Documento de Trabajo 09/01

Detecting earnings management in a

Spanish context
María del Mar Delgado
Luis Castrillo Lara
Universidad de Burgos

ABSTRACT

The aim of this study is to examine whether different accrual models

detect earnings management in a sample of Spanish firms that have
received GAAP audit qualifications.
The unexpected accruals obtained from the Jones model, the modified
Jones model, the Jones working capital model and the modified Jones
working capital model are significantly positive, supporting the
hypothesis that earnings have been managed upwards. However, the
unexpected accruals obtained from the margin model are not significantly
positive. We should be cautious in the interpretation of these results
because of the almost certainly bias towards the inclusions of the most
obvious and spectacular cases of earnings management. Potential extreme
financial performance bias can not be ruled out as explanation for these
results. An important limitation of this paper is the reduced number of
firms that constitutes the estimation sample and also the prediction
sample.
Keywords: earnings management, accruals, accrual models, GAAP audit
qualification.

DOCUMENTOS DE TRABAJO “NUEVAS TENDENCIAS EN DIRECCIÓN DE EMPRESAS” DT 09/01

dtecadem@eco.uva.es www2.eco.uva.es/ecadem
 1

1. INTRODUCTION

The measurement of accruals plays an important role in accounting literature.

This wide body of literature includes studies on the prediction of future cash
flows, the market´ s pricing and value relevance of accruals versus cash flows
(Bath et al. ,1999; Pfeiffer et al., 1998; Dechow, 1994), tests of earnings
management where there is a firm-specific event or a multiperiod strategic
approach (Healy, 1985; DeAngelo, 1986, 1988), McNichols et al., 1988; Jones,
1991; DeFond et al., 1994; DeFond et al., 1998; McCulloch, 1998; Erikson et
al., 1999), and market´s pricing of discretionary accruals versus non -
discretionary accruals (Subramanyam, 1996; Press et al., 1998). The focus of
this paper is the performance of different accrual models in testing earnings
management (hereafter EM).

Shipper (1989) defines EM as a purposeful intervention in the external financial

reporting process, with the intention of obtaining some private gain, as opposed
to a neutral participation. Some restriction on manager discretion over
accounting and other policies is expected, but some discretion will remain1.
Shipper (1999) has since proposed a modification to the definition that
identifies EM as “ implementation that impairs an element of decision
usefulness or implementation that is inconsistent with the intent of the
standards”.

Another interesting definition of earnings management is provided by Healy et

al. (1999). EM occurs when managers use judgement in financial reporting and
in structuring transactions to alter financial reports to either mislead some
stakeholders about the underlying economic performance of the company or to
influence contractual outcomes that depend on reported accounting numbers2.
Dechow et al. (2000) analyse different perceptions academics, practitioners and
regulators have of EM.

The problem researchers find when explaining accounting policy is that

managers have an informational advantage over researchers as well as incentive
to camouflage EM. Discretion exercised over accounting policy cannot be
directly observed. Different accrual models, through the estimation of what is

1
Positive accounting theory provides an interesting framework where the existence of accounting discretion can
be explained. Even though contracts that use accounting numbers to align managers and contracting parties´
interests would not be effective if managers had complete discretion over the reported accounting numbers,
managers are likely to know best which accounting procedures are optimal from the point of view of all
claimants, so some discretion will remain.
2
The use of accounting judgement to make financial reports more informative does not fall within this definition.

DOCUMENTOS DE TRABAJO “NUEVAS TENDENCIAS EN DIRECCIÓN DE EMPRESAS” DT 09/01

dtecadem@eco.uva.es www2.eco.uva.es/ecadem
2

called “non - discretionary, expected or normal accruals”3, provide earnings

management investigation with a benchmark for the exercise of accounting
discretion. So, the goal of these models is to separate total accruals into
discretionary and non discretionary components.

How well do “these models” work? This is a question that has been addressed
in several studies. In a market- based evaluation context, Guay et al (1996)
present evidence consistent with “considerable imprecision and / or
misspecification “ in existing actual models. New models or modification of
existing ones are often introduced by the improvements they make at generating
type I errors and type II errors4. The specification of the test statistic is
evaluated by examining the frequency with which they generate type I errors in
a randomly selected sample of firm years. The power of the test statistic is
evaluated by examining the frequency with which they generate type II errors in
a sample of firm years in which the researchers have artificially added a fixed
and known amount of accruals to each firm - year5.

The general finding of this evaluation is that the power of tests is low for EM of
economically plausible magnitudes and that the models produce reasonably well
specified tests for a random sample of event-years. When models are applied to
test firm years experiencing extreme financial performance, the models lead to
misspecified tests ( Dechow et al., 1995)6. This misspecification arises because
the magnitude of normal accruals is correlated with past and current
performance. This dependence occurs because firm performance conditional on
past performance does not follow a random walk and because both operating
accruals and operating cash flows are strongly mean reverting (Kothari, 2000).

In testing the power of various models, Kang (1999) points out that even though
simulation results are informative, there is no guarantee that accrual behaviour
of simulated data is reflective of real EM. Dechow et al. (1995) and Kang
(1999) evaluate the relative performance of different models in detecting EM on
a sample of firms for which they have strong prior beliefs that earnings have
been managed. These samples consists of firms that have been targeted by the

3
Due to the fact that the accrual models are really expectations models, Healy (1996) proposed a renaming of
some terms: “unexpected accruals” for “discretionary accruals” and “unexpected earnings ” for “discretionary
earnings“.
4
Dechow et al. (1995) and Kang et al. (1995) evaluate time - series models , and Jeter et al. (1999) and Peasnell
et al. (2000) evaluate cross - sectional models.
5
These simulations are similar to those performed by Brown and Warner (1985) in evaluating alternative models
for detecting abnormal stock price performance.
6
In the same way, Jeter et al. (1999) find evidence that the cross - sectional Jones model is misspecified for firms
whose cash flows deviate systematically from the industry median. However, extending the Jones Model to
explicitly control for cash flow from operations Jeter et al. (1999) show that the cash flow model is well specified
for all cash flow levels.

DOCUMENTOS DE TRABAJO “NUEVAS TENDENCIAS EN DIRECCIÓN DE EMPRESAS” DT 09/01

dtecadem@eco.uva.es www2.eco.uva.es/ecadem
 3

Securities and Exchange Commission (SEC) for allegedly overstating annual

earnings and firms suspected of earnings management to avoid losses7.

The objective of this study is to test whether different cross sectional models
detect earnings management in a sample of firms suspected of having managed
earnings. This sample consists of firms that have received audit qualification for
GAAP violation in Spain from 1991 to 1998. Our research is based on two
assumptions. The first is that auditors have correctly identify firms that violate
GAAP. And the second assumption is that these firms have first employed the
maximum discretion allowed within GAAP before using other accounting
procedures outside GAAP.

The economic determinants and the governance structure are key factors for
explaining the different accounting policies followed by firms. We are not
going to consider those factors that generate incentives and limitations to the
exercise of accounting discretion. Although we think they are very important,
the focus of this paper is limited to providing additional evidence on whether
different accrual models detect EM when it does really exist.

