The Difference between the Accounting Result and Taxable Income in Detecting Earnings Management and Tax Management : The Tunisian Case

In this study, we examine empirically whether the gap between the pre-tax income and the taxable income predicts simultaneously earnings management and tax management. Prior researches have begun to estimate several indictors of earnings quality. We extends this works by investigating whether the book-tax differences (BTD) provides information about managerial discretion. BTD can be used as an attribute of information quality and can contain incremental information for investors and users of financial reports of companies We compose a sample of 21 listed Tunisian firms over the 2003-2012 periods, we apply a statistical methodology implementing a linear panel regression. We develop a model in which we estimate abnormal BTD. All earnings and tax management activities are captures by our abnormal BTD indicator (ABTD). From this regression we are going to test the effect of the variables of the earnings and tax management on the ABTD variable. We find, as expected, a significant association between BTD and manager discretion proxies. Our results indicate that BTD contains a predictive power over the earnings and tax management. We find that Tunisian companies are concerned about tax reducing rather than improved information quality and increasing the financial result.


Introduction
In this study, we examine empirically whether the gap between the pre-tax income (hereafter called book income) and the taxable income predicts simultaneously earnings management (Note 1) (EM) and tax management (Note 2) (TM).Prior researches have begun to investigate the difference between the tax and accounting systems (Lamb et al., 1998;Hoogendoorn, 1996;Nobes & Schwencke, 2006).Considering the recent financial scandals, a large body of literature extends this works by investigating whether the book-tax differences (BTD) provides information about managerial discretion (Desai, 2002;Mills, 1998;Wilson, 2009;Tang & Firth, 2011;Hanlon, 2005).For example, a number of studies link BTD with earnings quality or earnings persistence (Hanlon, 2005;Blaylock et al., 2012), cost of capital (Dhaliwal et al., 2009;Goh et al., 2013), audit fees (Hanlon et al., 2012).Assessing the BTD is important because BTD can be used as an attribute of financial reporting quality and can contain incremental informations for investors and users of companies' financial reports.Recent research focus on either BTD are associated with earnings management (Joos et al., 2000;Mills & Newberry, 2001;Donohoe & McGill, 2010;Comprix et al., 2011) or tax planning (Wilson, 2009).Some studies attribute the BTD to both EM and TM (Tang & Firth, 2011;Chen et al., 2012).
We continue this recent innovation here and contribute further to the debate by analyzing the role of BTD in detecting the earnings manipulation in a Tunisian setting.We employ Tunisia as the setting for our study for two reasons.First, because Tunisia has a close relationship between financial and tax accounting, the level of financial information disclosure remains limited due to the pressures from the tax authorities (the most important user of financial reporting).While present literature is based on empirical data from Anglo-Saxon countries; the results from these studies must be applied in continental countries setting where the dominance of the public sector and the bad organization of financial market in the economy.Second, like most developing countries, Tunisia has highly family firms and characterized by less developed and emerging stock market.between earnings management proxy such as discretionary accruals and tax aggressiveness proxy such as effective tax rate (ETR).We assess the usefulness of BTD to detect both EM and TM.According to Tang and Firth (2011), BTD can be a powerful measure to detect manager's discretion.In order to identify the incremental information provided in BTD, we develop a model in which we decompose normal and abnormal components from total BTD.Normal BTD (NBTD) is designed to reflect differences between accounting and tax systems.All earnings and tax management activities are captured by our abnormal BTD indicator (ABTD).We estimate ABTD as the residual from the BTD regression model.The potential of BTD to inform about EM and TM has been examined previously.The closest to our study is Tang and Firth (2011), reporting a positive association between abnormal BTD and incentives for earnings and ax management.We introduce our measures of both EM and TM into established model of ABTD.First of all, we measure EM with accruals, real earnings management and changes in accounting methods.In addition, we evaluate the TM through tax evasion and tax fraud proxies.
We extend the contribution of this study in several directions.First, we focus on the degree of conformity between tax rules and accounting practices.According Dridi and Boubaker (2015), we select the major differences between GAAP and tax law of Tunisian setting.This difference will be used to measure NBTD due to relative proxies.Second, we focus on managerial discretion, including more reliable proxies for earnings and tax management activities that may increase the gap between book income and taxable income.
We examine a sample of Tunisian listed firms from the years 2003 to 2012, constituting 210 firm-years.Our two main set of tests examines firstly whether BTD can contain information about earnings quality and secondly whether companies based on a continental accounting system (where tax rules are closely related to GAAP) have also manager's discretion in order to improve the information quality.We find that, as expected, a significant association between BTD and manager discretion proxies.Our results indicate that BTD contains a predictive power over the earnings and tax management.We find that Tunisian companies are concerned about tax reducing rather than improved information quality and increasing the financial result.
The remainder of this study is organized as follows.The second section outlines the literature and presents our hypotheses.The two next sections describe our method and expose our results, the final section concludes.

