Knowledge that Transforms

[Professional standards for analytical procedures call for auditors to hypothesize likely causes of unexpected patterns in financial statement balances and to develop plans to investigate (SAS 56). Therefore, audit efficiency and effectiveness depend on competency in recognizing patterns in financial data and in hypothesizing likely causes of those patterns to serve as a guide for further testing. Yet our knowledge of how well auditors accomplish these crucial steps has been limited by the difficulty of developing research tasks with criteria for evaluating auditor performance. The purpose of this study is to relate processes of pattern recognition and hypothesis generation to quality of performance in an analytical task. To accomplish this purpose, we conducted a laboratory study in which 21 auditors were asked to think aloud while performing an analytical procedures task. The case contained a seeded error that caused a fairly complex pattern of discrepancies between projected and unaudited financial ratios and balances. Think-aloud verbal protocols provided a trace of subjects' reasoning processes, and the seeded error provided an outcome criterion to evaluate the hypotheses they generated. This design supplies evidence needed to evaluate pattern recognition and hypothesis generation processes that lead to correct and incorrect outcomes. Results indicated that three auditors made acquisition errors and four failed to combine crucial cues into a pattern. Of the 14 auditors who recognized the pattern, six proposed a hypothesis consistent with the pattern. Thus, hypothesis generation was the stage at which process errors most frequently occurred, and the least experienced auditors had the most difficulty at that stage. Specific process problems that inhibited generation of a correct hypothesis included: addressing only part of a recognized pattern, and/or fixating on a certain error type. Thus, protocol analysis allowed insight into which stages in the process proved most difficult for the auditors and why, providing a focus for further research and decision aid development. This research responds to the call for inclusion of criterion outcomes in audit judgment research (e.g., Libby 1981; Ashton 1982; Biggs 1985; Davis and Solomon 1989). The importance of criterion outcomes can be seen in the advances made in decision research in medicine (e.g., Johnson et al. 1982; Patel and Groen 1986; Lesgold et al. 1988). and physics (e.g., Chi et al. 1982; Robertson 1990). In medicine especially, the ability to relate diagnostic processes to outcome quality has contributed to the theory and practice of medical decision making (Kassirer 1989) in ways that have been unavailable in auditing.]

[Prior research has investigated firms' success in the political process and its relation to firm size. For example, effective tax rates have been used to measure the extent to which firms are able to influence corporate income taxation. Effective tax rates, however, are a biased proxy of political success because they may contain a factor, net operating losses (NOLs), attributable to firms' operating results and experience rather than to success in the political process. Moreover, if firm size is also correlated with NOLs, then NOLs may affect the relationship between firm size and tax rates in a systematic manner. If such an effect is substantial, the association of firm size and effective tax rates will be a biased basis for drawing inferences about firm size and political success. The purpose of this article is to assess the extent of the effect of NOLs on the overall relationship of firm size and tax rates. Path analysis is employed to isolate the effect of firm size on effective tax rates that is attributable to NOLs (i.e., the indirect effect of NOLs). Determining the indirect effect of NOLs allows an assessment of the bias in the overall firm size/tax rate relation attributable to NOLs. To estimate the indirect effect of NOLs, a path model with two structural equations is built to identify the indirect links between firm size and effective tax rates mediated by NOLs, and the remaining direct link between firm size and effective tax rates. To correct errors-in-variables problems and to provide a consistent interpretation of the results, variables in the structural equations are modeled by a latent-variable approach (measurement model). The path model and the measurement model are estimated by LISREL. Based on the parameter estimates of the structural equations, the path analysis results indicate that the indirect effect of NOLs has a statistically significant impact on the overall relationship between firm size and effective tax rates. This indicates that NOLs constitute an important omitted variable in prior studies. While effective tax rates may measure some degree of firms' success in the political process, conclusions regarding the relation between firm size and firms' success in the political process is confounded by the serious omission of NOLs.]

[A prominent stream of audit judgment research employs the policy-capturing paradigm and uses analysis of variance (ANOVA) to construct mathematical models of individual auditors' evaluations of internal control. Based on these "as if" representations, it may appear that auditors evaluate control systems by independently processing individual controls as opposed to the pattern (or configuration) of control features. Indeed, with a few noteworthy exceptions, virtually no evidence has been reported of auditors configurally processing information while evaluating controls or assessing control risk. One such exception is a recent paper by Brown and Solomon (1990). Therein, it was suggested that judicious usage of context-and task-specific knowledge is necessary to guide the development of experiments capable of detecting configural information processing. The present study generalizes the concept that auditors will exhibit configural information processing in situations in which domain-specific knowledge implies that it is appropriate. Two experiments using auditor-subjects and investigating configural information processing are described. The task of the first experiment is assessment of account misstatement risk given evidence produced by specified substantive audit procedures; and the task of the second experiment is such assessment given the results of specified analytical audit planning procedures. Domain-specific knowledge suggests configural information processing strategies that will be manifest (within ANOVA) as a compensatory-form ordinal interaction for the former task and as a disordinal interaction for the latter task. Experimental results are consistent with expectations and are especially striking for the second experiment. Twenty of the second experiment's 22 auditor-subjects apparently used the predicted configural strategy, and ANOVA models of these auditors' risk assessments attributed on average about 40 percent of total judgment variation to the single predicted interaction of analytical procedure results (trial balance figure comparisons). An implication of these results, in concert with Brown and Solomon (1990), is that auditor information processing is more complex than has been reported by prior studies. That is, across a wide variety of tasks, many auditors apparently are able to configurally process information. The results also imply considerable between-task and between-auditor configural processing differences (magnitude and prevalence), which anecdotal evidence suggests is due to domain-specific factors. Two factors, the rate-of-substitution between information cues (substitutable audit procedures or controls) and when auditors learn the domain-specific knowledge requisite for configural information processing, seem particularly important and could be worthwhile foci for future research.]

