The impact of IAS 36 on goodwill disclosure: Evidence of the writeoffs and performance effects
Gabriele D'Alauro
University of Genoa, (Italy)
Received March, 2013
Accepted October, 2013
D'ALAURO, G. (2013). The impact of IAS 36 on goodwill disclosure: Evidence of the writeoffs and performance effects. Intangible Capital, 9(3): 754799. http://dx.doi.org/10.3926/ic.415

Abstract
Purpose: This paper aims at examining the quality of corporate disclosure about goodwill impairment and its relationship with goodwill writeoffs and earnings performance, exploiting an accounting regulation that allows significant unverifiable estimates whilst requires a high level of information.
Design/methodology/approach: This study, based on a sample of Italian and British firms with market indications of goodwill impairment, verifies through a both univariate and multivariate analysis whether the level of disclosure is positively related to the magnitude of goodwill writeoff and to earnings performance, using a selfconstructed score of mandatory disclosure about goodwill impairment tests in accordance with IAS 36 requirements.
Findings: In a general context of insufficient information, we find that for Italian firms both the magnitude of goodwill writeoffs and earnings performance are significantly and positively associated to the level of mandatory disclosure about goodwill impairment tests. For the British firms, as companies more used to impairment test rules, the data does not confirm any significant association.
Research limitations/implications: The objective of this study is to test the initial impact of IAS 36 in the first years of its application, selecting a sample of firms belonging to limited but significant activity sectors. Future research could usefully analyse a wider sample of firms, also extending the time period of analysis. In any case, the findings of our study are consistent with the insights of earnings management theory, suggesting that the subjectivity inherent in impairment test assumptions could be used opportunistically by managers.
Originality/value: This research investigates questions still relatively unexplored, concerning the effects of goodwill writeoffs and accounting performance on corporate disclosure about goodwill impairment test. Based on this analysis, the study shows that corporate disclosure could be a “litmus paper” able to test the degree of good faith with which each firm has implemented IAS 36 requirements.
Keywords: Disclosure, Goodwill Writeoffs, Accounting Performance, Corporate Governance, Earnings Management
Jel Codes: M41, M48

