When does an ethical attribute matter for product evaluation? The role of warm‐glow feelings for low‐rated products

Funding information Swiss Commission for Technology and Innovation (CTI), Grant/Award Number: KTI 1155000154 Abstract Research on the effects of ethical attributes has recently gained traction. However, limited research has addressed consumer response to ethical attributes in the current context where product ratings have become of primary importance to make decisions. Specifically, this study examines the relative effect of ethical attributes on product evaluations across different product ratings. Building on cue consistency theory and the negativity bias, we suggest that ethical attributes gain weight when consumers evaluate a low‐rated product. This process leads consumers to anticipate more warm‐glow feelings, generating better evaluations for such low‐rated products featuring an ethical attribute (vs. another type of attribute). Two experiments provide consistent empirical support for this prediction, and demonstrate that, compared to other attributes or no attribute, an ethical attribute increases product evaluations to a larger extent when the product received low (vs. higher) ratings. We show that this effect occurs because of warm‐glow feelings: when product ratings are low, consumers anticipate more warm‐glow feelings from purchasing a product with an ethical attribute (vs. another type of attribute), leading to better product evaluations. These findings have direct managerial and ethical implications for practitioners.

)-in consumer decisionmaking. Such attributes have been increasingly used in the marketplace (e.g., Das, Agarwal, Malhotra, & Varshneya, 2019;Olsen, Slotegraaf, & Chandukala, 2014) but-in spite of recent attention in consumer research (e.g., Bodur, Tofighi, & Grohmann, 2016;Herédia-Colaço & do Vale, 2018;Schamp, Heitmann, & Katzenstein, 2019)-evidence about how consumers react to ethical attributes for products of different ratings is still lacking. Specifically, the recent literature has identified under which conditions-such as which product categories, types of brand, marketing tactics, or types of ethical claim (e.g., Andrews, Luo, Fang, & Aspara, 2014;Bodur et al., 2016;Herédia-Colaço & do Vale, 2018;Luchs et al., 2010)the effects of ethical attributes vary. This line of research provides interesting insights into the factors that moderate the effect of ethical attributes. Nevertheless, by often comparing the effect of products with and without ethical attributes in different contexts (e.g., Lin & Chang, 2012), it leaves aside the incremental effects of ethical attributes over other types of attributes.
However, recent research has started to investigate the effects of ethical attributes in relation to other types of attributes. For instance, consumers make trade-offs between ethical and utilitarian-those that purely signal performance, functionality, or effectiveness (Dhar & Wertenbroch, 2000;Okada, 2005)-attributes (Bridges, Schramm, & Roy, 2019), often resulting in a decrease of the perceived benefits of ethical attributes (Barone, Miyazaki, & Taylor, 2000). Those trade-offs suggest that utilitarian benefits are preferred to ethicality, even when the goal of the consumer requires a product with both ethical and utilitarian benefits (Berens, Riel, & Rekom, 2007;Luchs, Brower, & Chitturi, 2012). While these findings suggest that ethical attributes may not be much weighted in consumer decision-making process, two limitations of prior research should be highlighted here. First, trade-offs occur only when the consumer is aware of the performance of each attribute and can compare them. However, such situations are only occasional, as in many instances consumers face incomplete information. Even in cases where ratings for some product attributes are provided (such as on Consumer Reports or Tripadvisor), a rating of the ethicality of the product is rarely present. Hence, there is a need for studies that examine ethical benefits using limited information about the ethical benefit. Second, recent research identified that consumers may actually consider ethical benefits, especially when the self is threatened (Trudel, Klein, Sen, & Dawar, 2019). Therefore, the door seems open for a more in-depth understanding of the conditions in which ethical attributes outweigh other attributes in increasing consumer evaluations of the product.
To fill this gap, this research builds on cue consistency theory (Maheswaran & Chaiken, 1991) and the negativity bias (Skowronski & Carlston, 1989) to predict and test the effect of ethical attributes across different product ratings. Two reasons motivate the investigation of the intersection of ethical attributes and product ratings. First, quality cues seem to condition the effect of ethicality cues. For instance, in a corporate context, corporate ability interact with corporate social responsibility (CSR) to affect consumer perceptions and behaviors (Berens et al., 2007;Luo & Bhattacharya, 2006). In a service provider context, perceived ethicality and competence interact to affect consumers' evaluation of the provider (Kirmani, Hamilton, Thompson, & Lantzy, 2017).
However, limited research has focused on the interaction of quality and ethicality cues at the product level. Second, relying on ratings as quality cues appear appealing due to their widespread use and impact. Indeed, ratings have become one of the most important source of information for consumers to judge product quality (De Langhe, Fernbach, & Lichtenstein, 2016;Moe & Trusov, 2011;Sridhar & Srinivasan, 2012). Therefore, this research aims to examine the relative effects of ethical attributes for products receiving different ratings.
With these goals in mind, we propose and test that for a product with low ratings, an ethical attribute exerts a more positive effect on product evaluations than other attributes (Studies 1 and 2) and than no attribute (Study 2). When product ratings increase, this incremental effect of the ethical attribute weakens or disappears. We show that an ethical attribute will prompt more warm-glow feelings when the product is poorly rated, eventually leading to better product evaluations. We test and provide evidence for these propositions in two experiments. First, we present the theoretical background for our hypotheses, rooted in the literature on cue consistency, negativity bias, warm-glow feelings, ethical attribute effects, as well as product ratings. Then, we present two experiments using different product categories and providing consistent empirical evidence, which leads to the discussion of theoretical and managerial implications.
2 | THEORETICAL BACKGROUND 2.1 | Third-party product ratings as functional performance cues Consumers increasingly rely on product ratings to make their purchase decisions, in particular when shopping online (Sridhar & Srinivasan, 2012) and for product categories whose online reviews and expert opinions are easily available (Simonson & Rosen, 2014). Those ratings can emanate from other consumers (also referred to as peer-to-peer ratings or user ratings; Filieri, 2015) or from third-party organizations, such as Consumer Reports, Runners' World, PC Magazine, or Wine Advocate.
