Earnings Management during the Global Financial Crisis: Evidence from Australia

Transcription

1 Earnings Management during the Global Financial Crisis: Evidence from Australia Abu Taher Mollik, Monir Mir, Ronald McIver M Khokan Bepari This paper examines if Australian firms engaged in a higher level of earnings management during the global financial crisis (GFC) and if there was any industry effect on firms earnings management. During financial crises many firms experience a systematic decline in incomes, and majority of them report a fall in earnings or losses. The onset of a financial crisis may, therefore, trigger or magnify managerial motives to engage in earnings management, attributing the reduced earnings (or losses) to the macroeconomic shocks rather than to poor managerial performance. Earnings management has been defined in terms of the discretionary components of total accruals. We use parametric and non-parametric tests and panel data regression methods to analyse the impact of the GFC and industry effect on the discretionary accruals sample comprised of 149 Australian firms during the 2006 to 2009 period. We find that Australian firms engaged in a higher level of income-decreasing earnings management during the GFC. This finding is consistent with that of earlier research in the Asian Financial Crisis context. We also find that firms industry classification has statistically significant effect on their earnings management. Keywords: Earnings Management; Global Financial Crisis; Discretionary accruals, Panel data, Fixed-effect model. JEL Classification: G32; G01; M42 1. Introduction Empirical evidence suggests that firms engage in aggressive earnings management during periods of financial crisis (Chia et al. 2007; Johl et al. 2007). During financial crises many firms experience a systematic decline in incomes, and a majority will report a fall in earnings (or losses). Thus, the onset of a financial crisis may trigger (or magnify) managerial motives to engage in earnings management (Kim & Yi 2006). Management may attribute reduced earnings (or losses) to the macroeconomic shock rather than to poor managerial performance. Detection of such activities by investors requires that a firm s financial reports accurately communicate changes in its underlying economic position. Earnings management has remained a widely-researched area in accounting for the last two decades. In the accounting literature a variety of terms are synonymous with earnings management: creative accounting, cooking the books, earnings manipulation, accounts manipulation, income smoothing, etc. Consensus on the The authors gratefully acknowledge the financial support of CPA Australia for this research project. This is through their provision of CPA Australia Global Research Perspectives Scheme grant funding. Assistant Professor in Banking & Finance, Faculty of Business & Government, University of Canberra (contact author). Professor, Accounting, Banking & Finance, Faculty of Business & Government, University of Canberra. Lecturer in Financial Economics, Centre for Applied Financial Studies (CAFS), School of Commerce, Division of Business, University of South Australia. Sessional Academic staff in Accounting and Finance in the University of Canberra. 1

2 exact definition of earnings management is lacking. However, Healy and Wahlen provide a comprehensive definition: Earnings management occurs when managers use judgment in financial reporting and in structuring transactions to alter financial reports to either mislead some stakeholders about the underlying economic performance of the company or to influence contractual outcomes that depend on reported accounting numbers (Healy & Wahlen 1999 p. 368). This definition suggests that earnings management requires both motivation and opportunity. Diverse managerial motives for earnings management exist, as supported in prior research. These motives include: political cost minimization; financing cost minimization; managerial wealth maximization; debt contracts; compensation agreements; equity offerings; insider trading; reductions in the likelihood of wealth transfers; obtaining import relief; decreasing earnings during union negotiations; decreasing earnings in periods preceding management buyouts; and mergers (Watts & Zimmerman 1978; Jones 1991; Woody 1997; Louis 2004). However, managerial decisions to engage in earnings management are normally based on opportunistic reactions in response to incentives created by specific economic and financial conditions (e.g., economic downturn, or an unexpected fall in earnings/unexpected loss). One reason managers involve in earnings management is based on reflecting their opportunistic incentives ( Bernard and Skinner, 1996). A number of studies also document that earnings management is motivated by managers compensation contracts. There are few other circumstances where both the incentive and ability to manage earnings are as closely connected (i.e., through bonuses). Thus, the general proposition tested in these studies is whether firms with accounting-based bonus plans are more likely to adopt accounting methods that increase reported earnings (Watts & Zimmerman 1978). For example, Healy (1985) observes that managers will manage earnings upward if unmanaged earnings are between the level required to trigger the bonus (the lower bound) and that which gives the maximum bonus (the upper bound). The consequence of earnings management is that a firm s financial reports may not accurately communicate its underlying economic position. This is due to deliberate choices by management regarding financial reporting methods, estimates, and disclosures (Healy & Wahlen 1999). In the accounting literature, observed managerial earnings management behaviours have been widely explained using both the big bath and income smoothing hypotheses. The big bath hypothesis (Healy 1985) implies that when the firm s current period earnings are unexpectedly low depriving managers of the opportunity to receive a bonus or meet pre-specified targets managers will clear off future potential expenses by matching additional discretionary charges to current earnings, worsening the reported financial outcome. According to this hypothesis, the information content of a negative move in current earnings is the same regardless of the scale of the fall in earnings. Healy (1985) shows that if current earnings are 2