Introduction
There is an ongoing debate on the importance of goodwill accounting. The Financial Accounting Standards Board (FASB) has sought to improve the relevance of this issue by moving towards goodwill impairment rules with SFAS 142 – Goodwill and Other Intangible Assets (FASB, 2001). Similarly, IAS 36 – Impairment of Assets (IASB, 2004) eliminates goodwill amortization, requiring instead that goodwill be evaluated at least annually for possible impairment.
The shift from amortization to periodic reviews puts a new and continuous responsibility on business managers to determine the recoverable amount of goodwill and a new burden on auditors, regulators and investors to evaluate management determination (Hayn & Hughes, 2006).
In the light of IAS 36 requirements, as well known, an impairment test is based on a chain of significant assumptions, with reference, for instance, to cash generating units identification, discount rates estimate, growth rates appraisal. Such a degree of allowable discretion, as earnings management theory predicts, could be used opportunistically by managers (Watts, 2003; Quagli & Meini, 2007). On the other hand, the level of impairment disclosure required by IAS 36 is considerably high, regardless of whether goodwill writeoffs are recorded or not.
In the general context of IAS introduction, it is well documented that companies do not necessarily comply with accounting standards mandatory disclosure (Tsalavoutas, 2011), as the existence of legislation and enforcing bodies does not guarantee compliance (Yeoh, 2005) and, even when disclosures are mandatory, firms still have some flexibility in the way they report the information (Chavent, Ding, Fu, Stolowy & Wang, 2006; Chen & Jaggi, 2000). As a consequence, our analysis aims at measuring compliance with IAS 36 mandatory disclosure during the first years of its implementation, determining a significant change in the accounting treatment of goodwill.
Particularly, using a sample of Italian and British listed firms with market indications of goodwill impairment, this paper examines whether the quality of disclosure about a goodwill impairment test is related to the magnitude of goodwill writeoff.
We consider both Italian and British consolidated financial statements for the period 2006 – 2008, in order to check the initial impact of IAS 36 application for countries with a significantly different accounting tradition. OIC 24 – Intangible Assets (OIC, 2005) requires the systematic amortization of goodwill, while under FRS 10 – Goodwill and Intangible Assets (ASB, 1997) goodwill should be amortised only if it is regarded as having a limited useful life, otherwise it should not be amortised but reviewed for impairment at each periodend, providing an adequate disclosure in the notes to the accounts.
In a context of relevant unverifiable estimates which can seriously increase the likelihood of opportunistic behaviour, we assume that impairment disclosure could represent a relevant indicator of the degree of good faith with which management has implemented IAS 36 requirements on a goodwill impairment test. Accordingly, we test the positive relation between disclosure quality and magnitude of goodwill writeoffs.
A separate but related issue is whether the level of disclosure is related to earnings performance. A few prior studies provide evidence of a positive relation between the general level of corporate disclosure and accounting performance indicators. Hence, we also test the positive association between the quality of specific disclosure about goodwill impairment and accounting performance of the firm.
Our analysis also allows to check the impact of the quality of corporate governance on the degree of corporate disclosure, by including specific variables connected with the composition of the board of directors and the activity of the audit committee, which should monitor the integrity of the financial statements of the company (Financial Reporting Council, 2006; Borsa Italiana, 2006).
This paper contributes to the existing research in several ways. Firstly it provides empirical evidence on the magnitude of goodwill writeoffs, company profitability, the level of impairment disclosure and the quality of corporate governance with reference to a sample of firms of two countries with very different cultural contexts and accounting traditions. Secondly, it focuses on a question still relatively unexplored, finding a significant and positive relation between disclosure quality and goodwill writeoffs. Thirdly, it studies the relation between corporate disclosure and earnings performance in a new context, that of goodwill accounting, providing evidence of the effects of accounting performance on the degree of corporate disclosure inherent in goodwill impairment tests.
This paper is organised as follows.
The next section provides background on related research. Section 3 develops the hypotheses for the study. Section 4 describes our sample selection and research design. In Section 5 the empirical results are shown. Section 6 discusses the results. In Section 7 additional analyses and robustness test are conducted. The final section draws conclusions.
Previous Research
There is a long stream of research that examines relevant issues related to goodwill accounting; in addition, the adoption of SFAS 142 and IFRS 36 has encouraged new studies on writeoffs. Looking at the literature related to impairment of goodwill, there are three fundamental lines of research.
The first analyses the determinants of goodwill writeoffs, in order to verify if managers, according to agency theory prediction, use discretion in accounting standards to manage their earnings opportunistically. Especially, these studies test if nonimpairment decision increases in economic or financial characteristics that serve as proxies for greater unverifiable fair value based discretion.
The second related literature seeks to provide evidence on association between goodwill writeoffs, equity market values and stock returns. A subset of this literature examines the ability to predict goodwill impairment on the basis of accounting or financial performance ratios.
The third line of research, in some ways still relatively unexplored, aims to analyse the determinants of disclosure quality about goodwill impairment test. However, most of these studies only provide descriptive statistics about the content of corporate disclosure.
Overall, it should be added that the issues relating to these research areas are treated together in many works, as described below.
Beatty and Weber (2006) examine the factors affecting the decision to take an SFAS 142 writeoff and the percentage of the goodwill that is actually writtenoff. With reference to previous studies (Watts & Zimmerman, 1990; Francis, Hanna & Vincent, 1996), they find empirical evidence that firms’ debt contracting, bonus, turnover and exchange delisting incentives affect managers’ decisions to accelerate or delay expense recognition. In particular, they argue that managers with longer tenures are more likely to have been involved in the acquisitions that generated that goodwill. The data confirms that, to avoid reputation costs, such longtenure managers are less likely to take goodwill writeoffs.
Ramanna (2008) studies the evolution of SFAS 142, which uses unverifiable fairvalue estimates to account for acquired goodwill, in order to test what is argued in previous research (Holthausen & Watts, 2001; Watts, 2003; Roychowdhury & Watts, 2007). He assumes that FASB issues SFAS 142 in response to political pressure over its proposal to abolish pooling accounting, and the results are consistent with SFAS 142 impairment tests being due in part to firms opposed to abolishing pooling.
Another study of the same author, together with Watts (Ramanna & Watts, 2012), finds a low frequency of goodwill writeoffs in a sample of American firms with strong market indications of goodwill impairment (firms with book goodwill and two successive years of book to market ratios above one). The data does not confirm that the decision of avoiding impairment is due to management’s possession of favourable private information. On the contrary, the authors find evidence that nonimpairment is associated with agencybased motives: goodwill impairments decrease in CEO reputation and debtcovenant violation concerns. However, the results don’t confirm a significant association between goodwill writeoffs and firm capitalisation ratios.
Ahmed and Guler (2007) focus on the relationship between impairment of goodwill, stock returns and stock prices. In particular, contrary to Ramanna and Watts (2012), they find a significant negative association between goodwill writeoffs and stock returns in the post SFAS 142 period. Furthermore, they find evidence that such association is higher for firms that have a high number of segments, suggesting that goodwill numbers are more reliable for firms with a high number of segments relative to firms with a low number of segments.
Bens, Heltzer and Segal (2007) document a negative and significant stock market reaction to unexpected goodwill writeoffs. In particular, they find evidence that the market reaction is attenuated for firms with low information asymmetry, suggesting that the market impounds this information into price for these firms prior to the public announcement by the company. In contrast with Ahmed and Guler (2007), the authors find no variation in market based on the complexity of the firm’s structure (their proxy is the number of reported segments): however, in this study the sample is restricted to the firms with magnitude of goodwill impairment higher than 5% of total assets.
Hayn and Hughes (2006) find that the ability to predict goodwill impairment based on performance indicators and information provided in the financial statements is limited. This is also due to the fact that these are general indicators that pertain primarily to the firm as a whole rather than to a particular segment or reporting unit to which goodwill shall, from the acquisition date, be allocated. Indeed, certain characteristics of acquired companies such as the premium paid in the acquisition, the percentage of the purchase price assigned to goodwill and the use of stock as the primary mode of consideration, appear to contribute more to the prediction of goodwill writeoffs than available disclosures on the acquired entity in the years subsequent to the acquisition.
With regard to impairment disclosure, Paananen (2008), using a random sample of companies from France, Germany and the United Kingdom, examines the comparability of fair value accounting of goodwill under IFRS. The data confirms, as expected, that large companies operating in United Kingdom, which is considered an environment with a relatively higher level of investor protection, are more likely to provide more disclosure on a goodwill impairment test. However, the author recognizes that the results should be interpreted cautiously since the study has inherent limitation of a small sample size and a simplistic method is used to measure disclosure levels among the sampled companies.
The study of Verriest and Gaeremynck (2009) investigates the determinants of goodwill impairment decisions, finding empirical evidence that companies with stronger corporate governance mechanism and also firms exhibiting better accounting and market performance are more likely to impair. The authors, according to Francis et al. (1996), argue that better performing firms are more likely to engage in goodwill impairment as the signal they send out to investors of a lower profitability is weaker and of lower importance, provided that these firms are financially healthy.
In the same study Verriest and Gaeremynck (2009) also examine the determinants of disclosure quality on goodwill impairment. However, they find that ownership structure and corporate governance quality have a weak impact on the degree of impairment disclosure. In addition, the data doesn’t confirm the expected positive association between accounting performance indicators and the level of impairment disclosure.
There is also a more extensive research on the influence of earnings performance or earnings quality on the general level of mandatory or voluntary corporate disclosure, that is, disclosure not specifically inherent in goodwill impairment tests. In any case, some of these studies provide evidence of an increase in all types of disclosures during periods of increased earnings (Miller, 2002) and find that voluntary disclosure and earnings quality are positively related (Francis, Nanda & Olsson, 2008). In a broader context, other works investigate how firm disclosure activity affects the relation between current annual stock returns, contemporaneous annual earnings and future earnings (Lundholm & Myers, 2002) and examine the effect of voluntary disclosure on the use of discretionary accruals to smooth earnings and on the value relevance of earnings (LapointeAntunes, Cormier, Magnan & GayAngers, 2006).
Hypothesis Development
The results of studies of goodwill writeoffs determinants are generally mixed. Prior research does not find strong evidence on identifying specific factors able to have significant predictive ability for goodwill impairments. Indeed, analytical research provides conflicting predictions about how stock returns or accounting performance influence the magnitude of goodwill writeoffs.
This seems primarily due to the fact that the key indicators treated by most of literature pertain to the firm as a whole rather than to the specific cash generating unit to which each goodwill shall be allocated.
On the other hand, it should be noted that IAS 36 impairment tests allow significant unverifiable estimates which can seriously increase the likelihood of opportunistic disclosures. In estimating the recoverable amount of goodwill, management assesses the reasonableness of the assumption on which its current cash flow projections are based, with reference, for instance, to cash generating units identification, discount rates estimate, growth rates appraisal. In this context, managers could exploit the high degree of discretion in order to manage earnings, in line with the insights of earnings management theory (Watts, 2003; Quagli & Meini, 2007).
Relating to this issue, a few prior studies find evidence that nonimpairment is associated with agencybased motives and is increasing in financial characteristics, as number or size of reporting units and unverifiable net assets in reporting units, that serve as proxies for greater unverifiable fair value discretion.
However, the level of impairment disclosure required by IAS 36 is considerably high, regardless of whether goodwill writeoffs are recorded or not. Firms have to provide detailed information in the notes to the financial statements on any significant assumption used to determine goodwill fair value or value in use.
Hence, disclosure requirements seem to act as a significant counterweight to the several profiles of subjectivity inherent in impairment test assumptions. On the other hand, a low level of disclosure provided by the firm could reveal earnings management.
Prior research generally has not looked at the association between disclosure quality and magnitude of goodwill writeoffs. This study will address this issue, in order to investigate the existence of an interrelation between mandatory disclosure in accordance with IAS 36, recording of goodwill impairments and earnings performance.
In a sample of Italian and British listed firms with market indications of goodwill impairment, we firstly assume a lower level of disclosure provided by the firm in case of nonimpairment. Accordingly, we hypothesize a positive relation between disclosure quality and magnitude of goodwill writeoffs.
Hypothesis 1 (H1):
“The level of impairment disclosure provided by firms with market indications of goodwill impairment is positively associated to the magnitude of goodwill writeoffs”.
Our second hypothesis, consistent with the group of studies that provide evidence on positive relation between the general level of corporate disclosure and accounting performance indicators, is the following.
Hypothesis 2 (H2):
“The level of impairment disclosure provided by firms with market indications of goodwill impairment is positively associated to earnings performance”.
However, as discussed earlier, it should be noted that Italy and the United Kingdom are countries with very different cultural contexts and accounting traditions. In particular, only British firms have applied goodwill impairment tests before IAS 36 effective date, in accordance with FRS 10 requirements. As a result, we suppose that disclosure provided by British firms, as company more used to impairment test rules, is less influenced by goodwill writeoffs and accounting performance.
Hypothesis 3 (H3):
“The association between impairment disclosure and goodwill writeoffs and the association between impairment disclosure and accounting performance are higher for Italian firms compared to British firms”.
Research Design
Sample Selection
As noted in the literature review, the sample selection criteria adopted in previous related studies are not the same.
For instance, Bens et al. (2007) select firms with magnitude of goodwill impairment higher than 5% of total assets, Ramanna and Watts (2012) analyse only firms with two successive years of book to market ratios above one, Beatty and Weber (2006) restrict the sample to firms with a difference between the market and the book value of their equity that is less than their recorded goodwill, Paananen (2008) randomly selects firms with a positive gross value of goodwill.
In summary, most of the previous studies identify selection criteria, albeit variously configurable, in order to select firms with indication of likely goodwill impairment.
Similarly, in our research we analyse only firms that are expected to engage in goodwill impairment. We do this by investigating whether the firm market to book value (calculated before the effect of any goodwill impairment) is smaller than one or under the median of market to book value. The median of market to book value is calculated for each observation year and for each country.
The data used in the empirical tests is drawn from the consolidated financial statements of Italian and British sampled companies for the years 2006, 2007 and 2008. We choose this time period in order to investigate the initial impact of IAS 36 in the first years of its application: however, we exclude the year 2005 as the data could be influenced by the extraordinary effects caused by the transition to the different treatment of goodwill.
As discussed earlier, we consider both Italian and British companies in order to check the impact of IAS 36 application for countries with significantly different accounting traditions. Specifically, we choose, on the one hand, a country whose national accounting standard on intangible assets (the Italian OIC 24) is founded on the systematic amortization of goodwill, and, on the other hand, a country whose corresponding accounting standard (the British FRS 10) is based on the impairment test.
The sample selection criteria are detailed in Table 1.

Italian companies 
British companies 
Total 
First number of companies of selected sectors 
42 
39 
81 
First number of sampled consolidated financial statements 2006 – 2008 
126 
117 
243 
Less: 



