Earnings Management Dynamics in Portuguese Listed Firms

EARNINGS MANAGEMENT DYNAMICS IN PORTUGUESE LISTED FIRMS por Diogo Fernando Batista da Silva Tese de Mestrado em Finanças e Fiscalidade Orientada por António Cerqueira Elísio Brandão 2016 i Biographical note Diogo Fernando Batista da Silva finished his bachelor’s degree in Economics in the school of Management and Economics of the University of Porto in 2014, joining thereafter the Master in Finance and Taxation of the same institution. He worked in the Bank of Portugal as well as in the Ministry of Finance. At a research level, Diogo Silva developed analysis focusing on Portugal: “Can inventories predict production?”; “A survey-based measure of output gap for Portugal”; Qualitative Indicators of the Portuguese Financial Sector”; ii Abstract: This study examines earnings management dynamics in Portuguese listed firms. It provides evidence consistent with firms using accrual and real management as substitutes. Besides, the use of each strategy appears to depend on its relative cost. This investigation also points to the existence of a trade-off between derivatives use and the magnitude of earnings management. Furthermore, the impact of both earnings management and derivatives use on effective tax rates is examined. The results indicate that, if the tax function is convex, firms can reduce tax expenses by hedging with derivatives. This research also explores how earnings management practices are related with a set of market and financial incentives, while controlling for earnings management constraints. Meeting dividend thresholds seems to be the most relevant incentive for firms to manage earnings upward. Keywords: Earnings management; Derivatives; JEL classification: G32; M48; iii Resumo: Este trabalho analisa dinâmicas no âmbito da gestão de resultados nas empresas cotadas na bolsa de valores de Lisboa. Fornece evidência consistente com a existência de um efeito substituição entre 2 métodos alternativos, utilizados no âmbito da gestão de resultados: um baseado na flexibilidade das normas de relato financeiro e um outro relacionado com a estruturação da atividade da empresa. O recurso a cada um destes métodos está relacionado com o seu custo relativo. Adicionalmente, esta investigação aponta para a existência de uma relação inversa entre a utilização de produtos derivados e a magnitude da gestão de resultados. É ainda analisado o impacto destas práticas, bem como da utilização de derivados, nas taxas de imposto efetivas. Os resultados indicam que no contexto de convexidade fiscal, as empresas podem amenizar os encargos fiscais com recurso a derivados. Este estudo integra igualmente uma análise que visa examinar como é que a gestão de resultados está relacionada com incentivos financeiros e de mercado. De acordo com a análise empírica realizada, alcançar os resultados que permitem o pagamento do valor esperado de dividendos, é a motivação mais relevante neste contexto. Palavras-chave: Gestão de resultados; Derivados; Classificação JEL: G32; M48; iv Contents 1. Introduction ................................................................................................................ 1 2. Literature review and hypothesis development ......................................................... 4 2.1. Definition ................................................................................................................ 4 2.2. Accrual-based management vs real activities-based management ......................... 4 2.3. Tax incentives and derivatives ............................................................................... 5 2.4. Derivatives and earnings management ................................................................... 7 2.5. Financial and market incentives ............................................................................. 8 3. Methodology and sample selection .......................................................................... 10 3.1. Sample selection ................................................................................................... 10 3.2. Estimations at firm-level ...................................................................................... 10 3.3. Accrual-based management .................................................................................. 12 3.4. Real activities-based management ........................................................................ 14 3.5. Accrual-based management vs real activities-based management ...................... 15 3.6. Tax incentives ....................................................................................................... 19 3.7. Derivatives and earnings management ................................................................. 22 3.8. Financial and market Incentives ........................................................................... 23 4. Results ...................................................................................................................... 25 4.1. Accrual-based management vs real activities-based management ....................... 25 4.2. Tax incentives ....................................................................................................... 27 4.3. Derivatives and earnings management ................................................................. 32 4.4. Financial and market incentives ........................................................................... 34 5. Conclusion ............................................................................................................... 37 6. References ................................................................................................................ 39 7. Annex ....................................................................................................................... 49 v List of tables Table 1: The nominal tax rate and the marginal state tax rate by year (%) .................... 16 Table 2: Estimation results of equations (3.3) and (3.4) ................................................. 25 Table 3: Estimation results of equations (3.6) and (3.7) ................................................. 27 Table 4: Estimation results of equation (3.8) .................................................................. 30 Table 5: Estimation results of equation (3.9) .................................................................. 32 Table 6: Estimation results of equation (3.10) ................................................................ 34 Table 7: Variables measurement and sources (part 1) .................................................... 49 Table 8: Variables measurement and sources (part 2) .................................................... 50 Table 9: Variables measurement and sources (part 3) .................................................... 51 Table 10: Variables measurement and sources (part 4) .................................................. 52 Table 11: Descriptive statistics (part 1) .......................................................................... 53 Table 12: Descriptive statistics (part 2) ........................................................................... 54 Table 13: Pearson (below diagonal) and Spearman (above diagonal) pairwise correlation coefficients in percentage - variables in equation (3.3) ................................ 55 Table 14: Pearson (below diagonal) and Spearman (above diagonal) pairwise correlation coefficients in percentage - variables in equation (3.4) ................................ 56 Table 15: Pearson (below diagonal) and Spearman (above diagonal) pairwise correlation coefficients in percentage - variables in equation (3.6) and (3.7) (part 1) .... 57 Table 16: Pearson (below diagonal) and Spearman (above diagonal) pairwise correlation coefficients in percentage - variables in equation (3.6) and (3.7) (part 2) .... 58 Table 17: Pearson (below diagonal) and Spearman (above diagonal) pairwise correlation coefficients in percentage - variables in equation (3.8) ................................ 59 Table 18: Pearson (below diagonal) and Spearman (above diagonal) pairwise correlation coefficients in percentage - variables in equation (3.9) (part 1) ................... 60 Table 19: Pearson (below diagonal) and Spearman (above diagonal) pairwise correlation coefficients in percentage - variables in equation (3.9) (part 2) ................... 61 Table 20: Pearson (below diagonal) and Spearman (above diagonal) pairwise correlation coefficients in percentage - variables in equation (3.10) .............................. 