Signs promoting sustainably labelled seafood via social norm messaging were developed and employed in German and Norwegian supermarkets. Sales volumes were observed under different conditions.
The first hypothesis of this study is based on the finding that giving information about a certain issue can in itself influence behaviour [26
]. Providing information about sustainable seafood on a fish-shaped sign in the supermarket may in itself be prompt or a “nudge” that increases the likelihood of sustainable seafood consumption [26
]. Hence, our first hypothesis is
Hypothesis 1 (H1).
Placing a sign with information about sustainable seafood (i.e., a prompt) on the seafood freezer will significantly increase the sale of sustainably labelled seafood products compared to the baseline.
Ölander and Thøgersen [26
] argue that the combination of information and a social norms nudge should be even more effective than information or a social norms nudge alone. Thus, if the information about sustainable seafood was extended with social norm information on how many people buy sustainability-labelled seafood, where, and when, a bigger increase in the sale of sustainably labelled seafood should be obtained. However, an important question is how large does the reference group need to be to produce a significant change? It may be assumed that when the reference group is larger than 50%, the sale of sustainability-labelled seafood products should be significantly higher than the baseline or the information-only condition. If the reference group is smaller than 50%, it may be that less labelled seafood would be sold compared to the information-only condition, because participants adapt to the behaviour of the silent majority. Consequently, the second and third hypotheses are
Hypothesis 2 (H2).
Adaptation to the obvious majority: Bigger reference groups (>50%) in the social norms message lead to higher sales of labelled products compared to baseline and to a prompt about sustainable seafood only.
Hypothesis 3 (H3).
Adaptation to the silent majority: Smaller reference groups (<50%) lead to lower sales of labelled products compared to the prompt about sustainable seafood only.
2.2. Materials and Methods
In Norway, the study was carried out in five medium-sized supermarkets. All stores were located in Trondheim. In Germany, the study was carried out in one medium-sized supermarket and three discounters in Würzburg. In both cases, the supermarkets were located in different parts of the city to cover potential differences between consumers from different neighbourhoods. In each country, one supermarket served as a control condition, which means that no intervention was implemented there.
All participating supermarkets had a section with a sufficient selection of frozen seafood including labelled products and nonlabelled equivalents. Regarding labelling, products certified by the Marine Stewardship Council (MSC) for wild fish and the Aquaculture Stewardship Council (ASC) for farmed fish were selected, each being the most popular certification system in their domain. As many product pairs as possible were identified in each supermarket, resulting in 15 pairs in total of which 5 pairs were in German and 10 pairs were in Norwegian supermarkets. Products of one pair had similar characteristics regarding product type, taxa, price, weight, and packaging. One product in each pair carried an MSC or ASC sustainability label while the other one did not. Hence, it was possible to compare sales of labelled versus nonlabelled products while keeping the other product characteristics constant. Table A1
in Appendix A
lists all product pairs and their key characteristics.
Signs were placed on the glass cover in the frozen seafood section. The signs themselves were made from thick glossy paper in the colours blue and grey. The form of the signs resembled a fish (see Figure 1
). Eight different conditions were created. On all signs, the MSC/ASC seafood label was displayed with a short explanation of its meaning (“The MSC/ASC certification contributes to sustaining marine resources”). This may be considered a prompt, reminding consumers of the label at the point of purchase. The message can also be interpreted as an injunctive norm, implicitly reminding consumers of their responsibility to choose environmentally friendly consumption alternatives. Sign 1 only contained this prompt/injunctive norm and no descriptive norm information. Signs 2–8 displayed an additional text including the reference group size, type, and behaviour (4%, 11%, 28%, 52%, 69%, 82%, or 91% of all customers buying seafood in our shop yesterday chose MSC/ASC) (see Figure 1
). Hence, the reference group was defined as close as possible to the receiver regarding time (yesterday), location (same supermarket), and action (seafood purchase).
The text on the signs was in the national language. In Germany, both the MSC and the ASC certifications were indicated on the sign while in Norway only the MSC was mentioned because ASC labelling had not yet been introduced in Norway at that time. The eight conditions were rotated in a randomized order so that the same condition was never applied two days in a row in the same supermarket or at the same day in two different supermarkets in the same city. Sales numbers for the identified products were obtained after completion of the experimental period from the store managers. The period included two weeks before, two weeks during, and two weeks after the implementation of the signs.
