COVID-19 outbreak and earnings management practice: case of Tunisia

Purpose – Thisarticleattemptstoinvestigate theimpactofCOVID-19outbreakonthe earningsmanagement (EM) for listed Tunisian companies. Design/methodology/approach – The study focuses on both accrual-based and real EM (REM) practices. With panel data, the authors employ the multiple regression approach and the generalized least squares (GLS) estimate method. The sample is made up of 41 listed companies observed from the first half of 2016 to the second half of 2020. Findings – This study finds that, during the pandemic period, Tunisian firms use decreasing income discretionary accruals. Also, with regard to REM, the COVID-19 variable displays a negative response coefficient but of lesser magnitude. Researchlimitations/implications – Thisstudy ’ sfindingscanhelpTunisianauthorities,listedcompanies and market investors to better understand EM practices during a negative shock and to better understand the various internal and external factors influencing the quality of financial reporting. These findings may contribute, also, significant EM implications for scholars interested in other emerging markets. As limitations, the authors point out mainly to the small sample size used in this study and that the authors used a single model,namelythe modifiedJonesmodel(1995),tomeasurethe accountingEM.Also, theauthors usedabinary variable as a proxy for the COVID-19 pandemic. Originality/value – Tothebestofauthors ’ knowledge,itisthefirstinTunisia,ifnotinAfrica,toexaminethe impact of the COVID-19 pandemic on EM practices. Second, this study builds on previous work by examining both the accrual-based EM and the REM.


Introduction
On March 11, 2020, the World Health Organization (WHO) declared that the COVID-19 outbreak was a global pandemic.This news rocked global financial markets since COVID-19 was first released in China (more specifically in Wuhan) in December 2019.Since the WHO's declaration of the COVID-19 pandemic, there has been significant wide ranging global 2. Literature review 2.1 EM practices during crises Initially, positive accounting theory developed the debate on the incentives of EM.Firms are motivated essentially by contractual issues.These are: namely, debt contract (Defond and Jiambalvo, 1994); compensation contract (Gaver et al., 1995) and political costs (Watts and Zimmerman, 1986).According to agency theory, EM is explained by managers' opportunistic intentions (Jensen and Meckling, 1976).These behaviors are more important when companies are experiencing difficult times (Lisboa and Kacharava, 2018).More specifically, during a pandemic or natural disaster, managers, who possess moral hazard intentions, use this force majeure to strategically manipulate statements of financial position to serve their interests to such an extent that they risk losing them in the event of a pandemic.In addition to positive accounting theory, models, which explain EM during crises, are part of institutional theory.
According to this theory, the business environment, in which firms operate, is an important institutional component and has a significant and lasting effect on business practices (Campbell, 2007).This leads us to predict changes in accounting practices, such as EM, as well as responses to environmental pressures.Indeed, to react to exogenous shocks in their business environments, firms adopt certain practices such as conservative policies relating to the holding of liquid assets (Dessaint and Matray, 2017), capital investments (Wu et al., 2022), mergers and acquisitions (Bonaime et al., 2018) and EM (Chen et al., 2012).
An overview of the literature on EM in crises shows three different perspectives.The first perspective argues that firms manage their earnings upwards during crises.Indeed, in such times, managers are under severe financial pressure due to their firms' low profitability.They tend to manage earnings upward to improve reported performance and to survive the economic downturns (Arthur et al., 2015).Increased earnings help firms to mitigate the impact of the crisis and to sustain stakeholder relationships during these turbulent periods (Lisboa and Kacharava, 2018).More specifically, it increases the confidence of investors, who are troubled by the crisis (Arthur et al., 2015), and, depending on earnings, reduces the risk of breaching debt covenants (Defond and Jiambalvo, 1994).According to Charitou et al. (2007), since this affects their wages, managers are encouraged, also, to increase earnings to escape a sharp drop in the firm's stock price.The second perspective is that firms deflate their profits in times of crisis so that they can renegotiate better conditions to restructure their debts (Asquith et al., 1994).Alternatively, the firm may declare losses that can be seen as a sign of difficulties and, consequently in union negotiations, the firm makes concessions to its employees (Filip and Raffournier, 2014).In the same literature, some other authors consider that the troubled period can be favorable to large scale practices, which are used if losses cannot be avoided (Kjaerland et al., 2020;Kustono et al., 2021).According to Saleh and Ahmed (2005), this is an accounting technique that allows managers to clean up their accounts when the firm's operating performance is expected to be poor.A third perspective argues that, when compared to periods of expansion, periods of recession are less favorable to EM practices.Indeed, Chia et al. (2007) consider that, in times of crisis, firms are closely monitored by auditors, creditors and other stakeholders.Therefore, in such circumstances, the firm's managers have less latitude to manage earnings.Also, according to Ahmad-Zaluki et al. (2011), the firm's managers have no incentive to manage earnings because the market seems to tolerate the firm's poor performance.
Empirically, EM has been examined during several crisis periods.These are namely, the 1997 Asian crisis (Chia et al., 2007); the 2007-2009 subprime crisis; the 2014 oil crisis (Bugshan et al., 2020;Kjaerland et al., 2020); and, more recently, the 2020 crisis caused by the COVID-19 pandemic (Aljawaheri et al., 2021;Xiao and Xi, 2021;Lassoued and Khanchel, 2021;Ali et al., 2022).Their studies' findings are both mixed and inconclusive.Indeed, Aljawaheri et al. (2021) examined the COVID-19 impact on the EM of 87 firms listed on the Iraqi Stock Exchange from 2018 to 2020.The results show that firms manipulate earnings to sustain profitability over time.In the same year, by using a sample of 2,029 A-share firms listed on the Shanghai and Shenzhen Stock Exchanges, Xiao and Xi's (2021) findings show that, during the COVID-19 pandemic, there was an increase in accrual-based EM.Similarly, Lassoued and Khanchel (2021) investigated the effect of COVID-19 pandemic on EM practices.Based on a sample of 2,031 listed companies from 15 European countries and using OLS (ordinary least squares) regressions, their study's findings provide evidence of significant income-increasing EM in 2020 during the period of the COVID-19 pandemic.Therefore, their findings demonstrate that, by mitigating the level of reported losses, firms manage earnings upwards to restore the investor and stakeholder confidence needed to support the firm's economic recovery.More recently, Ali et al. (2022) aimed to investigate how the COVID-19 pandemic and investor protection influenced EM.By examining 5,519 firms registered in G-12 countries from 2015 to 2020, Ali et al.'s (2022) findings reveal that, contrary to the findings of previous studies, there is a negative and significant relationship between EM and the COVID-19 pandemic.

