The objective of this study is to determine the most effective PHWL content for Indonesia, including similarities and differences with research conducted in other countries. This study is also aimed to assess social and psychological factors that may influence PHWL impacts. The pattern of results from this study indicates that
Indonesians respond to PHWLs in ways that are comparable to findings from previous studies in MICs and HICs. In particular, the results indicate that PHWLs work better than the text only warnings, as has been found across social cultural settings (e.g., Cantrell et al., 2013; Fathelrahman et al., 2010; Hammond et al., 2012; McQueen et al., 2015; Nan, Dahlstrom, Richards, & Rangarajan, 2015; Noar et al., 2015; Rousu & Thrasher, 2014; Thrasher et al., 2012; Thrasher et al., 2012; Veer & Rank, 2012; Huang et al, 2016). In terms of PHWL content, PHWLs with graphic imagery evoked the strongest negative emotional responses, were rated as having the highest credibility and were perceived as the most effective, followed by PHWLs with suffering imagery, and, lastly, by PHWLs with symbolic imagery. These results are also consistent with findings from previous studies across cultural and socioeconomic contexts (Anshari et al., n.d.; Berg et al., 2011; Fathelrahman et al., 2010; Gravely et al., 2014; Hammond et al., 2012, 2004; Thrasher et al., 2010, 2007; Thrasher, Arillo-Santillán, et al., 2012; Thrasher, Carpenter, et al., 2012; Huang et al, 2016). In terms of textual type used in the warnings, we used a between-
from previous studies (Hammond, Thrasher, et al., 2012; Thrasher et al., 2012). This design allowed for a more controlled, systematic evaluation of this characteristic of warning content and had not been done in prior research. However, we found no significant difference between the use of didactic text and the use of brief testimonials, except that the didactic text seemed to work better among adolescents.
For the social psychological factors affecting the warning reactions, we found that smoker identity only negatively influence how participants perceived the effectiveness of PHWLs, while advertisings exposure, self-efficacy, and reactance were found to have positive effects on all outcomes. Such positive effects of advertisings exposure and reactance for two outcomes were in the opposite direction than what we expected. As for our moderation hypotheses, we found no support for moderation of smoker identity. Partial support for moderation by self-efficacy were found when examining association between image style and negative emotional responses. Results around moderation pointed in the opposite direction than we hypothesized when examining advertising exposure on message credibility and perceived effectiveness, and when examining reactance on negative emotional responses and perceived effectiveness. It is important to note that the pattern of relationships between the outcome variable and HWL image styles were the same across levels of the moderating variable: Graphic PHWLs has the strongest responses on all outcomes, followed by suffering, symbolic, and the text-only HWLs. Moderation did not change these relationships, but made them more or less evident. In general, our findings are consistent with the opinion that the most effective HWLs include graphic and suffering imagery and that smokers who may be more resistant to quitting do not appear to require different messaging strategies.
Interpretation of our findings should consider some limitations of this study. First, the study design involved a one-time, forced-exposure to our study stimuli, where
participants were shown the warnings on a computer screen. This approach may not adequately represent exposure in real life settings, which involves repeated exposure of up to 20 times a day for a pack-a-day smoker. However, our study results are largely consistent with previous studies using mock cigarette packs (Thrasher et al., 2012) and those that show the predictive validity of the approach that we used when determining which PHWLs perform best in pre- and post-market studies (Huang et al., 2016). Second, self-reported ratings might produce bias, although we do not expect the bias will be systematic across key areas of inquiry. Previous research has also shown that similar self-reported ratings of PHWLs produced a pattern of results for the type of PHWL imagery we examined that is generally consistent with the pattern of brain activity in visual and emotional processing (Newman-Norlund et al., 2014). Third, the use of between subject design for the textual condition may be less sensitive for assessing our testimonial manipulation because the large image is very prominent compared to the text, especially when the display size was relatively small (i.e., 40% of cigarette pack). Short version of testimonial text may need to be carefully tailored to better correspond with the image in PHWL.
Lastly, our convenience sample may not be representative of adult smokers and adolescent in Indonesia. Therefore we used intercept techniques to minimize selection bias and obtain a relatively heterogeneous sample, which has been shown to have external validity (Huang et al., 2016).
Recommendation
Based on our findings, we recommend the use of graphic and suffering imagery for PHWLs as they appear most effective for adult smokers and adolescents. As for the textual strategy, we recommend the use of didactic texts as they appear most effective among adolescents, although the use of testimonials may be more effective for adult smokers. Future research may need to more systematically examine types of testimonials, more carefully tailor the text to better fit with image, and may require more elaboration to make testimonials be more compelling. Furthermore, this type of strategy may become more effective after populations are exposed to PHWLs on packages in real life, and our findings have been considered by the Ministry of Health Republic of Indonesia for the selection of new content for the second rounds of PHWLs in Indonesia.