When do firms highlight their effective tax rate?

Figures & data

Table

Topic / questions	Topic-specific responses of five Global Heads of Taxes of major German multinational corporations
Sequence of ETR planning and disclosure strategy When do you actively report on your ETR? How do you determine how you report on the ETR? Do you report after the ETR is already established? Do you simultaneously plan the ETR and report on it? Do you determine the disclosure strategy regardless of the level (condition) of the ETR? Do you discuss, develop and readjust the disclosure strategy if the ETR is unusual?	The ETR is planned very carefully for two years and is permanently reviewed.
	The ETR is planned as well as possible over five years – of course, with high uncertainty.
	We regularly exchange information with the Investor Relations department.
	If we expect an unusual ETR, we reach out to the Investor Relations department at an early stage.
	On the basis of the outcome of the ETR the Head of Taxes provides all necessary explanations to the Investor Relations department.
	There is close coordination with the Investor Relations department as part of the regular exchange with the other parts of the finance function. This also includes the tax department reporting the outcome of the ETR and explaining deviations from the previous year to the finance function. Especially, it explains any special effects to the Investor Relations department
	I provide all necessary explanations to the Investor Relations department.
What is important? Does it matter to you that the ETR is particularly low (strong decrease) that the ETR is not particularly volatile (smooth), that your ETR is similar to that of your peers’, that your ETR lies in a specific pre-determined corridor?	The target rate is 25%. It is also important to us that the ETR does not fluctuate.
	A target that is ‘as low as possible’ is not reasonable. It is far too complicated and cannot be maintained in the long term. When a tax issue reverses over time, i.e. in years when the ETR is temporarily high, this can show the performance of the head of taxes in a negative light. Thus, for these and other reasons we do not actively plan for a very low ETR.
	There is no target tax rate. Due to our business model, the tax rate is far too volatile. Deferred taxes in particular fluctuate considerably.
	The level of the ETR is less important. It should not be too low. Too high or too low a ratio we explain. An accurate and transparent explanation is most important.
	A stable ETR as a perspective is important to us.
	We have had a fixed target ETR range of 2% for many years. We are always within this range, almost nothing changes over time.*
	The ETR is highly dependent on the ownership structure.
	Cash-effective taxes are the focus of planning. Deferred taxes are not of interest. If the ETR is very low, however, countermeasures are taken.
	In the past the ETR was very important, but today cash-effective tax rates are increasingly important.
	The ETR has de facto been around 20-30% in recent years.
	Downwards deviations of the ETR occur and are ok. They need to be explained.
	We have no lower ETR threshold. There are special effects from time to time that lead to low odds. We don't manage them artificially upwards, but we explain such low ETR. The ETR must always be explainable and understandable for stakeholders.
	We do not have deviations from the previous year that need to be explained. The ETR is more or less constant over time. This is very important to both the family and management. So there is nothing special to report about the ETR.*
	When we have acquisitions, we try to flatten out all aggressive structures before closing, whether these are multi-tier holdings, hybrid structures, transfer prices, Luxembourg companies without substance, etc. Everything artificial is removed and should not even come into the company.*
	We do not orientate ourselves towards peers at all.*
Reputation and unusual ETR Are you worried or have you experienced that an unusual ETR or a decreasing ETR is negatively perceived and thus has reputational effects?	We consider an ETR unusual if it is below 23% or above 28%.
	If we could plan the ETR, we would aim for about 25-35% as a rough target.
	An ETR is unusual if it is below 25% and above 35%. Then it is necessary to explain it.
	The ETR is unusual if it falls below 20%.
	Our tax planning is designed so that we cannot come into focus.*
	Media coverage has no influence on our ETR reporting. We have never been interviewed by media about the ETR either. Experience has shown that anyone who is involved with our company, including journalists, quickly understands how we deal with tax issues.*
Does this influence your approach to the way you report the ETR?	Reputation is an issue.
	Reputation effects are very important in Germany, but also in other countries.
	Reputation is definitely seen as a topic. For us, it is important not to plan aggressively, otherwise you will quickly be in the focus of NGOs. This also influences CSR reporting and the climate of a tax audit.
	The ETR must be justifiable and explainable, then you don't have any reputation problems.
	Reputational risk starts at a tax rate below 25–30%. But that's not that dramatic, you just have to explain it.
	A good relationship with the tax auditor is essential. Therefore, tax-aggressive planning is out of the question. Tax aggressiveness impairs the relationship with the tax auditor.
	We explain tax issues that we expect to be ambiguous to the broader public.*
	There is a regular exchange with the Investor Relations department. I provide this department with dedicated specialist information.*
Incentive scheme and ETR Is the ETR part of your incentive scheme?	The ETR has never been part of my incentive (bonus).
	The ETR has never been part of my incentive (bonus).
	The tax rate is not part of my incentive (bonus).
	Neither the ETR itself nor its stability is part of my incentive. This has never happened.*
Reporting channel How do you react in reporting terms (disclosure strategy)? Do you use the management report for ETR disclosures? Do you mention the ETR rather early or late in the report? Do you mention the ETR rather early or late in the report?	In order for Investor Relations to understand and communicate the many special effects incorporated in the ETR, intensive exchange takes place.
	Transparency and good explanations are the highest maxim and of central importance.
	The ETR is not important to us, so we keep the reconciliation as brief as possible.
	Reputation is also important for the tax audit climate. Good relations with the tax auditor and sound handling of tax audit issues are very important to us.
Crowding out Is the ETR disclosure decision influenced by the fact that not all information can be mentioned on the first page and that some prioritisation is required so that ETR information does not crowd out other information to be reported?	There is no competition between topics. We report on the ETR as required by law and explain any special features and conditions. There is always room for this tax information.

