This paper investigates the relationship between sampling risks and audit consequences under error and audit value projection testing approaches. The latter approach typically is presented in auditing textbooks in connection with classical variables sampling, while the former approach is shown to underlie both dollar-unit sampling methods (in which an upper precision limit for monetary error is computed) and statistical compliance testing methods. In particular, we demonstrate analytically that error projection approaches implicitly test null hypotheses which effectively are equivalent to an alternative hypothesis underlying audit value projection. We also present outcome matrices to identify the effect of such hypothesis interchange on the audit consequences of sampling risks, and thus provide a basis for clarifying sampling risk discussions in textbooks and the recently issued Audit and Accounting Guide: Audit Sampling.

This study examines the empirical relation between voluntary participation in the Internal Revenue Service's (IRS) Compliance Assurance Process (CAP) audit program, and tax uncertainty disclosed in financial statements pursuant to Financial Interpretation No. 48 (FIN 48). Based on the findings of prior analytical and empirical research, we formulate and test hypotheses about the likelihood of voluntary CAP participation and the resulting effect on FIN 48 tax reserves. We find that firms with moderate-sized FIN 48 reserves are more likely to participate in CAP than firms with either small or large reserves, indicating an inverted U-shaped relation between CAP participation rates and firms' tax reserves. After controlling for non-random sample selection, we find that CAP firms significantly reduce their FIN 48 reserves by about 16.5 percent relative to non-CAP firms. However, this reduction is concentrated among firms with moderate-sized FIN 48 reserves. These cross-sectional differences are consistent with FIN 48 reserves reflecting both tax uncertainty and tax aggressiveness. JEL Classifications: M41; M42; M48; H25. Data Availability: FIN 48 data and confidential tax data on CAP participants are obtained from the Internal Revenue Service (IRS) Large Business & International (LB&I), Planning, Analysis, Inventory, and Research Division (PAIR). The CAP data are not publicly available; the FIN 48 data were compiled and validated by the IRS and made available to one of the authors. All other data are available from public sources identified in this treatise. Because tax data are confidential and protected by data nondisclosure agreements under the Internal Revenue Code, all statistics are presented in the aggregate; no statistics with three or fewer observations are disclosed. Any opinions are those of the authors and do not necessarily reflect the views of the IRS.

[Tax law complexity and ambiguity may result in uncertainty about taxable income (Slemrod 1988) and are of concern to policy-making bodies such as the ABA, the AICPA, and the IRS (Sheppard and Evans 1990). Several studies have modeled the effects of uncertainty on taxpayer reporting and the role of tax practitioners in reducing uncertainty (Alm 1988; Shavell 1988; Beck and Jung 1989a, 1989b; Scotchmer 1989a, 1989b; and Scotchmer and Slemrod 1989). Empirical and experimental research, however, have not kept pace. This paper reports experimental tests of the effects of income uncertainty and other economic factors based on tax reporting models in Beck and Jung (1989a). Hypotheses were tested regarding the effects of changes in the uncertainty level, tax rate, penalty rate, and audit probability on reported taxable income. In addition, the explanatory power of the models was evaluated by comparing the taxable income reported by the subjects with model-based predictions. Subjects were endowed with a fictitious currency and were given a range of possible post-audit taxable incomes from which to report. A proportional tax was paid on reported income and, in the event of an audit, a monetary penalty was imposed when the actual taxable income was greater than the amount reported. Incentives were provided by making the subjects' post-experimental remuneration a function of the after-tax income retained from each experimental trial. Since previous theoretical research indicates that taxpayers' reporting decisions are sensitive to risk preferences, three seperate experiments were performed and the results were analyzed by repeated measures ANOVAs. In the first and second experiments, subjects' risk-taking attitudes were controlled by the Berg et al. (1986) mechanism. Risk-neutrality (risk-aversion) was induced in the first (second) experiment, while subjects' preferences in the third experiment were measured ex post, rather than controlled. Two measures of taxpayer reports were employed-the actual income reported by subjects and the corresponding reporting fractile. The experimental results provided support for risk-neutral predictions. First, risk-neutral subjects were found to report higher levels of income when penalty rates and audit probabilities increased. Second, the tax rate did not affect the reporting behavior of risk-neutral subjects. Third, income reports were affected by two interactions: audit probability with uncertainty and penalty rate with uncertainty. Specifically, a reduction in uncertainty led to higher (lower) levels of reported taxable income when penalty rates or audit probabilities were decreased (increased). In addition, the mean reporting fractile did not change with the uncertainty level and the deviation of mean observed reports from predicted levels was small. For the risk-averse model, the predicted tax rate effect was marginally significant for reports and insignificant for fractiles. Furthermore, only a small percentage of the variance was explained by the interaction of tax rate and uncertainty. Report fractiles, however, did increase significantly as predicted when uncertainty was elevated.]

