Applying the technology acceptance model to the WWW.

Johnson, Richard A. ; Hignite, Michael A.

INTRODUCTION

The WWW is a recently emerging and extremely important information technology. As evidence of the WWW's growth and influence, the value of goods and services provided via e-commerce was estimated at $3.0 billion in 1998 and is expected to mushroom to $75 billion in 2000 (Lieb, 1999). Online advertising expenditures in the U.S. alone will grow from $2.8 billion in 1999 to $22 billion by 2004 (Pastore, 1999). Two out of three American households now have access to the Internet at home, work, or school. Ninety-two million adults in North America (approximately 40% of the adult population) now use the Internet, with 55 million shopping online (Lieb, 1999). However, little research exists on understanding the psychological factors that contribute to the use of the WWW.

The technology acceptance model (TAM) (Davis, 1989) has experienced more than a decade of relative success in predicting and explaining the acceptance of many types of end user computing systems in work or school settings. However, it has not been rigorously applied to what may perhaps be the single-most important information technology of the new millennium--the WWW. It is vitally important to study the acceptance of the WWW since it is so far-reaching and should have a tremendous impact on the personal and economic lives of literally billions of individuals in the near and distant future (Hoffman et al., 1999; Graphic, Visualization, & Usability Center, 1999). Understanding more about the acceptance of the WWW can lead to significant improvements in the design of both software and hardware to increase its usefulness and ease of use. Applying the TAM to the WWW can also lead to a better theoretical understanding of possible important differences between the WWW and of other types of end user systems.

The purpose of this study is to explore the relationships among the factors that may influence the voluntary, personal use of the WWW. In order to apply the TAM to the WWW, a survey instrument that includes the twelve standard TAM items was administered to n=295 college students at a large Midwestern university. A rigorous analysis of the measurement and structural equation models was conducted on the collected data using LISREL (version 8.3) software, yielding some very unexpected and interesting results. Background on the WWW and TAM are now presented followed by a discussion of the research methodology and results. The paper concludes with an exploration of the study's implications and a call for additional research.

BACKGROUND

Usage of the WWW

A recent comprehensive survey (GVU Center, 1999) has brought to light many interesting aspects of WWW usage. Nearly two-thirds of all WWW users are between the ages of 21 and 45 and two-thirds of all users are male. Some 85% of all users live in the U.S. with a nearly equal percentage having attended or graduated from college. About half of all users have been online for less than three years, and half of all users access the WWW from home with another 35% connecting at work. Two-thirds of all WWW users connect via a 56k or slower modem. The two most frequent problems cited with the WWW are slow speed (61%) and broken links (57%). The overuse of graphics (48%) and being unable to find needed information (45%) were also among the top five problems cited. Gathering information for personal needs was the most frequently cited reason for using the WWW followed by work, entertainment, and education. Nearly 90% of all WWW users do so just to have fun and explore. There is no doubt that the WWW has already achieved a high rate of acceptance, but it is also clear that there is much room for improvement in both the quantity and quality of future use.

The Technology Acceptance Model (TAM)

The TAM (Davis, 1989) was developed to predict and explain the voluntary use of any type of end user computing system used within the context of one's job. As Figure 1 illustrates, TAM postulates relationships among three constructs, perceived usefulness (U), perceived ease of use (EOU), and actual (or self-reported) usage (USE) of the system. The individual items that constitute the U and EOU scales are provided in Appendix I where "X" represents the target system under consideration (such as electronic spreadsheets, word processors, or email). U and EOU are expected to directly influence USE while U may also mediate the effect of EOU on USE.

[FIGURE 1 OMITTED]

The original intent of the TAM was to explain and predict the voluntary usage of end user computing systems in a work environment. The results of nearly all previous studies indicate that U is the predominant determinant of USE, that EOU is usually a very weak or nonexistent influence on USE, but that EOU is a moderately strong influence on U. As a typical example, Davis (1989) uncovered the relationships shown in Figure 2 for use of a graphics package.