While both third-party and consumer ratings are used by consumers to evaluate the functional performance of a product, or "its ability to fulfill its purpose" (Homburg, Schwemmle, & Kuehnl, 2015, p. 44), user ratings are not reliable indicators of a product's objective quality because users (a) rely in part on esthetic and symbolic dimensions to rate a product, and (b) are subjectively biased by others' ratings (De Langhe et al., 2016).
By contrast, third-party ratings are often conducted by technical experts such as engineers or scientists, and represent a more accurate indicator of a product's objective quality (Aiken & Boush, 2006). In particular, expert reviews are judged as a more credible source of information than consumer testimonials for goods (De Maeyer & Estelami, 2011). Accordingly, consumers associate products with high (vs. low) third-party ratings with a high (vs. low) overall quality (Akdeniz, Calantone, & Voorhees, 2013).
While a third-party product rating influences consumers' overall appreciation of a product, it can also influence specific product attributes' evaluation. Golder, Mitra, and Moorman (2012) explain that 1572 | consumers evaluate the quality of an offering by comparing its "perceived attributes with their expectations to form summary judgments of quality […]" (p. 5), this process occurring at different levels of attribute aggregation. Consumers perceive attributes and form expectations based on stored knowledge accruing from different sources, including other consumers' experiences, media reports, or quality signals associated with each attribute (Golder et al., 2012).
Hence, as soon as consumers access product ratings, this information becomes part of their stored knowledge, which will affect the way they perceive product-specific attributes and form product expectations. For instance, high third-party ratings can reinforce the effect of a high price or a strong warranty on perceived product quality (Akdeniz et al., 2013).
Previous research highlights a decreased influence of ethical attributes, when consumers make trade-offs between ethical and utilitarian benefits (Barone et al., 2000;Luchs & Kumar, 2017).
Specifically, consumers prefer utilitarian over ethical value when they have access to third-party evaluation of each attribute (Luchs & Kumar, 2017;Luchs et al., 2012). These effects were tested in situations where participants had access to third-party evaluations of each attribute leaving limited margin to interpret how each attribute performs. In many situations, however, consumers do not have a comparable score associated with ethicality, and consumers make choices using their interpretation of ethical cues coming from third parties (e.g., labels) or marketing. Therefore, in this research, we investigate the role of third-party's product (and not attribute) ratings on consumers' evaluations of products with an ethical attribute (vs. a control consisting of either a utilitarian attribute or no attribute).
More specifically, we propose that (a) ethical attributes increase product evaluations more than control for low-rated products; and (b) this incremental effect of ethical attributes disappears for higher rated products. We develop below the logic underlying these propositions, which find their roots in the symbolic benefits associated with ethical attributes.
Contrary to utilitarian attributes-which contribute to the functionality, effectiveness, efficiency, or practicality of the product (e.g., Dhar & Wertenbroch, 2000;Okada, 2005;Strahilevitz & Myers, 1998;Voss, Spangenberg, & Grohmann, 2003)-ethical attributes have more symbolic benefits and are less related to functional performance. Consumers intuitively associate ethicality with less strength (Luchs et al., 2010), which can trigger doubts about the quality of a product that features an ethical attribute (Mai, Hoffmann, Lasarov, & Buhs, 2017). An exemplary illustration of this phenomenon is that consumers use more of a product (e.g., hand sanitizer or laundry powder) when it is labeled as ethical or sustainable than when it is not labeled as such, because they associate ethical products with less effectiveness (Lin & Chang, 2012). So, ethical attributes are associated with less functional performance, but why would ethical attributes affect consumer evaluations differently according to product ratings?
Consumers allocate different weights to ethical attributes depending on the stage of the decision-making journey (Schamp et al., 2019). In the context investigated here, we suggest that the weight of the ethical attribute may also vary depending on the ratings of the product. This contention builds on cue consistency theory (Maheswaran & Chaiken, 1991), which suggests that when people face multiple sources of information, congruent (vs. incongruent) information is more useful and likely to be mobilized jointly in evaluations that use information integration models (Miyazaki, Grewal, & Goodstein, 2005). As an extension to this theory-and already used as a theoretical basis in research on third-party ratings (Akdeniz et al., 2013)-negativity bias suggests that when facing incongruent information, consumers weigh the less positive information more heavily than the more positive information and the less positive cue becomes dominant in consumers' evaluation (e.g., Miyazaki et al., 2005;Skowronski & Carlston, 1989).
Our contention also builds on the notion that product quality is by definition related to product attribute performance, quality being defined as "a set of three distinct states of an offering's attributes' relative performance generated while producing, experiencing, and evaluating the offering" (Golder et al., 2012, p. 2). As a result, because consumers use ratings to assess the objective quality of the product, low ratings signal low performance. Ethical attributes (e.g., a child labor free statement) are not directly connected to the functionality of the product. They may however still negatively affect consumers' perception of effectiveness (Lin & Chang, 2012) or strength (Luchs et al., 2010;Mai et al., 2017). Thus, contrary to other types of attribute such as utilitarian attributes, ethical attributes are not incongruent with low product ratings. As a result, consistent with negativity bias and cue consistency theories, ethical attributes, unlike other types of attributes, may represent useful information when consumers evaluate low-rated products. They are likely to be mobilized jointly with product ratings in the evaluation process. Therefore, compared to controls including no attribute or a utilitarian attribute, an ethical attribute should increase product evaluation when product ratings are low: | 1573 H1 For low-rated (vs. higher rated) products, an ethical attribute (vs. control) will (vs. will not) increase product evaluations.

