Evaluating the impact of the documentary series Blue Planet II on viewers' plastic consumption behaviors

The global scale of the ocean plastics crisis demands a collective change in plastic consumption behaviors. The documentary series Blue Planet II has been praised for driving changes in consumer behaviors by raising awareness about this issue, yet there is little evidence that directly links the documentary to viewers' plastic consumption. We investigated the effectiveness of Blue Planet II as a behavior change intervention by conducting randomized control trials and used revealed preferences to measure plastic consumption behaviors. Although environmental knowledge was found to be positively influenced by Blue Planet II, this did not translate into a behavioral change among participants. Our results support the hypothesis that, due to the complexities of human behavior, exposure to a single documentary is unlikely to lead to a distinct increase in individual pro‐environmental actions. However, the potential for Blue Planet II to have an impact at a wider societal level, namely through influencing policy, remains unexplored.


| INTRODUCTION
Marine litter, particularly plastic debris, is an emerging and critical environmental issue (Hartley et al., 2018), occurring as a result of human actions (Pahl, Wyles, & Thompson, 2017). Therefore to address plastic pollution, alongside the implementation of environmentally conscious policy and infrastructure, conservationists must also promote pro-environmental behaviors surrounding the consumption and disposal of plastics (Steg & Vlek, 2009).
The topic of marine plastics gained public attention in the United Kingdom in the last decade, with the airing of Blue Planet II, a nature documentary series on marine life produced by the BBC, being reportedly one turning point (Hunt, 2017). Unlike the series' predecessor, The Blue Planet, the high level of conservation messaging within Blue Planet II has been credited with raising awareness of plastic pollution to a nationwide audience (Jones et al., 2019), with total viewing figures reaching 14.01 million across the United Kingdom (BARB, 2017). In line with the mission of the BBC, as a public service broadcaster, to "inform, educate and entertain" viewers (BBC, n.d.), Blue Planet II producers aimed for the show to provide a "platform that got the broader message (about plastic pollution) out" (Honeyborne, 2018). An analysis of Twitter activity relating to plastic waste found that conversations around this topic in the first quarter of 2018 had more than doubled compared to the same period in the previous year (Joyce, 2018). The airing of Blue Planet II was not only linked with a boost in public interest, but also allegedly with changes in plastic consumption behaviors (Collins, 2018), a phenomenon dubbed the "Blue Planet Effect" (Hunt, 2017). Although it is difficult to determine if the series had intended to influence plastic consumption behaviors, Blue Planet II's specific focus on human's environmental impacts and direct messaging highlighting a "responsibility to care for our blue planet" (Blue Planet II, Episode 7), suggests that the series hoped to create a call to action amongst viewers.
However, the current evidence relating to the influence of Blue Planet II on viewer's plastic consumption is largely anecdotal or based on self-reported behavior (Collins, 2018), which is often an unreliable measure of outcomes, with low correlations to observed behaviors (Kormos & Gifford, 2014). As marine plastics become a highly salient issue, individuals may also over-report proenvironmental behaviors to comply with subjective social norms . It is therefore important that assessments of Blue Planet II use a meaningful measure of behavior and a robust experimental design to promote evidence-based evaluation and understand the true impact of this series (Thomas-Walters, McNulty, & Veríssimo, 2019;Veríssimo, 2013).

