicating the original sadness–consumption findings, it introduces a timing confound across the two conditions.
Second, while they established that sadness-induced helplessness mediates the sadness–consumption relationship, they did not ascertain whether participants in the “high control” condition actually experience attenuation in their current sense of perceived helplessness (general helplessness) versus participants who did not receive a choice but were simply endowed with the gift. If sadness’ effect on consumption operates via helplessness, then we should see attenuation in felt helplessness, post choice.
Thus, the present research builds on and extends the research in the sadness–consumption domain by combining research examining sadness’ effect on consumption (Garg et al., 2007; Garg and Lerner, 2013; Tice et al., 2001) along with that on personal control (Langer, 1975). Specifically, we address the gaps in the literature and examine whether a theme of helplessness underlies sadness’ effect on consumption and whether this effect can be mitigated by offering an opportunity to choose a gift to individuals. In particular, we address the timing issue in Garg and Lerner (2013) and present the gift (no-choice, baseline) and the choice (high control) after the emotion manipulation in both conditions, thus eliminating potential confounds due to differential timing across the two conditions. Furthermore, in contrast to Garg and Lerner (2013), we test the robustness of this effect by examining how factors such as the nature of the gift (hedonic or utilitarian) and choice for self versus others impact sadness’ effect on consumption. Overall, we not only demonstrate the robustness of the effect but also provide a more comprehensive understanding of the links between sadness and hedonic consumption.
The no-choice (baseline) condition further enables us to test and replicate the prediction from existing literature. That is, sad individuals who simply receive the gift in the no-choice condition should replicate prior findings and over-consume compared to happy individuals in the no-choice condition. Thus, we predict that results demonstrated in prior research will emerge. This is because although participants in the no-choice condition receive the gift, we do not expect it to interfere with the sadness–helplessness connection:
H1: In the no-choice (gift) condition, sad individuals will consume significantly more of the hedonic food product, relative to happy individuals.
While replicating sadness’ effect on hedonic consumption establishes the robustness of this effect, we are more interested in comparing participants endowed with a gift in the no-choice condition with those in the choice condition, in which participants get a choice between a hedonic and a non-hedonic product. Existing literature shows that individuals prefer having a choice (vs no-choice) and that choice is linked to positive outcomes such as increase in intrinsic motivation, perceived control, as well as life satisfaction (Deci and Ryan, 1985, 1987; Langer, 1975; Langer and Rodin, 1976; Schulz and Hanusa, 1978). Langer (1975) further illustrates that the effect of choice is so powerful that participants in the studies acted as if they could affect the outcome (an “illusion of control”) of a chance-based lottery. Thus, we argue that when participants have a choice, it gives them a sense of increased control and empowerment. This increased control is in turn helpful in counteracting sadness-related (situational) helplessness and thus delinks the sadness–hedonic consumption relationship. That is:
H2: Choice of hedonic gift versus simple endowment (no-choice) will influence consumption, such that consumption will be lower for sad individuals in the choice condition than for sad individuals in the no-choice condition.
H3: Situational helplessness will mediate the emotion–consumption relationship, but only in the no-choice condition (and not in the choice condition).
2.2. Choice for self versus others
From a theoretical perspective, it is important to further understand the role that choice plays in the sadness–consumption link. Prior research points to the fact that choice increases a sense of personal control (Langer, 1975), which is attenuated when individuals experience sadness and is one of the key appraisals of sadness (Smith and Ellsworth, 1985). Furthermore, there is research to suggest that increased sense of control is valued and linked to greater intrinsic motivation and interest, less stress, and overall improved physical and psychological health. However, to the best of our knowledge, the literature is silent regarding whether the therapeutic effects of control are specific to choices made for the self (and therefore, relevant to self) or whether these benefits are obtained even when choices are made for others (and hence, irrelevant to self). By extension, will the link between sadness and helplessness be attenuated even when choice is made for others?
