3. Results
A total of 1335 participants completed the online experiment, of the 2219 sampled, for a 60% completion rate. The experiment consisted of two conditions (green tea–cancer/yukichi fruit juice–gastrocordialis) and seven QHC groups. Each group viewed one QHC in one of two conditions; seven QHCs were about the green tea–cancer relationship (n = 669) and the remaining seven were identically worded QHCs about the yukichi fruit juice–gastrocoridalis relationship (n = 666).
The majority of participants were between the ages of 55 and 74 years old (
n = 1123, 89.1%), White (
n = 1060, 79.4%), who held a high school degree or more (
n = 1233, 92.4%), with a household income under
$100,000 (
n =1028, 77.0%). Given the random assignment of participants, there were no differences between conditions or across groups by race and ethnicity, age, education, employment, household income or incidence of breast or prostate cancer (
Table 4).
Overall, the majority of participants reported that they were in good health (n = 1010, 76%) and had never received a cancer diagnosis from their doctor (n =1132, 84.8%). Participants were “a little” or “somewhat” worried about their health overall (n = 929, 69.9%) and fewer worried about becoming sick with cancer (“somewhat” n = 441, 39.2%; “not at all” n = 296, 26.3%). However, most had made a dietary change in the past year (n = 1,022, 76.8%) due to a health-related concern. Slightly more than half of the participants reported that they drink green tea (n = 691, 51.8%), and most claimed to drink it because they enjoy its taste (n = 215, 62.5%). Of the respondents who reportedly consumed green tea in the past year (n = 345), fewer than 10% drank it to reduce their risk of cancer (n = 32, 9.3%). The majority of the participants believe they are informed about diet and health (i.e., perceived nutrition knowledge) (n = 1298, 97.2%) and consider health claims important on food and dietary supplement product labels (n = 1168, 88.0%; n = 926, 91.5%, respectively). No differences were found between groups or conditions with respect to general health, perceived nutrition knowledge, green tea consumption in the past year, worry about cancer, or health-related dietary changes.
Overall, participants rated the level of evidence as a “2” (i.e., D-grade) in both conditions (
Table 3), and when the 13-point scale of evidence was collapsed into four grades of evidence, the results are consistent with a D level of evidence [
53]. More than half of the participants were not at all confident in the ability of green tea/yukichi fruit juice to reduce the risk of cancer/gastrocoridalis (
n = 709, 53.5%) (Mean (M)= 1.83, Standard Deviation (SD)= 1.10) but in both conditions, nearly two-thirds (
n = 757, 57%) think there would be at least a slight risk reduction for cancer/gastrocoridalis (M = 1.98, SD = 1.10) if they drank the beverage. Most also think they would need to drink green tea/yukichi fruit juice once a week or more (
n = 601, 80%) to reduce their risk. However, most would not buy green tea/yukichi fruit juice with the QHC (
n = 755, 57.1%).
The majority of participants in the green tea condition reported that they were familiar with the green tea–cancer relationship (n = 343, 51.9%) but had never seen the QHC on a label of a green tea product (n = 480, 73.5%), in an advertisement or on a website or in an article (n = 453, 70.9%).
3.1. Purchase Intentions for Green Tea and Yukichi Fruit Juice
The diet–disease condition had a significant effect on perceptions of confidence, evidence, risk reduction, and purchase intentions (Pillai’s Trace = 0.102, F = 36.896, df = 4, 1296, p < 0.0001). A statistically significant effect was also found by QHC group on perceptions of confidence, evidence, risk reduction, and purchase intentions (Pillai’s Trace = 0.138, F = 7.719, df = 24, 5196, p < 0.0001) as well as a significant interaction between the condition and group on the same variables (Pillai’s Trace = 0.034, F = 1.834, df = 24, 5196, p < 0.0001).