The average standardised prediction error is significantly positive for all models
except for the margin model, supporting the hypothesis that earnings have been
managed upwards. Although these results could serve as evidence on the
performance of alternative models measuring discretionary accruals in our
Spanish context, we should take into consideration when interpreting these
results that the prediction sample includes the more obvious and spectacular
cases of earnings management.

The remainder of the paper is organised as follows. The following section offers
a conceptual revision of existing models for testing EM. Firm sample is
described in section three. The results obtained are presented in section four and
section five provides the conclusion of the paper.

SECTION 2. ACCRUAL MODELS IN DETECTING EARNINGS

MANAGEMENT

Researchers interested in testing whether firms use discretionary accruals

(hereafter DA) to manage earnings must select a proxy for DA and they must
isolate situations (the event) where the direction of EM can be predicted
(McNichols et al., 1988). Although the ability of models to detect multi - period
EM is still an open issue (Kang, 1999), it is important to consider that the use of
accrual methology for detecting event-specific EM is just a part of its
applications. Accrual models have been used in tests of contracting and political

7
Burgstahler et al. (1997) find evidence of earnings management to avoid losses using a different methodology.

DOCUMENTOS DE TRABAJO “NUEVAS TENDENCIAS EN DIRECCIÓN DE EMPRESAS” DT 09/01

dtecadem@eco.uva.es www2.eco.uva.es/ecadem
4

cost incentives to manipulate accounting numbers, in tests of the efficient

contracting and opportunism hypothesis by correlating earnings with stock
returns8 and, in tests of the joint hypothesis of market inefficiency and accrual
manipulation with capital market motivation9.

In the first part of this section we present discretionary accruals (hereafter DA)
as a measure of the exercise of accounting discretion by managers in order to
alter the reported earnings (section 2.1.). As not all accrual decisions represent
cases of EM and as the discretionary exercise of accrual activity is not directly
observable by the researcher, they have to break down total accruals (hereafter
TA) into discretionary and non- discretionary elements. The discretionary
component is estimated by the difference between total accruals and the non -
discretionary accruals estimation from an expectation model (hereafter NDA).
This discretionary component will proxy for the real discretionary accruals
(section 2.2.). In section 2.3. we analyse conceptual differences between time
series and cross sectional models. We will finish this section by summing up
some statistical issues in regression analysis in order to better understand some
problems on inferences concerning EM.

2.1 Measures of earnings management

Managers attempting to arrive at a desired level of earnings, motivated from

event period or multiperiod incentives, can choose from a large set of
manipulation methods. Some of them may involve real decisions (i.e.,
operating, financing and investment decisions) and some are pure accounting
decisions (change in useful life of fixed assets, change in policy regarding
capitalizing - expensing repairs, changes in depreciation method). In some
cases, manipulation may be accomplished by a combination of a real decision
and an accounting decision (Jiambalvo 1996).

Once different ways used by managers for EM have been identified, the next
step is to find a proxy for measuring the discretion exercised by them. An
important issue is whether the discretion has been within GAAP or outside
them. In terms of GAAP choices, the first proxy for the exercise of accounting
policy discretion used in this research were accounting method choices or
changes in accounting method choices. Zmijewski et al. (1981) adopted what
they called “the income strategy approach” creating an EM weighted measure
from different accounting methods, trying to control for the portfolio nature of
income determination (Watts and Zimmerman,1990).

8
In these studies market efficiency is a maintained hypothesis.
9
See Kothari (2000) for more details of these three streams of research where accrual models have been used.

DOCUMENTOS DE TRABAJO “NUEVAS TENDENCIAS EN DIRECCIÓN DE EMPRESAS” DT 09/01

dtecadem@eco.uva.es www2.eco.uva.es/ecadem
 5

The use of operating total accruals, rather than specific accounting methods to
proxy for EM constitutes a great step forward in the EM literature. Healy (1985)
broke down earnings into cash flow from operations and TA, paying special
attention to the fact that accruals modify the timing of reported earnings, so that
it enables the manager to transfer earnings between periods. This author
suggests that it is more costly and more visible for managers to transfer earnings
between periods by changing accounting procedures than by changing accruals.
Moreover, managers appear to have greater flexibility to change accruals.
Another important feature of this measure is that it aggregates into a single
number the net effect of numerous recognition decisions, thereby takes into
consideration the portfolio nature of income determination as Zmijewiski et al.
(1981) hoped to do. However, accrual analysis constitutes a more
comprehensive measure in the sense that this measure includes both the effect
of accounting method choices and operating, financial and investment decisions
insofar as they affect accruals.

Manipulation outside of GAAP has been investigated in some studies. The

samples used in these studies consisted of firms subject to accounting
enforcement action by the SEC for alleged violation of GAAP (Dechow et al.,
1995, 1996; Beneish, 1997), firms that corrected overstating earnings in
previous financial statements (DeFond et al., 1991) or firms that disagreed with
their auditors over the use of income increasing accounting choices (DeFond,
1993).

The sample used in this paper consists of firms that have received audit
qualifications for GAAP violations. This sample is designed to provide
additional insight into the detection of EM by different accrual models.

Operating total accruals, defined as the difference between operating earnings

and Cash Flow from operations, does not necessarily include EM outside
GAAP. So, unless GAAP violations affect one of these components, TA will
remain unaffected. Even if operating total accruals includes some accruals
derived from GAAP violations, this is not the main focus of our investigation.
An implicit assumption made in this study is that managers of firms which
exercised accounting discretion outside GAAP, would have first used the
maximum discretion allowed within GAAP.

Unexpected accruals, defined as the prediction error from different accrual

models, will proxy for EM obtained through the discretionary application of
accruals. These abnormal accruals are hypothesised to be positive if accrual
models have correctly separated TA into discretionary and non discretionary
accruals.

DOCUMENTOS DE TRABAJO “NUEVAS TENDENCIAS EN DIRECCIÓN DE EMPRESAS” DT 09/01

dtecadem@eco.uva.es www2.eco.uva.es/ecadem
6

2.2. Accrual prediction models

The proxy for discretionary accruals is obtained after subtracting non -

discretionary accrual estimation from total accruals10. The parameters that shape
the NDA model are estimated through an “estimation period” during which no
systematic earnings management is predicted in time series accrual models, or
through the use of “an estimation sample “ from the same industry and year,
providing a benchmark for the exercise of accounting discretion.

The following models range from simple models in which DA are measured as
TA, to more sophisticated models that attempt to control for economic
circumstances, correlation structures of accruals and cash flows or statistical
issues.

2.2.1. The Healy Model

Healy (1985) tests EM comparing mean TA across different firms samples. His
partitioning variables divide the sample into different groups with earnings
predicted to be managed upwards or downwards. Inferences in EM are made
through pairwise comparisons of the mean TA. This approach is equivalent to
treating the mean accruals of a sample as the estimation period and the other as
the event period.
∑ TA t

NDAτ = t

T
where,
NDA = estimated nondiscretionary accruals;
TA = total accruals scaled by lagged total assets;
t = 1,2,… T is a year subscript for years included in the
estimation period; and
τ = a year subscript indicating a year in the event
period.

