Hyperlinking unaudited information to audited financial statements: effects on investor...

I. INTRODUCTION

This study investigates whether firms can influence investors' perceptions of their financial reports by hyperlinking information in the audited financial statements to unaudited information. (1) Specifically, I examine whether hyperlinking optimistic, unaudited information

about a firm's prospects to audited financial statements that reflect industry-average performance leads investors to (1) misclassify unaudited information as audited, (2) inflate the credibility of the unaudited information, and (3) judge the firm's earnings potential to be higher, relative to viewing the same information in a hardcopy (paper) format. I also test whether a simple disclosure rule reduces a firm's ability to manage user impressions through hyperlinks.

As the number of investors who conduct research and trade online grows (J. D. Power and Associates 2001; Brooker 1998), the demand for financial information on the World Wide Web (the Web) increases. A recent Financial Accounting Standards Board (FASB) survey reveals that 99 of the Fortune 100 companies have web sites, and that 93 of those firms include financial information on their sites (FASB 2000). Although the increased availability of information through the Web has helped to "promote transparency, liquidity, and efficiency in our capital markets" (SEC 2000, 2), the Securities and Exchange Commission (SEC) and the FASB recognize the potential drawbacks of disseminating financial information electronically (SEC 1995, 2000; FASB 2000). For example, the SEC's Division of Enforcement recognizes that hyperlinking materials on the Web can mislead investors:

   ... even a link can serve as an aid for an investment ruse. Just like
   visiting a site, using a link is usually free, so the fraudster might even
   provide a link to the home page of the SEC next to a representation that a
   particular security has received "approval" from the SEC or a link to a
   phony investment newsletter page which touts the investment as a tremendous
   investment opportunity. (Cella and Stark 1997, 822)

The SEC's enforcement action against a firm that linked fictitious investment information to its web site illustrates the SEC's concern about the misuse of hyperlinks to mislead investors (SEC v. Internet Solutions for Business, Inc., and Lawrence Shaw 2001). In addition, the SEC specifically forbids hyperlinking an analyst's research report to a company's preliminary prospectus, stating that "[t]his direct and quick access to [the] research report would be similar to the Company including the paper version of the research report in the same envelope that it is using to mail the paper version of the preliminary prospectus to potential investors" (SEC 1995, 1353; see also SEC 2000). As of March 2001, the SEC had instigated more than 200 web-related enforcement actions, and has established an office dedicated exclusively to web surveillance and enforcement (SEC 2001).

Standard setters are also concerned about how firms use the Web to communicate with investors (Simpson 1998; Upton 1998; FASB 2000). The Electronic Distribution of Business Reporting Information section of the FASB's Business Reporting Research Project states that "a company may inadvertently give visitors the impression that all information provided in other Web sites to which the Web site is linked is afforded the same level of accuracy and reliability" (FASB 2000, 3). To gain a better understanding of how investors might use web-based financial information, the FASB initiated the "FauxCom" project. (2) According to Wayne Upton, senior project manager of FauxCom at the FASB:

   There is no doubt that the Internet has changed the delivery of financial
   information.... [It] offers companies an opportunity to knit the several
   sections of an annual report together, to integrate them, by providing
   "hyperlinks." (Upton 1998, 4-5)

To provide the American Institute of Certified Public Accountants (AICPA) information on how CPAs view electronically published financial statements, the Illinois CPA Society surveyed its members. Respondents overwhelmingly (79 percent) support the development of AICPA standards governing the electronic presentation of financial statements, and believe that CPAs should review computer-based information covered by the audit report (71 percent). (3) Of the participants who were concerned about clients hyperlinking audited financial statements with other documents, 55 percent stated that doing so could confuse users (IACPA 1998).

I conduct an experiment to examine the effects on investor judgments of hyperlinking unaudited information to a firm's audited financial statements. In my experiment, graduate business students, acting as potential investors, assessed the earnings potential of a firm by evaluating the firm's audited financial statements and a subsequent optimistic unaudited letter to shareholders from the firm's management (subsequently referred to as "the letter"). Participants viewed the materials either on the Web with the option of using hyperlinks to navigate through the materials, or in a hardcopy format. I informed participants in both settings that the financial statements and letter were distinct documents, and that the financial statements were audited while the letter was not.

Using hyperlinks to quickly jump from one document to another tends to blend information from the documents together, giving users the impression that all of the information originated from one comprehensive document. I therefore expect investors in a web-based environment to have difficulty recalling whether information originated from the audited or the unaudited report. In addition, I expect web-based investors to misclassify more unaudited information as audited than the reverse. Investors who use the audited financial statements as their frame of reference when forming initial beliefs about a firm's financial condition are more likely to believe that the hyperlinked unaudited information originated from the audited statements than the reverse.

Mistakenly believing that information in the unaudited letter originated from the audited financial statements, when the statements are considered more credible than the letter, will inflate investors' credibility assessments of the unaudited information in the letter. When the unaudited information in the letter is optimistic relative to the audited information, inflating its credibility will increase investors' perceptions of the firm's earnings potential. Finally, if investors are unable to clearly distinguish the origin of information in hyperlinked environments, then a disclosure rule that requires firms to distinguish audited from unaudited information (i.e., an "AUDITED/NOT AUDITED" label) should attenuate the effects outlined above.

