Prior literature points out several drivers for the engagement in earnings management, including to raise external financing, to avoid debt covenants restrictions, and to increase managers’ job security and compensation (Dechow et al., 1996; Healy & Wahlen, 1999). For instance, Balsam et al. (2011) found that managerial compensation systems differ according to each BS. They identified that operational efficiency-oriented firms are more focused on short-term financial metrics with higher expectations for an increase in sales results, whereas innovation-oriented firms focus on non-financial measures to evaluate performance and place significantly lower weight on accounting measures.

Nevertheless, firms following a prospector strategy have higher agency costs as they face more risk and uncertainty due to the nature of their investments on intangible assets (Banker et al., 2014), leading them to deal more with undesirable financial results and also with a higher level of discretionary (Bentley-Goode et al., 2017). Habib and Hasan (2018) found that prospectors produce less readable narratives than defenders due to their higher information asymmetry and project uncertainty. Also, Bentley et al. (2013) argue that prospectors rely on external funds more than defenders. Thus, while prospectors need to reduce information asymmetry with capital providers, they might incur in avoiding debt covenants via REM.

Nonetheless, uniqueness and products valued by customers, or a strong brand name, enable firms to sustain their superior performance over time, as it is difficult for competitors to imitate such intangible resources (Banker et al., 2014). Also, Datta et al. (2013) argue that product differentiation can lead firms to enjoy a strong pricing power and competitive advantage, thus they are less likely to engage in earnings management to manipulate financial results. Moreover, previous research (e.g. Wu et al., 2015) indicate that innovation-oriented firms are less associated with REM than efficiency-oriented ones. Based on the arguments discussed above, it is proposed the following hypothesis:

H1: Prospectors are negatively associated with REM, whereas defenders are positively associated with this practice.

To better understand the practice of REM, prior literature indicates that MC is a key determinant. Some studies found that product market competition is a disciplinary mechanism forcing managers to act efficiently and to improve earnings quality (El Diri et al., 2020; Laksmana & Yang, 2014; Marciukaityte & Park, 2009). The research of Laksmana and Yang (2014) gives evidence that both AEM and REM activities are more prevalent among firms in low competitive environments than those in high competition. More recently, El Diri et al. (2020) empirically confirmed that both AEM and REM are greater in concentrated markets, that is, those with low competition.

In contrast, others suggest that such high competition intensifies agency problems and opportunistic behavior (e.g. Datta et al., 2013; Markarian & Santaló, 2014). Markarian and Santaló (2014) found that manipulating earnings by both accrual and real activities is mostly rewarding in highly competitive industries. They argue that competition increases the executive incentive to manipulate earnings since the stock market punishes or rewards in consonance with financial outcomes that indicates a competitive disadvantage/advantage.

There are three channels in which market competition can motivate managers to engage in earnings management activities, according to El Diri et al. (2020): (i) market pricing power, (ii) information disclosure, and (iii) disciplinary effect. First, firms with superior product pricing power engage less in earnings management due to their ability to pass on costs to costumers (Datta et al. 2013). Thus, firms in less (more) competitive environments tend to have less (more) difficulties in protecting their competitive advantages, consequently, they are less (more) motivated to manipulate earnings. The second channel can produce different effects on earnings management. When more companies are competing in a sector, more information is available and is required to reduce capital costs. However, more competition can lead to less companies willing to disclose information due to the threats of competitors and new entrants. Third, the disciplinary channel also causes different effects. As more information circulates in the market, more comparability among firms’ performance is possible. However, this dynamic diminishes a firm`s odds of survival, and managers are more exposed to punishment. Thus, to avoid threats that came with high competition, managers can engage in earnings management.

Based on the arguments discussed above, it is proposed the following hypothesis:

H2: Firms in more competitive markets are positively associated with REM, while firms in less competitive markets are negatively associated with REM.

Datta et al. (2013) argument that, for firms with a market pricing power, high market competition stimulates the engagement in earnings management as it is harder to sustain a competitive advantage facing more competitors. However, Widuri and Sutanto (2018) explored the interaction between differentiation strategy and MC on REM, and found that differentiators in more competitive markets are even less associated with REM than in less competitive markets. While, considering both strategies, Wu et al. (2015) showed that the level of opportunistic manipulation by cost leaders increases in high competition. But, for differentiators, the level of earnings management does not change due to differences in market competition. Thus, the third hypothesis is more exploratory and proposes that:

H3: The combined effect of BS and MC exerts influence on REM.

