1. MARCO REFERENCIAL
1.4. MARCO LEGAL
1.4.1. Estatuto de Vigilancia y Seguridad Privada
Study One has several limitations. The sample in Study One was designed to be representative of a selection of urban cities in China; therefore results may not generalize to the rural Chinese population or to other urban cities not included in the sample. The under-sampling of smokers aged 18-24 is probably due to absence at the time of interview.
Given loss to followup (n=949), the results from our study may be limited by attrition bias, one of the main threats to the external and internal validity of longitudinal cohort studies (Lavrakas, 2008; Schutt, 2012). Those lost to follow-up tended to be younger, have higher educational attainment, and have lower household income. This bias may weaken generalizability of our results to these subpopulations. Furthermore, internal validity may be compromised because study drop-outs may respond differently to campaigns than participants who were successfully followed up. For example, those who were successfully followed up had relatively lower educational attainment, and studies from high-income countries with long histories of tobacco control have found that low SES smokers have stronger responses to graphic, evocative cessation campaigns than smokers from high SES groups (Durkin et al., 2009, 2011; Wakefield, 2011). Hence, we may have overestimated campaign effects due to greater retention of less-well educated
smokers. However, greater campaign effects among lower SES smokers may occur only in societies where smoking has been concentrated in low SES groups, which is not the case in China. Furthermore, when examining the SES indicator of household income, higher income smokers were more likely to be followed up. Hence, our differential retainment of smokers with lower education but higher income makes it difficult to predict the direction of the bias.
The pre-and-post evaluation surveys were not conducted immediately before and after the GCGH campaign because this study utilized surveys from the ITC China Project, which was designed to measure the effectiveness of national-level tobacco control
policies (Wu et al., 2010), rather than the campaign alone. The lengthy period between pre-and-post campaign surveys may have introduced some biases. First, the pre-campaign survey was conducted one year before the campaign was broadcast, which is not optimal given that changes may have taken place between pre-campaign survey and campaign onset. Second, the short duration of the campaign and the lengthy time from the campaign’s end to post-campaign survey (three to seven months) may have missed the maximal impact of the campaign as campaign effects decay (Durkin et al., 2012). Studies show that the beneficial effect of mass media campaigns appears only within two to three months after exposure (Wakefield et al., 2008, 2011). Indeed, this likely helps to explain the low campaign recall rate and the relatively small or non-existent campaign effects. Third, this study is subject to internal validity threats such as history effects due to the lengthy evaluation timeframe. To address the possible influence of these events on study outcomes, our adjusted models included statistical controls for exposure to any anti- smoking campaigns in the last six months and to the SFO campaign which included mass
media campaigns promoting smoke-free environments took place in three of intervention cities, Beijing, Shanghai, and Shenyang, in August 2008. Despite these problems of timing and length, the biases are really conservative since we were able to detect campaign effects.
The potential non-comparability of the intervention cities and control cities such as tobacco industry activities and economic development may also have confounded associations between study variables. Although our regression analyses controlled for measured differences between comparison groups, unmeasured variables may also explain the results. For example, Changsha, a major base for Chinese tobacco industry, had significantly lower percentage of participants who thought that Chinese society disapproves smoking and cigarette gifts than any other five cities. Furthermore,
contamination is a potential internal validity threat since participants in the control cities could have been exposed to the campaign messages outside the city where they live or through satellite TV within their city of residence. Our assessment of campaign exposure within intervention cities helps overcome this limitation, but nevertheless may be limited by recall bias due to measurements of past events (Lavrakas, 2008). Participants had variations in their ability to remember their past experiences or events; thus recall bias may result in inaccurate or misleading results. In spite of these issues, our study is suggestive of campaign effects.
The Study Two has several limitations, including some that stem from the nature of focus group research. Participants in the study do not represent the broader population of 18 to 34 year old smokers, so the results may not generalize to Southern Taiwan or beyond. The sample size was small and may not catch the full range of variability in
participants’ responses to the ads. The lack of power to detect statistically significant differences due to small sample size was further magnified with multiple testing and analyzing participant responses to ads by educational attainment and quit intention. Although educational attainment strata and quit intention strata revealed similar patterns of quantitative responses, this may be a result of lack of statistical power to test responses to different ads across participant types. However, the limited meaningfulness of the quantitative results can be complemented and triangulated by qualitative results, uncovering and elaborating the underlying meaning of quantitative results. Participants self-selected into the study. Their propensity for participating in the study may be correlated with the research topic, causing a self-selection bias in the resulting data (Lavrakas, 2008). One participant expressed his enthusiasm about the study because he had strong opinions on smoking cessation ads broadcast in Taiwan.
Ad ratings involved self-reported responses, which may be subject to bias, and they did so under conditions of forced exposure, which is different from naturalistic campaign exposure. Actual effects on participants’ behaviors or attitudes may be quite different. However, other studies have shown how the perceived effectiveness of ads predicts actual changes in message-targeted attitudes and behaviors (Dillard et al., 2007; Brennan et al., 2013; Bigsby et al., 2013). Furthermore, our follow-up phone survey of ad recall
provided results that were consistent with those from the prior stage.
When comparing one particular characteristics of interest, the other ad
characteristics and demographic characteristics of people featured in the ad were not matched and held constant. The characteristics that were not hold constant for comparisons may as well explain the relative effectiveness of the ads. For example,
demographic characteristics of people may matter more when all other ad characteristics are held constant. Future research should select or create ads by controlling these
characteristics more closely and manipulating one particular characteristics of interest to examine the relative effectiveness of the ads. According to the ELM, an individual with high motivation and ability is more likely to take the central route to process message content (Petty and Cacioppo, 1986). Therefore, when an individual who is less motivated (i.e., without quit intention) takes the peripheral route to process message content,
peripheral cues (i.e., ad characteristics, demographic characteristics of people in ads) become more important in determining the persuasion of messages.
The adaption of the five foreign ads was not entirely consistent in terms of dubbing efforts, which may have impacted comprehension of ads. The decision to not dub the two testimonial ads was to present victims’ original voice and emotions, which could have been made inauthentic through dubbing. Indeed, most smokers commented that they preferred the original versions of the two ads because dubbing would have weakened their emotional impact and, in turn, their effectiveness; however, maintaining the
English-language audio may have impaired some smokers’ understanding of the ads even though Chinese subtitles were provided in the ads. Furthermore, the production value for selected ads was somewhat uneven and this or other ad characteristics for which we did not account may explain the difference in participant perceived effectiveness among ads. Nevertheless, our selection of foreign ads was based on the ad performance that was proven effective in previous studies or in their countries of origin, so our results are consistent with the notion that these ads would work best in Taiwan, as well. The study results also may have been confounded by novelty effects, since some participants had
seen the Taiwanese ads and even some of the foreign ads, whereas other ads were novel. The less novel nature of some ads may help explain their lower reported impact than the other ads which participants had not previously seen. However, the ad that most recalled having seen (i.e., Sponge) was also amongst the most well-evaluated ads.