| Warm-glow feelings as a response to ethical attributes for low-rated products
We propose that when anticipating the purchase of products featuring ethical attributes, consumers also anticipate warm-glow feelings, that may play a mediating role in the effect of those attributes on consumer evaluations. Warm-glow feelings are "feelings of pleasure and satisfaction derived from the cognitive appraisal of contributing to the wellbeing of society" (Hartmann, Eisend, Apaolaza, & D'Souza, 2017, p. 45). They are often referred to as "impure altruism" (Andreoni, 1989(Andreoni, , 1990) because they are self-focused, as opposed to pure altruism, which is only driven by others' needs. Warm-glow feelings are driven by the pursuit of the positive affective state that one feels following actions taken to help others (Andreoni, 1990). In other words, when one helps others, it is not only because they need help but also because the helper anticipates and experiences the warm glow of good feeling or moral satisfaction associated with giving aid. Nevertheless, warm-glow feelings do not only emerge when helping others directly. For instance, warm-glow feelings arise when acting environmentally friendly (Giebelhausen, Chun, Cronin Jr, & Hult, 2016;Taufik, Bolderdijk, & Steg, 2015), indicating that they affect consumer behavior in multiple domains. In the consumption domain, consumers may experience warm-glow feelings when they purchase a product with an ethical attribute because they know that the money spent on that product supports a good cause (Andrews et al., 2014;Chang & Chu, 2019;Habel, Schons, Alavi, & Wieseke, 2016).
In general, consumers should then anticipate more warm-glow feelings when the ethical attribute is given more weight in the evaluation process, as the feeling of doing something good for society should increase. As a result, building on our previous discussion on weight allocation during the evaluation process, we suggest that because the relative weight allocated to the ethical attribute (vs. controls such as a utilitarian attribute) should be greater when product ratings are low, consumers considering the purchase of a lowrated product should anticipate more warm-glow feelings when the product has an ethical attribute. Therefore, we expect: H2 For low-rated products, an ethical attribute (vs. control) will generate more (vs. less) anticipated warm-glow feelings. The incremental effect of an ethical attribute will decrease when product ratings increase.
An ethical attribute gives consumers a reason to feel good about the purchase of a product with low ratings. We suggest that the anticipation of these good feelings influence product evaluation. Indeed, previous research highlights that warm-glow feelings affect consumption-related attitudes and behaviors (Habel et al., 2016). More specifically, warmglow feelings mediate the effect of marketing initiatives such as cause marketing programs on service satisfaction (Giebelhausen, Chun, Cronin, & Hult, 2016) or purchase (Andrews et al., 2014). We previously argued that a low-rated product with an ethical attribute should increase the evaluation of the product. We suggest that this effect occurs through the mediating effect of warm-glow feelings. On the contrary, when consumers evaluate a product that received better ratings, they should focus more on attributes that relate to functional performance and less on ethical attributes. In this case, the relative effect of the ethical attribute on warm-glow feelings should decrease, removing the incremental effect of the ethical attribute on evaluation. Hence, we expect: H3 For low-rated products, anticipated warm-glow feelings will mediate the effect of an ethical (vs. control) attribute on product evaluations. This mediating effect will decrease when product ratings increase. These hypotheses are tested across two studies. Study 1 tests H1-H3 using experts to manipulate the ratings of a t-shirt and an ethical claim (Guaranteed child labor free) to manipulate the ethical attribute. Study 2 tests H1-H3 using a different product (sunscreen), and different manipulations of the ratings (Consumer Reports) and ethical attribute (EcoConscious claim). It also extends Study 1 by providing an additional control for green consumption values and an additional control condition to increase the validity of our results. In both studies, we anchor our manipulation of the ratings as low (as this is the focus of this article) and compare it to higher ratings, which do not correspond to top ratings (as our theory focuses on the difference between low and relatively higher ratings). Finally, we use different samples (UK and US) to extend the external validity of our results.