2.2.2. DeAngelo Model

Last period total accruals is the measure of NDA in this model. In contrast to
the following models, the Healy model and the DeAngelo use TA from the
estimation period to proxy for NDA, without taking into account the response of
normal accruals to changes in economic circumstances.

10
In order to make the nomenclature used in this paper simpler, we do not differentiate either discretionary
accruals from discretionary accruals proxy, or non - discretionary accruals from non discretionary accruals
estimation although it is important to know that researchers will never work with the real discretionary and non-
discretionary accruals because they are not observable.

DOCUMENTOS DE TRABAJO “NUEVAS TENDENCIAS EN DIRECCIÓN DE EMPRESAS” DT 09/01

dtecadem@eco.uva.es www2.eco.uva.es/ecadem
 7

NDA τ = TA τ -1

This model assumes that the non discretionary component of accruals follows a
random walk, which is inconsistent with the self reverting property of accruals.

2.2.3. The Friedlan Model

Since growth should affect certain aspects of a firm´s operations, including

accruals, the random walk assumption underlying the DeAngelo model is not
valid for growing firms. Friedlan (1994) introduces a modification to the
DeAngelo model that attempts to control for the effect of growth on total
accruals. This model assumes a constant proportionality between total accruals
and sales in successsive periods. The non discretionary accruals for the Friedlan
model are computed as:

NDAit = TAit-1 / Sit-1

2.2.4. The Jones Model

Jones (1991) determined NDA for each firm by regressing total accruals on the
change in sales revenues (REV) during the period and the gross level of
property, plant and equipment (GPPE) using a time - series of observations.
These variables were included to control for changes in the firm´s economic
circumstances, GPPE to control for the portion of TA related to non -
discretionary depreciation expense, and REV to control for the normal
(unmanaged) level of current accruals. Sales revenues is an objective measure
of the firm´s operation before manager´s manipulation, so that this variable will
control for the economic environment of the firm11.
NDA τ = α 1 (1/ A τ -1) + α 2 (∆REVτ )+ α 3 (GPPEτ )
where,
∆REVτ = revenues in year τ less revenues in year τ - 1 scaled
by total assets at τ-1;
GPPEτ = gross property plant and equipment in year τ scaled
by total assets at τ-1;
A τ -1 = total assets at τ-1; and
α 1 , α 2 ,α 3 = firm-specific parameters.
Estimates of the firm-specific parameters, α 1 , α 2 ,α 3, are
generated using the following model in the estimation period:
TA t = a 1 (1/ A t -1) + a 2 (∆REVt )+ a 3 (GPPEt ) + vt

11
Jones (1991) points out that this variable is not completely exogenous because revenues may be affected to
some extent by managers´ attempt to modify reported earnings.

DOCUMENTOS DE TRABAJO “NUEVAS TENDENCIAS EN DIRECCIÓN DE EMPRESAS” DT 09/01

dtecadem@eco.uva.es www2.eco.uva.es/ecadem
8

where, a 1 , a 2 , a 3 denote the OLS estimates of α 1 ,α 2 , α 3

and TA is total accruals scaled by lagged total assets.

The expected sign for the GPPE coefficient is negative because it is related to
depreciation expense. However, the expected sign for the change in revenue
coefficient in not obvious, since a given change in revenue can bring about
income - increasing changes in some working capital accounts but income
decreasing in others.

Another important assumption made in this model is that changes in current

assets and current liabilities are both driven by changes in revenues. Current
liabilities, such as payables, are more likely to be related to expenses than to
revenues. The omission of this relevant variable produces a potential bias
towards finding positive managed accruals in an economic upswing and
conversely in an economic downswing (Kang, 1999).

Systematic EM via the depreciation accruals is likely to have limited potential

because the effect of changes in asset lives or in the depreciation method are
required to be disclosed, so the associated earnings effects are comparatively
sample to adjust ex – post. Because of this some authors have tested the Jones
working capital model12.

2.2.5. The Modified Jones Model

Considering the possibility of EM contamination of revenues, the modified

Jones model assumes that discretion over revenue is not exercised in the
estimation period and that changes in credit sales in the event period result from
EM. This model provides a more powerful means of isolating revenue - based
manipulation than the Jones model (Beneish, 1998). But if no EM occurs in the
event year, unmanaged accruals will be understated by α2∆REC, so managed
accruals are overstated if ∆ REC > 0.
NDA τ = α 1 (1/ A τ -1) + α 2 (∆REVτ - ∆RECτ )+ α 3 (GPPEτ )
where
∆RECτ = net receivables in year τ less net receivables in
year τ - 1 scaled by total assets at τ - 1.

2.2.6. The Industry Model

The Industry model was used by Dechow et al. (1991). This model assumes that
variation in the determinants of NDA are common throughout firms in the same

12
Janin (2000) provides evidence of the importance of working capital accruals. Although depreciation and
amortisation explain the greatest part of total accruals, working capital accruals among all the total accruals
components, exhibit the greatest variability.

DOCUMENTOS DE TRABAJO “NUEVAS TENDENCIAS EN DIRECCIÓN DE EMPRESAS” DT 09/01

dtecadem@eco.uva.es www2.eco.uva.es/ecadem
 9

industry, without directly controlling for the determinants of NDA. In this

model NDA for the event period will be estimated as follows:
NDA τ = γ1 + γ2 median1 (TA τ ),
where,
medianI (TA τ ) is the median value of total accruals scaled by
lagged assets for all non-sample firms in the same 2-digit SIC
code.

2.2.7. The Kang and Shivaramakrishnan Model

Kang and Shivaramakrishnan (1995) developed the following instrumental

variable model:
[ACCBT/ AT-1] = Φ0 + Φ1[δ1REVT/ AT-1] + Φ2 [δ2EXPT/ AT-1] +
Φ3[δ3GPPET/AT-1]
where,
δ1 = ARTT-1/ REVT-1 , δ2 = OCAL T-1/ EXPT-1 ,
δ3 =DEP T-1/ GPPET-1
ACCBT = accrual balance, CAI, T - CASH I, T - CL I, T - DEPI, T
AT = net total asset
CA T = current assets
CASH T = cash
CL T = current liabilities excluding current portion of long term
debt
DEP T = depreciation and amortization
ARTT = receivables
OCAL T = all current asset less liabilities excluding accounting
receivable, CAI, T - ART I, T - CASH I, T - CL I, T
REVT = net sales revenues
EXPT = all expenses other than depreciation expense
GPPET = gross property plant and equipment

The three important features of this model are (Kang, 1999):

- This model solves three important statistical problems that OLS estimation of
Jones model has, simultaneity, errors in variables and omitted variables, by
including expenses in the regression and using an instrumental variable
approach.

- The estimated managed accruals are calculated using the level, rather than the
change of current assets and current liabilities.

DOCUMENTOS DE TRABAJO “NUEVAS TENDENCIAS EN DIRECCIÓN DE EMPRESAS” DT 09/01

dtecadem@eco.uva.es www2.eco.uva.es/ecadem
10

- The parameters δ1, δ2, δ3 can be interpreted as turnover ratios which

accommodate firm - specificity and compensate for the fact that the equation
will be estimated from a pooled sample13.