[FIGURE 2 OMITTED]

The voluntary use of the WWW may exhibit many characteristics that differ from those of the voluntary use of systems in a purely job-related or school-related environment. Using the WWW to search for specific information of personal interest (sports, weather, jobs, etc.) or to simply "browse" for entertainment may be quite different than using a system to schedule production in a factory or to complete a homework assignment. The goal of this research is to explore such differences, if they exist.

METHODOLOGY

The Instrument

The survey instrument (see Appendix II) consists of demographic items (age, gender, major) and questions concerning the extent of computer use in general and WWW use in particular. Questions about the use of the WWW include both the amount and purpose of WWW usage. The survey concludes with the twelve items that constitute the standard TAM scales for measuring perceived usefulness (U) and perceived ease of use (EOU). These twelve items were scrambled in order to avoid methods bias.

The wording of the TAM items in the survey differs from the original in two important ways. First, the term "the WWW" is substituted for the name of the target system, "X." It is a standard practice to replace "X" with the term used for the specific end user system under consideration. Second, references to the user's "job" are replaced with the user's "favorite activity" (an activity that the user can and does perform on the WWW, but can also be performed by conventional means, as explained in the survey instructions). Examples of such activities include searching for a job on the WWW instead of searching for one in the newspaper, or browsing for baseball cards on the WWW instead of at the card shop.

To help ensure content validity, the instrument was pilot tested on several faculty who teach WWW use and a class of 40 students who had recently been trained in WWW use. Several improvements were made in the original instrument prior to its final administration.

The Subjects

The subjects for this study originally consisted of 571 students from a large mid-western university. These students were nearing completion of their respective computer-related courses, thus ensuring that they were very computer literate. The students ranged from freshman to seniors and represented a wide cross-section of majors. The participants were requested to complete the survey to receive extra credit in their respective courses, and virtually all complied. Since the application of the TAM is predicated on the voluntary use of a system, a final sample of only n=295 subjects was used for analysis based on their response on the survey that at least 50% of their use of the WWW was to meet personal needs (as opposed to satisfying job or school requirements). Thus, the use of the WWW by these respondents was largely voluntary.

Analysis

The analysis method will involve first applying LISREL (version 8.3) structural equation modeling software to evaluate the appropriateness of the TAM measurement model for the three constructs under consideration (U, EOU, and USE) in a WWW context. After any required modifications of the measurement model, LISREL will be applied to the structural equation model as presented in Figure 1. The goodness of fit of the measurement and structural equation models will be evaluated using several standard measures provided by the LISREL output. Table 1 lists the specific measures to be used and their recommended cutoff values for satisfactory model fit (Bentler and Bonnet, 1980; Bollen, 1989; Hair et. al, 1992; Sharma, 1996).

The recommended technique for improving the fit of either the measurement model or the structural equation model is to perform successive iterations with LISREL to identify offending estimates, such as item loadings on constructs or path coefficients in the structural model (Joreskog and Sorbom, 1993). Of course, a sound theoretical basis is required before eliminating paths from models (Hair et. al, 1992). The final resulting structural model should shed light on how the acceptance of the WWW for personal use may differ from the acceptance of end user systems for work-related use.

RESULTS

Demographics of the Respondents

Respondents to the survey were fairly equally divided among four groups: freshmen enrolled in a required introductory IT course, sophomore business majors enrolled in a required intermediate IT course, senior business majors (non-IS majors) enrolled in an MIS course, and sophomore/junior IS majors. Of the n=295 students who completed the survey, 47% were at least 21 years old and 53% were age 18-20. With respect to gender, 54% were male and 46% were female. Forty-eight percent of the respondents were business majors (non-IS), 23% were non-business majors, 19% were IS majors, and 10% reported no major. The subjects spend an average of 6.4 hours per week on the WWW. An average of 67% of these students' time on the WWW was for personal reasons (about half of that just browsing) with 30% to satisfy school requirements. About 50% of their time on the WWW was accessed from home, 45% from school. This sample gave an average rating of 5.8 to their overall satisfaction with the WWW.