| Behavior change
In recent years, there has been an increasing focus across conservation organizations of using media-based intervention to encourage pro-environmental behaviors, in response to the growing disconnect toward environmental problems in urban population with limited access to nature . Documentary film has historically been used as a tool for promoting social change (Karlin & Johnson, 2011), and wildlife documentaries have been highlighted as a potential avenue for creating pro-environmental change, as they are able to articulate often complex ideas about environmental issues to a wide-reaching audience (Janpol & Dilts, 2016). Past studies have found nature documentaries to be able to promote positive change through influencing viewers' concern for the environment (Nolan, 2010), as well as increasing support for conservation (Lin, 2013). However, the ability of these programs to encourage specific behavioral actions has not been as well documented.
Human behavior is highly complex and influenced by an array of internal and external factors (Tonglet, Phillips, & Read, 2004). Several theoretical frameworks for identifying the influencers and barriers to pro-environmental behaviors have been developed (Davis, Campbell, Hildon, Hobbs, & Michie, 2015), one of the most widely used being the theory of planned behavior (TPB) (Ajzen, 1991). The TPB identifies the main constructs guiding behavioral intention to be: knowledge of the behavior and its outcomes; attitudes towards the behavior; subjective social norms around preforming this behavior; and the perceived ability to exert control over one's behaviors and the subsequent outcomes, or perceived behavioral control. We used the TPB to guide this research as this model has been previously found to be an appropriate predictor of pro-environmental choice behaviors such as plastic bag use (Sun et al., 2017) and recycling (Tonglet et al., 2004). However, previous research utilizing the TPB have often relied on inferred changes either using these constructs as behavior change proxies or relying on self-reported behaviors (Ajzen, 2011). There is therefore a need for studies to focus on measures of actual behavior change outcomes when utilizing the TPB.
In this paper, we aimed to evaluate the popular media hypothesis that the high levels of conservation messaging in Blue Planet II had significantly impacted the plastic consumption behaviors of viewers. We used a randomized control trial experimental design and actual observed behaviors to answer two main research questions: (i) what is the effect of the conservation messaging in Blue Planet II on the TPB constructs (i.e., knowledge, attitudes, subjective social norms and perceived behavioral control) when compared to a documentary on a similar topic that holds no conservation messaging and (ii) does the increased conservation messaging present in Blue Planet II influenced viewers' actual choice of single use plastics by increasing the likelihood of choosing a pro-environmental alternative when compared to a documentary on a similar topic that holds no conservation messaging. We used an active control group for this research design in order to provide a relative-effect estimate.

| METHODS
We used randomized controlled trials to test the relative influence of the conservation messaging within Blue Planet II on plastic choice behaviors compared to a control group. We designed an experiment with two conditions: the treatment condition (Blue Planet II, Episode 7, a nature documentary with high levels of conservation messaging) and a control condition (The Blue Planet, Episode 1, a marine focused documentary with no conservation messaging) to which participants were assigned randomly, unaware of which group they were in.
To select the specific episodes used for the control and treatment interventions, we used the program NVivo (12) to code the content of the scripts of each episode across Blue Planet II and The Blue Planet. We identified the recurring themes across the two documentary series to be: animal behavior, natural histories, anthropogenic impacts and conservation messaging (Table S3). We used this framework to code the transcripts from each episode of The Blue Planet and Blue Planet II and calculated the total percentage cover of these themes across each episode. From this analysis, episode seven of Blue Planet II was found to hold the highest cover of messaging on conservation and anthropogenic impacts, including specific messaging about ocean plastics, and was therefore selected as the treatment intervention. We designated episode one of The Blue Planet as the control intervention because it covered similar marine biology themes to the treatment episode but held no conservation messaging or information on anthropogenic impacts on the marine environment.
Trials were conducted in and around London and Oxford, United Kingdom, between June 2018 and July 2019 (Table S1). Participants were recruited through posters distributed around the local area and online. Those invited were offered to attend a screening of a nature documentary with the possibility to win a raffle prize worth £50. We conducted a priori power analysis in RStudio (version 1.2.5019) using the package pwr (Champely, 2018). We used the following parameters to determine our recommended sample size for testing the difference between two groups: alpha = 0.05, power = 0.8 and between group effect size = 0.25 (Cohen, 1977). Results determined that a minimum sample of 99 participants per group would be needed to detect this change. Our final sample size was 150 participants, which provides a statistical power of 0.7.
Participants completed a questionnaire before and after each intervention ( Figure 1). We designed this questionnaire to elicit information on the behavioral constructs identified within the TPB model, including: knowledge of environmental issues and behaviors; attitudes towards pro-environmental actions; subjective social norms around performing pro-environmental behaviors; and perceived control of wider environmental issues (Appendixes 1 and 2). The questions were structured as statements with a Likert scale response. We included both a mix of questions relating to the specific environmental issues in Blue Planet II, such as ocean plastic pollution and coral bleaching, as well as more general questions covering pro-environmental behaviors. We included demographic questions to elicit information on the participant's demographic background, including their formal education level, age and gender. This information was used to verify that the samples were comparable across the control and treatment groups, and to identify any possible confounding characteristics. In the final section of the questionnaire, we asked the participants to identify their previous exposure to either The F I G U R E 1 Flowchart of our randomized control trial experimental design including pre-experimental and post-experimental measures. The final n of the control and treatment groups did not add up to 150 sample due to incomplete surveys from four participants Blue Planet or Blue Planet II. We piloted our questionnaire on a sample of 13 postgraduate students studying environmental sciences from Imperial College London in order to test the general layout and clarity of the questions within the survey instrument.
The behavioral outcome was measured using revealed preferences towards plastic or paper packaging (with paper used as the pro-environmental alternative). We asked participants to swap an allocated voucher for a snack before and after the intervention screening ( Figure 1), with each snack option presented in both plastic and paper packaging ( Figure S1). The plastic packaging used for these snack and drink options was similar to that highlighted in the images of marine plastic pollution shown in Blue Planet II, Episode 7. In order to reduce any confounding variables in this choice, the flavors and sizes of the drink and snack choices were controlled to be consistent across the different packaging options (Appendix 3). We observed these choices, categorizing them into either plastic or paper and matched this back to the participants questionnaire through a corresponding number found on their voucher. This observation was carried out covertly in order to minimize social desirability bias. This experimental design was approved by Imperial College London research and ethics committee (IREC, case number: 2018-01383666-DUNN-M).
Four weeks following each experimental trial, we sent participants a follow-up email regarding the £50 raffle prize. The email required participants to select what snacks they would like to spend their prize voucher on from a mixture of options presented in both plastic and paper packaging (Appendix 4). Responses were coded into either a choice of paper or plastic and matched back to the individual participant to determine the retention of behaviors across both the control and treatment groups.