The Book-Tax Difference Debate
Changes in accounting and taxation are synchronized with the proliferation of dispersed ownership structures and stock firms.Reliable information becomes useful not only for shareholders but also to investors and outsiders.The application of accounting standards and tax rules involves considerable judgment and the use of GAAP and tax laws provide managers substantial discretion.Paradoxically, reliable information that objectively represents firm performance is generated by managers with a discretionary area which require a subjective behavior.Managers have a diversified portfolio of accounting choices and tax options to, legally, manipulate information reported to outsiders.The book-tax literature has mostly focused on aggressiveness in reporting book and tax incomes.The reason for the divergence between book income and taxable income are not fully due to conflicting goals of financial statements and fiscal reports.Reporting discretion may be reflected in the BTD.Previous research has suggested that BTDs are incrementally useful in detecting aggressive financial reporting.BTD can reveal earnings management in accounts such as depreciation, inventory allowance, warranty reserves and bad debts.
The literature on managerial discretion in financial and tax reporting is beginning with Watts and Zimmerman (1986).Since this study, several researches have tried to find the best proxy of this discretion.Hanlon (2005) finds evidence that large temporary BTD provide information about the persistence of earnings and concludes that investors interpret large BTDs as a "red flag" indicating lower earnings persistence.Tang and Firth (2012) find that regulatory and opportunistic sources of BTDs decrease the earnings persistence.Blaylock et al (2012) extend Hanlon's finding by looking through to the source of BTD.They provide evidence with large positive BTD being associated with earnings management.They establish a model of earnings persistence and interacted current earnings with tax avoidance and earnings management proxies.They find that large positive BTD capture discretion in the accrual process, but find tax shelters have no relation with the persistence of pre-tax earnings.Lev and Nissim (2004) show that BTD predict future earnings growth and Weber (2009) report evidence that BTD can explain analysts' forecast errors.Guenther (2011) conduct a similar analysis to investigate the results of Hanlon (2005) by using "data snooping" in order to examine "influential observations.He tries to investigate if the link between lower earnings persistence and large BTDs can be hold when he add several set of control variables (such as age of the firm, large Transitory items, large accruals, and high levels of pre-tax return on assets).He concludes that this relation is impacted by those factors.By investigating the association between temporary BTD and firm life cycle, Drake (2012) find also that life cycle capture the relation between BTD and earnings persistence and concludes that this relation is more complicated.Mills and Newberry (2001) focus on the relation between BTD and managers 'incentives to overstate earnings.Phillips et al. (2003) examine the association between BTD and various measures of earnings management.
Academic research has also claimed that BTD contain information about tax avoidance (Manzon & Plesko, 2002;Desai, 2003;Plesko, 2004;Desai & Dharmapala, 2006).A report issued by U.S. Treasury (1999) recognized the increasing in BTD as substantiation of the raise in tax shelters activities.This assertion gives rise to the usefulness of BTD as measurement of tax sheltering.Wilson (2009) investigates the characteristics of firms that can be engaged in tax sheltering activity.He provides evidence consistent with tax shelter participation positively associated with large book-tax differences.Frank et al. (2009) highlights that in the US which have a book-tax nonconformity managers may boost financial reporting and minimize income reported to tax authorities far away the book-tax trade-offs.He finds that tax reporting aggressiveness and financial reporting aggressiveness are strongly and positively associated.Similarly, Manzon and Plesko (2002) Desai andDharmapala (2006, 2009) use BTD as a proxy for corporate tax avoidance.
Corporate tax management can't be treated detached from earnings management.In fact, manager may have incentives to manipulate the two reports in order to avoid costs that may be incurred if he overlooks tax administrators or outside investors in the firm.Desai andDharmapala (2006, 2009) report that BTD is unable to reflect tax aggressiveness itself.Managers can be engaged in both earnings management and tax aggressiveness (Frank et al., 2009;Tang & Firth, 2011;Tang & Firth, 2012).
Considering the above, we propose to verify the following hypotheses.
Hypothesis 1: Earnings management by manager explains the difference between the accounting result and taxable income (BTD) presented by the listed companies.
Hypothesis 2: Tax management by manager explains the difference between the accounting result and taxable income (BTD) presented by the listed companies.