[This study investigates whether concern about reporting favorable trends in accounting net income influences decisions to invest in research and development (R&D). Using data for 438 U.S. industrial firms during the period 1977-1987, the analysis indicates that relative R&D spending is significantly less when spending jeopardizes the ability to report positive or increasing income in the current period. Unlike related studies that consider choices among alternative accounting practices (DeAngelo 1986; Healy 1985; Liberty and Zimmerman 1986; McNichols and Wilson 1989; Moses 1987), this study focuses on investment decisions as instruments for achieving income objectives. In most instances, choices among accounting practices have no direct cash flow consequences, but changes in R&D spending to satisfy current-period income objectives do alter cash flows. The evidence is consistent with assertions that U.S. manufacturing firms are not competitive internationally in part because U.S. managers are overly concerned about how their R&D investment decisions affect current-period earnings (see, e.g., Markoff 1990). We analyze capital spending to determine whether these R&D reductions reflect differences in investment opportunities or incentives rather than differences in accounting treatment. Unlike the results for R&D, differences in capitalized costs among firms in the sample are not statistically significant. We also consider whether the results can be attributed to accounting-based compensation arrangements. For this purpose, we examine a sample of firms with no accounting-based management compensation contracts in effect during the test period. Results are comparable to those obtained for the entire sample, which is inconsistent with an explanation that the observed differences in R&D spending are attributable only to explicit compensation arrangements. Our results are consistent with conclusions that compliance with SFAS No. 2 (FASB 1974) discouraged investment in R&D (Elliott et. al. 1984; Horwitz and Kolodny 1980). That is, the evidence suggests that managers are more likely to consider current-period income effects when making R&D decisions than when making capital-spending decisions, whose costs are amortized over a number of accounting periods.]

[Decision makers refer to their long-term memory to test the implications of evidence about a current problem (Birnberg and Shields 1984; Libby 1989). Given this reliance on long-term memory, biases in retrieval of previously encountered information may be an important source of decision error (Libby 1989), and differences in such biases may be one explanation for differences in auditor judgment performance across experience levels. The present study adopts a schema-based framework to examine some differences in the knowledge structures and judgments of experienced and inexperienced auditors and the relationship between these knowledge structures and judgments. The study examines the recall of typical and atypical information by experienced and inexperienced auditors within the context of a going-concern situation and then relates this measure of memory to the inferences and predictive judgments made by these auditors. Three experiments were conducted. In experiment 1, auditors read a description of a company that the audit partner-in-charge had suggested may have a going-concern problem. The description consisted of items that are considered typical of a company with going-concern problems, atypical items, and filler items. After an intervening period with a distractor task, all subjects were given a recall test, were asked to infer the likelihood of certain previously unstated items being true, and to estimate the probability that the firm would fail within a year. The first of six main findings showed that experienced auditors recalled more atypical items than inexperienced auditors, but there were no differences in the number of typical items recalled. Second, experienced auditors recalled more atypical than typical items, whereas inexperienced auditors did not. Third, experienced auditors were more likely than inexperienced auditors to infer that previously unstated atypical items were true. Fourth, for both experienced and inexperienced auditors, the ratio of atypical to typical items recalled was positively correlated with the inferences made, and the inferences were negatively correlated with the predictive judgments. This last correlation was much higher for the experienced than for the inexperienced auditors. Fifth, there was no direct relationship between recall and predictive judgments. Sixth, clustering of recall on the basis of atypical/typical items was significantly higher for experienced than for inexperienced auditors and was significantly correlated with inferences for experienced auditors only. In experiments 2 and 3, we collected additional data to examine some validity threats related to the first experiment. In experiment 2, we examined the relationship between recall and judgments, using audit managers who had worked on at least one audit with going-concern as an issue. We found results similar to those of experiment 1. In experiment 3, experienced auditors performed the recall and predictive judgments without the intervening inferences task. This provided a more direct test of the relationship between recall and predictive judgments. Again, no relationship was found.]