Notes: Topic-specific responses of five Global Heads of Taxes of major German multinationals. As defined in this paper, ETR reflects the ‘GAAP ETR’. In the interviews we used the term ‘GAAP ETR’ to avoid confusion with unspecific kinds of ETR or cash ETR. During the interviews, the discussion occasionally turned to cash ETR which we flag up accordingly in the table above. Responses marked with an asterisk are from family-oriented corporations.

Figure 1: Development of M_REPORT and PAGE (Unscaled and Scaled)

Notes: Development of the yearly average disclosure visibility measures M_REPORT and PAGE over the sample period 2005–2018. PAGE (unscaled) is the variable PAGE before dividing by the total number of annual report pages. Both PAGE variables are presented before multiplying by −1.

Table 1: ETR Conditions and Stakeholder Specific Assessments

Category	ETR Condition	Hypotheses	Favourable ETR Condition from Perspective of
			Shareholder	Other Stakeholder
1	Smooth Close to industry average	H1a H1b	+	+ / o
2	Decreasing	H2	+	-
3	Volatile Above industry average	H3a H3b	-	o / -

Notes: + indicates a positive assessment (benefit), o indicates a neutral assessment, and - indicates a negative assessment (cost). We restrict the ‘decreasing ETR’ condition to positive ETR and abstract from observations with negative pre-tax income or negative tax expenses and cases where tax expense exceeds pre-tax income. ‘Shareholder’ represents all stakeholders that primarily target after-tax cash flows. ‘Other Stakeholder’ primarily care about the societal role of the firm (e.g. fair share of taxes, reputation).

Table 2: Variable Definitions

Dependent Variables
ETRDISCL Alternatively, M_REPORT or PAGE
M_REPORT Indicator variable: one if ETR information is disclosed in management report, zero otherwise
PAGE Page of first ETR appearance in annual report (set to 1 if ETR is not mentioned in annual report), scaled by total pages, (*−1)
ETR Condition Variables
ETRCOND Alternatively, SMOOTH, BENCHM, DECR1, DECR2, VOLETR, or ABOVE_BENCHM
SMOOTH Coefficient of variation: Rolling firm specific ETR standard deviation of current and four previous years divided by the absolute mean ETR over the same five-year period, (*−1)
BENCHM (Absolute) deviation of firm-level ETR from lagged average (one-digit SIC) industry ETR within the sample (*−1)
DECR1 Indicator variable: one if ETRt < ETRt-1, zero otherwise
DECR2 Indicator variable: one if ETRt < ETRt-1 < ETRt-2, zero otherwise
VOLETR Indicator variable: one if ETR lies within the two highest deciles of the rolling five-year ETR standard deviation over all firms
ABOVE_BENCHM Indicator variable: one if ETR exceeds the lagged industry mean ETR by more than ten per cent
DECRLEVEL_x-y Indicator variable: one if ETR decreases in current relative to previous year and ETR of previous year is in the range from x (*100) to y (*100), zero otherwise
DECRDEGREE_x-y Indicator variable: one if ETR decreases in current relative to previous year and percentage points of decrease are in the range from x (*100) to y (*100), zero otherwise
Control Variables
ETR GAAP Effective Tax Rate: total income tax expense divided by pre-tax accounting income
SIZE Natural logarithm of sales
ROA Pre-tax accounting income divided by lagged total assets
AUD Indicator variable: one if firm is audited by Big4 auditor, zero otherwise
ARSCORE Disclosure score of overall annual report content quality
PPE Gross property plant, equipment divided by lagged total assets
R&D Research and development expenses divided by lagged total assets
LEV Long-term debt divided by lagged total assets
FAMILY Indicator variable: one if firm is listed on DAXplus Family 30 Index, zero otherwise