| Pretest
Given the aforementioned evidence showing that consumers make trade-offs between ethical and utilitarian attributes, Study 1 tests the hypothesized effect of ethical attributes by opposing them to utilitarian attributes. Before designing the experiment, a pretest was conducted to select suitable ethical and utilitarian attributes, as well as rating criteria. One hundred participants residing in the United Kingdom (58.0% female; M Age = 33.13; SD = 12.89) were recruited on F I G U R E 1 Theoretical model 1574 | the platform Prolific.co in exchange for monetary compensation. The objective was to ensure that the selected attributes were equally realistic but differed in terms of perceived ethicality and utilitarianism. We also wanted to ensure that product rating manipulation included realistic criteria. First, we identified a series of ethical and utilitarian attributes based on a review of clothing e-commerce websites. We pretested three ethical attributes ("100% recyclable," "Organic cotton," and "Guaranteed child labour free") as well as three utilitarian attributes ("Extra wash resistance," "Extra fade-resistant fabric," and "Guaranteed non-iron t-shirt").
Participants read that an attribute was utilitarian if it offered functional, instrumental, and practical benefits (Chitturi, Raghunathan, & Mahajan, 2008) and were asked to rate each attribute on a 7-point scale from "Not at all utilitarian" to "Very utilitarian." They also read that an attribute was ethical if it offered benefits to the society or the environment (Luchs et al., 2010), and were asked to rate each attribute on a 7-point scale from "Not at all ethical" to "Very ethical." They also rated how realistic it was to find each attribute on a tshirt's label or packaging (7-point scale from "Not at all realistic" to "Very realistic"). The order of attribute presentation was randomized.
The pair of attribute "Extra fade-resistant fabric" and "Guaranteed child labour free" emerged as ideal candidates. "Extra faderesistant fabric" was perceived as more utilitarian than "Guaranteed In the experiment, we planned to manipulate product ratings by showing how experts rated the t-shirt based on explicit criteria. We drew an initial list of possible criteria based on how consumers perceive clothing (Swinker & Hines, 2006). Participants rated how realistic it would be if experts rated the quality of a t-shirt based on these criteria (7-point scale from "Not at all realistic" to "Very realistic"). We selected the three criteria that received the higher value for the experiment: "Comfort" (M = 5.98), "Fabric holds shape" (M = 5.67), and "Design features" (M = 5.51). We also ensured that the utilitarian attribute and the criteria on which the t-shirt was rated were independent (i.e., the fact that the t-shirt does not fade does not impact experts' rating of its comfort, design features, and how the fabric holds shape).