The data used in this research were collected from Thomson Reuters DataStream database of listed firms in U.S. stock markets for a 33 year-period (e.g. Datta et al., 2013; Habib & Hasan, 2018) of 1987-2020. Initially, all industries identified by North American Industry Classification System (NAICS) code (2 digits) were selected. Following previous studies (e.g. El Diri et al., 2020; Habib & Hasan, 2017, 2018), firm year-observations from the regulated (NAICS code 22) and financial institutions (NAICS code 52-53) industries were not included in the final sample due to their unique accounting and financial practices. Also, observations with missing 2-digit NAICS code were eliminated. In line with Datta et al. (2013), firms with both total assets and net sales of less than US$1 million were removed from the database to avoid the effect of small firms. Finally, the continuous variables (REM and control) were winsorized at 1% and 99% levels to reduce the influence of outliers. Table 1 presents the sample procedures in Panel A and the industry distribution in Panel B.

Table 1

Panel A: Sample selection procedure
Description			Observations
Total number of firm-year observations from 1987 to 2020			255,918
Less: regulated industries (22 code) and financial industries (52-53 code) and missing 2-digit NAICS Codes values			(4,771)
Less: Observations without at least $1 million in net sales and total assets			(33,160)
Less: Observations with missing values for dependent and independent variables, including observations lost for estimating lagged variables.			(198,520)
Final sample			19,467
Panel B: Industry distribution
Code	Industry	# Observations	% Observations
72	Accommodation and Food Services	222	1.14%
56	Administrative and Support and Waste Management and Remediation Services	247	1.27%
71	Arts, Entertainment, and Recreation	31	0.16%
23	Construction	176	0.90%
61	Educational Services	61	0.31%
62	Health Care and Social Assistance	160	0.82%
51	Information	1,531	7.86%
31-33	Manufacturing	13,303	68.34%
21	Mining, Quarrying, and Oil and Gas Extraction	413	2.12%
54	Professional, Scientific, and Technical Services	1,735	8.91%
44-45	Retail Trade	1,004	5.16%
48-49	Transportation and Warehousing	177	0.91%
42	Wholesale Trade	407	2.09%
	TOTAL	19,467	100%

Source: Research data

Initially, the sample has 255,918 firm-year observations. Then, after excluding observations for regulated, financial, and missing 2-digit NAICS codes industries, small companies, and missing values of required variables to create dependent and independent proxies, the final sample was constrained to 19,467 firm-year observations and 5,127 unique companies. Due to the estimation of lagged variables for the BS composite score, the final sample period was constrained to 1992 to 2020.

3.2.1. Real earnings management

Based on Roychowdhury (2006), an aggregate variable based on three measures was used to proxy REM: the abnormal levels of production, sales, and discretionary expenditures. These were estimated via cross-section regressions, as specified in equations (1)-(3). The abnormal level of productio was assessed through the cost of goods sold (COGS) and changes in inventory. To capture the abnormal level of sales, caused by temporary price discounts or lenient credit terms, cash flows from operations (CFO) were considered, while the abnormal level of discretionary expenditures was measured using R&D and SG&A. Specifically, the following regressions were estimated for each year and two-digit NAICS industry code with at least 15 observations:

PRODitAit-1 = ∝0+ ∝11Ait-1 + ∝2SitAit-1+  ∝3∆SitAit-1+ ∝4∆Sit-1Ait-1+ εit  (1)

CFOitAit-1 = ∝0+ ∝11Ait-1 + ∝2SitAit-1+ ∝3∆SitAit-1+ εit (2)

DISXitAit-1 = ∝0+ ∝11Ait-1 + ∝2Sit-1Ait-1+ εit (3)

Where PROD_i,t is the sum of COGS and change in inventory; CFO_i,t is the cash flow from operations; and DISX_i,t is the sum of R&D and SG&A expenses for firm i in year t; A_i,t-1 is the total assets for firm i at the end of year t-1; S_i,t is the net sales for firm i in year t; S_i,t-1 is the net sales of firm i at the end of year t-1; ΔS_i,t is the change in net sales from year t-1 to t, while ΔS_i,t-1 is the change in net sales from year t-2 to t-1.