Table 3: Sample Selection (Sample Period 2005–2018)

Observations Firms
DAX30 and MDAX firms ^a) 1,070 80
Non-IFRS −21 -
Unusual values for pre-tax income or income tax expense ^b) −95 −1
Missing data −255 −9
Final Sample 699 70

Notes: This table presents the sample selection process for the sample period 2005–2018. The column ‘Observations’ presents the number of firm-years, the column ‘Firms’ indicates how many distinct firms our sample comprises (the number of years per firm varies due to the unbalanced panel structure). ^a)The initial number of observations is <1,120 (80 firms*14 years) due to limited annual report availability in the early sample years. ^b)Excluded unusual values for pre-tax income or tax expense are the following: pre-tax income<0, tax expense<0, tax expense > pre-tax income. We exclude these observations from our main tests but run additional tests in which we explicitly test these observations in the online appendix.

Table 4: Summary Statistics

Panel A: Regression Variables
	Mean	Median	Std. Dev.	Min	Max	N
M_REPORT	0.78	1	0.41	0	1	699
PAGE (unscaled) ^a)	−87.83	−72	61.40	−308	−2	699
PAGE ^a)	−0.42	−0.34	0.27	−1	−0.01	699
SMOOTH	−0.99	−0.20	4.98	−89.46	−0.01	699
BENCHM	−0.08	−0.05	0.10	−0.94	−0.00	699
DECR1	0.56	1	0.50	0	1	699
DECR2	0.26	0	0.44	0	1	699
VOLETR	0.14	0	0.34	0	1	699
ABOVE_BENCHM	0.25	0	0.43	0	1	699
ETR	0.29	0.28	0.11	0.00	0.96	699
SIZE	22.69	22.58	1.53	18.20	26.19	699
ROA	0.09	0.07	0.08	0.00	0.65	699
AUD	0.82	1	0.38	0	1	699
ARSCORE	0.59	0.59	0.09	0.23	0.85	699
PPE	0.63	0.56	0.39	0.00	2.21	699
R&D	0.02	0.02	0.03	0	0.16	699
LEV	0.20	0.18	0.15	0	1.09	699

Notes: This table presents descriptive statistics for the main variables for 699 observations over the sample period 2005–2018. PAGE (unscaled) is the variable PAGE before dividing by the total number of annual report pages. All other variables are defined in Table 2. ^a)For 71 observations of our final sample, PAGE is missing because the ETR is not mentioned in the annual report. We replace PAGE by −1 (PAGE unscaled by the number of total pages of the annual report) when the ETR in not mentioned, as this is the minimum possible value and indicates the least intent to highlight the ETR, similar to not reporting it at all.

Table

Panel B: Spearman Correlation Matrix
	1)	2)	3)	4)	5)	6)	7)	8)	9)	10)	11)
1) M_REPORT	1.000
2)PAGE	0.703	1.000
3)SMOOTH	0.395	0.280	1.000
4)BENCHM	0.202	0.133	0.394	1.000
5)DECR1	0.114	0.055	0.129	0.009	1.000
6)DECR2	0.119	0.082	0.134	−0.008	0.530	1.000
7)VOL	−0.271	−0.235	−0.531	−0.228	−0.088	−0.086	1.000
8)ABOVE_BENCHM	−0.014	0.051	−0.060	−0.345	−0.257	−0.171	0.172	1.000
9)ETR	0.088	0.091	0.069	0.141	−0.245	−0.199	0.193	0.645	1.000
10)SIZE	−0.092	0.047	−0.112	−0.095	−0.026	−0.074	0.010	0.045	−0.004	1.000
11)ROA	0.276	0.086	0.426	0.292	0.143	0.120	−0.223	−0.149	−0.050	−0.278	1.000
12)AUD	0.180	0.171	0.132	0.107	0.039	−0.013	−0.076	−0.026	−0.015	0.089	0.072
13)ARSCORE	0.054	0.112	−0.084	−0.078	0.066	0.075	0.019	0.071	0.042	0.268	−0.066
14)PPE	−0.216	−0.253	−0.292	−0.148	−0.046	−0.102	0.164	−0.055	−0.119	−0.035	−0.029
15)R&D	0.308	0.146	0.219	0.200	0.099	0.105	−0.150	−0.076	0.081	0.011	0.282
16)LEV	−0.109	−0.017	−0.145	−0.162	−0.010	0.007	0.098	0.071	−0.058	0.068	−0.220

