*Professor of mis school of Business Administration Sungkyunkwan University Seoul 110-745, Korea

B. Structural Equation Modeling (SEM) Analysis

The SEM results depicted in Figure 4 show that all the fit indices are successfully met. For example, value divided by degree of freedom is less than 3, and GFI is over .90, AGFI over .80, NFI over .90, NNFI over .90, and SRMR below .05 [34, 41]. Other fit indices also meet the theoretical threshold: CFI=.99, IFI=.99. This model, while very parsimonious, explains a significant portion of the variance in perceived extent of use (R² = .51), perceived website satisfaction (R² = .71), flow (R² = .56), and structural assurance (R² = .63). It can be concluded, then, that the proposed structural model is statistically sound. The structural model SEM results are as follows (Figure 4). First, offline trust impacts online banking constructs directly, with a significant effect on flow (.72, t=10.49**), structural assurance (.57, t=7.18**), perceived website satisfaction (.27, t=2.88**) and perceived extent of use (.64, t=5.03**). Second, structural assurance influences flow (.32, t=4.38**) and perceived website satisfaction (.41, t=4.00**), but not extent of use. Third, flow significantly influences perceived website satisfaction (.27, t=3.52**) and extent of use (.26, t=2.60**). Finally, perceived satisfaction did not influence extent of use.

^** p < 0.01 (t>1.96)

Figure 4. Structural Equation Model Results
The off-to-on portion of these results can be summarized by stating that in support of H1-H4, offline trust has a strong and positive influence on the online variables proposed. This supports the existence of Type 2 (Offline-to-Online) TTP. We may therefore conclude that Type 2 TTP exists and applies to real world banking e-commerce activities.
The questionnaire respondents include those who had never before used online banking (first-time users) and those who had (experienced users). While the experienced users reported based on online banking experience, the first-time users reported more based on cue-based trust [24, 72], in that they supplemented their offline bank trust with cues from the online website of the bank. To examine the differences between first-time users and experienced users, we split the sample into first time (n= 111) and experienced (n=88) and re-ran the model. Table 4 shows the model results split by first-time user data and experienced user data. For first-time users, offline trust influences all four online constructs significantly, whereas for experienced users, offline trust affects only three online constructs significantly—flow, structural assurance, and perceived extent of use, but not perceived website satisfaction. This likely means that since cue-based trust is replaced with experience-based trust over time, the experienced users’ perceived website satisfaction is not affected by offline trust itself. We also observe from Table 4 that the path coefficients from offline trust to the online constructs for first-time users are significantly greater (p<.01) than those for experienced users, except for perceived extent of use. This finding conforms with theory about cue-based trust and experience-based trust, in that first-time users are influenced more than experienced users by their trust in the brick-and-mortar bank to form their site flow, structural assurance, and satisfaction impressions. Experienced users know how online banking works, so they are less influenced by their offline bank trust. However, offline trust had a greater effect on perceived extent of use among experienced users, perhaps because these users have developed the skill set needed to effectively use the product. Thus, they could more clearly apply their offline trust to a projected use of the service.¹ We project that over time with repeated interactions using the online banking systems, experience-based trust will play a greater role than cue-based trust for all the dependent variables.
Table 4. Empirical results for first-time users and experienced users

	First-time user model	Experienced user model
Indexes of adjustment of the model	χ2=98.152(p=0.007), df=67, GFI=0.89, AGFI=0.83, NFI=0.92, NNFI=0.96 RFI=0.89, IFI=0.97, CFI=0.97, SRMR=0.04, RMSEA=0.06	χ2=82.983(p=0.090), df=67, GFI=0.89, AGFI=0.83, NFI=0.89, NNFI=0.97 RFI=0.85, IFI=0.98, CFI=0.98, SRMR=0.05, RMSEA=0.04
Flow: Structural Assurance Perceived Website Satisfaction Perceived Extent of Use	R^2 .33 .59 .72 .29	R^2 .28 .32 .43 .46
TrustFlow	0.55^**	0.34^** a
TrustStructural Assurance	0.72^**	0.40^** a
Trust  Perceived Website Satisfaction	0.54^**	n.s ^a
Trust Perceived Extent of Use	0.40^*	0.58^** a
Structural Assurance Flow	n.s	0.36^** a
Structural Assurance  Perceived Website Satisfaction	n.s	0.63^** a
Structural Assurance  Perceived Extent of Use	n.s.	n.s
Flow  Perceived Website Satisfaction	0.25^**	n.s
Flow Perceived Extent of Use	n.s	0.35^** a
Perceived Website Satisfaction  Perceived Extent of Use	n.s	n.s