62 1 1. Introduction If the main objective of financial reporting is to enlighten stakeholders so they can make efficient economic decisions, it is essential to assess the quality of financial information. Addressing earrings management practices is relevant in order to how the quality of information that is linked with executive discretionary options. Nowadays, given the complex context in which firms operate, earnings management may be seen has a continuous and iterative process characterized by a mixture of incentives and practices. There are several incentives to engage in earnings management, which managers need to weight. Likewise, there are different ways to drive earnings and cash flows in a certain direction. The choices between diverse types of earnings management are certainly not independent. Furthermore, puzzling earnings management gives the chance to anticipate the risk of a companies’ underlying information, given the presence of certain motivations and constraints. Therefore, this study is important to investors in general, borrowing institutions when assessing credit characteristics and auditors while evaluating audit-risk. It is also relevant to Portuguese market regulators, mainly CMVM1, but also ASF2 and Bank of Portugal since financial institutions are usually relevant stakeholders3 of Portuguese listed firms. In addition, it is pertinent for the Portuguese Tax Authorities given that this research addresses how earnings management and derivatives use are associated with effective tax rates. This study has increased relevance given that, regarding earnings management, literature that focused on Portuguese context is minor. Aiming Portuguese listed firms, it is possible to point the work of Mendes and Rodrigues (2006) that have analyzed earnings smoothing practices using Eckel approach (1981) and Alves (2011, 2012) that related discretionary accruals with board characteristics and ownership structure, respectively. An overall measure of real activities-based management have never been developed for Portugal and consequently never associated with accrual-based management. In a first stage of this research, it is tested the existence of a trade-off between the two strategies. Moreover, one tries to understand what drives the 1 Portuguese Securities Market Commission. 2 Insurance and Pension Funds Supervisory Authority. 3 They are usually both borrowers and shareholders of those firms. 2 preferences of the decision makers towards each earnings management scheme. Thereafter, a total measure of earnings management, that is, one that accounts for both strategies is applied in the examinations that follow. Then, it is analyzed how derivatives use may relate with earnings management decisions, because firms may smooth earnings by managing them (whether through accrual or real management strategies) or by hedging with derivatives. Furthermore, three kinds of earnings management drivers are studied, namely tax, financial and market incentives. Tax incentives are analyzed separately for methodological reasons. For that matter, effective tax rates are regressed in a set of determinants (firm specific and non-firm specific characteristics) to which proxies of earnings management, as well as a variable that captures the use of derivatives are added. Empirical literature analyzing hedging policies in Portuguese listed firms is also scarce. This study presents an empirically examination that clarifies why Portuguese listed firms hedge with derivatives. The main purpose of this analysis is to understand if tax convexity influence firms probability to use derivatives. In a final stage, it is investigated how earnings management are related with financial and market incentives. Specifically, it is tested if firms manage earnings upward to: report positive earnings (instead of losses); present increases in reported income (and not declines); diminish the cost of capital when contracting high amounts of long-term debt; meet dividend thresholds. This research contributes to the literature by computing both accrual and real activitiesbased management proxies at firm level. The typical industry-year estimations were not employed due to a small number of observations per industry. Besides, this is the first research in which a total measure of earnings management is related with derivatives use. Moreover, it is the first to present evidence that hedging policies and earnings management practices may be defined sequentially. It also contributes to the literature related with taxation by demonstrating empirically that firms can reduce tax expenses through the use of derivatives if the tax function is convex. Furthermore, this is the first investigation to provide empirical proof that firms’ hedging policies are influenced by tax convexity. In addition, this study looks at the impact of firms’ ability to manage earnings on effective tax rates, both through accruals and real activities. One methodological contribution of this research is the proxy for the cost of capital. It has never been used in corporate finance literature; however, it is a complete measure of the 3 costs of financing because it captures not only interest rates but also other credit characteristics, such as covenants and collateral requirements. Finally, when studying firms’ financial and market incentives to manage earnings, this study differentiates from other researches because both kinds of earnings management are considered, a reliable variable to control for earnings management constraints is used and the various motivations are tested at the same time, so that one can understand the relative importance of each one. To develop the analysis mentioned a panel data set with 38 non-financial Portuguese listed firms was used. The time period starts with the adoption of International Financial Reporting Standards (IFRS) and ends in 2014. This study is organized as follows: section 2 presents the literature review as well as the research hypothesis; in section 3 the methodology is described; section 4 provides the results and its analysis; per last, section 5 presents the concluding remarks. 10 3. Methodology and sample selection 3.1. Sample selection This study focuses on firms listed on Lisbon stock exchange in the year of 2014. The sample starts in 2004 and ends in 20141. Observations for 2004 are used only when data available is based IFRS2. Firms excluded from the study are: (1) financial institutions (sic code 60-67); (2) firms in which the main activity is related with servicesmembership sports and recreation clubs (sic code 7997); (3) firms with no available information in Datastream3; (4) Firms listed during only 1 year. As a result, this analysis uses 38 non-financial Portuguese firms listed in the Lisbon stock exchange in 2014. 3.2. Estimations at firm-level The abnormal component of both the accruals and the cash-flow from operations are used to proxy accrual and real management, respectively. Regarding this approach, this study has two limitations, which are immediate constraints of focusing the study on Portuguese listed firms. Firstly, there is a reduced number of firms, when compared with other studies, such as the ones of Roychowdhury (2006) and of Zang (2011), which were important benchmarks for this research. Secondly, the models used to calculate abnormal accruals (or abnormal cash-flows) are usually estimated for each industry-year. However, the number of listed firms for each industry is usually very small. Hence, the usual methodology cannot be employed, given the common five to twenty minimum observations threshold is not met. In fact, the industries with more observations are composed by 3 firms (when defining industries with 2 digits sic codes). These are two structural characteristics of the Portuguese capital markets. Thus, to study earnings management dynamics in Portugal, the usual approaches need to be adjusted. In order to surpass the constraints related with Portuguese environment, firm specific estimations are applied. Firms’ specific samples start when financial reporting based on 1 Variables definition, measurement and sources are available in the Annex (Table 7-10). 2 Indeed, the sample start is predetermined by the adoption of IFRS. 3 Datastream is the main data source of this study. 11 IFRS is available, which for most firms is 2005 (the year of mandatory adoption4). The sample ends in 2014, which implies 10/11 observations for most firms. Both firm and industry specific attributes play an important role in abnormal accrual estimations. There is a component of accruals that relates with firm characteristics, such as operations volatility (Dechow and Dichev, 2002), uncertainty (Palepu et al., 2000), internal control characteristics (Kinney and McDaniel, 1989) and other factors that may lead to misjudgments. Only a second component of abnormal accruals relates with possible opportunistic use of accounting rules. For example, omitted variables bias is an important issue in accrual-based estimations (Dechow et al., 2010). Therefore, there is value added in the use of firm specific estimations. The trade-off behind firm-level estimations relates with difficulties in capturing industry shocks. For example, if there is a relevant increase in a production input for a certain industry, the typical industry-year approach to estimate discretionary accruals, are more likely to be unresponsive to it. A firm specific procedure is probably weaker in this scenario, and would most likely predict abnormal accruals, which weakens the robustness of the estimations that use abnormal accruals to proxy for earnings management. Dechow et al. (2010) have also discussed this approach and stated: “We emphasize that all of the accruals models can be estimated at the firm level”. According to the authors, estimations at firm level have the advantage of allowing for variation in the coefficients of the determinants of the normal levels of accruals within firms, however, those models have invariant time parameters and possibly survivorship bias. They also highlight that, in the case of the usual industry-level estimations, some firms may be associated with higher residuals because of industry classification and not because of earnings management practices or errors. In addition, listed firms usually play a role in different sectors. Therefore, the use of industry classifications may also soften the power of the tests. A brief discussion of alternatives: (1) a possible substitute procedure to the firm specific approach would be to make industry-year estimations using 1 digit sic codes, which does not solve the usual observations threshold and still some firms need to be dropped. Moreover, it upturns industry-level estimations weaknesses, because one would have 4 In 2005 firms had to present financial reports based on IFRS, not only for 2005 but also for 2004, so that comparison with the previous year could be made. Nonetheless, Datastream does not have data for the year of 2004 for all firms. 