Printer Friendly

Similarity and precedent in tax authority judgment.

ABSTRACT

Tax professionals often examine past court decisions and administrative rulings in an effort to find precedent for the treatment of a specific transaction. We formulate and test a psychological model that describes how professionals evaluate precedent. We also test predictions made by the model about the influence of certain environmental variables on tax authority judgment. Results indicate that tax professionals' judgments regarding the presence of authority for a tax position are significantly related to judged similarity of precedent. However, contrary to requirements in the Regulations and principles of jurisprudence, only common features (and not distinctive features) impact similarity judgments and subsequent authority judgments. In addition, the direction of comparison and the relative amount of distinctive information known about the precedent and client case are observed to impact judgment. We also find that taxpayer advocacy influences professionals' views of both similarity and authority. These re sults provide valuable insights into the tax authority judgment process, and contribute theoretical rigor by extending a formal psychological model of similarity judgment to the tax setting.

INTRODUCTION

Tax professionals are required to evaluate the level of authority for tax-reporting positions that they advise their clients to take and this evaluation is central to the tax research process (Chow et al. 1989; Raby 1982). Poor performance in this task can be very costly to the professional. If a tax-reporting position lacks sufficient authority, then both the CPA and the taxpayer may be liable for penalties. (1) For example, the CPA may be liable for both preparer penalties under LRC [section] 6694 and taxpayer damages measured with respect to accuracy-related penalties imposed on the taxpayer. Further and more significant costs may be incurred through the loss of taxpayer goodwill, damage to the tax practitioner's reputation, potential loss of clients, and loss of right to practice before the IRS (Bain and Kilpatrick 1990; Raby 1982).

When evaluating the level of tax authority for a position, professionals often examine past court decisions and administrative rulings in an effort to find precedent for the treatment of a specific transaction. The regulations require the tax professional to weigh the conflicting authorities in the same manner as would a court of law when resolving an issue (Treas. Reg. [section] 1.6661-3(b)(3)). The regulations go on to state that:

the weight of authorities depends on their persuasiveness and relevance as well as their source. For example, a case or revenue ruling having some facts in common with the tax treatment at issue would not be considered particularly relevant if the authority is materially distinguishable on its facts....

Thus, relevance of a precedent depends on its similarity (judged on the basis of both common and distinctive features) (2) with the taxpayer's fact scenario. The central role played by similarity in judgments regarding the relevance of precedent has been well established in the jurisprudence literature (Ashley 1991, 12-13; Eisenberg 1988; Levi 1949, 2; Schauer 1987, 587).

While previous research on the tax authority evaluation task has examined the expertise demonstrated by tax professionals (Bain and Kilpatrick 1990; Chang and McCarty 1988; Chow et al. 1989) and the role of analogical transfer across domains (Marchant et al. 1991, 1993), little is known about the role played by similarity in the evaluation of tax authority. Research on the role of similarity is critical to a complete understanding of how tax professionals evaluate tax authority and the extent to which the legal requirements for such judgments are being met. Ultimately, this could lead to tax professional performance improvements and to educational enhancements.

This study formulates and tests a psychological model that describes how similarity is incorporated in the tax authority evaluation process. Specifically, we introduce and extend a psychological model of similarity (Tversky 1977) to the tax setting. We use this model to make specific predictions about the influence of certain environmental variables such as taxpayer advocacy and the impact of distinctive facts on evaluations. We test these predictions through the use of a case study, using CPAs in tax practice and IRS revenue agents as subjects.

As expected, our results indicate that tax professionals' evaluation of authority depends on similarity of the precedent. However, contrary to the legal requirements to consider "materially distinguishable" facts of precedents, only common features (and not distinctive features) in cases appear to impact perceived similarity and authority. In addition, while earlier work (Johnson 1993) suggests that advocacy affects evaluations of authority, our results indicate that this effect is attributable (at least in part) to an effect on the perceived similarity of the precedent. Finally, our model predicts that, in some conditions, perceived similarity (and evaluation of precedent) may depend on direction of comparison (e.g., instance a may be judged similar to instance b, and at the same time, b may not be judged similar to a). This lack of symmetry in judged similarity could lead to systematic violations of principles of jurisprudence. While we do not actually observe asymmetric similarity judgments, we do find th at the antecedent conditions necessary for asymmetric similarity judgments exist among our tax professional subjects. Our results support and extend the prior literature and provide valuable insights into the tax authority judgment process. The study contributes theoretical rigor by extending a formal psychological model of similarity judgment to the tax setting. The influence of taxpayer advocacy on tax judgment may be more narrowly and accurately predicted as a result of this study. Furthermore, new insights are provided into the manner in which similarity may influence tax authority judgment.

The remainder of this paper is organized as follows. The next section introduces the similarity model and applies the model to the tax authority evaluation process. Then the methods and design used in the study are presented, followed by the data analysis and a discussion of the results. Finally, our conclusions are presented.

MODEL AND ITS APPLICATION

A Descriptive Model of Perceived Similarity

Research in psychology has shown that objects, concepts, and events are categorized by people on the basis of their common and distinctive features. Models describing how people judge similarity in psychology follow this research, expressing the similarity of objects in terms of the measures of their unique and shared features (Suppes et al. 1989; Ritov et al. 1990; Tversky 1977; Tversky and Gati 1978). These "feature-mapping" models characterize each object, a, to be compared as a set of features, denoted A. The total feature set for an object includes its appearance and function, its relation to other objects, and its deducible properties given an individual's general knowledge about the world. When faced with the need to make a similarity judgment, the model presumes that the individual extracts from the feature set a list of relevant features with which to perform the task. The resulting similarity of two objects a and b is denoted by an interval scale s(a, b), expressed as a function of the objects' comm on and distinctive features. This function has three arguments: A [intersection] B, the common features of a and b; A - B, the distinctive features of a that are not shared by b; and B - A, the distinctive features of b that are not shared by a. Thus, similarity between objects is represented as a feature-mapping function (i.e., a function that measures the degree to which two sets of features match each other). (3)

The specific function used in the feature-mapping model to represent similarity is based on observations regarding people's behavior in numerous experiments. This function is a linear combination of the common and distinctive features noted above:

s(a, b) = [theta]f(A [intersection] B) - [alpha]f(A - B) - [beta]f(B - A), where [theta], [alpha], [beta] [greater than or equal to] 0.

The scale f is a nonnegative interval scale that measures the prominence or salience of features. The parameters [theta], [alpha], and [beta], represent the relative importance of the common features of the two objects ([theta]), the distinctive features of the source, a ([alpha]), and the distinctive features of the referent, b ([beta]) These parameters are assumed to be sensitive to context and task. Thus, s(a,b) defines a family of similarity indices that share specific properties rather than a unique index. The model is intended to be descriptive in nature and may be used to explain assumptions that are consistent with observations of individual judgment in the tax domain (e.g., taxpayer advocacy).