We argue that the sadness–helplessness consumption link will be decoupled in both conditions when the choice is relevant to the self and when it is made for others. The reason for this is that the effect of sadness on consumption tendencies is not usually salient (Cryder et al., 2008; Garg et al., 2007). When specifically asked post consumption, individuals state that they did not eat more food (Garg et al., 2007) or pay more for a product (Cryder et al., 2008) when sad. Given that the sadness–consumption link we are studying is subtle and not salient for most part (unlike retail therapy), the effect of choice on decoupling the sadness–helplessness link might also not be salient to the individual. If that is so, the nature of control should be less pertinent, as long as there is some level of control. In other words, we argue that the opportunity to exert control is the prevailing factor in the decoupling process. To test this argument, a new condition, choice-for-other, is introduced in which respondents are endowed with the gift but have to choose for another participant. We thus, expect the following:
H4: Choice of hedonic gift for others will influence the amount consumed by sad individuals such that consumption in the choice-for-other condition will be significantly lower versus the no-choice condition.
Another theoretically relevant question that prior research (e.g. Garg and Lerner, 2013) does not answer is whether current levels of the felt helplessness differ post choice. If our premise that offering choice to sad individuals increases their sense of control and thereby decreases the sense of helplessness leading to decreased consumption is true, then we should find lower levels of felt helplessness in choice conditions, post choice. Thus, the distinction between felt situational helplessness (related to the emotion-inducing event) and general helplessness (related to helplessness in general) at the current moment is crucial and provides an important step that is necessary to advance the theoretical arguments presented in this research. Specifically, general helplessness should capture the changes in the sadness-associated carryover effects in felt helplessness, post choice. In essence, in the absence of choice, general helplessness (which taps into felt helplessness at the moment) should mimic situational helplessness (associated with the emotion-inducing situation) patterns, but general helplessness should be attenuated when a choice is presented to individuals. More formally, we expect that:
H5: Choice of hedonic gift, both for oneself and others, will attenuate felt (general) helplessness in sad individuals versus those in the sad no-choice condition.
3. Study 1: role of choice in sadness’ effect on consumption
Study 1 focuses on the effect of choice versus no-choice on the sadness–hedonic food consumption relationship (e.g. Garg et al., 2007; Tice et al., 2001) while addressing the timing issue in Garg and Lerner’s (2013) study. Specifically, study 1 tests H1–H3. In study 1, half the participants receive a choice of products (a hedonic vs a non-hedonic gift) after the emotion induction, whereas others are simply endowed with the hedonic gift, also after the emotion induction. Thus, the timing of gift is the same across the no-choice versus choice conditions and cannot unintentionally influence the results.
3.1. Participants and experimental design
Participants from the undergraduate population of a large university in the United States were recruited in exchange for extra credit (n = 97). They were randomly assigned to one of the four conditions: 2 (emotion: sad, happy) × 2 (choice: no (gift), yes) between-subjects design.
Garg et al. (2007) tested the perceived hedonicity of various food products such as buttered popcorn, M&M’s, and raisins. The hedonic nature of each product was measured using five items—“makes me feel good,” “tastes great,” “lifts me up when I am down,” “is pleasurable,” and “is comforting.” Buttered popcorn and M&M’s candies were rated as the most hedonic products on the list. Thus, we use the amount of M&M’s consumed as the dependent variable in this study.
3.2. Procedure
Participants were seated in private cubicles. This ensured that they had no visual access to others in the lab. Each cubicle had a sealed brown bag in the corner of the desk. In the first part of the study, emotions were manipulated, while the second part consisted of reading and evaluating a narrative. We closely followed the procedure that has been used in prior emotion–consumption research (Garg et al., 2007).
3.2.1. Emotion inductions
Emotions were induced using a written manipulation (e.g. Lerner and Keltner, 2001). The manipulation consists of two questions where participants are first asked to describe three or four things that make them feel sad/happy, and then in the second question, they are asked to describe in detail the one thing that makes (or made) them feel most sad/happy (e.g. Garg et al., 2005). This emotion manipulation is well-established and has been successfully implemented by researchers in marketing as well as social cognition (e.g. Garg et al., 2017, 2007; Lerner et al., 2004).
3.2.2. Hedonic product after emotion induction
At this point, a note in the survey packet asked participants to open the brown bag. Participants in the no-choice condition found a bar of Ghirardelli chocolate inside the bag, while those in the choice condition were asked to choose one of the two gifts in the bag (a bar of Ghirardelli chocolate and a Bic ballpoint pen), as a token of appreciation for their participation. These options were particularly selected to effectively control for the product each participant ends up with. This is because almost all of the participants were expected to prefer the luxury chocolate bar compared to the inexpensive pen (the results are robust to the exclusion of participants choosing the pen). This procedure is similar to that used by Higgins et al. (2003), who controlled for choice by offering participants a choice between an inexpensive pen and a more desirable Columbia University coffee mug.