12 considerably different firms with the same coefficients for the determinants of both normal accruals and cash-flow from operations. (2) Estimating the models crosssectionally using the whole sample should boost to the same bias, since listed firms are significantly heterogeneous, which is clear when considering the distribution of firms by industries. Therefore, this procedure does not yield any of the advantages mentioned before. Nonetheless it can capture shocks that affect most firms, but this can also be achieved through time effects in the models that follow. Specifically, in this study, estimations of earnings management are used has inputs in other models, in which is possible to account for dynamics such as time related effects. (3) Another alternative is to include public firms from other European countries in the estimations of both abnormal accruals and cash-flows, in order to achieve the minimum number of observations. This way, the usual methods could be applied. Nonetheless, firms from other countries may operate in a very a different context due to economic, institutional and regulatory differences. Moreover, one of the main purposes of this study is to understand the Portuguese specific environment, providing insights that yield value for Portuguese researchers and policy makers, given the scarcity of empirically robust investigations regarding earnings management. (4) Another option corresponds to the use of the biggest non-listed Portuguese firms for each industry and apply industry-year estimations. Nevertheless, the differences among listed and non-listed are probably significant. Most financial reports of non-listed firms are based on Portuguese GAAPs, while listed firms use IFRS. Overall, none of the alternative options are expected to produce more robust estimates. The use of firm specific estimations has highlighted has both advantages and disadvantages when compared to the usual industry level analysis. Since recent literature that uses this approach is slight, a contribution is made, by presenting evidence on the use of firm-level estimations. 3.3. Accrual-based management Since Healy (1985), models to estimate the normal level of accruals have been developed. The Jones model (1991), extracts the abnormal component of accruals, using both changes in sales and property, plant and equipment. Later, the model was adjusted by Dechow et al. (1995), so that credit sales are not included as a determinant of normal 13 accruals, since credit sales may be related with earnings management. This is the model applied in this study. Other models have built on the modified version of the Jones model, such as the procedure presented in Kothari et al. (2005), which is not used in this analysis, mainly because it is not possible, given the heterogeneity between the firms under analysis, to identify a similar firm in the same industry with the closest return on assets. The Modified Jones with control variables concerning the asymmetric recognition of gains and losses (Ball and Shivakumar, 2006) is not applied since it leads, in some cases, to multicollinearity, if a firm-level approach is employed. The modified Jones model merged with the Dechow and Dichev (2002) methodology, as suggested by McNichols (2002), can only be used to compute normal current accruals. Moreover, it consumes more grades of freedom and implies that accruals must revert in the following year, which may not always be the case. Likewise, the estimations of abnormal accruals are used to regress in other variables, which gives the opportunity to account for measurement error with adequate controls later5. Discretionary accruals are proxied by the residuals of the following model, which is based on the modified version (Dechow et al., 1995) of Jones (1991) model: ttttt PPEARkREVTAC   *)*)1((* 210 (3.1) Where for each year t, TAC corresponds to total accruals6, ∆REV refers to the change in revenue between t-1 and t, ∆AR is the change in accounts receivable, PPE corresponds to the value of property, plant and equipment. The error term is represented by εt. All variables are scaled by lagged total assets. K corresponds to the slope coefficient of a regression of ∆AR on ∆REV, when the p-value of the coefficient is lower than 25% (0 otherwise)7. Thus, only a portion of credit sales are related to earnings management, but 5 This is not the case when researchers use abnormal accruals has an immediate measure of earnings quality. 6Total accruals are computed as the difference between net income and cash-flow from operations. Alternatively they can be calculated as the change in net current assets minus change in cash plus change in current portion of long term debt minus depreciations. However, not all firms disclose information about the current portion of long term debt, hence, it cannot be computed. 7 The purpose is to include a component of credit sales as a determinant of normal accruals, but only if credit sales are explained by changes in sales. Because the regressions are control lacking, the significance level defined is softer. 14 only if there is some evidence that changes in sales explain the change in credit sales. The residual of equation (3.1) is used to proxy for accrual management. 3.4. Real activities-based management It is usually proxied by overproduction, abnormal discretionary expenses or abnormal cash-flows from operations (Roychowdhury, 2006; Zang, 2011; Cohen and Zarowin, 2008; Cohen et al., 2008), even so, some researchers only use abnormal research and development expenses (Baber et al., 1991; Dechow and Sloan, 1991; Bushee, 1998; Bens, 2002; Kothari et al., 2012). Overproduction may not be adequate to compute for some firms in the services sector. One alternative is to drop firms from these industries, which is not possible given the number of firms under analysis is not vast. Likewise, in these circumstances, the conclusions of this research would yield lesser value. Discretionary expenses are commonly defined as the sum of research and development expenses, selling and general administrative expenses and advertisement expenses. Nonetheless, one factor that distinguishes Portuguese firms from others is the propensity to present the income statement in a nature based and not in a functional form. Therefore, the components of discretionary expenses, including selling and general administrative expenses, are generally not disclosed and consequently not available in Datastream. As a result, in this study, real management is proxied by abnormal cash-flow from operations. The normal level of cash-flow from operations is estimated with the following model (Roychowdhury, 2006): tttt REVREVCFO   ** 210 (3.2) Where for year t, CFO corresponds to cash-flow from operations, REV refers to revenue and ∆REV is the change in revenue from the previous year. All variables are scaled by lagged total assets. The residual of equation (3.2) equals the abnormal cash-flow from operations. 15 3.5. Accrual-based management vs real activities-based management The dynamics between both kinds of earnings management are studied using Zang (2011) setting as baseline. Specifically, the following regressions are applied: tty y txkt CONTROLSDEMRM   , , 2,,10 *_* (3.3) tttyytxkt RMCONTROLSDEMAM   *_* 3,,2,,10 (3.4) Where for year t and firm i, RM corresponds to real management and is proxied by the residual of equation (3.2). AM refers to the amount of accrual management and is quantified by the residual of equation (3.1). EM_D are a set of determinants of each type of earnings management. CONTROLS are control variables. Zang (2011) only included suspect firms-years, that is, the ones most likely to be managing earnings, while controlling for non-random sample bias with Heckman two step procedure. The author defined suspect as firm-years in which earnings are barely above last year reported income, just beating the null benchmark or the analysts’ consensus forecasts. Indeed, many researchers have found indications of discontinuity in the distribution of earnings, especially in firm-years with zero to small earnings (Hayn, 1995; Burgstahler and Dichev, 1997). Even so, studies such as the one of Dechow et al. (2003) were not able to provide reliable evidence that firms manage earnings to avoid losses. Therefore this “a priori” hypothesis of Zang (2011) may also be costly, mainly if firms just beating zero earnings benchmarks are firms managing earnings downwardly, in order to, for example, minimize tax expenses. Likewise, in the presented analysis, observations cannot restrict to firm years in these circumstances, given in this case the sample would be reduced to 64 observations8. 8 64 observations equals the sum of 35 firm-years in which the change in earnings per share are between 0 and 2 cents and 29 firm-years in which return-on-assets is between 0 and half percent. 