2.3. Data Analysis
Sales data for seafood products was used as the dependent measure on two different levels, nested into each other: the supermarkets and product pairs. Multilevel analyses recognize that product purchase in one supermarket systematically varies from purchases in other supermarkets and account for the fact that individual sales data would not provide independent observations. In this two-level analysis, the variation is split into two parts: one part that is constant across products and one part that is constant across supermarkets.
To analyze the data on two embedded levels, a multilevel mixed-effects linear regression analysis was conducted. First, relative sales of labelled versus nonlabelled seafood are regressed on the treatment variable (before, prompt only, 4%, 11%, 28%, 52%, 69%, 82%, 91%, after) and country (Norway, Germany) as well as their interaction. Robust standard errors were applied because full normal distribution of errors could not be ensured. Based on observed tendencies and to increase statistical power, the condition variable was also aggregated into five groups which were (1) before, (2) information only, (3) minority reference group <50%, (4) majority reference group >50%, and (5) after.
One product pair, Norwegian shrimps, had to be excluded from the analysis as it was only sold once during the whole experimental period, providing no variance. For the analysis there remained N = 3277 valid observations of product sales. Consequently, nine product pairs remained in Norway, and five in Germany. A preliminary look at the dataset revealed that the total number of frozen seafood products within the target categories sold per day per supermarket was significantly and substantially higher in Germany (M = 10.15, SD = 8.52) than in Norway (M = 4.03, SD = 3.69). Therefore, the following analysis focuses on the relative amount of labelled versus unlabelled products sold on each day. Before the intervention, relatively more labelled products than unlabelled products were sold in Norway (Mlabelled = 1.38, SD = 1.90; Munlabelled = 0.74, SD = 1.47) than in Germany (Mlabelled = 3.85, SD = 4.64; Munlabelled = 7.66, SD = 8.29).
The multilevel analysis applying all experimental conditions (Nproduct_pairs = 896) revealed significant (β = 0.30, p < 0.001) inter-country differences with Norwegian supermarkets selling a higher percentage of sustainability-labelled seafood (66.89%) than Germany (33.80%) during the experimental period, same as before. There was no significant effect on the sale of labelled seafood in any of the interventions. Interactions between country and condition were also not significant. Hence, in comparison to the control group or the baseline, there is no indication of any of the signs, irrespective of whether or not descriptive norms were included, or with higher or lower percentage, leading to more labelled seafood sales.
In order to simplify, and because the detailed analysis revealed no differences between different descriptive norms conditions, a new multilevel analysis was done with only five groups (before, prompt, low social norms, high social norms, after). Again, a significant main effect for country was found. This analysis revealed a significant increase of labelled seafood sales in the prompt-only condition in Norway and a significant decrease of labelled seafood sales in the low reference group descriptive norm condition in Germany. A marginally significant decrease was found in the high reference group descriptive norm condition in Germany as well (Table 1
and Figure 2
The total amount of studied types of seafood (labelled as well as unlabelled) sold increased significantly by 42% during the experiment (p
= 0.01) (see Figure 3
), 72% in Norway (p
= 0.04) and 29% in Germany (p
= 0.11). This further confirms that the interventions actually had an effect.
The main strengths of this study are that it was carried out in a field setting, in ordinary supermarkets in two different countries, and with the outcome measure being actual sales data, rather than self-reported behaviour. However, real-life settings are also “messy” and difficult to control, which may lead to unexpected, confounding influences. For example, repeat customers might recognize the signs and notice the strongly differing percentages in the descriptive norm messages. This could lead them to become suspicious and doubt the credibility of the messages and potentially affect their choices. More important is the fact that contextual factors can rarely be controlled in a field experiment. We know that participating supermarkets did not offer special discounts on seafood during the experiment. However, there could be indirect effects of discounts on other products or in other supermarkets, which we could not control. Also, due to data confidentiality issues, only sales numbers for the identified product pairs could be obtained. This leaves unknown how the sales developed of types of seafood products other than those included in the study.
More frozen seafood of the studied types was sold in the German than in the Norwegian supermarkets. This seems to contradict statistics showing that the total amount of seafood consumed in Norway is higher than in Germany [63
]. A possible explanation could be that the included German supermarkets are bigger and have more customers than the Norwegian supermarkets. Another reason might be that Norwegian consumers buy relatively more fresh seafood because of the shorter transport distances from catch to counter. Within the frozen seafood category, the number of different sustainability-labelled seafood products was higher in the Norwegian than in German supermarkets and so was their share of the sale within these categories. The higher share of sale in the Norwegian supermarkets might partly be due to the higher number of different labelled products leading to higher exposure to the labels in the Norwegian than in the participating German supermarkets. In addition, the participating supermarkets in Germany were mostly discounters, perhaps attracting consumers who put a stronger emphasis on low prices. In Norway, there is not necessarily a price premium connected to sustainability labelling whereas in Germany the labelled option is always the more expensive option in each pair (see Appendix A
). This also suggests fewer barriers for the Norwegian consumers to respond when prompted to buy sustainability-labelled seafood.