Book-Tax Difference in Tunisian Setting
BTD in Tunisia has been a long history since the disconnection of the accounting and the tax systems maintained even before the creation of the accounting system in 1997 (Dridi & Boubaker, 2015).However, the usefulness of BTD in detecting managerial manipulations (mainly the earning management) should be established in Tunisia especially after the increase of listed companies (24 firms introduced in the last 5 years which shows that the market is in emergence).
The Tunisian context is characterized by the concentrated bank-dominated system and the largest owner shareholder (Ben, Amar, & Abaoub, 2010).Despite attempts to companies' privatization from the 90 th , the size of public firms remains important.The Tunisian sector is dominated by micro, small and medium firms (83% of a total of 84,500 firms are micro companies).The Tunisian capital market can be characterized by a concentrated ownership structure.Until 2009, the stock exchange of Tunis has only 52 companies of which 23 companies are financial institutions (a number that reflects the mentality of Tunisian investors to funding from financial institutions).Most of shareholders represent important monitoring blocks (Kouki & Guisani, 2009).Tunisia creates its Stock Exchange in 1969.The Tunisian capital market role in financing the economy was insignificant due to the dominance of financial institutions and the State in the economy.The first half of the 1980s there has been a deterioration of the Tunisian economy with a historic slump in the real production (-1.86%) which caused the rise of privatization and the will of the State to renounce its role as an economic planner.These events have made many changes.
Since 1997, the accounting system has become an autonomous legal division.It applies to any person or entity in the private sector, which by law must keep accounting records.The Tunisian corporate accounting system is based on an Anglo-Saxon vision that meets the evolving needs of the Tunisian economy based on information transparency and opening market (Note 3).Benchmarking the Anglo-Saxon logic is characterized by the development of technical and specific accounting standards developed through a framework requiring a clear separation between accounting (which should be closer to economic and financial reality) and taxation (which requires legal formalism).The aim of this separation is the disclosure of high information quality to investors.Like many accounting systems in several countries, the Tunisian's system provides to manager a variety of accounting choice and operations structuring.The manager can use his discretionary latitude to report a high income quality.Evidence suggests that Tunisian managers opportunistically use the discretion in accounting standards.For example, Ben Amar and Abaoub (2010) provide evidence that Tunisian firms managed earnings to earnings decreases and to avoid losses.
Tunisian corporate taxation has developed with close relation to financial reporting.Tunisian tax system is a declarative system.The taxable income is resulting from the accounting income after making adjustments to comply with the tax law.Unlike the Anglo-Saxon countries, the use of these adjustments will not result in the production of two separate reports (tax statements and financial reports).The fiscal balance sheet is only a table which includes the integrations and deductions of some items forced by tax law to calculate the taxable income.The model of this table (prepared by the tax administration) takes the net income after accountings changes as base for calculating taxable income.In fact, the tax law requires taxpayers to calculate their taxable income on the basis of reporting income given by GAAP.Tunisian tax law requires only one tax rate of 30% of taxable income.Tunisian legislation give companies a number of tax options and tax benefits (through the investment incentives code by law n°93-120 of 27 December 1993) such as the income reinvestment abatement.These incentives give to the manager the opportunity of tax optimization.

