Emotional priming for sustainable consumption? The effects of social media content on the valuation of chocolate

Abstract

Current consumption patterns in the Global North are unsustainable across environmental, social, and economic dimensions. To promote more sustainable consumer behavior, emotional priming on social media has become an increasingly popular tool. In this paper, we mimic social media content to test the effectiveness of emotional priming in incentivizing sustainable consumption. Specifically, we examine whether emotionally primed information about sustainability aspects in cocoa production increases consumers’ valuation for sustainably produced products in the short and medium term. To achieve this, we conducted an online panel survey of 2,161 German consumers with random assignment to one of four different treatments (unprimed information or information with emotional prime) designed as social media posts (videos). We differentiate between environmental sustainability and social responsibility, and separate the products into sustainability certification, uncertified claims, and conventional chocolate. We find statistically significant positive effects of the emotional primes on consumers’ willingness-to-pay (WTP) for products with sustainability certifications and noncertified sustainability claims in the short term. These effects can be attributed to negative emotions triggered by the primes. Two weeks after the treatments, the WTP decreases for all groups. The difference in WTP between the primed and unprimed groups diminishes in the medium term. Primed participants value sustainability certifications and claims less than those who received unprimed information.

1. Introduction

Current food choices in the Global North are not sustainable: Most food products carry social and environmental costs that harm current and future generations worldwide (ILO et al. 2019; Pieper et al. 2020; Pendrill et al. 2022). The consumption of more sustainable alternatives remains low despite increasing efforts of policymakers and NGOs to inform about these externalities through media. Thus, there is a need for a better understanding of how consumption behavior can be changed to promote sustainable consumption and achieve the sustainable development goals (Turunen and Halme 2021).

Various studies analyzed the determinants of sustainable food choices and experimentally tested how to increase consumers’ willingness-to-pay (WTP) for sustainability certifications (White et al. 2019). Specifically, some studies investigated the role of information provision and source (Tobi et al. 2019; von Grafenstein et al. 2022) as well as the priming of social identity values (Hahnel et al. 2014; Andorfer and Liebe 2015; Loebnitz and Aschemann-Witzel 2016; Bullock et al. 2017; Zerbini et al. 2019; Tanford et al. 2020; Brečić et al. 2021; Kim et al. 2021). The former studies indicate that information provision alone has limited effects, while the latter studies demonstrate that activating consumers’ values tends to increase their WTP.

Besides values, individuals’ emotions can also be targeted. Research on charitable giving provides abundant evidence of how emotional appeals increase prosocial spending, regardless of participants’ predisposition toward it (Chang and Lee 2009; Chang 2012; Cavanaugh et al. 2015; Albouy 2017; Goenka and van Osselaer 2019). These studies identify specific emotions, such as guilt and love, which positively stimulate behavioral changes. Small and Verrochi (2009) combined information and an emotional prime (photo) and found that when participants had the opportunity to deliberately process the information, the presence of additional information reduced the primed sympathy. However, if participants were under a high cognitive load and unable to process the information, the effect of the emotional prime remained unchanged. This finding is relevant to our understanding of the role of emotions in consumption behavior because the shopping environment is filled with various signals and abundant information that increases consumers’ cognitive load. Priming emotions may have a lasting effect in such an environment compared to information-only interventions.

When we narrow our focus to the role of emotions in proenvironmental and prosocial consumption behavior, we can establish links that depend on the combination of emotion, consumer, and product. For example, Peloza et al. (2013) found that anticipated guilt, through selfaccountability, can positively influence sustainable consumption choices, whereas explicit guilt appeals do not. Another study that examined guilt, Guerreiro et al. (2015), discovered that triggering emotions in cause-related marketing is effective for hedonic products like chocolate but not for utilitarian products like detergent.¹ They explained that the impact of guilt reduction on choosing a sustainable product is larger in magnitude when the product itself is indulgent (see also Peloza et al. 2013). Regardless of the specific emotion, Matthes et al. (2014) showed that emotional and mixed-type messages increased positive brand attitude. In contrast, informative messages did not affect all participants but only those who had a high level of green involvement.

