Hostname: page-component-77c89778f8-rkxrd Total loading time: 0 Render date: 2024-07-16T13:57:11.007Z Has data issue: false hasContentIssue false
  • English
  • Français

Statistical Sampling in Tax Audits

Published online by Cambridge University Press:  27 December 2018

Joseph C. Bright Jr. ,
Joseph B. Kadane  and
Daniel S. Nagin
Get access Rights & Permissions [Opens in a new window]

Abstract

The courts, with some important qualifications, have been reluctant to uphold tax assessments based on a review of only a sample of all transactions. In this article we argue that audit assessments based on appropriately drawn and analyzed statistical samples do not suffer from the defects that the courts have correctly concluded mar assessments based on nonstatistical samples. We do, however, argue that because of the inherent imprecision of assessments based on a less-than-complete review of all records, the calculation of the assessment should include a factor to take into account the risk that the taxpayer has been overassessed. We suggest an assessment rule that does just this and also recommend guidelines for the use of statistical sampling in tax audits.

Type
Articles
Information
Law & Social Inquiry , Volume 13 , Issue 2 , Spring 1988 , pp. 305 - 333
Copyright
Copyright © American Bar Foundation, 1988 

Access options

Get access to the full version of this content by using one of the access options below. (Log in options will check for institutional or personal access. Content may require purchase if you do not have access.)

References

1. It is our understanding, however, that as a matter of practice the IRS typically restricts its use of statistical sampling to audits of large corporate taxpayers.Google Scholar

2. See W. G. Cochran, Sampling Techniques (3d ed. John Wiley & Sons, 1977).Google Scholar

3. For clarity, it is necessary to discuss some fine points of statistical theory. From the point of view of classical statistical theory, a 95% confidence interval means that if the same procedure is used many times, 95% of the uses will include the true tax deficiency. Confidence intervals do not purport to apply to any specific instance. By contrast, a 95% credible interval means that the probability is 95% that the true tax deficiency is in the interval specified in the given instance. Such intervals require added assumptions and a different, Bayesian statistical framework. Thus a confidence interval tells you what proportion of times you will bracket the right amount if you follow a set procedure and a credible interval tells you what the chances are that you bracketed the right amount this time. See L. J. Savage, The Foundations of Statistical Inference (New York: John Wiley & Sons, 1962), for a general discussion of the distinction. The relation between classical and Bayesian inference is treated in more detail in sec. 4.Google Scholar

4. State ex rel. Foster v. Evatr, 144 Ohio 65, 56 N.E.2d 265 (1944).Google Scholar

5. E.g., Schwegmann Bros. Giant Supermarkets, Inc. v. Mouton, 309 S.2d 686 (La. 1975); Bouchard v. Johnson, 170 A.2d 372 (Me. 1961); Ridolfi v. Director, 1 N.J. Tax 198 (1980); In re Babylon Milk & Cream Co., 5 A.D.2d 712, 169 N.Y.S.2d 124 (1957); King Drug of Dayton v. Bowers, 171 Ohio 461, 172 N.E.2d 3 (1961). Other cases are cited in H. Leib, Using Sampling Techniques to Assess State Taxes, 3 J. St. Tax (1985) and L. Fournier & W. Raabe, Statistical Sampling Methods in State Tax Audits, 2 J. St. Tax 115 (1983).Google Scholar

6. In re Grecian Square, Inc., 119 A.D.2d 948, 501 N.Y.S.2d 219 (1986); Pato Foods, Inc. v. Lindley, 7 Ohio App. 3d 22, 453 N.E.2d 1274 (1982); Torridge Corp. v. Commissioner of Revenue, 84 N.M. 610, 506 P.2d 354 (1972).Google Scholar

7. E.g., Mitchell Bros. Truck Lines v. Hill, 363 P.2d 49 (Ore. 1961) (fuel use tax case); W. T. Grant Co. v. Joseph, 2 N.Y.2d 196, 140 N.E.2d 244 (1957). Compare In re Hard Face Welding & Machine Co., 81 A.D.2d 967, N.Y.S.2d 744 (1981).Google Scholar

8. E.g., Yonkers Plumbing & Heating Supply Corp. v. Tully, 62 A.D.2d 18, 402 N.Y.S.2d 792 (1978). 674 P.2d 785 (Abs. 1983).Google Scholar

9. 5 A.D.2d 712, 169 N.Y.S.2d 124 (1957).Google Scholar

10. Id.Google Scholar

11. Id.Google Scholar

12. 54 A.D.2d 1023, 38 N.Y.S.2d 176, 177 (1976), af'd, 44 N.Y.2d 684. 405 N.Y.S.2d 454, 376 N.E.2d 927.Google Scholar