16 The earnings management determinants considered are: i.The marginal tax rate (MTR) is assumed to restrict real management because those practices are more likely to lead to higher costs related with taxation. When managing income upward through accounting-based methods, firms are able to increase earnings but not necessarily taxable income (TI). In addition, firms are able to defer possible additional tax expenses. Evidence of that is the study of Philips et al. (2003, 2004) which hints deferred tax expenses can be used to detect earnings management. MTR is set to zero if the firms’ pre-tax income is non-positive. In other cases, MTR equals the nominal tax rate (TR) plus the state tax rate (STR). Table 1: The nominal tax rate and the marginal state tax rate by year (%) Year TR STR 2005 27.5 0 2006 27.5 0 2007 25 0 2008 25 0 2009 25 0 2010 25 2.5 if TI ϵ ]2M, ∞[, 0 otherwise. 2011 25 2.5 if TI ϵ ]2M, ∞[, 0 otherwise. 2012 25 3 if TI ϵ ]1.5M, 10M], 5 if TI ϵ ]10M, ∞[, 0 otherwise. 2013 25 3 if TI ϵ ]1.5M, 7.5M], 5 if TI ϵ ]7.5M, ∞[, 0 otherwise. 2014 23 3 if TI ϵ ]1.5M, 7.5M], 5 if TI ϵ ]7.5M, 35M], 7 if TI ϵ ]35M, ∞[, 0 otherwise. M=€Millions; ii.Insider ownership (INSIDER) equals 1 if the ratio between the shares owned by insiders and common shares outstanding are above sample median (0 otherwise). The literature usually indicates that insider ownership restricts earnings management because it may reduce agency conflicts (Jensen and Meckling, 1976) and opportunistic behavior (Warfield et al., 1995). In practice, the process in which managers maximize their wealth is less likely to be costly to shareholders if managers’ incentives are aligned with shareholders’ interests. For example, Healy (1985) suggested that managers with low ownership may be more disposed to manage earnings upward to increase compensation (if compensation is a function of earnings). Nevertheless, the entrenchment hypothesis signals that insider ownership may also prompt self-interested actions (Cornett et al., 2008). In this analysis, the purpose is to understand how insider ownership relates with 17 each kind of earnings management. It is expected that higher insider ownership biases the preferences of managers towards accrual management, because real management are more likely to negatively impact on operational efficiency and future competitiveness, affecting managers that are relevant shareholders. Accrual management is a more fictitious approach to drive earnings. As enhanced, even its impact on tax expenses can be more easily dispatched or postponed. Since RM occurs during the year, insider ownership is lagged in equation (3.3). In equation (3.4), the contemporaneous value is included, since AM is expected to take place mainly at the end of the year or to be responsive to RM intensity. iii.Financial health (HEALTH) is projected to be related with higher operational flexibility (while the inverse also applies), therefore, low HEALTH should act as a real management constraint. In addition, firms closer to bankruptcy have more incentives to manage earnings upward, since they are trying to stay alive (although, in the sample used, there are not a large number of firms in a position of significant financial distress). Those firms are expected to use accrual management to increase reported earnings. This premise is consistent with evidence presented by Roychowdhury (2006) and Zang (2011). The variable is lagged in equation (3.3) and contemporaneous in equation (3.4). HEALTH is computed by the ZSCORE of Altman modified model (1968, 2000): it it it it it it it it it it it TL MCAP TA WC TA EARR TA REV TA NI Z*6.0*2.1 _ *4.1*0.1*3.0  (3.5) Where for each year t and firm i, TA is total assets, NI corresponds to net income, REV refers to revenue, R_EAR represents retained earnings, WC is working capital, MCAP denotes market capitalization and TL designates total liabilities; iv.The effect of board size (BOARD) on earnings management is rather unclear. Nevertheless, it is important to understand the outcome of the resources that are applied with the enlargement of the board. Bigger boards are supposed to sum more experience and knowledge. However, crowded boards may lead to inefficiencies related with difficulties of communication and coordination (Jensen, 1993). Mather and Ramsay 18 (2006) and Ching et al. (2006) concluded that board size is positively related with accrual management. It is expected that directors are particularly concerned about strategic decisions and less with accounting issues. As a result, they should more easily detect real management. In this study board size is the ratio between the number of directors and lagged total assets. v.In order to proxy for higher quality auditing a dummy variable (BIG4) that equals 1 if the firm is audited by a Big4 (0 otherwise) is employed. The literature has shown that Big4 auditing relates with lesser accrual management (Becker et al., 1998; Francis et al., 1999), even though, Lawrence et al. (2011) presented indications that differences in accrual management estimations of Big4 and non-Big4 clients are related with clients’ characteristics. Zang (2011) concluded that Big4 auditing is negatively related with real management and positively with accrual management and those results are expected to hold for Portuguese listed firms. Overall, BOARD is projected to be positively associated with accrual management. vi.Firms with lengthier operational cycles (CYCLE) should more easily rely on accrual management strategies since they will have both larger accrual accounts and longer periods for accruals to reverse9 (Zang, 2011). This variable is first quantified as in Dechow (1994)10 and then transformed into a dummy variable11 that equals 1 if a firm has an operational cycle above the whole sample median and 0 otherwise. The variable is lagged in equation (3.3) and contemporaneous in equation (3.4). vii.Barton and Simko (2002) indicated that net operating assets (NOA) incorporate previous accounting decisions, such as earnings management practices that rely on accounting principles’ flexibility. If a firm managed earnings upward in the past, current net 9 One must also highlight that firms with lengthier operational cycles may be more exposed to shocks, which may lead to higher volatility in cash-flow from operations and in the accruals. Therefore, firms with lengthier operational cycles may be related with more abnormal accruals (or cash-flow from operations) not because of earnings management. 10 The length of the operational cycle equals the days receivables are outstanding plus the number days products stay in stocks minus the number of days accounts payable are outstanding. 11 The use of dummy variables incorporates important advantages by automatically accounting for outliers and non-linear relations between the variables. 19 operating assets will be overstated. It is expected that overstated net operating assets act as a constraint on current accrual management strategies. It is measured as total assets minus cash minus total liabilities plus total debt. This measure is then scaled by lagged sales. NOA is a dummy variable that equals 1 if net operating assets are above sample median at the beginning of the period (0 otherwise). viii.CONC is a dummy variable that equals 1 if no shareholder owns more 25% of firms’ shares (0 otherwise). It is used to proxy for ownership concentration. Although large shareholders are the ones with more incentives to monitor managers (Dechow et al., 1995), which should constrain earnings management, the literature has usually found an empirical positive relationship between accrual management and ownership concentration (Choi et al., 2004; Alves, 2012; Rad et al., 2016). This is consistent with large shareholders pressuring managers to maximise their wealth, which may be translated in a higher magnitude of earnings management (Jaggi and Tsui, 2007). As a result, CONC is expected to be negatively related with accrual management. The variable is lagged in equation (3.3) and contemporaneous in equation (3.4) Per last, the CONTROLS considered are: the lagged market-to-book (MTB) ratio to contemplate firms’ growth opportunities; the natural logarithm of the sum of market capitalization with preferred stock, minority interests and total debt minus cash, at the beginning of the year, proxies for size (SIZE); Return-on-assets is included but it is net of accrual management in equation (3.4) and net of total earnings management in equation (3.3) (ROA_AM, ROA_EM); cash flow from operations scaled by lagged total assets is included in equation (3.3) to address possible measurement errors. In addition, both equations include year indicators and computed are with generalized least squares cross-section weights. 