COVID-19 outbreak and earnings management
Moreover, because the previous studies' findings are mixed and inconclusive, it is noteworthy that these studies have focused only on accrual-based EM and have ignored EM in its real aspect.

Accrual-based EM and real EM
Most previous EM studies used only one strategy: namely, either Arthur et al., 2015 accrualbased; or REM.The accrual-based EM allows managers to use financial report assessments to adjust the level of accruals and, subsequently, profitability.The REM, which is called, also, EM through real activity manipulation, is defined as management actions which vary from conventional business activities and which are conducted with the primary goal of achieving particular levels of earnings (Roychowdhury, 2006).Since Roychowdhury's (2006) theoretical work, certain authors, such as Cohen and Zarowin (2010) and Baker et al. (2018), have provided evidence that managers use both methods to achieve the desired profitability.According to Zang (2012), separate analysis of each approach can lead to misleading results.Furthermore, for Tunisian firms, Elleuch Hamza and Kortas (2019) investigated the relationship between accrual-based EM and REM (determined by discretionary expenditures and sales manipulation).Their findings suggest that, in addition to a substitutive interaction between accruals management and discretionary spending, there is a complementary relationship between accruals-based EM and sales manipulation.Finally, whether they are substitutable and/or complimentary, we believe that, to arrive at reliable conclusions, the accounting and REM need to be investigated together.