Although the sample of undergraduate students was not random, the group provides a good cross-section from several different demographic segments of the undergraduate population of a large midwestern university. The sample represents both genders nearly equally, a wide variety of experience in both computer and WWW use, and a diversity of both age and major areas of study.

The Measurement Model

A rigorous application of LISREL software was performed on the data supplied by the final sample of n=295 students. First, the measurement model for the two variables U and EOU was evaluated. While the factor loadings of all items on their respective constructs were all quite high (ranging from 0.86 to 0.96), many of the goodness of fit indicators were far below the recommended cutoff levels. The LISREL output includes modification indices that identify which items are offending estimates and are thus candidates for removal from the model. Following the recommended process of successive refinement of the measurement model for U and EOU (Joreskog and Sorbom, 1993; Hair et. al, 1992), a total of four of the twelve TAM items were removed from the U/EOU model. Table 2 lists those items removed and the rationale for their removal.

The measurement model for USE originally consisted of three indicators: HOURS/DAY, DAYS/WEEK, and SESSIONS/DAY. A LISREL analysis of this model resulted in the removal of SESSIONS/DAY due to a relatively low factor loading of 0.42 (with a recommended cutoff of 0.5) (Pedhazur and Schmelkin, 1991). The complete measurement model is displayed in Figure 3 showing the survey items that serve as indicators of the three TAM constructs: U, EOU, and USE. The item loadings on their respective constructs are also shown.

[FIGURE 3 OMITTED]

The Structural Equation Model

Once the best-fitting measurement model for WWW acceptance is established, the structural equation model may be analyzed to determine the relationships among model constructs. The very first attempt to establish the relationships was quite successful resulting in the structural model in Figure 4. All goodness of fit indicators fell within their acceptable ranges. The final model indicates that there is no significant relationship between U and USE, but there are substantial and significant relationships between EOU and U, and between EOU and USE.

[FIGURE 4 OMITTED]

DISCUSSION

The Measurement Model

Obtaining a measurement model that fits the collected data, as explained in Table 2 and displayed in Figure 3, points to several important findings. First, consider the removal of Item 12 (easy to become skillful) from the TAM scale. The ease with which one becomes skillful at using the WWW may be a very distinct issue from the ease with which one uses the WWW once the skill is obtained. It is a common perception that navigating the WWW effectively by using search engines (e.g., Yahoo, AltaVista, Infoseek) and by following hyperlinks can be a daunting task to a beginner. Also, the concept of skill may not apply to recreational or personal use of the WWW as much as it does to the work-related use of an end user system. For example, one does not normally think of skill in the context of such recreational activities as watching movies or window-shopping.

The wording of Item 8 in the TAM scale ("Using X makes it easier to perform Y.") may be confusing since Item 8 was originally intended as a measure of U. The term "easier" may invoke thoughts of EOU from the respondent. Item 5 (flexibility of the system) has a history of measurement problems (Davis, 1989) as respondents may not have a clear idea of exactly what is meant by the term "flexibility" in the context of using a software package. Also, flexibility may be considered a usefulness concept instead of an ease of use concept (as it was originally intended in the TAM). Finally, Item 2 ("Using X enables me to perform Y more quickly.") may not apply in the special case of the WWW, as it is known to be notoriously slow, especially when a slow modem is used as an access device (Overton, 2000).

The Structural Model

The structural model of Figure 4 also suggests many interesting relationships that may be unique to the WWW. The results of most applications of TAM are represented by Figure 2 where perceived usefulness (U) has a strong influence on actual use (USE), perceived ease of use (EOU) has a moderate influence on USE, and EOU has virtually no influence on USE. But the tables are turned dramatically when the WWW is considered, as evidenced in Figure 3. Perceived usefulness (U) has no significant influence on actual use. While one expects only a moderate influence of EOU on U, the WWW application exhibits a very strong influence of EOU on U. Finally, traditional TAM results predict virtually no effect of EOU on USE, but the WWW environment suggests that EOU is a moderately strong determinant of USE.

