Correlation test

Next, Social Advertising was coded as 0 = user-generated and 1 = brand-generated.

Advertised Product was coded as 0 = hedonic and 1 = utilitarian. Gender was coded 0 = female and non-binary/third gender and 1 = male. The variable annual household income had option 6 (i.e., ‘Prefer not to say’); this was transformed into a missing case since it is an ordinal variable.

The results of the correlation matrix gave a few interesting outcomes, as can be seen in Table 11. First of all, hypothesis 1 is not significant. Hypothesis 2 is significant since Social Advertising (reference group = brand-generated) is negatively correlated to Affective Appraisal (ρ = -.29). Hypotheses 3 and 4 are significant since Affective (ρ = .55) and Cognitive Appraisal (ρ = .58) are positively correlated to Purchase intention. Hypotheses 5 and 6 are also not significant in the correlation matrix. Social Advertising has a direct negative effect on Purchase Intention (ρ = -.29). The control variables age, gender, purchase experience on Instagram, and Time spent on Instagram gave significant results. Purchase experience on Instagram positively correlates with Purchase Intention (ρ = .28). Gender is negative correlated to Cognitive (ρ = -.12) and Affective Appraisal (ρ = -.18). Purchase experience on Instagram is positively correlated to Cognitive (ρ = .25) and Affective Appraisal (ρ = .28). Time spent on Instagram is negatively correlated to gender (ρ = -.19) and age (ρ = -.16). Purchase experience on Instagram is positively correlated to time spent on Instagram (ρ = .28) and negatively correlated to gender (ρ = -.26).

53 Note. N = 448. ***p < .001, **p < .01, *p < .05. Social Advertising is coded as 0 user-generated and 1 brand-generated. Advertised Product is coded as 0 hedonic and 1 utilitarian. Gender is coded as 1 male and 0 female and non-binary/third gender.

54 A Chi-square Test was performed to check if the manipulation was well executed in the experiment (Table 22, Appendix G2). The test showed that the respondents in the user-generated condition had more difficulties understanding the condition than respondents in the brand-generated condition, this was the same as in the first experiment. However, the results between the manipulation check and social advertising are significant χ² = 179.755, p <.001.

Moreover, based on the φ statistic, the association seems to be of large size (φ = .63).

5.2.8 Hypotheses testing

Hierarchical multiple regression was performed to investigate the ability of Purchase Intention to predict levels of Social Advertising, Affective Appraisal, Cognitive Appraisal, and Advertised Product after controlling for the control variables gender, age, time spent on Instagram, and Purchase Experience on Instagram (Table 24, Appendix G3). Again, the four assumptions should be met. First of all, as can be seen in the scatterplot in Appendix G3, Figure 13, a pattern can be identified. This pattern shows a weak linear relationship between social advertising (independent variable) and purchase intention (dependent variable). Thus, the linearity assumption is met. Next, the Tolerance values for all variables were > .20, and the VIF values < 5 (Daoud, 2018) (Appendix G3, Table 23), which concludes that the residuals are independent of each other. As mentioned before, there is a pattern in the scatterplot. However, the residuals are not wider spread out with larger fitted values, so the homoscedasticity assumption is also met. Lastly, as can be seen in Appendix G3, Figure 12, the data is following the line perfectly, which means the data is normally distributed.

In the first step of the hierarchical multiple regression, the control variables gender, age, time spent on Instagram, and Purchase Experience on Instagram were entered. This model was statistically significant, F(4,443) = 9.81; p <.001, and explained 8.1% of the variance in Purchase Intention. After entry of Affective Appraisal, Cognitive Appraisal, Social

55 Advertising, and Advertised Product in Step 2, the total variance explained by the model as a whole was 46.2%. The introduction of the four variables explained an additional 38.1%

variance in Purchase Intention, after controlling for gender, age, time spent on Instagram and Purchase Experience on Instagram (R² change = 46.2%; F(4,439) = 47.134; p < .001 ). In the final model five out of eight predictor variables were statistically significant, with Affective Appraisal recording a higher Beta value (β = .370, p < .001) than Cognitive Appraisal (β = .366, p < .001), Purchase experience on Instagram (β = .150, p < .001) and Advertised Product (β = -.086, p < .05). This value means that Affective Appraisal has the highest predictive power on Purchase Intention.

Next, the hypotheses were tested using PROCESS model 7 (Tables 12 and 13) by Hayes (2018). The effect of brand-generated Social Advertising on Cognitive Appraisal (a1) is not statistically significant, p = .721, which means that hypothesis 1 is not confirmed in this study.