Financial statements with zero goodwill value 
 14 
 17 
 31 
Financial statements with market to book value both > median and > 1 
 52 
 48 
 100 
Financial statements with negative book value of equity 
 1 
 1 
 2 
Final sample 
59 
51 
110 
Table 1. Sample Selection Procedure
We firstly selected all Italian and British consolidated financial statements of companies:
continuously listed from 2005 until 2008 in the stock market of their own country of origin (and not listed at the same time in the United States stock market);
not reporting under IAS/IFRS before year 2005.
Then, we calculated for each activity sector the standard deviation of the number of firms belonging respectively to Italy and United Kingdom and we selected the sectors with own standard deviation equal to the median value. This in order to analyse the representative areas of the typical difference in terms of number of companies between the two countries. As a result, we identified the next five sectors with the same standard deviation (equal to 2.12):
“Software and Computer Services”;
“Electronic and Electrical Equipment”;
“Automobiles and Parts”;
“Construction and Materials”;
“Household Goods and Home Construction”.
It should be added, in order to evaluate the significance of the selected sectors, that they play a great importance within the gross domestic product in both countries, and concern both “old economy” and “new economy” areas. A significant number of large companies, listed on their respective stock markets, belong to these sectors, which are characterized by high interrelations among themselves and with other important sectors of national and international economy. Moreover, these areas are not affected by very high fluctuations in prices, and, finally, the selected sectors are not characterized by forms of monopolistic or monopsonistic market, but at mostly by oligopolistic configurations close to imperfect competition.
The first number of sampled consolidated financial statements, referred to the years from 2006 to 2008, is 243; applying the selection criteria described in Table 1, we identify the final sample of 110 cases. This seems to be a justified reduction of the sample, having regard to the aim of this research. In addition, the final number of observations is aligned to the sample identifying in previous related studies, for instance the research of Ramanna and Watts (2012), reported to 124 observations, and the paper of Verriest and Gaeremynck (2009), focused on 47 statements.
We made a direct reading of each consolidated financial statements, without the use of any database, also in order to analyse the quality of disclosure about the goodwill impairment test reported in the notes to the accounts.
Empirical Models
In order to find evidence of our research hypotheses, we firstly develop a disclosure model concerning a goodwill impairment test. Disclosure is defined here as consisting of mandatory items of information provided in the notes to the consolidated financial statements in accordance with the requirements of IAS 36.
Table 2 describes ten items identified to measure corporate quality disclosure: the approach to scoring items is dichotomous in that an item scores one if disclosed and zero otherwise.
Score 
Requirement 
1 
Identification of each cash generating unit (CGU) over which goodwill is allocated 
1 
Goodwill allocation to CGUs 
1 
No changes in CGUs identification or in goodwill allocation to CGUs since the previous year 
1 
Information on the numbers of years covered by the budgets 
1 
Information on the growth rate used to extrapolate cash flow projections beyond the period covered by the budgets 
1 
Information on the numbers of years over which management has projected cash flow based on financial budgets 
1 
Information on the discount rates applied to the cash flow projections 
1 
Differentiation of the discount rates applied to each CGUs 
1 
Recourse to external sources of information or appraisal to verify the assumptions on which management has founded its impairment test 
1 
Sensitivity analysis for the units’ recoverable amount 
0 
Minimum score 
10 
Maximum score 
Table 2. Disclosure Score
Hence the maximum a company could achieve is 10 and the minimum zero.
The dichotomous method gives equal weight to the individual items required to be disclosed by the standard (Cooke, 1989; Cooke, 1992; Hossain, Tan & Adams, 1994) and therefore it enables to reduce the degree of subjectivity in the evaluation of mandatory information provided by each firm (Tsalavoutas, Evans & Smith, 2010).
The number of items is limited, but it results from the scope of our study, which concentrates on a single topic, that is disclosure regarding goodwill impairment tests. In addition, our number of requirements is aligned to the number of items identifying in some previous disclosure studies, for instance 11 items (Tai, AuYeung, Kwok & Lau, 1990), 9 items (Prencipe, 2004), 14 items (Chavent et al., 2006), 6 items (Deumes & Knechel, 2008), 12 items (Greco, 2010).
In short, our items refer to information about cash generating units, identification and allocation of goodwill, as well as each key assumption used by management to measure units’ recoverable amount, including information on recourse to external source of information and sensitivity analysis.
All the sampled firms have applied the “value in use” method in order to evaluate the recoverable amount of own goodwill.
Variable 
Definition 
DISC 
Score disclosure (from 0 to 10) concerning goodwill impairment test (see table 2) 
IMP / AST 
Goodwill impairment scaled by total assets (calculated before the effect of goodwill impairment) 
IMP / EQT 
Goodwill impairment scaled by equity (calculated before the effect of goodwill impairment) 
IMP / GDW 
Goodwill impairment scaled by goodwill (calculated before the effect of goodwill impairment) 
ROE 
Profit (loss) for year (calculated before the effect of goodwill impairment) scaled by equity (calculated before the effect of goodwill impairment) 
AVG.ROE 
Average ROE in the period 2005 – 2008 
A.C. MEET. 
Number of meetings of the audit committee held in year 
A.C.IND.DIR. 
Number of independent directors members of the audit committee scaled by total number of directors 
SIZE 
Natural logarithm of total assets 
YEAR.06 
Dummy variable set to one if the case concerns the year 2006 
YEAR.07 
Dummy variable set to one if the case concerns the year 2007 
YEAR.08 
Dummy variable set to one if the case concerns the year 2008 
SECT.A 
Dummy variable set to one if the case concerns the sector A “Software and Computer Services” 
SECT.B 
Dummy variable set to one if the case concerns the sector B “Electronic and Electrical Equipment” 
SECT.C 
Dummy variable set to one if the case concerns the sector C “Automobiles and Parts” 
SECT.D 
Dummy variable set to one if the case concerns the sector D “Construction and Materials” 
SECT.E 
Dummy variable set to one if the case concerns the sector E “Household Goods and Home Construction” 
COUNTRY 
Dummy variable set to one if the case concerns Italian companies 
Table 3. Variable Definitions
Table 3 includes the definitions of all variables used in this study, which apply to disclosure level (DISC), goodwill writeoffs, alternative ratios of company profitability, corporate governance variables and firm size. Most of the control variables have been commonly used in prior disclosure research studies (Cooke, 1991; Forker, 1992; Hossain, Tan & Adams, 1994; Wallace, Naser & Mora, 1994; Wallace & Naser, 1995; Chen & Jaggi, 2000; Eng & Mak, 2003; Cerbioni & Parbonetti, 2007).
In particular:
to ensure the robustness of the results, three different scalars are used in order to measure the goodwill impairment loss – total assets (IMP / AST), equity (IMP / EQT) and goodwill (IMP / GDW) – all calculated at financial yearend before the effect of any goodwill impairment;
the return on equity ratio (ROE), measured at financial yearend, represents the company profitability in the period;
the company profitability is also measured by the average value of return on equity ratio (AVG.ROE), calculated with reference to the period from 2005 to 2008;
the corporate governance related variables are represented by the number of meetings of the audit committee held in year (A.C.MEET.) and the number of independent directors members of the audit committee scaled by total number of directors (A.C.IND.DIR.);
the firm size is given in the natural logarithm of total assets (SIZE).
The dummy variables refer to each year (from YEAR.06 to YEAR.08) and each sector (from SECT.A to SECT.E) considered in this study.
Consequently, we firstly examine a series of descriptive statistics arising from the financial statements reviewed, using appropriate tests of significance in order to obtain early feedback to our assumptions.
Subsequently, correlation analysis will be done through the construction of Pearson correlation matrices, checking positive correlations between corporate disclosure and goodwill impairment as well as earnings performance.
Finally, we test the fundamental assumptions with a multivariate analysis, using a series of multiple linear regressions that are introduced below.
Multiple linear regression – Model 1:
DISC = a + b_{1}IMP/AST + b_{2}ROE + b_{3}A.C.MEET. + b_{4}A.C.IND.DIR. + b_{5}SIZE + b_{6}YEAR.07 + b_{7}YEAR.08 + b_{8}SECT.A + b_{9}SECT.B + b_{10}SETC.C + b_{11}SECT.D + b_{12}COUNTRY + e
Multiple linear regression – Model 2:
DISC = a + b_{1}IMP/EQT + b_{2}ROE + b_{3}A.C.MEET. + b_{4}A.C.IND.DIR. + b_{5}SIZE + b_{6}YEAR.07 + b_{7}YEAR.08 + b_{8}SECT.A + b_{9}SECT.B + b_{10}SETC.C + b_{11}SECT.D + b_{12}COUNTRY + e
Multiple linear regression – Model 3:
DISC = a + b_{1}IMP/GDW + b_{2}ROE + b_{3}A.C.MEET. + b_{4}A.C.IND.DIR. + b_{5}SIZE + b_{6}YEAR.07 + b_{7}YEAR.08 + b_{8}SECT.A + b_{9}SECT.B + b_{10}SETC.C + b_{11}SECT.D + b_{12}COUNTRY + e
Multiple linear regression – Model 4:
DISC = a + b_{1}IMP/AST + b_{2}AVG.ROE + b_{3}A.C.MEET. + b_{4}A.C.IND.DIR. + b_{5}SIZE + b_{6}YEAR.07 + b_{7}YEAR.08 + b_{8}SECT.A + b_{9}SECT.B + b_{10}SETC.C + b_{11}SECT.D + b_{12}COUNTRY + e
Multiple linear regression – Model 5:
DISC = a + b_{1}IMP/EQT + b_{2}AVG.ROE + b_{3}A.C.MEET. + b_{4}A.C.IND.DIR. + b_{5}SIZE + b_{6}YEAR.07 + b_{7}YEAR.08 + b_{8}SECT.A + b_{9}SECT.B + b_{10}SETC.C + b_{11}SECT.D + b_{12}COUNTRY + e
Multiple linear regression – Model 6:
DISC = a + b_{1}IMP/GDW + b_{2}AVG.ROE + b_{3}A.C.MEET. + b_{4}A.C.IND.DIR. + b_{5}SIZE + b_{6}YEAR.07 + b_{7}YEAR.08 + b_{8}SECT.A + b_{9}SECT.B + b_{10}SETC.C + b_{11}SECT.D + b_{12}COUNTRY + e
The level of disclosure is the dependent variable regression of the six functions above, which differ in the choice of independent variables likely to express the extent of impairment of goodwill and the corporate profitability.
We will indicate below the results of further regressions, calculated after carrying alternatives substitutions of dummy variables not yet included to assure the accuracy of the issues, namely the dummy variable for the year 2006 (YEAR.06) and the dummy variable for the sector E (SECT.E).
Goodwill impairment is scaled to total assets in Model 1 and Model 4, equity in Model 2 and Model 5, goodwill in Model 3 and Model 6. In Models 1, 2 and 3 the economic performance is represented by the return on equity of the correspondent year, whilst in Models 4, 5 and 6 ROE it is calculated as an average value for the whole period from 2005 until 2008.
The coefficient on goodwill impairment (b_{1}) and the coefficient on corporate profitability (b_{2}) are expected to be positive (see, respectively, H_{1} and H_{2}) and greater (see H_{3}) with respect to the Italian cases.
Furthermore, it is expected that the coefficient on size (b_{5}) is positive, since usually the costs of disclosure decrease with increasing firm size (Hossain, Perera & Rahman, 1995).
Concerning the issue of corporate governance, with regard to the scope of our study we selected a variable related to the level of activity of the audit committee, as its role and its responsibility include monitoring the integrity of the financial statements of the company (Sierra Garcia, Ruiz Barbadillo & Orta Perez, 2012). A few previous studies provide empirical evidence of the positive association between the level of voluntary disclosure and the number of meetings held by the audit committee (Menon & Williams, 1994; Karamanou & Vafeas, 2005; Greco, 2010), so we predict a positive sign of the corresponding coefficient (b_{3}).
In a broader context, with reference to the association between disclosure and corporate governance variables, the findings emerging from prior related research are not the same.
In fact empirical studies, mainly in the context of voluntary disclosure, show controversial results, in particular regarding the impact of independent directors.
Some works noted that the proportion of independent directors affects the quality of mandatory disclosure (Chen & Jaggi, 2000) while other analyses found no association between the two variables (Forker, 1992). A recent line of research only considers “qualified” classes of independent directors, for instance represented by those directors defined as community influential board members (Michelon & Parbonetti, 2012), showing positive associations with the level of voluntary disclosure.
Hence, we similarly consider in our analysis just the independent directors with an “active role”, identified by membership of the audit committee, assuming a positive sign of the corresponding coefficient (b_{4}).
Finally, we expect the level of disclosure to be significantly affected by the year as well as the sector (Cooke, 1992). With regard to variables related to years 2006 to 2008, it is assumed that the intensity of association to the level of disclosure is chronologically increasing, indicating a positive trend of progressive assimilation, especially for Italian companies, of the innovative provisions of IAS 36.
Results
Descriptive Statistics and Correlation Analysis
As noted earlier, there are 110 observations meeting our sample criteria. Of these, as shown in Table 4, only 25% record goodwill impairment. Given our sampleselection criteria (market to book value below one or below median), the relatively low frequency of impairment suggests that IAS 36 is not so effective in generating timely writeoffs.