Sample Selection and Data Source
Our sample originally consisted of 65 listed Tunisian firms over the 2003-2012 period.However, the final sample used for empirical testing consisted of 21 firms (210 firm-years) after excluding financial firms as well as firms with missing data.We chose the balanced panel to increase the number of observations since most companies have been newly introduced to the stock market.The choice of listed firms is justified by the availability of data.Financial data are collected from the financial statement available on the website of the Financial Market Council.

Regression Model Specifications
To verify our research hypotheses we apply a statistical methodology implementing a linear panel regression.From this regression we are going to test the effect of the variables of the EM and TM as well as the control variables on the ABTD variable.Our mode is as follows: Where: ABTD i,t : abnormal book-tax differences of firm i in year t; DA i,t : discretionary accruals of firm i in year t; RM_PROXY i,t : real activities manipulation of firm i in year t; INV i,t : inventory, this variable is measured by a binary value which takes 1 if the firm adopts the weighted average cost method and 0 otherwise; inventory of firm, DEP i,t : depreciation of the firm, this variable is measured by a binary value which takes 1 if the firm apply both accelerate and straight-line depreciation method and 0 otherwise; REINV i,t : reinvestment of profits, this variable is measured by a binary value which takes 1 if the firm practice the reinvestment of earnings and 0 otherwise ; ETR i,t : effective tax rate of firm i in year t ; SIZE i,t : natural logarithm of total assets of firm i in year t ; ROA i,t : profitability of the firm i in year t, this variable is measured by a binary value which takes 1 if the firm has a positive profit before tax and 0 otherwise.