The effect of Social Advertising on Affective Appraisal a2 = -0.658 means that two consumers that differ by one unit on brand-generated social advertisement are estimated to differ by 0.658 units on Affective Appraisal. The sign of a2 is negative, meaning that brand-generated social advertisements are estimated to have a lower Affective Appraisal. This effect is statistically different from zero, t = 6.611, p < .001, with a 95% confidence interval from -0.854 to -0.463, and is in support of H2.

The effect of b1 = 0.425 indicates that two consumers who experience the same level of brand-generated social advertising but differ by one unit in their level of Cognitive Appraisal are estimated to differ by 0.425 units on Purchase Intention. The sign of b1 is positive, meaning that high Cognitive Appraisal is estimated to have a higher Purchase Intention. This effect is statistically different from zero, t = 8.566, p < .001, with a 95% confidence interval from 0.328 to 0.523, and is in support of H3.

56 The effect of b2 = 0.498 indicates that two consumers who experience the same level of brand-generated social advertising but differ by one unit in their level of Affective Appraisal are estimated to differ by 0.498 units on Purchase Intention. The sign of b2 is positive, meaning that high Affective Appraisal is estimated to have a higher Purchase Intention. This effect is statistically different from zero, t = 8.492, p < .001, with a 95% confidence interval from 0.382 to 0.613, and is in support of H4.

The effects of a3 (effect of W on M1, p = .970) and a4 (effect of W on M2, p = .492) are not significant. The results indicate that the effect of Social Advertising (brand-generated) on Cognitive Appraisal is not moderated by the type of Advertised Product (utilitarian), as evidenced by an insignificant interaction between XW in the model of Y (a5 = -.226, p = .341), which means that H5 is not supported. On the contrary, the results indicate that the type of Advertised Product moderates the effect of Social Advertising on Affective Appraisal, as evidenced by a statistically significant interaction between XW in the model of Y (a6 = -.432, p <.05). However, a closer look shows that the effect of Social Advertisement (user-generated) on Affective Appraisal is higher for a Utilitarian advertised Product (effect = -.445, SE = .140, CI: -0.721 to -.170, p <.05) than for a Hedonic advertised Product (effect = -.877, SE = 142, CI: -1.156 to -0.598, p <.001). This effect is the opposite of what was hypothesized in H6, which stated that user-generated social advertising with a hedonic advertised product would lead to a higher Affective Appraisal, as shown in the interaction plot in Figure 3. Consequently, H6 is not supported.

The indirect effect of i1 = 3.576 means that two consumers who differ by one unit in their reported brand-generated Social Advertising are estimated to differ by 3.576 units in their reported (higher) Purchase Intention through higher reported Cognitive Appraisal. This indirect effect is statistically different from zero, as revealed by a 95% BC bootstrap confidence interval that is entirely above zero (2.942 to 4.210). The indirect effect of i2 = 4.129 means that two

57 consumers who differ by one unit in their reported brand-generated Social Advertising are estimated to differ by 4.129 units in their reported (higher) Purchase Intention through higher reported Affective Appraisal. This indirect effect is statistically different from zero, as revealed by a 95% BC bootstrap confidence interval that is entirely above zero (3.594 to 4.663). The direct effect of c’1 (p = .323) of Social Advertising on Purchase Intention is not statistically significant.

Lastly, two control variables gave significant results. First of all, gender has a negative statistically significant influence on Affective Appraisal (t = -3.196, p <.05, CI: -0.537 to 0.128) (reference group = male). This negative effect means that males have a lower Affective Appraisal than females. Additionally, gender has a positive statistically significant influence on Purchase Intention (t = 2.924, p <.01, CI: 0.107 to 0.546). This positive effect means the opposite; males have a higher Purchase Intention than females. Secondly, Purchase experience on Instagram has a positive statistically significant impact on Cognitive Appraisal (t = 4.636, p

<.001, CI: 0.216 to 0.534), Affective Appraisal (t = 2.159, p <.05, CI: 0.013 to 0.281) and on Purchase Intention (t = 3.899, p <.001, CI: 0.143 to 0.435). This positive effect means that if a consumer has experience with purchasing through Instagram, their Cognitive Appraisal, Affective Appraisal, and Purchase intention will be higher.

58 Table 13

Moderated mediation PROCESS study 2

59 5.2.9 Summary of the results

User-generated Social Advertising has a stronger positive effect on Affective Appraisal (H2). The higher the perceived Cognitive (H3) and Affective Appraisal (H4), the higher the Purchase Intention. The effect of Social Advertising (user-generated) is moderated by the type of Advertised Product, such that a utilitarian Advertised Product will have a stronger positive effect on Affective Appraisal, the opposite as was hypothesized in hypothesis 6. The more experience a consumer has in using Instagram to purchase a product, the higher the Cognitive Appraisal, Affective Appraisal, and Purchase Intention. Males have a lower Affective Appraisal than females, while their Purchase Intention is higher than that of females.