However, these measures are strongly correlated (
Table 5). A significant main effect of condition demonstrated that participants had greater purchase intentions for green tea than yukichi fruit juice
F(1, 1299) = 132.320,
p < 0.0001. There was also a significant main effect of QHC, such that claims written by Fleminger produced greater purchase intentions than claims written by the FDA,
F(6, 1299) = 8.047,
p < 0.0001. However, no interaction was found between condition and group,
F(6, 1299) = 1.713,
p = 0.114.
Significant differences were found between condition, with participants perceiving greater evidence for the green tea–cancer relationship than for the yukichi fruit juice–gastrocordialis relationship, F(1, 1299) = 25.407, p < 0.0001. There was also a significant main effect of the QHC group, with greater perceptions of evidence among claims written by Fleminger than those written by the FDA, F(6, 1,299) = 25.491, p < 0.0001. Irrespective of the QHC, participants reported greater evidence for the green tea–cancer relationship than for yukichi fruit juice and gastrocoridalis, suggesting there are other factors, in addition to the measures in the current study, that contribute to their perceptions. There was also a significant interaction between condition and QHC group on evidence ratings, meaning that participants who viewed a yukichi fruit juice–gastrocoridalis QHC rated the evidence lower than those who viewed a green tea–cancer QHC, regardless of the author, F(6, 1299) = 2.804, p = 0.10.
Similarly, there were main effects of both condition and QHC group on perceptions of disease reduction. Participants perceived greater reductions in cancer risk resulting from drinking green tea than gastrocoridalis from drinking yukichi fruit juice, F(1, 1299) = 36.369, p < 0.0001. Groups that viewed a Fleminger QHC also perceived greater reductions in disease risk than those who viewed a QHC written by others, F(6, 1299) = 19.263, p < 0.0001. A statistically significant interaction was also found between conditions and QHC groups on perceived risk reduction, F(6, 1299) = 3.085, p = 0.005. Participants’ perceptions of gastrocoridalis risk did not differ among yukichi fruit juice QHCs, while the perceived cancer risk was different among green tea QHCs.
Finally, participants in the green tea condition were significantly more confident in the claimed relationship than those in the yukichi fruit juice condition, F(1, 1299) = 78.922, p < 0.0001. A main effect was also identified between QHC groups. Participants who viewed a Fleminger claim reported greater confidence in the claimed relationship than in other groups, F(6, 1,299) = 16.919, p < 0.0001. There was also a statistically significant interaction between QHC group and condition. Participants’ confidence in the green tea condition was greater regardless of the QHC, whereas in the yukichi fruit juice–gastrocoridalis condition, participants were overall less confident, F(6, 1299) = 2.917, p = 0.008.
There were no statistically significant differences between the green tea and yukichi fruit juice conditions for dietary supplement use (p = 0.993), reported importance of health claims on dietary supplement or food labels (p =.505; p = 0.138, respectively), perceived nutrition knowledge (p = 0.837), worry about overall health (p = 0.979) that led to a dietary change in the past year (p = 0.360), worry about becoming sick with gastrocoridalis (p = 0.240) or cancer (p = 0.066), or a previous diagnosis of gastrocoridalis (p = 0.800) or cancer (p = 0.677).
3.2. Predictors of Purchase Intentions for Green Tea
To determine the factors’ predictive value for purchase intentions of green tea, consumer-specific and sociodemographic variables were individually tested for the green tea condition only (n = 666). The significant sociodemographic predictors are age and race/ethnicity and account for 3.5% of the variation in purchase intentions for green tea, F(5, 656) = 5.857, p < 0.0001. Participants reported lower purchase intentions with each incremental increase in age, b = −0.090, t(661) = −2.334, p < 0.05. Black (b = 0.126, t(656) = 3.274, p = 0.001). Hispanic (b = 0.141, t(656) = 3.670, p < 0.0001) participants reported statistically significantly greater purchase intentions for green tea than White participants.