3 below the earnings-based bonus plans floor, managers have incentives to lower earnings further using negative discretionary accruals. Empirical evidence shows that a managerial bonus plan is not the only situation when managers apply the big bath. It may also be applied during management changes (Strong & Meyer 1987; Wells 2002; Masters-Stout et al. 2008) and accounting policy changes (Beatty & Weber 2006). Lim and Matolcsy (1999) document the existence of significant income-reducing earnings management in Australian companies in the presence of price controls (established by the Australian government in the early 1970s) when subjected to high levels of scrutiny. The income-smoothing hypothesis implies that managers accounting choices are driven by their desire to remove (manage) large earnings fluctuations around predetermined target earnings levels. Thus, income smoothing reflects the propensity of managers to choose accounting policies that increase (decrease) reported earnings when unmanaged current period earnings are below (above) target earnings (Gill-de- Albornoz & Alcarria 2003). Income smoothing has been linked to dividend stability and higher share prices, reflecting managers efforts to signal positive private information about the firm s future performance (Wang & Williams 1994). Tse and Tucker (2009) find that managers time their earnings announcement (with earnings shocks) to occur soon after their industry peers warnings to minimize their apparent responsibility for earnings shortfalls. They conclude that the observed clustering of earnings shocks announcement is due to managerial herding behaviour. Karuna et al. (2012) examine the relationship between industry product market competition and earnings management. Their findings suggest that industry factors play a vital role in influencing the variation in extent of earnings management across firms. Interpreted together, the findings of Tse and Tucker (2009) and Karuna et al. (2012) imply that firms industry identity may affect firms earnings management behaviour. However, little evidence exists on the effect of industry identity on firms earnings management behaviour. In light of the above, we examine firms earnings management behaviour during a systematic financial crisis (e.g., Fiechter & Meyer 2010); and the industry effect on the firms earnings management. We focus on a single, developed economy Australia a country for which there is limited coverage in the literature. Following Miller et al. (2004), we focus on a single country context to avoid methodological and modelling problems frequently associated with multi-country studies. These arise in the forms of omitted and noisy variables, and differences in the cross-country effects of key factors, potentially complicating or preventing a clear identification of core concepts and an in-depth understanding of the event (problem) being examined. In line with these motivations, the present study embraces two research questions: (i) Did Australian firms engage in a greater level of earnings management during the global financial crisis? (ii) Does firms industry identity affect the earnings management behaviour? We follow Chia et al. (2007) and Johl et al. (2007) in examining firms earnings management behaviour in the context of a financial crisis. However, we differ from these two studies in that we examine aggressive earnings management behaviour during a period of financial crisis in a developed economy Australia. 3

4 The rest of the paper is as follows. Section 2 discusses and identifies the relevant time periods to be associated with the GFC and pre-crisis period (PCP) in Australia s case. Section 3 outlines the data and sample set from which the metrics in the paper are derived. Section 4 describes the methodology and basis for the construction of the models used in this study. Section 5 provides the results from the statistical analysis. Section 6 concludes the paper. 2. Defining the Global Financial Crisis (GFC) and the Pre-Crisis Period (PCP) Drawing an appropriate timeline for the GFC and the PCP in the Australian context is troublesome, given the GFC of 2008 and 2009 started in the US. The GFC began in the US market in mid-2007 but its impact was felt more broadly across the globe from the second quarter of The Australian financial market reflected little impact of the US sub-prime mortgage crisis during 2007 (Xu et al. 2011). Grosse (2010) suggests that the US financial crisis began to develop into a GFC from March 2008, with the first major event indicative of a spill-over effect being the near failure and then Bank of America purchase of Countrywide Financial in January Following this, the crisis spread around the world, as is apparent in the near shutdown of the inter-bank lending market from the end of March This continued well into 2009, with positive GDP growth returning for many countries affected by the crisis in the second half. Thus suggests that the years 2008 and 2009 be considered as the GFC period whereas 2006 and 2007 are considered as the pre-crisis period, PCP. Taking into consideration these issues, in the Australian context the GFC and the PCP are defined based on existing literature (Sidhu & Tan 2011; Xu et al. 2011; Spear & Taylor 2011) and movements in the ASX All Ordinaries Index. The ASX All Ordinaries Index was 3546 in 30 June 2004 which increased to 4347 as of 30 June2005, and to 5034 during June It soared to 6311 during June 2007 and up to 6779 during October The index began to decline thereafter, dipping to 5333 in June 2008 and then to 3297 by February At the end of June 2009, the index was 3948 (Yahoo finance). In Australia, the unemployed rate began to rise dramatically from October The reserve bank of Australia began to cut interest rates from 7.25 per cent in September 2008, to three per cent in Based on this and consistent with other studies (e.g., Mahmood et al. 2010), this paper, therefore, utilises 2006 to 2007 as the PCP period and 2008 and 2009 as the GFC period. 3. Data and Sample For the sample the 301 firms included in the S&P ASX 300 list as at September were identified. These firms represent 80+ per cent of the market capitalisation of ASX listed companies (S&P, 2007). Banks, other financial institutions, trust companies and utility companies were excluded because these companies have different legal and regulatory reporting requirements. This resulted in the elimination of 64 companies. Financial firms have different corporate governance structures and 4