2.2.8. The Margin Model

Like Jones - style procedures, Peasnell et al. (2000) model abnormal accruals on
accounting as prediction errors from an OLS regression of accounting accruals
on a vector of explanatory variables designed to capture “normal” accrual
activity. Working capital accruals is expressed as an explicit linear function of
two drivers, sales and cash received from customers.

WCAt = ω + γ1 REVt + γ2 CRt + et

The parameter estimates of this model γ1 and γ2 have clear economic

interpretations14:

γ1 reflects the average value of the sales margin and is predicted to be

positive

γ2 reflects the average value of the cash margin and is predicted to be

negative.

2.2.9. The McCulloch Model

The structure of this model is derived directly and explicitly from the structure
of financial statements15. The most important feature of this model is its
multiperiod approach. The effects of accrual reversals in future periods is an
important issue to be taken into account when initiating discretionary accruals
in the current period. Managers are thinking not only about the current period
but also about future earnings. The model is expressed as follows:
∆ WCt = β0 + β1 AccRec t + β2 RevRatiot + β3 CFOt +
H

∑θ NEWDA t-1 + NEWDA t

i =1
i

where,
AccRec = accounting receivables
Revratio = (Revt-1accRect)/AccRect-1
13
Kang and Shivaramakrishan (1995) based their estimation on the longest available time - series of data. Kang
(1999) based his estimation on a pooled cross - sectional firm´s data. Another difference is that the first work
excludes tax related accruals whereas Kang (1999) includes tax related accruals.
14
See Peasnell, Pope and Young (2000) for the formal analysis of the model. As these authors indicate, the
primary difference between the margin model and the Jones working capital model is that the margin model
disaggregates the change in revenue term into two components at the parameter fitting stage, substituting cash
receipts in the current period for revenues in the prior period.
15
See McCulloch (1998)

DOCUMENTOS DE TRABAJO “NUEVAS TENDENCIAS EN DIRECCIÓN DE EMPRESAS” DT 09/01

dtecadem@eco.uva.es www2.eco.uva.es/ecadem
 11

H
DAt = ∑θ NEWDA t-1 + NEWDAt
i =1
i

CFOt= cash flow from operations

The improvement of this model in accrual technology for investigating EM

using DA can be seen through the incorporation of:

the reversal of discretionary accruals into the estimation procedure, paying

more attention to the duration of the effect on EM and the time -series
properties of accruals,

a contemporaneous negative correlation between DA and NDA, this

correlation arises from the incentives that lead some managers to use DA to
offset unfavourable NDA and from researcher measurement error, and

a relation between accruals and Cash Flow, this relation arises from the
smoothing action of GAAP driving a negative relation between change in net
cash flow and unmanaged accruals, and as a result of discretionary smoothing
in which Cash flow will be negatively correlated with discretionary accruals.

These innovations cannot be accommodated by the ordinary least squares

approach to estimation because it requires the assumption that the regression
(modelling NDA) and the residuals (proxing for DA) are uncorrelated. The
author adopts a non linear instrumental design that allows the two components
to be (negatively) correlated, and that allows the residual to follow the moving -
average process.

2.2.10. The CFO Model

Shivakumar (1996) argues that a non - linear specification for cash flow from
operation (hereafter CFO) is desirable since CFO may vary between firms in the
estimation sample, due either to differences in the long - run level of return on
assets or to matching and/or timing problems in CFO. This possibility has been
implemented in the model allowing the slope coefficient on CFO to vary
between firms as follows:
NDAit = K0 + K1 ∆REVit / Ait-1 + K2 GPPEit / Ait-1 + K3 d1it
*CFOi / Ait-1 + κ4 d2it *CFOi / Ait-1 + Kκ5 d3it *CFOi / Ait-1 + Kκ6
d4it *CFOit / Ait-1 + Kκ7 d5it *CFOit / Ait-1.
where,
CFO = cash from operations

Firms in each estimation sample are sorted into quintiles based on CFOit/ Ai t-
1.

DOCUMENTOS DE TRABAJO “NUEVAS TENDENCIAS EN DIRECCIÓN DE EMPRESAS” DT 09/01

dtecadem@eco.uva.es www2.eco.uva.es/ecadem
12

This model has been tested by Jeter et al. (1999) using quarterly and annual
data. An important feature of this model is that it is well specified for all cash
flow levels.

2.2.11. The accounting process model

Dechow et al. (1995) and Guay et al. (1996) point out that the problems of the
accrual models may derive from ignoring the time series properties and
correlation structure of accruals and cash flows. The Cash flow model and the
Mc Culloch Model we have just referred to, constitute the first attempts to
incorporate cash flows to the modelling of non discretionary accruals.

Garza- Gómez et al. (1999) introduce a discretionary accrual model based on

the accounting process. “The Accounting Process Model” is designed to reflect
the cross correlation between earnings, accruals and cash flows as well as their
serial properties, which arise naturally from applying GAAP. This model is
expressed as follows:
NDA= φ0 (1/Ait-1) + φ1 (WKit-1 /Ait-2) + φ2(LTAit-1 / Ait-2) + φ3(
CFOit/At-1) + φ4 (CFO it-1 / Ait-2)
Where,
WK = working capital accruals
LTA= long term accruals
CFO = cash flow from operations

2.3. TIMES - SERIES AND CROSS SECTIONAL MODEL ESTIMATION

The sample used for estimating expected accruals should be taken into account
when interpreting parameter estimations. The estimation samples used in time
series models are past data of the same firms. During this estimation period no
systematic earnings management are expected to occur. Past accrual activity of
the firm is considered a benchmark to determine the DA for an event - specific
study16. This approach suffers from a survivorship bias (ten observations at least
are required to estimate time - series models). Moreover, this methology
introduces a selection bias as firms with such a long time series are more likely
to be large, mature firms with greater reputational capital to lose if EM is
uncovered (Jeter et al. , 1999). Apart from survivorship, other important aspects
of these approaches are:

- Some structural economic changes may have occurred in such a long time.
Although these models relax the assumption that NDA are constant over

16
Time - series models generate firm - specific parameter estimations.

DOCUMENTOS DE TRABAJO “NUEVAS TENDENCIAS EN DIRECCIÓN DE EMPRESAS” DT 09/01

dtecadem@eco.uva.es www2.eco.uva.es/ecadem
 13

time, they consider a stationary relation between NDA and its explanatory
variables over time.

- The self reverting property of accruals may introduce specification problems

in the form of serially – correlated residuals.

- The discretion available through this period is expected to have changed

because of changes in accounting rules and changes in firms’ governing
structures.

Cross sectional models do not make any assumption regarding systematic

earnings management for the estimation sample, but they implicitly assume that
the model parameters are the same in all firms in an estimation sample (Jeter et
al., 1999). The replacements of time – series models by their cross – sectional
counterparts introduces new problems. “Abnormal accruals” estimated from the
models should be interpreted as abnormal accruals relative to the industry
benchmark. If EM is expected to occur in several firms in an industry due to
specific situations, the cross - sectional approach will introduce a bias against
finding evidence of EM. So, EM tests will reject the null hypothesis only when
accrual activity is significantly different from that exercised by firms in the
same industry. Moreover, cross - sectional models are less likely to capture the
effects of a) mean reversion in accruals, and b) dynamic accrual management
strategies (Peasnell et al., 2000). Related to the number of firms in the sample,
cross - sectional models have the potential to generate larger samples than time
- series models.