Italian cases 
British cases 
Total 

n 
% 
n 
% 
n 
% 

No goodwill impairment 
46 
78% 
37 
73% 
83 
75% 
Goodwill impairment 
13 
22% 
14 
27% 
27 
25% 
Total 
59 
100% 
51 
100% 
110 
100% 
The chisquare statistic for the table has a pvalue of 0.510 
Table 4. Frequency of Goodwill Impairments
The frequency of goodwill nonimpairment in the Italian sample (59 cases) is 78%, while in the British sample (51 cases) the frequency is 73%. The chisquare statistic for the comparison of impairment frequency across firm country is not statistically significant (pvalue of 0.510).
Table 5 presents some descriptive statistics for all the variables, with reference to the Italian and British companies.
In Panel 5A we report on the results among sampled Italian cases, equal to 59. The mean value of disclosure score is 5.5, with an oscillation between the minimum possible score (equal to 0) to the maximum possible score (equal to 10).
The average amount of goodwill writeoffs, in terms of total assets, is 0.2%, reaching a maximum value of 9%. In relation to equity, the mean value of goodwill impairment is 1.3% and records very high peaks, at 66.7% of equity. Moreover, the mean value of goodwill writeoffs related to book value of goodwill is 0.6%, with a relevant maximum value equal to 14.9%.
Subsequent Panel 5B reproduces the same descriptive statistics for the British cases, equal to 51. The mean value of disclosure score is 5.0; the corresponding median value is equal to 5 and the maximum score reaches 8.
The average amount of goodwill impairment, scaled by total assets, is 0.9%, reaching a maximum value of 14.4%. In terms of equity, the mean value of goodwill impairment is 1.8%, with a maximum value equal to 26%. Moreover, the median value of goodwill writeoffs related to book value of goodwill is 7.0%: it is noticeable that the maximum value is 100%.

Panel 5A: Italian Companies
Variable
Mean
Median
Minimum
Maximum
Standard deviation
DISC
5.475
6.000
0.000
10.000
2.615
IMP / AST
0.002
0.000
0.000
0.090
0.012
IMP / EQT
0.013
0.000
0.000
0.667
0.087
IMP / GDW
0.006
0.000
0.000
0.149
0.022
ROE
0.059
0.070
 0.430
0.260
0.115
AVG.ROE
 0.033
0.071
 5.274
0.208
0.700
A.C.MEET.
3.373
3.000
0.000
10.000
2.399
A.C.IND.DIR.
0.221
0.222
0.000
0.429
0.118
SIZE
13.427
12.990
10.837
16.323
1.578
n = 59
Panel 5B: British Companies 

Variable 
Mean 
Median 
Minimum 
Maximum 
Standard deviation 

DISC 
4.980 
5.000 
0.000 
8.000 
2.549 

IMP / AST 
0.009 
0.000 
0.000 
0.144 
0.030 

IMP / EQT 
0.018 
0.000 
0.000 
0.260 
0.056 

IMP / GDW 
0.070 
0.000 
0.000 
1.000 
0.211 

ROE 
0.091 
0.099 
 0.181 
0.472 
0.107 

AVG.ROE 
0.163 
0.109 
 0.500 
1.063 
0.270 

A.C.MEET. 
3.647 
4.000 
2.000 
6.000 
1.146 

A.C.IND.DIR. 
0.406 
0.400 
0.200 
0.625 
0.101 

SIZE 
12.716 
12.948 
8.645 
15.550 
1.764 

n = 51 

Panel 5C: Italian and British Companies 

Variable 
Italian cases (n = 59) 
British cases (n = 51) 
Difference in mean value 
tstatistic ^{a} 


Mean 
Mean 



DISC 
5.475 
4.980 
0.495 
1.002 

IMP / AST 
0.002 
0.009 
 0.007 
1.619 

IMP / EQT 
0.013 
0.018 
 0.005 
0.398 

IMP / GDW 
0.006 
0.070 
 0.064 
**2.152 

ROE 
0.059 
0.091 
 0.032 
1.482 

AVG.ROE 
 0.033 
0.163 
 0.196 
1.985 

A.C.MEET. 
3.373 
3.647 
 0.274 
0.781 

A.C.IND.DIR. 
0.221 
0.406 
 0.185 
***8.854 

SIZE 
13.427 
12.716 
0.711 
**2.211 

^{a }Significantly different at the 0.10 level (*), 0.05 level (**) or the 0.01 level (***). 
Table 5. Descriptive Statistics
Panel 5C provides a comparison between the mean values of each variable, with reference, respectively, to the Italian and British cases.
In particular:
the mean score of disclosure among the Italian sample (equal to 5.5) is statistically indistinguishable from the correspondent mean value among the British sample (equal to 5.0);
the mean values of goodwill writeoffs scaled by goodwill and of number of independent directors members of the audit committee scaled by total number of directors, are significantly greater for the British cases;
the mean value of firm size is significantly greater for the Italian cases.
Of particular interest, the difference in terms of independent directors proportion is highly significant, as the percentage equal to 22.1% for the Italian cases rises to 40.6% with reference to the British observations.
Table 6 reports on the association between goodwill impairment and disclosure.
Panel 6A: Italian companies – Total cases 

Variable 
Cases without goodwill impairment (n = 46) 
Cases with goodwill impairment (n = 13) 
Difference in mean value 
tstatistic ^{a} 

Mean 
Mean 


DISC 
5.174 
6.538 
 1.364 
2.457 *** 
Panel 6B: British companies – Total cases 

Variable 
Cases without goodwill impairment (n = 37) 
Cases with goodwill impairment (n = 14) 
Difference in mean value 
tstatistic ^{a} 

Mean 
Mean 


DISC 
5.297 
4.143 
1.154 
1.364 
Panel 6C: Italian and British companies – Cases without goodwill impairment 

Variable 
Italian companies (n = 46) 
British companies (n = 37) 
Difference in mean value 
tstatistic ^{a} 

Mean 
Mean 


DISC 
5.174 
5.297 
 0.123 
0.215 
Panel 6D: Italian and British companies – Cases with goodwill impairment 

Variable 
Italian companies (n = 13) 
British companies (n = 14) 
Difference in mean value 
tstatistic ^{a} 

Mean 
Mean 


DISC 
6.538 
4.143 
2.395 
2.874 *** 
^{a}^{ }Significantly different at the 0.10 level (*), 0.05 level (**) or the 0.01 level (***). 
Table 6. Goodwill Impairment and Disclosure
Panel 6A shows that the mean score of disclosure for the Italian sampled firms is significantly greater (at the 0.01 level) with reference to the cases with goodwill impairment: this first finding is consistent with the positive assumed association between goodwill writeoffs and level of disclosure.
In contrast, as shown in Panel 6B, the level of disclosure among the British cases with goodwill writeoffs is statistically indistinguishable from the mean value of nonimpairing observations.
Moreover, Panel 6C reports that the difference in mean value of Italian and British disclosure score is not significant in nonimpairing cases, while Panel 6D shows that the difference in mean score of disclosure among the Italian cases with goodwill writeoffs is significantly (at the 0.01 level) greater than the mean value among the British impairment observations.
Table 7 provides Pearson correlation coefficients between variables for the Italian and British cases.
Pearson correlation coefficients between variables 

Variable 
Disc 
Imp / Ast 
Imp / Eqt 
Imp / Gdw 
Roe 
Avg. Roe 
A.C. Meet. 
A.C.Ind.Dir. 
Size 
DISC 
1.00 