Dependent Variable Measurement-Abnormal Book-Tax Differences
Prior studies have used different measures of BTD.Some researchers separated BTD into two components; permanent and temporary differences.The reason is that temporary differences are more informative about earnings manipulations because it represents the differences between accounting and tax treatment.Temporary differences capture items (income or expense) that are recognized at different times for book and tax purposes.Permanent differences include items that are accepted by accounting (taxation) but that are never taken into consideration by taxation (accounting).Therefore, temporary differences (generally identified by the deferred tax expense) are the essential component of the BTD which can provide information about earnings quality (Philips et al., 2003;Mills, 1998;Hanlon, 2005;Hanlon & Krishnan, 2006;Guenther, 2011).Further studies have used the total differences as a measure of BTD (Mills & Newberry, 2001;Lev & Nissim, 2004;Weber, 2009;Dhaliwal et al., 2009;Manzon & Plesko, 2002) (Note 4).Some other studies have analyzed the DCF through its decompositions into normal and abnormal differences (Tang & Firth, 2011;Tang & Firth, 2012;Formigoni et al., 2009).These studies have referred to Desai and Dharmapala (2006).The idea is to use the residual of a regression of the total BTD on the normal BTD.This approach consists of measuring by the residue the portion which is not observed in the difference between accounting and tax treatment.In this work, we choose this measure of BTD for three reasons.First, temporary differences are identified by the deferred tax expense.Unlike IFRS treatment, the deferred tax in Tunisia is totally ignored in practice, lack of specific standard in this issue.
Second, the measure of BTD through the total difference is not relevant since this difference can only results from a simple disconnection between accounting and tax systems.In fact, Dridi and Boubaker (2015) found a total disconnection between these two disciplines.Third, the residual approach allows taking the total difference and subtracting the different items that may increase the gap of BTD.We divide the BTD into two sets: a set of variables represents the difference treatments between accounting and taxation.These variables reflect the normal difference which is not related to managerial manipulations.The remaining gap (residual) is the abnormal difference part which is caused by earnings aggressiveness (Eq.2).The second set of variables represents the proxies of EM and TM.In fact, the residual which measures ABTD is regressed on variables related to earnings manipulations (Eq.1).
Where: BTD i,t : book-tax differences of firm i in year t, this variable is equal to income before tax minus taxable income scaled by total assets; GW i,t : goodwill of firm i in year t, this variable is equal to the gross value of goodwill scaled by total assets; INV i,t : change in investment in gross property, plants and equipment from year t-1 to year t, this variable is equal to the gross value of tangible assets in year t minus the gross value of tangible assets in year t-1; REV i,t : changes in revenues, this variable is equal to the revenue in year t minus the revenue in year t-1; FO i,t : the foreign operations of firm i in year t, this variable is measured by a binary value which takes 1 if the firm has make foreign operations in year t in 0 otherwise; SP i,t : change in stock portfolio from year t-1 to year t, this variable is equal to the value of stock portfolio in year t minus the value of stock portfolio in year t-1 scaled by total assets.
NBTD is a result of difference between accounting standards and tax laws.It is divergence in the treatment of goodwill (GW), investment in gross property (PPE), economic growth (REV), foreign operations (FO) and stock portfolio (SP).These items are inspired from that identified by Dridi and Boubaker (2015) as the items which represent a total disconnection in the accounting and tax treatment.All variables are scaled by total assets for year t except for FO to control for firm size.These variables are introduced to control the NBTD.ABTD will be determined based on equation (2) by the ordinary least squares method The GW is used to control the difference in treatment between the accounting and tax rules in Tunisia.On the fiscal side, the law doesn't allow the deduction of goodwill depreciation from taxable income.However, accounting standard for intangible assets recommends the depreciation of goodwill over a period not exceeding 20 years.This variable was adopted by Manzon and Plesko (2002).
The PPE is used to control the effect of investment growth on BTD.In fact, economic growth leads to value impairment of some assets that will be used and their life period will be limited.The value impairment will be recognized as the depreciation of assets.However, tax law allows only the deduction of depreciation of assets from taxable income.The REV is used to control the growth of revenue which leads to an increase in bad debt.Bad debt is fully provisioned in the accounting.On the fiscal side, this type of allowance is allowed only if it meets some conditions such as a lawsuit.These variables was adopted by Manzon and Plesko (2002) and Tang and Firth (2011).
The FO is introduced to control the difference in treatment of changes in foreign currencies.The Tunisian accounting standard requires the evaluation of evaluation of receivables and payables in foreign currencies at the end of year using the applicable exchange rate on that date.The loss or gain recorded is not allowed in tax rule since the receipt or the payment of the debt has not occurred.
Finally, the SP is used to control the growth of listed securities.Rise of the number of securities may increases the risk of impairment and consequently increases the allowance accounting record.Tax law provides for deduction of impairment of listed shares within the limit of 50% of taxable income (for illiquid shares).For highly liquid shares, accounting standard provides the recording of impairment (surplus) value a charge (gain).

Explanatory Variables Measurement  Earnings management measures
We improve the robustness of our results by employing three measures of EM as the independent variable in our study that may have association with BTD.These variables are discretionary accruals (DA), real earnings management (RAM) and changes in accounting methods (INV).
To estimate discretionary accruals, we chose Jones model modified by Dechow et al. (1995).Several models have been adopted (Healy, 1985;DeAngelo, 1986;Jones, 1991;Dechow et al., 1995;Dechow & Sloan, 1991).However, Guay (1996) concludes that the Jones model and his modified version are the most effective to give a reliable estimation of discretionary accruals.
The basic model to estimate discretionary accruals is a cross-sectional variation of Jones (1991) model: Where: TA i,t : total accruals of firm i in year t scaled by lagged total assets; A i,t-1 : assets of firm i in year t-1; ∆REV i,t : change in net sales, this variable is equal to the revenue in year t minus the revenue in year t-1; ∆REC i,t : variation in net receivable ; PPE i,t : the gross value of property, plant and equipment .
Total accruals (TA) are computed as: Where: BN i,t : net profit of firm i in year t; FGE i,t : funds generated by operations of firm i in year t.
Using Equation (3), discretionary accruals are estimated by subtraction the predicted nondiscretionary accruals.
Our second proxy for EM is real earnings management activities.As in Roychowdhury (2006) and Cohen and Zarawin (2010) we use three metrics to estimate the level of real earnings management: the abnormal levels of cash flow from operations (CFO), production costs (PROD) and discretionary expenses (DISEXP) (Note 5).
To estimate the first measure, we express normal cash flow as a linear function of sales and changes in sales of firm from year t-1 to t (Eq.5).
We mode the production costs by the sum of costs of goods sold (COGS) and the inventory growth (∆INV).
COGS is computed by the following We model change in inventory during the year as: We estimate the normal level of production costs using ( 6) and ( 7) as: We model the normal level of discretionary expenses as a function of lagged sales In the above equations CFO i,t : cash flow operations of firm i in year t; A i,t-1 : assets of firm i in year t-1 ; REV i,t : current sales of firm i in year t; ∆REV i,t : change in net sales; COGS i,t : costs of goods sold of firm i in year t; ∆INV i,t : change in inventory; PROD i,t : production costs of firm i in year t; DISEXP i,t : discretionary expenses of firm i in year t.
The abnormal component of CFO (D_CFO), PROD (D_PROD) and DISEXP (D_DISEXP) are computed as the difference between the actual values and the normal component estimated through equations ( 5), ( 8) and ( 9).These variables are used as our proxies for real earnings management.We created a single proxy that meets all these variables.We compute RM_PROXY as the sum of the variables (D_CFO), (D_PROD) and (D_DISEXP).
Our last proxy of EM is the changes in accounting methods.This variable is measured by the inventory (INV): this variable is measured by a binary value which takes 1 if the firm adopts the weighted average cost method and 0 otherwise.