The descriptive norm interventions did not have the intended effect on the proportion of sustainability-labelled seafood sold—on the contrary. In Germany, the descriptive norm messages led to a significant fall in the ratio of sustainability-labelled seafood sold, compared with the baseline. Hence, Hypothesis 1 was not confirmed. Further, in Norway, the increase in the proportion of sustainability-labelled seafood being sold that was registered when a prompt about sustainable seafood labels was provided on the counter was neutralized when descriptive norms information was added to the sign.
So, against all the expectations and the mounting evidence supporting the power of descriptive norms, only a negative (boomerang) effect was found for the employed descriptive norm interventions. Hence, hypothesis H2 was also not confirmed. The sign with reference group >50% did not lead to an increase in sales on labelled seafood. The expected negative effect of a <50% reference group (hypothesis H3) was confirmed in both countries compared with the prompt-only condition and also in Germany compared with the baseline. However, irrespective of the size of the reference group, the descriptive norms interventions produced a boomerang effect in this case.
It can hence be concluded that the employed type of descriptive norms communication was not effective at promoting sustainable seafood in these Norwegian and German supermarkets. Since consumers were not interviewed individually, we can only speculate as to why. The relatively strong increase in total sales of the covered types of seafood during the experimental interventions refutes the possibility that the consumers just did not notice the interventions. Consumers increased their general consumption of the covered types of seafood when there was a sign prompting sustainable seafood, with or without social norms information, installed on the frozen fish counter, whereas the proportion of labelled versus unlabelled seafood overall remained stable. Similarly, Payne and Niculescu [35
] found a general increase of fruit and vegetable purchase when displaying a message about descriptive norms regarding fruit and vegetable consumption. It seems that the message employed by Payne and Niculescu [35
] made the food group “fruit and vegetables” salient and prompted the consumption of fruit and vegetables, and the same in the present study regarding frozen seafood. Frozen seafood consumption was triggered irrespective of the message printed on the signs. Overall, consumers’ processing of the information on the sign seems to have been too shallow to make them change their usual patterns when purchasing seafood. In other words, consumers bought more of what they usually buy as a response to the fish-shaped sign installed on the counter.
After the intervention, the total sale of the covered types of seafood decreased again, back to the level before the intervention. This further suggests that the positive effect produced during the intervention was mainly due to the signs making seafood as a product group more salient.
The apparent insufficient attention to the actual message on the signs could be due to the information provided being too much and presented in too-small-lettered text. As in-store decisions are characterized by time pressure and the use of simple heuristics [55
], expecting consumers to read a relatively long text might not be realistic in this type of environment.
This may also partly explain why the sustainable seafood prompt had a stronger positive effect in the Norwegian than in the German supermarkets. In the Norwegian supermarkets, there were on average twice as many sustainability-labelled products and the customers here bought twice as high a proportion of sustainability-labelled seafood products as those in the German supermarkets. This suggests that the sustainability label was already more salient to the participating Norwegian than German consumers, making it likely that they were able to process the prompt more effortlessly.
However, the significant boomerang effect produced by the descriptive norms communication suggests that at least some consumers processed this information. This effect can hardly be explained by the fact that the interventions led to an increase in the total seafood sales. Rather, it suggests that those consumers who processed the descriptive norms information were discouraged by it. Prior research suggests that boomerang effects can result from social norms communication when receivers feel that it is pressing and potentially limits their freedom (i.e., psychological reactance) [64
]. However, this effect has until now only been reported in connection with injunctive norm communication. To the best of our knowledge, this is the first study finding that descriptive norm communication can also produce a boomerang effect, suggesting psychological reactance.
An important question that remains insufficiently answered is the extent to which the text on the sign was processed at all and thereby contributed to the found effects. Did the consumption of seafood as a product group increase because the sustainable label was made more salient, or because the shape of the sign prompted seafood in general? As the answer to this question would make an important difference to the communication of sustainable seafood, this question is investigated in Study 2. Also, individual characteristics of customers and possible interactions between customers could not be controlled in the field experiment. Therefore, it was decided to conduct Study 2 in a more controlled environment: a laboratory setting.