Available online at www.sciencedirect.

com

The International Journal of Accounting 43 (2008) 246 – 267

Capitalization of R&D costs and earnings management:

Evidence from Italian listed companies☆
Garen Markarian a , Lorenzo Pozza b , Annalisa Prencipe b,⁎
a
Department of Accounting Finance and Management Control, Instituto de Empresa, Spain
b
Università L. Bocconi, Accounting Department, Via Rontgen 1-20136 Milan, Italy

Abstract

The capitalization of research and development (R&D) costs is a controversial accounting issue
because of the contention that such capitalization is motivated by incentives to manipulate earnings.
Based on a sample of Italian listed companies, this study examines whether companies' decisions to
capitalize R&D costs are affected by earnings-management motivations. Italy provides a natural
context for testing our hypothesized relationships because Italian GAAP allows for the capitalization
of R&D costs. Using a Tobit regression model to test our hypotheses, we show that companies tend
to use cost capitalization for earnings-smoothing purposes. The hypothesis that firms capitalize R&D
costs to reduce the risk of violating debt covenants is not supported.
© 2008 University of Illinois. All rights reserved.
Keywords: Earnings management; Cost capitalization; R&D accounting; Earnings smoothing; Debt covenants;
Italian companies

1. Introduction

In the current era of globalization, a highly relevant issue facing regulators, academics,
and practitioners is the determination of an appropriate accounting treatment for research

☆
We would like to thank Antonio Parbonetti for his generous provision of the governance data, Giorgio Brunello,
Antonio Marra, Dimitrios Ghicas, two anonymous referees, and conference participants at the EAA annual meeting in
Dublin, 2006, and seminar participants at the University of Padova, for their helpful suggestions. All errors are our own.
⁎ Corresponding author. Tel.: +39 02 5836 2574; fax: +39 02 5836 2561.
E-mail addresses: garen.markarian@ie.edu (G. Markarian), lpozza@unibocconi.it (L. Pozza),
annalisa.prencipe@unibocconi.it (A. Prencipe).

0020-7063/$ - see front matter © 2008 University of Illinois. All rights reserved.
doi:10.1016/j.intacc.2008.06.002
 G. Markarian et al. / The International Journal of Accounting 43 (2008) 246–267 247

and development (R&D) costs. International Accounting Standards discuss accounting for
R&D costs in IAS No. 38 “Intangible Assets” (IASB, 2004a,b). Paragraph 54 of this
standard states that no intangible asset arising from research (or from the research phase of
an internal project) shall be recognized as an asset; and that research expenses shall be
expensed in the income statement when they are incurred. Concerning development costs,
paragraph 57 states that an intangible asset arising from development (or from the
development phase of an internal project) shall be recognized if, and only if, an entity can
demonstrate all of the following: (a) the technical feasibility of completing the intangible
asset so that it will be available for use or sale; (b) its intention to complete the intangible
asset and use or sell it; (c) its ability to use or sell the intangible asset; (d) how the intangible
asset will generate probable future economic benefits; (e) the availability of adequate
technical, financial, and other resources to complete the development and to use or sell the
intangible asset; and (f) its ability to measure reliably the expenditure attributable to the
intangible asset during its development. Although IAS 38 allows companies to capitalize
development costs, the inherent subjectivity of the validation process permits management
to exercise discretion in deciding whether the conditions of IAS 38 have been satisfied. In
essence, IAS 38 gives management considerable flexibility regarding the treatment of
development costs.
US GAAP takes a stricter approach to the issue. SFAS No. 2—Accounting for Research
and Development Costs (FASB, 1974)—requires that all R&D expenditures be expensed in
the current period. The only exception to the full expensing rule is stated in SFAS No. 86.
The exception relates to the capitalization of software development costs (FASB, 1985). At
the international level, certain national accounting standards (e.g., those of Italy) allow
flexibility for the capitalization of R&D costs when some conditions are satisfied. These are
conditions similar to those required by IAS.
The capitalization of R&D costs has always been a controversial accounting issue.
Supporters of capitalization report results suggesting that R&D is a long-lived asset that
influences future profitability (e.g., Bublitz & Ettredge, 1989; Sougiannis, 1994; Ballester,
Garcia-Ayuso, & Livnat, 2003). Also, R&D costs are positively related to market value
(Hirschey & Weygandt, 1985; Shevlin, 1991; Sougiannis, 1994) and yield value-relevant
information to investors (e.g., Aboody & Lev, 1998; Lev & Zarowin, 1999; Healy, Myers,
& Howe, 2002; Monahan, 2005).
Supporters of expensing are fewer. They stress the lack of reliable evidence of future
economic benefits (e.g., FASB, 1974; AIMR, 1993; Kothari, Laguerre, & Leone, 2002) or
refer to the benefits of consistency and comparability, pointing out that such benefits trump
the costs identified by the supporters of capitalization. Additionally, reliability and the risk
of earnings-management policies are underscored by supporters of the most conservative
accounting treatment. In particular, expensing is preferable to capitalization because it
increases the objectivity of financial statements. That is, it eliminates the opportunity for
managers to capitalize costs of projects that have low probabilities of success or to delay
impairment of R&D assets (Nelson, Elliott, Tarpley, & Tarpley, 2003; Schilit, 2002).
The debate surrounding the most effective accounting method for R&D costs supplements
other literature that examines the trade-off between relevance (i.e., the predictive ability) and
reliability (i.e., the representative faithfulness) of accounting information (FASB, 1980;
AICPA, 1994; IASB, 2004a,b). Thus far, empirical research on R&D costs has focused mainly
248 G. Markarian et al. / The International Journal of Accounting 43 (2008) 246–267

on the relevance side of the trade-off, while little has been written about the reliability side that
is, the possibility that R&D costs are subject to earnings management.
However, a few studies have indeed shown that R&D expenditures are subject to real
earnings management. In short, this means that companies cut their R&D investments in
order to achieve their earnings goals (e.g., Perry & Grinaker, 1994; Bushee, 1998; Mande,
File, & Kwak, 2000). But there is still a paucity of research that explores the motives behind
the accounting treatment of R&D costs within a setting where flexibility is allowed. Testing
whether companies engage in earnings management through R&D cost accounting can
significantly contribute to the debate around the best treatment for such costs. This debate
has recently been raised within the convergence project by US GAAP and IAS/IFRS.
Illustrating that R&D cost capitalization is motivated by incentives to manipulate earnings
would support the current U.S. GAAP position, which does not allow the capitalization of
such costs. On the contrary, showing that companies do not use R&D cost accounting for
earnings-management purposes would support the approach now stated by IAS/IFRS, in
which capitalization is allowed under certain conditions.
This study contributes to this debate by providing empirical evidence on the motivations
for R&D cost capitalization. We hypothesize that the decision to capitalize R&D
expenditures is related to two primary motivations: income smoothing and debt contracting.
We test our hypotheses using a sample of firms listed on the Milan Stock Exchange.
Multivariate results indicate that firms use capitalization of R&D costs to smooth earnings,
while there is no support for the debt-covenant hypothesis. These results are robust within a
variety of firm characteristics, such as firm size, risk, opportunities for growth, profitability,
governance characteristics, industrial membership, and time control.
The paper proceeds as follows. Section 2 introduces accounting in Italy and the
institutional background relating to R&D accounting. Section 3 discusses the previous
literature. Section 4 presents the hypotheses and is followed by the research methods in
Section 5. Section 6 presents the results and Section 7 concludes the study.

2. R&D accounting in Italy

Italian accounting regulation has always allowed for some flexibility in the capitalization
of R&D costs. This allowance is similar to that of IAS. Accounting for intangibles, including
R&D costs, is regulated by Principio Contabile n. 24 (Accounting Standard No. 24). This
standard distinguishes three different types of R&D costs as follows:

1) “Basic research,” which consists of studies, surveys, and experiments that do not refer to
a specific project; this type of R&D cost is normally carried out for the general utility of
a company (e.g., market research, updating, etc.);
2) “Applied research,” which consists of studies, surveys, and experiments that refer to
specific projects; and
3) “Development,” which consists of the application of research results to specific
materials, tools, products, and processes preceding production.