| Analysis
We calculated participant scores for knowledge, attitudes, subjective social norms and perceived behavioral control by summing the Likert scale scores across the questions related to each construct. Questions that were negatively worded were reverse scored and included in this calculation. We then coded participants' revealed preference behaviors before and after the intervention as either 0 for a choice of plastic or 1 for a choice of paper. We conducted all statistical analysis using RStudio (version 1.2.5019), and the packages epiR (Stevenson et al., 2020) and lme4 (Bates et al., 2015).
In order to examine the research questions: (i) what is the effect of the conservation messaging in Blue Planet II on the TPB constructs and; (ii) does the increased conservation messaging in Blue Planet II influenced viewers' choice of plastic, we first used odds ratio testing. This effect size measure was used to examine the influence of the control and treatment interventions on both the likelihood of change in questionnaire scores and preference choice behavior. Although this analysis is useful for identifying the size and direction of the observed effect, it does not take into account the influence of covariates on the observed outcome. Therefore, a linear mixed effects (LME) model and generalized linear mixed model (GLMM) were also employed to measure the influence of the interaction between the intervention group and pre-measure and post-measure on participant questionnaire scores for each TPB construct and participant behavior, respectively. For both models, participant ID and education level were also included as random effects.

| RESULTS
Of the 150 participants that took part in this study, 146 individuals completed the full experiment and were therefore included in the final sample, with 68 participants placed in the control group and 78 in the treatment group. Most participants were aged between 18 and 25 (44%) and the sample was skewed towards females (67%) ( Table S2). When comparing the two groups using Standard Mean Differences, education level was found to have a medium mean difference (SMD = 0.5-0.3) between the control and treatment groups, with more participants with university qualifications in the treatment group than the control (Table S2). This factor was therefore included alongside participant ID as a random effect within our mixed models.
Within the treatment group, scores for each TPB constructs (knowledge, attitude, subjective social norms, and perceived behavioral control) increased from pre-intervention to post-intervention. However, within the control group, participants' knowledge and perceived behavioral control scores were not found to change post the intervention ( Figure S1). Our analysis found participant's knowledge to be significantly positively influenced by the interaction between the intervention group and pre-poststage of the questionnaire (LME model, F = 7.1, 95% Cis [−1.7, 11.1], p < .01). Participants in the treatment group were 3.9 times more likely to increase their knowledge scores post the intervention than participants in the control group (odds ratio testing, odds ratio = 3.9, 95% CI [1.81, 8.51]) (Figure 2).
Despite results from odds ratio testing indicating a higher likelihood of participants perceived behavioral control and attitudes scores increasing post the treatment intervention than the control (Figure 2), the LME model did not find any of these constructs to be a significantly positively influenced by the interaction between the intervention group and pre-post-stage of the questionnaire (F = 17.2, 95% CI [4.7, 13.2], p < .1; F = 3.1, 95% CI [−8.7, 0.4], p < .1, respectively).
A total of 117 participants completed both the before and after revealed preference choice tests (54 participants from the control and 63 participants from the treatment group). Across both the treatment and control groups, before the intervention 63% of participants chose plastic packaging (Figure 3). Of the participants that displayed a behavior change from pre to post the intervention, 40% of participants within the treatment group went from choosing plastic to paper packaging compared to 28% within the control group (Figure 3). However, the log odd estimate of participants choosing paper over plastic was not found to be significantly influenced by the interaction between the intervention group and the pre-post-stage of the intervention within our GLMM (log odds estimate = −0.6, 95% CI [−1.8, 0.4], p = .2) ( Table S4).
A final sample of 72 individuals participated in the four-week follow-up choice experiment, a response rate of 53%. Thirteen participants were removed from the final count due to incomplete pre-intervention and postintervention choices creating a final sample of 59 (31 in the treatment group and 28 in the control). Due to this small sample size, we were only able to use descriptive statistics on this dataset. Of these participants, 46% were found to have retained their post-intervention revealed F I G U R E 2 Odds ratios of participants' questionnaire scores increasing post exposure to the treatment intervention for each TPB construct assessed. The dashed line indicated an odds ratio of 1.0. The odds ratios were calculated using odds ratio testing. TPB, theory of planned behavior preference, and 54% were found to have changed theirs. We found that a higher proportion of participants in the treatment group changed back to choosing plastic over paper packaging after four weeks (45%) compared with the control group (32%).