[An important differentiating attribute of the audit product is believed to be the credibility that the auditor is perceived to bring to an audit engagement. This study uses the context of the initial public offering (IPO) to investigate auditor credibility. It is contended that information asymmetry problems lead to a demand for credible auditors in companies going public. Entrepreneurs have incentives to signal their knowledge of favorable future earnings by selecting reputable auditors. Since there is limited information available on firms going public, employing credible auditors can convey monitoring cost advantages as well. Investment bankers also have a preference for credible auditors since they rely on audited financial statements in certifying the value of the firm and determining whether to underwrite the offering. In the present study, we consider auditor credibility in IPOs from the perspective of the client and the investment banker. If there is an increased demand for auditor credibility at the time of the IPO, there should be a significant number of credibility-increasing auditor changes prior to the offering. Furhter, if the investment banker benefits from having a more credible auditor sign off on statements prepared by the client, this should be reflected in the investment banker's fee structure. The empirical analysis is performed on companies that went public in 1985 and 1986. Relatively few auditor changes are observed prior to the offering. However, among those companies making auditor changes, there is a clear preference for more credible auditors. Logistic regression analysis shows that companies with prestigious investment bankers are more likely to change away from local auditors to more credible CPAs. The type of underwriting arrangement employed is also significant, consistent with an investment banker preference for credible auditors. A regression analysis is conducted, using the 1985 and 1986 IPOs, modeling investment banker compensation as a function of several factors, including type of auditor employed by the issuing firm. In the case of "firm commitment" offerings, the auditor type is found to be significant. Clients seem to be charged a smaller investment banking fee if they are associated with Big Eight auditors. There is no apparent auditor effect in the case of "best efforts" offerings. The evidence generally supports the hypothesis that investment bankers and their clients have a preference for credible auditors for the IPO.]

[This paper investigates whether capital market investors, in assessing market values of R&D firms' equity, view R&D limited partnerships (LPs) as increasing both the assets and liabilities of the R&D firms. The contract terms between the R&D firm and the LP suggest interpreting the LP as a call option held by the R&D firm and using option pricing theory to estimate the asset (the present value of the LP-funded R&D project) and liability (the present value of the exercise price) components of the option. Estimates of the LP variables are derived from information provided in footnote disclosures by R&D firms. However, several important assumptions are necessary to derive these estimates because the footnote disclosures do not provide all the data necessary to calculate the option values. The estimates of the LP variables are included as explanatory variables in a cross-sectional market value regression model that is based on the balance sheet identity. Other variables included in the regression model are the assets, liabilities, and in-house expenditures on R&D reported by the R&D firms in their financial statements. The results are consistent with market participants capitalizing in-house R&D expenditures (though SFAS No. 2 (1974) requires R&D expenditures to be expensed as incurred), viewing the call option feature of the LP as relevant information in assessing the market value of the R&D firm, and utilizing the footnote disclosures to assess the option values. The results support the SEC and the FASB positions that require R&D firms with LPs to disclose to investors the existence of and the amount raised through the LP, but the results do not necessarily support the SEC and FASB positions that this disclosure should be made on the balance sheet. The problems I encountered in estimating the LP variables suggest that more detailed disclosures of the terms of the LP option would help investors to assess the market value of R&D firms. More generally, the results suggest that the FASB should pay particular attention to the types of required disclosures and not be so concerned with whether the disclosures should be on or off-balance sheet in their ongoing consideration of financial instruments and transactions.]

[The relation between changing expectations of earnings and changing security prices is a central issue in accounting and finance. In this article, I reexamine common stock returns surrounding earnings forecast revisions, using a large database of individual analyst forecasts, and provide new evidence on market expectations of revisions, on cross-sectional differences in price effects, and on the influence of confounding events. In summary, my findings are that revisions affect prices, but prices do not immediately assimilate the information. Price reaction is greater when the percentage change in forecast is in the top or bottom five percent of the distribution of all forecast revisions. This price effect is not simply due to an association between revisions and earnings, dividend, or stock-split announcements. Surprisingly, prices continue to drift in the direction of the revision for about six months after the revision. Another surprise is that price reaction does not incorporate some publicly available information. Stock returns immediately after individual analyst forecast revisions suggest that an analyst's current outstanding forecast is a better measure of the market expectations of the analyst's next forecast than an updated version (i.e., the analyst's current forecast updated for information revealed after the date of the current forecast but before the date of the next forecast). I use this curious price reaction result to create an aggressive trading strategy that predicts changes in outstanding forecasts, in other words, a strategy that predicts price reactions. The difference in abnormal returns between securities predicted to perform best and worst is more than 13 percent every six months. Changes in beta do not explain these abnormal returns.]

A transaction costs framework is developed to explain the choice between lump-sum and per unit payment provisions in private timber-harvesting contracts. Predictions about which contract type minimizes the transaction costs of presale measurement and contract enforcement and monitoring are derived and tested using private timber sales contracts from North Carolina. The empirical results provide strong support for the transaction costs approach and also reject several predictions from a risk-based model. The transaction costs framework also provides insights into the choice between negotiated and competitive sales procedures.