The costs for basic research are to be expensed in the income statement. However, costs
related to applied R&D can be capitalized if the following conditions are met: a) the costs
 G. Markarian et al. / The International Journal of Accounting 43 (2008) 246–267 249

refer to a project for the realization of a clearly defined product or process; b) the costs are
identifiable and measurable; c) the project to which the costs refer is technically feasible; d)
the company owns the necessary resources to complete and exploit the project; and e) the
costs are recoverable through the revenues generated by exploiting the project.
It is evident that the conditions stated by the Italian accounting standards are similar to
those stated by IAS for development costs. In fact, the definition of applied research under
Italian standards also fits into the definition of development costs provided by IAS 38. The
Italian standards differ from IAS in that they do not require R&D capitalization when the
abovementioned conditions occur, leaving flexibility in the hands of the companies.
However, this difference is more formal than substantive. Given the subjectivity in
assessing the occurrence of some of the conditions, it seems that, even under IAS,
companies that prefer immediate expensing can easily justify this approach—even when
the aforementioned conditions are met.
Concerning the amortization of R&D costs, the Italian accounting standards require that
the amortization be carried out over a period of no longer than five years beginning from the
moment the outcome (product or process) is ready to be used. The Italian Civil Code (art.
2426) states that the capitalization of R&D costs shall be authorized by the collegio
sindacale (statutory auditors) and that it is not possible to pay dividends until there are
enough retained earnings to cover the carrying amount of the capitalized R&D costs. This
stipulation limits the incentive to capitalize R&D costs for the purpose of increasing the
amount of dividends paid. The Civil Code also requires that R&D activities be discussed in
the relazione sulla gestione (management discussion and analysis section); however, there
is no clear requirement as to what quantitative or qualitative disclosures should be relayed
with regard to the capitalization of R&D costs. Finally, the Civil Code states that
information regarding the amortization schedules of such R&D costs be provided in the
explanatory notes of the financial statements.

3. Earnings management and specific accruals

Earnings management is defined as a “purposeful intervention in the external financial-

reporting process, with the intent of obtaining some private gain” (Schipper, 1989, p. 92). In
generally accepted terms, earnings management occurs “when managers use judgment 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 numbers” (Healy &
Wahlen, 1999, p. 368).
The large amount of research carried out thus far indicates that managers exercise
discretion and manage earnings using a wide variety of methods, ranging from carrying out
special transactions (so-called real earnings management) to the manipulation of accruals.
Several of the main incentives for earnings management include debt covenants, bonus
plans, and income smoothing. The debt-covenant hypothesis suggests that managers have
an incentive to manage earnings in order to avoid violating covenants in debt contracts,
which are typically stated in terms of accounting numbers or ratios. The bonus-plan
hypothesis suggests that managers manage earnings in order to maximize compensation.
Healy (1985) shows that managers tend to reduce earnings if they fall either above or below
250 G. Markarian et al. / The International Journal of Accounting 43 (2008) 246–267

bonus-plan bounds. In contrast, they tend to increase earnings when they fall between the
two bounds. Finally, the income-smoothing hypothesis suggests that firms aspire to reduce
earnings fluctuations.
Empirical earnings-management studies find support for the abovementioned motives in
a variety of contexts. Many of these studies test the relationship between aggregate accruals
and incentives for earnings management (e.g., Healy, 1985; DeAngelo, 1986; Jones, 1991;
Dechow, Sloan, & Sweeney, 1995). As an alternative approach, other studies focus on
single items, suggesting that income from specific accruals is related in a systematic way to
earnings-management incentives. Among these latter studies, McNichols and Wilson
(1988) show that companies manage their bad-debt provisions according to the bonus-plan
hypothesis (Healy, 1985). Zucca and Campbell (1992) examine discretionary asset write-
downs, showing that companies use these accruals either for “big bath” strategies or for
earnings smoothing. Francis, Hanna, and Vincent (1996) confirm that earnings-manage-
ment incentives play a significant role in explaining goodwill write-offs and restructuring
charges. Other studies focus on allowances for deferred taxes (e.g., Miller & Skinner, 1998;
Schrand & Wong, 2003). These studies provide mixed results. Finally, Dowdell and Press
(2004) analyze the in-process R&D write-offs, but they do not find evidence to support their
bonus-plan hypothesis.
In line with the aforementioned studies on earnings management and specific accruals,
this study aims at testing whether the decision to capitalize or to expense R&D costs (when
flexibility exists) is affected by earnings-management motives.

4. Hypotheses development

Previous research investigates three main incentives for earnings management: earnings
smoothing, debt covenant, and bonus-plan incentives. In this study, we focus on the first
two since disclosure of data on the existence and structure of bonus plans by Italian
companies is limited.1
The income-smoothing hypothesis suggests that a manager's accounting discretion is
driven by his or her desire to reduce income-stream variability (Fudenberg & Tirole, 1995).
The process of smoothing serves to moderate year-to-year fluctuations in income by
shifting earnings from peak years to less successful periods. This process lowers the peaks
and makes earnings fluctuations less volatile (Copeland, 1968).
Income smoothing has been viewed both as a positive strategy, whereby managers
transmit private information to investors (e.g., Gordon, 1964; Beidleman, 1973; Ronen &
Sadan, 1981; Tucker & Zarowin, 2006), and as a manipulative practice driven by
opportunistic aims (Gordon, 1964; Imhoff, 1977; Kamin & Ronen, 1978). In this study, we
do not intend to argue for either one of these two views. Rather, we test whether R&D cost
capitalization is used for purposes of earnings smoothing.
Several explanations for earnings smoothing have been posited by prior studies. Some of
these explanations are related to capital-market incentives. Several authors suggest that
since current earnings are used as predictors for future earnings, managers have an incentive

1
Under the Italian regulation, companies have to provide only very general and aggregate figures regarding
executive compensation—hence our inability to examine this potentially interesting incentive.
 G. Markarian et al. / The International Journal of Accounting 43 (2008) 246–267 251

to smooth income as a signalling technique in order to deliver private information to the

market (Ronen & Sadan, 1981; Chaney & Lewis, 1995; Hunt, Moyer, & Shevlin, 1996).
Other authors stress that a primary motivation for earnings smoothing is to reduce perceived
risk by investors. Earnings variability is interpreted as an important measure of the overall
risk of a firm and has a direct effect on investors' capitalization rates—and thus has an
adverse effect on the value of a firm's shares (e.g., Wang & Williams, 1994; Barth,
Landsman, & Wahlen, 1995; Gebhardt, Lee, & Swaminathan, 2001).
Other explanations for income smoothing have been related to the existence of contracts
linked to accounting numbers. Trueman and Titman (1988) suggest that income smoothing
practices reduce debtholders' perceptions of a firm's risk of bankruptcy, and thereby
translate into a lower cost of capital for the firm. Political costs are also considered to be an
important motivation for earnings smoothing (Watts & Zimmerman, 1986; Wong, 1988;
Cahan, 1992; Godfrey & Jones, 1999). Income smoothing averts attention from “excessive”
high earnings that might attract adverse political attention. It also avoids concerns which
arise from low operating profits. Such concerns could come from employees concerned
about future employment prospects, from suppliers and customers assessing future stability,
or from governments that are investigating unviable industries.
Another primary explanation for earnings smoothing has been related to managers'
compensation and concerns over job security. Fudenberg and Tirole (1995) predict that
income smoothing occurs because the manager boosts the reported income in bad times in
order to raise the probability of keeping his job and tends to decrease reported income in
good times since “information decay” causes good current performance to be weighted less
in future performance. This expectation partially contradicts Healy's (1985) hypothesis on
bonus plans, according to which managers tend to decrease earnings (i.e., “take a big bath”)
when they are below the lower bound stated by the plans. However, more recent empirical
studies have contradicted the validity of Healy's hypothesis (Gaver, Gaver, & Austin, 1995;
Holthausen, Larcker, & Sloan, 1995; Dowdell & Press, 2004),2 suggesting that the
existence of bonus plans does not significantly affect the earnings-smoothing incentive.
The empirical evidence provided so far generally supports the existence of earnings
smoothing (e.g., Ronen & Sadan, 1981; Moses, 1987; Young, 1998; Chaney, Jeter, &
Lewis, 1998; Buckmaster, 2001). For empirical purposes, the income-smoothing
hypothesis has usually been characterized as the propensity of managers to choose
accounting policies that increase (decrease) reported earnings when current period pre-
managed earnings are below (above) the profitability measure of prior years. Consistent
with earlier studies (Barnea, Ronen, & Sadan, 1976; Moses, 1987; Godfrey & Jones, 1999),
we use operating profitability as the income-smoothing target.
While there is general support for the hypothesis that companies use accruals to smooth
earnings, there is no clear evidence about the use of R&D cost capitalization for such
purpose. However, a survey by Nelson et al. (2003) shows that cost capitalization is one of
the most common earnings-management strategies. Therefore, we expect companies tend
toward the capitalization of more R&D costs when their operating profitability (before

2
For possible explanations regarding the lack of empirical support of the bonus-plan hypothesis, see Dechow
and Schrand (2004).
252 G. Markarian et al. / The International Journal of Accounting 43 (2008) 246–267

R&D capitalization)3 is below the level of the past years, while they tend toward expensing
R&D costs when their operating profitability is above the level of that of past years. We
formulate the following hypothesis:

H1. There is a negative relationship between change in firm profitability and R&D capitalization.