Table

	12)	13)	14)	15)	16)
12)AUD	1.000
13)ARSCORE	0.039	1.000
14)PPE	−0.133	0.013	1.000
15)R&D	0.104	−0.028	−0.068	1.000
16)LEV	−0.072	0.106	0.173	−0.325	1.000

Notes: This table presents spearman correlations for disclosure visibility variables, ETR condition variables, and control variables. All variables are defined in Table 2. Figures in bold and italics indicate significance at 5% level.

Table 5: ETR Conditions and Disclosure Visibility

Panel A: Management Report
		1		2		3
Category		M_REPORT (1)	M_REPORT (2)	M_REPORT (3)	M_REPORT (4)	M_REPORT (5)	M_REPORT (6)
SMOOTH	+	0.041**
		(0.016)
BENCHM	+		3.238**
			(1.441)
DECR1	?			0.507***
				(0.196)
DECR2	?				0.624**
					(0.307)
VOLETR	-					−1.355***
						(0.364)
ABOVE_BENCHM	-						−0.013
							(0.344)
ETR		0.509	1.411	1.091	0.934	1.790	0.567
		(1.256)	(1.257)	(1.296)	(1.305)	(1.213)	(1.474)
SIZE		−0.189	−0.182	−0.185	−0.175	−0.204	−0.185
		(0.159)	(0.158)	(0.158)	(0.156)	(0.160)	(0.158)
ROA		8.357*	7.068	8.260*	8.418*	7.194	8.454*
		(4.504)	(4.358)	(4.344)	(4.344)	(4.498)	(4.445)
AUD		0.775	0.757	0.781	0.801	0.788	0.779
		(0.724)	(0.718)	(0.718)	(0.718)	(0.762)	(0.717)
ARSCORE		6.098**	6.403***	5.911**	5.775**	5.969**	6.147***
		(2.401)	(2.351)	(2.400)	(2.373)	(2.429)	(2.365)
PPE		−0.988**	−0.933*	−0.991**	−0.961**	−0.760	−1.020**
		(0.488)	(0.491)	(0.485)	(0.485)	(0.492)	(0.490)
R&D		32.019**	31.029**	31.792**	31.253**	29.901*	32.626**
		(15.100)	(14.821)	(15.057)	(14.762)	(15.530)	(15.228)
LEV		−0.200	−0.206	−0.413	−0.525	−0.103	−0.297
		(1.399)	(1.412)	(1.393)	(1.385)	(1.400)	(1.427)
Constant		0.345	−0.231	−0.241	−0.203	0.637	0.090
		(3.927)	(3.844)	(3.869)	(3.877)	(3.999)	(3.887)
N		699	699	699	699	699	699
Pseudo R²		0.200	0.208	0.202	0.202	0.224	0.195

Notes: This table presents regression results for the dependent variable M_REPORT with the different ETRCOND variables in columns (1) to (6). The models are estimated with Logit regressions. All variables are defined in Table 2. Standard errors clustered by firm are presented in parentheses. Industry-fixed effects (one-digit SIC level) and year-fixed effects are included but not reported. ***, **, and * indicate significance at 1%, 5%, and 10% levels respectively, in two-tailed tests.