| DISCUSSION
In addition to the adoption of government and industry level regulations, efforts to combat plastic pollution also rely on the collective actions of individuals reducing their consumption of single-use plastics (Jambeck et al., 2015). Therefore, in order for interventions to be successful, they must be wide-reaching but also effective in targeting demand reduction behaviors . As such, an important aspect of evaluating mass media interventions, and documentaries in particular, is to ensure that reach is distinguished from impact (Jones et al., 2019;Veríssimo et al., 2018). In the case of Blue Planet II, the series reached millions of viewers, but this offers no insight or guarantee into behavior changes taking place.

| Behavior change
Our results could not establish the behavioral influence of the conservation messaging in Blue Planet II on an observed change in revealed preference towards plastics among the viewers studied when compared with our control intervention. This finding is counter to common assertions that the series' messaging created a "Blue Planet Effect" in reducing plastic consumption behaviors F I G U R E 3 The proportion of participants (%) within the control and treatment groups that chose plastic pre and post the intervention, including 95% confidence intervals (Hunt, 2017), but is the first to use a robust experimental design as well as measure of observed behavior to test this hypothesis. Currently, there are few robust evaluations of nature documentaries (Thomas-Walters et al., 2019), but the best available evidence broadly supports our main findings. For example, Nolan (2010) found that despite the documentary An Inconvenient Truth increasing viewer's concerns about global climate change, this did not lead to a sustained adoption of behaviors to reduce greenhouse gas emissions. Similarly, the nature documentary series Planet Earth 2 was found to increase species awareness and stimulate engagement among audiences, however this was not found to lead to proactive actions such as donation behaviors (Fernández-Bellon & Kane, 2019). Both studies also highlight that awareness means very little in the context of behavior change.
Furthermore, mass media interventions such as Blue Planet II are vulnerable to rapid adoption and abandonment (Mascia & Mills, 2018), and it is therefore important to consider the future sustainment of any immediate impacts on behaviors. The findings from our follow-up choice experiment were consistent with those of previous studies which have reported high relapses in any immediate positive effects of an intervention (Howell, 2012). However, our low sample size cannot offer a statistically powerful conclusion to these findings. Additionally, as a consequence of using photographs of snack choices found at a cinema in order to fit the context of this follow-up choice, the saliency of the plastic and paper packaging options across these choices may have been diminished. Thereby any signals indicating a sustained influence of Blue Planet II on plastic preference behaviors may have been truncated by this factor alone.
The benefits of the highly controlled setting used in RCT, which limits the influence of confounding variable, can also double as a limitation of this methodology in reducing its real world replicability (Jadad & Murray, 2007). In this study, we controlled for factors such as how many episodes of Blue Planet II participants watched and who they watched this with, which may be counter to how viewers would have typically experienced the series. It is therefore important to recognize these limitations as well as the context of this experiment when further applying the results of this research.