Consistent with expectations, participants who viewed hyperlinked materials on the Web misclassified more unaudited information as audited, and assessed the unaudited information as more credible, than did participants who viewed hardcopy materials. Those participants who assessed the unaudited information as more credible also judged the firm's earnings potential to be higher. In general, results also suggest that clearly marking information as "AUDITED" or "NOT AUDITED" attenuated these effects. This evidence suggests that firms can use hyperlinks to influence financial-report users' perceptions, but that a notification aid diminishes this influence.

II. THEORY AND HYPOTHESES

Effect of Hyperlinking Information on Source Identification

In a hardcopy environment, when a firm issues a letter to shareholders after it issues its audited financial statements, the two documents are physically distinct. Distinctiveness helps individuals recall information accurately (Hunt and Elliott 1980; Clements and Wolfe 1997) and, hence, should help investors distinguish information presented in the audited financial statements from information presented in a separate unaudited letter to shareholders. Hyperlinking the two reports in a web-based environment, however, blends the documents into a cohesive network of information (Wright 1991; Mathew 1997) and eliminates the physical distinctiveness of the documents, making it more difficult for investors to identify correctly the origin of a given piece of information (Edwards and Hardman 1989; Foss 1989; McKnight et al. 1990). This blending can take one of two forms: (1) two distinct documents can blend equally together, or (2) one of the documents can blend into the other document (i.e., can be perceived as a subset). With respect to the latter form, when one of the two documents serves as a user's frame of reference, it is more likely to subsume the information in the other document than the reverse. (4)

Audited financial statements provide a common ground for investors to compare firms within, or across, time periods. In contrast, managers exercise discretion over the disclosure of additional unaudited financial information, and these voluntary disclosures vary across firms and over time. Therefore, I assume that investors evaluating a firm for the first time perceive the firm's audited financial statements as the baseline document and any additional, unaudited information as peripheral. If the audited report is their frame of reference, then investors are more likely to mistakenly recall a given piece of information as originating in the audited financial statements rather than in the unaudited letter.

These arguments suggest that investors evaluating a set of audited financial statements on the Web with hyperlinks to unaudited information will misclassify more unaudited information as audited, than the reverse. They also suggest that web-based investors will misclassify more unaudited information as audited than will investors who evaluate two physically distinct hardcopy documents.

H1a: Investors who view unaudited information hyperlinked to audited financial statements will misclassify more unaudited information as audited than they will misclassify audited information as unaudited.

H1b: Investors who view unaudited information hyperlinked to audited financial statements will misclassify more unaudited information as audited than will investors who view physically distinct hardcopy documents.

Effect of Misclassifying Information on Credibility Beliefs

Hyperlinking unaudited information to audited financial statements increases the likelihood that investors will mistakenly believe that the unaudited information originated from the audited statements. Prior research shows that investors generally consider audited information more credible than unaudited information (Winters 1975; Libby 1979; Reckers and Pany 1979; Pany and Smith 1982; Johnson et al. 1983; Epstein and Pava 1993). I therefore expect investors who misclassify unaudited information as audited to inflate the credibility of the unaudited information. Since hyperlinks facilitate misclassifying unaudited information as audited, I also expect web-based investors to believe that the unaudited information is more credible than will investors who receive hardcopy documents. Investors who have inflated beliefs about the credibility of the unaudited information, will in turn have inflated beliefs about the credibility of the entire information set. I refer to this phenomenon as a "credibility inflation effect."

H2a: The more unaudited information investors misclassify as audited, the higher they will assess the credibility of the entire information set.

H2b: Investors who view unaudited information hyperlinked to audited financial statements will assess the entire information set as more credible than will investors who view physically distinct hardcopy documents.

Effect of Credibility Inflation on Judgments of Earnings Potential

King et al. (1990) suggest that, ceteris paribus, investors should weight information in proportion to its perceived credibility. Accounting research (e.g., Hirst et al. 1995; Maines 1996; Hirst et al. 1999) and research in psychology (e.g., Birnbaum et al. 1976; Birnbaum and Stegner 1979; Anderson 1991; Eagly and Chaiken 1993) support this contention and find that people place more (less) weight on information from more (less) credible sources.

These arguments suggest that investors who consider audited information more credible than unaudited information should weight the audited information more heavily relative to the unaudited information when making judgments. (5) When these investors misclassify unaudited information as audited (H1a), thereby inflating the credibility of the unaudited information (H2a), I expect they will place more weight on the unaudited information when judging a firm's earnings potential. When the unaudited information is optimistic relative to the audited information, placing more weight on it should increase investors' judgments of the firm's earnings potential.

H3a: The more credible investors assess the entire information set to be, the higher they will judge the firm's earnings potential.

H3b: Investors who view optimistic unaudited information hyperlinked to audited financial statements will judge the firm's earnings potential as higher than will investors who view physically distinct hardcopy documents.

Effect of a Notification Aid

Psychology research shows that distinctiveness helps individuals accurately recall information (Hyde and Jenkins 1969; Clements and Wolfe 1997). Hyperlinks reduce distinctiveness by blending documents together (Wright 1991; Mathew 1997). I propose that alerting investors with an "AUDITED/NOT AUDITED" label as they navigate between the audited and unaudited information will mitigate this blending effect and restore the distinctiveness lost by hyperlinking the documents. (6) Specifically, I investigate whether an "AUDITED/NOT AUDITED" label reduces misclassification, credibility inflation, and the resulting increase in assessments of a firm's earnings potential.