| Design and procedure
One hundred and seventy-five participants residing in the United Kingdom (77.1% female, M Age = 36.39, SD = 12.01) were recruited on Prolific.co in exchange for monetary compensation. Participants were randomly assigned to one of four conditions in a 2 (product ratings: low vs. higher) × 2 (attribute type: ethical vs. utilitarian) between-subjects design and filled a questionnaire online. Participants read that "Trevy is a new brand that is specialized in casual clothing for women, men, and kids." The brand was fictitious to avoid prior attitudes toward and knowledge about the product. They also read the following: "The picture and the product description come from the brand. The expert ratings come from an independent magazine that evaluates clothing." Below, they saw a picture of a Trevy t-shirt with a description of its characteristics. Following the pretest, the attribute type conditions included either the ethical attribute ("Guaranteed child labour free") or the utilitarian attribute ("Extra fade-resistant fabric"). Product ratings were manipulated with a star system (two stars for the low product condition and four stars for the higher ratings). The criteria on which the magazine rated the t-shirt were explicitly stated as comfort, design features, and how the fabric holds shape, as supported by the pretest.
The stimuli are presented in Appendix A.
Our dependent variable, product evaluations, was measured using a 4-item, 7-point, semantic differential scale ("Please choose the option that best reflects your opinion toward this t-shirt"-Dislike/Like, Bad/Good, Unfavorable/Favorable, Negative/Positive; α = .95).
Participants expressed their anticipated warm-glow feelings by responding to the question "How would you feel if you bought this tshirt?" using a 4-item (ashamed/proud, in the wrong/in the right, wicked/ virtuous, unethical/ethical; α = .91; Giebelhausen et al., 2016), 7-point, semantic differential scale. As a manipulation check, we also measured product perceived quality on a scale from 0 (extremely poor quality) to 100 (excellent quality). (1, 171) = 9.04; p = .003). These results support the notion that the difference in anticipated warm-glow feelings generated by a product with an ethical (vs. utilitarian) attribute is greater when the product ratings are low, which supports H2.