This study extends the Beck, Davis, and Jung (1992) experimental study by incorporating opportunities for taxpayer subjects to purchase advice before making their tax reporting decisions. Tax advice has two roles in the model and experiment. First, tax advice allows taxpayers to reduce their uncertainty about the amount of taxes owed. This permits us to study the demand for tax advice in conjunction with its effect on tax reporting decisions. Second, the decision to purchase tax advice from an expert provides a means of signaling that may alter tax agency audit policies. The resulting audit policies, in turn, can affect taxpayers' decisions to purchase tax advice. These interdependencies are incorporated in a game‐theoretic model and several predictions are tested experimentally. Consistent with theory, subjects sorted themselves into three groups based upon their private information. Another prediction supported by the experiment is that the demand for tax advice increased with the magnitude of an uncertain tax deduction (amount‐at‐risk). We also found, as expected, that the effects of tax advice on pre‐audit tax revenues depend upon the amount‐at‐risk. When the amount‐at‐risk was high, subjects in settings with tax advice reported lower average incomes and paid lower taxes than did those subjects in settings without the opportunity to purchase tax advice. The opposite was true for the low amount‐at‐risk condition, supporting our prediction that tax advice and amount‐at‐risk would have an interactive effect on tax reporting. Analysis was also performed on post‐audit tax collections. Contrary to theory, post‐audit tax payments were found to increase when subjects had an opportunity to purchase tax advice. This result apparently was caused by a tendency to over‐purchase tax advice and, in some cases, to report a low income regardless of the advice received. Résumé. Les auteurs développent l'étude expérimentale de Beck, Davis et Jung (1992) en y incorporant la possibilité pour les sujets contribuables de solliciter des conseils avant de prendre leurs décisions relatives à l'information qu'ils fourniront aux fins fiscales. Les conseils fiscaux jouent deux rôles dans le modèle et l'expérience. Premièrement, ils permettent aux contribuables de réduire leur incertitude relative au montant des impôts auxquels ils sont assujettis. Les auteurs peuvent ainsi étudier la demande de conseils fiscaux en conjonction avec leurs conséquences sur les décisions relatives à l'information à fournir aux fins fiscales. Deuxièmement, la décision de solliciter les conseils fiscaux d'un expert est un signal pouvant influer sur les politiques de vérification des représentants du fisc. À leur tour, ces politiques peuvent influer sur les décisions des contribuables de solliciter des conseils fiscaux. Ces interdépendances ont été intégrées à un modèle de jeu théorique et les auteurs ont vérifié plusieurs prédictions par expérimentation. Conformément à la théorie, les sujets se sont répartis en trois groupes, en fonction de l'information privilégiée dont ils disposaient. L'expérimentation a confirmé une autre prédiction: la demande de conseils fiscaux augmente avec l'importance de la somme dont la déductibilité est incertaine (montant à risque). Les auteurs constatent également, comme ils s'y attendaient, que les effets des conseils fiscaux sur les recettes fiscales antérieures à la vérification dépendent du montant à risque. Lorsque le montant à risque est élevé, les sujets pouvant solliciter des conseils fiscaux font état de revenus moyens inférieurs et paient moins d'impôt que les sujets n'ayant pas la possibilité de solliciter des conseils fiscaux. L'inverse est vrai lorsque le montant à risque est faible, ce qui confirme la prédiction selon laquelle les conseils fiscaux et le montant à risque ont un effet interactif sur l'information fournie aux fins fiscales. Les auteurs analysent également les impôts perçus postérieurement à la vérification. Contrairement à la théorie, ils constatent que les paiements fiscaux postérieurs à la vérification augmentent lorsque les sujets ont eu la possibilité de solliciter des conseils fiscaux. Ce résultat semble être attribuable à une tendance à solliciter des conseils fiscaux avec excès et, dans certains cas, à faire état d'un faible revenu, malgré les conseils reçus.

