Source effects

Research in social psychology has demonstrated that the source of a message can have an impact on how it is processed. For example, one study demonstrated that source likeability influences the extent to which participants’ opinions change to align with that of the source [14]. It has been argued that individuals are more susceptible to persuasion if it comes from a credible source (with credible sources holding expertise, being trustworthy and being good willed) [15]. In the context of sales, if claims about products come from trusted sources, the claims are more likely to be accepted [16]. People often defer to a course of action recommended by someone with expertise [17] and knowing that a message comes from a trusted source works as a mental shortcut to simplify the decision-making process. As such, using trusted sources to endorse claims makes these claims more credible to the consumer [16, 18–21]. Supporting health claims on health labels with an endorsement from a credible source could therefore be an effective mental shortcut for consumers to rely upon when selecting healthier foods.

There is little direct evidence demonstrating the impact of source credibility on the effectiveness of food health labels in shaping consumer behavior. However, available indirect evidence is consistent with higher credibility sources having positive effects [19, 22–24]. In their study, Feunekes and colleagues tested whether endorsements by international and national health organizations have an influence on evaluations of health label credibility [23]. In an online study with over 1,600 participants across four European countries, smiley face health labels were presented with an endorsement from either the World Health Organization (WHO), the European Union (EU), European Food Manufacturers (EFM), or a national nutrition organization from the relevant country. Results showed that credibility was significantly higher in the presence of the endorsement. Furthermore, the less nutritionally targeted EU and EFM endorsements were evaluated as less credible than those from the WHO and the respective national organizations.

While these results suggest there are benefits to high source credibility labelling, they are limited by the use of attitude change as a measure rather than behavior. Source credibility might alter attitudes and intentions but not behavior. Without further insights on the impact health labels with credible sources have on purchasing behavior, it is difficult to estimate how effective they might be as an intervention, and as such, how cost effective such interventions might be.

In summary, health labels may serve as an easy and affordable method of nudging consumer behavior but they have shown only limited effectiveness. Attributing the labelling to a trusted, highly credible source of health information might boost the effect of the label. In this study, we test source credibility directly using purchasing behavior as the outcome measure.

Current study

We tested whether high source credibility improved labelling using vending machines. Vending machines are a useful tool for studying the effects of interventions on food choice as they offer exceptional levels of experimental control [8, 25, 26]. Sales data were collected from two vending machines at a hospital site over three months, with one machine in the A&E department and another at reception. The machines sold healthy products (e.g., baked crisps, cereal bars) and unhealthy products (e.g., standard crisps, chocolate bars) in equal proportions. Each week, the machines were independently and randomly assigned to one of three conditions, with the outcome measure being the number of products sold. There were two labelling conditions and a control, no label condition. In the low credibility condition, healthy products were labelled “lighter choices” (Fig 1). In the high credibility condition, the same label was appended with the National Health Service (NHS) logo (“NHS lighter choices”; Fig 2). In the UK, the NHS is one of the most trusted sources of health information, and its endorsement is often seen as a stamp of approval for products or treatments [27, 28].

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Low credibility label.

https://doi.org/10.1371/journal.pone.0296901.g001

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. High credibility label.

https://doi.org/10.1371/journal.pone.0296901.g002

If high source credibility improves the effects of labelling, purchasing behavior should shift from unhealthy products to healthy products. There should therefore be more healthy purchases and fewer unhealthy purchases in the high source credibility condition.

Participants

Participants voluntarily made purchases from vending machines without being aware that they were taking part in a study. Ethical approval was granted from the NHS research ethics committee (IRAS number: 231390) and the Cardiff University School of Psychology Ethics Committee. Permission was given not to collect consent from participants and to not inform them of the nature of the study.

Study design

Two vending machines were used (see Implementation in S1 Methods). One machine was in Accident and Emergency (A&E), the other in a reception area. Location and number of vending machines was determined by practical rather than scientific considerations.

The study lasted 90 days. Every six days, each machine was randomly allocated to one of the labelling conditions (no label, low credibility, high credibility) such that each machine was in each condition five times in total. Practical reasons limited the length of the study to 90 days rather than power considerations.