Application of the Model to Tax Authority Judgments Common and Distinctive Features in Similarity Judgments

The feature-mapping model predicts that increases in weights of common features will increase the judged similarity of the client fact situation (CFS) and precedent, while increases in the weights of distinctive features will decrease the judged similarity of the CFS and precedent. Assume a tax practitioner has been asked to advise a taxpayer on a tax issue. Assume further that the law is unclear and the tax practitioner locates a favorable court decision on the issue. According to the regulations, the tax professional should consider both facts common to the CFS and precedent and those facts that are distinguishable in determining the relevance of the precedent to the tax treatment at issue. The regulations may be viewed as requiring that the tax practitioner determine the similarity of the CFS and the precedent. The model suggests that this judgment may be formalized by the following hypothesis:

H1: The judged similarity of the CFS and precedent will depend upon the common feature weights and distinctive feature weights of the CFS and the precedent, assuming that both common and distinctive features are relevant to the similarity judgment.

Taxpayer Advocacy and Similarity Judgment

As discussed above, the regulations specify that the perceived similarity of precedent to the CFS is a central element in the tax authority evaluation process. One issue that can be addressed through the model is how a tax professional's degree of taxpayer advocacy might influence this perceived similarity. The conceptual definition of advocacy used in this study was:

Advocacy is a state of mind in which one feels one's primary loyalty belongs to the taxpayer. It is exhibited by a desire to represent the taxpayer zealously within the bounds of the law, and by a desire to be a fighter on behalf of the taxpayer.

Practitioner publications (Blaustein 1982), official pronouncements of the AICPA (Statements of Responsibility in Tax Practice No. 1, AICPA 1989), and academic publications (Johnson 1993) were relied upon when formulating this definition of advocacy.

In the feature-mapping model, a tax professional's level of advocacy will influence the relative weights assigned to the common and distinctive features. The effect depends on whether the outcome in the precedent is favorable or unfavorable to the client. For low levels of advocacy, the outcome of the precedent has little influence on the weights. Increasing advocacy causes the outcome of the precedent to have greater influence on the weights assigned to the common and distinctive features. Higher levels of advocacy cause the tax professional to place more weight on common (distinctive) features as opposed to distinctive (common) features when referencing a favorable (unfavorable) precedent. As a result, the perceived similarity of the CFS to a favorable (unfavorable) precedent to increases (decreases) for an advocate and consequently influences the tax professional's assessment of the authority for the taxpayer's position. Hence, the level of taxpayer advocacy and the outcome of the precedent are predicted t o "color" the perception of similarity and thereby impact the tax authority judgment, as follows:

H2: The relative weights assigned to the common and the distinctive features of the fact scenarios depend upon the joint effect of the level of advocacy exhibited by the tax professional and upon whether the precedent is favorable or unfavorable to the taxpayer's position.

This prediction, if supported, will add to our understanding regarding earlier findings on the impact of advocacy on performance in the research task. In particular, it explains a cause underlying the observation that tax professionals tend to weight a supportive authority more heavily than an unfavorable authority (Johnson 1993) and tend to focus their search for authority on cases that support the taxpayer's desired outcome (Cloyd and Spilker 1999). In particular, the contrast model predicts the process by which taxpayer advocacy will cause tax professionals to perceive that authority supportive of the taxpayer's positions are more similar to the CFS than authorities unfavorable to the taxpayer's position. This process involves an evaluation of the fact scenarios of the CFS and the precedent. In effect, the predicted impact on the weighting of the common and distinctive features of the CFS and the precedent may be partly responsible for the confirmation bias that has been documented in the literature.

Asymmetry and Similarity Judgment

In addition to providing an explanation of how advocacy influences tax professionals' similarity judgments, the feature-mapping model identifies conditions when tax professionals may make asymmetric judgments. (4) Specifically, the model predicts asymmetry in similarity judgments when the following two conditions are met: (1) the comparison is directional and (2) the amount of unique information known about the two objects is not equal. Both of these conditions are likely to occur in tax authority evaluations. Asymmetric judgments, if they occur in tax practice, could lead to problems in tax authority judgments.

In the feature-matching model, the ability to specify the direction of comparison (which object is the starting point, or subject, in a comparison) is important because people engage in a mapping of the features of the subject onto those of the referent. By using the subject as the starting point, people are attuned to its unique features (A - B), so that the unique features of the referent (B - A) are accessed and evaluated in reference to those of the subject. Features unique to the subject will be highlighted by this kind of comparison and play a significant role. Thus, the model predicts that the features of the subject control the agenda for the comparison. In the tax domain, this would occur if the tax professional either compares the CFS to the precedent or compares the precedent to the CFS. If the CFS is the starting point (subject), then the tax professional will be more aware of its features than those of the precedent, and vice versa if the precedent is the starting point. In summary, the starting point is predicted to control the comparison.

Tversky (1977) found that the choice of the subject and the referent depends on the amount of information known about each of the objects. Individuals, when making similarity judgments, tend to select the more salient object, or prototype, as the referent, and the less salient object, or the variant, as the subject (Tversky 1977; Tversky and Gati 1978). For example, Tversky observed that people tend to say "North Korea is like Red China" rather than "Red China is like North Korea" and that "the portrait resembles the person" rather than "the person resembles the portrait."

However, other research in psychology suggests that individuals develop habitual reference points that do not change as the measure of the objects change. Holyoak and Gordon (1983) reinterpreted the feature-matching model in terms of habitual or fixed reference points. They suggested that individuals develop habitual reference points within particular knowledge domains. For example in social cognition, it has been shown that the self operates as a habitual reference point in self/other judgments (Srull and Gaelick 1983).

This research raises two issues regarding tax professionals. First, tax professionals may develop habitual reference points when doing tax research. When developing a defensible argument to support a tax position, the tax professional may use the CFS as a referent and compare case law to it to determine if the case scenario is on point with the CFS or may be distinguished from it. Alternatively, the tax professional may have developed a habitual reference point such that the court case acts as the referent, and compare the CFS to it to determine if the taxpayer case may be distinguished from the court case. Regardless, when comparing the CFS with a court case, the comparison used by tax professionals may be directional and may involve a habitual reference point.

In the case of directional comparisons (where [alpha] [not equal to] [beta]), the feature-matching model predicts that similarity will be affected by the relative number of distinctive features possessed by each of the objects. In particular, similarity judgments will be asymmetric when the amount of distinctive features attributed to each of the objects differs. The nature of the asymmetry can be determined if the direction of the comparison and the relative amount of unique information about each of the objects is known. (5) More specifically, if two objects, a and b, are being compared directionally and object a has more unique features than does object b, judged similarity will vary depending on which object is the subject of the comparison. The pair will be judged less similar when a is the subject because the weight of the unique features of the two objects taken together will be greater when a is the subject than when b is the subject, even though the amount of unique information known about the two o bjects, in total, has remained unchanged (Tversky and Gati 1978).