The abnormal level of each activity was estimated as the residual from each regression. For equations 2 and 3 the residuals were multiplied by -1 to generate the abnormal level of sales and discretionary expenditures. In this sense, high values of all three variables indicate more engagement in real earnings management to enhance profits. Finally, an aggregate measure was formed by the sum of the three REM variables, with high values indicating intense engagement in income-increasing REM (Shi et al., 2018).

3.2.2. Business strategy composite measure

By using financial archival data, this study measured firm’s realized strategy rather than their intended one (David et al., 2002; Mintzberg, 1987). The intended strategy is the conception of strategy based on a statement of intent, whilst realized strategy is related to a pattern of actions in a stream of decisions which is found by objective indicators such as archival data (Snow & Hambrick, 1980). Thus, following Bentley et al. (2013), a discrete strategy composite score was applied to measure firms’ BS. Including variables from Ittner et al. (1997), this strategy composite score is based on the Miles and Snow`s (1978, 2003) framework, and it has been consistently employed by prior researchers (Bentley-Goode et al., 2017; Chen et al., 2017; Habib & Hasan, 2017, 2018). Table 2 presents variable descriptions and measurements, and the procedures for calculating the BS composite measure used in this study are detailed in Appendix I.

Table 2

Variable	Description	Variable measurement
(1) Research and development to sales (RDS5)	Company’s propensity to seek for new products.	Average of research and development expenditures to net sales over the prior five years.
(2) Employees to sales (EMPS5)	Company’s ability to produce and distribute products and services efficiently.	Average of the number of employees to net sales over the prior five years.
(3) Employee fluctuations (σ(EMP5))	Company’s organizational stability.	Standard deviation of the total number of employees [EMP] over the prior five years.
(4) Change in total revenue (REV5)	Company’s historical growth or investment opportunities.	Average of one-year percentage change in net sales over the prior five years.
(5) Marketing to sales (SGA5)	Company’s focus on marketing investments.	Average of selling, general and administrative expenses to net sales over the prior five years.
(6) Capital intensity (CAP5)	Company’s commitment to technological efficiency and production.	Average of net property, plant and equipment scaled by total assets over the prior five years.

Note: Each variable is measured per firm-year based on the rolling prior five-year average, except the variable σ(EMP5) that considers the standard deviation of the number of employees over the prior five years. In sequence, each of these average variables is ranked into quintiles per industry (two-digit NAICS code) and year. The observations in the highest quintiles are given a score of 5, while the ones in the lowest quintiles are given a score of 1 (except capital intensity which is reversed-scored, meaning that observations in the lowest (highest) quintile are given a score of 5 (1)). Within each firm-year, the scores are summed over the six measures, such that the maximum score that a firm could receive is 30 (prospector-type) and a minimum score of 6 (defender-type). Therefore, the discrete STRATEGY score ranges along a continuum in value from 6 to 30 with defender- and prospector-type companies closer to the endpoints (Bentley et al., 2013).

3.2.3. Market competition

Herfindahl-Hirschman Index (HHI) - Frequently used in prior literature (Datta et al., 2013; Marciukaityte & Park, 2009; Markarian & Santaló, 2014; Shi et al., 2018; Wu et al., 2015), the Herfindahl-Hirschman Index (HHI) is a consolidated indicator of product market competition. HHI reflects market concentration, in which lower values are closer to a monopoly representing higher market competition.

Concentration ratio (CR4) - Concentration ratio is the second measure for market competition used in this research. CR4 also captures the level of concentration, however it reflects high competition even in concentrated markets (El Diri et al., 2020). CR4 considers only the four firms with the largest market share in each industry. Nevertheless, the interpretation is the same as for HHI, and lower values mean highly competitive markets.

Hall Tideman Index (HTI) - HTI measures the variation of product substitutability, considering the absolute number of firms and their relative sizes, thus, reflecting the entry barriers of an industry (El Diri et al., 2020; Hall & Tideman, 1967). HTI includes the ranks of all firms in a particular industry based on their market share. Like the other two proxies, higher values of HTI indicate that a industry is less competitive. All three variables were multiplied by -1 to facilitate the interpretation of results and were calculated according to Appendix I.