3.2.3. Mood-maintaining filler task
To ensure that the manipulations (emotion and choice) were not concurrent to consumption, participants were offered the opportunity to consume, post manipulations. Specifically, after the choice manipulation, a note asked participants to shift to the adjoining cubicle to complete the remaining surveys. In this ostensibly unrelated “study 2,” participants read a two-page event narrative and then evaluated several statements related to it (e.g. how well could they relate to the story, how sad/happy/interesting was), using a 7-point scale (1 = strongly disagree; 7 = strongly agree). This filler task was introduced for the express purpose of giving participants time to consume M&M’s candies (from a pre-weighed bowl) which were offered to them as refreshments at this time, along with a bottle of water. The narratives were adopted from Garg et al. (2007). Their purpose was to maintain each participant’s affective state while giving the opportunity to consume. The sad story, presented to participants in the sadness condition, described the tragic loss of seven children in a fire and included the emotional responses of eyewitnesses, while the happy story, presented to participants in the happiness condition, described a chance reunion of four friends.
3.2.4. Emotion manipulation checks
Participants self-reported their felt emotions in response to the autobiographical event they were asked to write about (sadness or happiness). Of the 18 affective states measured, only 7 were of interest to us. “Upset,” “downhearted,” “depressed,” and “sad” comprised a sadness factor (α = .96) and “joyful,” “happy,” and “elated” comprised a happiness factor (α = .92). Response scales ranged from zero (“did not experience the emotion at all”) to eight (“experienced the emotion more strongly than ever before”).
3.2.5. Emotion-induced situational helplessness measures
Next, we collected measures that would help assess the difference in emotion-induced perceived “situational helplessness” that arises in response to the emotion event (sad or happy). Situational helplessness was ascertained by three items (α = .73; items 2–3 reverse coded), adapted from Hartline and Ferrell (1996), each beginning, “Looking back on it (refers to the autobiographical situation in the emotion manipulation) now, I feel …” (1) “I was helpless to change the situation,” (2) “I had the resources (time, energy, or skill) to handle the situation(s),” and (3) “I was well prepared for the situation(s) I faced.” These items were averaged to form a composite score for situational helplessness.
3.2.6. Additional measures
To control for potential confounds, participants were asked about the time elapsed since their last meal, the extent to which they felt hungry/full before and after the studies, their dieting status, and whether they tried to restrict or monitor their food consumption during the study. They were also asked about their average frequency of chocolate consumption (less than once a week, one to three times a week, or more than three times a week), whether they intended to keep the “token gift” for themselves or give it to someone else, and whether they considered Ghirardelli to be a brand of luxury chocolates. Finally, measures regarding gender, age, and race/ethnicity were collected, and participants were debriefed and dismissed. Each bowl of M&M’s was then individually reweighed to obtain the amount consumed (in grams).
3.3. Results and discussions
3.3.1. Manipulation checks
Individual analysis of variance (ANOVA) revealed strong emotion induction effects for both sadness (F(1, 93) = 140.82, p < .0001) and happiness (F(1, 93) = 206.38, p < .0001). Sad participants reported feeling significantly more sad than those in the happy conditions (Msad = 5.33, Mhappy = 0.99; t(93) = 11.87, p < .0001; Cohen’s d = 2.44). Similarly, those in the happy conditions reported feeling happier than those in the sad conditions (Msad = 0.62, Mhappy = 4.87; t(93) = 14.37, p < .0001; Cohen’s d = 2.95).
We also ran a 2 (emotion) × 2 (choice) ANOVA with situational (felt) helplessness associated with the emotion event described, as the dependent variable. As expected, only the main effect of emotion was significant (F(1, 93) = 52.59, p < .0001) such that higher levels of helplessness were reported in the sad versus happy conditions (Msad = 4.63 vs Mhappy = 2.73; t(95) = 7.24, p < .0001; Cohen’s d = 1.48) (see Figure 1).
Situational helplessness measures in study 1.