** p <0.01 (t>1.96), * p <0.1 (t>1.282)

^a: t-tests showed significant (p<.01) differences for these coefficients between first-time and experienced groups, using the formula: t = (PC₁-PC₂)/[ S_pooled x SQRT(1/N₁+1/ N₂)]; S_pooled = SQRT{[( N₁-1)/( N₁+ N₂ -2)] x SE₁² +[( N₂-1)/( N₁+ N₂ -2)] x SE₂²}; SE =Standard error of path in structural model; PC =Path Coefficient in structural model

V. Study Limitations
This study captures primarily a cross-sectional view of model constructs. Thus, a longitudinal study would be helpful. Because of the lack of longitudinal data, causality of the model is not proven. Reverse linkages or bi-directional linkages among the constructs are possible over time. For example, online satisfaction should lead to online trust over time, just as Harris and Goode [42] found about satisfaction and trust in the offline world. This constitutes a boundary condition for our model, in that the model works best when consumers are relatively new to online banking, such as those in our sample (see Table 1). Sample size is another limitation, although the size is adequate for the tests conducted. Note that while model fit degraded somewhat when the sample was split by experience level, yet the RMSEA remained in an acceptable range. Although our scales displayed acceptable psychometric properties, the items we used are adapted from (and subsets of) the items used in other studies. Using alternative items, the results may vary somewhat from ours. Generalizability of results is another weakness. The study results may be different if the model were tested in other offline/online domains or in other cultures. For comparison purposes, it is noteworthy that South Korea is regarded as a rapid adopter of the Internet. The model results may also differ depending on whether the websites serve an information-intensive or fulfillment-risk function. Our study’s results do not apply directly to those website vendors who attract customers through sources other than their offline business. Since offline trust cannot be used by these vendors to build online trust, they should build trust through other methods, such as through institutional assurances, website ease of use, and website design quality that signals to consumers the trustworthy attributes of the web vendor [34, 70].
VI. Managerial Implications
Based on the empirical results, this study arrives at the following implications. First, the TTP (trust transfer process) provides a unified view for understanding the effects of offline trust on online perceptions of flow and structural assurance. The empirical results show that consumer trust in an offline channel transfers rather easily to positive online channel perceptions, suggesting that vendors can leverage their offline trust to produce online flow and structural assurance. Hence, marketing strategies are best organized so that online perceptions like these are considered. For example, marketing materials should emphasize the good content and user control of the online system (to elevate flow perceptions) and the safeguards and protections of the system, such as SSL / sophisticated encryption (to elevate structural asurance). Marketing materials could subtly connect these positive online system attributes to aspects of the quality offline service the consumer already receives to increase the offline-to-online transference effect.
Second and related, offline trust is important in triggering positive online outcomes. Figure 4 shows that offline trust can trigger positive perceived website satisfaction and perceived extent of use. After the dot com bubble burst in 1998, financial analysts understood the importance of companies maintaining some offline activity. This confirms our proposed TTP: offline trust enables or facilitates the transfer process across channels. The result that offline trust positively influences these variables such as perceived website satisfaction and perceived extent of use of the website indicates that offline trust can be used as an enabling factor by which an online company that started from an offline channel can attract customers and make them more loyal to its website. Then an online company with strong offline trust can build up a high level of trust and reputation among online customers for a certain period, eventually triggering other types of TTP, for example, Type 3 (Online-to-Online) and Type 4 (Online-to-Offline).
Since offline trust has the impacts we found, several actions should be taken to increase offline trust in the bank. a) Banks should use marketing campaigns to try to improve perceptions about the reputation and size of the firm, because these have proven to increase trust [25, 50]. b) Banks should improve actual and perceived offline customer service, which should improve the benevolence and competence aspects of trust. c) Customer service employees should be trained to be more likable, which has been shown to influence offline trust [25]. d) Marketing should emphasize the values the bank shares with customers [75].
Third, the results not supportive of two hypotheses—H8 and H10—also have an important implication. Although structural assurance and perceived extent of use were significantly correlated (r = .56**), structural assurance did not predict perceived extent of use. This is largely because more powerful factors (i.e., offline trust and flow) outweighed the effects of structural assurance. Similarly, perceived website satisfaction correlated with perceived extent of use (r = .52**), but did not predict it because of the more dominant effects of flow and offline trust. This may also be because in the initial phase, website satisfaction is tentative and therefore is not heavily relied on as input for whether or not to use the online banking site. Offline trust, developed and reinforced over time, was relied on instead. These findings reinforce how important offline trust is to online outcomes.
Fourth, the questionnaire data confirmed the existence of Type 2 TTP, which was not studied previously in an explicit manner. Offline trust was undeniably transferred to online channels, in that it had a significant effect on online consumer perceptions.
Further, though offline trust was empirically proven here to be usually a starting point for TTP initiator, we believe from Figure 1 that TTP can be triggered not only from offline trust, but from online customer trust. TTP Types 3 and 4 in TTP are examples. As discussed in the previous section, among the four cells in TTP, Type 4 is very rare to find in real e-commerce applications. However, as e-commerce becomes omnipresent in business world and the Internet-infrastructure is getting more advanced and high-speed, Type 4-related businesses emerge in the market as a new trend in e-commerce to attract more customers regardless of offline and online. For example, WASSADA (http://www.wassada.com) is another typical Type 4-related company based in South Korea, which started from a pure Internet company selling audio electronic items and expanded into offline stores, successfully taking advantage of the high level of trust among online customers. MISSHA (http://www.beautynet.co.kr), mentioned earlier, is a case of another Type 4 business which is very successful and highly praised in various mass-media in South Korea. The authors feel that Type 4-related successes will most likely be found in those societies having high speed and broadband Internet infrastructure, and showing a mature stage of e-commerce, like South Korea, Hong Kong, etc.
Further implications of the empirical results of this study include the following two issues.
Issue 1: The Role of Offline Trust in Determining a Customer’s Online Behavior