The empirical models estimated to investigate the research hypothesis are presented in equations 4, 5, and 6. Ordinary least squares (OLS) regression were performed to test all the empirical models, including industry fixed effects to control for industry-wide common factors, and year fixed effects to control for cross-sectional effects. All regressions are based on standard errors adjusted at firm level (Petersen, 2009).

To test the first hypothesis, the current period REM was regressed on strategy and the control variables in the current period, as follows:

REMi,t= β0+β1STRATEGYi,t+ β2 ROAi,t+ β3 LEVi,t+ β4SIZEi,t+ β5GROWTHi,t+ IndustryDummyi,t+ YearDummyi,t+ εi,t (4)

Where REM_i,t is the level of REM for firm i in year t, proxied by the sum of APROD, ACFO, and ADISX, following equations 1-3 above. STRATEGY_i,t is the composite score (continuous), for firm i in year t, ranging from a minimum of 6 to a maximum of 30 following the procedures explained in 3.2.2. All models include dummies for industry and year, and control variables consistent with prior literature (Datta et al., 2013; Habib & Hasan, 2017; Wu et al., 2015), namely: return of assets (ROA), leverage ratio (LEV), size (SIZE), and growth (GROWTH). The variables definitions are summarized in Appendix I.

REMi,t= β0+β1STRATEGYi,t+ β2Market competitioni,t+ β3ROAi,t+ β4 LEVi,t+ β5SIZEi,t+ β6GROWTHi,t+ IndustryDummyi,t+ YearDummyi,t+ εi,t (5)

Next, equation 5 exhibits the model designed to test the second hypothesis, which regards to the relationship between REM and MC. Market competition is measured using three different indexes, namely HHI, CR4, and HTI, as detailed in 3.2.3. Three regression models were performed, one for each MC proxy. The STRATEGY composite was included in these models to control for its effect.

REMi,t= β0+β1STRATEGYi,t+ β2 Market competitioni,t+ β3 STRATEGYi,t*Market competitioni,t+ β4 ROAi,t+ β5 LEVi,t+ β6SIZEi,t+ β6GROWTHi,t+ IndustryDummyi,t+ YearDummyi,t+ εi,t (6)

Finally, the combined effect of firm’s BS and MC on REM is tested through the model exhibited in equation 6, as stated in hypothesis 3. As for the previous models, three different regression models were investigated with each market competition proxy interacting with the STRATEGY score.

Descriptive statistics of the variables used in the primary analysis are reported in panel A of Table 3. The dependent variable REM presented mean and median values of -0.052 and 0.004, respectively, indicating that, on average, the sample firms engage in low values of REM. The mean and median values of STRATEGY score are 17.75 and 18, respectively, which is in consonance with Habib and Hasan (2017, 2018) and indicate that, on average, the firms studied are not clearly positioned strategically. The values of MC variables are comparable with Markarian and Santaló’s (2014). The mean values of HHI (-0.051), CR4 (-0.042), and HTI (-0.022) indicate that MC in the industries analyzed is, in general, high, as values range from 0 and -1 and higher values of these variables indicate more competition within an industry.

Table 3

Panel A: Descriptive statistics
Variables	Full sample (n = 19.467)					Prospectors (STRATEGY range of 24-30) (n = 1.444)		Defenders (STRATEGY range of 6-12) (n = 1.756)
	Mean	Median	1st Quartile	3rd Quartile	Std. Dev.	Mean	Median	Mean	Median
REM	-0.052	0.004	-0.289	0.243	0.466	-0.326	-0.251	0.275	0.298
STRATEGY	17.750	18.000	15.000	21.000	3.904	25.046	25.000	10.886	11.000
HHI	-0.051	-0.026	-0.074	-0.020	0.054	-0.046	-0.026	-0.049	-0.026
CR4	-0.042	-0.020	-0.060	-0.012	0.054	-0.037	-0.020	-0.040	-0.019
HTI	-0.022	-0.009	-0.034	-0.007	0.023	-0.020	-0.009	-0.022	-0.009
ROA	-0.006	0.047	-0.012	0.089	0.216	-0.166	-0.045	0.019	0.045
LEV	0.501	0.483	0.300	0.638	0.298	0.462	0.406	0.564	0.522
SIZE	13.252	13.314	11.706	14.830	2.283	12.703	12.611	13.375	13.291
GROWTH	0.086	0.063	-0.028	0.166	0.270	0.265	0.178	0.023	0.023
Panel B: Correlation analysis
	1	2	3	4	5	6	7	8	9
1 REM	1.00
2 STRATEGY	-0.382	1.00
3 HHI	0.004	0.036	1.00
4 CR4	0.005	0.034	0.998	1.00
5 HTI	-0.010	0.057	0.895	0.876	1.00
6 ROA	-0.063	-0.211	0.047	0.047	0.014	1.00
7 LEV	0.107	-0.142	-0.049	-0.045	-0.057	-0.256	1.00
8 SIZE	0.106	-0.078	0.062	0.062	0.041	0.378	0.151	1.00
9 GROWTH	-0.155	0.215	-0.043	-0.043	-0.034	0.143	-0.061	0.038	1.00