Economic models of tax reporting were tested experimentally. Subjects were given endowments and made tax reporting decisions subject to monetary penalties for underpayment of taxes and uncertainty about the amount of taxable income and the tax agency's cutoff point for audit selection. Two types of tax audit regimes were considered. In the first (nonstrategic audit regime), subjects faced a fixed probability of audit selection. A second (strategic) audit regime also was examined in which the probability of audit selection varied in response to tax reports as in the Reinganum and Wilde (1988), Beck and Jung (1989b), and Jung (1991a) models. Five hypotheses based on the comparative statics predictions of the models were tested and four were supported by the experimental data. Among the noteworthy findings are that tax rate changes have significant effects on reporting decisions under both audit regimes, rather than just under the strategic audit regime as hypothesized. A reduction (increase) in taxable income uncertainty induces subjects to report a low level of income significantly more (less) often under strategic auditing, but does not have a significant effect on income reports when audit policies are nonstrategic. The effects of changes in the level of uncertainty about the audit cutoff point are also investigated and found to depend upon the benefit of conducting an audit. Finally, we find that subjects generally report a low income more frequently than predicted by either the strategic or nonstrategic models. Résumé. Les auteurs ont mis à l'épreuve les modèles économiques de présentation de l'information fiscale au moyen de l'expérimentation. Les sujets se sont vu attribuer une dotation et ont pris des décisions relatives à la présentation de l'information fiscale, décisions assujetties à des pénalités monétaires s'appliquant aux impôts payés en moins et à l'incertitude relativement au montant du revenu imposable ainsi qu'au seuil de démarcation utilisé par les autorités fiscales pour sélectionner les entreprises devant faire l'objet d'une vérification. Deux types de régimes de vérification fiscale ont été analysés. Dans le premier cas (régime de vérification non stratégique), les sujets avaient affaire à une probabilité fixe d'être sélectionnés en vue d'une vérification. Dans le second (régime de vérification stratégique), la probabilité de sélection variait selon les rapports présentés au fisc, comme dans les modèles de Reinganum et Wilde (1988), Beck et Jung (1989b) et Jung (1991a). Cinq hypothèses fondées sur les prédictions statiques comparatives des modèles ont été testées, et quatre d'entre elles ont été confirmées par les résultats de l'expérience. Parmi les conclusions dignes de mention figure le fait que les modifications du taux d'imposition ont une incidence significative sur les décisions relatives à la présentation de l'information dans les deux régimes de vérification, et non pas seulement dans le régime de vérification stratégique, comme les auteurs l'avaient supposé. Une réduction (augmentation) de l'incertitude relative au revenu imposable incite les sujets à faire état d'un faible niveau de bénéfices beaucoup plus (moins) souvent dans le cas de vérifications stratégiques, mais n'a pas d'incidence marquée sur l'information présentée relativement aux bénéfices lorsque le régime de vérification est non stratégique. Les conséquences des changements dans le niveau d'incertitude relatif au seuil de démarcation de la vérification fiscale ont aussi été analysées, ce qui a permis d'établir qu'elles dépendent des avantages que comporte la vérification. Enfin, les auteurs ont constaté que les sujets faisaient état, en général, de faibles bénéfices plus souvent que prévu par leurs modèles, qu'ils soient stratégiques ou non stratégiques.

The article reports on experimental tests of the effects of income uncertainty and other economic factors based on tax reporting models. Tax law complexity and ambiguity may result in uncertainty about taxable income and are of concern to policy-making bodies such as the ABA, the AICPA and the IRS. Hypotheses were tested regarding the effects of changes in the uncertainty level, tax rate, penalty rate, and audit probability on reported taxable income. In addition, the explanatory power of the models was evaluated by comparing the taxable income reported by the subjects with model-based predictions. Subjects were endowed with a fictitious currency and were given a range of possible post-audit taxable incomes from which to report. A proportional tax was paid on reported income and, in the event of an audit, a monetary penalty was imposed when the actual taxable income was greater than the amount reported. Incentives were provided by making the subjects' post-experimental remuneration a function of the after-tax income retained from each experimental trial.