IMP / AST 
0.03 
1.00 







IMP / EQT 
0.10 
0.77 
1.00 






IMP / GDW 
0.00 
0.66 
0.43 
1.00 





ROE 
0.12 
0.07 
0.03 
0.13 
1.00 




AVG.ROE 
0.15 
0.06 
0.07 
0.05 
0.25 
1.00 



A.C.MEET. 
0.07 
0.17 
0.20 
0.05 
0.08 
0.02 
1.00 


A.C.IND.DIR. 
0.14 
0.22 
0.15 
0.28 
0.00 
0.07 
0.26 
1.00 

SIZE 
0.35 
0.21 
0.17 
0.13 
0.14 
0.09 
0.18 
0.31 
1.00 
YEAR.06 
0.12 
0.05 
0.10 
0.09 
0.06 
0.13 
0.13 
0.05 
0.03 
YEAR.07 
0.09 
0.14 
0.13 
0.15 
0.22 
0.05 
0.07 
0.03 
0.03 
YEAR.08 
0.20 
0.09 
0.03 
0.23 
0.26 
0.07 
0.19 
0.02 
0.05 
SECT.A 
0.08 
0.30 
0.28 
0.11 
0.10 
0.12 
0.18 
0.28 
0.41 
SECT.B 
0.22 
0.12 
0.10 
0.12 
0.02 
0.06 
0.12 
0.02 
0.22 
SECT.C 
0.30 
0.08 
0.07 
0.07 
0.13 
0.30 
0.07 
0.05 
0.37 
SECT.D 
0.15 
0.12 
0.10 
0.08 
0.13 
0.01 
0.20 
0.46 
0.26 
SECT.E 
0.21 
0.02 
0.04 
0.16 
0.04 
0.02 
0.10 
0.25 
0.11 
Country 
0.10 
0.16 
0.04 
0.22 
0.14 
0.18 
0.07 
0.64 
0.21 
Pearson correlation coefficients between variables 

Variable 
Year. 06 
Year. 07 
Year. 08 
SECT.A 
SECT. B 
SECT. C 
SECT. D 
SECT. E 
Country 
YEAR.06 
1.00 








YEAR.07 
0.44 
1.00 







YEAR.08 
0.52 
0.54 
1.00 






SECT.A 
0.01 
0.00 
0.01 
1.00 





SECT.B 
0.02 
0.09 
0.07 
0.30 
1.00 




SECT.C 
0.05 
0.03 
0.02 
0.20 
0.17 
1.00 



SECT.D 
0.01 
0.02 
0.00 
0.37 
0.31 
0.21 
1.00 


SECT.E 
0.06 
0.09 
0.03 
0.26 
0.21 
0.14 
0.27 
1.00 

Country 
0.05 
0.05 
0.00 
0.15 
0.05 
0.13 
0.10 
0.11 
1.00 
Table 7. Univariate Correlations – Italian and British Companies
Two variables related to goodwill impairment are correlated with the disclosure score (with Pearson coefficients equal to 0.03 and 0.10 respectively for IMP/AST and IMP/EQT). Both profitability ratios are positively correlated with the disclosure score (with coefficients equal to 0.12 and 0.15 respectively for ROE and AVG.ROE).
We also note positive coefficients related to firm size and to the dummy variable referred to the year 2008. The results referred to the variables concerning the quality of corporate governance are controversial, as the first correlation is positive (A.C.MEET equal to 0.07), whilst the second is negative (A.C.IND.DIR. equal to 0.14).
Table 8 provides Pearson correlation coefficients between variables for the entire Italian sample.
Pearson Correlation Coefficients Between Variables 

Variable 
Disc 
Imp / Ast 
Imp / Eqt 
Imp / Gdw 
Roe 
Avg. Roe 
A.C. Meet. 
A.C.Ind.Dir. 
Size 
DISC 
1.00 








IMP / AST 
0.15 
1.00 







IMP / EQT 
0.14 
0.99 
1.00 






IMP / GDW 
0.14 
0.91 
0.88 
1.00 





ROE 
0.29 
0.09 
0.07 
0.08 
1.00 




AVG.ROE 
0.23 
0.04 
0.04 
0.06 
0.31 
1.00 



A.C.MEET. 
0.02 
0.22 
0.21 
0.32 
0.09 
0.00 
1.00 


A.C.IND.DIR. 
0.27 
0.20 
0.20 
0.19 
0.15 
0.03 
0.39 
1.00 

SIZE 
0.38 
0.12 
0.12 
0.13 
0.12 
0.02 
0.06 
0.48 
1.00 
YEAR.06 
0.08 
0.16 
0.18 
0.11 
0.01 
0.19 
0.17 
0.04 
0.02 
YEAR.07 
0.16 
0.11 
0.09 
0.18 
0.18 
0.08 
0.08 
0.05 
0.00 
YEAR.08 
0.23 
0.04 
0.08 
0.06 
0.15 
0.11 
0.24 
0.01 
0.02 
SECT.A 
0.16 
0.30 
0.28 
0.24 
0.01 
0.07 
0.20 
0.39 
0.52 
SECT.B 
0.12 
0.09 
0.08 
0.07 
0.04 
0.09 
0.20 
0.29 
0.37 
SECT.C 
0.29 
0.07 
0.06 
0.11 
0.17 
0.33 
0.08 
0.07 
0.36 
SECT.D 
0.15 
0.10 
0.10 
0.03 
0.26 
0.11 
0.20 
0.67 
0.53 
SECT.E 
0.49 
0.05 
0.05 
0.04 
0.27 
0.01 
0.21 
0.19 
0.02 
Pearson Correlation Coefficients Between Variables 

Variable 
Year. 06 
Year. 07 
Year. 08 
SECT.A 
SECT.B 
SECT.C 
SECT.D 
SECT.E 

YEAR.06 
1.00 








YEAR.07 
0.44 
1.00 







YEAR.08 
0.55 
0.51 
1.00 






SECT.A 
0.08 
0.05 
0.03 
1.00 





SECT.B 
0.02 
0.07 
0.08 
0.27 
1.00 




SECT.C 
0.04 
0.03 
0.01 
0.20 
0.21 
1.00 



SECT.D 
0.07 
0.04 
0.03 
0.35 
0.37 
0.27 
1.00 


SECT.E 
0.03 
0.11 
0.08 
0.19 
0.20 
0.15 
0.25 
1.00 
Table 8. Univariate Correlations – Italian Companies
Of particular interest, all the three variables related to goodwill impairment exhibit a positive correlation with the level of disclosure (equal to 0.15, 0.14 and 0.14 respectively for IMP/AST, IMP/EQT and IMP/GDW). As predicted, this provides a preliminary indication on a univariate basis that the magnitude of goodwill writeoffs is positively associated with the degree of firm disclosure on impairment test.
Moreover, both return on equity ratios are positively correlated with the disclosure score (with Pearson coefficients equal to 0.29 and 0.23 respectively for ROE and AVG.ROE). Also this finding is consistent with our hypothesis.
The level of disclosure is also positively correlated with our proxy for firm size, suggesting that the costs of disclosure decrease for larger firms, and with the dummy variable referred to the year 2008, confirming the predicted improvement in corporate disclosure before the first years of IAS 36 application.
Surprisingly, both the corporate governance related variables exhibit negative coefficients.
Pearson correlation coefficients between variables for the whole British sample are shown in Table 9.
Pearson Correlation Coefficients Between Variables 

Variable 
Disc 
Imp / Ast 
Imp / Eqt 
Imp / Gdw 
Roe 
Avg. Roe 
A.C. Meet. 
A.C.Ind.Dir. 
Size 
DISC 
1.00 








IMP / AST 
0.01 
1.00 







IMP / EQT 
0.04 
0.94 
1.00 






IMP / GDW 
0.01 
0.66 
0.68 
1.00 





ROE 
0.07 
0.20 
0.26 
0.26 
1.00 




AVG.ROE 
0.01 
0.05 
0.20 
0.03 
0.04 
1.00 



A.C.MEET. 
0.36 
0.22 
0.15 
0.02 
0.09 
0.22 
1.00 


A.C.IND.DIR. 
0.12 
0.15 
0.07 
0.27 
0.08 
0.18 
0.03 
1.00 

SIZE 
0.30 
0.23 
0.25 
0.10 
0.24 
0.18 
0.55 
0.07 
1.00 
YEAR.06 
0.17 
0.02 
0.04 
0.14 
0.17 
0.07 
0.04 
0.00 
0.01 
YEAR.07 
0.01 
0.19 
0.21 
0.22 
0.26 
0.03 
0.07 
0.09 
0.08 
YEAR.08 
0.17 
0.16 
0.24 
0.33 
0.40 
0.03 
0.11 
0.09 
0.08 
SECT.A 
0.02 
0.30 
0.30 
0.09 
0.27 
0.22 
0.15 
0.09 
0.27 
SECT.B 
0.33 
0.15 
0.15 
0.16 
0.01 
0.04 
0.10 
0.25 
0.09 
SECT.C 
0.30 
0.07 
0.07 
0.07 
0.01 
0.07 
0.08 
0.26 
0.38 
SECT.D 
0.56 
0.12 
0.11 
0.10 
0.00 
0.26 
0.19 
0.30 
0.08 
SECT.E 
0.08 
0.04 
0.05 
0.19 
0.32 
0.02 
0.11 
0.29 
0.27 
Pearson Correlation Coefficients Between Variables 