As already discussed, offline trust plays a crucial role in determining consumer online behavior. In the case of online banking, offline trust directly affects four key related online constructs. The flow customers experience on banking websites is especially influenced by offline trust, as it has a 0.72 path coefficient. Customers’ website satisfaction increases when offline trust is greater. Similarly, customers’ perceived extent of use is affected by offline trust, with a 0.64 path coefficient. The latter is especially striking since neither structural assurance nor perceived satisfaction affected perceived extent of use. Offline trust also influences customers’ structural assurance beliefs about the safety and security of the bank’s online website. Perceived website satisfaction is relatively less affected by issues of offline trust than other online constructs. As noted in [69], many factors influence perceived website satisfaction. Offline trust, therefore, is only part of the answer. Likewise, many factors affect the adoption of Internet banking [39], which indicates a need for much additional research. Internet marketing strategists have struggled to understand why various strategies prove effective on some websites and ineffective on others. Obviously, the transfer of offline trust to online channels should be given more attention by researchers and marketers before online marketing strategies are developed. Research should explore factors that facilitate transfer of offline trust to online website adoption.

Issue 2: How to secure trust in man-machine website interactions

One characteristic of the online channel is the tendency for customers to perform many human-computer interactions (“click-and-see” activities). Websites offer a hypermedia environment which is made up of text, images, voice, and animation, leading to an enriched environment for human-machine interactions. Nevertheless, the issue is whether consumers trust the information they get from the websites. Conventional wisdom is that the more users understand how the information originates, the more online trust increases [68]. Since websites provide a rather enriched hypermedia environment for customers, using the websites would probably secure online trust to some extent. However, we still need to turn to the importance of offline trust to understand how consumers feel secure about a website. There have been no studies thus far to clearly argue that offline trust is the key to predicting online behavior. Although a certain level of human-computer interaction is necessary to navigate a website, the results of this study suggest that mere management of the websites will not lead to the hoped-for results without the proper degree of offline trust.

VII. CONCLUDING REMARKS
This study’s results show that trust in an offline bank influences key factors in an online bank environment. Specifically, this paper contributes by showing that trust in an offline bank influences structural assurance, flow, consumer satisfaction and extent of use of the bank’s online system. In an era in which many companies are turning to the Internet as a way to expand their business, this study indicates that firms can leverage customer trust in their brick-and-mortar business to provide a similar customer-satisfying product line on the Internet. The extent to which offline trust affected online perceptions suggests that trust in the offline business may be a key factor of online business success.
There are further research issues that need to be addressed, including the link between offline trust in the bank and trust in its online banking system. This study addressed offline/online banking. Other types of businesses should be studied to see the extent to which TTP works in other domains besides banking. Also, it is important to understand the psychological mechanisms by which the offline to online trust transfer process operates. It is hoped that this study will provide a stepping stone to building more effective marketing strategies for e-commerce.
ACKNOWLEDGMENT

This paper was supported by Samsung Research Fund, Sungkyunkwan University, 2003.