The predictive consumer specific variables accounted for 4.1% of the variance in purchase intentions and included: Dietary supplement use and the perceived importance of health claims on food products and on dietary supplement products, F(3, 501) = 8.162, p < 0.0001. For each incremental increase in dietary supplement use in the past year, participants reported that they are less likely to intend to purchase green tea, (b = −0.114, t(503) = −2.582, p = 0.01). Yet, the more important participants consider health statements on dietary supplement products, the greater their purchase intentions (b = 0.135, t(501) = 2.209, p < 0.05).
Measures of perceived susceptibility that significantly predict green tea purchase intentions are: General health, worry about overall health, worry about becoming ill with cancer, and worry that led to a dietary change in the past year, F(4, 548) = 8.876, p < 0.0001. The better participants rate their general health, the greater their purchase intentions for green tea, b = 0.126, t(548) = 2.722, p < 0.0001. Similarly, participants who are more worried about their overall health (b = 0.105, t(550) = 2.228) or about becoming sick with cancer (b = 0.103, t(549) = 2.262) demonstrated greater purchase intentions, p < 0.05. Participants have greater purchase intentions with each incremental increase in dietary changes made in response to a health worry in the past year, b = 0.212, t(548) = 4.423, p < 0.0001. Non-explanatory predictors were removed from further analysis; sex, education, employment, income, perceived nutrition knowledge, and past cancer diagnosis.
3.3. Model 1
A block of the modifying sociodemographic, consumer-specific, and perceived susceptibility predictors was entered as independent variables into a regression model to predict green tea purchase intentions. These variables included: Race/ethnicity, age, perceived importance of health claims on food products and on dietary supplement products, dietary supplement use, and self-reported health status, worry about overall health, worry about becoming ill with cancer, and worry that led to a dietary change in the past year. The model significantly predicted purchase intentions for green tea and accounted for 9.3% of the variance, F(12, 406) = 4.581,
p < 0.0001 (
Table 6).
Participants who drank green tea in the past year indicated significantly greater purchase intentions (
M = 2.73,
SD = 1.49) than those who did not,
Welch’s F(1, 640.890) = 87.810,
p < 0.001. By contrast, participants who had not consumed green tea in the past year indicated that they are only “slightly likely” (
M = 1.76,
SD = 1.17) to buy it in the future, despite having seen a QHC about the green tea–cancer relationship. As a result, current consumption was added to the model at step 2, with a resulting significant R
2 change = 0.052,
F(13, 403) = 6.404,
p < 0.0001, adj. R
2 = 0.144 (
Table 6).
There were no differences between consumers who drank green tea in the past year and those who did not in terms of whether they had seen the claim on a label,
Welch’s F(2,114.967) = 2.826,
p = 0.063, website, advertisement, or article,
Welch’s F(2, 148.997) = 2.270,
p = 0.107. Yet those who drank green tea are significantly more familiar with the green tea–cancer relationship (M = 2.11, SD = 1.10) than those who did not (M = 1.61, SD = 0.87),
Welch’s F(1, 642.179) = 42.389,
p < 0.0001. Accordingly, familiarity with the green tea–cancer relationship was added to the model at step 3. Results showed that familiarity with the green tea and cancer is a significant predictor of purchase intentions for green tea,
F(14, 401) = 7.685,
p < 0.0001, and accounted for 18.4% of the variance in the model, R
2 change = 0.041 (
Table 6). At step 4, we then entered the QHCs, which were significant predictors of future behavior,
F(20, 395) = 6.263,
p < 0.0001, though they only accounted for a small amount of variance in the model, adj. R
2 = 0.202, R
2 change = 0.029 (
Table 6). In the full model, the significant predictors of purchase intentions for green tea are: The perceived importance of health statements on dietary supplements, worry about health that led to dietary change in the past year, having consumed green tea in the past year, familiarity with the diet–disease relationship, and exposure to a QHC written by Fleminger. Dietary supplement use is negatively related to purchase intentions for green tea.