Today, an increasing number of sustainability awareness campaigns focus on priming emotions through social media content, particularly through short videos comprising a few lines of text and imagery. Examples of such campaigns include the United Nations Development Program's initiative on carbon pricing (UNDP n.d.), the European Union's campaign “make it green” (European Union n.d.), or a program of WDR, a German public broadcaster, called Quarks, which addresses the issue of fishing (quarks.de n.d.). Priming emotions in social media awareness campaigns can be a powerful policy tool because it has the potential to reach a broad audience, including consumers who may not yet be interested in sustainability and who might be left out if only information is provided (see Matthes et al. 2014).

Yet, it is not well understood whether these campaigns indeed trigger emotional responses and ultimately lead to changes in consumption behavior. Another question that remains unanswered in this context is the time gap between exposure to social media campaigns (whether emotional or informative) and the actual purchasing decision. For instance, the study by Antonetti et al. (2018) suggests that priming negative emotions may lead to increased delayed compliance compared to not priming emotions. Participants might first reject the appeal but later accept its message (“delay hypothesis”). On the other hand, the opposite mechanism is possible: Emotions triggered by priming may have a strong, immediate impact but weaken over time as the emotional intensity fades (‘decay effect’). This effect is especially pronounced when emotions are linked to negative memories (Skowronski et al. 2014). Schwartz and Loewenstein (2017) provide evidence that the fading impact of emotional primes on proenvironmental behavior can occur as time passes. Consequently, we empirically examine the extent to which our treatments influence participants’ choices over time.

To investigate how emotional priming in social media campaigns affects consumers’ valuation of sustainable foods in the short- and medium-term, we developed treatments that simulate social media content, specifically Instagram Reels, similar to the posts by the German TV show Quer (2022). Our 1-min videos consist of animated slides presenting information on two aspects: (1) environmental sustainability and (2) social responsibility related to cocoa production. Each includes two different information treatments: (a) factual text with emotional primes and (b) factual text without emotional primes. The emotionally primed videos incorporate real-world images depicting child labor and deforestation to reinforce the conveyed information, while neutral imagery (cocoa beans) replaces these images in the nonprimed counterfactual slides. The treatments are randomly assigned to 2,161 German consumers through online surveys. To examine both short- and medium-term effects, we assess consumers’ WTP using a discrete choice experiment immediately after the treatments and again two weeks later. To assess whether the treatments influenced participants’ emotions through priming, we measured the impact on eight distinct emotions using the Discrete Emotions Questionnaire, developed by Harmon-Jones et al. (2016), in both waves of the study.

Our study contributes to the existing literature in three ways. First, we employ an innovative media format, specifically social media videos, as our treatment. Second, we provide evidence on both short-term and medium-term effects, acknowledging the time gaps between priming and the actual purchasing decision. Previous studies have predominantly focused on short-term effects only. Third, we clearly distinguish between sustainability certificates (Rainforest Alliance and Fairtrade) and uncertified sustainability claims by companies. Limited knowledge exists regarding consumer preferences for products featuring company-specific (uncertified) sustainability claims and third-party audited sustainability certifications. Most existing studies compare consumer preferences for governmental standards, such as the EU organic label, to preferences for private standards by sector organizations or control bodies (such as Soil Association or EcoCert) (Janssen and Hamm 2014) or differentiate standard sponsors among governmental agencies, business groups or NGOs (Iweala and Sun 2022). To the best of our knowledge, no study has examined the effects of treatments on consumer valuation and perceptions of company-specific claims, particularly when compared to certified alternatives.

Our results provide several insights: Firstly, the videos featuring emotional primes indeed induce stronger levels of emotions such as anger, anxiety, disgust, and fear in the short term. However, these effects diminish in the medium term, as all participants exhibit lower levels of emotional response. Secondly, the emotional primes also lead to increased WTP for most of the sustainably sourced alternatives, indicating positive short-term effects. Nevertheless, these effects diminish in the medium term, suggesting that priming is more effective when the time gap between exposure and purchasing decision is minimal. Additionally, the difference in WTP between the primed and unprimed groups decreases over time, with primed participants valuing sustainability certifications and claims less than those who received unprimed information.

