Do Auditors Perceive Non-articulation
between Financial Statements as a Source of
Audit Risk
Daniel W. Collins (University of Iowa)
Hong Xie (University of Kentucky)
Kai Zhu (Shanghai University of Finance and Economics)
1
Research Question
Background…
 Bahnson et al. (1996) and Hribar and
Collins (2002) document the prevalence of
non-articulation, i.e.,


Changes in noncash C.A. and in C.L. accounts on
comparative B/S ≠ their corresponding changes on
the statement of cash flows
Non-articulation amounts (NARTA) = changes in
noncash C.A. and in C.L. accounts on comparative
B/S – their corresponding changes on the statement
of cash flows
2
Research Questions – continued
 Why does non-articulation arise?


Unusual transactions that are non-operating but affect
operating accounts (hereafter, non-articulation
transactions or events)
Case 1: Delphi Corp (Mulford and Comiskey 2005, p. 142)
(1)
(2)
(3)

a non-articulation transaction—arranging
for GE Capital to pay off its $287 million A/P
OCF is artificially inflated by $287 million
a positive NARTA of $287 million
Case 1(a): Delphi Corp - hypothetical
(1)
(2)
(3)
a non-articulation transaction—transferring
a piece of land to pay off its $287 million A/P
OCF is artificially inflated by $287 million
a positive NARTA of $287 million
 An otherwise financing (Case 1) or investing (Case 1(a))
cash inflows are classified as operating cash inflows,
thereby inflating OCF
3
Research Questions – continued
 Why does non-articulation arise?

Case 2: Lesco
(1)
(2)
(3)




a non-articulation transaction—selling most of its accounts
receivable to GE Capital
OCF is artificially deflated because Lesco reported the proceeds
as financing cash flows
a negative NARTA up to the amount of A/R factored
McAfee was sued and identified in a SEC AAER for multiple
counts of frauds during 1989-2000, which fraudulently inflated the
company’s revenues
To conceal the buildup of A/R that have little chance to be
collected, McAfee sold approximately $261 million A/R to banks
for cash during 1998-2000
McAfee’s sum of absolute NARTA during 1998-2000 is $73.429
million, about 28% of the A/R sold
McAfee’s attempt to conceal its frauds leaves a trail (nonarticulation) that is captured by NARTA
4
Research Questions – continued
 Why does non-articulation arise?

Case 3: Apple
(1)
(2)
(3)



a non-articulation transaction—employee stock options
accounting per SFAS 123 or APB 25 engenders NARTA because
the tax benefit reduces Income Taxes Payable on B/S but is
credited to APIC (Additional Paid-in Capital)
OCF is artificially inflated because most firms report the tax
benefit as part of OCF. Unlike other components of OCF, the tax
benefit is unlikely to recur (Siegel 2006; Hribar and Nichols 2007)
a positive NARTA equal to the tax benefit
Apple was sued and identified in a SEC AAER for misreporting in
2001 and 2002 related to backdating of stock options awarded to
Steve Jobs
Apple’s tax benefit from stock option exercise in 2002 = $28
million and its NARTA = $15 million (using current accounts only)
or $25 million (using current accounts + noncurrent accounts
shown in the operating section of Apple’s SCF)
Apple’s backdating results in non-articulation and thus can be
captured by NARTA
5
Research Questions – continued
 What’s the punch line?


As long as a cash flow management technique results in
non-articulation, that cash flow management technique
can be captured by NARTA
NARTA appears to be a nature measure of the amount
of managed OCF
6
Research Question – continued
 Do auditors perceive non-
articulation as a source of audit
risk?
7
Literature and Hypothesis
 1. Non-articulation between financial
statements



Drtina and Largay (1985) – many events lead to
non-articulation, including the changes in the
reporting entity (i.e., M&A and divestiture)
Huefner et al. (1989) – foreign currency
translation gains or losses lead to nonarticulation
Bahnson et al. (1996) – 75% of their large
sample of Compustat firms report non-articulated
changes between comparative B/S and
statement of cash flows (SCF).
8
Literature and Hypothesis –
continued
 1. Non-articulation between financial
statements