The question naturally arises as to why the results of applying the TAM to the WWW should be so decidedly contrary to the results of several other previous studies. The answer possibly lies in the focus on personal use of the WWW in contrast to the work-related use of end user systems in previous studies. If one's goal in using the WWW is primarily personal (browsing, entertainment, etc.), there may be little time or performance pressure involved. Additionally, "user comfort" may be of much higher concern. Therefore, the ease with which one is able to use the WWW should be a strong determinant of actual use. When a particular task is mandated (jobs must be scheduled or homework assignments must be completed), the usefulness of a system is of paramount importance and ease of use is secondary (Davis, 1989).

The perceived ease of use of the WWW has a very strong influence on perceived usefulness. Being able to quickly and easily navigate the WWW to accomplish personal goals is certainly a worthwhile objective. However, perceived usefulness has a very small (negative) influence on use, and cannot even be meaningfully addressed since the relationship is not statistically significant. Personal use of the WWW may be useful in the sense that it satisfies one's curiosity or entertains, while the use of end user systems in the workplace may be useful in the sense that they allow one to keep a job or get a promotion. The latter type of utility, where there are clear-cut goals and expected beneficial outcomes resulting from the achievement of such goals, should have a particularly strong influence on future use, as indicated in previous studies of TAM.

Implications

While it may be tempting to conclude that the personal use of the WWW is not as important as the work-related use of other types of end user systems, it is important to remember the tremendous economic impact that the WWW is now having and will continue to have in the future. This research suggests that in order to enhance the usage of the WWW, developers of Web sites, network administrators, and hardware providers must pay particular attention to the ease of use issue. Web sites must be well designed, network downtime must be minimized, and connection speed must be improved. As recent surveys indicate (GVU Center, 1999), ease of use of the WWW is often a sore spot with users.

This study also suggests that enhancing perceived ease of use should have a very strong effect on perceived usefulness. Although perceived usefulness had no effect on use in this study, this situation may change in the future. As WWW users become more familiar with this new technology and use it to achieve more concrete and more economically vital personal goals (such as business-to-consumer e-commerce or job hunting), one would expect that perceived usefulness would have a more substantial impact on use.

Limitations and Future Research

One limitation of this study is the use of student subjects. Many students may not have the financial means to perform some important personal activities on the WWW, such as making credit card purchases. However, of all users of the WWW, a very sizable portion (nearly 25%) falls in the age range of 18-25 (GVU Center, 1999). The subjects were also not selected at random, so the results are not generalizable to the population of university students, let alone the general population of WWW users. However, the goal of this study is not to uncover TAM relationships for a general population, but simply to do so for a particular target group.

To overcome these limitations, a random sample of all WWW users should be surveyed. A distinction should be made between the uses of the WWW for personal, as opposed to work-related, reasons. The logistics of obtaining such a random sample could, however, be formidable, but worthwhile. A follow-up study on the personal use of the WWW would be able to confirm or refute the results obtained in this research.

The TAM could also be applied to the use of specific Web sites or to specific Internet hardware configurations. Such research could highlight conditions of good or poor site design, or the effect of connection speed on acceptance.

CONCLUSION

A survey of n=295 university students, who use the WWW primarily for personal reasons, revealed that the dynamics of WWW acceptance are very different from those of the acceptance of end user systems utilized for work-related purposes. Previous research involving the technology acceptance model (TAM) (Davis, 1989) has found that the perceived usefulness of a system has a strong effect on actual use and that perceived ease of use has little direct impact on actual use, although perceived ease of use wields moderate influence over perceived usefulness. In the case of personal use of the WWW, perceived ease of use has a moderate effect on actual use and a very strong effect on perceived usefulness. However, perceived usefulness has virtually no effect on actual use of the WWW. The major implication of these findings is that designers and developers in the field should strive to dramatically improve the perceived ease of use of the WWW to boost actual use in the future. This will also be increasingly important as users of the WWW begin to develop an enhanced perception of its usefulness in terms of e-commerce. Researchers should also realize that the dynamics of technology acceptance might depend heavily on the primary purpose of its application.