Tversky (1977) demonstrated this result using college students. In one task, participants rated the similarity of 21 pairs of countries. Each pair consisted of a prominent (i.e., Russia) and a nonprominent country (i.e., Poland). (6) Participants were asked to rate the degree to which one member of the pair (i.e., Poland) was similar to the other member of the pair (i.e., Russia). When the subject of the comparison was the more prominent country, the perceived similarity was less than when the less prominent country was the subject of the comparison. (7)

If these results extend to the similarity judgments performed by tax professionals, then several unexpected factors may affect tax professionals' judgments regarding the aptness of precedent and resulting judgments regarding the legal authority for a position. In particular, the direction of the comparison employed by the tax professional and the relative amounts of unique features of the CFS and the precedent may have a substantial impact on the tax professionals' judgments.

For example, a tax professional may receive a research request that includes a brief summary of a client's fact scenario (CFS). The tax professional's research however, may reveal a court case with an extensive description of the facts of the case. Alternatively, the tax professional may have extensive information about a CFS, and a relatively brief description of facts in a court opinion. Both represent commonly encountered examples in tax practice where asymmetric similarity judgments might be expected. If the fact scenario with the largest amount of unique features is treated as the referent in the comparison, then the perceived similarity will tend to be greater than if it were treated as the subject. Thus, tax professionals who treat the fact scenario with the larger number of unique features as the subject ([alpha]) would view the CFS and the court case as less similar to each other than would the tax professional who treated the scenario with the larger number of unique features as the referent ([beta] ), all other things being held constant. These observations lead to the following hypothesis:

H3: The weights assigned to the distinctive features of the CFS and court case will depend upon the joint effect of the direction of comparison and the amount of relevant distinctive information known about each case, such that the total weight of the distinctive features of the two cases, will be larger (smaller) when the fact scenario with the greatest (smallest) number of unique features is the subject (referent) of the comparison.

Similarity and the Tax Authority Evaluation

As discussed previously, the regulations and the jurisprudence literature indicate that tax authority judgments should not depend solely on whether the outcome of the precedent is favorable or unfavorable. The similarity of the facts of the precedent with the CFS should also play a role. As a precedent's similarity to the CFS increases (decreases), the impact of the outcome of the precedent on a professional's tax authority evaluation should increase (decrease). A joint effect of similarity and the outcome of the precedent is expected. To test this joint relationship, the

following hypothesis is offered:

114: As the degree of reported similarity between the CFS and the precedent increases (decreases), the influence of the outcome of the precedent on the tax authority judgment will increase (decrease).

EXPERIMENT

Subjects

Ninety-one CPA firm tax practitioners and 59 Internal Revenue Service agents were the participants in the study. Participants were randomly assigned to conditions. The CPA firm participants were paid $10.00 in advance for their participation. The IRS participants were CPAs. All IRS participants were either revenue agents or appeals officers. Due to government ethics restrictions, IRS personnel were not remunerated for their participation. (8)

Case Development

The task in this experiment required a determination of whether a class of workers was properly classified as employee or independent contractor. (9) This issue was chosen because it is an ambiguous area that has generated substantial controversy and litigation (Stratton 1995) and the resolution of the majority of the litigation in the area has turned upon an analysis of facts and circumstances. Participants were given the CFS, a short discussion of the law and analysis related to the issue, and then the court opinion (precedent). Participants then performed the following tasks:

Part A: indicate similarity judgment on a seven-point Likert scale;

Part B: indicate level of tax authority (likelihood of prevailing if position is challenged) on a 101-point scale where 100 = certain victory and 0 = certain defeat;

Part C: indicate the importance of individual features to judgments in Part A and Part B respectively on seven-point Likert scales (features were grouped as common, distinctive to CFS, and distinctive to court case);

Part D: provide demographic information including employer, experience, etc.;

Part E: complete a nine-item taxpayer advocacy scale; and

Part F: complete a free recall knowledge test.

Actual case law was used to develop the two narratives that were employed in the study. The high-fact scenario contained 24 facts and the low-fact scenario contained 16 facts. Eight of the facts were common to both, resulting in 16 facts that were unique to the high-fact scenario and eight facts that were unique to the low-fact scenario. One of the narratives was used as the CFS and the other was used as precedent. One half the participants received the high-fact case as the CFS and the low-fact case as the precedent. The other half of the participants received the low-fact case as the CFS and the high-fact case as the precedent. The participants' task was to evaluate the legal authority for the taxpayer's position and then judge the likelihood of success of the position given the decision in the precedent. One half of the participants were told the outcome of the precedent was favorable to the taxpayer and the other half were told the outcome of the precedent was unfavorable to the taxpayer. Subjects were ra ndomly assigned across experimental conditions.

The instrument was developed from court cases that addressed the independent contractor status of workers (gypsy chasers) who follow interstate truckers around the country to assist in unloading of their cargo. The facts that the courts relied upon for their decisions and the general background information presented in the cases were modified and adapted for the initial development of the two narratives. Dicta from the court decisions were not included in the narratives. (10) The narratives were constructed so that they had different amounts of distinctive features. After the initial fact patterns were created, two tax professors at a large midwestern university listed the individual features independently. Differences were resolved with an author acting as arbitrator.

Based on a pretest given to a tax professor, a tax attorney, a tax practitioner, and three Ph.D. students with extensive tax practice experience, several changes were made to the instrument. The degree of ambiguity was increased by reducing the number .of common features from ten to eight, increasing the number of features unique to the low-fact case from six to eight, and increasing from 14 to 16 the number of features unique to the high-fact case. These changes increased the total number of features from 30 to 32.

The individual features in the two fact patterns were again listed and validated by the two raters and classified as either common to both cases, unique to the low-fact case, or unique to the high-fact case. The Kappa coefficient (Cohen 1960) for interrater reliability was 0.92 with overall agreement of 95 percent. Each of the facts was also evaluated as either favorable to the taxpayer's desired reporting position, unfavorable to the taxpayer's desired reporting position, or neutral. On this scale, the Kappa coefficient was 0.89 with overall agreement of 93 percent. As before, differences were resolved with one author acting as arbitrator.

Independent Variables

Two independent variables were manipulated across participants in the experiment and two independent variables were measured. The favorableness of the outcome of the precedent and the assignment of the high- and low-fact cases to the taxpayer CFS were both manipulated as between-subject factors resulting in a fully crossed 2 X 2 design. Taxpayer advocacy and direction of comparison were measured variables. Taxpayer advocacy and the outcome of the precedent relate to H2. The assignment of the high- or low-fact cases to the taxpayer scenario and outcome of precedent permit a test of H3. Reported similarity and outcome of the precedent were used to test H4. The measurement or manipulation of each independent variable is discussed in detail below.

Taxpayer Advocacy

A nine-item scale was used to measure advocacy in the study (Mason and Levy 2001). (11) The procedures used to derive the advocacy scale closely adhered to recommended psychometric practice (Churchill 1979; Nunnally and Bernstein 1994). After generating a number of possible scale items and screening each item for construct validity (using two tax professors at a large midwestern university), a preliminary advocacy scale consisting of 22 items was administered to 33 tax CPAs and 20 IRS revenue agents. The preliminary scale was subjected to reliability analysis and 13 items were eliminated, resulting in an alpha coefficient of 0.88. Using principal components analysis, the remaining nine items loaded (at the 0.40 level or greater) onto a single factor explaining 51.6 percent of the variance. This nine-item advocacy scale was administered to participants in the main study.