Hribar and Collins (2002) – accruals estimated
using the B/S approach (ACCbs) ≠ accruals
estimated using the SCF approach (ACCcf),
when a non-articulation event is present.
Recall: non-articulation events are nonoperating transactions that affect operating
accounts
We define NARTA = ACCbs – ACCcf
Hribar and Collins (2002) find that NARTA is
predictably large in magnitude when there are
M&As, divestitures, or foreign currency
translations
9
Literature and Hypothesis –
continued
 1. Non-articulation between financial
statements

The literature up to this point shows:
1)
2)
3)

Non-articulation is wide spread
Non-articulation is caused by non-articulation events or
transactions
NARTA is particularly large in magnitude when major
non-articulation events (i.e., M&A, divestiture, and
foreign currency translations) are present, although
NARTA is still present when none of the three major
non-articulation events is present
However, the literature up to this point does
not investigate whether NARTA contains useful
information
10
Literature and Hypothesis –
continued
 1. Non-articulation between financial
statements

Gong et al. (2014) are the first to conjecture
based on anecdotal evidence similar to Cases 1,
2, and 3 discussed earlier:
1)
2)

Managers can use non-articulation events to
manage (inflate or deflate) OCF
NARTA resulting from the non-articulation events
captures managed portion of OCF
Their findings support the above insights
 NARTA captures managed portion of OCF
and |NARTA| contains useful information
about a firm’s risk.
11
Literature and Hypothesis –
continued
 2. Audit risk and auditor decisions

The audit risk model: Audit risk = Inherent risk
× Control risk × Detection risk
 Prediction of the audit risk model: when
auditors perceive an increase in inherent
risk or control risk, they must try to reduce
detection risk by doing more substantive
testing  increase audit fees to
compensate for more effort and risk
12
Literature and Hypothesis –
continued
 2.1 A large volume of studies support the
prediction of the audit risk model



Gul and Tsui (1998) - firms of low growth with
high free cash flows are more likely to engage
in non-value-maximizing activities  auditors
perceive higher audit risk for these firms 
charge higher fees
Bedard and Johnstone (2004) - auditors plan
increased effort and billing rates for clients with
earnings manipulation risk
Hanlon et al. (2012) - larger absolute BTD
indicates high earnings management and low
reporting quality  auditors charge higher fees
for firms with larger absolute BTD
13
Literature and Hypothesis –
continued
 2.1 A large volume of studies support the
prediction of the audit risk model

Hogan and Wilkins (2008) - firms with internal
control weakness have high control risk 
auditors charge higher fees for firms with internal
control weakness
14
Literature and Hypothesis –
continued
 2.2 Auditors can also mitigate increased
audit risk by issuing a modified or goingconcern audit opinion



Carcello and Palmrose (1994) - MAOs issued
prior to bankruptcy reduce both the incidence
and magnitude of litigation if bankruptcy
subsequently occurs
Francis and Krishnan (1999) – larger absolute
accruals pose higher audit risk  auditors lower
the threshold for issuing MAOs
Kaplan and Williams (2013) – issuing goingconcern opinions reduces auditor litigation
probability
15
Literature and Hypothesis –
continued
 3. Hypotheses
To summarize, prior literature suggests the
following:


NARTA captures the managed portion of OCF
Firms with larger absolute NARTA are more
likely to be involved in cash flow management
activities  lower financial reporting quality
 Hypothesis 1: A firm’s financial reporting
quality is negatively associated with
absolute NARTA.
16
Literature and Hypothesis –
continued
 3. Hypotheses
Prior literature also suggests the following:

Auditors respond to an increase in perceived
audit risk by charging higher audit fees and
issuing MAOs or going-concern audit opinions
 Hypothesis 2a: Audit fees that a firm pays
are positively associated with absolute
NARTA.
 Hypothesis 2b: The probability for a firm to
receive a modified or going-concern audit
opinion is positively associated with
absolute NARTA
17
Research Design
 Measurement of non-articulation amounts
ACCbs = (ΔCA – ΔCASH) – (ΔCL – ΔSTDEBT) – DEP
(1)
ACCcf = – (CHGARcf + CHGINVcf + CHGAPcf + CHGTAXcf + CHGOTHcf) – DEPcf
(2)
|NARTA| = |ACCbs – ACCcf|
(3)
18
Research Design – continued
 Absolute NARTA and Reporting
Quality
1) SEC AAERS as a proxy for reporting Quality
following Armstrong et al. (2013)
AAER = a0 + a1|NARTA| + a2MKTCAP + a3BM + a4LEV + a5ROA + a6AGE
+ a7RETURN + a8STDRET + a9CAPEXP + a10INTANG + a11SGRW
+ a12RECINV + Industry dummies + ε
(4)
a1 > 0 is consistent with H1
19
Research Design – continued
 Absolute NARTA and Reporting
Quality
2) Accruals quality as a proxy for reporting
Quality following Butler et al. (2004)
AQ = c0 + c1|NARTA| + c2MKTCAP + c3BM + c4LEV + c5LagROA + c6LagROA2
+ c7CRATIO + c8BigN + Industry dummies + ε
(6)
c1 > 0 is consistent with H1
20
Research Design – continued
 Absolute NARTA and Auditor
Decisions
1) Audit fees following DeFond et al. (2002)
AUDFEE = d0 + d1|NARTA| + d2SIZE + d3BigN + d4ROA + d5RETURN + d7STDRET
+ d8LEV + d9SPE + d10RECINV + d11INSTN + d12BM + d13SEG
+ d14REPLAG + d15FINANCE + d16SGRW + d17FIRST2YRS
+ Industry dummies + ε
(7)
d1 > 0 is consistent with H2a
21
Research Design – continued
 Absolute NARTA and Reporting
Quality
2) MAOs or GC following DeFond et al. (2002)
MAO or GC = e0 + e1|NARTA| + e2BKRPTZ + e3SIZE + e4AGE + e5RETURN
+ e7STDRET + e8LEV + e9CLEV + e10LagLOSS + e11BigN + e12ROA
+ e13INVESTM + e14RECINV + e15SGRW + Industry dummies + ε
(8)
e1 > 0 is consistent with H2b
22
Research Design – continued
 Sample selection