3.6. Tax incentives To investigate how both kinds of earnings management and derivatives use impact on tax expenses, the following models are applied: 26 The results (Table 2) point that insider ownership (INSIDER) acts mostly as a real management constraint, still it is positively related with accrual management. This indicates that when executives have a relevant fraction of the firms’ stock, they prefer not to engage in practices that may affect their future wealth and is in accordance with the agency theory. The signal and significance of the coefficients for MTR are consistent with accrual management being more desirable from a tax-based standpoint. When managing earnings upward, through accrual management, firms may not affect tax expenses or be able to defer additional costs. The results point that firms with lower HEALTH are more likely to use real management stratagems, which is the opposite to the predictions presented. Firstly, it is important enhance that the variable HEALTH does not differentiate firms in financial distress from others. It actually compares the 50% better with the 50% worst. The outcome hints that firms with lower HEALTH are more motivated to increase cash resources than just to report higher earnings. For example, if a firm is closer to be in distress, then it may be more interested in making sales in which receivables will be outstanding for shorter periods, giving its clients less time to pay, at the cost of a lower turnover. Interestingly, boards with more directors (BOARD) are associated with accrual-based strategies. This hints that an increase in the number of directors may lead to an increase in bureaucracy, which diminishes directors’ monitoring abilities, generating space for accrual-based approaches. One should highlight that more directors do not meaningfully imply less real management, just more accrual management. Thus, the evidence suggests that from an earnings management perspective, on average, more directors do not entail better monitoring. It was projected that firms with lengthier operational cycles (CYCLE) would have more space to engage in accrual management. Nonetheless, it also stimulates real management, according to the results obtained. Still, firms with lengthier operational cycles, such as firms from the real-estate sector, may have more volatile cash-flows and as a result, being associated with higher levels of earnings management1. As expected Big4’ clients have lower levels of discretionary accruals, that is, rely more in accounting-based strategies. Moreover, this variable relates positively with real management. Net operating assets (NOA), does not relate with earnings management in the way it was anticipated. Nevertheless, the variable is not statistically relevant in both 1 This highlights the relevance of Dechow et al. (2012) methodology to assess earnings quality. 27 equations, amid of the specification design presented2. Per last, the purpose of equations (3.3) and (3.4) was to test the existence of a trade-off between accrual and real activities-based management (H1). In fact, the coefficient of RM in equation (3.4) is negative, which confirms H1, at 99% confidence level. As a robustness check, AM and RM were transformed in dummy variables (AM_D and RM_D) that equal 1 if accrual or real management are positive (0 otherwise), respectively. In this scenario, H1 is still confirmed. The model was also re-estimated with robust least squares to check if measurement errors in the endogenous variables could be driving the main results. The conclusions hold both qualitatively and quantitatively. Tough, these results do not imply that there is not also a simultaneous decision process. Probably, both dynamics play a role. However, managers are likely to prefer to firstly use real management while being able to apply accrual management strategies if necessary. Overall, the results show the relevance of considering both strategies in subsequent researches. Moreover, regulation that restrains accounting-based strategies may be compensated for more intensive realmanagement strategies. 4.2. Tax incentives Table 3: Estimation results of equations (3.6) and (3.7) Predicted sign ETR1 ETR2 CONSTANT 0.1695*** 0.0006 CAP_INTt - -0.0793*** -0.0598* INV_INTt + -0.0584 0.0604 LEV_Dt-1 - -0.0219*** -0.0329*** INSIDER t-1 + 0.0004 -0.0283** CONC t-1 + 0.0063 0.0348*** BOARDt - -361.22 -1227.1*** BIG4t - 0.0109 -0.0452*** ROAt + 0.1864** 0.4596*** 2 In untabulated results, it was possible to confirm EM is negatively related with a dummy variable that equals 1 if net operating assets are on the top quartile (0 otherwise). This implies that net operating assets acts as an earnings management constraint only when net operating assets are significantly overvalued. 28 SIZEt-1 + 0.0059** 0.0129*** MTBt-1 + 0.0021 7.32E-5 AM_FLEXt ? -0.6959*** 0.6422** RM_FLEXt ? 1.3171*** 0.5878 CONVEXt*DERIVt - -0.0080*** -0.0034*** Year dummies Yes Yes Adjusted R-squared 0.8146 0.2189 Prob (F-stat) 0.0000 0.0000 DW Stat 1.2234 1.3319 Total panel obs. 342 342 *, **, and *** indicate significance at the 10%, 5% and 1% level, respectively; According to the results presented in Table 4, capital intensive (CAP_INT) firms have lower effective tax rates. Nevertheless, this variable is less relevant when tax expenses are consider as proportion of cash-flow from operations. INV_INT relates differently in the two equations, while not statistically significant in none of them. This outcome may be the result of not considering expenses related with research and development. Leverage (LEV_D) is negatively associated with effective tax rates in both equations, which translates tax savings related with indebtedness. As expected, profitability (ROA) relates positively with ETRs. These results are also consistent with bigger firms facing political costs that can be perceived through the tax system, since SIZE’s coefficient is positive and statistically significant in both equations. Interestingly, it is possible to verify that when tax expenses are measured as a proportion of pre-tax income, non-firm specific characteristics (BOARD, CONC, INSIDER, and BIG4) are not statistically relevant. However, they appear to play an important role when cash ETR is used (ETR2)3. This may be indicating that firms are primarily concerned with the impact of taxation on their cash resources. The results point that, at 99% confidence level, boards with more directors are able to soft cash ETRs. In addition, firms in which no shareholder has a large influence (CONC), tend to have higher cash ETRs. This suggests that in those circumstances, managers are less likely to be compelled by shareholders to 3 By using the cash-flow from operations as denominator when computing ETRs, one can control for differences that arise from the use of different accounting methods. 29 engage in tax management or that minority shareholders have less incentives to monitor management’s efficiency. In opposite, firms in which insiders have higher ownership (INSIDER) have lower cash ETRs (ETR2). This is consistent with managers being motivated to reduce taxation impact on firms’ cash when they are more relevant owners, directly benefiting from the reduction of tax expenses. Overall, the impact of BOARD, CONC and INSIDER on cash ETR is in line with the results of Ribeiro et al. (2015) when studying firms listed on the London stock exchange. As projected, by providing tax services, Big4s can induce relevant tax savings. On average, Big4s are able to reduce their clients cash ETRs by 4.5%. It could be important for the Tax Authority to understand how this is achieved, that is, how BIG4’ clients were being benefited. RM_FLEX is positively related with both measures of effective tax rates, which points that when managing earnings through real activities, firms are usually not motivated by tax incentives. The interpretation of the coefficient of AM_FLEX is not that clear, since the variable is statistically significant in both equations but with opposite signs. Therefore, it is not possible to conclude if firms engage in accrual management strategies with the purpose of paying fewer taxes. Per last, the coefficient of CONVEX*DERIV is negative and statistically significant (99% confidence level) in both equations. This shows that as the tax function becomes more convex, tax savings related with the use of derivatives increase, as suggested by Smith and Stulz (1985). If DERIV and CONVEX are included in the equation separately, that is, by not making an interaction between them, none of the two has a statistical relevant explanatory power. In other words, according to the results, firms can directly4 reduce tax expenses by hedging, if the tax function is convex (H2). 4 As an example, firms can indirectly reduce tax expenses with derivatives by reducing risk due to the reduction in the volatility of cash-flows. This way, firms are able to reduce the probability of bankruptcy. As a result, firms may reduce the cost of capital and increase debt capacity, generating tax savings related with indebtedness (Graham, and Rodgers, 2002). 30 Table 4: Estimation results of equation (3.8) Predicted sign DERIV Coefficients Marginal effects CONSTANT -4.9852 -1.9888 LTDEBTt + 2.45E-5*** 9.76E-6 STDEBTt - -2.24E-5 8.94E-6 ISALESt + 0.9719* 0.3874 HEALTHt-1 - 1.1060*** 0.4412 CYCLEt-1 + 0.8474** 0.3381 BOARDt + -30190* -12044 DYt-1 + 9.8936** 3.9470 ROAt ? 3.6622 1.4610 SIZEt-1 ? 0.1525 0.0609 MTBt-1 ? 