3.3.2. Main results
A 2 (emotion: sad, happy) × 2 (no-choice (gift), choice) ANOVA (r2 = .19) with the amount of M&M’s consumed as the dependent variable was conducted to test the hypotheses (n = 93 for this analyses due to missing values). The results support H1–H3 (see Figure 2). As predicted, the interaction between emotion and gift condition was significant (F(1, 81) = 4.03, p < .05). Specifically, sad (vs happy) participants consumed significantly more M&M’s (Msad = 22.14 vs Mhappy = 5.92 grams; t(81) = 2.56, p = .01; Cohen’s d = 0.40) in the no-choice (baseline) condition, supporting H1. Importantly, this difference does not manifest in the presence of choice (Msad-choice = 12.15 vs Mhappy-choice = 8.90 grams; t(81) = 0.52, ns; Cohen’s d = 0.11). Control variables such as an individual’s dieting status, intensity of felt sadness, whether they tried to restrict or monitor their consumption of M&M’s, and whether they meant to keep the chocolate bar for self or give it to someone were also included in this analyses to account for individual and situational differences. None of them was significant and will not be discussed further.
Moderating effect of choice on the amount of M&M’s consumed in study 1.
H2 predicts that when individuals have an option of choosing the gift, the effect of sadness will be attenuated and the amount of M&M’s consumed will be lower compared to that for the no-choice sad condition. That is, when individuals have an option of choosing the gift, the effect of sadness will be attenuated. Our results support this prediction (Msad-no-choice = 22.14 vs Msad-choice = 12.15 grams; t(81) = 2.26, p < .05; Cohen’s d = 0.53).
In addition, we examined the effect of gender on consumption. In keeping with prior research in the emotion and consumption domain (see Garg et al., 2007; Garg and Lerner, 2013), the effect of gender was found to be insignificant (F(1, 80) = 0.04, ns). Thus, our results support the premise that in the general population (e.g. non-dieters), gender does not influence the effect of emotion on hedonic consumption.
3.3.3. Situational helplessness as a mediator
Garg and Lerner (2013) argue that it is the innate helplessness associated with sadness rather than loss that drives the effect of sadness on consumption. However, in their study, they examined sadness versus neutral, and as we discussed earlier, there is the critical issue of timing across the control conditions. Thus, we test whether emotion-induced situational helplessness acts as a mediator when sadness is contrasted with happiness and when both the choice and no-choice conditions receive the gift post emotion induction.
As reported above, sad individuals felt more helpless (lower scores) compared to happy individuals (Msad = 4.63 vs Mhappy = 2.73; t(95) = 7.24, p < .0001). If our thesis is correct that it is choice rather than receiving the gift that attenuates sadness’ effect, then situational helplessness should mediate sadness’ effect on consumption but only in the no-choice condition and not in the choice (high control) condition (H3)—moderated mediation. To test this prediction, we conducted bootstrapping analysis (model 4, Preacher and Hayes, 2004) separately for the no-choice and choice conditions (i.e. two sets of one factor (emotion: sad, happiness) analyses). This is similar to a procedure adopted by other researchers (e.g. Garg and Lerner, 2013; Govind et al., 2018; Lerner et al., 1998).
As hypothesized, in the no-choice conditions where participants are endowed with the gift but without a choice, the 90% confidence interval (CI) around the indirect effect of 4.03 did not include zero (0.0031 to 9.4996). This supports the argument that emotion influences the amount consumed through situational helplessness in the no-choice condition. However, in the choice condition, the link between sadness and consumption is no longer predicted by situational helplessness, as expected (−1.76, 90% CI includes zero: −7.2146 to 4.1302). Statistically, these two effects are significantly different at p < .05.
3.4. Alternative explanation: does the nature of the gift matter?
The results of study 1 establish that helplessness is a key mediator of the sadness–consumption relationship and that choice can attenuate sadness’ deleterious effect on consumption. However, we do not know whether the effect of choice is dependent on the nature of the gift (hedonic or utilitarian) that participants receive. That is, will the effect of choice on sadness and helplessness still hold if the participants are offered a utilitarian gift rather than a hedonic one (chocolate bar)? We ran a small pilot study to address this issue. A total of 47 participants were recruited and a 2 (emotion: sad, neutral) × 2 (no-choice, choice) between-subjects design was implemented. We contrast sadness with neutral rather than happiness in this pilot study. This is to increase the generalizability of our results and to examine how they compare to other studies that have compared sadness’ compensatory tendencies to neutral rather than a positive emotion (Cryder et al., 2008; Garg and Lerner, 2013).