3.4. Model 2
To understand if perceptions of evidence for the green tea–cancer relationship, risk reduction for cancer, and confidence in the claimed relationship predict purchase intentions for green tea, three additional steps were added to the model: Ratings for evidence, risk reduction, and confidence in the green tea–cancer relationship. Next, a second model was created to isolate the effects of these variables from the QHCs and included steps 1 (i.e., sociodemographics, consumer-specific, and perceived susceptibility variables), 2 (i.e., past behavior), and 3 (i.e., familiarity), and the dummy-coded QHCs were removed. The following analysis represents steps 4, 5, and 6 for the second model.
Perception of evidence was found to be a strong and significant predictor of purchase intentions for green tea, F(15, 399) = 19.781,
p < 0.0001, adj. R
2 = 0.315, R
2 change = 0.128,
p < 0.0001. Confidence in the green tea–cancer relationship is also a significantly strong predictor of future intentions to purchase green tea, F(16, 398) = 16.345,
p < 0.0001, adj. R
2 = 0.372, R
2 change = 0.056,
p < 0.0001. Entry of the third variable, perception of cancer risk reduction, showed that the overall model was significant for predicting purchase intentions for green tea, F(1, 400) = 18.736,
p < 0.0001, although the variable did not add any more to the model, adj. R
2 = 0.371, R
2 change = 0.000,
p = 0.896 (
Table 6).
However, evidence, risk, and confidence are multicollinear and demonstrate strong positive correlations (
Table 5), which likely explains the weak predictive value of perceived risk reduction on purchase intentions. Therefore, a backwards stepwise regression model tested the three predictors in a block. The model removed risk reduction and maintained evidence perceptions and confidence in the green tea–cancer relationship, indicating that perceived disease risk reduction did not account for any additional variance and so was removed.
The two significant predictors were re-entered into the model (i.e., perceived evidence, confidence) at Step 4. As expected, this block was a strong predictor of purchase intentions,
F(16, 398) = 16.345,
p < 0.0001, adj. R
2 = 0.372, R
2 change = 0.185 (
Table 6). For every incremental increase in perceived evidence, there was a 0.130 increase in purchase intention for green tea, and for every incremental increase in confidence in the green tea–cancer relationship, there was a 0.369 increase in purchase intention. While dietary supplement use remained negatively related to purchase intentions, the importance of health statements on supplements, worry that led to dietary change, past green tea consumption, and familiarity with the claimed relationship were predictive of purchase intention (
Table 6).
Neither dose nor taste were tested in a predictive model, since no association was detected between perceived dose necessary to achieve risk reductions and purchase intentions, χ2(24) = 29.845, p = 0.190 or with the taste of green tea, χ2(6) = 7.284, p = 0.295.
To summarize, the significant individual predictors for purchase intentions were entered into the models at step (1) and include: Sociodemographics (race/ethnicity, age); consumer-specific variables (dietary supplement use, importance of health claims on food and supplement labels); and perceived susceptibility (general health, worry about health, worry about cancer, worry that led a dietary change). Model 1 significantly predicted purchase intentions and included steps (2) green tea consumption, (3) familiarity with green tea and cancer relationship, and (4) the seven QHCs, F(16, 400) = 7.841, p < 0.0001, adj. R2 = 0.208, R2 change = 0.028. Model 2 included steps (1–3), removed step 4 (i.e., QHCs from Model 1) and replaced it with (4) evidence perceptions and (5) confidence in the green tea–cancer relationship, which led to a statistically significant increase in R2 of 0.185, F(13, 401) = 20.225, p < 0.0001, adj. R2 = 0.376.
4. Discussion
With respect to their future intentions to purchase green tea, the data suggest that older consumers are influenced by their past behaviors and preferences. However, even after controlling for those preferences and experiences, exposure to the claims written by Fleminger in 2004, 2008, and 2010 demonstrated greater intentions to purchase green tea than other claims. Our results also demonstrated that some QHCs increase consumer perceptions of evidence for the diet–disease relationship, which affects their confidence in the relationship and purchase intentions for green tea.
Research has identified several sociodemographic characteristics that are associated with greater acceptance of functional foods with health claims. We expected some overlap between these same characteristics of consumers who are receptive to health claims, and QHCs.