Outcome of Precedent

The outcome of the precedent was manipulated as a two-level between-participant factor (favorable or unfavorable). One half of the participants were instructed that the court ruled the workers were independent contractors (favorable outcome), and one half of the participants were instructed that the court ruled the workers were employees (unfavorable outcome). The favorable condition was coded as (+1) and the unfavorable condition was coded as (-l). (12)

Assignment of High- and Low-Distinctive-Feature Cases

To test H3, which predicts that differing amounts of distinctive information known about the taxpayer CFS and the precedent may influence similarity judgment, a two-level between-subjects manipulation was performed. One half of the participants were given the high-distinctive-feature case as the taxpayer CFS and the low-distinctive-feature case as the precedent. The other half of the participants received the low-distinctive-feature case as the taxpayer CFS and the high-distinctive-feature case as precedent. Regardless of condition, the taxpayer scenario was always presented before the precedent in the experimental packet. The high-distinctive-feature condition was coded (-1) and the low-distinctive-feature condition was coded (+1).

Direction of Comparison

The direction of comparison was measured by having participants indicate whether they compared the precedent to the taxpayer CFS or compared the taxpayer CFS to the precedent. Although some prior psychological studies were able to manipulate the direction of comparison by the framing of the task, the success of those procedures is suspect when habitual reference points may exist (Read 1987). Because of the concern raised previously that habitual reference points may be an issue in this study, and since the prior literature in tax has not addressed this issue, a question in the experimental materials requested participants to indicate the direction of comparison they used in the study.

Control Variables

Tax Expertise

It is possible that level of tax expertise may influence perceptions of similarity. For example, Sanbonmatsu et al. (1992) suggest that individuals who are highly knowledgeable about the target object are much more likely to be influenced by omissions and by number of features relative to the evaluations of less knowledgeable subjects. To control for any expertise effects, we introduced two measures of expertise as control variables in the experiment. We also attempt to control for expertise via random assignment.

Both experience-based and knowledge-based measures of expertise were employed. For the knowledge measure, a task-specific knowledge test was used. A task-specific experience-based measure was employed to capture aspects of knowledge not addressed in the knowledge test.

The knowledge test involved a free recall task that required subjects to recall from memory factors employed by the IRS when making independent contractor/employee determinations. (13) The number of correct factors recalled constituted participants' scores on the knowledge test, with higher scores representing greater knowledge. The maximum possible score on this task was 20 and the minimum possible score was zero. Two raters with substantial tax background independently scored participants' free recall of the Rev. Rul. 87-41 factors. The Kappa coefficient (Cohen 1960) for interrater reliability was 0.91. The mean score on the free recall task was 5.48 with a standard deviation of 3.75, minimum score of zero and maximum score of 16. (14) Differences were reconciled with an author acting as arbitrator.

IRS/Tax Practitioner Indicator

An additional variable was included in the design to control for other potential differences among participants due to their employment status. Working as a tax practitioner vs. working as an IRS revenue agent may result in other differences not captured by the other variables in the study; thus, participants indicated their employment status on the instrument (public accounting firm or IRS).

Dependent Variables

Several dependent variables were used in this study. Participants were asked to assess the importance of each of the common and distinctive features for the similarity judgment and for the likelihood of success judgment. In addition, they were asked to assess the similarity of the two fact scenarios and perform a likelihood of success judgment (tax authority evaluation).

Importance of Features

The model suggests that the similarity of the CFS and the precedent is dependent upon the weighting of the common and distinctive features of the stimuli. In the model's terminology, participants' weighting of the common and distinctive features of the two cases will determine the level of similarity between the two cases. The importance (weighting) of each feature was measured by listing all the individual features in the narratives and having participants rate the importance of each of the features employing a seven-point Likert scale. Total common feature weights were the sum of the feature weights of the individual common feature weights of the features common to the CFS and the precedent. Total distinctive feature weights of the CFS and precedent were measured by computing the sum of the individual distinctive feature weights of the distinctive features of the CFS and precedent, respectively.

In addition to rating the importance of each feature to their similarity judgments, participants also rated the importance of each feature to their likelihood of success assessment. This additional measure was included to validate the aptness of the similarity model in the tax authority judgment context.

Similarity

Similarity was measured by asking participants to indicate the degree of similarity of the court case to the taxpayer scenario on a seven-point Likert scale. The endpoints of the scale were labeled "very similar" (7) and "not similar at all" (1).

Likelihood of Success

For the likelihood of success judgment, participants were asked to indicate the likelihood of success of the taxpayer-favored tax treatment prevailing if it was judicially challenged. The subjects responded with a number from zero to 100 to indicate their likelihood of success judgment. The endpoints of the scale represented no authority for favorable treatment (zero) and complete authority for favorable treatment (100), respectively. The no-authority endpoint referred to the absence of any authority or zero likelihood of winning judicially. The complete authority endpoint represented a position in which the tax professional believed the likelihood of prevailing was certain.

RESULTS

Common and Distinctive Features on Similarity Judgment (Hypothesis 1)

Hypothesis 1 predicts that both the distinctive and common features of the cases may influence similarity judgments. The model predicts that, ceteris paribus, the weights of the common (distinctive) features should be positively (negatively) related to the similarity judgment such that as the weights of the common (distinctive) features increase the similarity judgment should increase (decrease). The results of the analysis, presented in Table 1, reveal this hypothesis was only partially supported. Specifically, while common feature weights were significantly and positively related to similarity judgments features (t = 3.48, p < 0.01), there was no significant relationship between the weighting of the distinctive features and the similarity judgments of the participants.

To determine if this lack of influence was merely due to a lack of variability in the weights of the distinctive features, the descriptive statistics for the distinctive feature weights were examined. The mean feature weight score was 100.12 and the standard deviation was 20.05 with a range of 101. The maximum possible score was 216 and the minimum possible score was 24 in the experiment. The lack of an effect for the distinctive feature weights on similarity would not appear to be due to a lack of variability of the distinctive feature weights. Therefore, even though there was adequate variability in the distinctive feature weights, the distinctive feature weights did not affect the participants' similarity judgments.

These results suggest that differences in similarity judgments were due to differences in the weighting of the common features, not to differences in the weighting of the distinctive features. These results suggest that tax professionals were not performing as required by the regulations with respect to their evaluation of the distinctive features.

Expertise Effects

Two measures of expertise were employed in the experiment to control for any potential relationship between tax professionals' judgment and expertise. Analysis of the effect of expertise on feature weights, in Table 3, shows that the knowledge-based expertise measure was related to the distinctive feature weights (t = 2.66, p < 0.01). Furthermore, both expertise measures were significantly related to the combined feature weights (t = -2.91, p < 0.01 for the knowledge measure and t = 2.01, p < 0.05 for the experience measure). Neither measure of expertise was significantly related to the weighting of the common features of the scenarios.