Machines housed the same stock in the same positions within the machines throughout. There were 12 healthy products and 11 unhealthy products (see Vending in S1 Methods). Healthy snacks satisfied the Welsh Hospital Healthy Vending directive constraints [29] whereas unhealthy products did not. All products were sold for the same price (£1).

Purchasers could use either cash or credit card in the vending machines.

Labels

Labels were placed underneath the product referred to (see Labels in S1 Methods).

Analysis

The data were analysed as a mixed model with the lme4 package in R [30]. Sales volume was the outcome variable e.g., to evaluate whether the high credibility label influenced purchases we compared total sales in the high credibility condition against total sales in the no label condition. Individual products (N = 23) and time (N = 15) were included as random effects. This allowed for generalisation across time and product range. Model specification was maximal [31], in that all possible random effects parameters were included.

We computed an omnibus regression model that included labelling (no label, low credibility, high credibility), product health value (healthy, unhealthy), machine (A&E, reception), and payment type (cash, credit card) as fixed effects, and product and time as random effects. The model is shown below in R pseudo-code:

Backfire effect

We found that the high source credibility label increased sales. The effect was restricted to the A&E machine and to credit card sales, but the effect was nonetheless significant. An intervention designed to reduce consumption that leads to an increase in consumption, albeit under restricted conditions, is the reverse of the intended result, i.e., a backfire effect. We consider three explanations for this.

The first is that there was psychological reactance against the label [33]. Theories of psychological reactance postulate that imposing rules or constraining freedoms can result in individuals reacting against them and showing increased desire for the restricted choice [33]. While health labels do not enforce choices, they suggest courses of action, which some may interpret as an intrusion of their personal freedom [34]. There is evidence of reluctance to accept health claims by health promoters [35], and attempts to influence diets are sometimes perceived as paternalistic violations of agency and freedom of choice by the government [36]. In the context of this study, consumers might have objected to the suggestion that they should be making certain food choices [33, 34], with the high credibility label being perceived as particularly paternalistic. An alternative possible source of reactance is the fact that individuals in A&E may have been waiting a long time to receive medical attention. They may have felt irritated with the NHS about their predicament and reacted by deliberately choosing the products not endorsed by the NHS.

The second explanation for the backfire effect is based on goal fulfillment. Evidence suggests that the presence of a healthier option within a selection of less healthy options can increase unhealthy consumer behavior [37]. In their study, Wilcox and colleagues showed that high self-control individuals were more likely to select the most indulgent side option available within a range when the range included a salad option, compared to when it did not. The authors hypothesized that exposure to healthier items vicariously satisfied consumers’ existing health goals (regardless of whether the product was chosen) [37]. The presence of NHS labelling may have indirectly supported goal fulfillment. Making a selection from a range that had a strong NHS branding presence could have been interpreted as making sufficient progress towards consumers’ health goals.

The final explanation is that the labelling may have negatively influenced product perceptions. While for some, healthier foods are more attractive, there is evidence that healthy foods can be perceived as less tasty, less satiating, and more expensive [38–40]. The NHS label may have drawn attention to the healthy features of the products, and in doing so, inadvertently made the products appear less attractive for consumption, e.g., less tasty. Future studies can investigate which sorts of healthy messaging give rise to negative product perceptions and whether the labels used here fall into that category.

Variability across location and payment type

Backfire effects were larger in the A&E machine than the reception machine. Why was there such a difference between the two locations? We consider two possibilities. First, if consumers were reacting against the label, they would have been more likely to do so in A&E than in reception. Consumers would have been more stressed in A&E, they may have been waiting for a long time, concerned about those admitted to hospital, hungry or tired. These factors are likely to maximize irritation with the authority who they viewed responsible for their wait (the NHS).