| Knowledge and attitudes
Despite the lack of evidence to support Blue Planet II as a behavior change intervention, our analysis did reveal that exposure to the show had a significant influence on increasing viewers' knowledge of environmental issues. Although our results demonstrate the value of increased conservation messaging in media creating a far-reaching environmental education opportunity (Barbas, Paraskevopoulos, & Stamou, 2009), it also postulates that understanding alone cannot drive action (Kollmuss & Agyeman, 2002). This conclusion is supported by previous research (Abrahamse et al., 2005;Howell, 2012;Janpol & Dilts, 2016), which argues that interventions solely focused on information deficit are over-simplistic (Kollmuss & Agyeman, 2002), and is in-line with the TPB model which hypothesizes that multiple influences beyond understanding are necessary for a behavior to be implemented (Ajzen, 1991).
Our results did not support previous studies that found environmental documentaries to have an overall positive influence on individual attitudes towards the environment (Janpol & Dilts, 2016). Perhaps this is because study participants had pre-existing high levels of environmental attitude, with average pre-intervention scores of 83% across both groups. These high baselines could have resulted from self-reported bias of participant's scoring their own attitudes , or a self-selection bias of the sample whereby those with already positive attitudes about the environment were interested in taking part in the study (Howell, 2011). As a result of these pre-existing positive environmental attitudes across both experimental groups, any difference that was observed in the treatment group as a result of the intervention could have been underestimated. It is therefore not possible to assume our findings would be true of an audience more similar to the wider UK population, limiting the study's external validity.
Although the TPB has previously been found to be a good framework for predicting pro-environmental behavior change (Sun et al., 2017), additional factors could exist outside of this model that also influence individuals' preference towards choosing plastic over paper. For example, Triandis' theory of interpersonal behavior incorporates the function of a habitual response in addition to social and affective factors in influencing a given behavior (Triandis, 1980), arguing that habits are mediators of behaviors. Although this model has not been as widely used as the TPB model, where it has been applied, studies have found the addition of habits to have an increased explanatory value, for example in the study of food waste behaviors (Russell et al., 2017) and car use (Bamberg & Schmidt, 2003).

| Wider impacts
Although individual change is an important measure of intervention impact, research into the evaluation of films and documentaries have called on expanding beyond this to measure wider societal influences (Karlin & Johnson, 2011). In the case of Blue Planet II there is evidence that the series resulted in much media attention and increased conversation around the issue of marine plastics, which may have led to upstream changes in allowing the topic of marine plastics to become more salient and therefore creating a window of opportunity for policy change. For example, the series was referenced in a speech launching the UK governments' 25 year environmental plan by former Prime Minister Theresa May, in which she praised the show for "vividly highlighting" the problem of ocean plastics (May, 2018). Within this environmental plan, the UK government committed to taking increased legislative action to tackle plastic pollution. This included the extension of a charge on using plastic carrier bags, which has been credited with a decrease in their use by 83% (HM Government, 2018) as well as the setting of an ambitious target to eliminate avoidable plastic waste by 2041 (HM Government, 2018). Additionally, past research has also recognized interpersonal communications to be important motivators in the adoption of new behaviors (Green et al., 2019). Therefore, subsequent conversations about plastic consumption and pollution triggered by Blue Planet II could still have influenced a change in behavior, but this context was not something that we accounted for within the experiment.
Previous research has found that an integration of different intervention strategies may be the most effective way to sustain behavior change in an environmental context (Salazar, Mills, & Veríssimo, 2019). Therefore, further studies into the impact of documentaries and other mass media interventions should consider both measures of individual behavior change as well as broader societal changes, for example, at the interpersonal or government policy level, which could provide a more holistic and nuanced understanding of impact.

| Conclusion
Despite the pressing issues of marine plastics and its link to individual's behavioral choices, there is currently a lack of empirical research on how interventions can be used to effectively target plastic consumption behaviors (Hartley, Thompson, & Pahl, 2015). This study offers an important example of how to utilize and apply impact evaluation methods to better understand intervention impacts on proenvironmental behavior change. As nature documentaries become an increasingly popular television genre (Koblin, 2020), the resurgence of these programs has also seen a change in narrative for the shows, shifting their focus towards more conservation themes (Jones et al., 2019). This study is the first to present an understanding of the impacts of this narrative shift on viewers pro-environmental behaviors using a robust experimental design. We call on researchers to further develop and deliver impact evaluation research focused on nature documentaries and other mass media interventions concerning biodiversity.