Biographical Information John Null, CMI, Sr. Manager Indirect Tax, - - PDF document

• Workshop NN

Statistical Sampling in Sales & Use Tax Audits … Best Practices and How States are Using These Techniques

Wednesday, January 29, 2020 2:00 p.m. to 3:00 p.m.

Biographical Information

John Null, CMI, Sr. Manager – Indirect Tax, The Home Depot 2455 Paces Ferry Rd., B12, Atlanta, Georgia 30339 (770) 384-4555 Mobile (404) 245-4217 JOHN_A_NULL@homedepot.com John Null, CPA, CMI, is the Senior Tax Manager of Indirect Tax at The Home Depot in Atlanta,

• GA. He has 25+ years of state sales & use tax experience in the telecommunications, retail and

manufacturing industries with both private and public firms. Prior to joining The Home Depot, John was a consulting manager at Deloitte Tax LLC working in the Tax Management Consulting group. He has worked for several telecommunications companies throughout his career and has assisted telecommunications companies with the implementation and maintenance of various transaction tax software systems. John is a Vertex Certified Professional and has worked with TaxWare, Avalara, and other tax calculation engines. He was a frequent presenter at the Telestrategies Tax Conference. In his current role, John has shifted into the retail industry and manages the audit, compliance and systems areas for The Home Depot. John earned a BS in Accounting from The Ohio State University. Bradley W. Tomlinson, Senior Manager (non-attorney professional) Zaino Hall & Farrin LLC, 41 South High Street, Suite 3600, Columbus, OH 43215 614-349-4818 Mobile: 614-580-9149 Fax: 614-754-6368 (Fax) btomlinson@zhftaxlaw.com Brad has more than 36 years of experience in state and local taxation, specializing in statistical sampling techniques for multiple taxes. From 2007 until joining the firm, Brad was an assistant administrator for the Ohio Department of Taxation, Audit Division, while also serving as the manager of the Computer Assisted Audit Group for more than 11 years. In that position he was responsible for the formulation, implementation, and oversight of the Department's statistical auditing practices; including the approval of data populations, sample designs, and the review of all statistical sampling procedural agreements. Prior to that, Brad spent several years auditing fortune 500 companies in manufacturing, telecommunications, computer services, and retail for sales and use tax compliance. Brad was instrumental in the design and implementation of the Audit Division's computerized auditing program used by multiple divisions to audit personal property, corporate franchise, employer withholding, pass-through entity, and sales and use taxes; as well as managing quarterly updates and releases of the application. Brad's other responsibilities included the establishment of single rate reporting procedures for taxpayers as well as assisting in reviewing and approving the Department's PARSA (Previous Audit Representative Sampling Analysis) agreements. During his tenure, Brad was a frequent presenter on statistical sampling at the Ohio Tax Conference and the Federation of Tax Administrators national computer and technology

• workshops. He is a graduate of Columbus State Community College (Computer Programming)

and The Ohio State University (Business Administration).

Statistical Sampling in Sales and Use Taxes

Best Practices and How States Are Using These Techniques

John Null, CMI

Senior Manager – Indirect Tax/The Home Depot Atlanta, GA

john_a_null@homedepot.com

Bradley Tomlinson

Senior Manager

Zaino Hall & Farrin LLC Columbus, OH

btomlinson@zhftaxlaw.com

LINK

Learning Objectives

• Obtain a basic understanding of

common statistical sampling terminology

• Understand the impact of differences in

applications and policies among states

• See actual examples of audit scenarios

among various states

Run Through

• In discussing sampling practices, three

states’ sampling methods are explored

• Ohio, Texas, & Tennessee (design only)
• Agreements
• Sample Plans
• Estimation Methods
• Sample Error Assessment

When to Sample

• Sampling is appropriate when:
• Records are voluminous
• Retrieving supporting documents is costly
• Audit period includes some unavailable records
• Auditor and taxpayer reach agreement on method
• f sampling and projection
• Sampling is not appropriate when:
• Cost of sampling exceeds cost of complete

examination (detail)

• Unable to identify population to be sampled
• No agreement on method of sampling and

projection

Common Terms

• Non‐statistical sampling methods:
• No measure of sample risk is provided (e.g., Texas)
• Statistical sampling requires assessment of the associated sample

risk:

• Confidence interval, confidence bound, relative precision, etc.
• Example for statistical sampling terms using $100,000:
• Confidence interval: We are 95% confident population

assessment is between $80,000 and $120,000

• Confidence bound: We are 95% confident population

assessment is at least $84,000

• Relative precision: We are 95% confident that projected

assessment is within plus or minus 10% of population assessment

6

The Home Depot

Statistical Sampling Audit Policy

7

The Home Depot Statistical Sampling Policy

• Large Volume Retailer
• Hundreds of millions of transactions
• Brick & Mortar store sales – 2,000 U.S. locations
• Online Sales
• Installed services – remodels and appliance installations
• Pro Customers (remodelers, carpenters, plumbers,

contractors,…)