In the current study, race or ethnicity was a strong predictor of purchase intentions for green tea when entered at step one in the regression model. Black and Hispanic consumers intended to purchase green tea more than White consumers. This finding is in contrast to previous research, which has pointed to White adults as more accepting of functional food products [
40,
42,
43,
44,
55] although this difference may be explained in part by the presence of a QHC which is unique to the current study. However, race or ethnicity did not remain a significant predictor of purchase intentions for green tea in the full regression model(s).
Other research has indicated that women are more receptive to functional foods and to nutrition information in health claims [
37,
42]. However, the current study did not find sex to significantly predict purchase intentions for green tea. Further, no other sociodemographic variables demonstrated predictive value including education, employment, or income.
Previous research also demonstrates mixed results for age as a predictor of functional food acceptance with health claims [
40,
42,
43,
44]. For example, one study found that functional foods were accepted by women 35–54 years old [
42], whereas another study identified a greater range of accepting adults, 30–70 years old [
41]. Still, two other studies did not find any relationship between age and functional food acceptance [
42,
44].
Our study initially indicated age as a strong predictor for green tea purchase intentions in adults aged 55–64 but not for those over 65 years old. However, in the full model, age was not a significant predictor of purchase intentions. More than half of our sample already drinks green tea for the taste, not to reduce their risk of cancer. While there is strong evidence that taste is a predictor for purchase intentions and behavior related to health claims [
38,
52], it was not correlated with purchase intentions for green tea after viewing a QHC.
Purchase intentions for yukichi fruit juice were significantly lower than for green tea. Participants could not have had prior experience with yukichi fruit juice or gastrocordialis due to their fictitious or novel nature. Therefore, consumers considered other information when responding to questions. In addition, existing green tea consumption and familiarity with the green tea–cancer relationship were both significant predictors of behavioral intentions. These findings are consistent with previous research. Familiarity and experience with a product and/or a health condition have been found to significantly predict consumer behaviors [
12,
38,
42,
47,
48,
49], including purchase intentions for functional foods.
Dietary supplementation was negatively associated with the intention to purchase green tea, after viewing a QHC. This is in contrast to other research that indicates consumers who regularly take supplements are more accepting of functional foods [
55], such as green tea [
56].
Claim language is regarded as a predictor of purchase intentions for functional products with health claims [
38], and our results support this theory. While the results of the current study cannot isolate the language or wording in QHCs that may have led to greater purchase intentions for green tea, the results show that QHCs written by Fleminger led to greater purchase intentions compared with other QHCs. However, the FDA concluded that the claims written by Fleminger overstate the level of scientific support for the green tea–cancer relationship [
19,
21]. Our results provide support for the agency’s conclusion, demonstrating that these same QHCs produced greater perceptions of evidence and greater intentions to purchase green tea, compared with the other claims. The 2004 Fleminger QHC indicates, “There is scientific evidence supporting this health claim although the evidence is not conclusive”, while the 2008 and 2010 Fleminger claims suggest the evidence is “credible but limited”. Further, the 2010 claim identifies the FDA as the reviewer of evidence which, altogether, aimed to increase the trustworthiness of the claim. The 2005p and 2005b QHCs reference the number and quality of studies behind the claimed relationship which is similar to other diet–disease claims that the court found to be too technical for consumers to understand [
20]. Although technically accurate, the claims are not consumer-friendly. Since QHCs were required by case law, it is possible that the agency sought to protect the consumer from being misled about the health value of a product by drafting overly technical claims [
20,
21]. The 2011 claim, which states that the FDA does not agree with the claimed relationship, provides insight into the agency’s perspective [
21]. The 2012 FDA QHC suggested by the federal court as a compromise between companies’ abilities to market the health benefit of green tea and the FDA’s mission to prevent consumer confusion about scientific certainty for the relationship [
21] led to greater perceptions of evidence than the FDA QHCs in 2005p/b and 2011. The 2012 QHC also produced greater purchase intentions for green tea than the FDA QHCs, but the results were not statistically significant, which suggests the current QHC may strike a balance between interests of making profit for companies and the interest of the FDA in preventing consumer confusion.