H4a: Investors who view hyperlinked documents with a notification aid will misclassify less unaudited information as audited than will investors who view hyperlinked documents without a notification aid.

H4b: Investors who view hyperlinked documents with a notification aid will assess the entire information set as less credible than will investors who view hyperlinked documents without a notification aid.

H4c: Investors who view hyperlinked documents with a notification aid will assess the firm's earnings potential as lower than will investors who view hyperlinked documents without a notification aid.

III. EXPERIMENT

Participants

Online traders are typically more open to new technologies, more self-directed and aggressive, have higher incomes and assets, are younger, and are more highly educated than their counterparts who do not trade online (Fidelity Investments 2000). Graduate business students possess many of these characteristics: typically, they are self-motivated, they work with and have an understanding of financial statements and the role of auditing, they use the Web to retrieve information, and they are interested in making their own investment decisions. (7) Thus, in my experiment I use graduate business students as surrogates for online investors.

Forty-seven M.B.A. students from a large state university participated in the experiment. On average, participants had completed four accounting and three finance courses. Ninety-two percent of the participants plan to invest in a company's stock in the next five years, and 64 percent of those who plan to invest indicated that they intend to make the investment themselves rather than through a broker. Ninety-eight percent of the participants had previously evaluated a company's performance by analyzing financial statements. Each participant earned a flat wage of $10 for completing the experiment, and all participants who completed the materials were entered into a random drawing for $100.

Design

I used a 2 x 1 between-subjects design with a benchmark. The levels of the independent variable were "WEB--with links" and "WEB--with aid." The benchmark was a "HARDCOPY--no links" condition that mirrors firms' current practice of providing investors with physically distinct hardcopy financial reports.

Participants received a firm's audited financial statements and an unaudited letter from management that contained pro forma financial information. In both the web and hardcopy conditions, I informed participants that the financial statements and the letter were distinct documents, and that the financial statements were audited while the letter was not. In the WEB--with links condition, the unaudited information was hyperlinked to the audited financial statements. The WEB--with aid condition was identical to the WEB--with links condition, except that each page of information was labeled "AUDITED" or "NOT AUDITED." Exhibit 1 illustrates the WEB--with aid interface.

All participants received a case describing Precision, Inc., a firm in the navigation and guidance technology industry (SIC 3812). Participants in the HARDCOPY--no links condition received all materials in a hardcopy (paper-based) format. Participants in the two web-based conditions viewed all materials on computers.

The materials, in the order presented, included instructions for completing the case; a letter of introduction outlining the study's purpose, sequence, and expected duration; background information on Precision's business; Precision's audited financial statements; and a letter to shareholders from Precision's management discussing the company's prospects for the next three years. (8) The letter to shareholders was an optimistic one-page description of Precision's projected future performance, followed by pro forma balance sheets, income statements, and three-year forecasts of selected financial ratios. In the letter, Precision announced a major expansion that was expected to double the company's manufacturing and distribution capacity and significantly increase sales. Precision's audited financial statements indicated average financial condition and performance over the last three years as defined by Dun & Bradstreet's 14 key financial ratios. Participants used the information provided to evaluate Precision's earnings potential over the next three years.

I examine three dependent variables--classification error rates, assessments of information credibility, and judgments of earnings potential. Participants identified whether each of nine pieces of information originated from the financial statements or from the letter to shareholders/pro forma statements. The nine pieces of information shared a common trait: each could have been discussed in the financial statements or in the letter. (9) A participant's classification error rate is the percentage of audited and unaudited pieces of information that he or she misclassified.

Participants assessed credibility by responding to the question, "For the purpose of judging Precision's earnings potential, the information provided was --," using an 11-point scale with endpoints labeled "not credible/very credible." Similarly, participants assessed earnings potential by responding to "I believe Precision's earnings potential over the next three years is --," using an 11-point scale with endpoints labeled "very low/very high."

Procedure

I conducted the experiment in two settings. Forty-one percent of the participants completed the experiment in a controlled classroom setting ("in-lab" participants). In the controlled setting, participants in the web-based conditions completed the materials in a computer lab, and participants in the hardcopy condition completed the materials in a regular classroom. The remaining participants (59 percent) completed the experiment at their convenience ("out-of-lab" participants). Out-of-lab participants in the web-based conditions and the hardcopy condition received the same experimental materials as their in-lab counterparts. The ratio of in-lab to out-of-lab participants for each condition was WEB--with links, 9/9, WEB--with aid, 7/6, HARDCOPY--no links, 5/11.

To examine whether in-lab and out-of-lab participants completed the materials in a similar manner (i.e., in one sitting and without looking back through the materials when answering questions) I compared the average amount of time each group spent completing the materials, and each group's average classification error rate. If out-of-lab participants did not complete the materials in one sitting, then the average amount of time they spent completing the materials should be greater than that of the in-lab participants. If out-of-lab participants looked back through the materials while completing the classification task, then their classification error rates should be lower than the error rates of in-lab participants. Results reflect no differences across these two measures. In-lab (out-of-lab) participants took 41 (38) minutes on average to complete the case ([t.sub.45] = 0.89, p = 0.376, two-tailed), and in-lab (out-of-lab) participants on average misclassified 22 percent (24 percent) of the information ([t.sub.44] = 0.42, p = 0.680, two-tailed). Thus, I conclude that both in-lab and out-of-lab participants followed the directions in completing the materials.