5 disclosure requirements (Khan et al. 2008), and have also been excluded in previous studies (Baxter & Cotter 2009; Habib 2008). Of the remaining 237 firms, 9 firms were excluded because their annual reports were not available from the Connect Four annual report collection. A further 73 firms were excluded either because the required variables for the calculation of the proxies for earnings management discretionary accruals (DACs) were not available, or they belong to an industry with less than 10 firms (the calculation of DACs requires that there be at least 10 firms in the industry). The above processes left a sample comprised of 149 firms. These firms were then traced back from 2009 to 2006 to be included in the final sample, which consists of 576 firm-year observations. Observations are 149, 147, 141 and 139 for the years 2009, 2008, 2007 and 2006, respectively. Financial accounting data has been collected from the Datastream Advance data base. Data has been collected from companies annual reports, sourced from the Connect 4 data base. Table 1 outlines the sample selection procedure and the industry distribution of the final sample. Table 1: Sample selection procedure and industry distribution of the sample Panel A: Sample selection procedure Firms included in the S&P ASX 300 index as at September Less: GICS 60: A-REIT 22 Less: GICS 65: Financial x-a-reit 29 Less: GICS 50: Utilities companies 13 Remaining firms 237 Required data not available 6 Annual report not available via the Connect Four data base 9 Less: Financial data not available to calculate the proxy for earnings management 73 (DAC) or there are less than 10 firms in the industry Total firms remaining in sample 149 Panel B: Industry distribution of the sample Total Consumer Discretionary Consumer Staples Energy Health Care Industrials Information Technology Materials Telecommunications Services Total Common firms for all four years Research methods 4.1 Proxies for earnings management discretionary accruals Prior research has used a variety of measures of DACs to proxy earnings management. Since DACs are not observable, many proxies and estimation techniques have been suggested in the literature to capture the DACs. For example, Healy (1985) uses total accruals as the proxy for DACs, whereas DeAngelo (1986) uses the change in total accruals as the proxy for DACs. Jones (1991) uses a more 5

6 sophisticated approach to estimate DACs. This is to decompose total accruals into explained (non-discretionary accruals, or NDAs) and unexplained components (the DACs) via regression methods. More commonly, total accruals are assumed to be the sum of both DACs and NDAs. To determine NDAs, total accruals are regressed on changes in revenue or sales during the year, the firm s property, plant and equipment, and current, one-period lag and lead cash flow from operations. The unexplained portion of total accruals from this regression is considered as providing a measure of DACs. Dechow et al. (1995) evaluate the relative performance of five DAC models including those of Healy (1985), DeAngelo (1986), the industry model proposed by Dechow and Sloan (1991), the Jones (1991) model, and their own proposal for a Modified Jones model. The performance of the five models is evaluated using four samples: (a) a random sample; (b) a sample of firm-years experiencing extreme financial performance; (c) a sample of firm-years with artificially induced earnings management; and (d) a sample of firm-years for which the SEC alleged earnings were overstated. Dechow et al. (1995) conclude that the Jones (1991) and Modified Jones models perform better than the other models tested. Guay et al. (1996) provide a similar assessment of the five DAC models, finding that only the Jones and Modified Jones models appear to have potential to provide reliable estimates of DACs. Consistent with the evidence provided by Dechow et al. (1995) and Guay et al. (1996), it follows that the Jones (1991) and Modified Jones models are the models most frequently used in estimating DACs in a wide range of previous studies (e.g.,defond & Jiambalvo 1994; Dechow et al. 1995; Subramanyam 1996; Beneish 1997; Becker et al. 1998; Francis et al. 1999; Bartov et al. 2000; Krishnan 2003; Davidson et al. 2005; Kothari et al. 2005; Chia et al. 2007; Ahmed et al. 2008; Baxter & Cotter 2009; Iqbal & Strong 2010). Within this literature a cross-sectional version of the Modified Jones model is common. The absolute value of DACs in the Modified Jones (1991) model serves as a proxy for accrual-based earnings management (consistent with Francis et al. 1999; Krishnan 2003). A time-series approach would normally be expected in estimating firm-specific accruals, which is assumed to capture the accruals-generating process. Indeed the original Jones model (1991) requires 10 time series observations for each sample firm for reliable measures of NDAs and DACs. Thus, data non-availability is one problem for this model, as is survivorship bias (DeFond & Jiambalvo 1994; Rees et al. 1996). Additionally, a firm-specific approach needs to assume that the process generating the accruals is constant over time (Richardson 2000, p. 331). However, this assumption is likely to be violated during periods of changing economic conditions especially those associated with a crisis period such as the GFC (Ahmed et al. 2008). For these reasons the literature suggests that cross-sectional versions of the Jones and Modified Jones models will have better power in decomposing accruals into NDAs and DACs and thus in detecting DACs than the time-series versions (Subramanyam 1996; Jeter & Shivakumar 1999; Bartov et al. 2000). In light of the above, three alternative specifications are used to estimate DACs: the cross-sectional version of Jones (1991) model; the Modified Jones model proposed 6

