*2.5. Audit Committee*

The Public Companies Act and the Stock Exchange Regulations stipulates whether Norwegian public companies are required to establish an audit committee or not. The members of the audit committee are elected by and among the board members and at least one of the members of the committee must be independent with regards to NUES' (2018) recommendations (Lekvall et al. 2014). According to the Public Companies Act, the audit committee's primary mission is to prepare the supervision of the financial reporting process and monitor the systems for internal control and risk management. The committee should further meet regularly with the firm's external auditor and internal financial managers to produce balanced and accurate reports. Accordingly, audit committees complement existing internal governance practices by improving the monitoring function and reduce agency conflicts (Cai et al. 2015). Prior studies have found a significant relation between earnings management and audit committee practices (Bedard et al. 2004; Wan Mohammad et al. 2016). Klein (2002) found that the existence of an audit committee will reduce earnings management. Similarly,

Dechow et al. (1996) and Purat Nelson and Devi (2013) found that firms manipulating earnings were less likely to have an audit committee. The last hypothesis is formulated as follows:

**Hypothesis 6** (**H6**). *There is a negative relation between the presence of an audit committee and earnings management*.

#### **3. Data and Methodology**

#### *3.1. Data And Sample Selection*

Our initial dataset consisted of quarterly financial statements from 168 companies listed on the Oslo Stock Exchange in the period 2014 to 2017. Due to difficulties in defining abnormal accruals in the financial service industry, 16 bank and insurance companies were eliminated from the sample. In addition, there is an exclusion of 18 companies that had not been listed for the entire period, 83 firms due to lack of data and 2 firms due to mergers and acquisitions in the period (see Table 2). The financial data was collected through the Thomson Reuters Eikon database, while the corporate governance data was collected from companies' annual reports. If the reports lacked data, it was retrieved directly from the companies through e-mails and phone calls.



In Das et al.'s (2009) study on quarterly earnings patterns and earnings management, they find that firms performing poorly in interim quarters may attempt to increase earnings in the fourth quarter to achieve a desired annual earnings target. Accordingly, this study used data from quarterly reports in the analyses to catch more of the fluctuations in earnings. Further, interim reports are often unaudited, which allows greater managerial discretion and require less detailed disclosure than annual financial statements (Jeter and Shivakumar 1999). Using quarterly financial data in the analysis could thus increase the likelihood of detecting earnings management.

## *3.2. Measurement of Earnings Management*

In the existing earnings management literature, a commonly used approach for detecting earnings management is by examining accruals. The literature distinguishes between two widely used approaches in defining total accruals: the balance sheet-based approach (Healy 1985; Jones 1991) and the cash flow-based approach (Vinten et al. 2005). The cash flow approach measures accruals directly from the statement of cash flows which mitigate the danger of measurement errors. Consequently, this study used the cash flow approach to define total accruals. The cash flow approach measures total accruals as the difference between the earnings of an entity and its cash flow generated from operating activities. Thus, to calculate total accruals using the cash flow approach the following formula has been used:

$$\text{TA}\_{\text{it}} = \text{NI}\_{\text{it}} - \text{CFO}\_{\text{it}}$$

where TAit = total accruals for company i in quarter t, NIit = net income for company i in quarter t and CFOit = cash flow from operating activities for company i in quarter t.

Total accruals consist of a discretionary component and a nondiscretionary component. Nondiscretionary accruals represent changes in a company's underlying performance, while discretionary accruals represent changes due to management's accounting decisions (Ronen and Yaari 2008). When estimating earnings management, it is the discretionary accruals that are of interest. A fundamental issue is however the challenge of separating the discretionary and nondiscretionary components of earnings (Elgers et al. 2003), since they cannot be directly observed. Several methods have been developed to estimate the discretionary component of accruals. A widely used approach is to benefit regression techniques, where total accruals are regressed on variables that are proxies for normal accruals. Discretionary accruals were thus the unexplained component of total accruals.