Although the results of this test suggest that both measures of tax expertise had a significant relationship to the weights of the distinctive features in the experiment, analysis regarding Hl indicates that the weights of distinctive features had no significant influence on participants' similarity judgment (t = 0.29). Thus, there was no evidence that expertise influenced the similarity judgments of the participants even though it influenced the evaluation of the distinctive features. These results suggest that participants were attuned to the importance of the distinctive features, but were unable to incorporate that importance into their judgment process.

Employer Effects

The demographic variables did not suggest any differences between IRS participants and CPA firm participants that were not captured by the expertise and the advocacy measures (descriptive statistics for the advocacy scale are presented in Table 2). However, it is possible that IRS agents have different objectives in performing their work duties as administrators of the tax laws and these differences may influence their tax authority evaluations. To control for these other potential differences between IRS and CPA firm participants, an employer variable was included in the analysis. However, as can be seen from Table 3, this variable was not significant for any of the analyses. Thus, it appears that employment status, IRS vs. CPA firm, does not effect the tax authority evaluations in this study. (15)

Taxpayer Advocacy (Hypothesis 2)

Hypothesis 2 predicts that the relative weights assigned to the common and distinctive features of the fact scenarios depend upon the joint effect of the level of advocacy exhibited by the tax professional and whether the precedent is favorable or unfavorable to the taxpayer's position. Because the common features, the distinctive features, or both may be influenced by this interaction, three different regression analyses of this hypothesis were performed for the data in the experiment.

The three analyses use the same set of independent variables, but each uses a different dependent variable. The total feature weights assigned to the common features by the participants were used initially. The second analysis employed the total feature weights assigned by participants to the distinctive features as the dependent variable. The difference between the common and distinctive feature weights was the dependent variable in the third analysis. The results of these tests are presented in Table 3.

As can be seen from Table 3, the interaction of the outcome of the court case and taxpayer advocacy was significant with respect to the weighting of the common features in the experiment (t = 2.32, p < 0.05). However, this interaction was not significant with respect to the weights of the distinctive features or with respect to the combined weighting of the common and distinctive features. This suggests that the joint effect of advocacy and outcome predicted by H2 is supported, but only with respect to the common feature weights.

The positive partial regression coefficient for the outcome by advocacy variable indicates the influence of this variable is in the predicted direction. Specifically, the influence of a favorable outcome for a court ruling on the common feature weights increases (decreases) as the level of taxpayer advocacy increases (decreases) and the influence of an unfavorable court ruling decreases (increases) as the level of taxpayer advocacy increases (decreases). Figure 1 illustrates this effect using data from the experiment. (16)

Asymmetry (Hypothesis 3)

Hypothesis 3 predicts that the total weights assigned by participants to the distinctive features of the two cases (CFS and precedent) would be influenced by the joint effect of the allocation of the distinctive features to each of the cases and the direction of comparison (an asymmetry effect). The regression analysis of the distinctive feature weights used in the test of H2 was used to test this hypothesis.

Table 3 reveals a significant effect for the hypothesized (direction by distinctive) interaction (t = -2.84, p < 0.01) for the distinctive feature weights, as predicted by the model. When the taxpayer is the subject of comparison, total distinctive feature weights are greater when the CFS has more distinctive features. Alternatively, when the taxpayer is the referent of the comparison, total distinctive feature weights are less when the CFS has more distinctive features. Figure 2 illustrates this effect. Although a main effect for allocation of distinctive features is present with the interaction included in the model (t = -2.03, p <0.05), as mentioned before, the appropriate test for the main effect is with the interaction removed. When this is done the main effect for allocation of the distinctive fact patterns is not significant (t = 0.79). Thus, tax professionals weight distinctive features asymmetrically. This behavior could lead to asymmetric similarity judgments when distinctive features impact perceiv ed similarity. However, our results regarding Hl showed that distinctive features did not affect the similarity judgment.

The interaction is of the form predicted by the model. When the taxpayer is the subject of the comparison, the total distinctive feature weights are higher for the condition when the taxpayer was the high-fact case than for the condition when the taxpayer was the low-fact case. Alternatively, when the taxpayer was treated as the referent of the comparison, the total distinctive feature weights were lower for the taxpayer condition with the higher amount of unique features than for the taxpayer condition with the lower amount of unique features. The results suggest that tax professionals are subject to asymmetry in their weighting of the distinctive features of the cases. These results provide additional evidence that participants were cognizant of the distinctive features. However, the results of H1 indicated no significant effect of distinctive features on the performance of the similarity judgment process.

Influence of Similarity on Tax Authority Judgments (Hypotheses 4)

The relationship between the participants' likelihood of success judgment and the joint effect of participants' similarity judgment and the outcome of the court case is predicted by H4. To test for this relationship, a regression analysis was performed with likelihood of success judgments as the dependent variable. The regression analysis contained three independent variables: similarity judgments, a contrast-coded discrete variable representing the outcome of the court case, and the interaction of these two variables. Table 4 presents the results of the analysis.

The interaction predicted by H4 was significant (t = 7.47, p < 0.01) and in the direction predicted. The influence of a favorable outcome for a court ruling increases (decreases) as the similarity of the facts increases (decreases) and the influence of an unfavorable court ruling increases (decreases) as the similarity of the facts increases (decreases). This result suggests that tax professionals are performing normatively in terms of this aspect of the tax authority evaluation process. Figure 3 illustrates this effect using data from this experiment.

Additional Analysis

In addition to the assignment of feature weights for the similarity judgment, participants also indicated the importance of the features to their likelihood of success assessment. Based on the model, no differences were expected. However, to determine whether our result regarding the failure to consider distinctive features extends from similarity to the tax authority judgment, an additional regression was performed using subjects' rankings of the relative importance of each factor in judging tax authority. A comparison of descriptive statistics for the common and distinctive feature weights revealed there were some consistent differences in the weighting of the features for the similarity judgment and the likelihood of success judgment. In particular, the common features were generally considered less important to the likelihood of success judgment and more important to the similarity judgment. Conversely distinctive features were relatively less important to similarity judgment and relatively more important to the likelihood of success judgment.

Nonetheless, the results of the additional regression analysis revealed that the distinctive feature weights for the likelihood of success judgment were not significant (t = 0.53). In contrast, common feature weights were significantly and positively related to the likelihood of success judgment (t = 2.79, p < 0.01). Thus, the results of this study provide no support for the influence of the distinctive features on either similarity or likelihood of success judgment.

Another potential explanation of the lack of an effect for distinctive feature weights on similarity judgment is that the impact of precedent on tax authority may depend on the character of the distinctive features of the precedents and the CFS. If the distinctive features of the first precedent all support the favorable position, then the impact of the precedent on a tax authority judgment would be limited. This is because the precedent is easily distinguished from the CFS by arguing that the favorable decision in the precedent was largely attributable to the favorable distinctive features of that case. In contrast, if the distinctive features of the second precedent uniformly failed to support the favorable decision (i.e., the judge reached his or her decision despite the unfavorable distinctive features of the precedent), then the impact of the second precedent on the professional's tax authority evaluation would be increased. Thus, favorable or unfavorable distinctive features of the CFS may have a simil ar mediating role in the impact of similarity on tax authority judgments. Common features, by their nature, will always reduce the risk of a precedent being distinguished from the CFS. Consequently, they will not mediate the impact of similarity on tax authority judgments. (17)

DISCUSSION AND IMPLICATIONS OF THE RESEARCH

The purpose of this study was to gain insight into the tax authority judgment process through the application of the feature-matching model to tax authority evaluations. In addition to testing the normative requirements for tax authority evaluations (i.e., the ability of tax professionals to perform evaluations as specified in the regulations), the study used a descriptive model of similarity to test predictions that tax authority judgments would be influenced by taxpayer advocacy and the relative amount of detail known about the CFS and precedent and asymmetry.