The procedure of the study is the same as that of study 1 with the exception of the gift offered. Instead of offering a hedonic gift, we offer a utilitarian one (coffee mug). 1 Thus, one half of the participants are endowed (simply gifted) with a coffee mug, whereas the other half are given a choice between a coffee mug and a Bic ballpoint pen.
To examine emotion effects on situational helplessness, we ran a 2 (emotion) × 2 (choice) ANOVA with situational helplessness associated with the emotion event described, as the dependent variable. As in study 1, only the main effect of emotion was significant (F(1, 43) = 10.53, p < .01) such that higher levels of helplessness were reported in the sad versus the neutral conditions (Msad = 4.42 vs Mneutral = 3.29; t(45) = 3.38, p < .01; Cohen’s d = 1.04).
Next, a 2 (emotion: sad, neutral) × 2 (no-choice, choice) ANOVA (r2 = .41) with the amount of M&M’s consumed as the dependent variable was run. Results find a significant main effect of emotion (F(1, 33) = 4.69, p < .05) but more importantly, they find a significant interaction between emotion and choice condition (F(1, 33) = 5.75, p < .05). In keeping with H1, a simple effects test shows that sad individuals in the no-choice condition consumed more than those in the neutral condition (Msad-no-choice = 33.17 vs Mneutral-no-choice = 6.08 grams; t(33) = 3.20, p < .01; Cohen’s d = 0.90). Furthermore, the amount of M&M’s consumed across sad conditions is lower when a choice of a gift is offered to participants versus the no-choice condition (Msad-no-choice = 33.17 vs Msad-choice = 22.07 grams; t(33) = 1.42, p = .08, one-tailed; Cohen’s d = 0.22). Thus, H2 is supported.
Overall, study 1 supports our thesis that choice affords control and a certain sense of empowerment to sad individuals, regardless of the nature of the gift involved. This in turn attenuates sadness’ effect on subsequent consumption.
4. Study 2: effect of choice for self or others on sadness–consumption relationship
Study 1 shows that (1) providing a choice of gift decouples the sadness–helplessness link and (2) emotion-induced situational helplessness is significantly higher for sad than happy conditions. In study 2, we examine whether the decoupling of the sadness–helplessness consumption link also obtains when the choice is made for others (H4). Furthermore, we test H5 by examining whether general (but not situational) helplessness at the current moment is attenuated when participants are given an opportunity to exert control via a choice.
4.1. Respondents and experimental design
Participants were recruited from the undergraduate population of a large university in the United States for participation in exchange for course credit (n = 140). To test the hypotheses, a 2 (emotion: sad, happy) × 3 (no-choice, choice-for-other, choice-for-self) between-subjects design was run.
4.2. Procedure
The procedure of study 2 is the same as study 1 with the exception of an additional choice condition and general helplessness measures. These are described in detail below.
4.2.1. Choice manipulation after emotion induction
After the emotion induction task, a note instructed one-third of the participants to open the brown bag on their desk which contained a bar of Ghirardelli chocolate (no-choice condition). The other third of the participants were asked to open the bag and choose one of the two gifts in the bag (a bar of Ghirardelli chocolate and a Bic ballpoint pen; choice-for-self condition), while the remaining one-third in the choice-for-other condition were asked to open the brown bag which contained a bar of Ghirardelli chocolate, but were then instructed to make a selection for participants in the next session (choice was between Ghirardelli chocolate bar and Bic ballpoint pen). This was the choice-for-other condition. After this, all participants completed the rest of the measures and checks as in study 1.
4.2.2. Situational and general helplessness measures
Post choice, perceived helplessness across sad and happy individuals was measured, vis-à-vis the original emotion-inducing event (situational helplessness) as well as in general in life at that moment (general helplessness). That is, measures of helplessness related to the emotion situation and felt helplessness at the current time (7-point scales) were collected. As in study 1, emotion-related situational helplessness was measured by three items (α = .73). Post-choice measure of felt general helplessness at that moment was assessed by four items (α = .76; items 2–4 reverse coded), each beginning, “In general,” (refers to their experiences in life) (1) “I feel I am helpless to change my situations,” (2) “I feel I have the resources (in terms of time, energy, or skill) to handle most situation(s),” (3) “I feel that I am well prepared for the situation(s) I face,” and (4) “I feel I am in control of the situations I am in.” Thus, the two measures, situational and general helplessness, tap into two different constructs: the innate helplessness evoked by the emotion and associated with the situation described (as part of the emotion manipulation) and the felt helplessness at the current moment, respectively. The latter (general helplessness) helps us assess whether the presence or absence of choice influences the level of current felt helplessness.