Participants in the hardcopy condition received three envelopes: one containing instructions, an introductory letter, and the case; one containing the main questionnaire; and one containing the post-experiment questionnaire. After evaluating the materials, participants in the hardcopy condition placed them back in Envelope 1. They then opened Envelope 2 and completed the main questionnaire by providing their earnings potential judgments and completing the information classification task. After completing this questionnaire and returning it to Envelope 2, they opened Envelope 3 and completed the post-experiment questionnaire. The post-experiment questionnaire asked participants how useful and credible they found the information, and how objective Precision's management was in creating the audited and unaudited reports. It also contained several manipulation checks and concluded with a demographic survey.

Participants in the two web-based conditions received a single piece of paper listing a web address for the case. After typing in the address they viewed the same instructions, letter of introduction, and case that hardcopy participants received. After evaluating the materials, participants in the two web-based conditions clicked on a "Ready to Begin Questionnaire" button that presented them with the main questionnaire. After completing this questionnaire, participants sent their responses to me via email by clicking on a button labeled "Submit Questionnaire." Participants in the web-based conditions then clicked on a "Questionnaire #2" button that presented them with the post-experiment questionnaire, which they completed and submitted in the same manner.

IV. RESULTS

Manipulation Checks

Data collected in the post-experiment questionnaire reveal that all participants recognized that Precision's financial statements were audited and that the letter was not. Participants also reported that they weighted the information in the financial statements more heavily relative to the information in the letter when making their judgments ([t.sub.46] = 6.59, p < 0.001). ANOVA results indicate that these self-reported weights do not differ among the three experimental conditions ([F.sub.1,44] = 0.19, p = 0.831). In addition, participants across conditions judged the unaudited information in the letter/pro forma statements as more optimistic than the audited information in the financial statements ([t.sub.45] = 10.10, p < 0.001). Finally, the data reflect no significant differences among conditions in the number of accounting courses taken ([F.sub.1,43] = 0.57, p = 0.571), the number of finance courses taken ([F.sub.1,43] = 1.43, p = 0.251), or the number of times that participants had evaluated a set of financial statements ([F.sub.1,44] = 1.11, p = 0.339).

Hypotheses Tests

Panel A of Table 1 reports the descriptive statistics for the three dependent measures. (10)

Classification Error Rates

Participants classified whether each of nine pieces of information originated from the financial statements or the letter to shareholders. A programming error rendered data from two of the nine classification questions useless. The hypotheses tests reported in this section rely on data from the remaining seven questions. (11)

Hypothesis 1a predicts that participants in the WEB--with links condition will misclassify more unaudited information as audited than the reverse. Panel B of Table 1 reports that participants in the WEB--with links condition mistakenly classified 37 percent of the unaudited items as audited and only 22 percent of the audited items as unaudited ([t.sub.17] = 1.90, p = 0.038, one-tailed). These results support H1a, and suggest that hyperlinking the unaudited and audited information led investors to blend the information in the unaudited letter into the audited financial statements, rather than the reverse. Additional analysis reveals no significant difference in the direction of classification errors committed by participants in the HARDCOPY--no links condition--they mistakenly classified 22 percent of the unaudited items as audited, and 15 percent of the audited items as unaudited ([t.sub.14] = 0.93, p = 0.367, two-tailed).

Hypothesis 1b predicts that hyperlinking the audited and unaudited information will lead participants in the WEB--with links condition to misclassify more unaudited information as audited than participants in the HARDCOPY--no links condition. Panel C of Table 1 reports that participants in the WEB--with links condition misclassified more unaudited information as audited than did participants in the HARDCOPY--no links condition (37 percent vs. 22 percent, [t.sub.31] = 1.62, p = 0.058, one-tailed). These results support H1b, and suggest that hyperlinking the unaudited and audited information made it more difficult for participants to classify the unaudited information correctly.

Information Credibility

Hypothesis 2a predicts that the more unaudited information participants misclassify as audited, the higher they will assess the credibility of the entire information set. Results reported in Panel D of Table 1 support H2a [r.sub.45] = 0.25, p = 0.047, one-tailed). Panel E of Table 1 reports that participants in the WEB--with links condition assessed the entire information set as more credible than did participants in the HARDCOPY--no links condition (6.47 vs. 5.13, [t.sub.31] = 2.23, p = 0.017, one-tailed), consistent with H2b.

Immediately following the overall credibility question, Participants rated how objective Precision's management was in creating the financial statements and the letter on 11-point scales, with endpoints labeled "not at all objective/very objective." If participants' frame of reference was the audited financial statements, then participants in the WEB--with links and HARDCOPY--no links conditions should perceive management as equally objective in creating the statements. If a credibility inflation effect exists, then participants in the WEB--with links condition should perceive management as more objective in creating the letter than participants in the HARDCOPY--no links condition. Untabulated results reveal that participants in the WEB--with links condition believed that Precision's management was more objective in creating the unaudited letter than did participants in the HARDCOPY--no links condition ([t.sub.32] = 2.49, p = 0.009, one-tailed). In contrast, participants in the WEB--with links and HARDCOPY--no links conditions did not have different beliefs about how objective management was in creating the audited financial statements ([t.sub.32] = 0.18, p = 0.855, two-tailed). These results suggest that hyperlinking the two documents inflated the perceived credibility of the information in the unaudited letter, without impairing the credibility of the information in the financial statements.