The second hypothesis is based on implications of the literature regarding agency theory
(Jensen & Meckling, 1976). This literature suggests that contracts between debtholders and
owner-managers contain covenants restricting management behavior in order to prevent actions
from being taken that may negatively affect the debtholders' wealth position. Many such debt
agreements rely on the accounting numbers. To the extent that the accounting standards allow
flexibility in the choice of accounting methods, managers have an opportunity to choose those
procedures that allow them to avoid violating restrictive debt covenants—specifically methods
that increase income. Even if default cannot be avoided by manipulation of accounting
information, managers are still expected to make asset- and income-increasing accounting
choices in the hope of improving their bargaining position in the event of renegotiation (DeFond
& Jiambalvo, 1994). Extant empirical evidence on the debt-covenant hypothesis is mixed (e.g.,
DeAngelo, DeAngelo, & Skinner, 1994; Sweeney, 1994; DeFond & Jiambalvo, 1994).
The most frequently used proxy to test the debt-covenant hypothesis is the leverage ratio
(e.g., Hunt, 1985; El Gazzar, Lilien, & Pastena, 1986; Watts & Zimmerman, 1986;
Rusbarsky, 1988). A number of studies are specifically dedicated to test whether the
leverage ratio is able to reliably proxy for debt-covenants tightness (Duke & Hunt, 1990;
Press & Weintrop, 1990). The majority of these studies support the use of the leverage ratio.
The debt-covenant hypothesis suggests that managers of more leveraged companies have
stronger incentives to make income-increasing accounting decisions, e.g.—in our case—to
capitalize more R&D costs. Such capitalization “loosens” debt-covenant restrictions in two
ways. First, if debt covenants are based on measures of profitability, then capitalization will
increase earnings. Second, if the debt covenant is based on the ratio of total debt to assets,
then capitalization will lower this ratio. Of course, this is true under the assumption that
lenders do not adjust a firm's earnings and total assets by excluding the effects of the
capitalization (Duke & Hunt, 1990). Thus, our second hypothesis is formulated as follows:

H2. There is a positive relationship between a company's financial leverage and its R&D
capitalization.

5. Research methodology

5.1. Sample selection

Under Italian disclosure rules, all firms that undertake R&D activities during a fiscal
year have an obligation to report such activities in the management's discussion and

3
By “before R&D capitalization,” we mean before the effects of the capitalized R&D costs, net of the related amortization.
 G. Markarian et al. / The International Journal of Accounting 43 (2008) 246–267 253

analysis section (MD&A) of their annual reports. Therefore, we searched the annual reports
of all non-financial firms listed on the Borsa Valori di Milano (Milan Stock Exchange) that
disclosed R&D activities for the years 2001, 2002, and 2003. We hand-collected and
analyzed the original financial statements since electronic databases did not report all the
information required for our analysis.4 Unfortunately, even using the original documents,
we were unable to collect the complete set of information needed. This was because many
companies did not specify the amount of total R&D expenditures or R&D capitalization, or
they provided only aggregate data for R&D and advertising expenditures.5 We found a total
of 130 firm-years that disclosed data needed for our statistical analysis across the sample
period (43, 43, and 44 firms for the years 2001, 2002 and 2003 respectively). In our
regression model (discussed in section 6.2), we require lagged data to be available for each
firm. Hence, in our statistical tests, we use 86 observations related to 43 unique firms.

5.2. Variables

For each firm in our sample, we calculated a number of variables to be used in the
statistical tests. Our variable of interest is the R&D capitalization variable, Capitalization,
which is calculated as total R&D capitalization (net of the yearly depreciation) divided by
the total assets of the firm.6 To test our first hypothesis, we use ChROA, which is the change
in return on assets over the average of the previous two fiscal years. Later, we also report
results on a simple year-to-year change in ROA.7 What we test is whether the accounting
treatment of R&D is used—in conjunction with other techniques—to reach earnings goals.
To test our second hypothesis (i.e., the relationship between firm debt and
Capitalization), we calculated Leverage, which is a firm's total debt divided by total
assets (calculated before the effect of R&D capitalization). Additionally, in order to
examine the effects of earnings-management incentives on capitalized R&D costs, we
included in our statistical tests a number of control variables that had previously been linked
to R&D capitalization (see Aboody & Lev, 1998). Our first control variable is the total
amount of R&D expenditures undertaken by the firm divided by total assets in the current
fiscal year, R&Dtotal. It is reasonable to expect a relationship between the amount of total
4
We could not collect data prior to 2001 since full annual reports for most Italian companies are available only
for the last five years.
5
According to Italian law, capitalized R&D costs and advertising costs are included on the same item in the
legal balance sheet format.
6
The choice of total assets as a deflator is due to a number of reasons. First, it provides for a direct relationship to
firm leverage (a variable of interest in this study); it also is in line with other earnings-management studies, as the
original Jones (1991) study deflates accruals by total assets. A possible alternative could be deflating by income in
order to provide a clearer relationship with firm earnings. However, this tactic would require the exclusion of firm-
years with negative earnings, which could bias the results. Finally, to deflate by total R&D expenditures is not
theoretically desirable, since a firm that capitalizes a large (small) percentage of its R&D might lead to no significant
(a large) impact on earnings. An example of this is large, non R&D intensive firms capitalizing most of their R&D
expenditures. In this case, the percentage of R&D capitalized would be high while impact on earnings would be
small. Later, we include the total expenditures in R&D as a control variable in our model.
7
In the calculation of ChROA, current ROA is assumed “pre-managed” (that is to say, the ROA before the
effects of R&D capitalization). We are aware that companies use several earnings management mechanisms (not
solely R&D cost capitalization) in order to reach their goals. Therefore, we do not assume that our “pre-managed”
ROA is unaffected by other earnings-management policies.
254 G. Markarian et al. / The International Journal of Accounting 43 (2008) 246–267

R&D investments undertaken by the firm and the amount of capitalization. On one side, the
higher a firm's total R&D investments, the higher the probability that a larger portion of the
undertaken R&D meets the conditions stated by the standards. But it should be pointed out
that the standards allow flexibility in this decision—even when all the above-mentioned
conditions are present. Moreover, for companies undertaking a large number of R&D projects
(which companies are typically those that have a higher R&Dtotal), checking whether the
conditions required by the standards are met is more difficult than for companies undertaking
only a few small projects. Therefore, we included R&Dtotal as a control variable—without
having any ex ante prediction concerning its relationship to Capitalization.
Our second control variable is related to the profitability of the firm. Current profitability
can be considered an indirect control for the existence of the economic resources necessary
to complete and exploit the project. It should be pointed out, however, that the relationship
between current profitability and the extent of the capitalization of R&D costs can be at
least partially affected by earnings-management purposes. Companies can be motivated to
capitalize R&D costs when their current profitability is either negative or low in order to
improve their accounting performance. Similarly, companies with high profitability can be
motivated to expense R&D costs in order to reduce political costs. These two effects
suggest a negative relationship between current profitability and the level of R&D cost
capitalization. On the other hand, current profitability can also be considered a proxy for
expected future profitability (see McNichols & Wilson, 1988; Francis et al., 1996; Miller &
Skinner, 1998). Expected future-operating profitability can be considered a strong rationale
for the company to successfully recover its capitalized costs. We expect that the higher the
future profitability, the higher the probability that the last condition required by the
accounting standards—that companies can recoup the capitalization through expected
increases in revenues—will be met. Therefore, we expect a positive relationship between
future-operating profitability and the extent of capitalized R&D costs. In order to control for
all these effects, we introduced ROA as a control variable of a firm's profitability, measured
as the firm's operating income (before R&D capitalization) divided by the firm's total assets.
We use firm size as a control for firm visibility, political costs and media attention (Watts
& Zimmerman, 1986; Bhushan, 1994). We proxy for firm size by calculating the natural
logarithmic form of the firm's total assets, Logassets. We also control for the materiality of
the capitalization by calculating High-Capitalizer: a dummy variable equal to one if the
amount capitalized (normalized by the absolute value of earnings) is above the median for
capitalizing firms. Moreover, we control for the amount of capitalized R&D during the prior
year by calculating LagCapitalization,8 which is equal to Capitalization of the prior year.9
Not controlling for the prior year's capitalization represents a serious, correlated, omitted
variable.10 Therefore, in our statistical analysis, we eliminated all firms that do not have
capitalization data for the prior year, reducing our final sample size to 86. We also
calculated MB, which is the market value of equity divided by its book value, as a control
for future growth opportunities. Beta is included in our multivariate analysis as a control for

