C. Yoo, D MacInnis /Journal of Business Research 58(2005)1397-1406 Table 1 Standardized LISREL estimates I ad format 1l(PF→ Credibility) 35 2.80 11(Credibility-+PF) 212(NF-Credibility) 2.18 72l( Credibility→NF) 2l(PF→ Beliefs) 07 45 731( Credibility-Beliefs) 722(NF→ Beliefs) 741(Credibility-Aad) 31(PF→Aad) 阝41(PF一Aa 32(NF→Aad) 2.12 B42(NF→Aad) 664 21(Cre → Beliefs) 2.67* B51(PF→Ab) 40029953 阝31(Cre 45* B52(NF→Ab) →Ab) 33323 3.45* B53( Beliefs→Aad 25 2.25* B42( Beliefs→Ab) B54(Aad→Ab 2.96 B43(Aad→Ab) 3.06* φ12(PF…NF x2(d=46)=3592,p=201 x2(d=48)=6472,p=.084 GFI= 91 RMSR=04 RMSR=056 CFI=952 PNFI=.646 PNFI=.66 *p<,01 *p<001 single indicator measure(see MacKenzie and Lutz(1989)or (731=.14, p=ns). Thus, the results support Hla only for the advantages of a single indicator). Constructs were not negative feelings assumed to be perfectly indicated. The measurement error Consistent with H2a and H2b, credibility thoughts terms were fixed at one minus the square root of their significantly affected Aad (B31=35, P<.01)and Ab reliability coefficients(Hayduk, 1987). We tested models in (B41=34, p<.01). Aad also significantly affected Ab Fig. 1 by ad execution conditions (B43=.30, P<.05). Consistent with H3a, credibility thoughts Overall fits of the models for the emotional ad execution have a significant positive effect on beliefs (B21=37, (x2(46)=35.92,p<.201;GFI=950;AGF1=926;RMSR p<.05). However, no significant effects of positive or 047; CFI=972; PNFI=646) and model for informational negative feelings on beliefs (721=.07, p=ns: 731=.15 ad execution (z"(48)=64.72, P<084; GFI=916: AGFI- p=ns)emerged. Thus, feelings seem to influence beliefs 863; RMSR=056: CFI=952; PNFI=665) were accept- only through the mediational effect of credibility. Beliefs able. The Q Plot, the low number of high normalized significantly influence brand attitudes (B42=23, P<.05), residuals (i.e, only two exceed 2.0), the t-tests, and range supporting H4 of values(e. g, no negative variances) suggest that each model is acceptable. Thus the hypothesized model appears to fit the data. Structural equation model estimates are summar- ized in table 2 Summary results of confirmatory factor analyses Construc No of Reliability(variance Lowest t-value 5.3. Test of hypotheses items extracted estimate)(average of 4) methodology by ad execution condition. Results are Belies elings7 We tested our hypotheses using structural equations .92(.641) Phi-values(standard error of the estimates) 531. Emotional ad execution condition Consistent with HIb both positive feelings (711=35, p<o1)and negative feelings(7=-27,p<.05)signifi- 2.Negative feelings -.253(060) cantly affect credibility. Credibility, in turn, influences Aad 3. Credibility thoughts -.521(.048)-.239(.061) (B31=35, p<.05). Thus HIb is supported. Negative4Beliefs -.276(077)-234(079)379(.075) feelings also influence Aad directly (732=-.26, P<.05) Overall fit indices: x'(345)=480.42(P<.000 However, positive feelings have no direct effects on Aad Comparative fit index=952single indicator measure (see MacKenzie and Lutz (1989) or the advantages of a single indicator). Constructs were not assumed to be perfectly indicated. The measurement error terms were fixed at one minus the square root of their reliability coefficients (Hayduk, 1987). We tested models in Fig. 1 by ad execution conditions. Overall fits of the models for the