The reported confidence in the diet–disease relationship was a very strong predictor of purchase intention. On its own, the perceived extent of the reduction in the risk of cancer was associated with intentions to purchase green tea. However, it did not contribute additional predictive value to the model of purchase intentions after accounting for the measure of confidence in the green tea–cancer relationship. Our results support the hypothesis that specific claims differentially affect the perceived health benefits of green tea products since perceptions of evidence and confidence in the green tea–cancer relationship were associated with variations in the QHC.
The overall proportion of variance in purchase intentions that can be explained by the independent variables was nearly 40% (
Table 6). There are several additional measures in the HCF and HBM that were not included in the current study and so, it is possible that the addition of these measures could lead to models with greater predictive power of purchase intentions for green tea with a QHC.
There is considerable potential for QHCs in the market, since most US consumers believe that functional foods can reduce the risk of becoming sick with a specific disease and believe they have some control over their personal health [
57]. Participants in our sample of older consumers had greater intentions to purchase green tea if in the past year, they worried about their health and if their worry led to a dietary change. Indeed, more than three-quarters of our sample changed their diet to address health concerns. Since US consumers perceive cancer as a greater risk than other diseases, including heart disease, type 2 diabetes, and stroke [
58], there is likely a considerable marketing opportunity for products associated with reducing the risks of cancer. Future research could examine the complexities of QHCs through a more interdisciplinary lens. More specifically, collaborating with market researchers could expand the application of our results to general consumer behavior and consider its place within the theory of consumption.
However, it should be noted that the current study examines QHCs for the green tea–cancer relationship, a D-grade claim, which meets the lowest level of evidence for a diet–disease relationship [
19]. Future study might examine science communication in B and C grade QHCs to gain greater understanding about existing claims as well as how federal lawsuits may impact the language of QHCs, which ultimately affects consumers. Although claims do not necessarily distinguish categorically healthy versus unhealthy foods [
59], the perceived health quality of products by consumers may yield different perceptions of evidence and purchase intentions [
60] with QHCs.
Strengths and Limitations
Few recent studies have examined how consumer perceptions change in response to different QHCs. This study is unique in its contribution to the understanding of the influence of QHCs on purchase intentions. Only one other study has explored purchase intentions with respect to product labels with QHCs but was limited in that the format tested has never been allowed for use on labels [
12]. A major strength of this study is that it included QHCs written by different stakeholders, including Fleminger, the federal court, and the FDA, many of which have been used on labels or on the Internet.
One potential limitation of the study is that it did not include a control group that did not view any QHC. However, the existence of QHCs associated with green tea and a significant degree of prior familiarity with the green tea–cancer relationship among the participants made inclusion of a control group problematic. Instead, prior familiarity with the green tea–cancer relationship and with the QHC seen by the participant was used as a within-subjects control variable, creating a more ecologically valid design. That is, those who were unfamiliar with the relationship and had no prior familiarity with the QHC to which they were exposed serve as comparisons for those who were familiar with the relationship and/or the QHC. In addition, the study aimed to evaluate the differences in consumer perceptions among actual QHCs proposed or used by stakeholders and to test the assertions made by the various stakeholders about those claims. Unfortunately, members of a traditional control group would be unable to respond to the key questions about the level of perceived evidence and confidence or cancer risk reduction, based on a QHC. However, future research could test the differences between a control group and the current green tea QHC that is enforced by the FDA.
This study used a 2 × 7 between-subjects design. An alternative within-subjects study design was tested in which participants viewed several QHCs and responded to the same battery of questions. However, the time to complete the survey was considerably lengthened, threatened primacy effects, and increased participant burden, and so this design was rejected as impractical.
Finally, the study results are based on self-reported data which, in general, have a low to moderate correlation with observed behaviors [
61,
62]. There are likely differences between self-reported intentions and actual purchases of green tea with a QHC.