Earnings Potential Judgments

Panel F of Table 1 reports evidence supporting H3a's prediction that the more credible participants believe the entire information set to be, the higher they will judge the firm's earnings potential ([r.sub.46] = 0.25, p = 0.048, one-tailed). (12) Hypothesis 3b predicts that participants in the WEB--with links condition will judge the firm's earnings potential as higher than will participants in the HARDCOPY--no links condition. Panel G of Table 1 reports that the earnings-potential judgments of WEB--with links participants are slightly higher than those of HARDCOPY--no links participants (6.17 vs. 5.56, [t.sub.32] = 1.26, p = 0.108, one-tailed). Hypothesis 3b is at best marginally supported.

Notification Aid

Hypotheses 4a through 4c predict that notifying participants with an "AUDITED/NOT AUDITED" label as they navigate between the audited and unaudited information will reduce misclassification, credibility inflation, and the resulting increase in assessments of a firm's earnings potential. To examine these predictions, I compare the responses of participants in the WEB--with aid condition to those in the WEB--with links condition. If the notification aid is effective, then participants in the WEB--with aid condition should misclassify fewer unaudited items as audited than will participants in the WEB--with links condition, and their credibility and earnings-potential judgments should be less than the credibility and earnings potential judgments of participants in the WEB--with links condition.

Table 2 reports that participants in the WEB--with aid condition did not misclassify significantly fewer items of unaudited information as audited than did participants in the WEB--with links condition (33 percent vs. 37 percent, [t.sub.29] = 0.32, p = 0.374, one-tailed). These results do not support H4a. In contrast, untabulated results reveal that the notification aid was effective in helping participants correctly classify the audited information: participants in the WEB--with aid condition misclassified only 14 percent of the audited information as unaudited, whereas participants in the WEB--with links condition misclassified 22 percent of the audited information as unaudited ([t.sub.29] = 1.44, p = 0.081, one-tailed).

Table 2 shows that participants in the WEB--with aid condition assessed credibility as significantly lower than did participants in the WEB--with links condition (4.69 vs. 6.47, [t.sub.28] = 2.57, p = 0.008, one-tailed), consistent with H4b. Similarly, Table 2 reports that the earnings-potential judgments of participants in the WEB--with aid condition are significantly lower those of participants in the WEB--with links condition (5.00 vs. 6.17, [t.sub.29] = 2.06, p = 0.025, one-tailed), consistent with H4c.

One possible explanation for why results support H4b and H4c, but not H4a, is that I necessarily limited the classification task to a small number of potentially relevant pieces of information. The pattern of results for H4a through H4c could result if the notification aid helped individuals correctly identify enough unaudited items to influence their subsequent judgments, but the classification task did not include these items. It is also possible that a stronger notification aid, such as the example presented in Exhibit 2, might have been more effective in reducing participants' classification errors.

V. CONCLUSIONS, IMPLICATIONS, AND LIMITATIONS

Eighty-five percent of all participants responded in the post-experiment questionnaire that their judgments would not depend on whether they viewed financial information on the Web or in hardcopy format. However, this study's empirical results contradict these beliefs and indicate that participants who viewed hyperlinked materials on the Web misclassified more optimistic, unaudited information as audited and believed the optimistic, unaudited information to be more credible than did participants who viewed hardcopy materials. Results also indicate that inflating the credibility of the optimistic unaudited information increased participants' judgments of the firm's earnings potential. Finally, results indicate that notifying web-based participants with an "AUDITED/NOT AUDITED" label as they navigated between the audited and unaudited information reduced their credibility assessments and their earnings-potential judgments.

The AICPA and the Canadian Institute of Chartered Accountants jointly developed a service called WebTrust, aimed at providing assurance for electronic commerce. My results suggest that a similar service aimed at providing assurance to financial-report users that a firm's web-based financial disclosures meet certain criteria, such as explicitly labeling, or prohibiting, direct links between audited and unaudited information, might benefit investors.

This study is subject to several limitations. First, I limit the amount of information participants receive to a subset of that available on most corporate web sites so that participants can complete the task in a reasonable amount of time. Most firms' web sites are more complex than my experimental setting, with multiple links between and within financial and nonfinancial information. Although this reduced complexity biases against my predictions, it also diminishes the generalizability of my findings. Future research could examine how the complexity of hyperlinked environments affects investor judgments.

Second, due to low sample size, this study does not disentangle how much of the effect of presenting hyperlinked information on the Web is due to presenting information on a computer screen vs. presenting it in hardcopy form, and how much is due to hyperlinking the information in a web-based environment.

Third, because 59 percent of participants completed the experiment in an "out-of-lab" setting, I cannot be completely sure that participants complied with my instructions to complete the case in one sitting and to refrain from looking back through the case materials when answering questions. Although I did not detect evidence that participants violated these instructions, my out-of-lab data-gathering technique cannot eliminate the possibility that uncontrolled, extraneous factors affected my results.

Fourth, I use graduate business students as surrogates for online investors. Although M.B.A. students exhibit demographic characteristics similar to those of online investors, many have limited investment experience and therefore may not reflect the beliefs, or use the analytical techniques, of actual investors. Despite these limitations, this study provides descriptive ex ante evidence to policymakers, auditors, and investors that firms can influence the perceptions of financial-report users by hyperlinking unaudited information to information in the audited financial statements, and that a practical, low-cost notification aid attenuates these influences.