APPENDIX I

Davis' (1989) U and EOU Items ("X" represents the target system)

Perceived Usefulness (U) Items Perceived Ease of Use (EOU) Items

Using X in my job would enable me Learning to operate X would be
to accomplish tasks more quickly. easy for me.

Using X would improve my job I would find it easy to get X to
performance. do what I want it to do.

Using X in my job would increase My interaction with X would be
my productivity. clear and understandable.

Using X would enhance my I would find X to be flexible to
effectiveness on the job. interact with.

Using X would make it easier to do It would be easy for me to become
my job. skillful at using X.

I would find X useful in my job. I would find X easy to use.

APPENDIX II

[ILLUSTRATION OMITTED]

REFERENCES

Bentler, P.M., and Bonnet, D.G. (1980). Significance tests and goodness of fit in the analysis of covariance structures. Psychological Bulletin, 88, 588-606.

Bollen, K.A. (1989). Structural equations with latent variables. Wiley (New York).

Lieb, J. (1999). 1999 U.S./Canada Internet Demographic Study. Retrieved July 21, 2000 from the World Wide Web: http: //www.commerce.net/research/stats/analysis/99-USCandaStudy.pdf.

Davis, F.D. (1989). Perceived usefulness, perceived ease of use, and user acceptance of information technology. MIS Quarterly, 13:3 (September), 319-340.

Graphic, Visualization, & Usability Center (1999). GVU's 10th WWW user survey. Retrieved November 10, 1999 from the World Wide WWW: http://www. cc.gatech. edu/ gvu/user_surveys/survey-1998-10.

Hair, J.F., Jr., Anderson, R.E., Tatham, R.L. and Black, W.C. (1992). Multivariate data analysis with readings. Macmillan (New York).

Hoffman, D., Kalsbeek, W. & Novak, T. (1999). Internet and WWW use in the United States: baselines for commercial development." Retrieved November 10, 1999 from the World Wide Web: http:// ecommerce.vanderbilt.edu/baseline/Internet.demos.july9.1996.html.

Joreskog, K. and Sorbom, D. (1993). LISREL 8: structural equation modeling with the SIMPLIS command language. Erlbaum (Hillsdale, NJ).

Overton, R. (2000). Broadband or bust. PC World, 18:5 (May), 102-110.

Pastore, M. (1999). Global ad spending: $33 billion by 2004. Retrieved July 21, 2000 from the World Wide Web: http:// adres.internet.com/stories/print/0,1401,7561_183641,00.html.

Sharma, S. (1996). Applied multivariate techniques. Wiley (New York).

Steers, R. and Porter, L. (1983). Motivation and Work Behavior, McGraw-Hill (New York).

Richard A. Johnson, Southwest Missouri State University

Michael A. Hignite, Southwest Missouri State University

Table 1
Goodness-of-fit Indicators and Recommended Cutoff Values

Goodness of fit (GFI) > 0.90
Adjusted goodness of fit (AGFI) > 0.80
Normed fit index (NFI) > 0.90
Comparative fit index (CFI) > 0.90
Incremental fit index (IFI) > 0.90
Root mean square residual (RMR) < 0.05

Table 2
Offending Estimates in U/EOU Measurement Model

Items Removed (in order) Rationale for Removal

12. It was easy for me to become The concept of "skill" may not be
skillful at using the WWW. especially germane in the context
 of personal WWW use.

8. Using the WWW makes it easier Originally intended as a U item,
to perform my favorite activity. the term "easier" may cause
 confusion with EOU issues.

5. I find the WWW flexible to The "flexibility" concept has been
interact with. noted as problematic in previous
 research.

2. Using the WWW enables me to The WWW is notorious for poor
perform my favorite activity more performance, especially when using
quickly. slow modems.