The motivation for the use of a model of similarity to explain judgments performed by tax professionals in tax authority evaluations was due in part to the recent focus by policymakers on the effects of taxpayer advocacy on tax judgment. Specifically, the model predicts that factors such as taxpayer advocacy and the relative amount of detail known about the CFS and precedent will influence tax professionals' judgment regarding the similarity of the precedent to the CFS. In turn, tax authority assessments may be affected indirectly. Since the prior tax literature on the effect of similarity on tax authority evaluation has been exploratory in nature, the introduction of a model of similarity judgment to the area suggested a significant and important contribution to the literature. This model, in particular, provided the means to explain and test potential deviations by tax professional judgment from the normative expectations of the regulations and the jurisprudence literature.

The results of the experiment provided mixed support for the hypotheses that were tested in this study. Participants' likelihood of success judgments were influenced by the joint effect of similarity of the precedent to the CFS and the outcome of the precedent, as predicted by H4. The outcome of the court case had a stronger influence on the likelihood of success judgment as the similarity of the CFS and precedent increased. Thus, tax professionals seem to be sensitive to the importance of similarity to their likelihood of success judgments as required by the regulations.

However, contrary to our expectations, the results suggest that participants' similarity judgments were sensitive to only the common features of the CFS and precedent. Judgments regarding similarity did not appear to be affected by the distinctive features of the CFS or precedent. Exploratory data analysis suggests that participants generally considered the distinctive features more important for likelihood of success judgments than for similarity judgments. However, the distinctive features did not significantly influence likelihood of success judgments. Thus, based upon the results of this study, there appears to be an inability on the part of tax professionals to incorporate the notion of distinctive information into the similarity judgment task, which is contrary to the behavior required under the law.

The results also suggest an explanation of how taxpayer advocacy may influence tax professional judgment. The model predicted that taxpayer advocacy would influence similarity judgments through its joint influence with the outcome of the court case on the weights assigned to the features of the cases when judging similarity. The results of this study support the prediction and suggest this influence is only through the common features.

A precursor to the asymmetry effect that has been reported to influence similarity judgments in the psychology literature was observed in this experiment. This effect arises from the differential weighting of the distinctive features resulting when comparison is directional and the amount of distinctive information known about the CFS and precedent differ. However, while distinctive features were differentially weighted, as predicted, no subsequent asymmetry effect was observed in similarity judgments. This was perhaps because the distinctive information in the case was not properly considered by subjects, as noted above. With respect to the influence of expertise on judgment, both knowledge-based and experience-based measures of expertise were observed to influence the weighting of the distinctive features of the cases. But, as previously mentioned, the distinctive features of the cases had no subsequent influence on similarity judgment.

The results of this study suggest several avenues for future research. First, the lack of significance of the distinctive features of the cases calls for further investigation. As discussed previously, tax professionals are subject to tax preparer penalties and many other potential sanctions for inadequate performance on the tax authority evaluation task. The results of this study suggest some concerns about the ability of tax professionals to perform this task as required by the regulations and the jurisprudence literature. Further research should be pursued to attempt replication of the current findings to determine the pervasiveness of this result. (18) Other contextual factors may also impact perceived similarity through the feature-matching model. These factors should be examined as well. Finally, other potential models of similarity judgment and other tax domains should be studied. Should the observed results persist, investigation of how improvements to training might be achieved should be pursued.

[FIGURE 1 OMITTED]

[FIGURE 2 OMITTED]

[FIGURE 3 OMITTED]
TABLE 1

TEST OF HYPOTHESIS 1: REGRESSIONS OF SIMILARITY JUDGMENTS OF TAX
PROFESSIONALS ON COMMON AND DISTINCTIVE FEATURE WEIGHTS OF THE CFS AND
PRECEDENT

                                    Regression
Source                              Coefficient  t-statistic

Sum of common feature weights          0.04        3.48 *
(Hypothesis 1)
Sum of distinctive feature weights     0.00        0.29
(Hypothesis 1)
Adjusted [R.sup.2]                                 0.09

* p < 0.01.

TABLE 2

DESCRIPTIVE STATISTICS FOR RESPONSES TO THE NINE-ITEM TAXPRAYER ADVOCACY
SCALE (a) FOR 255 TAX PROFESSIONALS

                                                    Standard
Group                                    n    Mean  Deviation  Minimum

CPA firm participants                    89  46.03    7.74       24
IRS participants                         58  36.29    7.52       15
CPA firm and IRS participants combined  147  42.19    9.00       15


Group                                   Maximum

CPA firm participants                     63
IRS participants                          56
CPA firm and IRS participants combined    63

(a) Taxpayer advocacy increases as the advocacy score increases. The
maximum (minimum) possible advocacy score on the scale is 63 (9).

TABLE 3

TESTS OF HYPOTHESES 2 AND 3: REGRESSION OF MEASURES OF IMPORTANCE OF
FEATURE WEIGHTS ON TAXPAYER ADVOCACY (HYPOTHESIS 2), ASYMMETRY
(HYPOTHESIS 3), EXPERTISE, AND EXPERIMENTAL FACTOR VARIABLES

                                          Coefficients (t-statistics)
                                                for Regression
                                              Models, by Feature
                                                Weigh Measures

                                             Common      Distinctive
Source                                      Features      Features

Taxpayer advocacy measure (advocacy)          0.01          0.76
                                            (-0.08)        (3.27) **
Tax knowledge (tax)                           0.17          5.77
                                            (-0.21)        (2.66) **
Tax experience (experience)                   0.03         -0.49
                                             (0.26)       (-1.79)
Favorableness of precedent (outcome)         -6.50          0.12
  (+1) Favorable to taxpayer                (-2.15) *      (0.02)
  (-1) Unfavorable to taxpayer

Number of distinctive features in CFS
 relative to precedent (distinctive)
  (-1) High CFS                              -0.88         -4.29
  (+1) Low CFS                              (-1.10)       (-2.03) *

Direction of comparison (direction)
  (+1) Court case to taxpayer                 1.20          2.02
  (-1) Taxpayer to court case                (1.51)        (0.96)
Employer (+1) IRS, (-1) CPA Firm             -0.28          0.10
                                            (-0.29)        (0.04)
Direction by distinctive (Hypothesis 3)      -0.54         -6.05
                                            (-0.67)       (-2.84) **
Outcome by distinctive                        0.95          1.00
                                             (1.50)        (0.60)
Outcome by advocacy (Hypothesis 2)            0.16          0.05
                                             (2.32) *      (0.28)
Adjusted [R.sup.2]                            0.12          0.18