TABLE 1
Participants' Classification Error Rates, Credibility Assessments,
and Earnings Potential Judgments

Panel A: Descriptive Statistics: Mean [standard deviation]

                                          Dependent Variable

                                Classifi-
                                 cation                      Earnings
                                  Error        Credibi-     Potential
Presentation Condition          Rates (a)      lity (b)         (b)

Web with links (n = 18) (c)                   6.47 [1.81]   6.17 [1.58]
  Unaudited error rate        37.0% [32.1%]
  Audited error rate          22.2% [16.9%]
  Overall error rate          28.6% [18.3%]

Web with aid (n = 13)                         4.69 [1.97]   5.00 [1.53]
  Unaudited error rate        33.3% [30.4%]
  Audited error rate          13.5% [16.5%]
  Overall error rate          22.0% [15.0%]

Hardcopy no links (n = 16)                    5.13 [1.65]   5.56 [1.14]
    (d)
  Unaudited error rate        22.2% [16.3%]
  Audited error rate          15.0% [22.8%]
  Overall error rate          18.1% [13.7%]

Panel B: Paired-Sample Comparison of Classification Error Rates within
the WEB--with Links Condition

                                         df   t-statistic   p-value (e)
H1a: Unaudited error rate (37.0%) vs.
Audited error rate (22.2%)               17       1.90        0.038

Panel C: Planned Comparison of Unaudited Classification Error Rates

                                         df   t-statistic   p-value (e)

H1b: WEB--with links (37.0%) vs. HARD-
COPY--no links (22.2%)                   31       1.62        0.058

Panel D: Correlation Analysis

                                             Pearson Correlation

                                         n         r        p-value (e)

H2a: Unaudited error rate and
credibility                              45       0.25        0.047

Panel E: Planned Comparisons of Credibility Assessments

                                         df   t-statistic   p-value (e)

H2b: WEB--with links (6.47) vs.
HARDCOPY -- no links (5.13)              31       2.23        0.017

Panel F: Correlation Analysis

                                              Pearson Correlation

                                          n         r       p-value (e)

H3a: Credibility and earnings
potential                                46       0.25        0.048

Panel G: Planned Comparison of Earnings Potential Judgments

                                         df   t-statistic   p-value (e)

H3b: WEB--with links (6.17) vs. HARD-
COPY--no links (5.56)                    32       1.26        0.108

(a) In the classification task, participants identified whether each of
seven pieces of information originated from the financial statements or
the letter to shareholders. Overall error rate is the percentage of
audited and unaudited pieces of information a participant
misclassified. Unaudited error rate is the percentage of items from the
letter to shareholders that a participant misclassified and audited
error rate is the percentage of items from the financial statements
that a participant misclassified.

(b) Participants provided their credibility assessments and earnings
potential judgments using 11-point scales with endpoints labeled "not
credible/very credible" and "very low/very high," respectively.

(c) Of the 18 participants in the WEB--with links condition, only 17
answered the credibility measure.

(d) Of the 16 participants in the HARDCOPY--no links condition, only
15 completed the classification task.

(e) p-values are one-tailed.
TABLE 2
The Effect of an AUDITED/NOT AUDITED Notification Aid on Participants'
Classification Error Rates, Credibility Assessments, and Earnings
Potential Judgments

Planned Comparisons of:                  df   t-statistic   p-value (a)

Unaudited Classification Error Rates
  H4a: WEB--with aid (33.3%) vs. WEB--   29      0.32          0.374
  with links (37.0%)

Credibility Assessments
  H4b: WEB--with aid (4.69) vs. WEB--    28      2.57          0.008
  with links (6.47)

Earnings Potential Judgments
  H4c: WEB--with aid (5.00) vs. WEB--    29      2.06          0.025
  with links (6.17)

(a) p-values are one-tailed.

This manuscript is based on my dissertation completed at Indiana University. I thank my committee members, Ed Hirt, Pat Hopkins, Jim Wahlen, and especially my chair, Jamie Pratt, for their invaluable guidance and support. I also thank the following for their helpful comments and suggestions: Steve Baginski, Michael Bamber, David Barrett, Walt Blacconiere, Bob Bowen, Tony Chen, Randy Elder, Karl Hackenbrack, Abigail Hodge, Ivo Jansen, Diane Janvrin, Kathryn Kadous, Jane Kennedy, Michael Kimbrough, Barbara Lougee, Laureen Maines, Roger Martin, Gary McGill, Shivaram Rajgopal, David Randolph, Jerry Salamon, Lisa Sedor, Jim Seida, Terry Shevlin, Nathan Stuart, Lynette Wood, an anonymous reviewer, and workshop participants at Arizona State University; the University of California. Irvine; Emory University; the University of Florida; Indiana University; and the University of Washington. I gratefully acknowledge the financial support of the Kelley School of Business, the Accounting Department, and the University Graduate School at Indiana University.

Submitted August 2000 Accepted May 2001

(1) A hyperlink functions as a "go to" command in electronic documents. When a user clicks on hyperlinked text (usually displayed in a different color [e.g., blue] from main text [e.g., black]), the computer displays the linked file. This command allows the user to quickly retrieve linked information from another section of the same report or from a report located anywhere on the Web.