8
The Durbin Watson statistic and the calculated first-order autocorrelation indicate that we do not have
problems of serial correlation between Capitalization and LagCapitalization.
9
Results remain unchanged if we use a 1/0 dummy variable.
10
For a similar treatment, see Boone, Field, Karpoff, and Raheja (2007).
 G. Markarian et al. / The International Journal of Accounting 43 (2008) 246–267 255

Table 1
Comparison of selected characteristics of sample firms to all firms available through Datastream for the year 2003
Sample firms All firms t-test Wilcoxon
# Observations 42 182
Variable Mean Mean Difference Pr N t Pr N z
ROA 0.046 0.018 0.028 0.13 0.21
ChROA − 0.060 0.008 0.068 0.45 0.68
Assets 913 million 364 million 549 million 0.09 0.02
Leverage 0.624 0.631 0.007 0.82 0.84
ROA is measured as the firm's operating income divided by the total assets of the firm. ChROA is the change in
return on assets over the average of the prior two fiscal years. Assets is the total assets of the firm at fiscal year-end.
Leverage is total debt divided by total assets.

firm risk. Finally, all multivariate analyses include controls for years and for industrial
membership based on the Milan Stock Exchange classification.

6. Data analysis and results

6.1. Sample statistics, descriptives, and univariate analysis

In Table 1, we compare our sample firms to the population of non-financial firms listed
on the Milan Stock Exchange to see whether our sample firms are biased on certain
dimensions.
We compare our sample firms for the year 2003 with all firms whose data are available
through Datastream.11 We see that both sample firms and control firms are of about equal
leverage, profitability, and change in profitability. The only difference is that our sample
firms are larger in size. This difference in size of sample firms is an expected occurrence, as
we are studying R&D issues. Since size is the only firm characteristic that is different
between our sample firms and the population of firms listed on the Milan Stock Exchange,
and since we specifically control for firm size in our statistical tests, selection bias does not
appear to be a concern. Also, in our tests, we control for a variety of additional firm and
institutional characteristics (e.g., governance and ownership structure, firm risk and growth
opportunities, etc.).
Table 2 presents the descriptive statistics for our sample of 86 firms. Mean Capitalization
is 0.19%, with 0.69% being capitalized at the final decile (ninetieth percentile). In general,
firms spend an equivalent of 2.1% of their total assets on R&D, with the largest decile being
6.2% for the most research-intensive firms. Comparing Capitalization to R&Dtotal, we find
that firms capitalize roughly 10% of their R&D expenditures. In general, the firms are
profitable, with an average ROA of about 4.4% and with a decrease in their profitability
from the average of the prior two years to about −4.7%. The firms are levered at about 50%,

11
We make comparisons for 2003 only, since including 2001 and 2002 would bias the comparisons by
examining data across multiple years. Using Datastream data for both the sample and the other companies avoids
the risk of inconsistent calculations between the two samples. In relation to year 2003, Datastream reports data on
42 out of 43 sample firms and on a total of 182 firms listed on the Milan Stock Exchange.
256 G. Markarian et al. / The International Journal of Accounting 43 (2008) 246–267

Table 2
Descriptive statistics on selected variables
Variable Mean Median SD P10 P90
Capitalization 0.002 0.000 0.004 0.000 0.007
R&Dtotal 0.021 0.013 0.023 0.001 0.062
ROA 0.044 0.045 0.076 − 0.018 0.120
ChROA − 0.047 −0.046 0.061 − 0.120 0.004
Leverage 0.512 0.559 0.157 0.281 0.714
Assets 1.33 billion 513 million 2.68 billion 62 million 67 billion
P10 (P90) signifies the tenth (ninetieth) percentile of the variable distribution. Capitalization is total R&D
capitalization (net of the yearly depreciation) divided by the total assets of the firm. R&Dtotal is the total
investment in R&D undertaken by the firm divided by total assets. ROA is measured as the firm's operating income
(before capitalization) divided by the total assets of the firm. ChROA is the change in return on assets (before
capitalization) over the average of the prior two fiscal years. Leverage is total debt divided by total assets (before
capitalization). Assets is the total assets of the firm at fiscal year-end.

indicating that debt financing is a prime source of funds. As concerns their size, the firms
are relatively large, with about 1.3 billion euros in assets.
Since we examine R&D capitalization as a means of earnings management, we provide
descriptive statistics disaggregated by industry in Table 3. Since our sample-selection
procedure involves gathering data on all firms that report R&D activities, our sample
contains a broad representation spread across many industries. The majority of R&D
activity is conducted by new economy, electronics, and automotive firms. A smaller
amount of R&D transpires in utilities firms. In terms of capitalized R&D, the plants &
machinery and automotive industries have larger capitalization rates as compared to the
textile industry.12
In Table 4 we conduct both parametric t-tests and non-parametric Wilcoxon rank-sum
tests in order to examine differences between capitalizing and non-capitalizing firms. We
see that at the univariate level, without controlling for multivariate effects, capitalizing and
non-capitalizing firms have similar levels of change in profitability and firm size while they
have different levels of leverage. In particular, capitalizing firms are more highly levered.
R&Dtotal is larger in the non-capitalizing group (statistically significant at 0.01 and 0.10 in
the t-test and Wilcoxon test, respectively). This is an interesting observation because it
indicates that capitalization is not a simple function of a firm's R&D activities. This result,
combined with results in Table 3, suggests that alternate underlying motivations could be
involved.
Table 5 presents the table of correlations (Pearson). Here, we see that ChROA is negatively
related to Capitalization, even though the level of significance is not particularly high
(P b 0.10), while Leverage shows no significant correlation. Capitalization is also not related to
R&Dtotal, and is only marginally and negatively related to Assets. The only variable showing
a significant correlation is ROA, which is negatively related to Capitalization (P b 0.01). In

12
Inspecting Table 3 is interesting because there are two main patterns that can be observed. Examining means
and medians, we find a large variability between industries in terms of both the R&D undertaken and the ensuing
capitalization. This finding provides confidence that our ensuing statistical analysis is not driven by industrial
patterns.
 G. Markarian et al. / The International Journal of Accounting 43 (2008) 246–267 257

Table 3
Descriptive statistics on R&D capitalization variables, by industry
Industry N Variable Mean Median SD
Automotive 8 R&Dtotal 0.022 0.022 0.005
Capitalization 0.007 0.004 0.010
Chemicals 14 R&Dtotal 0.015 0.018 0.013
Capitalization 0.000 0.000 0.000
Electronics 14 R&Dtotal 0.024 0.019 0.023
Capitalization 0.001 0.000 0.002
Media 6 R&Dtotal 0.003 0.004 0.001
Capitalization 0.000 0.000 0.001
Metals and oil 2 R&Dtotal 0.015 0.014 –
Capitalization 0.006 0.007 –
New Economy 14 R&Dtotal 0.048 0.059 0.034
Capitalization 0.003 0.000 0.006
Plants and machinery 8 R&Dtotal 0.003 0.001 0.004
Capitalization 0.020 0.012 0.021
Textile 4 R&Dtotal 0.008 0.011 0.007
Capitalization 0.000 0.000 0.000
Utilities 8 R&Dtotal 0.002 0.002 0.001
Capitalization 0.000 0.000 0.000
Miscellaneous 8 R&Dtotal 0.007 0.004 0.010
Capitalization 0.011 0.006 0.010
Capitalization is total R&D capitalization (net of the yearly depreciation) divided by the total assets of the firm.
R&Dtotal is the total investment in R&D undertaken by the firm divided by total assets.

sum, we see that at the univariate level, there is some (even if not overtly clear) support for H1
and H2, as Capitalization is negatively related to ChROA, while Leverage is higher for
capitalizing firms in the Wilcoxon and t-tests. However, these results should be interpreted
with caution as at the univariate level we do not control for cross-correlations (e.g., size is
significantly related to three of our variables), and we do not control for industry membership,
etc. Therefore, we next turn our attention to the more formal multivariate regressions.