                                         Coefficients
                                         (t-statistic
                                              s)
                                             for
                                          Regression
                                          Models, by
                                           Feature
                                            Weigh
                                           Measures

                                         Combined
Source                                   Features

Taxpayer advocacy measure (advocacy)       0.76
                                          (3.51) **
Tax knowledge (tax)                        5.93
                                          (2.91) **
Tax experience (experience)               -0.52
                                         (-2.01) *
Favorableness of precedent (outcome)       6.61
  (+1) Favorable to taxpayer              (0.88)
  (-1) Unfavorable to taxpayer

Number of distinctive features in CFS
 relative to precedent (distinctive)
  (-1) High CFS                           -3.41
  (+1) Low CFS                           (-1.72)

Direction of comparison (direction)
  (+1) Court case to taxpayer              0.82
  (-1) Taxpayer to court case             (0.42)
Employer (+1) IRS, (-1) CPA Firm           0.38
                                          (0.16)
Direction by distinctive (Hypothesis 3)   -5.51
                                         (-2.76) **
Outcome by distinctive                     0.06
                                          (0.04)
Outcome by advocacy (Hypothesis 2)        -0.11
                                         (-0.63)
Adjusted [R.sup.2]                         0.19

*, ** p < 0.05 and p < 0.01, respectively.

TABLE 4

TEST OF HYPOTHESIS 4: REGRESSION OF SIMILARITY JUDGMENTS AND
FAVORABLENESS OF PRECEDENT ON TAX AUTHORITY (LIKELIHOOD OF SUCCESS)
JUDGMENTS

                                        Regression
Source                                  Coefficient  t-statistic

Tax Professionals' Similarity Judgment     0.05         0.65
 (similarity)
    Seven-point Likert scale
Favorableness of precedent (outcome)      -0.28        -5.48 *
    (+1) favorable to taxpayer
    (-1) unfavorable to taxpayer
Similarity by outcome (Hypothesis 4)       0.09         7.47 *
Adjusted [R.sup.2]                                      0.48

* < 0.01.


We acknowledge the contributions made by workshop participants at University of Notre Dame, University of Wisconsin-Madison, Universite Laval, Boston College, University of Memphis. University of Illinois at Urbana-Champaign, the American Accounting Association Annual Meeting, Virginia Polytechnic Institute & State University, and Michigan State University. We especially thank Betty Jackson, Barry Lewis, Ira Solomon, Mark Peecher, K. Dianne Jackson, and Reid Hastie for their input and time spent discussing many issues in this paper with us.

Submitted: September 2001

Accepted: September 2002

(1.) In response to criticisms that preparers and taxpayers were exploiting ambiguous tax issues and also in an attempt to keep them from playing the "audit lottery," Congress introduced a "substantial authority" standard in the Tax Equity and Fiscal Responsibility Act of 1982 (TEFRA) and a "realistic possibility" standard in the 1989 Revenue Reconciliation Act (IMPACT). These provisions introduced taxpayer and tax preparer penalties for tax return positions that lack the requisite level of tax authority.

(2.) The term "distinctive feature" refers to a feature that the precedent does not share with the taxpayer's current fact situation (CFS). Similarly, "common feature" refers to a feature that is shared by the precedent and the CFS. For example, if one case involved a green taxi and another case involved a blue taxi, the color of the taxi in each case would be a distinctive feature (possibly a "materially distinguishable" fact per the regulations), while the presence of a taxi in both cases would be a common feature.

(3.) Early research in psychology characterized similarity as a metric distance between points in a coordinate space. However, numerous experiments led to rejection of this notion (e.g., see Tversky 1977).

(4.) If one judges item a to be like item b when a is the subject of comparison and b to be like a when b is the subject, then such judgments are said to be symmetric. In contrast, similarity judgments are asymmetric if a is judged to be like b when a is the subject but b is judged not to be like a when b is the subject. Such directional differences, or asymmetric similarity judgments, are contrary to general notions of normative decision making in jurisprudence in general, and tax research in particular. In particular, the direction of comparison should not influence judgments regarding similarity.

(5.) For example, the scale value f(A) associated with stimulus a is regarded as an overall measure of that stimulus.

(6.) more prominent country in each comparison had more unique features. The terms "salience" and "prominence" have been used interchangeably in the literature to denote the relative measure (amounts of unique information) of the objects (Houston et al. 1989). In discussing the prior literature in this area, salience and prominence will be used interchangeably to describe the relative amounts of the unique features of the objects.

(7.) It has been shown that the formulation of the instructions to the subjects can control the direction of comparison (Houston et al. 1989; Tversky 1977; Tversky and Gati 1978). When subjects are told to compare object a to object b, subjects are more likely to treat object a as the subject and object b as the referent.

(8.) IRS employees were included in the study for two reasons. First, given that the IRS is responsible for administering the tax system, it seemed important to know if IRS employees' similarity judgments and tax authority evaluations differed significantly from those of other tax professionals. Second, it was believed, a priori, that the level of taxpayer advocacy exhibited by IRS employees would be significantly different from other tax professionals. Thus, including IRS employees in the study would allow for more variability on the advocacy scale. Surprisingly, results indicated that IRS employees' similarity judgments and tax authority evaluations were not significantly different from other tax professionals. Furthermore, there was a large amount of variability across both groups on the advocacy scale.

(9.) An employer must generally withhold income taxes, and withhold and pay social security, Medicare, and unemployment taxes on wages paid to employees. An employer does not generally have to withhold or pay employment tax on payments to independent contractors. Thus, employers minimize their tax liability by treating workers as independent contractors rather than as employees.

(10.) Schauer (1987) notes that consideration of dicta can confound the evaluation of precedent. While limiting the cases only to relevant facts loses some richness, for purposes of the experiment, more was gained by simplifying the task faced by the subjects.

(11.) A complete discussion of the procedures followed in the development of the client advocacy construct used in this study (and in a second study, Levy [1996]) and the importance of IRS agents to the development of the construct may be found in Mason and Levy (2001).

(12.) Judd and McClelland (1989) recommend the use of a complete set of contrast codes when group sample size is unequal to improve the interpretability of the results. Another advantage of contrast codes when cell sizes are equal is that it eliminates redundancy or correlations among the predictors.

(13.) The IRS has extracted from prior court cases a list of 20 factors (Rev. Rul. 87-41, 1987-1 C.B. 296) to be considered in making employee/independent contractor determinations.

(14.) To correct for violations of the normality and homogeneity of variance assumptions, a square root power transformation was performed on the knowledge test data. This approach to dealing with a joint violation of normality and homogeneity of variance is recommended by Judd and McClelland (1989). Subsequent specification tests of the transformed data suggested no significant violations of these assumptions.

(15.) Using a full effects model, none of the interactions between the employer variable and the other variables were statistically significant. Consequently, they were dropped from subsequent analysis.