Description of data
To investigate the effect of the COVID-19 pandemic on EM, we used semiannual data in this study.We chose such data because it provides a more accurate view of an event by better capturing fluctuations in earnings and, therefore, increasing the likelihood of detecting EM (Kjaerland et al., 2020).To carry out our empirical study, we began with all the Tunisian firms listed on the Tunisian Stock Exchange that had published their semi-annual financial reports in the period from 2016 to 2020.For the sake of maintaining a balanced sample, we used as an initial sample the 79 firms listed in 2016.Next, we removed 33 financial institutions due to their accounting specificities and 5 firms with missing data).As shown in Table 1, our final sample comprises 41 companies monitored during the period from 2016 to 2020.This study period includes the prepandemic period spanning the first half of 2016 to the second half of 2019 and then the period of the COVID-19 pandemic from March to December 2020.Consequently, there are, in total, 410 observations.We collected by hand the semi-annual and annual financial statements data available on the Financial Market Council website (http://www.cmf.org.tn) and on the Tunisian Stock Exchange website (http://www.bvmt.com.tn).
3.1.1Dependent variable.The dependent variable is "earnings management, EM" And in this respect, our study covers both accounting and REM.(1995), to determine the nondiscretionary level of total accruals (TAC).The measure of discretionary accruals (DA) is the residuals from firm-specific regression of changes in noncash sales and gross level of property, plant and equipment.Discretionary or abnormal accruals (lagged by total assets) serve then as a proxy for AEM and are calculated as follows: To begin, calculate TAC: Where, i and t denote respectively the firm and the semester during the period from 2016 to 2020; TA is the total assets; ΔREV is the variation in revenues; and ΔREC is the variation in net accounts receivables.
Next, we used the parameter estimates from equation ( 1) to generate the non-DA.DA are then calculated as follows: Clearly, high DA reflect an upward accounting EM.For our empirical analysis, AEM 5 DA.
3.1.1.2The Real EM (REM).We measured REM by using three proxies.These are namely, operating cash-flows; discretionary spending and cost of production.According to Roychowdhury (2006), an abnormal low level of cash-flow from operating activities, an abnormal low level of discretionary expenses, such as administrative expenses and R&D expenses, and an abnormal high level of production costs can be used to indicate REM occurrence.Due to data availability issues, we decided to remove the third proxy.Hence, we designed the following: where, i and t denote respectively the firm and the semester during the period from 2016 to 2020; and CFO (Chief Financial Officer) is the operating cash flow; and DEx is the discretionary expenses.In terms of the total of advertising expenses, R&D expenses and selling, general and administrative expenses, TA is the total assets and the residual term ε refers to the sum of abnormal operating cash flow and abnormal discretionary spending.
Then we used the parameter estimations from equation (3) to generate the abnormal operating cash flow (AbnCFO) and abnormal discretionary expenses (AbnDEx).
Finally, we multiplied (AbnCFO þ AbnDEx) by a negative one to obtain the extent of REM.For our empirical analysis, REM 5 À (AbnCFO þ AbnDEx).The higher values suggest that, to manage earnings upwards, the firms decreased more operating cash flows and discretionary costs.In this work, the REM is combined with the accrual-based EM to derive total EM.We then get the following: COVID-19 outbreak and earnings management 3.1.2Independent variable.To investigate the influence of the COVID-19 pandemic on EM, the independent variable in our modeling is a binary variable (COVID) which takes the value of 1 if the observation concerns the year 2020 (the pandemic period) and otherwise is zero.
3.1.3Control variables.The modeling includes six control variables that may influence EM.These are namely, return on assets (ROA), market-to-book (MTB), financial leverage (LEV), audit quality (AUD), ownership concentration (BLOC) and firm size (SIZE).
All of these variables are summarized in Table 2.

Specification of empirical approach
To fulfill our study's objective, we examined not only the whole EM but, also, separately the real and the accrual-based EM.We used the multiple regression methodology with panel data.The data has two dimensions: namely, an individual dimension (1) representing the firm and a (t) index representing the period of study.Given the particular nature of panel data, certain tests need to be performed to obtain reliable estimates.To test for the presence of individual effects, we performed first an 'F-statistic'.Significant chi-square values are displayed that indicate the presence of individual effects and confirm sample heterogeneity.Then, we implemented the Hausman test to choose between fixed and random effects models.
The results, which are not reported here, indicate that the random-effects model outperforms the fixed-effects model.Accordingly, we applied the generalized least-squares (GLS) estimation method.We design the following models: where, i and t denote respectively the firm and the semester during the period 2016-2020.

Preliminary data analysis
Table 3 displays the descriptive statistics for all study variables.It shows that, on average, the total EM is 0.016; the accrual-based EM is 0.012; and the REM is 0.003.According to these statistics, Tunisian listed firms generally use both real and accounting EM strategies to value their earnings.As for the control variables, the average profitability is 0.046 with a minimum of À0.254 and a maximum of 0.789.On average, the MTB ratio is 1.832.Tunisian firms appear, also, to be heavily indebted.Indeed, the ratio of debt-to-total-assets (mean 5 0.606) averages well over half of total assets and varies from 0.008 to 2.265.The binary variable, "Audit quality", shows an average of 0.185 and the variable, "Ownership concentration", shows that in more than half of Tunisian firms the average capitalization, three is 5% or more block holder participation (0.547).
To identify any potential multicollinearity issue, Table 4 displays the correlation matrix between the different explanatory variables.According to Gujarati et al. (2004), such problems can occur when the correlation between two variables exceeds a value of 0.8.This

COVID-19 outbreak and earnings management
study's results show very low correlations between the independent variables (best 0.67) and no multicollinearity problems.Furthermore, Table 5 shows the mean difference in AEM, REM and EM between 2020 and other years.This study's results show negative values for 2020 which is contrary to all the other years.This difference is statistically significant for the three variables.A priori, the COVID-19 pandemic tends to have a negative influence on EM; this needs to be validated by multivariate analysis.