• Tool & Vehicle rentals
• Seasonal Sales – garden center, Christmas decorations,

…

• Customer friendly tax exemption platform

8

The Home Depot Statistical Sampling Policy

• Examples of bad auditing outcomes
• Insufficient population size

– 4 weeks selected for a 4 year audit – Only initial data provided was used, later data provided was ignored – One transaction distorted the entire population

• Insufficient number of strata

– Only 2 strata resulted in nonrepresentative results

• States not able to handle data size
• States without adequate statistical sampling experience

9

The Home Depot Statistical Sampling Policy

• Rules
• Require the population to include entire audit

period

– If not possible, no less than one full year of data

• No block sampling
• Populations must be homogenous

– Problem areas must be placed in separate populations

• No less than 5 strata

10

The Home Depot Statistical Sampling Policy

• Sales populations segregated by error rates
• Cash & Carry / POS Register

– Highest volume (90+%), very low error rate – Do not want to “contaminate” with higher error rate populations

• Online Sales

– Low error rate, but slightly higher than POS sales – Delivery address errors – Larger volume of products for tax mapping

11

The Home Depot Statistical Sampling Policy

• Sales populations segregated by error rates
• Problem areas must be placed in separate

populations

• Installed Services (Remodeling) & Deliveries

– Higher error rates; ambiguity in interpretation of capital improvements; sales agents remove taxes

• Exempt Customers

– Highest error rates; customer friendly platform, fraudulent intent, missing certificates

Sampling Agreements

• Agreements used to jointly prepare and agree on

sampling plan, sample design, estimation method, and treatment of special cases (e.g., missing documentation)

• Planning and agreements can be useful to insure

agreement on the sample plan and a key to a successful sample audit

• Types of agreements
• Binding (e.g., Ohio)
• Non‐binding (most other states using agreements)
• Model agreement (California – CDTFA form 472)
• http://www.cdtfa.ca.gov/formspubs//cdtfa472.pdf

Sampling Agreements

Be sure to read and understand all components of the agreement Excerpts from Ohio agreements:

• Taxpayer has opted to forego allowing use tax overpayment
• ffsets in the sampling procedure. Therefore, tax overpaid

items will be classified as non‐errors and comprehensive refunds must be submitted for transactions where tax was remitted in error.

• Tax overpaid to vendors or sellers will not be offset and will

require a separate refund claim to be filed by the taxpayer with documentation based on a comprehensive review of all tax overpaid

• Tax overpaid items will be classified as negative errors and

will offset positive errors if the provisions stipulated in the "Overpayments Procedure" of this agreement are all met.

Polling Question 1

• The best policy for signing sampling

agreements is:

• A. Always sign agreements to show cooperation
• B. Never sign agreements because too many bad

things can happen

• C. Only sign them if they are not binding
• D. Base every decision on the agreement content

and specific scenario

14

tEST

Know Your Data

• Availability of supporting documents is your best

defense in any audit of sales and use tax, unclaimed property, or types of other transactions

• Do a preliminary investigation before auditors arrive
• Conduct a “walk‐through” to discover and

document what supporting documents are available and retrievable

• Examine electronic and paper files
• Identify potential exposure areas

Negative Values

• Negative amount transactions are considered a special

situation

• With purchase data this includes returns, price changes, voids,

and adjusted entries

• For the treatment of negative values, there are two common

methods:

• (a) Match negative items to their positive counterparts,

(b) set aside remaining negatives, (c) sample the positive transactions

• (a) Match negative items to their positive counterparts,

(b) create an absolute value amount, (c) construct strata

• n absolute value
• Most states use the first method (and is statistically preferred)

Missing Documentation

• Missing documentation is considered a special

situation and should be addressed in sampling agreements

• There are three common methods for documents
• Treat as 100% taxable (i.e., in error)
• Use of alternative information such as a like kind

invoice from the same vendor

• Sampled item replacement
• Statistically, sample replacement is not recommend as

it does not preserve the random order

Estimation Methods

• Projecting sample results to the population
• There are six standard estimation techniques
• Two most commonly used methods are
• Separate Ratio often call the Percentage Method
• Alabama, Arizona, Arkansas, Florida, Georgia, Hawaii,

Kentucky, Louisiana, Maine, Maryland, Massachusetts, Michigan, Mississippi, Nevada, New Jersey, North Carolina, Ohio, Pennsylvania, Rhode Island, South Carolina, South Dakota, Texas, Vermont, West Virginia, Wyoming

• Some of the above states other methods
• Mean Per Unit frequently (and incorrectly) referred to as

the Difference Method

• California, Iowa, New York, North Dakota,

Tennessee, Wisconsin

Estimation Methods

• High level example of the two common

estimation methods

A B C D E F G H Population Count Population Taxable Amount Population Average Taxable Amount Sample Size Sample Taxable Amount Sample Average Taxable Amount Sample in Error Amount Sample Average in Error Amount 50000 10,000,000.00 $ 200.00 $ 100 22,000.00 $ 220.00 $ 4,400.00 $ 44.00 $ Estimation Method Formula Point Estimate Percent =G/E*B 2,000,000.00 $ Mean per Unit =G/D*A 2,200,000.00 $

Polling Question 2

• Which of these elements is necessary for statistical

sampling?