Table 4
Univariate small-sample tests examining differences among capitalizing and non-capitalizing firms
Variable Non-capitalizing firms Capitalizing firms Difference t-test Wilcoxon
Mean Mean Pr N t Pr N z
R&Dtotal 0.027 0.014 0.013 0.01 0.07
ROA 0.052 0.035 0.017 0.28 0.45
ChROA − 0.052 −0.041 − 0.011 0.30 0.25
Assets 1.11 billion 1.17 billion 59 million 0.80 0.60
Leverage 0.469 0.573 0.104 0.01 0.01
# of obs. 44 42
R&Dtotal is the total investment in R&D undertaken by the firm divided by total assets. ROA is measured as the
firm's operating income before capitalization divided by the total assets of the firm. ChROA is the change in return
on assets over the average of the two prior fiscal years. Leverage is total debt divided by total assets before
capitalization. Assets is the total assets of the firm at fiscal year-end.
258 G. Markarian et al. / The International Journal of Accounting 43 (2008) 246–267

Table 5
Pearson correlations among selected variables
Capitalization R&Dtotal ROA ChROA Leverage
R&Dtotal 0.054 (0.61)
ROA −0.362 (b0.01) −0.291 (b0.01)
ChROA −0.178 (0.09) 0.103 (0.33) −0.786 (b0.01)
Leverage 0.069 (0.50) − 0.119 (0.36) 0.008 (0.90) 0.072 (0.46)
Assets −0.194 (0.06) −0.323 (b0.01) 0.276 (b0.01) − 0.151 (0.32) 0.243 (b0.05)
Correlation coefficient is reported in each cell, significance level is in the parenthesis.
Capitalization is total R&D capitalization (net of the yearly depreciation) divided by the total assets of the firm.
R&Dtotal is the total investment in R&D undertaken by the firm divided by total assets. ROA is measured as the
firm's operating income before capitalization divided by the total assets of the firm. ChROA is the change in return
on assets over the average of the prior two fiscal years. Leverage is total debt divided by total assets (before
capitalization). Assets is the total assets of the firm at fiscal year-end.

6.2. The regression model

Since Capitalization is a left-truncated variable—i.e. companies that don't capitalize

R&D have Capitalization = 0, while companies that capitalize have a positive value of less
than one—we use a Tobit regression for our analysis (see Kennedy, 2003).13 The statistical
equation has the following general form (the subscript i denotes each firm in our sub-
sample, as does t for multiple time periods):

Capitalization i ¼ b0 þ b1 ChROAi þ b2 Leveragei þ b3 R&Dtotal i þ b4 ROAi

þ b5 Logassetsi þ b6 M Bi þ b7 Beta i þ b8 HighCapitalizer i
þ b9 LagCapitalization i þ Industry Controls
þ Year Dummies þ uit
Per H1 and H2, we expect the coefficients on b1 and b2 to be negative and positive, respectively.
Model 1 of Table 6 presents the tests on H1 and H2. We see that, as predicted by H1,
ChROA is negative and significant, indicating that the higher (lower) the profitability this
year compared to the average of the previous two years, the smaller (larger) the amount of
capitalized R&D expenditures. For H2, we see that, inconsistent with our predictions,
Leverage is not significant. A possible interpretation of this result is that debt covenants are
structured so that companies are not motivated to capitalize costs in order to avoid
covenants violation. Actually, debt covenants are often designed so that reported earnings
and total assets are adjusted to eliminate the effects of particular accounting treatments,
such as cost capitalization (see Duke & Hunt, 1990).
Regarding our control variables, we find that R&Dtotal is insignificant (P b 0.96),
indicating that the amount of capitalized R&D is not a function of a firm's R&D
expenditures. This insignificant relationship between R&D expenditures and capitalization
indicates that a firm's decision to capitalize is independent of its R&D functions; rather, it
could signify that the decision to capitalize is related to other firm characteristics.
Concerning ROA, we find a significant negative relationship with Capitalization, indicating

13
Using OLS regression instead of the Tobit model does not qualitatively affect our inferences.
 G. Markarian et al. / The International Journal of Accounting 43 (2008) 246–267 259

Table 6
Tobit regression analysis examining the relationship between R&D capitalization and hypothesized variables
Dependent variable: Capitalization
Parameter Model 1 Model 2
Estimate Pr N |t| Estimate Pr N |t|
Intercept 0.004 0.42 0.005 0.30
ChROA − 0.021 0.01
ChROA-2 −0.023 0.01
Leverage − 0.002 0.27 −0.002 0.26
R&Dtotal 0.001 0.96 0.003 0.84
ROA − 0.027 0.01 −0.010 0.065
Logassets − 0.000 0.37 −0.000 0.32
MB 0.001 0.04 0.000 0.12
Beta 0.001 0.22 0.001 0.17
High-Capitalizer 0.007 0.01 0.008 0.01
LagCapitalization 0.180 0.01 0.190 0.01
N 86 86
Schwartz Criterion − 707 − 706
Pseudo R-squared 0.73 0.67
Capitalization is total R&D capitalization (net of the yearly depreciation) divided by the total assets of the firm.
ChROA is the change in return on assets (before capitalization) over the average of the prior two fiscal years.
ChROA-2 is the change in return on assets (before capitalization) over the previous fiscal year. Leverage is total
debt divided by total assets (before capitalization). R&Dtotal is the total investment in R&D undertaken by the firm
divided by total assets. ROA is measured as the firm's operating income before capitalization divided by the total
assets of the firm. Logassets is the logarithmic transformation of the total assets of the firm at fiscal year-end. MB is
the market value of equity divided by the book value. Beta is the sensitivity of the asset's returns to market returns.
High-Capitalizer is a dummy variable equal to one if the amount capitalized (normalized by the absolute value of
earnings) is above the median for capitalizing firms. LagCapitalization is equal to Capitalization of the prior year.
Dummy variables for year and industry membership are not reported.

that more profitable (unprofitable) firms capitalize less (more) of their R&D expenditures.
If we considered ROA as an indicator of the financial health of the company and of the
availability of economic resources necessary to complete and exploit the project, or
alternatively as an indicator of future expected profitability, we would expect a positive
relationship. Instead, it seems that R&D cost capitalization is used to counterbalance high
or low earnings, which indirectly confirms that such capitalization is used for earnings-
smoothing purposes, as stated by H1. An alternative possible explanation is that companies
tend to reduce political costs arising from too high or too low profitability, or that more
profitable companies tend to invest in higher risk development projects, which typically do
not meet one or more of the conditions required for the capitalization, while less profitable
companies tend to invest in projects whose result is more predictable and less risky and,
therefore, more likely to meet the conditions stated by the standards. In relation to the other
control variables, we note that firm size is not related to R&D capitalization and that
opportunities for firm growth are positively related to R&D capitalization, indicating that
firms that are expected to grow are more likely to capitalize—perhaps because they face
more pressure to provide a positive outlook on firm performance or because growth firms are
more likely to meet the conditions of capitalization. High-Capitalizer, as expected, is
260 G. Markarian et al. / The International Journal of Accounting 43 (2008) 246–267

positively related to Capitalization, possibly due to a mechanical effect or because the higher
the effect of capitalization on earnings, the more likely that a capitalization decision will be
made. LagCapitalization is also positively related to Capitalization, since firms are expected
to take capitalization decisions with uniformity, and firms that previously capitalized are
more likely to capitalize again, otherwise risking a dip in reported performance. Regarding
the statistical diagnostics, we see that the Schwartz Criterion is − 707, indicating that the
Tobit algorithm has converged and our model is well specified. The pseudo R-squared of
73% indicates that the dependent variable is explained well by our regressors.
In Model 2, we utilize an alternative specification of change in profitability: ChROA-2.
This is calculated as the year-to-year change in profitability. We perform this action because
managers (or the market) could perceive change in profitability either as a change of “base”
profits (calculated as the average of the prior two years) or as a change from the last
earnings figure. ChROA-2 is still highly significant. However, the adjusted R-squared
drops by about six percentage points, providing some evidence that changes over a “base”
profit number calculated as the change in profitability over the average of the prior two
years are more important in the income smoothing decision as opposed to the yearly change
in profitability. All in all, the results in Table 6 provide support for H1 but not for H2. The
results are especially robust because in a small sample—after controlling for year and
industrial effects, for the materiality of the capitalization and for the prior year
capitalization—the change in profitability is consistently significant and is in the expected
direction.
In Table 7, we undertake a number of robustness tests in order to supplement our main
findings. Our first test introduces a dummy variable that tests for earnings thresholds.
Specifically, it controls if R&D capitalization is able to turn a negative ROA into a positive
one. Our second test replicates our main regression model (Table 6) with the addition of a
number of variables that control for characteristics of firm corporate governance. Finally,
our last robustness test involves a larger sample drawn from Datastream and is limited to the
variables available on that database. In each of the above tests, our results indicate that our
income smoothing hypothesis is robust to earnings thresholds and various governance
characteristics, it is not idiosyncratic to our sample and it is extended over multiple years.
More specifically, in Model (1) of Table 7 we replicate our multivariate analysis with the
addition of a new variable, Manage, which is a dummy variable equal to one if
capitalization helps a firm achieve a positive change in ROA from a negative one. We carry
out this analysis because a firm could expense R&D even if it is below its target, i.e., it is
below the average earnings level of the prior two years. This could be the case when
achieving an increase in profitability is not attainable. We expect this dummy variable to
have a positive relationship with Capitalization, since prior research identifies benchmark
beating as a motivation for earnings management (e.g., Burgstahler & Dichev, 1997).
Controlling for Manage enables us to test whether the inability to beat the benchmark
affects the income-smoothing motivation, allowing a better test of H1.
As expected, our results indicate that Manage is significantly and positively related to
Capitalization, indicating that a firm's decision to capitalize is motivated by profitability
concerns, that is to increase profit. Additionally, our original performance variable ChROA
is significant (P b 0.04) in the hypothesized direction, indicating that the incentive to
smooth earnings still holds after controlling for benchmark-beating ability.
 G. Markarian et al. / The International Journal of Accounting 43 (2008) 246–267 261