2.4. STATISTICAL ISSUES

The real discretionary component of TA is not possible to be observed by the

researchers, so they are forced to estimate the discretionary element by
imposing on TA an expectation model of the non - discretionary component.
Another source of problems in event - EM investigation is the omission of
relevant variables for explaining DA and NDA( which induces measurement
error in DA proxy).

Tests of event - specific EM suffer from important statistical issues, which

consequences Dechow et al. (1995) sum up in :

Incorrectly attributing EM to Part (dummy variable, 1 for the event year and
zero otherwise) / unintentionally extracting EM caused by Part. This will
occur when Part is correlated with the variables omitted explaining DA or the
error in the researcher´s proxy for DA.

Low power test because of the exclusion of relevant (uncorrelated) variables.

DOCUMENTOS DE TRABAJO “NUEVAS TENDENCIAS EN DIRECCIÓN DE EMPRESAS” DT 09/01

dtecadem@eco.uva.es www2.eco.uva.es/ecadem
14

Young (1999) has paid special attention to the problems arising from
measurement error in DA proxy and the omission of relevant variables
explaining DA. The source and magnitude of this measurement error depends
on the effectiveness with which the expectation model of NDA controls for the
factors that determine the non - discretionary component of TA. Results suggest
that the five models he analyses induce systematic measurement error in DA as
a function of operating cash flow performance, sales growth and asset
structure. Moreover, this author controls for relevant variables at explaining
discretionary exercise of accrual activity, that is variables that control for
differences in the propensity for earnings management, such as leverage,
managerial equity ownership, size and income smoothing. The results obtained
for some of these variables demonstrate the possibility for erroneous EM
inferences when the DA proxy contains predictable measurement errors.

Hansen (1999) has also examined measurement error in discretionary accrual

models. He finds that structural changes are associated with the direction and
magnitude of discretionary accrual estimates and that the measurement error
associated with these structural changes is correlated with the level of earnings.
Since the level of earnings is often used by researchers to partition EM
incentives or is correlated with the partitioning variable, this bias in
discretionary accrual models generated by structural changes could cause
researchers to reach incorrect EM conclusions.

Focusing on the non - discretionary accrual estimation by ordinary least

squares, Kang (1999) identifies three statistical problems applying to the Jones
model, simultaneity, errors in variables and omitted variables. Simultaneity
arises because of failure to incorporate the identity constraint “accruals =
earnings - cash flow from operations”. Errors in variables are due to the fact that
regressors can be manipulated, and the omitted variable problem results when
key determinants of unmanaged accruals are omitted. These problems will be
solved by introducing expenses as a variable in the model and by applying an
instrumental variable approach.

3. SAMPLE

The sample of firms used to evaluate the models consists of 27 firms listed on
the Madrid Stock Exchange that have first received audit qualification for
GAAP violation from 1991 to 1998. As results from these violations earnings
have been overstated17. Financials are excluded from the sample because of
fundamental differences in the nature of their accruals and cash flows that are
not captured by expectation models of normal accrual activity. The main source
of information has been the database “Auditoría de cuentas anuales” published
17
One firm of this sample has an audit qualification whose effect is an overstatement in equity.

DOCUMENTOS DE TRABAJO “NUEVAS TENDENCIAS EN DIRECCIÓN DE EMPRESAS” DT 09/01

dtecadem@eco.uva.es www2.eco.uva.es/ecadem
 15

by the “National Stock Market Commission” on a computer support, which

includes the annual accounts and audit reports of most of the companies under
the control of this commission.

TABLE 1

Data available has severely limited empirical testing of accrual models. A time
series approach cannot be applied since consistent definitions of earnings,
accruals and cash flows would restrict portfolio estimation to the years after
199018.

The cross sectional approach we adopt requires a sufficient number of

observations in the industry - year portfolio, so all firms with fewer than six
were excluded. Only firms with unqualified opinions from auditors are included
in the sample estimation. As this sample remains very small ( mean 12 and
median 10) , we tried to solve this problem by including all industry portfolio
from the same year 19 and introducing a set of industry dummy variables. The
final sample comprises 241 firm year observations spanning 7 industry groups.
The mean (median) number of firms in the year portfolio was 26 (27). Annual
sample sizes are in table 1. Finally, we have estimated NDA from a pooled
sample, considering an industry - year set of dummy variables and an industry
set of dummy variables.

Data available has restricted the models that can be tested in this study to the
Jones model, The modified Jones model, The Jones working capital model, The
modified Jones working capital model and the margin model. The information
needed to compute TA definition and its explanatory variables is obtained from
firms’ annual reports.

4. DATA ANALYSIS AND RESULTS

Consistent with previous studies of EM, TA is computed as:

TA t = (∆ Current Asset - ∆ Cash )- (∆ Current Liabilities - ∆ Short - term debt)

- Depreciation and amortization expense

WAt = (∆ Current Asset - ∆ Cash )- (∆ Current Liabilities - ∆ Short - term debt)

Contingency estimates are taken into account in the computation of TA through

the variable Dep. Tax related accruals are not included in this definition, except

18
Apellániz et al. (1995) find that the change in Spanish accounting regulation, the 1990 general accounting plan,
reduced management´s opportunities to manipulate earnings.
19
Only industry portfolios where there is a firm with GAAP audit qualification have been considered in the
estimation sample.

DOCUMENTOS DE TRABAJO “NUEVAS TENDENCIAS EN DIRECCIÓN DE EMPRESAS” DT 09/01

dtecadem@eco.uva.es www2.eco.uva.es/ecadem
16

from tax contingencies. Another important feature of this measure is that it does
not include accruals from extraordinary items.

Ordinary least squares is used to obtain the coefficient estimates for different
models. Prediction errors represent the level of discretionary or abnormal
accruals.

Ujp = TAjp / Ai p-1 - NDA

Significance tests are computed using standardised prediction errors which are
computed as:

Vjp = ujp / s (ejt ),

where

s (ejt) is the standard deviation of the error term from the cross sectional model
estimated for firm j.

Parametric significance tests of the standardised prediction errors are computed

as
Zvp = ∑ Vjp / [∑ (Tj - k)/ (Tj – ( k + 2))] 1/2
Where,
Tj is the number of firms in the estimation sample
K is the model degree of freedom

Table 2 provides descriptive statistics on the parameter estimates and the test
statistic generated (except for the independent term) for the Jones model, the
Jones working capital model and the margin model20.

TABLE 2

The GPPE coefficient presents the expected sign in seven out of eight years in
the Jones model. The two coefficients in the margin model also present the
expected sign but they cannot reflect the sales margin or the cash margin
because they are too high in most of the years. The two variables of this model,
revenues and cash received, are highly correlated21.