4.3. Results and discussions
4.3.1. Manipulation checks
As in study 1, “upset,” “downhearted,” “depressed,” and “sad” comprised a sadness factor (α = .96) and “joyful,” “happy,” and “elated” comprised a happiness factor (α = .92). Individual ANOVAs established strong emotion effects: sadness (F(1, 137) = 235.21, p < .0001) and happiness (F(1, 136) = 308.55, p < .0001). Participants felt significantly more sad in the sad conditions than in the happy conditions (Msad = 5.06, Mhappy = 0.72; t(137) = 15.34, p < .0001; Cohen’s d = 2.62) and significantly more happy in the happy conditions than in the sad conditions (Msad = 0.77, Mhappy = 5.27; t(136) = 17.57, p < .0001; Cohen’s d = 2.98). We also ran a 2 (emotion) × 3 (choice) ANOVA with self-reported situational helplessness related to the emotion-inducing event as the dependent variable. As in study 1, only the main effect of emotion was significant (F(1, 134) = 92.35, p < .0001) such that sad individuals experienced higher levels of situational helplessness than did those in the happiness conditions (Msad = 4.68 vs Mhappy = 2.51; t(138) = 9.75, p < .0001; Cohen’s d = 1.66). This replicates our findings in both study 1 as well as the pilot study, regarding emotion-related situational helplessness.
4.3.2. Does choice alleviate general helplessness?
We next ran a 2 (emotion) × 3 (choice) ANOVA on self-reported general helplessness which captures felt helplessness, at the current moment. Contrasting the patterns across general helplessness and situational helplessness (related to the emotion event) is important to assess whether choice of a gift makes a difference to how helpless participants feel at that point in time and thus delink the sadness–helplessness consumption relationships. As predicted, emotion effects on general helplessness are not significant (F(1, 134) = 1.41, ns). Importantly, simple effects analyses yield further insights and allow us to test H5. As predicted by H5, general helplessness scores are significantly higher for the sadness no-choice condition (M = 2.80) versus sadness choice-for-self (Mchoice-for-self = 2.20; t(134) = 2.24, p < .05; Cohen’s d = 0.60) and sadness choice-for-others conditions (Mchoice-for-other = 2.32; t(134) = 1.80, p = .07; Cohen’s d = 0.50). Furthermore, in the choice conditions, differences in general helplessness between happy and sad participants are not statistically significant (Msad-choice-for-self = 2.20 vs Mhappy-choice-for-self = 2.36; t(134) = 0.60, ns; Cohen’s d = 0.16; Msad-choice-for-other = 2.32 vs Mhappy-choice-for-other = 2.05; t(134) = 0.96, ns; Cohen’s d = 0.35). Thus, when a choice is offered (for self or for others), felt helplessness goes down and sad participants feel more in control, which is consistent with our hypotheses. Figure 3 presents the details for situational and general helplessness across the six conditions.
Situational and general helplessness measures in study 2.
4.3.3. Main results
To test the hypotheses, a 2 (emotion: sad, happy) × 3 (no-choice, choice-for-other, choice-for-self) ANOVA (r2 = .18) with the amount of M&M’s consumed as the dependent variable was run (n = 138 due to missing values). As in study 1, control variables were used in the analyses such as intensity of felt sadness, whether individuals tried to monitor their consumption of M&M’s, and their level of hunger to account for individual and situational differences. Only the level of hunger participants’ felt when they arrived at the session was significant (F(1, 129) = 15.54, p < .01). None of the other covariates was significant and will not be discussed further. Figure 4 illustrates the means across the conditions. Neither the main effect of emotion nor the main effect of choice condition was significant, but as expected, the interaction between emotion and the choice condition was significant (F(2, 129) = 3.35, p < .05).