 Tax management measures
To empirically test the relation between BTD and TM, we employ three variables.We use depreciation of the firm (DEP) this variable is measured by a binary value which takes 1 if the firm applies both accelerate and straight-line depreciation method and 0 otherwise.This variable is used to detect TM through tax options.Berg et al (2001) argue that firms can reduce the present value of the future tax payment through the choice of a particular method of depreciation among those admitted in the tax law.
The second variable is the reinvestment of profits (REINV) this variable is measured by a binary value which takes 1 if the firm practice the reinvestment of earnings and 0 otherwise.This variable allows controlling TM by tax incentives.To encourage and steer some areas and favor certain regions, the Tunisian legislation may be required to incite firms through its tax power.Tax legislation provides for the firms many incentives such as exemptions, reductions in tax rates or accelerated depreciation.These incentives may encourage managers to engage in activities to avoid tax.
The third variable is the effective tax rate (ETR) measured by the ratio of cash taxes paid to pretax financial accounting income.This variable is used in several studies (Holland, 1998;Gupta & Mills, 2002;Rego, 2003;Blaylock et al., 2012;Dyreng et al., 2008, Minnik & Noga, 2006;Richardson, 2013) to control the effectiveness of tax avoidance.In fact, ETR represents the alternative measure of tax aggressiveness.

Control Variables Measurement
SIZE: Firm size measured by the natural logarithm of total assets at the fiscal year-end; ROA: Profitability of the firm Dummy variable taking value 1 if the firm has a positive profit before tax and 0 otherwise.