20
The estimation results for the modified Jones model are the same that estimation results for the Jones model.
As Beneish (1998) points out ∆REVt - ∆RECt can be rewritten as (CRt - REVt-1) where CRt equals Cash
Received in period t and re – specification of the modified Jones model in this way highlights the absence of any
strong economic intuition for the purposed adjustment in the estimation sample.
21
We cannot pay too much attention to the coefficient estimates in 1997 since the estimation sample consists
only of 7 firms. The Z statistics remain similar without including the firm that has received GAAP audit
qualification in 1997.

DOCUMENTOS DE TRABAJO “NUEVAS TENDENCIAS EN DIRECCIÓN DE EMPRESAS” DT 09/01

dtecadem@eco.uva.es www2.eco.uva.es/ecadem
 17

When referring to the pooled sample, all coefficient estimates present the
expected sign and are significant at one percent in the margin model. The GPPE
coefficient is also significant at one percent in the Jones model and the revenue
coefficient is significant at five percent in the Jones working capital model.
However, revenue variable is not significant in the Jones model22.

As we can see in table 3, the Z statistic is positive and statistically significant at

conventional levels for all models but for the margin model, so we can reject
the null hypothesis that the average standardized prediction error is less or equal
to zero. The results obtained from the modified Jones model and the modified
Jones working capital model are very similar to the Jones model and the Jones
working capital model, respectively23.

TABLE 3

Although the results obtained could serve as evidence on the performance of

alternative models measuring discretionary accruals in our Spanish context, we
should be cautious when interpreting them because the prediction sample
includes the more obvious and spectacular cases of earnings management and
also because of the small number of firms available for the models estimations.

As contingency estimates may be subject to important manager´ judgement and

although they are part of long term accruals they are related to current variables,
we have also estimated the working capital accrual models considering them.
However, the results are not different from these shown in this paper24.

Accounting literature has shown us that existing accrual models are not useful
for firms experiencing extreme financial performance. The accrual models
tested by Dechow et al. (1995) reject the null hypothesis of no earnings
management at rates exceeding specified test levels when applied to samples of
firms with extreme financial performance. High rejection rates arise because
firm – years with low (high) earnings also tend to have low (high) total accruals
and accruals models attribute part of the lower (higher) accrual to negative
(positive) discretionary accruals. On the other hand, low (high) cash flow from
operation samples have high (low) total accruals resulting in an over – rejection
( under – rejection) of the null hypothesis that earnings management is less or
equal to zero (more or equal to zero). The results of the investigation carried out
by Peasnell et al. (2000) indicate that the margin model generates relatively
22
The results remain similar when considering an industry set of dummy variables instead of an industry - year of
dummy variables for the pooled sample. Data are available upon request to the authors.
23
That is, the modified Jones model and the modified Jones working capital model do not outperform the Jones
model and the Jones working capital model. Although the revenue coefficient estimated in these models is
positive, these results are due to the fact that the change in receivables is not very important and sometimes the
level of receivables is even less than the year before.
24
Data are available upon request to the authors.

DOCUMENTOS DE TRABAJO “NUEVAS TENDENCIAS EN DIRECCIÓN DE EMPRESAS” DT 09/01

dtecadem@eco.uva.es www2.eco.uva.es/ecadem
18

better specified estimates when cash flow performance is extreme than the
Jones working capital model and the Modified Jones working capital model.
These findings are of particular interest for tests of EM in response to a
stimulus. If the stimulus selected by the researcher is correlated with firm
performance, false causal determinants of EM may be attributed to the stimulus.

Since the results of this paper could be driven by a potential extreme

performance sample bias, we have analysed whether our sample belongs to the
highest earnings decile or to the lower cash flow decile. Although only one firm
belongs to the highest earnings decile, nine firms could be classified in the
lowest cash flow decile, suggesting that our results could be due to the problem
of the over – rejection of the null hypothesis that EM is less or equal to zero as
has been pointed in Dechow et al. (1995).

Some researchers worried about the limited usefulness of accrual models for
firm with extreme performance, have developed a model that provides a means
of assessing the likelihood of opportunistic reporting among firms with large
discretionary accruals. This is the case of Beneish (1997)25. This model adds to
the modified Jones model two variables, lagged total accruals and a measure of
past price performance, improving the specification of this model.

Despite these models having a general application, with the exception of firms
experiencing extreme financial performance, it is important to jointly consider
the situation and form in which earnings management is expected to occur and
the models of non discretionary accruals. This implies taking into account the
different models´ ability to detect specific forms of earnings management26 and
some particularities of a country’s accounting regulations. The findings in
Peasnell et al. (2000) suggest that the Jones working capital model and the
modified Jones working capital model are substantially more powerful at
detecting subtle revenue and bad debt manipulations. However, the margin
model outperforms the Jones working capital model and the modified Jones
working capital model. So, these authors suggest that the use of several models
in combination may improve the detection of accrual management, the specific
form of which is impredictable.

Young (1999) evaluates discretionary accruals estimated by different models

using a method that controls for potential variation in the propensity for EM
activity. If firms in the industry present a high propensity to manage earnings,
the benchmark provided by cross sectional models may also be too high, so that
it will be difficult to detect earnings management.
25
This author proposes the use of this model where extreme financial performance is likely, such as security
offerings and financial distress.
26
McNichols et al. (1988) model the provision for bad debts in the absence of EM and use the expected
provision for that model to test EM.

DOCUMENTOS DE TRABAJO “NUEVAS TENDENCIAS EN DIRECCIÓN DE EMPRESAS” DT 09/01

dtecadem@eco.uva.es www2.eco.uva.es/ecadem
 19

5. SUMMARY AND CONCLUSION

Several cross sectional models (the Jones model, the modified Jones model, the
Jones working model, the modified Jones working capital model and the margin
model) have been examined in a Spanish context. The sample of firms
suspected of having managed earnings consists of firms that have received
GAAP audit qualifications, so these firms represent the most obvious and
spectacular cases of earnings management. We select this sample in order to test
the ability of different accrual models to detect EM when it does really exist in a
Spanish context. The abnormal accruals detected are significantly positive for
all models except for the margin model. Potential extreme financial
performance bias can not be ruled out as an explanation for these results. An
important limitation of this paper is the estimation and the prediction sample
size, so these results should be interpreted cautiously.

By adding evidence on the usefulness of different accrual models as a tool to

detect EM, the results of this paper contribute to the EM literature in two ways.
First, we provide additional evidence on whether different accrual models detect
EM by using a sample of firms suspected of having managed earnings instead
of a sample where a fixed and known amount of accruals have been added to
each firm. Second, this paper provides additional evidence on the use of accrual
models in a different accounting regime, such are the Spanish accounting
regulation, characterised by a high degree of conservatism introduced by the
principle of prudence.

Future investigation into earnings management must seek to improve accrual

models technology to detect earnings management. Young (1999) has identified
the source of systematic measurement error but it remains to be seen how to
control for them in new models.

An issue of a great importance to be considered in EM investigation is the

relation between the context in which EM is hypothesized and the NDA model.
This context includes special GAAP features in countries too. Models have
different ability to detect specific forms of EM activity so this ability and the
specific form of EM one expects to find, should be taken into account when
choosing among different models. Since the discretionary accrual model
provides us with a benchmark for the determination of the exercise of
discretionary accruals activity, the propensity and restriction for EM of sample
estimation firms is specially relevant in cross sectional models.