Effect of choice for self/others on the amount of M&M’s consumed in study 2.
The no-choice conditions are the baseline emotion conditions and were used once again to test H1. Comparing the means between the sadness and the happiness no-choice conditions, we find that participants in the sadness condition consumed more than those in the happiness condition (Msad-no-choice = 27.11 vs Mhappiness-no-choice = 8.34 grams; t(129) = 2.74, p < .01; Cohen’s d = 0.76), supporting H1.
H2 predicted that the amount of M&M’s consumed across sadness conditions would be lower when a choice of a gift was offered to participants versus the no-choice condition. Results support this prediction (Msad-no-choice = 27.11 vs Msad-choice-for-self = 16.73 grams; t(129) = 1.96, p = .05; Cohen’s d = 0.34).
Finally, study 2 tests the prediction that the choice need not be relevant for the self in order to effectively attenuate sadness’ effect (H4). Specifically, it predicts that consumption in the choice-for-other condition would be lower than that in the no-choice condition. As predicted, sad individuals consumed significantly less in the choice-for-other condition than in the no-choice condition (Mchoice-for-other = 14.90 vs Mno-choice = 27.11 grams; t(129) = 2.31, p < .05; Cohen’s d = 0.53).
Overall, this study replicates the results of study 1, establishing the robustness of the sadness–consumption relationship. Furthermore, it supports our premise that sad individuals do not discriminate between choice for self versus one for others. Moreover, these results shed light on the underlying appraisal themes of helplessness and loss of control innate to sadness and demonstrate that having a choice lessens feelings of general helplessness at that moment which in turn helps delink sadness from increased hedonic consumption.
5. General discussion
That incidental emotions impact consumers’ judgment and decision-making is a well-established fact in the literature. Our objective was to develop a deeper understanding of how the two are connected and the implications thereof. Given the concerns with unhealthful (hedonic) consumption, this is especially important in the domain of food consumption.
5.1. Conceptual contributions
Across two studies, our research contributes to two key aspects of the literature on emotions and their influence on judgment and decision-making. First, rather than focusing on the global negativity or positivity of emotions, it contributes to the growing research on the effect of discrete emotions. Second, we not only replicate existing research that shows the carryover effect of sadness (and emotions, more generally) on consumption (Garg et al., 2007; Garg and Lerner, 2013; Lerner et al., 2004) but also extend it by explicating its underlying mechanism. To test the relationship between sadness and consumption, we manipulate sadness and happiness/neutral emotions across the studies and test choice as a moderator in a variety of forms (for self and for others). It is important to note that we distinguish between the incidental effects of emotions, sadness in particular, and the more conscious form of mood management such as retail therapy. Our focus lies in understanding the effect of sadness in situations where participants engage in over-consumption (e.g. overeating and overspending) even though they are not aware of doing so and actually deny having done so (Cryder et al., 2008; Garg et al., 2007).
Our findings suggest that an opportunity to exert some degree of choice (or control) over consumption, rather than mere endowment with consumption opportunities, shapes how sad individuals respond to subsequent consumption tasks. Study 1 supports the notion that only when choice is afforded and provides an opportunity for individuals to exert some degree of control, via choice of gift, does it override the effect of sadness on the amount of food consumed. That is, giving sad individuals a choice of what they can take as a gift, rather than simply giving them the gift, determines whether sadness’ effect is attenuated. The follow-up pilot study further demonstrates that it is the opportunity to make a choice rather than the nature of the gift one chooses (hedonic or utilitarian) that impacts sadness’ influence. In doing so, study 1 extends not only the literature in emotion but also that in choice. It validates the psychological impact of choice and suggests that choice is not only a way of increasing a sense of perceived control but that it can also be an important tool in self-regulation by making consumers more responsible for their consumption. Thus, the current research might be of interest to academics studying emotions as well as those interested in the choice literature.