(2) FauxCom is a fictitious company the AICPA's Special Committee on Financial Reporting (the Jenkins Committee) developed to illustrate its vision of a business-reporting package (AICPA 1994). The FASB staff took the Special Committee's vision a step further by using hyperlinks to integrate different sections of the package.

(3) Interpretation No. 4 of Statement of Auditing Standards (SAS) No. 8 states that "auditors are not required by AU Section 550 to read information contained in electronic sites, or to consider the consistency of other information in electronic sites with the original documents" (AICPA 1997, AU 9550.17). A legal consensus concerning auditor liability related to information in electronic sites has not been established (Pacini and Sinason 1999a, 1999b).

(4) This expectation is based on psychology research that shows new information is organized around familiar categories (Cohen 1981: Roskos-Ewoldsen and Fazio 1992; Smith et al. 1996).

(5) The weight individuals place on audited vs. unaudited information depends not only on the information's credibility, but also on its relevance. Audited information may be more credible, but less relevant (e.g., timely), than unaudited information--especially for high-tech firms. However, as reported in Section IV, data gathered from the post-experiment questionnaire support my assumption that when investors form initial beliefs, they assign more weight to audited information than to unaudited information.

(6) Intel's web site uses this type of design (see Exhibit 2), which the CPA Letter's Practice Alert 97-1 advocates (Quinn and Goldstein 1997).

(7) Maines (1990), Hirst et al. (1995), and Hirst et al. (1999) report similar demographic information for graduate business students, and show that these students are sensitive to information credibility differences.

(8) I did not manipulate the order of presentation due to the small number of participants in each condition. In a pilot study with 86 undergraduate accounting students, I manipulated the order in which participants received the financial statements and the letter to shareholders. The results of the pilot study are inferentially similar to the reported results, and do not differ based on the order of presentation.

(9) The nine were (1) a discussion of revenue recognition, (2) a discussion of navigation and guidance equipment, (3) industry ratios, (4) a quick ratio equal to the industry average, (5) a discussion of new products, (6) a debt-to-equity ratio above the industry average, (7) a discussion of inventory valuation, (8) a discussion of computer equipment, and (9) a dividend payment.

(10) I examined the effect of extreme values in two ways. First, I created box plots for the variables used in each test. There were no extreme values. Second, I computed Cook's distance, which measures how much the coefficients change when any particular observation is deleted. Cook and Weisberg (1982) suggest a value greater than I deserves closer scrutiny. Across all tests, the highest Cook's value was 0.23, which suggests that no single observation significantly influences the reported results. Thus, I conclude that outliers are not driving the reported results.

(11) The two excluded pieces of information were (1) a discussion of navigation and guidance equipment, and (2) a debt-to-equity ratio above the industry average. Of the seven remaining pieces of information, four were from the financial statements and related footnotes (a discussion of revenue recognition, industry ratios, a discussion of inventory valuation, and a dividend payment), and three were from the letter/pro forma statements (a quick ratio equal to the industry average, a discussion of new products, and a discussion of computer equipment).

(12) Consistent with the results of H3a, the more objective participants perceived the letter to be the higher they judged the firm's earnings potential ([r.sub.47] = 0.35, p = 0.007, one-tailed). Furthermore, mediation analysis suggests participants' beliefs about how objective management was in creating the unaudited letter accounts for the relation between how participants received the information (web vs. hardcopy) and their judgments of the firm's earnings potential. Results using participants' overall credibility assessments as the mediating variable are inconclusive.

REFERENCES

American Institute of Certified Public Accountants (AICPA). 1994. Report of the Special Committee on Financial Reporting: Improving Business Reporting--A Customer Focus (Jenkins Committee). New York, NY: AICPA.

--. 1997. Other Information in Documents Containing Audited Financial Statements. Interpretation No. 4 of Statement of Auditing Standards No. 8. New York, NY: AICPA.

Anderson, N. 1991. Contributions to Information Integration Theory. Volume II: Social. Hillsdale, NJ: Lawrence Erlbaum Associates.

Birnbaum, M., R. Wong, and L. Wong. 1976. Combining information from sources that vary in credibility. Memory and Cognition 4: 330-336.

--, and S. Stegner. 1979. Source credibility in social judgment: Bias, expertise, and the judge's point of view. Journal of Personality and Social Psychology 37: 48-74.

Brooker, K. 1998. Online investing: It's not just for geeks anymore. Fortune (December): 89+.

Cella, J., and J. Stark. 1997. SEC enforcement and the Internet: Meeting the challenge of the next millennium. The Business Lawyer 52 (May): 815-849.

Clements, C., and C. Wolfe. 1997. An experimental analysis of multimedia annual reports on nonexpert report users. Advances in Accounting Information Systems 5: 107-136.

Cohen, C. 1981. Person categories and social perception: Testing some boundaries of the processing effects of prior knowledge. Journal of Personality and Social Psychology 40: 441-452.

Cook, R., and S. Weisberg. 1982. Residuals and Influence in Regression. New York, NY: Chapman and Hall.

Eagly, A., and S. Chaiken. 1993. The Psychology of Attitudes. Orlando, FL: Harcourt Brace Jovanovich College Publishers.