2. Study design and data

2.1. Treatments

We tested four information treatments that address issues within the cocoa value chain. Each treatment consisted of a 1-min video comprising a series of animated slides. The treatments differed based on the sustainability objective, providing information on (1) environmental sustainability concerns and (2) social responsibility concerns related to cocoa production. The information in all videos is presented as factual text using neutral language, with data sourced from the Cocoabarometer (Fountain and Huetz-Adams 2020). The videos vary in their background images: The video addressing social responsibility appears in two versions—one featuring photos of child labor (emotionally primed) and another with images of cocoa beans (nonprimed). Similarly, the video on environmental sustainability also has two versions—one with images of deforestation (emotionally primed) and another with cocoa beans (nonprimed). The text between the two versions of each video does not differ. To ensure comparability, the information content, animation, and slide order remained consistent for treatments with the same sustainability objective.² We designed the videos to emulate popular social media formats, such as those found on platforms like Instagram. Specifically, we aimed to replicate the social media strategies employed by large awareness campaigns (e.g. UN agencies) and information outlets (e.g. public broadcasters).³ The treatments were randomly assigned to all respondents, ensuring that each participant viewed a video, either with unprimed information or with emotionally primed information. Rather than including a traditional control group, our study specifically aimed to examine the contrast between information-only and emotionally primed information. The questionnaire was programmed to ensure that participants could not skip the video; it had to play fully on their device.

2.2. Discrete Choice Experiment

To assess consumers’ valuation of sustainability attributes, we used an unlabeled discrete choice experiment that focuses on the trade-offs of the different levels of the attributes (Hensher et al. 2015). Each participant was presented with a randomized set of choices, in which we offer 100-g chocolate bars of unknown brand and flavor, but with varying attributes (Table 1). The sustainability attributes included (1) the Rainforest Alliance certification, (2) an uncertified company claim related to environmental sustainability, (3) the Fairtrade certification, and (4) an uncertified company claim related to social responsibility.⁴ The fifth attribute that varied was the price, with four different levels. The lowest and highest price levels were determined based on a study conducted by the consumer advice center of Lower Saxony (Verbraucherzentrale Niedersachsen 2018): In 2018, the WTP for 100-g of Fairtrade certified chocolate ranged from 0.49 to 3.11 Euro. We adjusted the prices using the consumer price index for the entire country of Germany as of October 2021.⁵ Consequently, the lowest price was set at 0.52 Euros, and the highest price was set at 3.33 Euros. We constructed the second and third highest prices, 1.46 Euros and 2.39 Euros, to ensure that the interval between each price had an equal range of approximately 0.94 Euros.

We derived priors from the pilot data collection for the coefficients of the labels, the claims, and prices. We have excluded the dominant choice of a product with all labels and claims and lowest price. Further, no product was shown that offered chocolate for free or had no labels and claims and the highest price. We focused on the ability of discrete choice experiment to elicit the main effects of attributes to focus on the main objective of the study, the general effect of emotional priming on labels and claims. We employed a d-efficient design based on a modified Fedorov-algorithm (Carlsson and Martinsson 2003). We calculated the design of the attribute levels in each choice set with the STATA command dcreate (Hole 2017). To manage the considerable number of choice sets, we randomly assigned participants to one of the seven blocks. Each block consisted of a fixed group of four choice sets. In order to mitigate any potential order effects, we randomized both the order of the choice sets and the order of the choices within each set. In each choice set, participants could choose between two chocolate bars or the option to not purchase any of the chocolate bars.

2.3. Measurement of emotions

Following the treatments and the choice experiment, we assessed whether the treatments triggered emotional responses in participants, aiming to understand the mechanisms behind these emotional reactions and how they relate to the valuation of certified and uncertified sustainability claims. To do so, we measured participants’ emotions after they watched the videos, using the final section of the questionnaire in both waves. For this purpose, we employed the Discrete Emotions Questionnaire by Harmon-Jones et al. (2016), which measures eight emotions—anger, disgust, fear, anxiety, sadness, desire, relaxation, and happiness—using four items for each. In the first wave we asked, “While you watched the video, to what extent did the following emotions/feelings apply to you?”, and in the second wave, “Imagine cocoa cultivation in West Africa. To what extent do the following emotions/feelings apply to you?”. In the appendix, we provide a flow chart of the study setup including both waves (Fig. A.2).