Study 2 builds on this foundation and clarifies that because the influence of sadness on consumption is not always salient, participants do not distinguish between choice for self versus others, and that sadness’ influence is attenuated regardless. Study 2 specifically corroborates prior research (e.g. Cryder et al., 2008; DeSteno et al., 2000; Garg et al., 2007), which suggests that incidental emotion effects operate at a non-conscious level and that most people are not aware of them. We argued that this is one reason why sad individuals are not sensitive to whether the choice is for the self or for others, and that the mere presence of choice attenuates sadness’ effect. Prior research has examined the difference in judgments and decisions when self versus others are involved especially in risk assessments by examining the differences where the locus of the decision is the self or others (Raghubir and Menon, 1998; Taylor and Brown, 1988; Weinstein, 1980). In the health-risk literature, for example, these biases known as self-positivity (underestimating risk to self vs others) and self-negativity (overestimating risk to self vs others) have been found to lead to distinct patterns. In contrast, the current work finds that in situations where the main effect (emotion in our case) is incidental and not salient, the distinction between self and others might be less relevant.
Overall, given the growing movement advocating replication of effects (e.g. Earp and Trafimow, 2015) to test their robustness and to establish theoretically relevant moderators of the same, we believe that our work examining sadness’ effect on hedonic consumption under a variety of contexts makes a contribution by giving credence to the larger body of work on emotion effects on consumer decision-making in general and to sadness’ effects in particular.
5.2. Practical implications
The current studies answer the key questions about sadness and its appraisals and offer a more complete understanding of sadness and its links to compensatory consumption. Furthermore, they elucidate the underlying mechanism responsible for this effect and how to attenuate it. The studies also support and extend prior research suggesting that in the general population (and not restrained eaters), gender effects are not significant and both males and females seem to respond to sadness with compensatory behaviors (see Garg et al., 2007). This is important for stakeholders such as policy makers and consumers who might be interested in developing more generally applicable strategies to control over-consumption.
Given the increasing concern with over-consumption, especially overeating, the current study provides important insights and deeper theoretical knowledge into one aspect driving this behavior. This has implications for stakeholders such as policy makers and consumers to work toward mitigating such harmful, non-salient behaviors. For example, one implication for both consumers and policy makers is that by being aware of the tendency to engage in deleterious consumption when experiencing sadness, individuals can learn to empower themselves via making small (innocuous) choices as in our opening scenario, thereby attenuating sadness’ effects.
Another implication of our findings would be to recommend building in small choices in difficult decisions which cause stress and negative affect in consumers such as retirement planning and buying life-insurance. Such complex decisions are often fraught with negative affect (Luce, 1998) and can thus lead to unintended decision-making by consumers. Thus, building in trivial (or important) choices in such decisions can impart a sense of control to consumers which in turn will enable them to better manage the consequences of negative affect such as sadness which might be generated as a result of engaging in these complex but unavoidable decisions.
5.3. Limitations and future research
While we have established the key mediator and moderators that link to the theoretical structure of sadness, existing research suggests that other variables such as relevant information (Garg et al., 2007) and emotion salience (DeSteno et al., 2000) can also moderate emotion effects. It would be a fruitful direction for future research to examine such moderators as well and to test additional boundary conditions (e.g. self-efficacy). For instance, will choice always lead to attenuation of sadness’ effects or are there conditions under which it will be ineffective? While our research in conjunction with Garg et al. (2018) suggests that it is the attenuation of felt helplessness rather than compensating for the sense of loss that ameliorates sadness’ effect on hedonic consumption, future research could also examine whether making a choice for others (study 2) triggers “gift-giving” connotations and potential positive affect which in turn contributes to the attenuation of sadness’ effect. Furthermore, while we focus on moderators that attenuate sadness’ effect, understanding factors that might amplify sadness’ effects will also be useful.
We find that the attenuating effect of choice emerges regardless of the type of product (hedonic or utilitarian) involved, as in both cases, felt helplessness is attenuated. Exploring other potential mediators of the sadness–consumption relationship and ruling out alternative explanations will be crucial to gain a more nuanced understanding of emotions’ effects on consumer decision-making and behavior. Furthermore, our research studied sadness’ effect on food consumption. It would be interesting to replicate these findings across other domains such as movie consumption (Mishra et al., 2016) or retail overspending. This will also serve to inform us about the robustness of sadness’ effects on consumption more generally.
Finally, our research does not answer the question about whether it is the unique combination of loss and helplessness that dictates sadness’ effect on consumption or whether it is helplessness and low control, in general, that drive this effect. If it is the latter, then other low control emotions such as fear and anxiety should also lead to similar results. Thus, it will be important to examine other positive and negative low control emotions to answer this issue.