REFERENCES
[1] S. L. Ahire, D. Y. Golhar and M. A. Waller, "Development and valuation of quality management implementation constructs," Decision Sciences, vol. 27, no. 1, pp. 3-56, 1996.
[2] A. M. Aladwani, "Internet banking: A field study of drivers, development challenges, and expectations," International Journal of Information Management, vol. 21, no. 3, pp. 213-225, 2001.
[3] J. C. Anderson and D. W. Gerbing, "Some methods for respecifying measurement models to obtain unidimensional construct measures," Academy of Management Review, vol. 19, no. 3, pp. 47-59, 1982.
[4] E. M. Awad, "The structure of e-commerce in the banking industry," Proceedings of the 2000 ACM SIGCPR Conference on Computer Personnel Research, pp. 144-150, 2000.
[5] S. Ba and P. A. Pavlou, "Evidence of the effect of trust building technology in electronic markets: price premiums and buyer behavior," MIS Quarterly, vol. 26, no. 3, pp. 243-268, 2002.
[6] J. B. Baty and R. M. Lee, "Intershop: enhancing the vendor/customer dialectic in electronic shopping," Journal of Management Information Systems, vol. 11, no. 4, pp. 9-31, 1995.
[7] J. R. Bettman, "Relationship of information-processing attitude structures to private brand purchasing behavior," Journal of Applied Psychology, vol. 59, pp. 79-83, February 1974.
[8] A. Bhattacherjee, "An empirical analysis of the antecedents of electronic commerce service continuance," Decision Support Systems, vol. 32, no. 2, pp. 201-214, 2001.
[9] K. A. Bollen, Structural Equations with Latent Variables, New York: John Wiley, and Sons, 1989.
[10] A. Boomsma, K. G. In, Jorekog and H. Wold (eds.), The robustness of LISREL against Small Sample Sizes in Factor Analysis Models, in Systems under Indirect Observation: Causality, Structure, Prediction, Amsterdam, North Holland, pp. 149-173, 1982.
[11] W. Boulding, A. Kalra, R. Staelin and V. A. Zeithaml, "A dynamic process model of service quality: From expectations to behavioral intentions," Journal of Marketing Research, vol. 30, pp. 7-27, February 1993.
[12] M. C. Campbell and R. C. Goodstein, "The moderating effect of perceived risk on consumers’evaluations of product incongruity: Preference for the norm," Journal of Consumer Research, vol. 28, pp. 439-449, December 2001.
[13] G. Chakraborty, V. Lala and D. Warren, "An empirical investigation of antecedents of B2B websites’ effectiveness," Journal of Interactive Marketing, vol. 16, no. 4, pp. 51-72, 2002.
[14] A. Chaudhuri and M. B. Holbrook, "The chain of effects from brand trust and brand affect to brand performance: The role of brand loyalty," Journal of Marketing, vol. 65, pp. 81-93, April 2001.
[15] Cheskin Research, eCommerce Trust Study, January 1999.
[16] G. A. Jr. Churchill and C. Surprenant, “An investigation into the determinants of customer satisfaction,” Journal of Marketing Research, vol. 19, pp. 491-504, November 1982.
[17] J. Conklin, "Hypertext: An introduction and survey," IEEE Computer, vol. 20, no. 9, pp. 17-41, 1987.
[18] A. D. J. Cooke, T. Meyvis and A. Schwartz, "Avoiding future regret in purchase-timing decisions," Journal of Consumer Research, vol. 27, pp. 447-459, March 2001.
[19] A. T. Coughlan, E. Anderson, L. Stern and A. El-Ansary, Marketing Channels. Prentice Hall, NJ: Englewood Cliffs, 2001.
[20] M. Csikszentimihalyi, Flow: The Psychology of Optimal Experience, New York: Harper & Row Publishers, 1990.
[21] W. DeLone and E. McLean, "Information systems success: The quest for the dependent variable," Information Systems Research, vol. 3, no. 1, pp. 60-95, 1992.
[22] A. R. Dennis, "Lessons from three years of web development," Communications of the ACM, vol. 4, no. 7, pp.112-113, 1998.
[23] S. Devaraj, M. Fan and R. Kohli, "Antecedents of B2C channel satisfaction and preference: Validating e-commerce metrics," Information Systems Research, vol. 13, no. 2, pp. 316-333, September 2002.
[24] R. R. Dholakia and B. Sternthal, “Highly Credible Sources: Persuasive Facilitators or Persuasive Liabilities?” Journal of Consumer Research, vol. 3, pp. 223–232, March 1977.
[25] P. M. Doney and J. P. Cannon, "An examination of the nature of trust in buyer-seller relationships," Journal of Marketing, vol. 61, pp. 35-51, April 1997.
[26] G. R. Dowling and R. A. Staelin, "Model of perceived risk and intended risk activity," Journal of Consumer Research, vol. 21, pp. 119-134, June 1994.
[27] R. F. Dwyer, P. H. Schurr and S. J. Oh, "Developing buyer seller relationships," Journal of Marketing, vol. 51, pp. 11-27, April 1987.
[28] J. Eighmey and L. McCord, "Adding value in the information age: Uses and gratifications of sites on the world wide web," Journal of Business Research, vol. 41, no. 3, pp. 187-194, 1998.