7 by Dechow et al. (1995); and, as proposed by Kothari et al. (2005), a Modified Jones model with an alternative control variable return on assets (ROA) used to capture firm performance (see also Choi et al. 2011). This performance adjusted discretionary accruals model has also been used by Lawrence et al. (2013). In each case DACs are represented by model residuals with NDAs being the explained component of the model. Jones (1991) model: TAC 1 REV PPE it TA TA TA TA it it it it 1 it 1 it 1 it 1 (eqn. 1) Modified Jones model (Dechow et al., 1995): TAC 1 REV REC PPE it TA TA TA TA it it it it it 1 it 1 it 1 it 1 (eqn. 2) Modified Jones model adjusted for firm performance (Kothari et al., 2005) TAC 1 REV REC PPE ROA it it TA TA TA TA it it it it it 1 it 1 it 1 it 1 (eqn. 3) Where: TAC it TA it-1 REV it REC it PPE it ROA it ε it Total accruals (measured as net income cash flow from operations) for firm i at time t Total assets at the beginning of the year for firm i at time t Changes in total revenue for firm i at time t Changes in receivables for firm i at time t Property plant and equipment for firm i at time t Return on assets for firm i at time t Error term (discretionary accruals component) for firm i at time t The residuals from equations 1, 2 and 3 provide the three alternative measures of DACs used in this paper (hereafter identified as DAC1, DAC2 and DAC3, respectively). DACs can be either positive or negative, as discretionary accruals can be used either to conceal poor performance or to save current earnings for a future time period (Gul et al. 2003). Negatively-signed DACs represent income-decreasing discretionary accruals (DECDACs), while positive-signed discretionary accruals are considered as income-increasing discretionary accruals. Absolute values of the DACs from the modified Jones (1991) model serve as proxies for accruals-based earnings management (Francis et al. 1999; Krishnan 2003), with the absolute values of the residuals from equations 1, 2 and 3 (hereafter ABSDAC1, ABSDAC2 and ABSDAC3, respectively) providing alternative proxies for earnings management in this paper. 7

8 4.2 Measuring total accruals Total accruals are calculated by subtracting cash flows from operations (OCF) from income before extraordinary items. Hribar and Collins (2002) argue that the difference between net income and cash from operations is the correct measure of total accruals and that the use of a balance sheet approach may in some circumstances lead to a systematic bias in the estimated discretionary accruals. Hence, consistent with Choi et al. (2011) and Ahmed et al. (2008) total accruals in this paper are calculated as the difference between income before extraordinary items and tax and the OCF. 4.3 Model for examining research questions 1 and 2 To test research questions 1 and 2, we identify the sub-sample of firms with income-decreasing discretionary accruals. The effect of the GFC on earnings management is examined using the following cross-sectional models: DECDAC it 0 1CRISIS 2OCFit 3LEVit 4SIZEit 5NEGit iii it i 1 (eqn. 4) ABSDAC CRISIS OCF LEV SIZE NEG I (eqn. 5) it it 3 it 4 it 5 it i i it i 1 Where: DECDAC it Income decreasing or negative accruals for firm i at time t ABSDAC it Absolute value of discretionary accruals for firm i at time t CRISIS Crisis dummy taking the value of 1 for the year 2008 and 2009, 0 otherwise OCF it Cash flow from operations for firm i during year t deflated by the lag of total assets LEV it Leverage ratio for firm i during year t measured as total liabilities to total debt SIZE it Natural log of the market value of firm i during year t NEG it Indicator variable taking the value of one if firm i has negative earnings for year t, 0 otherwise Industry dummy for firm i I i Industry dummies were used to test research question 2, that is to examine the industry effect on earnings management during GFC. Industry dummies are defined as: Consumer Discretionary (CD); Consumer Staples (CS); Energy (Energy); Home Construction (HC); Industrial (IND); Information Technology (IT); Materials (MAT); and Telecommunication Services (TS). Separate industry estimation also permits the coefficients to reflect systematic variations in economic and accounting environments across industries (Barth et al. 1999). A statistically significant 1 (the coefficient of the crisis variable) would imply that firms displayed different earnings management behaviour during the GFC period as compared to the PCP. Equations 4 and 5 are estimated to examine if firms engage in an increasing level of income decreasing earnings management during the GFC compared to the PCP. A statistically significant p i (the coefficient of industry dummies) would suggest that company s industry identification has significant impact on its earnings management behaviour. j j 8