2.3. Data

We conducted an online survey using a panel provider between May and July 2022.^6,7 Two weeks after participating in the first wave, each respondent received a second invitation to a slightly revised and shortened questionnaire without any treatment. Our final sample consists of 2,161 individuals who completed the questionnaire in both survey waves and met our data quality insurance measures, which screened out participants displaying inattentive answering behavior, such as speeding and straight-lining. In the first data collection round, we collected information of 3,754 participants and in the second round of 3,019 participants. We used a linear probability model to assess selective and differential attrition (Table A.2 in the appendix). The treatments were not statistically significant predictors for attrition in contrast to age, education, and income. For interactions of the treatment of environmental information with income, we could see statistically significant differences in the model using the treatments of social information or environmental primed information as comparison group. However, joined F-test of each treatment and interactions with other predictors did not display joined statistical significance. Thus, we do not correct for attrition in our model. We determined the approximate sample size based on power calculations for the main effects of the utility outcomes, following the rule of thumb by Johnson and Orme (2003). The required sample size deemed sufficient to determine the minimal detectable effect size of 0.10 Euro for two-sample t-tests to assess the differences in WTP.

Quotas were set to approach representation of the German population in terms of age, education, gender, and income.⁸ The socio-economic characteristics of our sample are well-balanced across the four treatment groups (refer to Table A.4 in the appendix), indicating successful randomization. The only exception is that a slightly larger proportion of participants in the primed information group reported an income between 2,000 and 3,599 Euros compared to the information group. However, these differences are not statistically significant in the other four income groups, suggesting relatively minor variations in income.

2.4. Possible biases

To minimize potential biases induced by our study setup, we implemented mitigation strategies for three types of biases.

Expectancy bias. In experimental setups and surveys, respondents’ behavior can be influenced by their expectations of the purpose of the experiment (Zizzo 2010). Although the treatment videos were not explicitly defined as interventions, respondents might have anticipated that the study aims to examine their reaction to the videos. To address this risk, we employed a between-subjects design and compared the effects across the different videos. Thus, we do not quantify the absolute changes in consumers’ valuation of chocolate resulting from our treatments. Since each respondent received only one treatment and was unaware of the other treatments, we anticipated minimizing this bias (Bougherara and Combris 2009).

Hypothetical bias. A widespread bias in consumer research is the discrepancy between stated sustainable intentions and actual behavior (Cerri et al. 2018; Lusk 2018). Compared to incentive-compatible experiments, such as binding choice experiments or experimental auctions, our hypothetical setup is more susceptible to this bias. However, to ensure a sufficiently large sample size across the four treatment groups and because our focus was on the differences between treatments rather than the absolute WTP values, we accepted this trade-off, but addressed this bias by implementing two measures: First, our choice-based experiment was conducted anonymously online because this reduces the risk of respondents overstating their willingness to act compared to face-to-face interviews (Murphy et al. 2005; Krumpal 2013). Second, we included a cheap talk script to draw participants’ attention to the potential bias in their answering behavior due to the hypothetical nature of the study (Lusk 2003; Fifer et al. 2014). This additional step aimed to mitigate hypothetical bias.

Social desirability bias. The respondents received information on a social or environmental issue. As a response, they might want to present themselves in a favorable light by exaggerating their consumption behavior toward products with certifications or uncertified sustainability claims. In real-world scenarios, social desirability plays a role in the disparity between public opinion favoring socially acceptable fair-trade products and the relatively small market share of these products (Carrington et al. 2014; Cerri et al. 2018; Lusk 2018). However, we believe that the impact of social desirability bias in an anonymous online setting is negligible or, at the very least, not significantly different across the four treatments, as all treatments provide information on socially desirable topics.

3. Analysis

We hypothesize that changes in consumption behavior in response to the emotional prime are mediated by shifts in negative emotions. To measure respondents’ emotions, we used the Discrete Emotions Questionnaire developed by Harmon-Jones et al. (2016). We conducted a confirmatory principal component analysis to validate the presence of the eight emotion factors, replicating the analysis conducted by Harmon-Jones et al. (2016). To assess differences across treatment groups, we used Wilcoxon rank sum tests for both the principal component factors and the mean scores of each factor. Thereby, we can determine whether our treatments lead to different emotional elevation among respondents (results presented in Section 4.1).

