Transcript Slide 1

Ed Tobias, CISA, CIA
May 12, 2010
 Expectations
 Background
 Why
it works
 Real-world examples
 How do I use it?
 Questions
 How
many have heard of it?
◦ All over the professional journals
 J. of Accountancy – 2003, 2007
 J. of Forensic Accounting – 2004
 Internal Auditor – 2008
 ISACA Journal – 2010
 Fraud Magazine - 2010
 As
of 2004, over 150 articles
have been written about
Benford’s Law
 1881
– Simon Newcomb,
astronomer / mathematician
 Noticed
that front part of
logarithm books was more used
 Inferred
that scientists were
multiplying more #s with lower
digits
 1938
– Frank Benford,
Physicist at GE Research labs
 Front
part of the log book was
more worn out than the back
 Analyzed
20 sets of “random
numbers” – 20,299 #s in all
 Tested
random #s and
random categories
 Areas of rivers
 Baseball stats
 #s in magazine articles
 Street addresses - first 342
people listed in “American Men
of Science”
 Utility Bills in Solomon Islands
 Benford’s
Law:
◦ Random #s are not random
◦ Lower #s (1-3) occur more
frequently as a first digit than
higher numbers (7-9)
 In a sample of random numbers:
 #1 occurs 33%
 #9 occurs 5%
 What
are “random numbers”?
◦Non-manipulated numbers
 Population stats, utility bills,
 Areas of rivers
◦NOT human-selected #s
 Zip codes, SSN, Employee ID
 What’s
the practical use?
◦ 1990s – Dr. Mark Nigrini, college
professor
 Tested insurance costs (reim. claims), sales
figures
 Performed studies detecting under/overstmts
of financial figures
 Published results in J. of Accountancy (1990)
and ACFE’s The White Paper (1994)
◦ Useful for CFEs and auditors
 What
about financial txns?
◦“Random data” = nonmanipulated numbers
 AP txns, company purchases
◦NOT human-selected #s
 Expense limits (< $25)
 Approval limits (No sig < $500)
 Hourly wage rates
 How
will it help me with nonrandom data?
◦Aid in detection of unusual
patterns
 Circumventing controls
 Potential fraud
 You
won the lottery – invest
$100M in a mutual fund
compounding at 10% annually
◦ First digit is “1”
◦ Takes 7.3 yr to double your $
◦ At $200M, first digit is “2” ...
 At
$500M … First digit is “5”
◦ Takes 1.9 yr to increase $100MM
 Although time is decreasing,
there are more years that start
with lower digits
◦ Eventually, we will reach $1B
 First digit is “1”
 Seems
reasonable that the
lower digits (1-3) occur more
frequently
◦ These 3 digits make up approx.
60% of naturally-occurring digits
 Scale
invariant
◦ 1961-Roger Pinkham
◦ If you multiply the numbers by the
same non-zero constant (i.e., 22.04
or 0.323)
 New set of #s still follows Benford’s
Law
 Works
with different currencies
 $2M
Check Fraud in AZ
 $4.8M Procurement fraud in
NC
 Check
fraud in AZ
◦ #s appear random to untrained
eyes
◦ Suspicious under Benford’s Law
◦ Counter-intuitive to human
nature
 Wrote
23 checks (approx. $2M)
 Many amts < $100K
◦ Tried to circumvent a control that
required a human signature
 Mgr
tried to conceal fraud
 Human choices are not random
 Avoided
common indicators:
◦ No duplicate amounts
◦ No round #s – all included cents
 Mistakes:
◦ Repeated some digits / digit
combinations
◦ Tended towards higher digits (7-9)
 Count of the leading digit showed high
tendency toward larger digits (7-9)
 Anyone familiar with Benford’s Law would
have recognized the larger digit trend as
suspicious
 Benford’s
Law can be
extended to first 2 digits
◦ Allow examiner to focus on
specific areas
◦ High-level test of data
authenticity
 Procurement
fraud in NC
◦ 660 invoices from a vendor
◦ Years 2002-2005
◦ Total of $4.8M submitted for
payment
 Run
the 660 txns through
Benford’s Law …
See any suspicious areas?
Drilling down in the “51” txns
 Over
a 3-year period, at least
$3.8M in fraudulent invoices for
school bus and automobile parts
were submitted.
 The
investigation recovered
$4.8M from the vendor and
former school employees.
 Data
Analytics software
◦ ACL / IDEA
 Excel
◦ Add-Ons
◦ Built-in Excel Functions
 Expectations
 Background
 Why
it works
 Real-world examples
 How do I use it?
 Ed
Tobias
 [email protected]
◦ LinkedIn
 http://www.linkedin.com/in/ed3200

Benford’s Law Overview. n.d. Retrieved March 10, 2010 from
http://www.acl.com/supportcenter/ol/courses/course.aspx?cid=010&ver=9&mod=1&nodeKey=3

Browne, M. Following Benford’s Law, or Looking Out for No.
1. n.d. Retrieved March 10, 2010 from
http://www.rexswain.com/benford.html



Durtschi, C., Hillison, W., and Pacini, C. The Effective Use of
Benford’s Law to Assist in Detecting Fraud in Accounting
Data. 2004. Journal of Forensic Accounting. Vol. V. Retrieved
March 10, 2010 from http://www.auditnet.org/articles/JFA-V-1-17-34.pdf
Managing the Business Risk of Fraud. EZ-R Stats, LLC. 2009.
Retrieved March 10, 2010 from http://www.ezrstats.com/CS/Case_Studies.htm
Kyd, C. Use Benford’s Law with Excel to Improve Business
Planning. 2007. Retrieved March 10, 2010 from
http://www.exceluser.com/tools/benford_xl11.htm

Lehman, M., Weidenmeier, M, and Jones, T. Here’s how to
pump up the detective power of Benford’s Law. Journal of
Accountancy. 2007. Retrieved March 10, 2010 from
http://www.journalofaccountancy.com/Issues/2007/Jun/FlexingYourSuperFinancialSleuthPower.htm

Lynch, A. and Xiaoyuan, Z. Putting Benford’s Law to Work. 2008.
Internal Auditor. Retrieved March 10, 2010 from
http://www.theiia.org/intAuditor/itaudit/archives/2008/february/putting-benfords-law-to-work/

Nigrini, M. Adding Value with Digital Analysis. Internal Auditor.
1999. Retrieved March 10, 2010 from
http://findarticles.com/p/articles/mi_m4153/is_1_56/ai_54141370/

Nigrini, M. I’ve Got Your Number. Journal of Accountancy. 1999.
Retrieved March 10, 2010 from
http://www.journalofaccountancy.com/Issues/1999/May/nigrini.htm

Rose, A. and Rose, J. Turn Excel Into a Financial Sleuth. 2003.
Journal of Accountancy. Retrieved March 10, 2010 from
http://www.systrust.us/pubs/jofa/aug2003/rose.htm

Simkin, M. Using Spreadsheets and Benford’s Law to Test
Accounting Data. ISACA Journal. 2010, Vol. 1. Pp. 47-51.

Stalcup, K. Benford’s Law. Fraud Magazine. 2010, Jan/Feb. Pp
57-58.