9 LEV it, SIZE it, NEG it, OCF it are used as control variables, given previous studies have found these variables to be significantly related to the firm s level of accruals, which may influence the level of DACs as well. The models used in the present study are modified from Baxter and Cotter (2009), Johl et al. (2007) and Chia et al. (2007). 5 Results and discussion 5.1 Descriptive statistics Table 2 provides descriptive statistics for alternative dependent variables and for the main independent variables used in equations 4 and 5. Table 2: Descriptive Statistics Panel A: Pooled sample Variable N Minimum Maximum Mean Standard Skewne Kurtosis Deviation ss DAC DAC DAC ABSDAC ABSDAC ABSDAC SIZE NEG OCF LEVERAGE CRISIS Variable definitions are as per equations 4 to 8. The means of the signed DACs are negative for all three proxies of earnings management. The means of the absolute value of DACs are 0.188, and 0.161, for ABSDAC1, ABSDAC2 and ABSDAC3, respectively. These measures compare to the means of the absolute value of DACs of 0.18 found by Baxter and Cotter (2009) and found by Davidson et al. (2005) for the Australian market. The skewness and kurtosis statistics suggest that some variables are not normally distributed. Therefore to remove any heteroskedasticity problems arising out of the non-normal distributions, all regressions are estimated with White-adjusted standard errors and t- statistics. 5.2 Correlation coefficients Table 3 presents the correlation coefficients between the dependent and independent variables used in each of the models. The Pearson s correlation coefficients are shown to the upper right of the diagonal, and the Spearman s Rho coefficients are shown to the lower left of the diagonal. The correlations between the three proxies for earnings management (ABSDAC1, ABSDAC2 and ABSDAC3) are positive and statistically significant. This implies that these proxies are capturing similar phenomena. The CRISIS dummy variable is positively correlated with all three measures of earnings management. This provides initial support for the proposition that the absolute levels of DACs were high during the GFC (compared to the PCP). 9

10 OCF has a negative correlation (although not significant) with all three measures of earnings management. This implies that if OCF increases, firms discretionary accruals decrease. Of particular note is the fact that, with the exception of OCF and NEG, the independent variables are not highly correlated to each other. This reduces concerns about multicollinearity problems impacting the precision of estimated coefficients. 5.3 Univariate test results Table 4 presents univariate test results for differences in the three measures of earnings management. Results for both the signed DACs and absolute values of DACs (ABSDAC) are reported. For the signed DACs, DAC1 and DAC2 and DAC1 and DAC 3 are not significantly different in terms of Wilcoxon signed-rank tests. Results from the paired sample t-tests for equality of the means suggest that DAC1 and DAC2 are significantly different. For the ABSDACs, both the non-parametric Wilcoxon signed-rank tests and the parametric paired-sample t-tests suggest no statistically significant difference between ABSDAC1 and ABSDAC2. However, ABSDAC3 is significantly different from each of ABSDAC1 and ABSDAC2. This may suggest that ABSDAC3 has different distributional properties than ABSDAC1 and ABSDAC2, and that the mean of ABSDAC3 is significantly different from the respective means of ABSDAC1 and ABSDAC2. Table 5 presents the univariate parametric and non-parametric tests results for the differences in the variables between the GFC and PCP. The means of all the three proxies of earnings management (ABSDAC1, ABSDAC2 and ABSDAC3) are higher during the GFC those during the PCP. The parametric independent sample t-test suggests that the difference in the proxies of earnings management between the GFC and PCP are significant, although the non-parametric Mann-Whitney test suggests that the differences are not significant for ABSDAC2 and ABSDAC3. 10

11 Table 3: Correlation coefficients between variables (Pearson correlation coefficients to the upper right of the diagonal and Spearman s Rho coefficients to the lower left of the diagonal) DAC1 DAC2 DAC3 CRISIS SIZE NEG OCF LEV DAC ** ** * ** ** DAC ** ** * ** ** DAC ** ** ** ** ** CRISIS * SIZE ** ** ** * ** NEG ** ** ** ** * OCF ** ** ** ** ** LEV ** ** ** ** ** ** ** ** Significant at the 1 per cent level; * Significant at the 5 per cent level. Variable definitions are as per equations 3 to 8. Note that DACs in the above refers to the absolute values of DACs (ABSDACs) Table 4: Differences in the different proxies for earnings management Wilcoxon signed rank tests: z-statistics Paired sample t-tests for the equality of the mean Signed DACs DAC1- DAC DAC1- DAC * DAC1-DAC DAC1- DAC DAC2- DAC ** DAC2- DAC Absolute value of DACs ABSDAC1- ABSDAC ABSDAC1 - ABSDAC ABSDAC1-ABSDAC ** ABSDAC1 - ABSDAC * ABSDAC2- ABSDAC ** ABSDAC2 -ABSDAC * ** Significant at the 1 per cent level; * Significant at the 5 per cent level. Variable definitions are as per equations 3 to 8. 11