Note.. The continuous variables REM, ROA, LEV, SIZE and GROWTH were winsorized at 1% and 99% levels to reduce the influence of outliers. Bold and italics variables are significant at p < 0.01, bold only variables are significant at p < 0.05, and italics for variables significant at p < 0.1.

In addition, firms were classified in accordance with their STRATEGY score. Thus, a sample with prospectors was formed by firms with scores ranging from 24 to 30 (the maximum), whereas, there was a sample of firms with scores ranging from 6 (the minimum) to 12 for defenders (Bentley et al., 2013; Habib & Hasan, 2018). Prospectors represent 7.41% of the full sample, and defenders represent 9.02% of it. The mean value of STRATEGY for the prospector group was 25.046, while for the defender group the mean was 10.889, which is similar to Bentley-Goode et al. (2017) and Bentley et al. (2013). Moreover, for prospectors, the mean value of REM was -0.326, while for defenders, it was 0.275, which indicates initial evidences that the former engages in earnings management less than the latter.

Panel B of Table 3 exhibits the Pearson correlation of the variables used in the final analyses. The correlation analysis indicates that REM is negative and significative correlated with STRATEGY, giving an initial picture that firms pursuing a prospector strategy engage less in REM than firms following a defender strategy. In contrast, the correlation between REM and HHI, CR4, and HTI are not significant, which might be an indicative that MC is not directly associated with REM.

Table 4 presents the principal regression results. Model 1 exhibits the findings related to the association between BS and REM. The results show that the coefficient of STRATEGY is negative and significant (coefficient of -0.050, significant at p < 0.01), which confirms H1. High scores of STRATEGY represent firms pursing a prospector strategy, while low scores indicate defender strategy. In this sense, this result supports the research hypothesis H1, as it suggests that firms following an innovation-oriented strategy are less likely to engage in REM, whereas firms following an efficiency-oriented strategy are more likely to engage in this practice. That is, as more a firm pursues a prospector strategy, less levels of REM it presents.

Table 4

	Dependent variable: Real Earnings Management
	(1)	(2)	(3)	(4)	(5)	(6)	(7)
STRATEGY	-0.050***	-0.050***	-0.050***	-0.050***	-0.212***	-0.209***	-0.209***
	(0.002)	(0.002)	(0.002)	(0.002)	(0.011)	(0.010)	(0.011)
HHI		-0.104			-0.149
		(0.259)			(0.268)
CR4			-0.103			-0.150
			(0.251)			(0.260)
HTI				-0.065			-0.203
				(0.820)			(0.842)
STRATEGY*HHI					-0.399***
					(0.141)
STRATEGY*CR4						-0.413***
						(0.142)
STRATEGY*HTI							-0.740***
							(0.347)
SIZE	0.034***	0.034***	0.034***	0.034***	0.034***	0.034***	0.034***
	(0.005)	(0.005)	(0.005)	(0.005)	(0.005)	(0.005)	(0.005)
LEV	-0.053	-0.053	-0.053	-0.053	-0.055	-0.054	-0.055
	(0.037)	(0. 037)	(0.037)	(0.037)	(0.037)	(0.037)	(0.037)
ROA	-0.502***	-0.503***	-0.503***	-0.502***	-0.502***	-0.501***	-0.504***
	(0.051)	(0.051)	(0.051)	(0.051)	(0.051)	(0.051)	(0.051)
GROWTH	-0.067***	-0.067***	-0.067***	-0.067***	-0.069***	-0.069***	-0.067***
	(0.020)	(0.020)	(0.020)	(0.020)	(0.020)	(0.020)	(0.020)
Industry FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Year FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Observations	19,467	19,467	19,467	19,467	19,467	19,467	19,467
Adjusted R2	0.206	0.206	0.206	0.206	0.208	0.208	0.207