Second, the effects of labelling more generally may be greater in A&E than in the reception area. Consumers in A&E are captive–many are waiting with little to occupy themselves, allowing time for using the labels and absorbing their content. In comparison, the reception area had high passing footfall and minimal seating. Participants in the reception area may not have attended to the labels to the same degree as those in A&E. The difference in mental states of consumers between the two locations might also contributed to greater use of the label in A&E. Mental shortcuts are more likely in scenarios where individuals wish to exert minimal cognitive effort when making their choices [14, 41, 42] and indeed stress has been shown to increase reliance on mental shortcuts [43]. Participants in A&E may have sought to ease the decision process as much as possible by using cues available within the environment to help them make a selection.

There was also variability across payment type. Participants paying by credit card were influenced by the labels to a greater extent than those paying by cash (there was no evidence that cash purchases were influenced by the label). It is not clear why. Perhaps those who made multi-purchases (presumably over-represented in credit card payments) were more likely to use the label as a source of information because they had to make more choices within the fixed range. Multi-purchases might also have been a sign of greater stress and greater irritation with the environment, and therefore reactance was more likely.

Limitations and future work

One limitation of this study was that it used only a single type of high credibility label—the NHS endorsement. While the NHS is a trusted source of health messaging [27, 28], it is possible that endorsement by the NHS may not have provided the credibility that we assumed. At least two issues could be problematic. First, the NHS not only provides health advice, but it also manages health sites such as hospitals. It can therefore be seen as responsible for the predicaments of the consumer, e.g., long waiting times in A&E. A high credibility label divorced from the management of the health site may not lead to the same negative effects as the NHS label and may even promote healthier sales. Future research needs to investigate why we observed the negative effects of the NHS label and whether these effects generalise to other high credibility sources.

Second, trust in the NHS may not span across all health-related messaging. It may be most suited towards messaging around health services, rather than messaging that promotes certain dietary behaviours in lieu of others. NHS England commissioned a qualitative investigation of the NHS brand identity, which highlighted how NHS branding was recognisable and brought confidence about the services being delivered [44]. Whether this same confidence would expand beyond NHS services and into social marketing practices that aim to promote healthier behaviors is unknown. Further work should investigate the degree which the NHS is trusted to deliver dietary advice as well as services.

A related issue concerns perceptions of the healthiness of the products. The products sold in the vending machine were selected based on whether they were compliant or non-compliant with Welsh Hospital Healthy Vending directive constraints [29]. However, this does not guarantee that consumers perceived these products as intended (i.e., healthy vs unhealthy). Any incongruency between consumers perceptions of product health, and what is stated on a label, could impact on attitudes and in turn behaviour.

Finally, there were two limitations with respect to experimental design. First, practical constraints limited our work to a single hospital across a thirteen-week time window. A greater sample of sites and longer study would have increased power to detect differences between labels (e.g., French and colleagues tested at 24 sites [8]). Furthermore, the hospital environment might have been unusual in that health is prominent and stress levels are high. Generalising our findings to sites other than hospitals will therefore require widening the range of sites tested.

Secondly, participants were not individually randomly allocated to a labelling condition. Instead, they were randomly allocated as a group, based on the six-day period in which they were at the hospital e.g., those who arrived at A&E in period 7 were allocated to the high credibility condition. This meant that observations (participants) were less likely to be independent than if we had allocated each participant individually to a labelling condition. This issue was partly alleviated by allocating each labelling condition to multiple, randomly selected weeks within the study period, rather than just one, but an improvement would be to randomly allocate at an even finer level of temporal granularity e.g., each day rather than each week.

Conclusion and implications

The purpose of this study was to determine whether greater source credibility would improve the performance of health advice labelling in promoting healthy snack choices. We found no evidence that it does: There were no positive effects of the labels, with or without high source credibility. Indeed, we observed a significant backfire effect whereby consumers selected a greater proportion of unhealthy products when healthy products were accompanied by a high credibility label, the NHS endorsement, at least for participants who paid by credit card. While testing with a wider range of high credibility endorsements is needed to establish to what extent our results generalise, the health implications for labelling in hospitals are clear: high source credibility labelling is not effective in high stress environments and may even cause harm.