Edwards, D., and L. Hardman. 1989. Lost in hyperspace: Cognitive mapping and navigation in a hypertext environment. In Hypertext: Theory into Practice, edited by R. McAleese, 105-125. Oxford, U.K.: Intellect.

Epstein, M., and M. Pava. 1993. The Shareholder's Use of Corporate Annual Reports. Greenwich, CT: JAI Press.

Fidelity Investments. 2000. Today's Wired Investor: The Attitudes and Behaviors of Online and Traditional Traders. Boston, MA: Fidelity.

Financial Accounting Standards Board (FASB). 2000. Business Reporting Research Project: Electronic Distribution of Business Reporting Information. Norwalk, CT: FASB.

Foss, C. 1989. Detecting Lost Users: Empirical Studies on Browsing Hypertext. INRIA Technical Report No. 972. Sophia-Antipolis, France: INRIA.

Hirst, D., L. Koonce, and P. Simko. 1995. Investor reactions to financial analysts' research reports. Journal of Accounting Research 33 (Autumn): 335-351.

--, --, and J. Miller. 1999. The joint effect of management's prior forecast accuracy and the form of its financial forecasts on investor judgment. Journal of Accounting Research 37 (Supplement): 101-124.

Hunt, R., and J. Elliott. 1980. The role of nonsemantic information in memory: Orthographic distinctiveness effects on retention. Journal of Experimental Psychology: General 109: 49-74.

Hyde, T., and J. Jenkins. 1969. The differential effects of incidental tasks on the organization of recall of a list of highly associated words. Journal of Experimental Psychology 82: 472-481.

Illinois CPA Society's Accounting and Assurance Services Network (IACPA), Steering Group. 1998. Summarized Findings of Survey Regarding Electronically Published Financial Statements. Chicago, IL: IACPA.

J. D. Power and Associates. 2001. March 2001 Online Trading Investor Satisfaction Study. Agoura Hills, CA: J. D. Power and Associates.

Johnson, D., K. Pany, and R. White. 1983. Audit reports and the loan decision: Actions and perceptions. AUDITING: A Journal of Practice & Theory, (Spring): 38-51.

King, R., G. Pownall, and G. Waymire. 1990. Expectations adjustment via timely management forecasts: Review, synthesis, and suggestions for future research. Journal of Accounting Literature 9: 113-144.

Libby, R. 1979. Bankers' and auditors' perceptions of the message communicated by the audit report. Journal of Accounting Research 17 (Spring): 99-122.

Maines, L. 1990. The effect of forecast redundancy on judgments of a consensus forecast's expected accuracy. Journal of Accounting Research 28 (Supplement): 29-47.

--. 1996. An experimental examination of subjective forecast combination. International Journal of Forecasting 12: 223-233.

Mathew, K. 1997. A comparison of the influence of interactive CD-ROM storybooks and traditional print storybooks on reading comprehension. Journal of Research on Computing in Education (Spring): 263-275.

McKnight, C., A. Dillon, and J. Richardson. 1990. A comparison of linear and hypertext formats in information retrieval. In Hypertext: State of the Art, edited by R. McAleese and C. Green, 10-19. Oxford, U.K.: Intellect.

Pacini, C., and D. Sinason. 1999a. Auditor liability for electronic commerce transaction assurance: The CPA/CA WebTrust. American Business Law Journal 36 (Spring): 479-514.

--, and --. 1999b. The law and CPA WebTrust. Journal of Accountancy 187 (February): 20-24.

Pany, K., and C. Smith. 1982. Auditor association with quarterly financial information: An empirical test. Journal of Accounting Research (Autumn): 472-481.

Quinn, P., and E. Goldstein. 1997. Members in public accounting firms--Practice Alert 97-1. CPA Letter 77 (January/February).

Reckers, P., and K. Pany. 1979. Quarterly statement reliability and auditor association. Journal of Accountancy (October): 97-100.

Roskos-Ewoldsen, D., and R. Fazio. 1992. On the orienting value of attitudes: Attitude accessibility as a determinant of an object's attraction of visual attention. Journal of Personality, and Social Psychology 63: 198-211.

Securities and Exchange Commission (SEC). 1995. Use of electronic media for delivery purposes. SEC Docket Volume 60, No. 8 (November 7). New York, NY: SEC.

--. 2000. SEC Interpretation: Use of Electronic Media. File number S7-11-00. New York, NY: SEC.

Securities and Exchange Commission v. Internet Solutions for Business, Inc., and Lawrence Shaw. Civil Action No. CDS-01-0225 (DH)(USDC)/District of Nevada, 2001.

Simpson, R. 1998. Author's notes from a presentation on business reporting research. FASB Ph.D. Program, Norwalk, CT, July.

Smith, E., R. Fazio, and M. Cejka. 1996. Accessible attitudes influence categorization of multiple categorizable objects. Journal of Personality and Social Psychology 71: 888-898.

Upton, W. 1998. The Internet and financial information--Is there a future here? Financial Accounting Series No. 183 (March): 4-6.

Winters, A. 1975. Banker perceptions of unaudited financial statements. The CPA Journal (August): 29-34.

Wright, P. 1991. Cognitive overheads and prostheses: Some issues in evaluating hypertexts. In Proceedings of the Third ACM Conference on Hypertext, edited by R. Furuta, and D. Stotts, 1-12. New York, NY: ACM Press.

Frank D. Hodge
University of Washington