Several widely used regression techniques have their origin in the original Jones model from 1991. This study used 2 modified versions of the original model; the Modified Jones model proposed by Dechow et al. (1995) and a performance-matched model introduced by Kothari et al. (2005). The Modified Jones model was designed to eliminate the assumed tendency of the Jones model to measure discretionary accruals with error when discretion was exercised over revenues (Dechow et al. 1995). The modification made from the original Jones model is that changes in revenues are adjusted for the changes in receivables in the event period. When applying the Modified Jones model, the nondiscretionary and the discretionary components of total accruals can be calculated by the following equation (Dechow et al. 1995):

$$\frac{\text{TA}\_{\text{it}}}{\text{A}\_{\text{it}-1}} = \beta\_0 + \beta\_1 \frac{1}{\text{A}\_{\text{it}-1}} + \beta\_2 \frac{\Delta \text{REV}\_{\text{it}} - \Delta \text{REC}\_{\text{it}}}{\text{A}\_{\text{it}-1}} + \beta\_3 \frac{\text{PPE}\_{\text{it}}}{\text{A}\_{\text{it}-1}} + \varepsilon\_{\text{it}}.\tag{1}$$

where

TAit = total accruals deflated by lagged total assets for company i in quarter t

Ait−<sup>1</sup> = lagged total assets for company i in quarter t

ΔREVit = changes in total sales deflated by lagged total assets for company i in quarter t ΔRECit = changes in account receivables deflated by total assets for company i in quarter t PPEit = net value of property, plant and equipment deflated by lagged total assets for company i in quarter t

Kothari et al.'s (2005) performance matched model is an extended version of the Modified Jones model, where return on assets (ROA) is added as an additional variable. The following equation is used:

$$\frac{\text{TA}\_{\text{it}}}{\text{A}\_{\text{it}-1}} = \beta\_0 + \beta\_1 \frac{1}{\text{A}\_{\text{it}-1}} + \beta\_2 \frac{\Delta \text{REV}\_{\text{it}} - \Delta \text{REC}\_{\text{it}}}{\text{A}\_{\text{it}-1}} + \beta\_3 \frac{\text{PPE}\_{\text{it}}}{\text{A}\_{\text{it}-1}} + \beta\_4 \frac{\text{ROA}\_{\text{it}}}{\text{A}\_{\text{it}-1}} + \varepsilon\_{\text{it}} \tag{2}$$

where

ROAit = net income after tax deflated by lagged total assets for company i in quarter t

Kothari et al. (2005) claim that economic intuition, empirical evidence and extant models of accruals suggest that accruals are correlated with a firm's present and past performance. Hence, to control for performance on discretionary accruals, ROA is added as a control variable. Further, because of the nonlinear relationship between accruals and performance, Kothari et al. (2005) argue that a performance matched approach is better specified to test discretionary accruals than by using a linear regression-based approach.

In both models the variables are deflated by lagged total assets to control for firm size effect (Healy 1985; DeAngelo 1986) and to mitigate heteroscedasticity in the residuals (White 1980). Further, nondiscretionary accruals are estimated using ordinary least squares (OLS). The prediction from the

OLS estimation in model (1) and model (2) represents nondiscretionary accruals while the residuals represents discretionary accruals. Discretionary accruals can be both positive and negative. In the analysis, the study used the absolute value of discretionary accruals as a proxy for earnings management (as a normal procedure—see Hribar and Nichols (2007) for elaboration). Higher levels of discretionary accruals indicate greater levels of earnings management.

The Modified Jones model (1) showed an explanatory power of 0.1139 (Table A1), while the Kothari model (2) showed an explanatory power of 0.4334 (Table A2). The higher the explanatory power, the closer the estimated regression equation fits the sample data (Brooks 2019). Hence, the measure of discretionary accruals following the Kothari model (2) was used as the dependent variable for the further corporate governance analysis.