We used an extended version of the random utility maximization model, following the methodology outlined by Hole (2007a), to analyze the responses from our unlabeled discrete choice experiment. While McFadden (1972) established that discrete choice experiments are grounded in random utility theory, we employed a random parameters multinomial logit (RPL) model, also known as a mixed logit model, to address limitations in the standard conditional logit approach (Train 2009). The RPL model allows us to relax the assumptions of independence of irrelevant alternatives and to account for variations in individual preferences. Our focus is on estimating the respondents’ average WTP for specific attributes of the chocolate bars. We calculate this by taking the ratio of the coefficients for each attribute to the price coefficient. To estimate the 95 per cent confidence intervals, we employ the delta method (Hole 2007b).

In the following, we present our WTP estimation based on the utility model (Train 2009). We estimate the results separately for each time period.

The outcome |${U_{njt}}$| stands for the utility the consumer n gains from choice alternative|$\,\,j$| at choice situation t. |${\beta _1},\,\,\ldots {\beta _K}$| capture the coefficients for attributes, while |${X_1},\,\,\ldots ,\,\,{X_K}$| and |${\beta _C}\,\,$|represent the coefficients for the cost of the alternatives. The random term |${\varepsilon _{njt}}$| is extreme value distributed and its variance varies by consumer n. The fitted models exclude alternative-specific constants because of the unlabeled discrete choice data (Hole 2007a). The coefficients are specified to be normally distributed including price. In our main estimation, we specified a normal distribution for the random WTP coefficient in the main specification, given its flexibility and ability to capture a wide range of individual heterogeneity. The normal distribution's symmetry around the mean allows for positive and negative WTP values, which can realistically reflect variation in preferences but may also imply negative WTP in some cases. To ensure robustness, we also tested a triangular distribution, which is bounded by defined minimum and maximum values, constraining the range of variation and reducing the likelihood of extreme or implausible values. This alternative distribution yielded results (Table A.11) very similar to those obtained with the normal distribution, supporting the robustness of our findings.

After taking the total derivative of |${U_{njt}}$| with respect to the changes in attribute |${X_k}$| and cost C, we can express the WTP as the change in cost conditional on a change in |${X_k}$| with constant utility:

4. Results

4.1 (Primed) information and emotions

First, we discuss the association between our treatments and the measured emotions of participants. The success of our emotional prime can be determined by comparing the emotional responses of the primed information groups to those of the information groups. Figure 1 illustrates the emotional responses for the treatments related to environmental sustainability, while Fig. 2 depicts the emotional responses for treatments related to social responsibility. In both figures, the group average scores on the respective emotions are presented along with 95 per cent confidence intervals (for further details, refer to Tables A.5–A.7).

In the short term, both primed information groups exhibit stronger emotional reactions compared to their respective information-only groups. The differences in both comparisons are statistically significant (P < 0.05) for anger, anxiety, disgust, fear, and happiness. In the case of the environmental treatments, the primed information group also shows statistically significant differences in the levels of sadness and relaxation. For the social treatments, there is no statistical difference in sadness and relaxation, but there is for desire. Overall, the mean differences tend to be larger for the environmental treatments compared to the social treatments.

In the medium term, statistically significant differences between the primed information group and the information group persist for anxiety, sadness, disgust, and fear in the environmental treatments. However, we do not observe similar statistically significant differences for the social treatments. Overall, and not surprisingly, the mean differences in emotional responses between the primed information group and the information group decrease over the two-week period between the first and second wave of data collection.

In this context, we observe that the emotional responses in the primed information group for the environmental treatment are statistically significantly lower in the medium term compared to the short term for anger, anxiety, sadness, disgust, and fear. Conversely, the emotional responses for relaxation and happiness are higher in the medium term. We observe similar patterns for the primed information group regarding social responsibility, with the exception of disgust and fear. Additionally, we find that the emotional response of the information-only groups decreases between the short term and medium term for anger, sadness, and disgust.

Overall, our emotional prime was successful because the primed information groups show stronger emotional reactions than the information-only groups, especially in the short term. This finding is more pronounced in relation to environmental sustainability. In the medium term, we observe a decline in the mean difference of emotions across all groups, indicating a less-pronounced emotional response.