Note. All variables are described in Appendix I. REM and control variables were winsorized at 1% and 99% levels to reduce the influence of extreme values. FE regards to fixed effects of industry and year. To control for multicollinearity, STRATEGY was centered around the mean for the interaction models, and all VIF values were below 10. ^***p<0.01; ^**p<0.05; ^*p<0.1.

Models 2 to 4 (Table 4) show the results of the regression models that tested the relationship between the MC variables and REM. The coefficients HHI (coefficient of -0.104), CR4 (coefficient of -0.103), and HTI (coefficient of -0.065) were not found to be significant. Thus, it does not confirm the research hypothesis H2, that the level of MC is associated with REM. Further, the coefficient of STRATEGY is also negative and significant (coefficient of -0.050, significant at p < 0.01), which also confirms H1.

The outputs of models 5 to 7 (Table 4) show that the coefficient of the interaction between STRATEGY and MC - proxied by HHI (coefficient of -0.399, significant at p < 0.01), CR4 (coefficient of -0.413, significant at p < 0.01), and HTI (coefficient of -0.740, significant at p < 0.01) - are negative and significant. Supporting H3, these results confirm the statistical significance of a moderated effect of MC on the relationship between STRATEGY and REM. Specifically, it can be inferred that firms with higher scores of STRATEGY operating in markets with high competition engage less in REM. In this sense, MC can affect the relationship between STRATEGY and REM.

In summary, the results evidence that, when considering only STRATEGY, firms with high STRATEGY score are less prone to engage in REM practices than firms with low STRATEGY score. Also, prior research has found innovation-oriented strategy firms to be associated with low values of REM (Widuri & Sutanto, 2018; Wu et al., 2015), whereas cost efficient-oriented strategy firms were found associated with higher values of REM (Wu et al., 2015). These results are in line with the idea that efficiency-oriented firms have greater incentives to engage in earnings management, while prospectors have less motivation to engage in the management of real activities. As defenders tend to focus more on the short term and, also, are more prone to imitation and obsolescence of their processes and resources for operational efficiency, they may engage in higher levels of REM.

Furthermore, when considering the combined effect of BS and MC, the findings reveal that the level of market competition affects the effect of STRATEGY on REM, with negative net effect. These results are partially in line with prior research findings, as Wu et al. (2015) found a positive relationship between efficiency-oriented cost leaders and REM in more competitive markets. However, Wu et al. (2015) found an insignificant effect of MC on the relationship between innovation-oriented firms and REM, whereas Widuri and Sutanto (2018) found that differentiators in highly competitive markets have lower levels of REM.

4.3.1. Alternative measure of BS

For additional analyses, an alternative measure of BS was used to test H1 and H3. From the composite measure of BS, firms with STRATEGY score ≥ 24 were classified as PROSPECTORS, and firms with STRATEGY score ≤ 12 were classified as DEFENDERS. Next, a dummy variable was created, in which firm-year observations classified as PROSPECTORS took a value of 1 and zero otherwise (Bentley et al., 2013; Habib & Hasan, 2018). The findings are presented at Table 5 and show that the coefficients of PROSPECTORS are negatively associated with REM, supporting the results of the main analysis for H1. Moreover, the interaction analysis of the alternative measure of BS and each competition measure confirms a negative and significant coefficient for PROSPECTORS. All three models, (2)-(4), confirm MC as an influencer in the relationship between PROSPECTORS and REM. These results are aligned with the main analysis and confirm the research hypothesis H3. Finally, STRATEGY was also replaced by two indicator variables for prospectors and defenders, where analyzers were used as the benchmark. Again, the results supported the findings reported above.