Empirical results
Table 6 displays the results for the estimated models M1, M2 and M3 using the GLS method.It is noteworthy that the dependent variables for these models are respectively total EM, accrual-based EM and REM.In this regard, the Wald test indicates that models M1, M2 and M3 significantly fit the data.Table 5 shows that the Wald χ 2 global significance test is significant at the 1% level for all models.
As expected, the market-to-book (MTB) and Financial leverage (LEV) variables as control variables show clearly positive coefficients in the three models.On the one hand, firms with higher market's expectations for future profitability and those, which are greatly leveraged, tend to engage in more upward EM practices.On the other hand, EM is influenced negatively by the profitability variable (ROA) and BLOC.However, the (BLOC) variable does not have a significant effect on REM (see Model 2).For the AUD and SIZE, neither appears to have any significant effect on the EM practices.
The COVID-19 pandemic (COVID) is the variable of interest in our regressions.As shown in Table 6, this variable has a negative and significant impact (À0.014) at the 5% threshold on accrual-based EM (see model 1).Therefore, during the period of the COVID-19 pandemic, Tunisian listed companies used decreasing income DA.With regard to REM (see Model 2), the variable (COVID) displays, also, a negative and significant response coefficient at the 10% threshold, but of lesser magnitude (À0.009).A priori, during the COVID-19 pandemic, Tunisian firms have deferred their revenues by delaying sales and/or accelerating spending on R&D, advertising campaigns and pension contributions.By combining accounting and REM, Model 3 estimates show that, at the 1% threshold, the COVID-19 pandemic has had a negative and significant impact on total EM.The magnitude of this effect is larger (À0.021) when compared to models M1 and M2.Tunisian listed firms adopted both accrual-based and REM downwards during the period of the COVID-19 pandemic.These findings support Kjaerland et al.'s (2020) and Kustono et al.'s (2021) hypotheses that troubled periods can be favorable to large scale practices.According to Habib et al. (2013), large scale accounting practices during the crisis period allows firms to report positive earnings in the post-crisis period.Investors may expect firms to report losses and, therefore, it is a rational managerial response to manage earnings downward by adopting large scale accounting practices during such times.Since firms' managers realize that the market tolerates poor performance during a crisis, they have greater incentives to reduce earnings further and to reverse the performance in the post-crisis period.Certainly, Tunisian firms have been greatly affected by the lockdown and restriction measures announced on March 9, 2020 in the hope of overcoming the health crisis.Economic activity has been affected mainly through multiple transmission channels, such as halting production, closing stores, halting business-to-business recovery and suspending international trade operations.It was expected then that these measures would result in firms' poor operating performance.In this sense, a survey conducted in March 2020 by the Arab Institute of Business Leaders in Tunisia (IACE, Institut Arabe des Chefs d'Entreprises), shows that 96.7% of the questioned business leaders feared that the COVID-19 pandemic would have a negative and significant influence on their activities for the next six months.This could explain the tendency of managers to clean up their accounts to make future performance seem better.

Additional analysis
We conducted an additional interaction analysis to strengthen the empirical analysis.We tested the interaction effects of each explanatory variable and the COVID-19 pandemic on EM.
When compared to the M3 estimation results, we note first that in their interactions with the COVID variable, the ROA and BLOC variables no longer have a statistically significant impact on EM practices.However, the AUD variable, which has no significant effect in the estimation of M3, shows a negative and statistically significant interaction coefficient at the 1% threshold.As for the MTB variable, its contribution to explaining EM is much greater (coefficient of 0.037 versus 0.013) when combined with the COVID variable.

Conclusion
This study examined the impact of the COVID-19 pandemic on the Tunisian listed firms' EM practices.The findings show that these firms used decreasing income DA during the period of the COVID-19 pandemic.Also, with regard to REM, the COVID variable displays a negative and significant response coefficient but of lesser magnitude.Tunisian listed firms adopted both accounting and REM downwards during the period of the COVID-19 pandemic, indicating that the troubled period can be favorable to large scale practices.These findings present strong implications.In fact, this study's findings can help Tunisian authorities, listed companies and market investors to better understand EM practices during a negative shock and to better understand the various internal and external factors influencing the quality of financial reporting.These findings may contribute, also, significant EM implications for scholars interested in other emerging markets.As limitations, we point out mainly to the small sample size used in this study and that we used a single model, namely the modified Jones model (1995), to measure the accounting EM.Also, we used a binary variable as a proxy for the COVID-19 pandemic.Going forward, we recommend that future research studies rely more on annual data.This would certainly increase the sample size since all firms publish annual reports.In addition, we recommend that future studies evaluate the impact of the severity of the COVID-19 epidemic on EM by making use of other COVID-19 indicators such as (Ali et al., 2022) either the mortality rate or the hospitalization rate.It would also be interesting to integrate the financial distress (Ali et al., 2022) variable and to study the effect of its interaction with the COVID-19 pandemic on EM.

Main References
Ahmad-Zaluki, N.A., Campbell, K. and Goodacre, A. (2011) The accrual-based earnings management (AEM).Consistent with most EM research studies, we used the modified Jones model, introduced by Dechow et al.

Table 2 .
Summary of study variables