4.2. (Primed) information and WTP

Next, we analyze the effect of the emotional prime on consumers’ WTP. We present the results for each wave separately because we are interested in the changes of the average effects rather than in an average effect across waves. The findings for the environmental sustainability treatments are illustrated in Figs 3 and 4.⁹ The left side of the figures displays the short-term effects immediately following the treatment, while the right side represents the medium-term effects two weeks after the treatment. We tested the reported relationships with t-tests and only report comparatively if there are statistically significant differences (see Tables A.8 and A.9 for results). We differentiate WTP based on sustainability certification (Rainforest Alliance) and uncertified environmental claims. Across all groups, participants exhibit a preference for purchasing chocolate with certification and claims rather than no chocolate at all.¹⁰

We find that in the short term, emotionally primed respondents exhibit a higher WTP than respondents with the information-only treatment for environmental certification. Similarly, the WTP of environmental claims appear also higher. A simple comparison between average WTP values indicates a difference of 43 Euro-Cents for the certification and 10 Euro-Cents for the claim. Within the primed information group, the WTP for certified chocolate and chocolate with an uncertified claim was relatively similar. Interestingly, the WTP for the claim appears higher in the information-only group. This indicates that the primed information does not only influence the WTP for more environmentally sustainable chocolate but also affects the preferred type of sustainability attribute (claim versus certification). In the medium term, the WTP differences between the primed group and the information-only group declines for certified chocolate. The WTP for the environmental claim is lower in the primed group compared to the information-only group.

In Fig. 4, we compare the changes between the short term and medium term within each treatment group. The left-hand side illustrates the differences for the respondents who received information without any emotional prime. We observe an increase in WTP for the certification and a decrease for the claim in the information-only treatment. The first observation suggests positive medium-term effects of information treatments on the valuation of the environmental certification. Within the primed information group, the WTP for both products appears to decrease, with a significant drop for the claim. These findings align with the changes observed in consumers’ emotional reactions in the short term and medium term, as discussed in Section 4.1.

In the context of social responsibility, the effects of the primed and unprimed information treatments are depicted in Figs 5 and 6. Notably, the effects are not as straightforward as those observed for environmental sustainability. In the short term, we observe that the WTP for the social certification (Fairtrade) is slightly lower in the primed information group compared to the information-only group. However, the primed information group has higher WTP for the social claim compared to the certification. Notably, the WTP for chocolate with the social claim appears generally higher than for the certified product. In the medium term, we observe partly reversed effects. The WTP for chocolate with the social claim remains higher in both groups than for the certified chocolate, but in the primed information group, the WTP appears higher for the certification and lower for the social claim compared to the information-only group.

Figure 6 illustrates the differences over time within each respondent group and product. Again, we observe a decrease in product valuation across all groups and products within the two-week period, which aligns with the changes in emotional responses.

We focus not only on the main effects in our RPL model but also estimate interactions between the social certificate and claim and between the environmental certificate and claim (results in Table A.10 in the Appendix). The interactions show negative values, suggesting a diminishing returns effect: The combined impact of a certificate and a corresponding claim is lower than the sum of their individual effects, likely because consumers see the combination as redundant. Despite these negative interactions, the main effects of social and environmental attributes become more strongly positive, indicating that certificates alone are highly valued as standalone indicators. Additionally, the price coefficient increases in magnitude, suggesting greater price sensitivity when both certificates and claims are present, as consumers reassess the added value relative to cost.

5. Discussion

Overall, our study reveals three main findings. First, the emotional prime largely enhances the valuation of sustainable product choices in the short term. Second, the effects of the prime diminish in the medium term. In some cases, the differences between the primed information group and the information-only group are implicitly negative, because unprimed respondents value certain products higher. These findings indicate that information accompanied by an emotional prime is particularly effective in the short term, but not necessarily more effective than information-only in the medium term. These findings align with the observation that emotional intensity fades over time, particularly when emotions are triggered by negative memories (Skowronski et al. 2014). This pattern is reflected in our results, as the overall intensity of emotions decreased in the second wave. However, for environmental sustainability, we still observe slightly higher levels of anxiety, sadness, disgust, and fear in the primed group. Since participants were specifically asked about their emotions concerning the cocoa sector in West Africa, we believe this helped trigger a recall of images related to deforestation, which may explain the sustained emotional difference in this group after a two-week delay.