Table 7
Tobit regression analysis examining the relationship between R&D capitalization and hypothesized variables and
other control variables
Parameter Model 1 Model 2 Model 3
Estimate Pr N |t| Estimate Pr N |t| Estimate Pr N |t|
Intercept 0.001 0.77 0.003 0.70 − 0.000 0.85
ROA − 0.023 0.01 − 0.012 0.14 0.005 0.09
ChROA − 0.014 0.04 − 0.037 0.01 − 0.016 0.01
Manage 0.008 0.01
Leverage − 0.002 0.25 0.002 0.71 0.002 0.17
R&Dtotal − 0.007 0.62 − 0.059 0.06
Logassets 0.000 0.96 − 0.000 0.27 − 0.000 0.77
MB 0.001 0.04 0.000 0.41
Beta − 0.000 0.98 0.002 0.38
High-Capitalizer 0.007 0.01 0.009 0.01 0.010 0.01
LagCapitalization 0.087 0.16 0.193 0.07 0.495 0.01
BoardSize 0.000 0.42
Family − 0.002 0.06
BIndependence 0.005 0.07
Largest-Owner − 0.000 0.93
N 86 86 890
Schwartz Criterion − 719 − 213 − 3431
Pseudo R-squared 0.71 0.69 0.66
Model (1) Controls for the ability to reach positive earnings through capitalization. Model (2) adds a number of
control variables related to corporate governance characteristics. Model (3) tests the hypothesized variables on a
larger sample drawn from Datastream, using Capitalization-2 (which is an approximation of Capitalization) as the
dependent variable.
Capitalization is total R&D capitalization (net of the yearly depreciation) divided by the total assets of the firm.
ChROA is the change in return on assets (before capitalization) over the average of the prior two fiscal years.
Manage is a dummy variable equal to one if capitalization helps a firm achieve a positive change in ROA (from a
negative one). Leverage is total debt divided by total assets (before capitalization). R&Dtotal is the total investment
in R&D undertaken by the firm divided by total assets. ROA is measured as the firm's operating income before
capitalization divided by the total assets of the firm. Logassets is the logarithmic transformation of the total assets
of the firm at fiscal year-end. MB is the market value of equity divided by the book value. Beta is the sensitivity of
the asset's returns to market returns. High-Capitalizer is a dummy variable equal to one if the amount capitalized
(normalized by the absolute value of earnings) is above the median for capitalizing firms. LagCapitalization is
equal to Capitalization of the prior year. BoardSize is the number of board members. BIndependence is the fraction
of independent directors sitting on a board. Largest-Owner is the fraction of shares owned by the largest owner.
Family is a dummy variable equal to one if a firm is family controlled or family influenced, zero otherwise. Dummy
variables for year and industry membership are not reported.

In Model (2) of Table 7, we explicitly control for firm-governance characteristics.

Specifically, we control for firm-ownership structure, whether a firm is family controlled,
the proportion of independent directors on the board, and board size. We control for
ownership structure by calculating the percentage of shares owned by the largest
shareholder (Largest-Owner). This is done because concentrated ownership by a single
dominant shareholder creates a unique set of agency problems; this characteristic is
prevalent in the Italian capital market system. Brunello, Graziano, and Parigi (2003,
p.1029) document that “in more than half of listed firms (on the Milan Bourse) one
shareholder owns the absolute majority of common shares.” Concentrated ownership
262 G. Markarian et al. / The International Journal of Accounting 43 (2008) 246–267

makes the agency relationship between owners and managers less critical, but raises a new
agency problem between the controlling shareholders and the minority ones. Fan and Wong
(2002) argue that concentrated ownership limits the flow of accounting information to
outside investors, and that this limitation translates into greater opportunities for earnings
management. Introducing Largest-Owner as a control variable allows us to control for this
potentially relevant governance characteristic.
We also control for family ownership by employing a dummy variable, Family, if the firm is
majority owned (or heavily influenced) by a family.14 This action is taken since several Italian
firms are controlled (or strongly influenced) by one or more closely related families. This
characteristic presents a unique situation that also extends to financial-reporting decisions.
Prior studies (Chen & Radhakrishnan, 2005; Wang, 2006) found that family ownership,
ceteris paribus, is associated with higher earnings quality. This association is attributed
both to the ability of family firms to monitor and curtail opportunistic management's behavior
and to the stronger demand for higher quality earnings from outsiders. Therefore, Family
allows us to further control for this relevant governance characteristic.
Finally, we calculate the percentage of independent directors sitting on the board,
BIndependence, and the size of the board, BoardSize. Both of these characteristics have
been associated with the effectiveness of shareholder monitoring. In particular, a higher
level of board independence is related to higher levels of monitoring (Dechow, Sloan, &
Hutton, 1996; Klein, 2002), and smaller and more compact boards are better monitors
(Vafeas, 2000; Yermack, 1996). As a consequence, BIndependence and BoardSize have
been associated, respectively, with a lower and higher extent of earnings management.
Controlling for both of these factors allows for better tests of our hypotheses.
Model (2) of Table 7 presents results on our income-smoothing variable controlling for
firm governance characteristics. We see that ChROA is negatively related to Capitalization
(P b 0.01), indicating that the income-smoothing hypothesis still holds with the inclusion of
firm-governance characteristics. Two of the governance variables (Largest-Owner,
BoardSize) are not significant, indicating that in the Italian capital market system, these
governance structures are not related to the decision of R&D cost capitalization. However,
Family and BIndependence are weakly significant, indicating that these governance
characteristics are related to the R&D capitalization decision. Finally, ROA is no more
significant, which could indicate that controlling for governance structures, only the change
in profitability is related to R&D cost capitalization. In sum, Model (2) of Table 7 indicates
that governance variables play a marginal role in the capitalization decision; however, the
profitability results are not affected.
We also conduct several additional tests (results unreported), in which we control for:
bank ownership in firms (as a proxy for monitoring); whether the CEO is also the chair of
the board (as a further indicator of board independence); and the percentage of independent
directors on audit committees (as a measure of audit quality). These variables are not
significant and their inclusion does not change our primary findings.
In the last robustness test (Model (3), Table 7), we extend our sample over multiple years
in order to show that our results are not year- and sample-specific. Given that R&D-related