(16.) The only other significant result from Table 3 with respect to the common feature weights is a main effect for the outcome of the court case (t =-2.15, p < 0.05). However, this effect must be interpreted with caution since it only provides information about the effect of a variable on the weighting of the common features when the other component with which it is multiplied to form the interaction equals zero (Judd and McClelland 1989). Furthermore, Judd and McClelland (1989) recommend that testing the effect of a component across levels of some other component should be done by using a model that does not include the interaction term. Following this approach, there was no significant main effect for outcome. Thus, these results suggest that on average, the main effect for outcome of the court case and advocacy do not influence the weighting of the common features of the two scenarios.

(17.) To gather some exploratory data on this potential explanation a second experiment was run in which similarity was regressed on likelihood of success using subsets of the data in the two experiments. The cases in the second experiment dealt with categorization of financial instruments as debt or equity (this second study was also used to evaluate the robustness of the results from the study reported in this paper). In this second experiment, the two narratives were designed so that the high-fact case contained only features that were either unfavorable or neutral with respect to the desired tax-reporting position and the low-fact case contained a mix of favorable, unfavorable, and neutral features. The regression coefficient for each of the two conditions in experiment one (mixed distinctive features and either favorable or unfavorable court outcome) were compared with the related coefficients in experiment two (negative distinctive features assigned to the precedent or CFS). Statistical comparison of th e regression coefficients showed no significant result. Thus, these results do not provide an explanation that the lack of an effect of distinctive features was somehow due to the mix of favorable/unfavorable distinctive features in the CFS and the precedent.

(18.) As mentioned previously, a replication of this study was performed addressing the categorization of financial instruments as debt or equity. As with the independent contractor/employee issue, the debt-equity issue typically turns upon an analysis of facts and circumstances. The methods employed to develop the narratives were identical to those used in the first study. The results of the second study generally replicated the results reported in this paper.

REFERENCES

American Institute of Certified Public Accountants (AICPA). 1989. Statements of Responsibilities in Tax Practice. New York, NY: AICPA.

Ashley, K. 1991. Modeling Legal Argument: Reasoning with Cases and Hypotheticals. Boston, MA: MIT Press.

Bain, C., and B. Kilpatrick. 1990. A note on professionals' judgments of tax authority. The Journal of the

American Taxation Association 12 (Fall): 78--86.

Blaustein, R. 1982. How to Do Business with the IRS: The Complete Guide for Tax Professionals. Englewood Cliffs, NJ: Prentice Hall, Inc.

Chang, O., and T. McCarty. 1988. Evidence on judgment involving the determination of substantial authority: Tax practitioners versus students. The Journal of the American Taxation Association (Fall): 26-39.

Chow, C., M. Shields, and G. Whittenburg. 1989. The quality of tax practitioners' judgments regarding substantial authority: An exploratory empirical investigation. Advances in Taxation 2:165-180.

Churchill, G., Jr. 1979. A paradigm for developing better measures of marketing constructs. Journal of Marketing Research 16 (February): 64-73.

Cloyd, C., and B. Spilker. 1999. The influence of taxpayer preferences on tax professionals' search for judicial precedents, subsequent judgments, and recommendations. The Accounting Review 74 (July): 299-322.

Cohen, J. 1960. A coefficient of agreement for nominal scales. Educational and Psychological Measurement 20 (Spring): 37-46.

Eisenberg, M. 1988. The Nature of the Common Law. Cambridge, MA: Harvard University Press.

Holyoak, K., and P. Gordon. 1983. Attitudes and social cognition. Journal of Personality and Social Psychology 44 (May): 881-887.

Houston, D., S. Sherman, and S. Baker. 1989. The influence of unique features and direction of comparison on preferences. Journal of Experimental Social Psychology 25 (March): 121-141.

Johnson, L. 1993. An empirical investigation of the effects of advocacy on preparers' evaluations of judicial evidence. The Journal of the American Taxation Association 15 (Spring): 1-22.

Judd, C., and G. McClelland. 1989. Data Analysis: A Model-Comparison Approach. Orlando, FL: HBJ, Inc.

Levi, E. 1949. An Introduction to Legal Reasoning. Chicago, IL: University of Chicago Press.

Levy, L. 1996. Evidence evaluation and aggressive reporting in ambiguous tax situations. Ph.D. dissertation, Department of Accounting, University of Colorado at Boulder.

Marchant, G., J. Robinson, U. Anderson, and M. Schadewald. 1991. Analogical transfer and expertise in legal reasoning. Organizational Behavior and Human Decision Processes (April): 272-290.

-----, -----, -----, and -----. 1993. The use of analogy in legal argument: Problem similarity, precedent, and expertise. Organizational Behavior and Human Decision Processes (June): 95-119.

Mason, J., and L. Levy. 2001. The use of the latent constructs method in behavioral accounting research: The measurement of taxpayer advocacy. Advances in Taxation 13: 123-139.

Nunnally, J., and I. Bernstein. 1994. Psychometric Theory. Third edition. New York, NY: McGraw-Hill.

Raby, W. 1982. TEFRA: Impact on taxpayer compliance. Journal of Accountancy 154 (November): 154-164.

Read, S. 1987. Similarity and causality in the use of social analogies. Journal of Experimental Social Psychology 23 (May): 189-207.

Ritov, I., I. Gati, and A. Tversky. 1990. Differential weighting of common and distinctive components. Journal of Experimental Psychology: General 119 (March): 30-41.

Sanbonmatsu, D., F. Kardes, and P. Herr. 1992. The role of prior knowledge and missing information in multi-attribute evaluation. Organizational Behavior and Human Decision Processes 51: 76-91.

Schauer, F. 1987. Precedent. Stanford Law Review 39: 571-605.

Srull, T., and L. Gaelick. 1983. General principles and individual differences in the self as a habitual reference point: An examination of self-other judgments of similarity. Social Cognition 2:108-121.

Stratton, S. 1995. Worker classification issues hound employers, IRS, and the courts. Tax Notes (November 6): 673-675.

Suppes, P., D. Krantz, R. Luce, and A. Tversky. 1989. Foundations of Measurement. Volume 2. San Diego, CA: Academic Press, Inc.

Tversky, A. 1977. Features of similarity. Psychological Review 84 (July): 327-352.

-----, and I. Gati. 1978. Studies of similarity. In Cognition and Categorization, edited by E. Rosch, and B. Lloyd. Hillsdale, NJ: Erlbaum.

Jon S. Davis is an Associate Professor at the University of Wisconsin-Madison and J. David Mason is an Associate Professor at Southern Illinois University Edwardsville.
COPYRIGHT 2003 American Accounting Association
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 2003 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Author:Davis, Jon S.; Mason, J. David
Publication:Journal of the American Taxation Association
Geographic Code:1USA
Date:Mar 22, 2003
Words:10757
Previous Article:Embedded options and tax decisions: a reconsideration of the traditional vs. Roth IRA decision.
Next Article:The impact of floors and phase-outs on taxpayers' decisions and understanding of marginal tax rates.
Topics:

Terms of use | Privacy policy | Copyright © 2018 Farlex, Inc. | Feedback | For webmasters