0.1710** 0.0682 CONVEXt-1 + 0.0204** 0.0082 Year dummies Yes McFadden R-squared 0.3982 Prob (LR stat) 0.0000 Obs. correctly predicted 83.11% Total panel obs. 296 *, **, and *** indicate significance at the 10%, 5% and 1% level, respectively; If firms can reduce tax expenses by hedging with derivatives, when the tax function is convex, one shall analyse if firms hedge due to tax convexity. That is the purpose of estimating equation (3.8). Table 5 shows firms’ probability to use derivatives increases with long term debt (LTDEBT). This implies that firms’, during the sample period, have been concerned with interest rate risk. In addition, international sales (ISALES) are positively related with hedging, which is consistent with firms managing exchange rate risk. Strangely, healthier firms are the ones more likely to hedge with derivatives and HEALTH’s coefficient is statistically significant at 99% confidence level. From a theoretically point of view, the more distressed firms would be the ones expected to 31 hedge, because they are more sensible to shocks and by hedging, they could diminish bankruptcy probability. In addition, by hedging those firms could smooth earnings and consequently creditors’ perception of risk. The results may, however, be linked with transaction costs associated with these instruments and with the existence of alternative strategies to smooth earnings, which might be less costly. Indeed, firms may smooth earnings through earnings management strategies (this will be discussed later, Table 5). As projected, firms with lengthier operational cycles (CYCLE), face more risks, thus, hedge more. The variable BOARD relates negatively with hedging, which is opposite to the predictions. That implies that the effect of additional experience and knowledge on the board is more than compensated by the increase in bureaucracy and in difficulties to approve complex decisions. Nonetheless, once more, one should highlight that there may be other procedures to reduce earnings volatility. The results presented in Table 2, indicated that firms with more directors on the boards tend to use more accrual management. Besides, the correlation between BOARD and both AM_FLEX and RM_FLEX, is significant and positive (see Annex, Table 11). Overall, the signal of the coefficients of the variables HEALTH and BOARD demonstrate the importance of studying the existence of a trade-off between earnings management and hedging policies. The results obtained show firms that pay higher dividends (DY), tend to use derivatives. Those firms need to maintain higher levels of cash, so they can pay higher dividends. They are probably hedging to safeguard that their cash resources do not fall short of dividend payments. The importance of paying expected dividends is also discussed later (Table 6). Finally, the main purpose of the regression was to understand if hedging policies were responsive to tax convexity. At 95% confidence level, H3 is confirmed. An increase of one percentage point in convexity increases the probability of derivatives use by 0.8%5. Overall, the results point that firms consider tax convexity when deciding to hedge. Furthermore, the model appears to be reliable given it correctly classifies firms-years as hedgers or non-hedgers 83.11% of the time. 5 Marginal effects are computed at the mean. 32 4.3. Derivatives and earnings management Table 5: Estimation results of equation (3.9) Predicted sign |EM| C 0.0261** HEALTHt-1 - -0.0025 CYCLEt-1 - -0.0017 BOARDt - -201.8** PAYOUTt - -0.0002 CONVEXt + 8.58E-5* LEV_Dt-1 - -0.0029** AM_FLEXt + 0.3324*** RM_FLEXt + 0.2161** ROA_EMt - -0.0566*** MTBt-1 ? 0.0004 SIZEt-1 - -0.0003 |CFO|t + 0.0036 DERIVt - -0.0076*** Year dummies Yes Adjusted R-squared 0.5256 Prob (F-stat) 0.0000 DW Stat 2.0317 Total panel obs. 299 *, **, and *** indicate significance at the 10%, 5% and 1% level, respectively; Table 6 should be examined together with Table 5, because it is expected that firms first decide on the use of derivatives and then adjust earnings management strategies to smooth earnings. Therefore, equation (3.8) (Table 5) represents the first step of the decision making process and equation (3.9) (Table 6) the second one. If taken together, the results in Table 5 and 6 suggest the more distressed firms are more likely to manage earnings to smooth reported income than to use derivatives, which may be explained, 33 for example, by distressed firms’ disability to pay transaction costs. BOARD is statistically relevant, but negatively associated with smoothing practices, whether based on earnings management or derivatives, that is, crowded BOARDs are less likely to use strategies to smooth earnings, whether through earnings management engagements or by hedging with derivatives. This hints the existence of limitations related with more populated boards. Furthermore, this does not appear to be the result of better monitoring because BOARD is positively associated with accrual management practices to increase earnings (Table 2). The more leveraged firms (LEV_D) hedge more. As a result, they rely less in earnings management practices to reduce earnings volatility. In addition, as expected, tax convexity (CONVEX) is positively associated with earnings management practices to smooth earnings. Intuitively, firms able to manage earnings, whether through accruals (AM_FLEX) or real activities (RM_FLEX), tend to use more these strategies to smooth reported income. The main purpose of this analysis was to test the hypothesis that when firms use derivatives, they are likely to have smoother earnings, and consequently, will need to manage earnings to a lesser extent, with the objective of smoothing reported income (H6). DERIV is statistically relevant (at 99% confidence level), which implies that, when firms use derivatives, on average, total absolute earnings management, measured as a proportion of total assets (at the beginning of the year), decline by 0.7 percent. Still, it would be rather normal that, due to higher volatility, non-hedging firms have more abnormal accruals or cash-flows. Nevertheless, in the estimations, |CFO| and ROA_EM should already be controlling for that. Likewise, if a dynamic approach is employed, that is, when the lagged value of |EM| is included as an explanatory variable the results do not change meaningfully. 34 4.4. Financial and market incentives Table 6: Estimation results of equation (3.10) Predicted Sign EM CONSTANT -0.0028* ROA_EMt - -0.3079*** CFO_RMt - 0.0505 LEVt-1 - -0.0262*** SIZEt-1 - 0.0017** MTBt-1 ? 0.0008 CONSTRAINTSt - -0.0032*** BEATERSt + -0.0016 SMOTHERSt + -0.0055* BORROWERSt*CREDITt + 0.0006** DIVIDt + 0.0108*** Year dummies Yes Adj. R-squared 0.2808 Prob (F-stat) 0.0000 DW Statistic 1.7789 Total panel obs. 300 *, **, and *** indicate significance at the 10%, 5% and 1% level, respectively; Equation (3.10) was estimated (Table 6) with the purpose of answering the question: why firms manage earnings upward? First of all, the variable CONSTRAINTS appears to be effectively capturing earnings management restrictions. Its coefficient is negative and the variable is statistically significant at 99% confidence level. Ceteirs paribus, the difference in total earnings management between a non-constrained and a full constrained firm equals on average, 1.92% of total assets. Controlling for earnings management constraints is of first importance, since firms with specific motivations to manage earnings may not be able to do it because they are in some way restrained. This should increase the power of the tests. The most typical market incentives to manage 35 earnings, proxied in this case by BEATERS and SMOOTHERS appear not to be of first relevance for Portuguese listed firms. This study does not present evidence of firms managing earnings to avoid losses or earnings decreases6. The results presented show that firms with higher increases in long term debt, manage earnings upward in response to an increase in banks terms and conditions. This is consistent with firms minimizing the cost of capital through earnings management. On average, when those firms celebrate new debt contracts, an increase of one standard deviation in banks terms and conditions leads to an increase in earnings management of 0.6% of total assets. If the variables BORROWERS and CREDIT are included separately, they do not appear to be important from a statistical point of view. In other words, firms manage earnings to increase the cost of capital if financial conditions get worse, that is, when the marginal benefit of managing earnings increases. The variable BORROWERS equals 1 if the increase in long-term debt is on the top quartile (within firms with increases in longterm debt) and 0 otherwise. If the observations on the third quartile are also set to equal one, then the coefficient of the variable BORROWERS*CREDIT becomes less significant. In this sense, the results are methodology sensitive. Theoretically, this may just be pointing, again, to the idea that firms risk to manage earnings upward only when they can significantly benefit from it, in this case, when the amount of credit obtained is high enough. Still, the main incentive for firms to manage earnings seems to be the payment of expected dividends, since DIVID is statistically significant at 99% confidence level. The results suggest that when earnings are falling short of the dividends payed in the previous year, firms on average manage earnings upward by 1.08% of total assets. This outcome is in accordance with the literature that highlights the importance of dividends to shareholders. This is also consistent with the results of Naveen et al. (2008) when studying Standard & Poor’s 1500 firms. Overall, the results presented are robust because both kinds of earnings management are considered; most studies focus only in only one incentive while this investigation incorporates different ones; this study has a complete measure of earnings management constraints, whereas 6 As a robustness check, the variable beaters was set to equal to 1 if the variation in earnings per share is between 0 and 1 cent while smoothers was computed to equal 1 if return on assets was above 0 and lower than half percent. Both variables were set to 0 in all other cases. In this scenario, the signal of the coefficient of those variables is still negative, even if not statistically significant. 