| Mediation through warm-glow feelings
A moderated mediation analysis (Hayes, 2017;model 8;5,000 bootstrap samples) was performed to test our hypothesis that the indirect effect of attribute type (coded as ethical attribute = 1 and utilitarian attribute = 0) on evaluations through anticipated warm-glow feelings is moderated by product ratings (H3). Consistent with our prediction, the indirect effect of attribute type was significant in the low ratings condition (ab = 0.98; 95% CI = 0.060 to 1.41). The indirect effect of attribute type was also significant in the higher ratings condition (ab = 0.43; 95% CI = 0.14 to 0.75). Further, the index of the moderated mediation (index = −0.55; 95% CI = −1.00 to −0.15) indicated that the indirect effect of the ethical (vs. utilitarian) attribute on product evaluations was significantly greater in the low rating condition than in the higher rating condition, which supports H3.

| Discussion
Results of Study 1 support H1 by showing that an ethical attribute compared to a utilitarian attribute, increases consumers' evaluations of a low-rated product, whereas it does not increase consumers' evaluations in case of higher rated products. Also, consumers anticipate more warm-glow feelings with a product that includes an ethical (vs. utilitarian) attribute for low-rated products. This difference is smaller for higher rated products (H2). Finally, results indicate that anticipated warm-glow feelings mediate the effect of the ethical attribute in the low rating condition and that the mediation effect weakens when rating increase (H3). Study 2 replicates and extends these findings in another context.

| Design and procedure
To manipulate attribute type, Study 2 used an ethical attribute, a utilitarian attribute, as well as a "no attribute" condition to provide an additional secondary baseline to compare the ethical attribute effect.
This choice resulted in a 3 (attribute: ethical vs. utilitarian vs. no attribute) × 2 (product ratings: low vs. higher) between-subjects de- Then, participants were exposed to Consumer Reports ratings of this new product. The report mentioned on which criteria the sunscreens were evaluated (in particular UVA and SPF, which were explained) and also mentioned that "The worst score of our test program was 5 and the best 95." Participants were randomly allocated to the low rating condition (5/100) or to the higher rating condition (55/100).
We then measured the dependent variable, sunscreen evaluations (α = .99), the mediator, anticipated warm-glow feelings (α = .94), and perceived quality of the sunscreen with the same measures as in Study 1. We also measured participants' green consumption values, using a 6-item, 7-point, Likert scale (e.g., "It is important to me that the products I use do not harm the environment"; Haws, Winterich, & Naylor, 2014), the believability of the Consumer Reports review ("In your opinion, how believable is the review from Consumer Reports?" with answers ranging from 1-"Extremely unbelievable" to 7-"Extremely believable"), familiarity with Consumer Reports ("How familiar are you with Consumer Reports?" with answers ranging from 1-"Not familiar at all" to 7-"Extremely familiar"), and awareness of the brand Daylong ("Did you know Daylong brand before taking this survey?" with a dichotomous answer, yes or no). (2, 609) = 3.39; p = .034; η p 2 = 0.011), was also observed.

| Manipulation checks confounds and brand awareness
We observed no impact of the independent variables and their interaction on the believability of the Consumer Reports reviews (all p > .1), meaning that the low and higher rating conditions generated the same level of believability (which could have been a potential confounding effect). Only six participants mentioned that they knew the brand Daylong before taking part in the study. Below, we present the analyses including those six participants. However, we ran the same analyses without them and found no difference in significance levels.