14
In order to distinguish family-controlled or -influenced companies from the others, we use the classification
proposed by Corbetta and Minichilli (2005).
 G. Markarian et al. / The International Journal of Accounting 43 (2008) 246–267 263

disclosures are difficult to find prior to 2001, we are not able to extend our hand-collection
in order to obtain a larger sample size and to conduct more powerful statistical tests. In
order to sidestep this problem, we use data available through Datastream. Unfortunately,
Datastream reports R&D and advertising capitalized costs as one data item and does not
report total R&D investments and amortization expenses related to the capitalized costs.
Nevertheless, we calculate Capitalization-2 as an approximation of our original dependent
variable based on the assumptions that follow. First, we consider all the capitalized costs to
be related to R&D. We believe this to be a realistic assumption since, based on our hand-
collected sample, we observe that capitalized advertising costs are rare in Italian companies
due to the strong limits of such capitalization set forth by the Italian accounting standards.
Second, we estimate the amortization of R&D costs by assuming a useful life of five years,
which is the maximum time allowed by the Italian accounting standards and the most
common policy among the companies in our hand-collected sample. Therefore,
Capitalization-2 is calculated as capitalized costs minus estimated annual amortization
divided by total assets.
We draw Datastream data from 1995 to 2004 for all the firms listed on the Milan Stock
Exchange that belong to the same industries as those from our hand-collected sample (in
order to reduce the risk of selecting companies not performing R&D activities). Our final
sample contains 890 observations. In Model (3) of Table 7, we see that ChROA is still
negatively related to our estimate of R&D cost capitalization (P b 0.01) and, similar to
before, we do not find evidence for our leverage hypothesis. These findings confirm the
robustness of the results obtained in the prior tables, strengthening our smoothing
hypothesis by applying it to a larger sample and a longer period of time.
We perform additional tests on our original sample to control for other occurrences that
might affect income smoothing by introducing two dummy variables (results unreported).
We first test the big-bath hypothesis (see Huson, Wiedman, & Wier, 2003) where we
control for changes in top management (whether the CEO or chair of the board changes
during the year being observed); results remain unchanged. Then, we control whether the
firm had an IPO in the year being observed or in the year prior to it. Results still remain
unchanged.
Consequently, our findings indicate strong support for our income-smoothing
hypothesis, given our results in Tables 6 and 7. This result is robust to alternate measures
of change in profitability and to various covariates such as firm size, risk, growth
opportunities, profitability, industrial membership and time controls, governance
characteristics, and other extraordinary issues that could affect income smoothing (e.g.,
IPOs and executive turnover). Finally, we augment our results by an approximation method
by which an estimated R&D capitalization variable is utilized in order to confirm our results
over a longer period of time and on a larger dataset.
Throughout all of our analyses, we do not consider real-earnings management.
As discussed before, earlier studies show that R&D costs are subject to real earnings-
management policies (Dechow & Sloan, 1991; Perry & Grinaker, 1994; Bushee, 1998).
This finding could partially affect our results and should be kept in mind when
interpreting them. Our study shows that—over and above real earnings-management
strategies—companies use accounting flexibility to affect earnings through accruals. The
existence of real-earnings management can weaken the relationship between the
264 G. Markarian et al. / The International Journal of Accounting 43 (2008) 246–267

capitalization of R&D costs and the variables related to earnings-management incentives,

since some companies may prefer to reduce the R&D expenditures more than they do the
related expenses through capitalization. Notwithstanding this effect, we find empirical
support for our hypothesis and, from this, our conclusions are strengthened regarding the
existence of accrual-earnings management as it relates to R&D costs.

7. Conclusions and limitations

This study examines the relationship between the choice of R&D cost accounting and
earnings-management incentives. We hypothesize that the decision to capitalize R&D costs
is related to a firm's change in profitability. Our results indicate that firms that have a lower
return on assets (compared to the average of the previous two years) are more likely to
capitalize R&D expenditures, while firms that have improved performance are more likely
to expense, consistent with the earnings-smoothing hypothesis. We also hypothesize that a
firm's level of debt financing is related to capitalizing decisions. We do not find support for
this prediction—probably due to the fact that financial institutions tend to adjust reported
earnings by eliminating the effect of any cost capitalization in order to limit the risk of
misleading manipulation.
Our income-smoothing-related tests are robust to the introduction of a number of control
variables and to a lengthening of the time period under analysis.
Our results have several policy implications. The convergence project recently started by
IASB and FASB has raised a debate surrounding what constitutes an optimal accounting
standard for R&D costs. Currently, the two bodies have different positions on this issue and
the limited empirical research carried out thus far does not help in finding a common solution.
This study contributes to this debate by providing empirical evidence on the use of R&D cost
capitalization for purposes of earnings management. The results indicate that managers use
R&D cost capitalization in order to smooth earnings. To the extent that earnings smoothing is
considered an opportunistic strategy, our conclusions move in favor of the current FASB
position, which does not allow for flexibility and requires all R&D costs to be expensed. On
the other hand, we should not forget that earnings smoothing can also be an effective and
efficient way to signal and communicate important information to the market and that prior
literature generally supports R&D cost capitalization in terms of relevance for those who
utilize financial statements. Therefore, whether the best accounting treatment for R&D costs
is full expensing or capitalization (subject to some conditions) remains a controversial
accounting issue that again leads us to the traditional relevance/reliability trade-off.

References

Aboody, D., & Lev, B. (1998). The value relevance of intangibles: The case of software capitalization. Journal of
Accounting Research, 36(3), 161−191.
AICPA (1994). Improving financial reporting—a customer focus. AICPA Special Committee on Financial Reporting.
Association for Investment Management and Research (1993). Financial Reporting in the 1900s and beyond.
Ballester, M., Garcia-Ayuso, M., & Livnat, J. (2003). The economic value of the R&D intangibile assets. European
Accounting Review, 12(4), 605−633.
Barnea, A., Ronen, J., & Sadan, S. (1976, January). Classificatory smoothing of income with extraordinary items.
The Accounting Review, 110−122.
 G. Markarian et al. / The International Journal of Accounting 43 (2008) 246–267 265

Barth, M. E., Landsman, W. R., & Wahlen, J. M. (1995). Fair value accounting: Effects on banks' earning
volatility, regulatory capital, and value of contractual cash flows. Journal of Banking and Finance, 19(3/4),
577−605.
Beidleman, C. R. (1973). Income smoothing: The role of management. The Accounting Review, 48(4), 653−667.
Bhushan, R. (1994). An informational efficiency perspective on the post-earnings announcement drift. Journal of
Accounting and Economics, 18(July), 45−65.
Boone, A. L., Field, L. C., Karpoff, J. M., & Raheja, C. G. (2007). The determinants of corporate board size and
composition: An empirical analysis. Journal of Financial Economics, 85(1), 66−101.
Brunello, G., Graziano, C., & Parigi, B. M. (2003). CEO turnover in insider-dominated boards: The Italian case.
Journal of Banking & Finance, 27, 1027−1051.
Bublitz, B., & Ettredge, M. (1989, January). The information in discretionary outlays: Advertising, research and
development. The Accounting Review, 108−124.
Buckmaster, D. a. l. e. (2001). Development of the Income Smoothing Literature 1893–1998: A Focus on the
United States. Elsevier Science Ltd.
Burgstahler, D., & Dichev, I. (1997). Earnings management to avoid earnings decreases and losses. Journal of
Accounting and Economics, 24, 99−126.
Bushee, B. J. (1998). The influence of institutional investors on myopic R&D investment behavior. The
Accounting Review, 73(3), 305−333.
Cahan, S. F. (1992). The effect of antitrust investigations on discretionary accruals: A refined test of the political-
cost hypothesis. The Accounting Review, 67(1), 77−95.
Chaney, P. K., & Lewis, C. M. (1995). Earnings management and firm valuation under asymmetric information.
Journal of Corporate Finance, 1, 319−345.
Chaney, P. K., Jeter, D. C., & Lewis, C. M. (1998). The use of accruals in income smoothing: A permanent earnings
hypothesis. Advances in Quantitative Analysis of Finance and Accounting, 6, 103−135.
Chen, T., & Radhakrishnan, S., (2005). Family firms, audit and non-audit fees and earnings management. Working
Paper, Hong Kong University of Science and Technology.
Copeland, R. M. (1968). Income smoothing. Journal of Accounting Research, 6(3), 101−116.
Corbetta, G., & Minichilli, A. (2005). Il governo delle imprese italiane quotate a controllo familiare: I risultati di
una ricerca esplorativa. Economia & Management(6), 59−77.
DeAngelo, L. (1986). Accounting numbers as market valuation substitutes: A study of management buyouts of
public stockholders. The Accounting Review, 61(3), 400−420.
DeAngelo, H., DeAngelo, L., & Skinner, D. (1994). Accounting choice in troubled companies. Journal of
Accounting and Economics, 17, MS−143.
DeFond, M. L., & Jiambalvo, J. (1994). Debt covenant and manipulation of accruals. Journal of Accounting and
Economics, 17, 145−176.
Dechow, Patricia M., & Schrand, Catherine M. (2004). Earnings quality. Research Foundation of CFA Institute.
Dechow, P. M., & Sloan, R. G. (1991). Executive incentives and the horizon problem: An empirical investigation.
Journal of Accounting and Economics, 51−89.
Dechow, P. M., Sloan, R. G., & Sweeney, A. P. (1995). Detecting earnings management. The Accounting Review,
70(2), 193−226.
Dechow, P., Sloan, R. G., & Hutton, A. P. (1996). Causes and consequences of earnings manipulation: An analysis
of firms subject to enforcement actions by the SEC. Contemporary Accounting Research, 13(2).
Dowdell, T., & Press, E. (2004). The impact of SEC scrutiny on financial statement reporting of in-process research
and development expense. Journal of Accounting and Public Policy, 23(3), 227−244.
Duke, J. C., & Hunt, H. G., III (1990). An empirical examination of debt covenant restrictions and accounting-
related debt proxies. Journal of Accounting and Economics, 12, 45−63.
El Gazzar, S., Lilien, S., & Pastena, V. (1986). Accounting for leasing by lessees. Journal of Accounting and
Economics, 8, 217−237.
FASB (1974). SFAS No. 2: Accounting for research and development costs. Financial Accounting Standards Board.
FASB (1980). Qualitative characteristics of accounting information. Financial Accounting Standards Board.
FASB (1985). SFAS No. 86: Accounting for the costs of computer software to be sold, leased or otherwise
marketed. Financial Accounting Standards Board.
Fan, J., & Wong, T. J. (2002). Corporate ownership structure and the informativeness of accounting earnings in
East Asia. Journal of Accounting & Economics, 33, 401−425.
266 G. Markarian et al. / The International Journal of Accounting 43 (2008) 246–267