Another important issue is that TA definition does not include extraordinary

items. The accounting discretion exercised through these items may be

DOCUMENTOS DE TRABAJO “NUEVAS TENDENCIAS EN DIRECCIÓN DE EMPRESAS” DT 09/01

dtecadem@eco.uva.es www2.eco.uva.es/ecadem
20

considered in this kind of investigation as complementary means of EM

(McCulloch, 1998).

References
Apellániz, P. and Labrador, M. (1995) “El impacto de la regulación contable en
la manipulación del beneficio. Estudio empírico de los efectos del PGC
de 1990”, Revista Española de Financiación y Contabilidad, 24(82): 13 –
40.
Barth, M.E., Beaver, W.H., Hand, J.R.H., and Landsman, W.R. (1999)
”Accruals, Cash Flow and Equity Values”, Working Paper Standford
University and University of North Carolina.
Beneish, M.D. (1997) “Incentives and penalties related to earnings
overstatements that violate GAAP”, The Accounting Review, 74 (4): 425 -
457.
Beneish, M.D. (1998) “Discussion of “Are accruals during initial public
offerings opportunistic?”, Review of Accounting Studies, 3 (1-2): 209-
215.
Brown, S.J., and Warner, J.B. (1985) “Using daily stock returns: the case of
event studies”, Journal of Financial Economics, 14: 3 – 31.
Burgstahler, D., and Dichev, I. (1997) “Earnigns management to avoid earnigns
decreases and losses”, Journal of Accounting and Economics, 24 (1): 99 -
126.
Collins, D.W. and Hribar, P. (1999) “Errors in estimating accruals: implications
for future research”, . Working paper, University of Iowa.
DeAngelo, L.E. (1986) “Accounting numbers as market valuation substitutes: a
study of management buyouts of public stockholders”, The Accounting
Review, 61(July):400-420.
DeAngelo, L.E. (1988a) “Discussion of evidence of earnings management from
the provision for bad debts”, Journal of Accounting Research,
26(supplement): 32 – 40.
DeAngelo, L.E. (1988b) “Managerial competition, information costs, and
corporate governance: the use of accounting performance in proxy
contests”, Journal of Accounting and Economics, 10 (January): 3-36.
Dechow, P. (1994) “Accounting earnings and cash flows as measures of firm
perfomance: the role of accounting accruals”, Journal of Accounting and
Economics, 18 (1): 3-42

DOCUMENTOS DE TRABAJO “NUEVAS TENDENCIAS EN DIRECCIÓN DE EMPRESAS” DT 09/01

dtecadem@eco.uva.es www2.eco.uva.es/ecadem
 21

Dechow, P., R. Sloan, and Sweeney, A. (1995) “Detecting earnings

management”, The Accounting Review, 70 (April): 193-225
Dechow, P.M., Sloan R.G and Sweeney, A.P. (1996) “Causes and
consequences of earnings manipulation: an anlysis of firms subject to
enforcement actions by the SEC”, Contemporary Accounting Research,
13(1):1-36.
Dechow, P.M., and Skinner, D.J. (2000) “Earnings managment: reconciling
view of accounting academics, practitioners, and regulators. Working
Paper, University of Michigan.
Dechow, P.M., and Sloan, R.G. (1991) “Executive incentives and horizon
problem: An empirical investigation”, Journal of Accouting and
Economics 14 (1): 51 - 89.
DeFond, M.L. and Jiambalvo J. (1993) “Factors Related to Auditor-Client
Disagreements over income-increasing accounting methods”,
Contemporary Accounting Research, Spring: 415-431.
DeFond, M.L. and Jiambalvo J. (1994) “Debt covenant violation and
manipulation of accruals”, Journal of Accounting and Economics, 17 (1-
2):145-176.
DeFond, M.L. and Jiambalvo, J. (1991) “Incidence and Circumstances of
Accounting Errors”, The Accounting Review, July: 643-655.
DeFond, M.L. and Subramanyam, K.R. (1998) “Auditor changes and
discretionary accruals”, Journal of Accounting and Economics , 25 (1):
35-67
Erikson M. and Wang, S. (1999) “Earnings management by adquiring firms in
stock for stock mergers”, Journal of Accounting and Economics, 27(2):
Friedlan J.M. (1994) “Accounting choices of issuers of Initial Public
Offerings”, Contemporary Accounting Research, 11(Summer): 1-31.
Garza- Gómez X., Okumura, M., and Kunimura, M. (1999) “Discretionary
accrual models and the accounting process”, Working Paper , Nagoya
City University.
Gaver, J.J., Gaver, K.M. and Austin, J. (1995) “Additional evidence on bonus
plans and income management”, Journal of Accounting and Economics,
19 (February): 3-28.
Guay, W.R., Kothari, S.P. and Watts, R.L. (1996) “A market-based evaluation
of discretionary accrual models”, Journal of Accounting and Economics,
34 (Supplement): 83-105

DOCUMENTOS DE TRABAJO “NUEVAS TENDENCIAS EN DIRECCIÓN DE EMPRESAS” DT 09/01

dtecadem@eco.uva.es www2.eco.uva.es/ecadem
22

Hansen, G.A. (1999) “Bias and measurement error in discretionary accrual

models”, Working Paper, Pen State University.
Healy, P. (1985) “The effect of bonus schemes on accounting decisions”,
Journal of Accounting and Economics, 7 (April): 85-107
Healy, P. (1996) “Discussion of a market based evaluation of discretionary
accrual models”, Journal of Accounting and Economics, 34 (supplement):
107-115.
Healy, P.M., and Whalen, J.M. (1999) “A review of the earnigns managemnet
literature and its implications for standard setting”, Accounting Horizons,
13 (4): 365 - 383.
Healy, P.M., and Whalen, J.M. (1999) “A review of the earnings management
literature and its implications for standard setting”, Accounting Horizons,
13 (4): 365 – 383.
Holthausen, R. W., Larcker, D. And Sloan, R. (1995) “Annual bonus schemes
and the manipulation of earnings”, Journal of Accounting and
Economics, 19 (1): 29-74
Janin, R. (2000) “Accrual – based models for detecting earnings management
and the French case”, 23rd European Accounting Association Annual
Congress.
Jeter, D.C. and Shivakumar, L. (1999), “Cross sectional estimation of abnormal
accruals using quarterly and annual data: effectiveness in detecting event
- specific earnings management”, Accounting and Business Research,
29(4): 299 – 319.
Jiambalvo, J. (1996) “Discussion of “causes and consequences of earnings
manipulation: an anlysis of firms subject to enforcement actions by the
SEC”, Comtemporary Accounting Research, 13 (1):37 –47.
Jones, J. (1991) “Earnings management during import relief investigations”,
Journal of Accounting Research, 29 (Autumn): 193-228
Kang, S. H. and Shivaramakrishnan, K. (1995) “Issues in testing earnings
management and an instrumental variable approach”, Journal of
Accounting Research, 33(2): 353-367
Kang, S.H. (1999) “A conceptual and empirical evaluation of accruals
prediction models”, Working paper, Yale School of Management.
Kothari, S.P. (2000) “Capital markets research in accounting”, Working paper,
Massachusetts Institute of Technology.
Mc Culloch, B.W. (1998) “Multi – period incentives and alternatives dials for
earnings management”, Working paper, University of Washington.