42 Cohen, D. A., A. Dey and T. Z. Lys (2008), “Real and accrual-based earnings management in the pre-and post-Sarbanes-Oxley periods”, The Accounting Review, Vol. 83, Nº3, pp 757-787. Cohen, D. and P. Zarowin (2008), “Economic consequences of real and accrual-based earnings management activities”, Leonard Ster School of Business & New York University, Working Paper. Cornett, M. M., A. J. Marcus and H. Tehranian (2008), “Corporate governance and payfor-performance: The impact of earnings management”, Journal of financial economics, Vol. 87, Nº2, pp. 357-373. Daniel, N. D., D. J. Denis, and L. Naveen (2008), “Do firms manage earnings to meet dividend thresholds?” Journal of Accounting and Economics, Vol. 45, Nº1, pp. 2-26. DeAngelo, H. and L. DeAngelo (2006), “The irrelevance of the MM dividend irrelevance theorem”, Journal of financial economics, Vol. 79, Nº2, pp. 293-315. DeAngelo, H., L. DeAngelo and R. M. Stulz (2006), “Dividend policy and the earned/contributed capital mix: a test of the life-cycle theory”, Journal of Financial economics, Vol. 81, Nº2, 227-254. Dechow, P. M., A. P. Hutton, J. H. Kim and R. G. Sloan (2012), “Detecting earnings management: A new approach”, Journal of Accounting Research, Vol. 50, Nº2, pp. 275334. Dechow, P. M. (1994), “Accounting earnings and cash flows as measures of firm performance: The role of accounting accruals”, Journal of accounting and economics, Vol. 18, Nº1, pp. 3-42. Dechow, P. M. and D. J. Skinner (2000), “Earnings management: Reconciling the views of accounting academics, practitioners, and regulators”, Accounting horizons, Vol. 14, Nº2, pp. 235-250. 43 Dechow, P. M., I. D. Dichev (2002), “The quality of accruals and earnings: The role of accrual estimation errors”, The accounting review, Vol. 77, Nº1, pp. 35-59. Dechow, P. M. and R. G. Sloan (1991), “Executive incentives and the horizon problem: An empirical investigation”, Journal of accounting and Economics, Vol. 14, Nº1, pp. 51-89. Dechow, P. M., R. G. Sloan and A. P. Sweeney (1995), “Detecting earnings management”, The Accounting Review, pp. 193-225. Dechow, P. M., S. A. Richardson and I. Tuna (2003), “Why are earnings kinky? An examination of the earnings management explanation”, Review of accounting studies, Vol. 8, Nº2-3, 355-384. Dechow, P., W. Ge and C. Schrand (2010), “Understanding earnings quality: A review of the proxies, their determinants and their consequence”, Journal of Accounting and Economics, Vol. 50, Nº2, pp. 344-401. Desai, M. A. (2005), “The degradation of reported corporate profits”, The Journal of Economic Perspectives, Vol. 19, Nº4, pp. 171-192. Donaldson, G. (1961), “Corporate debt capacity”. Donohoe, M. P. (2015), “The economic effects of financial derivatives on corporate tax avoidance”, Journal of Accounting and Economics, Vol. 59, Nº1, pp. 1-24. Dyreng, S. D., M. Hanlon and E. L. Maydew (2008), “Long-run corporate tax avoidance”, The Accounting Review, Vol. 83, Nº1, pp. 61-82. Dyreng, S. D., M. Hanlon and E. L. Maydew (2010), “The effects of executives on corporate tax avoidance”, The Accounting Review, Vol. 85, Nº4, pp. 1163-1189. Easterbrook, F. H. (1984), “Two agency-cost explanations of dividends”, The American Economic Review, Vol. 74, Nº4, pp. 650-659. 44 Eckel, N. (1981), “The income smoothing hypothesis revisited”, Abacus, Vol. 17, Nº1, pp. 28-40. Francis, J. R., E. L. Maydew and H. C. Sparks (1999), “The role of Big 6 auditors in the credible reporting of accruals”, Auditing: a Journal of Practice & theory, Vol. 18, Nº2, pp. 17-34. Gao, L. and J. H. Zhang (2015), “Firms’ earnings smoothing, corporate social responsibility, and valuation”, Journal of Corporate Finance, Vol.32, pp. 108-127. Géczy, C., B. A. Minton and C. Schrand (1997), “Why firms use currency derivatives”, The Journal of Finance, Vol. 52, Nº 4, pp. 1323-1354. Graham, J. R. (2003), “Taxes and corporate finance: A review”, Review of Financial studies, Vol. 16, Nº4, pp. 1075-1129. Graham, J. R. and C. W. Smith (1999) “Tax incentives to hedge”, The Journal of Finance, Vol. 54, Nº6, pp. 2241-2262. Graham, J. R. and D. A. Rogers (2002), “Do firms hedge in response to tax incentives?”, The Journal of finance, Vol. 57, Nº2, pp. 815-839. Graham, J. R., C. R. Harvey and S. Rajgopal (2005), “The economic implications of corporate financial reporting”, Journal of accounting and economics, Vol. 40, Nº1, pp. 3-73. Grullon, G. and R. Michaely (2002), “Dividends, share repurchases, and the substitution hypothesis”, The Journal of Finance, Vol. 57, Nº4, pp. 1649-1684. Gupta, S. and K. Newberry (1997), “Determinants of the variability in corporate effective tax rates: Evidence from longitudinal data”, Journal of Accounting and Public Policy, Vol. 16, Nº1, pp. 1-34. Hayn, C. (1995), “The information content of losses”, Journal of accounting and economics, Vol. 20, Nº2, pp.125-153. 45 Healy, P. M. (1985), “The effect of bonus schemes on accounting decisions”, Journal of accounting and economics, Vol. 7, Nº1, pp. 85-107. Healy, P. M. and J. M. Wahlen (1999), “A review of the earnings management literature and its implications for standard setting”, Accounting Horizons, Vol. 13, Nº4, pp. 365383. Jaggi, B. and J. Tsui (2007), “Insider trading, earnings management and corporate governance: empirical evidence based on Hong Kong firms”, Journal of International Financial Management & Accounting, Vol. 18. Nº3, pp. 192-222. Jensen, M. C. and W. H. Meckling (1976), “Theory of the firm: Managerial behavior, agency costs and ownership structure”, Journal of financial economics, Vol. 3, Nº4, pp. 305-360. Jensen, M. C. (1993), “The modern industrial revolution, exit, and the failure of internal control systems”, The Journal of Finance, Vol. 48, Nº3, pp. 831-880. Jones, J. (1991), “Earnings management during import relief investigations”, Journal of accounting research, pp. 193-228. Jung, B., N. Soderstrom and Y. S. Yang (2013), “Earnings smoothing activities of firms to manage credit ratings”, Contemporary Accounting Research, Vol. 30, Nº2, pp. 645676. Kalay, A. (1982), “Stockholder-bondholder conflict and dividend constraints”, Journal of financial economics, Vol. 10, Nº2, pp. 211-233. Kinney, W. R. and L. S. McDaniel (1989), “Characteristics of firms correcting previously reported quarterly earnings”, Journal of accounting and economics, Vol. 11, Nº1, pp. 71-93. Kirschenheiter, M. and N. D. Melumad (2002), “Can “Big Bath” and Earnings Smoothing Co‐exist as Equilibrium Financial Reporting Strategies?”, Journal of Accounting Research, Vol. 40, Nº3, pp. 761-796. 46 Kothari, S. P., N. Mizik S. Roychowdhury (2012), “Managing for the moment: The role of real activity versus accruals earnings management in SEO valuation, Research in Progress. Kothari, S. P., A. J. Leone and C. E. Wasley (2005), “Performance matched discretionary accrual measures”, Journal of accounting and economics, Vol. 39, Nº1, pp. 163-197. Kraft, A. (2014), “What Really Affects German Firms' Effective Tax Rate?”, International Journal of Financial Research, Vol. 5, Nº3, pp. 1-19. Lawrence, A., M. Minutti-Meza and P. Zhang (2011), “Can Big 4 versus non-Big 4 differences in audit-quality proxies be attributed to client characteristics?”, The Accounting Review, Vol. 86, Nº1, pp. 259-286. Lintner, J. (1956). “Distribution of incomes of corporations among dividends, retained earnings, and taxes”, The American Economic Review, Vol. 46, Nº2, pp. 97-113. Mather, P. and A. Ramsay (2006), “The effects of board characteristics on earnings management around Australian CEO changes”, Accounting Research Journal, Vol. 19, Nº2, 78-93. McNichols, M. F. (2002), “Discussion of the quality of accruals and earnings: multiples”, Journal of Accounting Research, Vol. 40, pp. 135-172. Mendes, C. A. and L. L. Rodrigues (2006), “Estudo de práticas de earnings management nas empresas portuguesas cotadas em bolsa: Identificação de alisamento de resultados e seus factores explicativos”, Tékhne-Revista de Estudos Politécnicos, Vol. 5-6, pp. 145-173. Minnick, K. and T. Noga (2010), “Do corporate governance characteristics influence tax management?”, Journal of corporate finance, Vol. 16, Nº5, pp. 703-718. Myers, S. C. (1984), “The capital structure puzzle”, The journal of finance, Vol. 39, Nº3, pp. 574-592. 47 Myers, S. C. and N. S. Majluf (1984), “Corporate financing and investment decisions when firms have information that investors do not have”, Journal of financial economics, Vol. 13, Nº2, pp. 187-221. Nance, D. R., C. W. Smith and C. W. Smithson (1993), “On the determinants of corporate hedging”, The Journal of Finance, Vol. 48, Nº1, pp. 267-284. Healy, P. M. and K. G. Palepu (2012). Business Analysis Valuation: Using Financial Statements. Cengage Learning. Pincus, M. and S. Rajgopal (2002), “The interaction between accrual management and hedging: Evidence from oil and gas firms”; The Accounting Review, Vol. 77, Nº1, pp. 127-160. Rad, S. E. M., H. Salehi and H. V. Pour (2016), “The Impact of Audit Quality and Ownership Structure on Earnings Management of Listed Firms on Tehran Stock Exchange”, International Business Management, Vol. 10, Nº10, pp. 1827-1832. Roychowdhury, S. (2006), “Earnings management through real activities manipulation”, Journal of accounting and economics, Vol. 42, Nº3, pp. 335-370. Shaw, K. W. (2003), “Corporate disclosure quality, earnings smoothing, and earnings' timeliness”, Journal of Business Research, Vol. 56, Nº12, pp. 1043-1050. Smith, C. W. and R. M. Stulz (1985), “The determinants of firms' hedging policies”, Journal of financial and quantitative analysis, Vol. 20, Nº4, pp. 391-405. Stickney, C. P. and V. E. McGee (1983), “Effective corporate tax rates the effect of size, capital intensity, leverage, and other factors”, Journal of accounting and public policy, Vol. 1, Nº2, pp. 125-152. Swiston, A. J. (2008), “A US Financial Conditions Index: Putting Credit Where Credit is Due”, IMF Working Papers, pp. 1-35. 