Zenhausern (1978) proposed 20 items for heuristic-systematic thinking style scale which was formed in two versions, yes/no questions version and 10-point Liket-type scale.
McElroy and Seta (2003) took this scale in marketing research to analysis consumers’
thinking styles. After pretest, 14 items, 7 items for heuristic thinking and the other 7’s for systematic thinking, were included in this research. The anchor arranged in 5-point Likert-type (1=strongly disagree, and 5=strongly agree).
Information Evaluation in Purchase Decisions on the Internet
Pervious researches mentioned that there are three online information sources:
information from retailers or manufacturers, from other consumers, and from neutral information. We modify Bei, Chen, and Widdows’(2004) measurement of perceived importance of information sources. We separated the information from retailers or
manufacturers into image and advertisement. After pretest, 9 items, 3 items for information from other consumers, each 2 items for it from neutral source, ad, and image, were included in this research. The anchor arranged in 5-point Likert-type (1=strongly disagree, and 5=strongly agree).
Data Analysis Method
This research adopted two-step structural equation procedure (Jöreskog and Sörbon, 1989) proposed by Anderson and Gerbing (1988). Two-group confirmatory factor analysis (CFA) was conduct to evaluate construct validity and to test the equality of the measurement models across two product categories. Moreover, we used CFA with partial disaggregation (Bagozzi and Heatherton, 1994; Dabholkar, Thorpe, and Rentz, 1996). The partial
disaggregation modified the traditional structural equations approach which used each item as a separate indicator of the related construct. Although the traditional approach provides the more detailed level of analysis for construct testing, it is unwieldy to deal with the condition of likely high levels of random error in typical items and the many parameters that must be estimated (Bagozzi and Heatherton, 1994).
Partial disaggregation is accomplished by randomly aggregating items that relate to a given construct so that there are fewer combined indicators instead of original single-item indicators (Dabholkar, Thorpe, and Rentz, 1996). Bandalos (2002) called this technique
“item parceling,”and revealed that the use of item parcels resulted in better fitting solutions.
We took sample 1 (Nsearch goods= 200, Nexperience goods= 201) to test measurement model, and sample 2 (Nsearch goods= 200, Nexperience goods= 200) to test structural model. We
performed two-group structural analysis to test whether the structural coefficients were unequal between these two groups.
RESULTS
Confirmatory Factor Analysis (CFA)
Since heuristic-systematic thinking style scale and collectivistic-individualistic orientation scale are two factor constructs, we conducted CFA of these two construct in advance, and then conducted the CFA of all constructs. Moreover, we adopted two-group CFA to accomplish the purpose of this study which is comparison of two different product categories.
Heuristic-systematic thinking style scale included heuristic and systematic factors, and each factors included 7 items. Partial disaggregation of this model yielded acceptable fit, χ2(14)= 21.09, Root Mean Square Error of Approximation (RMSEA) = 0.053, Normed Fit Index (NFI) = 0.90, Non-Normed Fit Index (NNFI) = 0.87, Comparative Fit Index (CFI) = 0.93. We also conduct partial disaggregation of collectivistic-individualistic orientation scale, and yielded acceptable fit (χ2(6)= 25.35, RMSEA=0.13, NFI=0.82, CFI=0.84). After checking these two constructs have passed the second-order CFA tests, we then conducted CFA which included individualistic-collectivistic orientation, self-monitoring,
heuristic-systematic thinking, and information source from others customers, neutral report, and ad and image of retailers and manufacturers . Since self-monitoring scale includes 14 items, we deleted 3 items which was high cross-loading with other constructs and left 11 items in this scale. The results also showed acceptable measurement model fit, χ2(166)= 296.43, RMSEA=0.063, NNFI=0.82, CFI=0.87.
Equivalence of Measurement Models: Two Groups
In order to test factor loading invariance, we constrained the factor loading equal across groups. The nonsignificant difference in chi-square between this model (model 2 in Table 1) and baseline model indicates that the factor loading of the two measurement models were invariant, ∆χ2(9)= 5.39, ns at p < .05. The next step was to test the equality of error variances of the latent variables across the two groups. The nonsignificant difference in chi-square between this model (model 3 in Table 1) and model 2 indicates that the error variances of the two measurement models were invariant, ∆χ2(16)= 9.60, ns at p < .05. The above results reveal that the measurement models of factor loading and error variances between search goods and experience goods were invariance. This measurement model with loading and error invariance was subsequently used in the structural model analysis (Chiou, Droge, and Hanvanich, 2002).