Variable 
Year. 06 
Year. 07 
Year. 08 
SECT.A 
SECT.B 
SECT.C 
SECT. D 
SECT. E 

YEAR.06 
1.00 








YEAR.07 
0.43 
1.00 







YEAR.08 
0.49 
0.57 
1.00 






SECT.A 
0.12 
0.06 
0.06 
1.00 





SECT.B 
0.06 
0.11 
0.05 
0.33 
1.00 




SECT.C 
0.03 
0.00 
0.03 
0.18 
0.12 
1.00 



SECT.D 
0.13 
0.10 
0.03 
0.39 
0.26 
0.14 
1.00 


SECT.E 
0.08 
0.07 
0.01 
0.35 
0.23 
0.12 
0.27 
1.00 
Table 9. Univariate Correlations – British Companies
Contrary to the Italian findings, the coefficients report no significant relation between goodwill impairment ratios and the level of disclosure (with Pearson coefficients equal to 0.01, 0.04 and 0.01 respectively for IMP/AST, IMP/EQT and IMP/GDW).
Furthermore, both return on equity ratios are not significantly correlated with the disclosure score (with Pearson coefficient equal to  0.07 for ROE and 0.01 for AVG.ROE).
However, similarly to the Italian results, the level of disclosure is positively correlated with firm size and with the dummy variable referred to the year 2008.
The two corporate governance variables show positive coefficients, in contrast with the Italian cases.
As discussed earlier, we conducted regression tests by including only one proxy at a time for goodwill impairment and profitability, also in order to mitigate possible effects of multicollinearity (Chen & Jaggi, 2000). In Tables 7, 8 and 9 it has been observed, as a result, that the highest simple correlation between the independent variables considered in each regression model does not exceed 0.70, suggesting that multicollinearity is unlikely to pose a serious problem in the interpretation of the results of the multivariate analysis (Hossain & Hammami, 2009).
Regression Results
Table 10 reports on multivariate tests of the determinants of corporate disclosure in the Italian and British sample.
Dependent Variable: Disc 

Variable 
Model 1 
Model 2 
Model 3 

Coefficient 
tstatistic 
pvalue 
Coefficient 
tstatistic 
pvalue 
Coefficient 
tstatistic 
pvalue 

INTERCEPT 
**5.336 
2.138 
0.035 
**5.315 
2.168 
0.033 
*4.956 
1.979 
0.051 
IMP / AST 
*16.805 
1.746 
0.084 






IMP / EQT 



**7.037 
2.458 
0.016 



IMP / GDW 






1.672 
1.117 
0.267 
ROE 
**3.892 
1.995 
0.049 
*3.685 
1.916 
0.058 
**4.058 
2.055 
0.043 
AVG.ROE 









A.C.MEET. 
0.086 
0.714 
0.477 
0.110 
0.915 
0.362 
0.050 
0.413 
0.680 
A.C.IND.DIR. 
2.966 
1.314 
0.192 
3.111 
1.397 
0.166 
3.192 
1.386 
0.169 
SIZE 
***0.729 
4.517 
0.000 
***0.741 
4.662 
0.000 
***0.697 
4.312 
0.000 
YEAR.07 
0.430 
0.829 
0.409 
0.500 
0.975 
0.332 
0.340 
0.654 
0.515 
YEAR.08 
***1.541 
3.070 
0.003 
***1.617 
3.261 
0.002 
***1.427 
2.781 
0.007 
SECT.A 
***1.934 
2.785 
0.006 
***1.868 
2.728 
0.008 
***2.124 
3.042 
0.003 
SECT.B 
***3.013 
4.182 
0.000 
***3.006 
4.242 
0.000 
***3.062 
4.160 
0.000 
SECT.C 
***2.620 
3.072 
0.003 
***2.608 
3.102 
0.003 
***2.721 
3.142 
0.002 
SECT.D 
0.118 
0.164 
0.870 
0.156 
0.220 
0.826 
0.046 
0.064 
0.949 
Country 
0.399 
0.719 
0.474 
0.532 
0.971 
0.334 
0.414 
0.739 
0.462 
n 
110 
110 
110 

Adjusted R^{2} 
0.352 
0.371 
0.340 
Variable 
Model 4

Model 5 
Model 6 

Coefficient 
tstatistic 
pvalue 
Coefficient 
tstatistic 
pvalue 
Coefficient 
tstatistic 
pvalue 

INTERCEPT 
**5.388 
2.176 
0.032 
**5.371 
2.207 
0.030 
**5.022 
2.019 
0.046 
IMP / AST 
**15.985 
1.671 
0.098 






IMP / EQT 



**6.797 
2.385 
0.019 



IMP / GDW 






1.390 
0.937 
0.351 
ROE 









AVG.ROE 
**0.896 
2.317 
0.023 
**0.851 
2.228 
0.028 
**0.915 
2.343 
0.021 
A.C.MEET. 
0.083 
0.688 
0.493 
0.106 
0.887 
0.377 
0.049 
0.411 
0.682 
A.C.IND.DIR. 
2.703 
1.203 
0.232 
2.858 
1.290 
0.200 
2.867 
1.251 
0.214 
SIZE 
***0.752 
4.745 
0.000 
***0.762 
4.888 
0.000 
***0.722 
4.549 
0.000 
YEAR.07 
0.397 
0.768 
0.445 
0.467 
0.915 
0.363 
0.307 
0.593 
0.554 
YEAR.08 
**1.224 
2.475 
0.015 
***1.314 
2.684 
0.009 
**1.118 
2.205 
0.030 
SECT.A 
**1.798 
2.599 
0.011 
**1.740 
2.550 
0.012 
***1.969 
2.828 
0.006 
SECT.B 
***3.019 
4.223 
0.000 
***3.014 
4.283 
0.000 
***3.051 
4.170 
0.000 
SECT.C 
***2.856 
3.308 
0.001 
***2.832 
3.328 
0.001 
***2.940 
3.359 
0.001 
SECT.D 
0.001 
0.001 
0.999 
0.044 
0.063 
0.950 
0.063 
0.088 
0.930 
Country 
0.367 
0.664 
0.508 
0.496 
0.909 
0.365 
0.386 
0.693 
0.490 
n 
110 
110 
110 

Adjusted R^{2} 
0.361 
0.379 
0.349 

* Significant At The 0.10 Level; ** Significant At The 0.05 Level; *** Significant At The 0.01 Level. 
Table 10. Multivariate Linear Regressions – Italian and British Companies
The analysis of the residuals, founded on the exam of histograms and normal probability plots, confirms in all the models that the Ordinary Least Squares regressions are the most suitable techniques for testing our hypotheses. In addition, we calculated the Variance Inflation Factor (VIF) score for each independent variable, in order to evaluate whether multicollinearity may be a cause of concern. VIF scores higher than 10 are commonly taken as an indication of multicollinearity: in our cases, the highest VIF obtained is only 2.715.
The specification for the six multivariate regressions (Model 1 to 6) is provided in Section 4.
Model 1 and Model 4 scale goodwill impairment by total assets; the performance indicators are, respectively, ROE and AVG.ROE.
The multiple linear regressions show that the association between goodwill impairment and the dependent variable DISC is significant at the 0.10 level in the predicted sense; the disclosure level is significantly and positively associated at the 0.05 level with the alternative profitability ratios. The model adjusted R2 are equal to about 0.35.
Model 2 and Model 5 scale goodwill impairment by total assets. As expected, the multiple linear regressions shows that the dependent variable DISC is significantly and positively associated (at the 0.05 level) with goodwill impairment, with the model adjusted R2 equal to about 0.37. The positive association between DISC and performance indicators is significant respectively at the 90% and 95% confidence level.
Finally, when in Models 3 and 6 goodwill impairment is scaled by goodwill, the association between goodwill writeoffs and disclosure is not significant, whilst the alternative profitability ratios are significantly associated with disclosure score in the predicted sense. The model adjusted R2 is equal to about 0.34.
In summary, addressing the first research question of the positive association between goodwill writeoffs and disclosure (see H1), the findings are significant in the predicted sense in four models.
Looking at the research question of the positive association between earnings performance and corporate disclosure (see H2), in all six models ROE and AVG.ROE coefficients are statistically significant in the predicted direction.
In all six regressions the model adjusted R2 is higher than 0.3.
Furthermore, all six models show that the level of disclosure is highly significantly associated in the prediction sense with firm size and with the dummy variable referred to the year 2008, confirming the findings of the univariate tests. As predicted, also the coefficients of the dummy variables for each sector, only except for sector D, are highly significantly associated with corporate disclosure.
Other similar regression models, not presented in Table 10, show negative but insignificant coefficients associated with dummy variable YEAR.2006, and positive but insignificant coefficients on dummy variable SECT.E.
In all six models the coefficients referred to the quality of corporate governance are negative but statistically insignificant. Similarly, the dummy variable for the country exhibit coefficients always insignificant, showing that firm nationality does not influence the general level of disclosure about goodwill impairment test.
Next, we examine separately the two samples of Italian and British firms, in order to test if the results have been affected by inclusion of Italian company, having regard to our third research question (see H3).
Table 11 reports on multivariate tests of the determinants of corporate disclosure in the Italian sample.
In order to test whether relevant multicollinearity is affecting the results, we performed the Variance Inflation Factor for the Italian cases: the maximum VIF is equal to only 4.394, so confirming that multicollinearity among the predictor variables is not a problem. Moreover, the analysis of the residuals through the DurbinWatson test does not provide evidence of autocorrelation.
Model 1a scales goodwill impairment by total assets; the performance indicator is ROE.
As expected, the multiple linear regression shows that the dependent variable DISC is significantly and positively associated (at the 0.05 level) with goodwill impairment and (at the 0.10 level) with return on equity ratio. The model adjusted R2 is equal to 0.47.
Similarly, in Model 4a, when goodwill impairment is scaled by total assets again and the performance indicator is AVG.ROE, the level of disclosure is significantly associated in the predicted direction with goodwill impairment (at the 0.05 level) as well as return on equity ratio, with the model adjusted R2 equal to 0.50.
In the Italian sample, addressing the first research question of the positive association between goodwill writeoffs and disclosure (see H1), this finding of the valuerelevance of goodwill impairment is confirmed when alternative scalars are used. Actually, as can be seen from Table 11, when impairment goodwill is scaled by equity (Model 2a and Model 5a), the magnitude of goodwill writeoffs is positively associated with corporate disclosure at the 95% confidence level. Finally, in Models 3a and 6a, the level of disclosure is positively associated (respectively at the 90% and 95% confidence level) with goodwill impairment scaled by goodwill.
Dependent variable: DISC 