| Mediation through warm-glow feelings
To test H3, we ran a moderated mediation analysis (Hayes, 2017;model 8;5,000 bootstrap samples) with attribute type as the independent variable (0 = Ethical attribute, 1 =Utilitarian attribute, 2 = No attribute), warm-glow feelings as a mediator, product ratings (0 = low ratings, 1 = higher ratings) as a moderator, and product eva- Consumer Reports as covariates rendered the results slightly neater.
All significance levels stayed constant but for one exception: the index of the moderated mediation when the ethical attribute was compared to the no attribute condition (which was only marginally significant) became significant (Index = 0.42; 95% CI = 0.00 to 0.82).
Overall, these results support H3.

| Discussion
Using different samples, different products, as well as different manipulations of ratings and attribute type, Studies 1 and 2 provide evidence for the positive effect of an ethical attribute when product ratings are low. This effect of the ethical attribute on product evaluation is not observed when the ratings are higher. This effect is explained by the anticipated warm-glow feelings that are generated in the low rating condition. The mediating effect of anticipated warmglow feelings for the ethical attribute is significantly weaker when ratings increase. In addition to Study 1, Study 2 controls for green consumption values, the believability of the ratings, familiarity with the rating agency, and brand awareness. Controlling for these variables yields similar results. Also, Study 2 provides an additional control condition (no attribute), which again yields similar results.
At first sight, a ceiling effect could be present and provide an alternative explanation to our effects (i.e., with higher ratings, product evaluations increase, providing limited margin for an ethical attribute to have a positive effect). However, this hypothesis is highly unlikely. We deliberately operationalized the higher rating conditions Overall, our two studies support our hypotheses.

| GENERAL DISCUSSION
The objective of this research was to test the prediction that ethical attributes have a stronger effect on product evaluations when product ratings are low. In this regard, we show in two studies that when a product receives low ratings, the effect of an ethical attribute on evaluations is positive and more specifically stronger than that of a utilitarian attribute. Also, we show that this positive effect of an ethical attribute weakens as product rating increase.
This research also documents the mechanism that explains the effects of ethical attributes according to product ratings. Specifically, we show that an ethical attribute fuels warm-glow feelings, making consumers anticipate to feel good about their purchase. Interestingly, we find that these warm-glow feelings are moderated by product ratings: when product ratings are low, an ethical attribute makes consumers feel better about their purchase, but this incremental feelgood effect of the ethical attribute decreases with higher ratings.
Importantly, we find that these effects are consistent across different product categories, as well as different ethical attributes.

| Theoretical contributions
The current research provides some important theoretical contributions. The first relates to the stream of literature analyzing the impact of ethical attributes on consumer response. In line with cue consistency theory (Maheswaran & Chaiken, 1991) and the negativity bias (Skowronski & Carlston, 1989), this research brings further empirical support to the notion that when consumers face incongruent cues (for instance when product ratings are low and a utilitarian attribute is present), the negative cue gains importance, that is, consumers disregard utilitarian attributes and rely more strongly on the product ratings. On the contrary, when low product ratings are associated with an ethical attribute (i.e., the cues are not incon-  (Luchs et al., 2010), attribute type (Bodur et al., 2014), price discounts (Andrews et al., 2014), type of brand (Arora & Henderson, 2007), the need communicated with the ethical attribute and the justice restoration potential (White, MacDonnell, & Ellard, 2012)  The third stream of contributions lies in the trade-off that consumers make related to ethical attributes. Luchs and Kumar (2017) show that consumers prefer to trade off hedonic (vs. utilitarian) benefit for ethical benefit. Given the body of research that emphasizes the trade-off between ethical and utilitarian attributes, this research used utilitarian attributes for the control condition of the experiments, and thus focused more specifically on the utilitarian versus ethical attribute comparison to understand when ethical attributes are preferred. In this regard, previous research highlights functional precedence effects whereby functional performance is preferred over ethicality when both are valued (Luchs et al., 2012).
We supplement such research by showing that when the performance of the attributes is not rated, and consumers infer it from product information and product ratings-which is a pervasive situation online-consumers will evaluate a product with ethical attributes more positively than similar products with utilitarian attributes when product ratings are low.
Finally, our results also contribute to the green marketing lit- used in this research such as "Guaranteed child labour free" could be considered an instance of CSR communication) are not only dependent on the brand concept, but could also depend on the ratings and potentially more generally on product or brand perceived quality. It would be for low-quality brands that ethical attributes reveal the most effective.