• A. Statistical formula to determine sample size
• B. Report a measure of sampling risk (e.g., relative

precision, confidence interval)

• C. Approval by a statistician
• D. Baffle the taxpayer with lots of statistical formulas

20

LINK

Sample Plan Simulation

• AP Data File
• Sample Design
• Estimation
• Comparison
• Resources

AP Data File

• Line items = 494,683
• Total dollars = $349,427,402.69
• Total vendors = 8,320
• Vendors with tax accrued = 100
• Vendors charging tax = 1,644
• Total invoices = 265,205
• Credit items = 2,385

AP Data File

• Use 7% tax rate
• Total use tax paid = $5,407,951.21
• Total sales tax paid = $3,189,398.61
• Total Tax Paid = $8,597,349.82

Design ‐ Ohio

Design ‐ Texas

Design ‐ Tennessee

Estimation ‐ Ohio in Detail

Sample Projection

Stratum Sample Count Sampled Amount Sum of Taxable Amount with Offsets Sum of Taxable Amount without Offsets Population Total Error Factor with Offsets Error Factor without Offsets Audit Taxable with Offsets Audit Taxable without Offsets 1 250 6,496.24 864.40 910.73 2,526,437.87 13.3062% 14.0193% 336,172.88 354,188.90 2 250 23,587.43 2,534.00 2,637.55 8,285,523.48 10.7430% 11.1820% 890,113.79 926,487.24 3 250 64,428.35 7,363.62 7,663.25 21,649,656.10 11.4292% 11.8942% 2,474,382.49 2,575,053.40 4 250 213,333.74 21,460.65 22,445.75 70,109,450.63 10.0597% 10.5214% 7,052,800.41 7,376,495.74 5 250 1,013,517.35 44,627.50 50,418.80 87,418,514.60 4.4032% 4.9746% 3,849,212.03 4,348,721.43 6 250 3,796,377.65 202,937.72 244,969.81 40,880,302.80 5.3456% 6.4527% 2,185,297.47 2,637,883.30 7 1571 120,348,211.41 (914,757.88) 825,121.22 120,348,211.41 ‐0.7601% 0.6856% (914,766.75) 825,107.34 Grand Total 3071 125,465,952.17 (634,969.99) 1,154,167.11 351,218,096.89 15,873,212.32 19,043,937.35 <= $10.00 383,144.70 13.3062% 14.0193% 50,982.00 53,714.20 Total Taxable Amount 15,924,194.32 19,097,651.55 Tax Due at 7% $ 1,114,693.60 $ 1,336,835.61

Estimation ‐ High level TX & OH

Category TX OH - w/ offsets OH - w/o offsets Method Percent Percent Percent Point Estimate $1,104,088.88 $1,077,729.60 $1,416,688.94 <= $10.00 $1,081,298.34 $1,420,257.68 Standard Error $323,035.42 $162,989.59 $151,968.44 Precision 48.3% 24.9% 17.7% Margin of Error $533,008.44 $268,443.85 $250,292.01 Lower Bound $571,080.44 $809,285.75 $1,166,396.92 Upper Bound $1,637,097.32 $1,346,173.45 $1,666,980.95 degrees of freedom 271 651 631 t-value 1.650 1.647 1.647 Confidence Interval 90.0% 90.0% 90.0% Confidence Bound 95.0% 95.0% 95.0% Data level line item line item line item Sample Size 607 3071 3071 # of Errors in Sample 51 371 Projected # of Error % 13.6% 16.4%

Actual Numbers

Total Taxable Amount 135,650,558.53 Total Tax Due @ 7% 9,495,539.10 Sales Tax Paid 3,189,398.61 Use Tax Paid 5,407,951.21 Tax Paid 8,597,349.82 Net Tax Due 898,189.28 Sales Tax Paid in Error 136,009.60 Use Tax Paid in Error 291,870.21

Comparison

State Point Estimate Precision Difference from Actual Ohio with Offsets $1,081,298.34 17.7% $183,109.06 Ohio without Offsets $1,420,257.68 24.9% $522,068.40 Texas $1,104,088.88 48.3% $205,899.60

Polling Question 3

• The most logical reason that the Ohio

precision is better than Texas is:

• A. Pure luck – it was a 50/50 chance
• B. Texas was better since it was 48% compared to

24% for Ohio

• C. Texas used absolute value for the negatives
• D. Ohio’s sample size was significantly larger

31

LINK

State Resources

Tennessee:

https://www.tn.gov/content/dam/tn/revenue/documents/taxes/sales/statist icalsampling.pdf

Texas:

https://comptroller.texas.gov/taxes/audit/docs/sampling‐manual.pdf

Polling Question 4

• This presentation on statistical

sampling was:

• A. The best presentation I have ever

attended

• B. All of the above

33

LINK

Thank You!

Questions?