Table 1 insert here
The validity of the final model was then evaluated. An examination of the final model (see Table 2) indicated that substantial amounts of variance in the measures were captured by the latent constructs because all loading except one path were significant. This shows the convergent validity of this measurement is acceptable.
Table 2 insert here
The Structural Model and Hypothesis TestingIndividualistic-collectivistic orientation and heuristic-systematic thinking scale were composition of two factors. In order to test the hypothesis and explain the relationships with other variables, we took the ratio of individualism and collectivism to instead the original raw score of culture dimension. In the same token, the ratio of heuristic score and systematic scores was taken to instead the original raw scores of thinking style dimension. The results of the baseline two-group model with no structural constrains were χ2(90)= 125.39,
RMSEA=0.044, NNFI=0.89, CFI=0.93. After checking the model which is achieved good model fit, we then tested equality of structural paths by systematically constraining
coefficients to be equal across groups and comparing the constrained model with the baseline model for a significant increase in the chi-square (see Table 3).
Table 3 insert here
First, we tested the moderated effect by constraining each path equal across groups separately. Table 3 shows all paths except γSM → ADwere significant difference between search goods and experience goods. This indicates the moderated effect exist in this structural model. H3is supported. SinceγCulture→ TSin both search and experience goods are positive, it demonstrates that individuals of collectivistic orientation are more likely to use a heuristic thinking style in both condition for the purchase of search goods and experience goods. H1is supported. Because γCulture→ SMis negative in search goods but positive in experience, it demonstrates that individuals of collectivistic orientation have higher self-monitoring traits in both condition for the purchase of search goods and experience goods. H2is partially supported. Table 3 also demonstrated the coefficients of γSM →
Consumerand γSM → IMwere bigger in search goods group than in experience goods group. It means that high self-monitoring consumers value the sources of information as higher level in search goods than in experience goods. H3cis partially supported.
Moreover, we tested the mediated effect of self-monitoring and thinking style between culture and evaluation of information. Four paths from culture to information from other consumers, image, ad, and neutral reports were added to compare with baseline model. The results of the competing model with full gamma matrix were χ2(82)= 119.67, RMSEA=0.048, NNFI=0.87, CFI=0.92. Compared this model with baseline model, though they were not significant different from each other, ∆χ2(8)= 5.72, ns at p < .05, we conclude the baseline model was better than the full gamma matrix model. According to the parsimony principle, the path numbers of baseline model is fewer than full gamma matrix model, baseline model is selected. Moreover, the goodness of index was better in baseline model. The figures showed mediated effect exist. H5is supported.
Furthermore, we tested the strength of self-monitoring and thinking style on different information sources. Constrained γTS → Consumer, γTS → IM, γTS → AD, and γTS → Neutral, and then compared the chi-square with not constrained model. In search goods, the strength of thinking style on different information sources was different, ∆χ2(3)= 20.27, significant at p
< .05. Table 3 showed the strength of other consumers was larger than other source.
Heuristic thinking orientated consumers value other consumers’information as the most important information among other online information for search goods. H3ais supported.
In experience goods, the strength of thinking style on different information sources was nonsignificant , ∆χ2(3)= 2.39, ns at p < .05. H3bis not supported.
DISCUSSION
Online shoppers’searching sources and processes has received less attention because it is difficult to detect. Base on prior studies and practical observation, we developed a model to detect the searching sources and processes between search and experience goods.
Although not every hypothesis was supported, we obtain some evidence from this study.
Further research would put more effort on studying this topic. Different moderated variables might attain the same conclusion as current study. Moreover, this study was conducted in Taiwan. Further research might take different culture context or different country to test this model. The function of the Internet as an information source is more importantforconsumers’decision making. Thecontribution ofthisprojectwillnotonly enhancetheunderstanding ofconsumers’information searching and processing behavior,but also help the marketing practitioners to reach their online consumers efficiently. We expect
this conclusion might provide marketing researchers more insight and practical suggestions about online sellers.
REFERENCE
Aaker, Jennifer L., and Durairaj Maheswaran (1997). The Effect of Culture Orientation of Persuasion, Journal of Consumer Research, 24 (December), 315-328.