13. Id.Google Scholar

14. 43 Ohio 2d 5, 330 N.E.2d 699 (1975).Google Scholar

15. See also Zapitelli v. Lindley, 1981 Westlaw 4376 (Ohio App. N.E.2d 1981).Google Scholar

16. 119 A.D.2d 948, 501 N.Y.S.2d 219 (1986) (citation omitted).Google Scholar

17. Id.Google Scholar

18. Farrar Brown Co. v. Johnson, 207 A.2d 406 (Me. 1965).Google Scholar

19. Id.Google Scholar

20. Id.Google Scholar

21. Yonkers Plumbing & Heating Supply Corp. v. Tully, 62 A.D.2d 18, 402 N.Y.S.2d 792 (1978).Google Scholar

22. Id.Google Scholar

23. 429 N.Y.S.2d 755 (App. Div. 1980).Google Scholar

24. Names in the News v. New York State Tax Comm., 429 N.Y.S.2d 755 (App. Div. 1980) (emphasis added).Google Scholar

25. 102 A.D.2d 352, 477 N.Y.S.2d (1984).Google Scholar

26. In re Korba, 84 A.D.2d 655, 444 N.Y.S.2d 312 (1981), appeal denied 56 N.Y.2d 502, 435 N.E.2d 1099; In re Chartair, Inc., 65 A.D.2d 44, 411 N.Y.S.2d 41 (App. Div. 1978).Google Scholar

27. Id. Google Scholar

28. Sears, Roebuck & Co. v. City of Inglewood (Los Angeles Super. Ct. 1955), discussed in Sprowls, Admissibility of Sample Data into a Court of Law, 4 U.C.L.A. L. Rev. 222, 226 (1957).Google Scholar

29. 38 Ohio 2d 135, 311 N.E.2d 1 (1974).Google Scholar

30. Id. at 141-43.Google Scholar

31. 26 174 P.2d 785 (Alaska 1983).Google Scholar

32. IdGoogle Scholar

33. While the courts have placed tight constraints on its use, sampling is routinely used in sales and use tax audits. The reasons are that taxpayers without complete records are prime audit targets and that many taxpayers with complete records consent to the use of sampling. To illustrate sampling's contribution to efficient enforcement, consider the case of Pennsylvania's Department of Revenue. In fiscal year 1984-85, the department conducted about 6,500 sales and use tax audits that identified over $44 million in assessed deficiencies. The total direct personnel commitment charged to these audits was nearly 150 person years. In about 85% of these audits, some sort of test-period procedure was used. A restriction of test period auditing to instances where taxpayer records were incomplete would have required a 15-fold increase in auditors at an annual cost of about $75 million-a commitment of 2,500 additional auditors, which is more than the department's entire personnel complement. Alternatively, the no-sampling audit strategy would require a reduction in the number of audits of about 8596, resulting in a direct revenue loss of about $35 million. The direct reduction in audit productivity would have resulted in deficits for the Commonwealth in four of the past eight fiscal years.Google Scholar

34. Names in the News v. New York State Tax Comm., 429 N.Y.S.2d 755 (App. Div. 1980).Google Scholar

35. M. DeGroot, Probability and Statistics (Reading, Mass.: Addison-Wesley, 1978).Google Scholar

36. See W. L. Felix & R. Roussey, Statistical Inference and the IRS, 159 1. Accountancy 38 (1985).Google Scholar

37. New York State Dep't of Tax. & Finance, EDP Systems Audit Bureau, letter to Pennsylvania Dep't of Revenue, Bureau of Audit, at 4.Google Scholar

38. R. A. Fisher, Statistical Methods for Research Workers (14th ed. New York: Hafner Publishing Co., 1973).Google Scholar

39. H. Raiffa & R. Schlaiffer, Applied Statistical Decision Theory viii (Cambridge, Mass.: MIT Press, 1961).Google Scholar

40. Pa. Dep't of Revenue, Board of Appeals Docket #5O4817 SUT.Google Scholar

41. Pa. Dep't of Revenue, Board of Appeals Docket #713816 SUT.Google Scholar

42. See also M. DeGroot, Optimal Statistical Decisions 261 (New York McGraw-Hill Book Co., 1970).Google Scholar

43. M.H. DeGroot, Probability and Statistics 577 (Reading, MA: Addison-Wesley Publishing Co., 1975).Google Scholar