12 Table 5: Differences in the earnings management Variable Mean Standard deviation Independent sample t- test (t-stat) Between the GFC and the PCP GFC PCP GFC PCP GFC vs. PCP Mann- Whitney test (zstat) GFC vs. PCP DECDAC ** DECDAC DECDAC ** ABSDAC ** ** ABSDAC ** ABSDAC *** SIZE * ** NEG OCF LEV * *** *** Significant at 1 the per cent level; ** Significant at the 5 per cent level; * Significant at the 10 per cent level. Variable definitions are as per equations 3 to 8. Similar results emerge when the mean statistics of DECDAC1, DECDAC2 and DECDAC3 are compared for the GFC and PCP. Firms engaged in a higher level of income-decreasing earnings management via negative DACs. However, the nonparametric tests suggest that the increase in the level of income-decreasing earnings management during the GFC is not statistically significant. This contrasts with the results of the independent sample t-tests, which suggest that the differences in the income-decreasing earnings management between the GFC and the PCP are significant for two of the three proxies of earnings management. With respect to the other control variables, the differences between the GFC and the PCP in the means of SIZE and LEV are significant. The significantly higher SIZE during the PCP is consistent with the decline in firms market value during the GFC as compared to the PCP. The significantly lower mean for leverage during GFC as compared to the PCP implies a decline in debt financing during the GFC relative to that undertaken during the PCP. 5.4 Multivariate regressions results Research question 1 Research question 1 examines if Australian firms engaged in a greater level of earnings management during the global financial crisis. To examine this question, equation 4 is estimated. The results for equation 4 are presented in Table 6. The model is significant at the 1 per cent level (F-statistic values are significant at the 1 per cent level for all the three proxies of earnings management). The Durbin-Watson statistics are within the 1.50 to 2.50 range, which suggests that no serious autocorrelation problem is present. To remove heteroskedasticity, the model has been estimated with White adjusted standard errors and t-statistics. 12

13 Table 6: Effect of the GFC on firms earnings management Variable DECDAC1 DECDAC2 DECDAC3 Intercept 1.254*** 1.303*** 0.763*** CRISIS *** ** ** Control variables OCF *** SIZE *** *** *** NEG LEV Industry dummies CD ** ** ** CS ENERGY *** *** *** HC IND ** *** * IT * * MAT ** ** ** Adjusted R-square F-statistic 5.323*** 4.507*** 5.240*** Durbin-Watson t-statistics appear under each coefficient. ***Significant at the 1 per cent level; **Significant at the 5 per cent level; * Significant at the 10 per cent level. Variable definitions are as per equations 3 to 8. The coefficients of the CRISIS test variable are significant and negative for all three proxies for income-decreasing earnings management. The statistically significant and negative coefficients of the CRISIS dummy variables suggest that the sample firms engaged in more income-decreasing earnings management during the GFC than in the PCP. The results from the estimation of equation 5 using the absolute value of DACs (ABSDAC1, ABSDAC2 and ABSDAC3) as the dependent variable are presented in Table 7. Statistically significant F-statistics suggest that the models are significant at the 1 per cent level. The values for the Durbin-Watson statistics suggest that there is no serious autocorrelation problem with any of the models. Results are similar to 13

14 those obtained from the use of the income-decreasing earnings management models. The coefficients of the CRISIS dummy variables are positive and statistically significant for all three proxies for earnings management. Table 7: Effect of the GFC on firms earnings management Variable ABSDAC1 ABSDAC2 ABSDAC3 Intercept *** *** *** CRISIS 0.072*** 0.067** 0.062*** Control variables OCF ** LOGMV 0.116*** 0.120*** 0.087*** NEG LEV ** Industry dummies CD 0.260*** 0.250*** 0.254*** CS 0.181** 0.161** 0.175** ENERGY 0.262*** 0.266*** 0.260*** HC IND 0.284*** 0.296*** 0.219*** IT 0.170** 0.165*** 0.164*** MAT 0.307*** 0.290*** 0.282*** Adjusted R-square F-statistics 7.200*** 6.479*** 7.604*** Durbin-Watson t-statistics appear under each coefficient. *** Significant at the 1 per cent level; ** Significant at the 5 per cent level; * Significant at the 10 per cent level. Variable definitions are as per equations 3 to 8. Amongst the control variables SIZE has positive and significant coefficients for all three earnings management proxies. This suggests that larger firms have a higher level of income-decreasing earnings management. This may be due to the higher level of operating activities for larger firms than for smaller firms. Pooling the cross-section and time series data increases the likelihood of violating the assumption of independence in the error terms. Panel data regressions are used to deal with the problem. Hausman specification tests suggest that a fixed-effects model is appropriate against the alternative of a random-effects model. Because the CRISIS dummy variable controls for period effects, a further robustness test for research question 1 is performed. This is by estimating equation 4 with a one-way cross-section fixed effect. Minutti-Meza (2013) suggests that firm specific fixed effect model mitigates the unobservable firm specific characteristics that are stable over 14