Francis, J., Hanna, J. D., & Vincent, L. (1996). Causes and effects of discretionary asset write-offs. Journal of
Accounting Research, 34(Supplement), 117−134.
Fudenberg, D., & Tirole, J. (1995). A theory of income and dividend smoothing based on incumbency rents.
Journal of Political Economy, 103(1), 75−93.
Gaver, J. J., Gaver, K. M., & Austin, J. R. (1995). Additional evidence on bonus plans and income management.
Journal of Accounting and Economics, 19, 3−28.
Gebhardt, W. R., Lee, C. M., & Swaminathan, B. (2001). Toward an implied cost of capital. Journal of Accounting
Research, 39(1), 135−176.
Godfrey, J. M., & Jones, K. L. (1999). Political cost influences on income smoothing via extraordinary item
classification. Accounting and Finance, 39, 229−254.
Gordon, M. C. (1964, April). Postulates, principles and research in accounting. The Accounting Review, 251−263.
Healy, P. M. (1985). The effect of bonus schemes on accounting decisions. Journal of Accounting and Economics,
7, 85−107.
Healy, P., & Wahlen, J. (1999). A review of the earnings management literature and its implications for standard
setting. Accounting Horizons, 13(14), 365−383.
Healy, P., Myers, S. C., & Howe, C. D. (2002). R&D accounting and tradeoff between relevance and objectivity.
Journal of Accounting Research, 40(3), 677−710.
Hirschey, M., & Weygandt, J. (1985, Spring). Amortization policy for advertising and R&D expenditures. Journal
of Accounting Research, 326−335.
Holthausen, R., Larcker, D. F., & Sloan, R. G. (1995). Annual bonus schemes and the manipulation of earnings.
Journal of Accounting and Economics, 19, 29−74.
Hunt, H. G., III (1985, Autumn). Potential determinants of corporate inventory accounting decisions. Journal of
Accounting Research, 448−467.
Hunt, A., Moyer, S. E., & Shevlin, T. (1996). Managing interactive accounting measures to meet multiple
objectives: A study of LIFO firms. Journal of Accounting and Economics, 21(3), 339−374.
Huson, M., Wiedman, C., & Wier, H., (2003). Cash Compensation and Earnings Components Around CEO
Turnover. Working Paper University of Alberta.
IASB (2004). Framework for the preparation and presentation of financial statements. International Accounting
Standards Board.
IASB (2004). International Accounting Standards 38: Intangible Assets. International Accounting Standards Committee.
Imhoff, E. A. (1977, Spring). Income smoothing: A case for doubt. Accounting Journal, 85−101.
Jensen, M. C., & Meckling, W. H. (1976). Theory of the firm: Managerial behaviour, agency costs and ownership
structure. Journal of Financial Economics, 4, 305−360.
Jones, J. F. (1991). Earnings management during import relief investigations. Journal of Accounting Research, 29(2),
193−228.
Kamin, J., & Ronen, J. (1978). The smoothing of income numbers: Some empirical evidence on systematic
differences among manager-controlled and owner-controlled firms. Accounting, Organisations and Society, 3(2).
Kennedy, P. (2003). Econometrics. Prentice–Hall.
Klein, A. (2002). Audit committee, board of director characteristics and earnings management. Journal of
Accounting and Economics, 33, 375−400.
Kothari, S. P., Laguerre, T., & Leone, A. (2002). Capitalization versus expensing: Evidence on the uncertainty of
future earnings from current investments in PPE versus R&D. Review of Accounting Studies, 7(4), 355−382.
Lev, B., & Zarowin, P. (1999). The boundaries of financial reporting and how to extend them. Journal of Accounting
Research, 37(2), 353−385.
Mande, V., File, R. G., & Kwak, W. (2000). Income smoothing and discretionary R&D expenditures of Japanese
firms. Contemporary Accounting Research, 17(2), 263−302.
McNichols, M., & Wilson, G. P. (1988). Evidence of earnings management from the provision for bad debts.
Journal of Accounting Research, 26(1988), 1−40.
Miller, G. S., & Skinner, D. J. (1998). Determinants of the valuation allowance for deferred tax assets under SFAS
No. 109. The Accounting Review, 73(2), 213−233.
Monahan, S. (2005). Conservatism, growth and the role of accounting numbers in the fundamental analysis
process. Review of Accounting Studies, 10(2–3), 227−260.
Moses, O. D. (1987, April). Income smoothing and incentives: Empirical tests using accounting changes. The
Accounting Review, 358−377.
 G. Markarian et al. / The International Journal of Accounting 43 (2008) 246–267 267

Nelson, M., Elliott, J., & Tarpley, R. (2003). How are earnings managed? Examples from auditors. Accounting
Horizons(Supplement), 17−35.
Perry, S., & Grinaker, R. (1994). Earnings expectations and discretionary research and development spending.
Accounting Horizons, 8(4), 43−51.
Press, E., & Weintrop, J. (1990). Accounting-based constraints in public and private debt agreements: Their
association with leverage and impact on accounting choice. Journal of Accounting and Economics, 12, 65−95.
Ronen, Joshua, & Sadan, Simcha (1981). Smoothing income numbers: Objectives, means and implication.
Addison-Wesley.
Rusbarsky, M. (1988). Motivations for discretionary accounting changes: the case of the switch from accelerated to
straight-line depreciation, Working Paper, Graduate School of Management, University of California,
Riverside, CA.
Schilit, Howard (2002). Financial shenanigans. McGraw–Hill.
Schipper, K. (1989, December). Commentary on earnings management. Accounting Horizons, 91−102.
Schrand, C. M., & Wong, M. H. F. (2003). Earnings management using the valuation allowance for deferred tax
assets under SFAS No. 109. Contemporary Accounting Research, 20(3), 579−611.
Shevlin, T. (1991, January). The valuation of R&D firms with R&D partnerships. The Accounting Review, 1−21.
Sougiannis, T. (1994, January). The accounting based valuation of corporate R&D. The Accounting Review,
44−69.
Sweeney, A. P. (1994). Debt-covenant violations and managers' accounting responses. Journal of Accounting and
Economics, 17, 281−308.
Tucker, J. W., & Zarowin, P. A. (2006). Does income smoothing improve earnings informativeness? The
Accounting Review, 81(1), 251−270.
Trueman, B., & Titman, S. (1988). An explanation for accounting income smoothing. Journal of Accounting
Research, 26(Supplement 3).
Vafeas, N. (2000). Board structure and the informativeness of earnings. Journal of Accounting and Public Policy, 19(2).
Wang, D. (2006). Founding family ownership and earnings quality. The Accounting Review, 44(3), 619−656.
Wang, Z., & Williams, T. H. (1994). Accounting income smoothing and stockholder wealth. Journal of Applied
Business Research, 10(3), 96−104.
Watts, R., & Zimmerman, J. (1986). Positive Accounting Theory. Prentice–Hall.
Wong, J. (1988). Political costs and an intraperiod accounting choice for export tax credits. Journal of Accounting
and Economics, 10(1), 37−51.
Yermack, D. (1996). Higher market valuation of companies with a small board of directors. Journal of Financial
Economics, 40, 185−212.
Young, S. (1998). The determinants of managerial accounting policy choice: Further evidence for the UK.
Accounting and Business Research, 28(2), 131−143.
Zucca, L., & Campbell, D. (1992, September). A closer look at discretionary writedowns of impaired assets.
Accounting Horizons, 30−41.