| Practical and ethical implications
Our research also has managerial and ethical implications for marketers and policymakers. As could be expected, our results show that third-party ratings always have a main effect on consumer evaluations. For companies willing to improve evaluations, the first obvious step to consider is to improve the quality of their product and act on the issues that potential ratings pointed at. When quality improvement appears too costly or too complex, managers could consider an ethical positioning to increase consumer evaluations. Such consideration may intervene at the earlier stage of the product BEZENÇON ET AL.
| 1579 development process in order for the product and the brand to be legitimate if the ethical positioning-and thus the use of ethical attributes-is actual. Managers who anticipate their products to be poorly rated may benefit from the adoption of such positioning as consumers may value the ethical attribute and eventually form positive evaluations of the products. An ethical positioning could occur at a further stage of the product life cycle. For instance, if a product or service received low ratings, marketers may reposition the product or service and highlight its actual ethical aspects to increase consumer evaluations. For example, a restaurant that received low ratings could promote, in its communications and marketing, its ethical labor practices and/or initiatives it is undertaking (or planning to undertake) to reduce its carbon footprint (e.g., sourcing sustainable products for its meals, reducing food waste, etc.). However, as stated earlier, the first and most obvious step is to fix the issues that led third parties to rate the product poorly.
Nevertheless, when a product receives poor ratings after its market launch, brands may be tempted to design and use unsubstantiated ethical attributes, a phenomenon termed greenwashing when applied to superficial environmental actions (W. S. Laufer, 2003). Given the deceptive nature of such techniques, policymakers may intervene and either support the use of authentic ethical attributes designed and attributed to the product by independent third parties, or consider increasing the regulation regarding the use of brand-designed ethical attributes that do not correspond to reality.

| Limitations and future research
As with every research, some limitations and subsequent avenues for further research may be highlighted. First, this research focused on product ratings coming from experts. Although our theoretical framework could also be applied to consumer ratings (such as Tripadvisor ratings), the effects could vary due to factors such as social influence. In addition, it is less clear whether consumer ratings account for the ethicality of the product, whereas in the expert ratings we used, we explicitly mentioned the underlying criteria of the ratings (which did not include ethicality). Nonetheless, further exploration of ethical attributes and consumer ratings provide interesting future research potential. Also, product ratings represent only one major cue of product quality. Future research could investigate other quality cues and test whether the interaction effect of quality and ethical attribute replicates.
Future research could also test whether the effect of ethical attributes for low-rated products differ based on whether consumers perceive the source of the ethical attribute as an independent thirdparty certification company or the brand. Previous research shows that consumers easily engage in attributional processes in response to product quality perceptions (D. Laufer & Gillespie, 2004;Lei, Dawar, & Gürhan-Canli, 2012). The increase in warm-glow feelings and evaluations shown in our studies might therefore vary when the ethical attribute is perceived to be designed by the brand itself.
Further, the two empirical studies presented in this research used quantitative product ratings. However, product quality may not only be assessed through ratings but also through reviews. To this regard, Archak, Ghose, and Ipeirotis (2011) suggest that the text provided in product reviews might represent an important determinant of consumers' evaluations. Hence, further research could investigate whether the current results replicate with poorly reviewed products instead of poorly rated products.
Finally, this research limits the outcome under study to product evaluations. Further research could examine whether a low-rated product with an ethical attribute also benefits from the "benevolent halo of corporate social responsibility," whereby products associated with prosocial activities are perceived as performing better (Chernev & Blair, 2015). In that case, ethical attributes may contribute to the perceived overall performance of the product.

ACKNOWLEDGMENT
The research is part of the activities of SCCER CREST, which is financially supported by the Swiss Commission for Technology and Innovation (CTI) (Grant No. KTI 1155000154).