Additionally, the finding that primed information and information-only is similar effective in the medium-run is consistent with Schwartz and Loewenstein (2017), who found that priming sadness led to increased donations in the short term, but after an hour's delay, the effect was no more significant than providing information alone. This led them to conclude that emotional priming may be more effective for one-time behavior changes rather than sustained behavioral shifts. While grocery shopping typically falls into the latter category, it is important to consider that, in reality, consumers are likely to encounter repeated exposure to such campaigns on social media. While our study does not examine the effects of repeated exposure, our respondents, who are frequent and active online users, accurately represent the potential target audience of social media campaigns.

Thirdly, the WTP decreases over the two-week period—irrespective of whether participants received information-only or the emotional prime. Only the environmental certification shows a higher valuation in the information group in the medium term compared to the short-term. This effect could be attributed to the initially lower awareness and knowledge level of the selected environmental certification, Rainforest Alliance, among consumers compared to other certifications like Fairtrade or organic certifications (Grunert et al. 2014). The information treatment may have heightened the problem awareness among our participants, which they then associated with the Rainforest Alliance certification. This connection between problem awareness and certification could have resulted in a more enduring positive effect on their valuation.

Nevertheless, when discussing the three main findings of our treatments on the emotions and valuation of participants, it is important to acknowledge two limitations in our study design: The factual text provided to the information-only group also elicited negative emotions caused by sensitive issues such as child labor, potentially diminishing the contrast between the emotional prime and information-only treatments. Including a control group would have provided a clearer baseline, so the observed effects of the emotional prime should be interpreted as a lower bound on its potential impact.

Additionally, we did not directly verify whether participants fully watched or paid attention to the treatment videos. While this limitation applies to both treatment groups, it reflects real-world conditions, where individuals may engage with social media campaigns with varying levels of attention. Including an attention check could have allowed us to better control for participants’ engagement with the content. Together, these limitations suggest that while our results provide valuable insights, they may slightly underestimate the potential impact of a fully engaged emotional prime.

Overall, there is an exception in the pattern of higher WTP in the short term and a larger decrease from short term to medium term when comparing the primed information group to the information-only group. This exception pertains to the chocolate with the social certification, Fairtrade. The primed information group valued it less than the information-only group in the short term but more in the medium term. This rather puzzling finding could be explained by the phenomenon of psychological reactance. The emotional prime induced stronger negative emotions, particularly anger, anxiety, and sadness. These negative emotions may have caused participants to feel a sense of blame for supporting child labor in the cocoa sector through their chocolate consumption. When faced with the choice of chocolates in the discrete choice experiment, participants may have intended to push back against this guilt. Psychological reactance occurs when individuals resist attempts to influence their behavior because they perceive such attempts as threats to their autonomy (Brehm and Brehm 2013). The effect in the medium term can be attributed to delayed compliance (Antonetti et al. 2018). The initially triggered emotions such as anxiety and fear may have resulted in psychological reactance in the short term. In the medium term, these emotions could have transformed into behavior that aligns with the message conveyed. The primed emotions might have facilitated a deeper understanding and acceptance of the message over time, leading to a more compliant behavior in the medium term (ibid.).

We argue that the pattern of psychological reactance is evident only with Fairtrade, and not with the social claim (where the primed information group values it higher), due to the difference in awareness between the two. Fairtrade is a well-established certification in the market, and we know that consumers already associate it with the specific social cause it represents (von Grafenstein et al. 2022). Prior to the study, participants may have already associated the Fairtrade certification with the objective of reducing child labor. Consequently, the emotionally charged treatment may have triggered psychological reactance in these participants. As the uncertified social claim does not have a pre-established association, the emotionally loaded treatment may not have evoked the same reactance. Nevertheless, since we did not measure awareness directly, this explanation remains theoretical.

6. Policy implications

Consumption choices need to become more sustainable. Messages and campaigns promoting sustainable production and consumption, while appealing to consumers’ emotions, have the potential to effectively reach the average consumer and counterbalance factors like shopping-related mental load, which may generally lead to less involvement in sustainability topics (Small and Verrochi 2009; Matthes et al. 2014). Social media platforms provide various options in terms of format, frequency, and timing to deliver these emotional messages. Therefore, gaining insights into the most effective ways for policymakers, civil society, and the private sector to reach consumers with their messages on these platforms is becoming increasingly important.