Variable 
Model 1a 
Model 2a 
Model 3a 

Coefficient 
tstatistic 
pvalue 
Coefficient 
tstatistic 
pvalue 
Coefficient 
tstatistic 
pvalue 

INTERCEPT 
*7.270 
1.968 
0.055 
*7.305 
1.972 
0.054 
**7.876 
2.117 
0.040 
IMP / AST 
**48.907 
2.127 
0.039 






IMP / EQT 



**6.510 
2.055 
0.045 



IMP / GDW 






*25.334 
1.986 
0.053 
ROE 
*4.233 
1.714 
0.093 
*4.393 
1.779 
0.082 
4.227* 
1.701 
0.096 
AVG.ROE 









A.C.MEET. 
0.010 
0.072 
0.943 
0.010 
0.071 
0.944 
0.022 
0.155 
0.878 
A.C.IND.DIR. 
4.600 
1.332 
0.189 
4.647 
1.341 
0.186 
4.610 
1.327 
0.191 
SIZE 
***0.766 
2.836 
0.007 
***0.771 
2.848 
0.007 
***0.819 
3.006 
0.004 
YEAR.07 
0.054 
0.081 
0.936 
0.033 
0.050 
0.960 
0.047 
0.069 
0.945 
YEAR.08 
*1.202 
1.887 
0.065 
*1.241 
1.931 
0.060 
*1.112 
1.752 
0.086 
SECT.A 
***3.046 
2.867 
0.006 
***3.068 
2.882 
0.006 
***3.231 
3.055 
0.004 
SECT.B 
***4.327 
4.198 
0.000 
***4.292 
4.152 
0.000 
***4.325 
4.173 
0.000 
SECT.C 
***3.643 
3.280 
0.002 
***3.596 
3.228 
0.002 
***3.617 
3.238 
0.002 
SECT.D 
1.682 
1.513 
0.137 
1.636 
1.466 
0.149 
1.519 
1.353 
0.183 
n 
59 
59 
59 

Adjusted R^{2} 
0.466 
0.463 
0.460 

Variable 
Model 4a 
Model 5a 
Model 6a 

Coefficient 
tstatistic 
pvalue 
coefficient 
tstatistic 
pvalue 
Coefficient 
tstatistic 
pvalue 

INTERCEPT 
*6.837 
1.915 
0.062 
*6.917 
1.927 
0.060 
**7.459 
2.075 
0.044 
IMP / AST 
**50.008 
2.259 
0.029 






IMP / EQT 



**6.509 
2.123 
0.039 



IMP / GDW 






**25.922 
2.112 
0.040 
ROE 









AVG.ROE 
**0.969 
2.539 
0.014 
**0.970 
2.526 
0.015 
**0.967 
2.516 
0.015 
A.C.MEET. 
0.017 
0.129 
0.898 
0.018 
0.138 
0.891 
0.015 
0.114 
0.910 
A.C.IND.DIR. 
4.426 
1.325 
0.191 
4.477 
1.332 
0.189 
4.437 
1.320 
0.193 
SIZE 
***0.732 
2.804 
0.007 
***0.741 
2.820 
0.007 
***0.787 
2.986 
0.004 
YEAR.07 
0.007 
0.011 
0.991 
0.028 
0.043 
0.966 
0.014 
0.022 
0.982 
YEAR.08 
0.851 
1.370 
0.177 
0.880 
1.399 
0.168 
0.760 
1.228 
0.226 
SECT.A 
***3.284 
3.298 
0.002 
***3.336 
3.339 
0.002 
***3.473 
3.510 
0.001 
SECT.B 
***4.618 
4.840 
0.000 
***4.606 
4.799 
0.000 
***4.616 
4.808 
0.000 
SECT.C 
***4.441 
3.999 
0.000 
***4.397 
3.935 
0.000 
***4.412 
3.947 
0.000 
SECT.D 
**2.200 
2.103 
0.041 
**2.174 
2.064 
0.045 
*2.032 
1.917 
0.061 
n 
59 
59 
59 

Adjusted R^{2} 
0.501 
0.495 
0.495 

* Significant At The 0.10 Level; ** Significant At The 0.05 Level; *** Significant At The 0.01 Level. 
Table 11. Multivariate Linear Regressions – Italian Companies
With reference to the second research question, concerning the positive association between earnings performance and corporate disclosure (see H2), in all six models ROE and AVG.ROE coefficients are statistically significant in the predicted direction, at the 90% (Models 1a, 2a and 3a) or 95% (Models 4a, 5a and 6a) confidence level.
In all six regressions the model adjusted R2 is higher than 0.45.
It should be noted that the similarity of the coefficients on both goodwill impairment and performance indicators give confidence in the robustness of the results across the different scalars.
Furthermore, all six models show that the level of disclosure is highly significantly associated in the prediction sense with firm size and with the dummy variables for each sector (only except for sector D).
Other similar regression models, not presented in Table 11, show negative but insignificant coefficients associated with dummy variable YEAR.2006, and negative coefficients on dummy variable SECT.E, statistically significant at the 0.05 level.
The coefficients related to the quality of corporate governance are not significant, while the positive coefficient referring to the dummy variable for the year 2008 is significant at the 90% level in the first three models.
Table 12 reports on multivariate tests of the determinants of corporate disclosure in the British sample.
Also in the British cases, the VIF scores calculated for each independent variable confirm that multicollinearity does not represent a problem, as the largest VIF score is equal to only 3.444. Similarly to the Italian results, the analysis of the residuals through the DurbinWatson test does not provide evidence of autocorrelation.
As can be seen, the multiple linear regressions shows that in all six models (Model 1b to Model 6b) the coefficients on goodwill impairment are statistically insignificant, with each model adjusted R2 around 0.42.
Furthermore, no relation is found between return on equity ratios and disclosure score: in all six models the coefficients are negative but not statistically significant.
However, all six regressions show that the level of disclosure is significantly associated in the prediction sense with the dummy variable for the year 2008, while the positive coefficients related to firm size are not significant. Looking at the dummy variables referred to firm sector, only SECT.D is significantly associated to score disclosure.
Other similar regression models, not presented in Table 12, show negative coefficients on dummy variable YEAR.2006, statistically significant at the 0.10 level, and positive coefficients associated with dummy variable SECT.E, significant at the 0.01 level.
Dependent variable: DISC 

Variable 
Model 1b 
Model 2b 
Model 3b 

Coefficient  tstatistic  pvalue  Coefficient  tstatistic  pvalue  Coefficient  tstatistic  pvalue  
INTERCEPT 
1.403 
0.374 
0.710 
1.017 
0.270 
0.788 
1.461 
0.410 
0.684 
IMP / AST 
0.206 
0.018 
0.985 






IMP / EQT 



1.708 
0.290 
0.773 



IMP / GDW 






0.359 
0.239 
0.812 
ROE 
1.210 
0.384 
0.703 
1.154 
0.366 
0.716 
1.336 
0.418 
0.678 
AVG.ROE 









A.C.MEET. 
0.311 
0.784 
0.438 
0.270 
0.715 
0.479 
0.314 
0.882 
0.383 
A.C.IND.DIR. 
1.304 
0.400 
0.691 
1.223 
0.374 
0.710 
1.184 
0.359 
0.721 
SIZE 
0.195 
0.674 
0.504 
0.228 
0.812 
0.422 
0.191 
0.728 
0.471 
YEAR.07 
1.219 
1.683 
0.100 
*1.240 
1.722 
0.093 
1.213* 
1.690 
0.099 
YEAR.08 
*1.464 
1.940 
0.060 
*1.447 
1.914 
0.063 
1.495* 
1.955 
0.058 
SECT.A 
0.168 
0.161 
0.873 
0.134 
0.129 
0.898 
0.206 
0.197 
0.845 
SECT.B 
1.400 
1.197 
0.239 
1.504 
1.288 
0.205 
1.343 
1.168 
0.250 
SECT.C 
2.232 
1.634 
0.110 
2.218 
1.626 
0.112 
2.177 
1.573 
0.124 
SECT.D 
***2.980 
2.974 
0.005 
***2.923 
2.909 
0.006 
***3.022 
3.006 
0.005 
n 
51 
51 
51 

Adjusted R^{2} 
0.418 
0.419 
0.419 

Variable 
Model 4b 
Model 5b 
Model 6b 

Coefficient 
tstatistic 
pvalue 
Coefficient 
tstatistic 
pvalue 
Coefficient 
tstatistic 
pvalue 