15 time and impact the outcome variable. Fixed effect model isolates the effect of the time-invariant characteristics from the predictor variables in order to assess the predictors net effect. Because the cross-sectional fixed-effects model controls for any firm-specific effects, the industry dummy variables are not included in this estimation of equation 4. Results are presented in Table 8. The adjusted R-square (in Table 8) of the fixed-effects model is higher than the ordinary least square models (in Tables 6 and 7). The CRISIS dummy variable is positive and statistically significant for all the three proxies of the dependent variable (earnings management). The above analysis suggests that the sample firms engaged in more incomedecreasing earnings management via discretionary accruals during the GFC than during the PCP. The results of the multivariate regressions corroborate those of the univariate parametric and non-parametric tests presented in Table 5. Table 8: Effect of the GFC on firms earnings management Variable ABSDAC1 ABSDAC2 ABSDAC3 Intercept CRISIS 0.071*** 0.060*** 0.057*** OCF 0.003*** 0.002*** 0.010*** SIZE NEG 0.046* LEV * Adjusted R-square F-statistics 2.397*** 2.917*** 2.160*** Durbin-Watson Cross-section fixed effect Chi square *** *** *** t-statistics appear under each coefficient. *** Significant at the 1 per cent level; ** Significant at the 5 per cent level; * Significant at the 10 per cent level. Variable definitions are as per equations 3 to 8. The findings relating to research question 1 are also consistent with those in the existing literature. For example, Chia, et al. (2007), examine a sample of firms in Singapore in the context of 1997 Asian financial crisis and find that their sample firms engage in income-decreasing earnings management during the Asian crisis period. Similarly, in the context of 1990 Persian Gulf crisis, Han and Wang (1998) investigated whether firms that expect increases in earnings resulting from product price increases use accounting accruals to reduce earnings. Their findings suggest that oil firms that expected to profit from the oil price increase during the Gulf crisis used accounting accruals to reduce their reported quarterly earnings. van Zalk (2010) examines the effect of the GFC on earnings management for a sample firms applying the earnings distribution approach proposed by Burgstahler and Dichev (1997) in the context of the code law France and common law UK. The 15

16 findings suggest that some firms in both countries adopted income-decreasing earnings management during the GFC as compared to the PCP. Hence, the findings of the present study are consistent with earlier studies in other crisis and other country contexts. Research question 2 Research question 2 examines the effect of firms industry identity affect the earnings management behaviour. With respect to the different industry dummies, consumer discretionary (CD), ENERGY, Industrial (IND) and material (MAT) have significantly higher level of income-decreasing earnings management during the GFC than the PCP (Table 6). Results from Model 5 (presented in Table 7) are similar to those reported in Table 6, excepting that the signs of the coefficients are positive. This is consistent with prior expectations, because of the use of the absolute values of DACs as the dependent variables. Thus our findings suggest that firms industry identity affects their earnings management behaviour. Our findings are consistent with Tse and Tucker (2009) who find that managers time their earnings announcement (with earnings shocks) to occur soon after their industry peers warnings to minimize their apparent responsibility for earnings shortfalls. Our findings are also consistent with the findings of Karuna et al. (2012) that industry factors play a vital role in influencing the variation in extent of earnings management across firms. Our findings imply that managers herd in managing earnings when peer managers in the same industry also manage earnings. Thus earnings management by one firms in an industry will act as warnings of earnings management by other firms in the same industry. 6. Conclusions The study is important for Australian investors, academics and policy makers. This study contributes to the earnings management literature. It extends the earnings management literature by examining whether managers engage in earnings management during a financial crisis. It also provides evidence of earnings management behaviour in a developed market, that of Australia, during the GFC. The analysis in the paper suggests that firms engaged in more income-decreasing earnings management during the GFC (2008 and 2009) than during the PCP (2006 and 2007). This was through greater use of discretionary accruals. We also find that firms industry classification has statistically significant effect on their earnings management. An objective of future research should be to examine whether these income-decreasing discretionary accruals were justified by prudence, in light of uncertainties regarding the impact of the GFC, or reflected opportunistic behaviour the taking of a big bath. However, this is beyond the scope of the current study. 16

17 References Ahmed, A., S. Rasmussen and S. Tse "Audit quality, alternative monitoringmechanisms, and cost of capital: An empirical analysis." Barth, M. E., W. H. Beaver, J. R. M. Hand and W. R. Landsman "Accruals, cash flows, and equity values." Review of Accounting Studies 4(3): Bartov, E., F. A. Gul and J. S. L. Tsui "Discretionary-accruals models and audit qualifications." Journal of accounting and economics 30(3): Baxter, P. and J. Cotter "Audit committees and earnings quality." Accounting & Finance 49(2): Beatty, A. and J. Weber "Accounting discretion in fair value estimates: An examination of SFAS 142 goodwill impairments." Journal of Accounting Research 44(2): Becker, C. L., M. L. DeFond, J. Jiambalvo and K. Subramanyam "The Effect of Audit Quality on Earnings Management." Contemporary accounting research 15(1): Beneish, M. D "Detecting GAAP violation: Implications for assessing earnings management among firms with extreme financial performance." Journal of Accounting and Public Policy 16(3): Bernard, V. L. and D. J Skinner "What motivates managers' choice of discretionary earnings management?" Journal of Accounting and Economics 22(1-3): Burgstahler, D. and I. Dichev "Earnings management to avoid earnings decreases and losses." Journal of accounting and economics 24(1): Chia, Y. M., I. Lapsley and H. W. Lee "Choice of auditors and earnings management during the Asian financial crisis." Managerial Auditing Journal 22(2): Choi, J. H., J. B. Kim and J. J. Lee "Value relevance of discretionary accruals in the Asian financial crisis of " Journal of Accounting and Public Policy 30(2): Davidson, R., J. Goodwin Stewart and P. Kent "Internal governance structures and earnings management." Accounting & Finance 45(2): DeAngelo, L. E "Accounting numbers as market valuation substitutes: A study of management buyouts of public stockholders." Accounting review: Dechow, P. M. and R. G. Sloan "Executive incentives and the horizon problem:: An empirical investigation." Journal of accounting and economics 14(1): Dechow, P. M., R. G. Sloan and A. P. Sweeney "Detecting earnings management." Accounting review: DeFond, M. L. and J. Jiambalvo "Debt covenant violation and manipulation of accruals." Journal of accounting and economics 17(1-2): Fiechter, P. and C. Meyer Big bath accounting using fair value measurement discretion during the financial crisis, Working Paper, University of Zurich. Francis, J., E. Maydew and H. Sparks "The role of Big 6 auditors in the credible reporting of accruals." Guay, W. R., S. Kothari and R. L. Watts "A market-based evaluation of discretionary accrual models." Journal of Accounting Research 34:

18 Gul, F. A., C. J. P. Chen and J. S. L. Tsui "Discretionary Accounting Accruals, Managers' Incentives, and Audit Fees*." Contemporary accounting research 20(3): Han, J. C. Y. and S. Wang "Political costs and earnings management of oil companies during the 1990 Persian Gulf crisis." Accounting review: Healy, P. M "The effect of bonus schemes on accounting decisions." Journal of accounting and economics 7(1-3): Healy, P. and J. Wahlen "A review of the earnings management literature and its implications for standard setting." Hribar, P. and D. W. Collins "Errors in estimating accruals: Implications for empirical research." Journal of Accounting Research 40(1): Iqbal, A. and N. Strong "The effect of corporate governance on earnings management around UK rights issues." International Journal of Managerial Finance 6(3): Jeter, D. and L. Shivakumar "Cross-sectional estimation of abnormal accruals using quarterly and annual data." Accounting and Business Research 29: Johl, S., C. A. Jubb and K. A. Houghton "Earnings management and the audit opinion: evidence from Malaysia." Managerial Auditing Journal 22(7): Jones, J. J "Earnings management during import relief investigations." Journal of Accounting Research 29(2): Karuna, C., K. R. Subramanian and F. Tian "Industry product market competition and earnings management." Paper presented at American Accounting Association 2012 mid-year conference. Khan, A. R., B. Balachandran and P. Mather "Managerial Share Ownership and Operating Performance: Do Independent and Executive Directors have Different Incentives? Australian Journal of Management. Volume 24. Kim, J. B. and C. H. Yi "Ownership structure, business group affiliation, listing status, and earnings management: Evidence from korea*." Contemporary accounting research 23(2): Kothari, S., A. J. Leone and C. E. Wasley "Performance matched discretionary accrual measures." Journal of accounting and economics 39(1): Krishnan, G. V "Does Big 6 auditor industry expertise constrain earnings management?" Accounting Horizons 17(1). Lawrence, A., M Minuti-Meza and P. Zhang (2013). "Can Big-4 versus Non-Big-4 Differences in Audit Quality Proxies Be Attributed to Client Characteristics?" The Accounting Review 86(1): Lim, S. and Z. Matolcsy "Earnings management of firms subjected to product price controls." Accounting & Finance 39(2): Louis, H "Earnings management and the market performance of acquiring firms." Journal of Financial Economics 74(1): Masters-Stout, B., M. L. Costigan and L. M. Lovata "Goodwill impairments and chief executive officer tenure." Critical Perspectives on Accounting 19(8): Miller, D. P., S. A. Mansi and W. F. Maxwell "Does auditor quality and tenure matter to investors? Evidence from the bond market." Journal of Accounting Research 42(4):

19 Minutti-Meza, M "Does Auditor Industry Specialization Improve Audit Quality?" Journal of Accounting Research 51(4): Sidhu, B. and H. C. Tan "The performance of equity analysts during the global financial crisis." Australian Accounting Review 21(1): Strong, J. S. and J. R. Meyer "Asset writedowns: Managerial incentives and security returns." The Journal of Finance 42(3): Subramanyam, K "The pricing of discretionary accruals." Journal of accounting and economics 22(1): Tse, S. and J. W. Tucker "Within industry timing of earnings warnings: do managers herd?" Review of Accounting Studies 15(): Watts, R. L. and J. L. Zimmerman "Towards a positive theory of the determination of accounting standards." Accounting review: Watts, R. L. and J. L. Zimmerman "Positive accounting theory: a ten year perspective." Accounting review: Wells, P "Earnings management surrounding CEO changes." Accounting & Finance 42(2): Woody, W "Management buyouts and earnings management." Xu, Y., A. L. Jiang, N. Fargher and E. Carson "Audit reports in Australia during the global financial crisis." Australian Accounting Review 21(1):