Our findings suggest that social-media-like sustainability campaigns that elicit emotional responses, particularly in the context of environmental sustainability and uncertified environmental and social claims, have positive short-term effects on consumers’ WTP for sustainability attributes. These results suggest that the timing of exposure to these messages is crucial, as they should ideally coincide with the imminent consumption choice—or as Schwartz and Loewenstein (2017) put it, ‘strike while the iron is hot’. To achieve this proximity, messages could be strategically placed at the point of sale, such as through collaborations with retailers in public campaigns. Additionally, the constant exposure to social media content through smartphones offers opportunities to time posts according to peak shopping hours. However, the medium-term effects indicate the need for different strategies adjusted to different time frames. Information-only campaigns may ultimately be more or—at least—not less effective in the medium term.

Besides timing, our findings suggest that posts or campaigns do not need to focus on a specific certification. Our video highlighted social and environmental issues associated with cocoa production without specifically mentioning any certifications that address these problems. Therefore, various actors from the public and private sector can jointly create campaigns to raise awareness about the negative externalities of consumption among the population. Awareness raising can increase the valuation of sustainably produced foods without emphasizing specific claims or certifications.

At the same time, our findings highlight that one-size-fits-all approach may be inadequate. Posts or campaigns intended to trigger certain emotions in the audience can lead to unintended consequences, as demonstrated by our experiment with Fairtrade chocolate. As the exact mechanism is unknown, it is crucial to test for possible psychological reactance and/or delayed effects when designing policies, including public awareness campaigns that aim to evoke emotional responses. This will help ensure that the intended impact is achieved while minimizing any potential unintended consequences.

5. Conclusion

A large body of literature has investigated the role of emotions in sustainable behavior, particularly in the context of charitable giving. However, the evidence on the effects of emotional primes in social-media awareness campaigns remains limited although the popularity in the context of sustainable consumption increases. This paper aims to address this research gap by examining the effects of emotional primes, conveyed through imagery in short videos, on consumers’ emotional responses and WTP for sustainable food products.

Our findings show that emotional primes effectively trigger emotional reactions that translate into differences in WTP. This highlights their potential as a useful policy tool for altering consumption behavior. Yet, we find that the timing of the treatments matters, and that a one-size-fits-all approach may not be suitable, as different sustainability characteristics can yield varying effects.

Building on these findings, we propose two main areas for future research. Firstly, regarding timing, our study examined two specific time points: immediately after the treatment and two weeks later. This time gap allowed us to gain initial insights into the relevance of timing, as our results showed variations between the two waves. Our findings indicate that immediate exposure to emotional primes elicits stronger emotional reactions and leads to higher WTP. An important policy implication is to strategically target these primes during peak shopping hours. Future studies should further investigate the validity of this finding and explore the optimal frequency at which such campaigns should be targeted toward consumers.

Regarding the sustainability characteristics, our findings demonstrate that the same emotional primes lead to increased WTP for both claims and certifications. However, we also observe unintended effects, as in the case of the Fairtrade label. Future research should explore the interrelations between existing sustainability attributes and consumers’ valuation, considering factors such as source (company, government, NGO), verification mechanisms, and the existing level of awareness. A related open question that warrants further investigation is the extent to which emotional primes in one sustainability dimension may influence sustainable choices in other dimensions of sustainability. Future studies could incorporate additional attributes in choice experiments to explore whether such adverse effects emerge.

Acknowledgments

We thank the German Federal Ministry for Economic Cooperation and Development for funding this study. We also want to thank Professor Jann Lay and the Chair of Marketing for Food and Agricultural Products led by Professor Achim Spiller at Georg August University of Göttingen, especially Dominic Lemken, for their continuous feedback for our study. We also thank one anonymous reviewer of our registered report draft. This study received ethical approval by the Ethics committee of the University of Göttingen and was registered in the AEA RCT Registry with the unique identifying number: ‘AEARCTR-0008689’.

Conflict of interest

None declared.

Funding

This work was funded by the German Federal Ministry for Economic Cooperation and Development through the initiative ‘Research Network Sustainable Global Supply Chains’ [project number: 2020.9591.7]. We acknowledge support by the Open Access Publication Funds of the Göttingen University.

Data availability

The raw data, processed data, and the code used for the estimation will be provided by the corresponding author upon request.

Footnotes

Appendix A: Tables and Figures

References