DOCUMENTOS DE TRABAJO “NUEVAS TENDENCIAS EN DIRECCIÓN DE EMPRESAS” DT 09/01

dtecadem@eco.uva.es www2.eco.uva.es/ecadem
 23

McNichols M., and Wilson, G.P. (1988) “Evidence of earnings management

from the provision for bad debts”, Journal of Accounting Research, 26
(supplement): 1 – 31.
Peasnell, K.V., Pope, P.F. and Young, S. (1998) “Detecting earnings
management using cross – sectional abnormal accruals models”,
Accounting and Business Research, 30 (4): 313 – 326.
Pfeiffer, R.J., Elgers, P.T., Lo, M.H. and Rees, L. (1999) “Additional evidence
on the incremental information content of Cash Flows and Accruals: the
impact of errors in measuring market expectations”, The accouting
Review, 74(3): 373 – 385.
Press, E.G. and Paek, W.P. (1998) “Discretionay accruals, manager´incentives
and stock prices”, Working Paper, Temple University and Sung Kyun
Kwan University.
Schipper, K. (1999) “Implications of requirement to discuss quality (not just
acceptability) of reporting on accounting research and the decision
making of standars setter, regualtors, management, audit committees, and
auditors”, Adress at the 1999 AAA/FASB Financial Reporting Issues
Conference.
Shipper, K. (1989) “Commentary on earnings management”, Accounting
horizons, 3 (December): 91 – 102.
Shivakumar, L. (1996) “Essays related to equity offerings and earnings
management” Ph. Dissertation, Vanderbilt University.
Subramanyam, K.R. (1996) “The pricing of discretionary accruals”, Journal of
Accounting and Economics,22 (1-3): 249-281.
Teoh, S.H., Welch, I. and Wong, T.J. (1998) “Earnings management and the
post issue underperformance of seasoned equity offerings”, Journal of
Financial Economics, 50 (October):63 – 99.
Teoh, S.H., Wong, T.J. and Rao, G.R. (1998) “Are accruals during initial publi
offerings opportunistic?”, Review of Accounting Studies, 3 (1-3): 175 -
208.
Watts, R. and Zimmerman, J.L. (1990) “Positive accounting theory: a ten year
perspective”, The Accounting Review, 65 (January): 131-156.
Young, S. (1999) “Systematic measurement error in the estimation of
discretionary accruals: an evaluation of alternative modelling
procedures”, Journal of Business Finance and Accounting, 26(7): 833 –
861.

DOCUMENTOS DE TRABAJO “NUEVAS TENDENCIAS EN DIRECCIÓN DE EMPRESAS” DT 09/01

dtecadem@eco.uva.es www2.eco.uva.es/ecadem
24

Zmijewski, M.E. and Hagerman, R.L. (1981) “An income strategy approach to
the positive theory of accouting standard setting choice”, Journal of
Accounting and Economics, 3 (August): 129 –149.

DOCUMENTOS DE TRABAJO “NUEVAS TENDENCIAS EN DIRECCIÓN DE EMPRESAS” DT 09/01

dtecadem@eco.uva.es www2.eco.uva.es/ecadem
 25

TABLE 1: Distribution of sample estimation firms by industry group

91 92 93 94 95 96 97 98
Cement, glass and
building material 12 10
Chain stores and
other services 7

Building 6 7

Energy and water 13

Property
companies 10 7 8 10 10
Other
transformation 19 15 14 15 18 23
industry
Metal
transformation 8 7 22
Total (241) 49 22 29 25 31 46 7 32

DOCUMENTOS DE TRABAJO “NUEVAS TENDENCIAS EN DIRECCIÓN DE EMPRESAS” DT 09/01

dtecadem@eco.uva.es www2.eco.uva.es/ecadem
26

Table 2: Ordinary Least Squares estimates for accrual models ( annual

sample estimation and pooled sample estimation)
The Jones model TA= α1 + α2 ∆ REV + α3 GPPE+ e
The Jones working capital model: WCA = α1 + α2 ∆ REV+ e
The Peasnell Pope and Young model: WCA = ω + γ1REV + γ2 CR + e

Jones adj R2 α2 α2
tα α3 α3
tα
1991 0.17 0.06 0.80 -0.09 -3.08
1992 0.48 0.18 1.86 -0.12 -2.83
1993 0.49 0.30 4.69 -0.04 -1.33
1994 -0.09 0.03 0.30 -0.01 -0.17
1995 0.26 -0.07 -0.41 -0.18 -3.30
1996 0.29 0.08 0.93 -0.02 -0.70
1997 -0.16 -0.26 -0.57 0.45 0.82
1998 0.01 0.03 0.50 -0.03 -0.96

Jones WK Adj. R2 α2 α2
tα
1991 0.34 0.22 4.13
1992 0.25 0.23 2.17
1993 0.49 0.30 4.97
1994 -0.01 0.15 1.25
1995 0.04 -0.04 -0.41
1996 0.21 -0.14 -1.43
1997 -0.09 -0.31 -0.71
1998 -0.03 0.05 0.81

Margin M. Adj. R2 γ1 tγ1 γ2 tγ2

1991 0.54 0.68 6.96 -0.68 -6.93
1992 0.68 0.89 5.97 --0.89 -5.89
1993 0.73 0.94 8.26 -0.94 -8.08
1994 0.10 0.33 1.76 -0.31 -1.62
1995 0.57 0.95 5.77 -0.96 -5.82
1996 0.84 0.92 12.72 -0.92 -12.80
1997 0.20 0.95 1.88 -1.18 -1.85
1998 0.76 0.96 9.17 -0.97 -9.44

Pooled sample estimation

Adj. R2 α2 / γ1 α2 / tγ1
tα α3/ γ2 α3/ tγ2
tα

Jones 0.14 0.01 0.27 -0.05 -3.36

Jones WK 0.07 0.07 2.04
Margin model 0.56 0.91 16.42 -0.91 -16.41

DOCUMENTOS DE TRABAJO “NUEVAS TENDENCIAS EN DIRECCIÓN DE EMPRESAS” DT 09/01

dtecadem@eco.uva.es www2.eco.uva.es/ecadem
 27

TABLE 3: Z – statistic for 27 GAAP audit qualification firms

Z- statistic Z- statistic
(annual sample (pooled sample
estimation) estimation)
Jones model 2.4729* 1.9845**
Modified Jones model 2.4564* 1.9666**
Jones Working Capital model 2.5198* 2.0988**
Modified Jones Working Capital 1.8129** 1.9942**
model
Margin Model -0.8207 -0.7124
* Statistically significant at the 0.01 level ( one - tailed)
** Statistically significant at the 0.05 level (one – tailed)

DOCUMENTOS DE TRABAJO “NUEVAS TENDENCIAS EN DIRECCIÓN DE EMPRESAS” DT 09/01

dtecadem@eco.uva.es www2.eco.uva.es/ecadem