emotional ad execution (v2 (46) = 35.92, p <.201; GFI = .950; AGFI =.926; RMSR = .047; CFI = .972; PNFI =.646) and model for informational ad execution (v2 (48) = 64.72, p <.084; GFI = .916; AGFI = .863; RMSR =.056; CFI =.952; PNFI =.665) were accept￾able. The Q Plot, the low number of high normalized residuals (i.e., only two exceed 2.0), the t-tests, and range of values (e. g., no negative variances) suggest that each model is acceptable. Thus the hypothesized model appears to fit the data. Structural equation model estimates are summar￾ized in Table 2. 5.3. Test of hypotheses We tested our hypotheses using structural equations methodology by ad execution condition. Results are summarized in Fig. 2. 5.3.1. Emotional ad execution condition Consistent with H1b both positive feelings (c11 = .35, p <.01) and negative feelings (c12 =
.27, p <.05) signifi￾cantly affect credibility. Credibility, in turn, influences Aad (b31 =.35, p < .05). Thus H1b is supported. Negative feelings also influence Aad directly (c32 =
.26, p <.05). However, positive feelings have no direct effects on Aad (c31 =.14, p = ns). Thus, the results support H1a only for negative feelings. Consistent with H2a and H2b, credibility thoughts significantly affected Aad (b31 =.35, p < .01) and Ab (b41 =.34, p < .01). Aad also significantly affected Ab (b43 = .30, p < .05). Consistent with H3a, credibility thoughts have a significant positive effect on beliefs (b21 =.37, p < .05). However, no significant effects of positive or negative feelings on beliefs (c21 =.07, p = ns; c31 =.15, p = ns) emerged. Thus, feelings seem to influence beliefs only through the mediational effect of credibility. Beliefs significantly influence brand attitudes (b42 =.23, p <.05), supporting H4. Table 2 Summary results of confirmatory factor analyses Constructs No. of items Reliability (variance extracted estimate) Lowest t-value (average of k) Positive feelings 14 .88 (.588) 3.64 (.69) Negative feelings 7 .92 (.641) 4.01 (.76) Credibility thoughts 6 .89 (.639) 4.23 (.74) Beliefs 1 Phi-values (standard error of the estimates) 1 2 34 1. Positive feelings – 2. Negative feelings
.253 (.060) – 3. Credibility thoughts
.521 (.048)
.239 (.061) 4. Beliefs
.276 (.077)
.234 (.079) .379 (.075) – Overall fit indices: v2 (345) = 480.42 ( p <.000). Comparative fit index =.952. Table 1 Standardized LISREL estimates Emotional ad format Coefficient t-value Informational ad format Coefficient t-value c11 (PFYCredibility) .35 2.80* c11 (CredibilityYPF) .43 3.51** c12 (NFYCredibility)
.27
2.18*** c21 (CredibilityYNF)
.40 3.18** c21 (PFYBeliefs) .07 .45 c31 (CredibilityYBeliefs) .10 .96 c22 (NFYBeliefs) .09 .69 c41 (CredibilityYAad) .02 .06 c31 (PFYAad) .14 1.22 b41 (PFYAad) .49 3.66** c32 (NFYAad)
.26
2.12*** b42 (NFYAad)
.29
2.41* b21 (CredibilityYBeliefs) .37 2.67* b51 (PFYAb) .25 2.06*** b31 (CredibilityYAad) .35 2.45* b52 (NFYAb)
.13
1.25 b41 (CredibilityYAb) .34 3.45* b53 (BeliefsYAad) .25 2.25* b42 (BeliefsYAb) .23 2.23 b54 (AadYAb) .32 2.96* b43 (AadYAb) .30 3.06* /12 (PF6NF) .08
.58 v2 (df = 46) = 35.92, p =.201 v2 (df = 48) = 64.72, p =.084 GFI =.950 GFI =.916 AGFI =.926 AGFI =.863 RMSR =.047 RMSR=.056 CFI =.972 CFI =.952 PNFI =.646 PNFI =.665 PF=positive feeling. NF=negative feeling. * p <.01. ** p <.001. *** p <.05. 1402 C. Yoo, D. MacInnis / Journal of Business Research 58 (2005) 1397 – 1406