INTERCEPT 
1.558 
0.409 
0.685 
1.153 
0.305 
0.762 
1.623 
0.445 
0.658 
IMP / AST 
0.328 
0.029 
0.977 






IMP / EQT 



2.501 
0.414 
0.681 



IMP / GDW 






0.218 
0.147 
0.884 
ROE 









AVG.ROE 
0.465 
0.394 
0.696 
0.576 
0.479 
0.635 
0.450 
0.383 
0.704 
A.C.MEET. 
0.279 
0.681 
0.500 
0.220 
0.556 
0.581 
0.290 
0.792 
0.433 
A.C.IND.DIR. 
1.582 
0.466 
0.644 
1.559 
0.461 
0.647 
1.486 
0.431 
0.668 
SIZE 
0.189 
0.658 
0.515 
0.232 
0.827 
0.413 
0.181 
0.694 
0.492 
YEAR.07 
1.221 
1.687 
0.100 
*1.244 
1.731 
0.091 
*1.214 
1.691 
0.099 
YEAR.08 
**1.577 
2.230 
0.032 
**1.547 
2.183 
0.035 
**1.601 
2.203 
0.034 
SECT.A 
0.012 
0.012 
0.991 
0.030 
0.030 
0.976 
0.035 
0.035 
0.973 
SECT.B 
1.460 
1.262 
0.214 
1.583 
1.375 
0.177 
1.416 
1.252 
0.218 
SECT.C 
*2.332 
1.746 
0.089 
*2.304 
1.727 
0.092 
*2.307 
1.713 
0.095 
SECT.D 
***3.002 
2.961 
0.005 
***2.959 
2.932 
0.006 
***3.026 
2.986 
0.005 
n 
51 
51 
51 

Adjusted R^{2} 
0.418 
0.420 
0.418 

* Significant at the 0.10 level; ** significant at the 0.05 level; *** significant at the 0.01 level. 
Table 12. Multivariate Linear Regressions – British Companies
As in the Italian sample, the corporate governance related indicators exhibit insignificant associations.
Discussion
In a sample of firms with market indications of goodwill impairments, we find a relatively low frequency of writeoffs. Only 27 of the 110 observed cases (25% of the entire sample) record goodwill impairments, with a statistically insignificant difference between the Italian and the British cases.
Similarly, the level of disclosure on impairment test provided in the notes to the consolidated financial statements is relatively low, albeit the items identified to measure disclosure score are mandatory in accordance with the requirements of IAS 36.
Regarding the whole sample, with a maximum score that a company could achieve equal to 10, the mean value of the level of disclosure does not reach 6. There is no significant difference between 59 Italian cases (with a mean disclosure score equal to 5.5) and 51 British cases (with a mean disclosure score equal to 5.0).
However, as expected, the degree of disclosure tends to improve from 2006 to 2008, and it is influenced by firm sector.
Looking at the first research question (see H1), the results confirm that Italian companies provide more information about impairment test when goodwill is impaired, with the strength of the positive association between the level of disclosure and the magnitude of goodwill writeoffs across a variety of model specifications.
This finding suggests that the level of disclosure, as assumed, could represent a relevant indicator of the degree of reliability with which Italian companies have implemented IAS 36 requirements on goodwill impairment test.
In addition, to examine the second research question (see H2), the Italian data confirms, both in univariate and multivariate analysis, our hypothesis, finding a significant association between disclosure score and accounting performance in the predicted direction. In other words, Italian firms with lower return on equity ratio tend to provide less information about goodwill impairment test.
For British cases we assumed (see H3) a lower degree of positive association between the level of disclosure and the magnitude of goodwill impairments as well as the return on equity ratios.
Indeed, beyond expectation, the data of British sampled companies does not confirm that the level of disclosure is significantly associated with goodwill writeoffs or accounting performance.
In summary, it should be noted that the general level of disclosure about goodwill impairment test provided by British companies, although not more satisfactory respect to the Italian cases, seems to be smoother, as it is less affected by managers’ decision to take goodwill writeoffs and less influenced by corporate accounting performance.
Our results suggest that despite the mandatory nature of disclosure requirements, there are significant variations in disclosure score across firms. Overall, the degree of information about goodwill impairment, in both the Italian and the British cases, is not influenced by the quality of the corporate governance. Both the level of activity of the audit committee, measured by the number of meeting held in year, and the percentage of the independent directors with an active role, identified by membership of the audit committee, do not exhibit significant associations with corporate disclosure of impairment tests.
These findings are consistent with those studies that specifically noted that the proportion of directors who are independent does not affect the quality of mandatory (Forker, 1992) or voluntary (Ho & Wong, 2001; Brammer & Pavelin, 2006) disclosure.
Indeed, as discussed earlier, the results of previous research regarding the association between board independence and the quality of corporate information are heterogeneous: some works provide evidence of a positive association between the two variables (Chen & Jaggi, 2000), whilst others found a negative relationship (Eng & Mak, 2003).
Thus the question remains whether independent directors contribute to increase mandatory or voluntary disclosure or whether they are ineffective (García–Meca & Sánchez–Ballesta, 2010). Moreover, it should be noted that in practice it is rather difficult to classify independent directors as truly independent from management: as a consequence, some nominally independent directors may be valuable, while others are not (Di Pietra, Grambovas, Raonic & Riccaboni, 2008).
On the other hand, with regard to the object of our research, an alternative or more specific explanation could be that the significant estimates required in order to assess the recoverable amount of goodwill stress the importance of director’s inside perspective for improving the quality of control and disclosure. In particular, the certification of such information disclosure, also by the audit committee, necessarily requires some firmspecific expertise on the part of directors. Therefore, the independence of directors may reduce the ability to acquire firmspecific information inherent in impairment test assumptions (Biondi, Giannocolo & Reberioux, 2010).
Additional Analysis: The Last Years of IAS 36 Application
Despite the objective of our study is to test the initial impact of IAS 36 in the first years of its application, it seems interesting to extent the analysis of the goodwill disclosure with reference to the last years of IAS 36 application. These additional research also represents a robustness test with reference to our previous findings.
As a consequence, we examine the 2011 and 2012 financial statements of the same Italian and British companies included in our sample, as defined in Section 4, also in order to verify if the positive trend of progressive improvement of goodwill disclosure is confirmed.
Table 13 details the sample selection criteria for the years 2011 and 2012: our final sample consists of 82 financial statements.

Italian companies 
British companies 
Total 
Number of the same companies included in the final sample of our previous analysis 
27 
21 
48 
First number of consolidated financial statements 2011 – 2012 
54 
42 
96 
Less: 

Financial statements of companies not more listed 
 2 
 6 
 8 
Financial statements with zero goodwill value 
0 
 5 
 5 
Financial statements with negative book value of equity 
 1 
0 
 1 
Final sample 
51 
31 
82 
Table 13. Sample Selection Procedure for The Years 20112012
Of our 82 observations, as shown in Table 14, only 26% record goodwill impairment. These results are very similar to the percentage (25%) of nonimpairment described in Section 5 with reference to the first years (20062008) of IAS application.

Italian cases 
British cases 
Total 

n 
% 
n 
% 
n 
% 

No goodwill impairment 
38 
75% 
23 
74% 
61 
74% 
Goodwill impairment 
13 
25% 
8 
26% 
21 
26% 
Total 
51 
100% 
31 
100% 
82 
100% 
The chisquare statistic for the table has a pvalue of 0.975 
Table 14. Frequency of Goodwill Impairments
Also in this additional time period of analysis we note that the chisquare statistic for the comparison of impairment frequency across firm country is not statistically significant (pvalue of 0.975), as the frequency of goodwill nonimpairment in the Italian (51 cases) and British (31 cases) samples is almost the same (respectively, 25% and 26%).
Table 15, by analogy with Table 5, presents some descriptive statistics for all the variables, with reference to the Italian and British companies. We specify that now the variable AVG.ROE refers to the average ROE in the period 20112012.
As regards the disclosure quality, Panel 15A and Panel 15B provide evidence of improvement, confirming our hypothesis. The mean value of disclosure score is 6.9 in the Italian sample (equal to 5.5 in the period 20062008) and 6.1 with reference to the British observations (equal to 5.0 in the period 20062008). However, these results are still unsatisfactory, considering that all the ten items (Table 2) identified to measure the quality of corporate information about goodwill are mandatory.
Panel 15A: Italian Companies 

Variable 
Mean 
Median 
Minimum 
Maximum 
Standard deviation 

DISC 
6.922 
8.000 
0.000 
10.000 
2.407 

IMP / AST 
0.001 
0.000 
0.000 
0.025 
0.004 

IMP / EQT 
0.003 
0.000 
0.000 
0.280 
0.049 

IMP / GDW 
0.016 
0.000 
0.000 
0.280 
0.049 

ROE 
0.010 
0.035 
 1.202 
0.825 
0.320 

AVG.ROE 
0.013 
0.041 
 0.957 
0.448 
0.232 

A.C.MEET. 
4.157 
4.000 
0.000 
12.000 
2.595 

A.C.IND.DIR. 
0.234 
0.231 
0.000 
0.429 
0.099 

SIZE 
13.225 
12.813 
10.594 
16.258 
1.522 

n = 51 

Panel 15B: British Companies 

Variable 
Mean 
Median 
Minimum 
Maximum 
Standard deviation 

DISC 
6.129 
6.000 
0.000 
9.000 
1.928 

IMP / AST 