Descriptive and Univariate Statistics
The tables below present the descriptive and univariate statisctics for variables used in the analysis of our regression model.Table 1 presents descriptive statistics for variables in ABTD model.The results show that only 13% of firms opt to depreciation method which allows a tax saving.Most of these firms do not practice the reinvestment of profit yet the results show a positive result for 80%.However, the inventory valuation by the weighted average cost method is greater than a half.The results can be explained by the trend toward earnings management by accounting choices rather than tax options.This trend comes back to the complexity of tax rules and the need for professional in this field.
In order to control the effect of qualitative variables on the ABTD, we conducted an average analysis.Firstly, we assumed a normal distribution and we compared the average of the ABTD under the different modalities.This analysis is done by the test for the equality of average through the Student-t.We conducted secondly a nonparametric test (Kruskal and Wallis test) since the first one assumes that variables have normal distributions.
The Table 1 shows that both tests led to the same results of significance.These tests are significant for the variables (REINV) and (ROA).(Tang & Firth, 2011).The mean of PPE and REV is positive, indicating a growth in investment in gross property, plants and equipment and therefore a growth in revenues.The mean of SP is 0.013 expressing the increase in investment in the stock market.Panel B allows observing the correlation of variables.The results show that the variables of BTD model are weakly correlated with each other and consequently an absence of multicollinearity.This Panel shows that BTD has a significantly positive association with PPE and SP, consistent with our predictions.
Panel C shows that the mean of ABTD is negative at 0.8% of total assets.The discretionary accruals have a negative mean (-0.65).This reveals that Tunisian companies manage their results down.The minimum and the maximum are -25 and 0.21 respectively which conclude (consistent with Kouaib & Jarboui, 2014) that in Tunisia, discretionary accruals have a significant important impact on the level of published results.The average for RM_PROXY our proxy for real earnings management for sample firms is 0.07 which takes small value comparing to the absolute value of accrual-based earnings management.This observation is consistent with Cohen et al. (2008) and Graham (2005) survey result.They suggest that real earnings management is more costly.This Panel shows also that the mean of the effective tax rate is 0.12 indicating a considerable performance of tax management by Tunisian companies.In fact, the average ETR is very small comparing to the statutory corporate tax rate (30%).This means that in average Tunisian firms pay less than half of statutory rate.This result can be explained by the tax benefits and options granted to business which allow tax avoidance.
Panel D shows the correlation matrix between metric variables of ABTD model.The results show the absence of a strong correlation between variables which may bias our estimates.ABTD has a significantly association with DA and RM_PROXY, consistent with our predictions.Discretionary accruals and real earnings management explain ABTD.The results show that estimates coefficients on the variables AD and RM_PROXY are negative and significant (p-v=0.00).This finding is consistent with our predictions.Discretionary accruals and real earnings management; proxies of the EM can explain ABTD.The negative sign can be explained by the fact that the Tunisian companies manage their result in order to reduce the tax expense.The manager is interested with taxable income and therefore manages earnings by accruals or cash flows.These practices decrease the reintegration and deduction of items.In fact, the manager uses the accounting rules for tax purposes and therefore the gap between accounting result and taxable income is low.This result confirms that the corporate taxation has close relation to financial reporting.The Tunisian capital market role in financing the firm is insignificant comparing to the importance of the tax authority in the managerial optimization practices.Tunisian companies are concerned about tax reducing rather than improved information quality and increasing the financial result.

ABTD Estimation
The coefficient on INV is not significant, suggesting that changes in accounting methods cannot explain ABTD.The coefficient on REINV is negative and significant which confirms our predictions.Tax benefits are an effective means of TM that generates an impact on ABTD.This finding can be explained by the fact that the Tunisian companies use the reinvestment earnings for an economic purpose but also for reducing the tax burden.
Likewise, the coefficients on ETR which measure the tax aggressiveness is negative and significant.The negative sign explains that when ABTD increase the effective tax rate decrease which reflects a better tax optimization.The coefficient on DEP is not significant; suggesting that this variable cannot reflects manager manipulations.

Conclusion
We document whether the ABTD can detect both earnings and tax management.Our results indicate that BTD contains a predictive power over managerial manipulation.The latter is detected from the discretionary part of the BTD.We develop a multiple regression model in which estimated BTD is regressed on a set of variables that proxy for normal difference (NBTD).This difference is designed to reflect differences between accounting and tax systems.All earnings and tax management activities are captures by our abnormal BTD indicator (ABTD estimated as the residual from the BTD regression model).The main objective of this study is to identify whether BTD detect managerial discretion in the listed companies in Tunisia.Multiple regression analyses yielded evidence consistent with hypotheses: earnings and tax management by manager explains the difference between the accounting result and taxable income (BTD) presented by the listed companies.The analyses adduced evidence of a significant relationship between ABTD and discretionary accruals and real earnings management.We find that ABTD is overall significantly associated with EM and TM proxies who can be used as a substitute of both earnings aggressiveness and tax shelters indicators.BTD can be used as an attribute of financial reporting quality and can contain incremental information for investors and users of financial reports of companies.Companies based on a continental accounting system (where tax rules are closely related to GAAP) have also manager's discretion.Unlike in developed countries, Tunisian companies are concerned about tax reducing rather than improved earnings quality and increasing the financial result.
This finding is in line with previous studies that have shown that BTD provide information about the earnings quality (Guenther, 2011;Tang & Firth, 2011;Tang & Firth, 2012;Hanlon, 2005;Formigoni et al., 2009;Blaylock et al., 2012).The reduced number of listed Tunisian companies on the Tunisian Stock Exchange limits the generalization of this finding.We propose extending the number of firms and the period study to allow a more specific pooled analysis study.We suggest also applying this study to other continental countries to allow assessment with other researches.These results are important for the development of manager behavior detection by outsiders and particularly by financial analysts and investors.BTD can be an incremental indicator of earnings quality.This result allows reflecting on the adverse effect of the disconnection between accounting and tax systems and their negatives effects on the information quality.