48 Tufano, P. (1996), “Who manages risk? An empirical examination of risk management practices in the gold mining industry”, The Journal of Finance, Vol. 51, Nº4, pp. 10971137. Warfield, T. D., J. J. Wild and K. L. Wild (1995), “Managerial ownership, accounting choices, and informativeness of earnings”, Journal of accounting and economics, Vol. 20, Nº1, pp. 61-91. Warren, A. C. (2004), “US income taxation of new financial products”, Journal of Public Economics, Vol. 88, Nº5, pp. 899-923. Watts, R. L. and J. L. Zimmerman, (1990), “Positive accounting theory: a ten year perspective”, The Accounting review, pp. 131-156. Zang, A. Y. (2011), “Evidence on the trade-off between real activities manipulation and accrual-based earnings management”, The Accounting Review, Vol. 87, Nº2, pp. 675703. Zimmerman, J. L. (1983), “Taxes and firm size”, Journal of accounting and economics, Vol. 5, pp. 119-149. 49 7. Annex Table 7: Variables measurement and sources (part 1) Variables Measurement Source AM Residual of equation (3.1) for each firm and year Datastream, AM_FLEX Sum squared residuals of equation (3.1) for each firm Datastream, BEATERS 1 if the change in earnings per share is between 0 and 2 cents, 0 otherwise. Datastream BIG4 1 if the firm audited by a Big4, 0 otherwise Amadeus, financial reports BOARD Directors on the board / lagged total assets Amadeus, financial reports BORROWERS *CREDIT CREDIT equals the change in the first principal component of the qualitative data of Bank Lending Survey about Portuguese banks’ terms and conditions regarding firms’ credit, after recalling it to have mean 100 and standard deviation 10. BORROWERS equals 1 if LTDEBT is on the top quartile (among firms with increases in long-term debt), 0 otherwise. Bank of Portugal, Datastream, Financial reports CAP_INT Net property, plant and equipment / total assets Datastream CFO Cash-flow from operations / lagged total assets Datastream CFO_RM CFO - RM Datastream CONC 1 if no shareholder holds more than 25% of outstanding shares, 0 otherwise Amadeus, financial reports 50 Table 8: Variables measurement and sources (part 2) Variables Measurement Source CONSTRAINTS Sums 1 in each of these circumstances: if firms’ operational cycle or Z-score (Altman, 2000) is in the first quartile of the sample; if insider ownership (closely held shares / common shares outstanding) and net operating assets (scaled by lagged total sales) are in the top quartile in the beginning of the year; if the firm is audited by a Big4 or if CONC equals 1. Datastream, financial reports, Amadeus CONVEX Nominal marginal tax rate - effective tax rate Datastream, Tax authority CYCLE Average days receivables are outstanding + average days stocks are held – average days payables are outstanding Datastream DERIV 1 if the firm uses derivatives, 0 otherwise Financial reports DIVID 1 if net income net of earnings management are above dividends paid in the previous year and the firm pays dividends in the current year, 0 otherwise Datastream DY Dividends / market capitalization Datastream EM AM + RM Datastream ETR1 Tax expenses / pre-tax income Datastream ETR2 Tax expenses / cash-flow from operations Datastream HEALTH 1 if the Z-score, calculated as in Altman (2000), is above sample median, 0 otherwise Datastream 51 Table 9: Variables measurement and sources (part 3) Variables Measurement Source INSIDER 1 if the ratio between closely held shares and common shares outstanding is above sample median, 0 otherwise; Closely held shares denotes shares owned by insiders which implies officers, directors and their families, shares held in trust, shares of the company held by another corporation (except shares held in a fiduciary capacity by financial institutions), shares held by pension/benefit plans and shares held by individuals who hold 5% or more of the outstanding shares Datastream INV_INT Inventories / total assets Datastream ISALES sales from production made abroad scaled by lagged total sales Datastream LEV Total debt / total assets Datastream LEV_D Equals 1 if LEV is above sample median, 0 otherwise Datastream LTDEBT Long term debt / logarithm of lagged total assets Datastream MTB Market capitalization / shareholders’ funds Datastream MTR Nominal marginal tax rate (includes marginal state tax) Tax authority NOA 1 if net operating assets (scaled by lagged total sales) are above sample median, 0 otherwise; Net operating assets equals: (total assets - cash - total liabilities + total debt) Datastream PAYOUT Dividends / net income net of earnings management Datastream 58 Table 16: Pearson (below diagonal) and Spearman (above diagonal) pairwise correlation coefficients in percentage - variables in equation (3.6) and (3.7) (part 2) (9) (10) (11) (12) (13) (14) (15) (16) AM_FLEXt (1) -12** 3 -1 15*** -12** 73*** -25*** -63*** BIG4t (2) 7 6 13** 2 10* -11* 14** 36*** BOARDt (3) -23*** -2 -7 4 -24*** 56*** -34*** -95*** CAP_INTt (4) 0 3 23*** -1 11* -29*** 8 36*** CONCt-1 (5) 8 -24*** -6 18*** 17*** -10* 17*** 8 CONVEXt (6) -38*** -10* 8 23*** -16*** -13** -20*** -4 DERIVt (7) 21*** 10* 13** -12** 21*** -24*** 31*** 54*** ETR1t (8) 40*** 10* -7 -24*** 17*** 10* 24*** 8 ETR2t (9) 2 -5 -11** 6 -7 29*** 24*** INSIDERt-1 (10) 5 20*** -17*** 3 0 12** 6 INV_INTt (11) 1 21*** -5 -2 -1 -5 6 LEV_Dt-1 (12) -5 -17*** -7 -4 1 -37*** 0 MTBt-1 (13) 0 -1 -5 0 -10* 35*** 36*** RM_FLEXt (14) 1 2 -8 -2 2 -17*** -55*** ROAt (15) 9 15*** -1 -35*** 23*** -10* 37*** SIZEt-1 (16) 12** 4 -7 0 13** -49*** 35*** *, **, and *** indicates significance at the 10, 5 and 1% level, respectively. 59 Table 17: Pearson (below diagonal) and Spearman (above diagonal) pairwise correlation coefficients in percentage - variables in equation (3.8) (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) BOARDt (1) 5 16*** -57*** -51*** -14** -30*** -89*** -24*** -34*** -95*** -76*** CONVEXt-1 (2) -3 -3 0 -6 -14** -3 -7 -15*** -22*** -6 3 CYCLEt-1 (3) 15*** -4 4 0 -11** 34*** -13** -26*** -11* -23*** -14** DERIVt (4) -38*** 0 4 49*** 19*** 27*** 51*** 21*** 31*** 54*** 42*** DYt-1 (5) -22*** -1 3 34*** 16*** 12** 46*** 8 44*** 50*** 42*** HEALTHt-1 (6) -22*** -12** -11** 19*** 6 16*** 0 26*** 39*** 16*** -6 ISALESt (7) -22*** 0 33*** 28*** 7 12** 30*** 2 8 27*** 28*** LTDEBTt (8) -23*** -1 -4 31*** 31*** -10* 13** 23*** 21*** 91*** 72*** MTBt-1 (9) 6 -16*** -15*** 9 -5 7 0 -1 35*** 36*** 16*** ROAt (10) -21*** -28*** 0 26*** 24*** 31*** 10* 12** 23*** 37*** 13** SIZEt-1 (11) -65*** -6 -23*** 53*** 33*** 15*** 26*** 63*** 13** 35*** 76*** STDEBTt (12) -26*** 1 -5 29*** 30*** -9 16*** 89*** 1 12** 63*** *, **, and *** indicates significance at the 10, 5 and 1 % level, respectively. 60 Table 18: Pearson (below diagonal) and Spearman (above diagonal) pairwise correlation coefficients in percentage - variables in equation (3.9) (part 1) (1) (2) (3) (4) (5) (6) (7) AM_FLEXt (1) 64*** -4 18*** -31*** 40*** BOARDt (2) 46*** 1 16*** -57*** 29*** |CFO|t (3) -25*** -23*** -16*** -28*** 31*** -27*** CONVEXt (4) 9 -10* -14** -4 -3 2 CYCLEt-1 (5) 0 15*** -28*** -4 4 2 DERIVt (6) -27*** -38*** 24*** -3 4 -24*** |EM|t (7) 57*** 29*** -20*** 2 -4 -25*** HEALTHt-1 (8) -23*** -22*** 39*** -8 -11** 19*** -15*** LEV_Dt-1 (9) 19*** 0 -23*** 22*** -3 -12** 9* MTBt-1 (10) -18*** 6 31*** -12** -15*** 9 -9* PAYOUTt (11) -3 -2 5 8 10* 6 -4 RM_FLEXt (12) 52*** 69*** -15*** -23*** 20*** -24*** 39*** ROA_EMt (13) -1 -6 25*** -12** -14** 11** 4 SIZEt-1 (14) -45*** -65*** 39*** -3 -23*** 53*** -32*** *, **, and *** indicates significance at the 10, 5 and 1% level, respectively. 61 Table 19: Pearson (below diagonal) and Spearman (above diagonal) pairwise correlation coefficients in percentage - variables in equation (3.9) (part 2) (8) (9) (10) (11) (12) (13) (14) AM_FLEXt (1) -21*** 15*** -12** -32*** 73*** -28*** -63*** BOARDt (2) -14** 4 -24*** -41*** 56*** -34*** -95*** |CFO|t (3) 38*** -24*** 42*** 51*** -28*** 60*** 48*** CONVEXt (4) -9 23*** -16*** -3 -13** -23*** -4 CYCLEt-1 (5) -11** -3 -26*** -8 25*** -18*** -23*** DERIVt (6) 19*** -12** 21*** 38*** -24*** 30*** 54*** |EM|t (7) -11** 7 -5 -31*** 34*** -24*** -30*** HEALTHt-1 (8) -47*** 26*** 28*** -3 32*** 16*** LEV_Dt-1 (9) -47*** -4 -28*** 1 -37*** 0 MTBt-1 (10) 7 0 18*** -10* 32*** 36*** PAYOUTt (11) 10* -12** -6 -17*** 55*** 41*** RM_FLEXt (12) -14** -2 2 1 -12** -55*** ROA_EMt (13) 13** -16*** 7 2 -1 37*** SIZEt-1 (14) 15*** 0 13** -2 -49*** 15*** *, **, and *** indicates significance at the 10, 5 and 1% level, respectively. 62 Table 20: Pearson (below diagonal) and Spearman (above diagonal) pairwise correlation coefficients in percentage - variables in equation (3.10) (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) BEATERSt (1) -5 -8 -3 -4 4 7 -7 -7 -3 14** -3 BORROWERSt (2) -5 0 0 4 1 -8 16*** 7 1 5 2 CFO_RMt (3) -7 2 19*** 9* -5 -24*** 39*** 63*** 50*** -24*** 26*** CONSTRAINTSt (4) -2 0 13** 6 -12** 33*** 13** 5 17*** -11* 12** DIVIDt (5) -4 3 12** 6 25*** 4 16*** -18*** 12** -2 15*** EMt (6) 5 0 -6 -10* 20*** -6 9* -25*** -3 -2 -8 LEVt-1 (7) 4 -8 -26*** 34*** 0 -4 -3 -36*** 4 -2 26*** MTBt-1 (8) -5 4 25*** 12** 14** 11** 2 32*** 36*** -3 16*** ROA_EMt (9) -4 3 27*** 8 -7 15*** 4 7 37*** -13** 15*** SIZEt-1 (10) -3 0 46*** 16*** 11** 0 -1 13** 15*** -12** 76*** SMOOTHERSt (11) 14** 10* -21*** -9* -2 0 -4 4 -3 -13** -6 STDEBTt (12) -1 4 12** 10* 6 0 6 1 5 63*** -8 *, **, and *** indicates significance at the 10, 5 and 1% level, respectively.