Table 5

	Dependent variable: Real Earnings Management
	(1)	(2)	(3)	(4)
PROSPECTORS	-0.651*** (0.035)	-0.769*** (0.044)	-0.751*** (0.041)	-0.743*** (0.045)
HHI		0.958* (0.534)
CR4			0.968* (0.527)
HTI				1.865 (1.562)
PROSPECTORS* HHI		-2.567*** (0.522)
PROSPECTORS* CR4			-2.675*** (0.523)
PROSPECTORS* HTI				-4.406*** (1.299)
SIZE	0.030*** (0.008)	0.025*** (0.008)	0.025*** (0.008)	0.027*** (0.008)
LEV	-0.017 (0.048)	-0.016 (0.045)	-0.016 (0.045)	-0.020 (0.047)
ROA	-0.432*** (0.074)	-0.436*** (0.072)	-0.435*** (0.072)	-0.439*** (0.073)
GROWTH	-0.060* (0.031)	-0.068** (0.031)	-0.069** (0.031)	-0.060* (0.031)
Industry FE	Yes	Yes	Yes	Yes
Year FE	Yes	Yes	Yes	Yes
Observations	3,200	3,200	3,200	3,200
Adjusted R2	0.381	0.393	0.394	0.388

Note. All variables are described in Appendix I. REM and control variables were winsorized at 1% and 99% levels to reduce the influence of extreme values. FE refers to the presence of industry and year fixed effects. To control for multicollinearity, STRATEGY was centered around the mean for the interaction models, and all VIF values were below 10. ^***p<0.01; ^**p<0.05; ^*p<0.1.

4.3.2. Alternative measure for earnings management

As firms can engage in both real and accrual earnings management, AEM was tested as a dependent variable for all the three models (see section 3.3) as robustness. Following Kothari et al. (2005), the modified version of Jones’ model (Dechow et al., 1995; Jones, 1991) was used to capture discretionary accruals. Table 6 exhibits the results that indicate STRATEGY (coefficient of 0.003, significant at p < 0.01) as a determinant of earnings management. However, the singular impact of MC on AEM was found not significant. In turn, the combined effect of STRATEGY and MC on AEM is partially confirmed since only the combined effect of STRATEGY and HTI was found significant (coefficient of 0.112, significant at p < 0.01). On the one hand, the findings indicate that prospectors during the high competition are more likely to engage in AEM. On the other hand, it confirms the importance of MC in the relationship between STRATEGY and earnings management in general.

Table 6

	Dependent variable: Accurals Management
	(1)	(2)	(3)	(4)	(5)	(6)	(7)
STRATEGY	0.003*** (0.000)	0.003*** (0.000)	0.003*** (0.000)	0.003*** (0.000)	0.012*** (0.002)	0.012*** (0.002)	0.013*** (0.002)
HHI		0.044 (0.035)			0.049 (0.035)
CR4			0.040 (0.034)			0.045 (0.034)
HTI				0.003 (0.101)			0.039 (0.101)
STRATEGY *HHI					0.036 (0.022)
STRATEGY *CR4						0.035 (0.022)
STRATEGY *HTI							0.112*** (0.041)
SIZE	-0.010*** (0.001)	-0.010*** (0.001)	-0.010*** (0.001)	-0.010*** (0.001)	-0.010*** (0.001)	-0.010*** (0.001)	-0.010*** (0.001)
LEV	0.039*** (0.008)	0.039*** (0.008)	0.039*** (0.008)	0.039*** (0.008)	0.039*** (0.008)	0.039*** (0.008)	0.039*** (0.008)
ROA	-0.089*** (0.014)	-0.089*** (0.014)	-0.089*** (0.014)	-0.089*** (0.014)	-0.089*** (0.014)	-0.089*** (0.014)	-0.089*** (0.014)
GROWTH	0.044*** (0.007)	0.044*** (0.007)	0.044*** (0.007)	0.044*** (0.007)	0.045*** (0.007)	0.045*** (0.007)	0.044*** (0.007)
Industry FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Year FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Observations	20,509	20,509	20,509	20,509	20,509	20,509	20,509
Adjusted R2	0.166	0.166	0.166	0.166	0.166	0.166	0.166

Note. All variables are described in Appendix I. The control variables were winsorized at 1% and 99% levels to reduce the influence of extreme values. FE refers to the presence of industry and year fixed effects. To control for multicollinearity, STRATEGY was centered around the mean for the interaction models, and all VIF values were below 10. ^***p<0.01; ^**p<0.05; ^*p<0.1.