Base sample
Accruals quality sample

Audit fees and going-concern sample

23
Findings – Table 4 Probit regression of
AAERs on |NARTA|
Variable
Intercept
Expected
Sign
|NARTA|
+
MKTCAP
+
BM
+
LEV
+
ROA
–
AGE
–
RETURN
–
Other controls
Industry dummies
N
Pseudo R2
Full
Sample
-6.574***
(-23.14)
0.680**
(2.510)
0.173***
(8.254)
0.081*
(1.677)
-0.297***
(-3.364)
0.239*
(1.823)
-0.147***
(-4.261)
-0.043*
(-1.647)
Included
Included
75,452
0.0870
AAER
Without
MNAE
-5.738***
(-20.48)
1.312***
(3.003)
0.159***
(7.233)
0.167***
(3.131)
-0.240*
(-1.756)
0.254*
(1.732)
-0.228***
(-4.814)
-0.009
(-0.287)
Included
Included
42,348
0.0923
With
MNAE
-6.786***
(-18.09)
0.171
(0.398)
0.189***
(8.266)
-0.014
(-0.222)
-0.378**
(-2.233)
0.230
(1.451)
-0.085*
(-1.867)
-0.072**
(-2.102)
Included
Included
31,032
0.0846
24
Findings – Table 5 OLS regression of
AQ on |NARTA|
Variable
Intercept
Expected
Sign
|NARTA|
+
MKTCAP
–
BM
–
LEV
–
LagROA
–
LagROA2
–
CRATIO
–
Other controls
Industry dummies
N
Adj. R2
Full
Sample
0.089***
(22.47)
0.146***
(16.08)
-0.007***
(-25.09)
-0.011***
(-9.906)
-0.025***
(-14.17)
-0.031***
(-6.339)
0.022***
(3.810)
-0.001***
(-4.964)
Included
Included
54,200
0.370
AQ
Without
MNAE
0.082***
(15.31)
0.181***
(14.76)
-0.007***
(-31.06)
-0.011***
(-10.48)
-0.023***
(-10.30)
-0.036***
(-6.998)
0.015***
(2.870)
-0.001***
(-3.755)
Included
Included
29,398
0.379
With
MNAE
0.094***
(216.7)
0.115***
(11.85)
-0.006***
(-24.28)
-0.011***
(-8.113)
-0.026***
(-10.93)
-0.025***
(-3.940)
0.032***
(3.651)
-0.001***
(-3.347)
Included
Included
24,802
0.342
25
Findings – Table 6 OLS regression of
Audit fees on |NARTA|
Variable
Intercept
Expected
Sign
|NARTA|
+
SIZE
+
BigN
+
ROA
–
RETURN
–
STDRET
+
LEV
+
Other controls
Industry dummies
N
Adj. R2
Full
Sample
-6.239
(0.000)
0.532***
(4.095)
0.598***
(41.08)
0.030
(0.405)
-0.586***
(-7.297)
-0.043
(-1.115)
-4.674
(-1.590)
-0.294***
(-6.603)
Included
Included
27,784
0.732
AUDFEE
Without
MNAE
-6.016
(0.000)
0.641***
(5.117)
0.553***
(30.62)
0.071
(0.970)
-0.555***
(-7.707)
-0.045
(-1.207)
-4.056
(-1.615)
-0.277***
(-4.688)
Included
Included
15,010
0.682
With
MNAE
-6.343***
(-31.30)
0.078
(0.376)
0.606***
(39.90)
-0.022
(-0.265)
-0.587***
(-5.614)
-0.048
(-1.162)
-5.338
(-1.520)
-0.307***
(-5.694)
Included
Included
12,774
0.735
26
Findings – Table 7 Probit regression of
MAOs on |NARTA|
Variable
Intercept
Expected
Sign
|NARTA|
+
BKRPTZ
+
SIZE
–
AGE
–
RETURN
–
STDRET
+
LEV
+
Other controls
Industry dummies
N
Pseudo R2
Full
Sample
-2.026***
(-7.596)
1.199***
(6.832)
0.202***
(3.124)
0.105***
(5.600)
0.145***
(5.103)
-0.004
(-0.127)
-0.618
(-0.245)
0.160
(1.603)
Included
Included
76,938
0.0518
MAO
Without
MNAE
-2.097***
(-5.300)
1.511***
(6.289)
0.275***
(4.113)
0.096***
(4.618)
0.132***
(4.284)
-0.013
(-0.361)
-0.231
(-0.110)
0.083
(0.791)
Included
Included
43,636
0.0506
With
MNAE
-1.878***
(-4.602)
0.764***
(4.283)
0.104
(0.830)
0.100***
(5.388)
0.160***
(5.204)
0.006
(0.170)
-1.186
(-0.364)
0.258**
(2.046)
Included
Included
33,302
0.0487
27
Findings – Table 8 Probit regression of
AAERs on |NARTA|
Variable
Intercept
Expected
Sign
|NARTA|
+
BKRPTZ
+
SIZE
–
AGE
–
RETURN
–
STDRET
+
LEV
+
Other controls
Industry dummies
N
Pseudo R2
Full
Sample
-1.959***
(-7.599)
1.239***
(3.665)
0.050
(0.312)
-0.150***
(-5.370)
0.119**
(2.534)
-0.278***
(-5.536)
12.049***
(8.399)
0.721***
(3.708)
Included
control
27,784
0.412
GC
Without
MNAE
-1.710***
(-6.794)
1.121**
(2.242)
0.128
(0.773)
-0.184***
(-6.453)
0.087
(1.637)
-0.300***
(-4.506)
11.033***
(7.200)
0.591***
(3.128)
Included
control
14,992
0.417
With
MNAE
-2.257***
(-6.190)
1.355**
(2.329)
-0.110
(-0.385)
-0.123***
(-3.470)
0.150**
(2.340)
-0.253***
(-2.843)
13.628***
(7.016)
1.020***
(2.756)
Included
control
12,698
0.417
28
Conclusion
 Absolute NARTA is positively associated with the
probability of a firm receiving AAERS.
 Absolute NARTA is negatively associated with
accruals quality.
 Auditors charge higher audit fees for firms with larger
absolute NARTA.
 Auditors are more likely to issue MAOs or going-
concern for firms with larger absolute NARTA.
 Larger absolute NARTA indicates lower financial
reporting quality
 Auditors charge higher audit fees and/or issue MAOs
to compensate for higher audit risk stemming from
cash flow management as captured by NARTA.
29
Future Research
 Relation between absolute NARTA and
properties of analyst earnings forecasts
and cash flow forecasts.
 Relation between absolute NARTA and the
cost of equity capital.
30
The End
Thank You!
31