Anderson, James and C. Gerbing (1988). Structural Equation Modeling in Practice: A Review and Recommended Two-Step Approach, Psychological Bulletin, 103 (3), 411-423.
Bagozzi, Richard P. and Todd F. Heatherton (1994). A General Approach to Representing Multifaceted Personality Constructs: Application to State Self-Esteem, Structural
Equation Modeling, 1(1), 36-67.
Bandalos, Debroah L. (2002). The Effects of Item Parceling on Goodness-of-Fit and
parameter Estimate Bias in Structural Equation Modeling, Structural Equation Modeling, 9 (1), 78-102.
Bei, Lien-Ti, Etta, Y. I. Chen, and Richard Widdows (2004). Consumers’Online Information Search Behavior and the Phenomenon of Search vs. Experience Products, Journal of
Family and Economic Issues, 25 (4), 449-467.
Bickart, B. and R. M. Schindler (2001). Internet Forums as Influential Sources of Consumer Information, Journal of Interactive Marketing, 15 (3), 31-52.
Chaiken, Shelly (1980). Heuristic versus Systematic Information Processing and the Use of Source versus Attribute Cues in Persuasion, Journal of Personality and Social
Psychology, 39 (November), 752-766.
Chaiken, Shelly (1987). The Heuristic Model of Persuasion, In M. P. Zanna, J. M. Olson and C. P. Herman (Eds.), Social Influence: The Ontario Symposium, 5 (pp. 3-39). Hillsdale, NJ: Erlbaum.
Chiou, Jey-Shen, Cornelia Droge, and Sangphet Hanvanich (2002). Does Customer Knowledge Affect How Loyalty Is Formed?, Journal of Service Research, 5 (2), 113-124.
Cousins, Steven (1989). Culture and Selfhood in Japan and the US, Journal of Personality
and Social Psychology, 56 (January), 124-131.
Dabholkar, Pratibha A., Dayle I. Thorpe, and Joseph O. Rentz (1996). A Measure of Service Quality for Retail Stores: Scale Development and Validation, Journal of the Academy of
Marketing Science, 24 (1), 3-16.
The Dieringer Research Group (2003). September. Internet Impacts More Spending Offline than Online. Received May 30, 2005, from
http://www.thedrg.com/d/nws/pr/
aics_spending_20030908.pdf.
Epstein, S., A. Lipson, C. Holstein, and E. Huh (1992). Irrational reactions to negative outcomes: Evidence for two conceptual systems, Journal of Personality and Social
Psychology, 38, 889-906.
Evans, J. S. & D. E. Over (1996). Rationality and Reasoning. Psychology Press.
FIND (2005). December. Survey of online population in Taiwan. Retrieved June 18, 2006, from http://www.find.org.tw/find/home.aspx?page=many&id=134.
Hill, D. J., M. F. King, and E. Cohen (1996). The Perceived Utility of Information Presented via Electronic Decision Aids: A Consumer Perspective. Journal of Consumer Policy, 19, 137-166.
Hoffman, D. L., and T. P. Novak (1995). Marketing in Hypermedia Computer-Mediated Environments: Conceptual Foundations, Journal of Marketing, 60 (3), 50-68.
Hofstede, Geert (1980). Cu
l t ur e ’ s Cons equences, Beverly Hills, CA: Sage.
Hofstede, Geert (1990). Culture and Organizations: Software of the Mind, Landon:
McGraw-Hill.
Jöreskog, Karl and Dag Sörbon (1989). LISREL 7: A Guide to the Program and Applications, Chicago: SPSS. Inc.
King, M. F. and S. Balasubramanian (1994). The Effects of Expertise, End-goal Adoption and Product Type on Preference Formation Strategy Adoption, Journal of Academy of
Marketing Science, 22, 146-159.
Klein, L. R. (1998). Evaluating the Potential of Interactive Media through A New Lens:
Search Versus Experience Goods, Journal of Business Research, 41, 196-203.
Markus, Hazel and Shinobu Kitayama (1991). Culture and the Self: Implications for Cognition, Emotion and Motivation, Psychological Review, 98 (April), 224-253.
McElroy, Todd and John J Seta (2003). Framing effects: An analytic-holistic perspective.