Notes
Note 1.Following Leuz et al. (2003, p. 506) we define earnings management as being the "alteration of firms' reported economic performance by insiders either to mislead some stakeholders or to influence contractual outcomes".
Note 2. we define aggressive financial reporting as upward earnings management that may or may not be within the confines of generally accepted accounting principles (GAAP) and aggressive tax reporting as downward manipulation of taxable income through tax planning that may or may not be considered fraudulent tax evasion (Frank et al., 2009).
Note 3. The Tunisian reconciliation to the Anglo Saxon model and especially international standards remains defaulting and few developed.Until today, the Tunisian normalize still attached to the historical cost, which puts the reliability of financial reporting before the information relevance (which implies the use of fair value).It is the main difference between the Tunisian standards and international standards.
Note 4. Mills and Newberry (2001) have focused on the differences between taxable income and accounting income before tax Lev and Nissim (2004) developed a tax-based fundamental defined as the ratio of net taxable income on the net accounting income.This ratio captures the total BTD.They found that this ratio is associated with earnings growth.
Note 5. We include a scaled intercept (1/A i,t-1 ) as it is suggested by the literature when estimating the nondiscretionary part.

Table 1 .
Descriptive statistics and mean comparison of dichotomous variables DEP: depreciation of the firm; REINV: reinvestment of profits; INV: inventory, this variable; ROA: profitability of the firm.*represents significance.

Table 2 .
Descriptive statistics and correlations GW: goodwill INV: change in investment in gross property, plants and equipment REV: changes in revenues; FO: the foreign operations; SP: change in stock portfolio; ABTD: abnormal book-tax differences; DA: discretionary accruals; DA: discretionary accruals; RM_PROXY: real activities manipulation; INV: inventory; DEP: depreciation of the firm; REINV: reinvestment of profits; ETR: effective tax rate; SIZE: natural logarithm of total assets of firm; ROA: profitability of the firm.

Table 2
presents the statistics of numeric variables and Pearson correlations for variables used in the BTD and ABTD models.Panel A presents descriptive statistics for variables in BTD model.The mean of BTD is positive 0.022 which is consistent with aggregate BTD in the United States and opposite to aggregate Chinese BTD

Table 5 .
Estimated coefficients from the BTD model ABTD i,t =  0 + 1 DA i,t +  2 RM_PROXY i,t +  3 INV i,t + 4 DEP i,t +  5 REINV i,t +  6 ETR i,t + 7 SIZE i,t +  8 ROA i,t + ε i estimate of ABTD, EM and TM made by MCG is presented in Table 5.It shows a Wald chi2 (8) is 165.43 and p-value is 0.000 indicating overall a significant model.The majority of the independent variables are individually significant in explaining the dependent variable.The software sends a strictly smaller p-value at a threshold of 5% for the coefficient relating to the variables ‫|‬AD‫,|‬ RM_PROXY, ETR, REINV, ROA, the coefficient relating to the variables SIZE, DEP, INV are not significant.We can conclude that the model contains a good explanatory power.
ABTD: abnormal book-tax differences; DA: discretionary accruals; RM_PROXY: real activities manipulation; INV: nventory, this variable; DEP: depreciation of the firm; REINV: reinvestment of profits; ETR: effective tax rate; SIZE: natural logarithm of total assets of firm; ROA: profitability of the firm.Our