Table

Panel B: Page
		1		2		3
Category		PAGE (1)	PAGE (2)	PAGE (3)	PAGE (4)	PAGE (5)	PAGE (6)
SMOOTH	+	0.007***
		(0.002)
BENCHM	+		0.395***
			(0.143)
DECR1	?			0.039**
				(0.019)
DECR2	?				0.048**
					(0.023)
VOLETR	-					−0.180***
						(0.040)
ABOVE_BENCHM	-						0.023
							(0.037)
ETR		0.018	0.096	0.079	0.066	0.171	−0.027
		(0.145)	(0.141)	(0.151)	(0.150)	(0.129)	(0.168)
SIZE		−0.009	−0.009	−0.008	−0.008	−0.011	−0.008
		(0.018)	(0.018)	(0.018)	(0.018)	(0.017)	(0.018)
ROA		−0.028	−0.063	−0.015	−0.019	−0.080	0.003
		(0.264)	(0.271)	(0.261)	(0.263)	(0.266)	(0.264)
AUD		0.069	0.066	0.069	0.071	0.072	0.069
		(0.090)	(0.090)	(0.091)	(0.091)	(0.090)	(0.091)
ARSCORE		0.630***	0.658***	0.622***	0.609***	0.642***	0.631***
		(0.226)	(0.220)	(0.227)	(0.229)	(0.216)	(0.226)
PPE		−0.133**	−0.123*	−0.137**	−0.133**	−0.107*	−0.137**
		(0.063)	(0.062)	(0.063)	(0.063)	(0.061)	(0.064)
R&D		0.814	0.794	0.780	0.761	0.711	0.894
		(0.830)	(0.837)	(0.844)	(0.842)	(0.803)	(0.851)
LEV		0.074	0.062	0.054	0.048	0.078	0.061
		(0.162)	(0.163)	(0.164)	(0.164)	(0.153)	(0.169)
Constant		−0.584	−0.611	−0.644	−0.638	−0.593	−0.612
		(0.433)	(0.426)	(0.429)	(0.432)	(0.411)	(0.433)
N		699	699	699	699	699	699
Adj. R²		0.166	0.167	0.156	0.157	0.195	0.152

Notes: This table presents regression results for the dependent variable PAGE with the different ETRCOND variables in columns (1) to (6). The models are estimated with OLS regressions. All variables are defined in Table 2. Standard errors clustered by firm are presented in parentheses. Industry-fixed effects (one-digit SIC level) and year-fixed effects are included but not reported. ***, **, and * indicate significance at 1%, 5%, and 10% levels respectively, in two-tailed tests.

Table 6: Family Firms

Panel A: Management Report
		1		2		3
Category		M_REPORT (1)	M_REPORT (2)	M_REPORT (3)	M_REPORT (4)	M_REPORT (5)	M_REPORT (6)
SMOOTH	+	0.039
		(0.025)
BENCHM	+		2.990***
			(1.032)
DECR1	?			0.516**
				(0.219)
DECR2	?				0.583**
					(0.269)
VOLETR	-					−1.240***
						(0.283)
ABOVE_BENCHM	-						−0.037
							(0.292)
FAMILY		1.763**	2.183*	2.351	1.612*	1.665*	1.579*
		(0.861)	(1.248)	(1.456)	(0.862)	(0.861)	(0.873)
SMOOTH*FAMILY		0.025
		(0.048)
BENCHM*FAMILY			17.023
			(18.484)
DECR1*FAMILY				−1.359
				(1.668)
DECR2*FAMILY					−0.510
					(1.675)
VOLETR*FAMILY						−1.649
						(1.893)
ABOVE_BENCHM*FAMILY							−0.322
							(1.715)
Controls		Yes	Yes	Yes	Yes	Yes	Yes
N		699	699	699	699	699	699

Notes: This table presents regression results for the dependent variable M_REPORT with the different ETRCOND variables in columns (1) to (6), including an indicator for family-run firms and interactions with the ETRCOND variables. All control variables from the main model (including industry- and year-fixed effects) are included but not reported. The models are estimated with Logit regressions. All variables are defined in Table 2. Standard errors clustered by firm are presented in parentheses. ***, **, and * indicate significance at 1%, 5%, and 10% levels respectively, in two-tailed tests.

Table

Panel B: Page
		1		2		3
Category		PAGE (1)	PAGE (2)	PAGE (3)	PAGE (4)	PAGE (5)	PAGE (6)
SMOOTH	+	0.007***
		(0.002)
BENCHM	+		0.400***
			(0.147)
DECR1	?			0.045**
				(0.021)
DECR2	?				0.058**
					(0.027)
VOLETR	-					−0.181***
						(0.040)
ABOVE_BENCHM	-						0.019
							(0.040)
FAMILY		0.052	0.026	0.086*	0.077	0.051	0.045
		(0.048)	(0.049)	(0.046)	(0.049)	(0.047)	(0.051)
SMOOTH*FAMILY		−0.005***
		(0.002)
BENCHM*FAMILY			−0.660
			(0.467)
DECR1*FAMILY				−0.049*
				(0.028)
DECR2*FAMILY					−0.072*
					(0.037)
VOLETR*FAMILY						0.101*
						(0.057)
ABOVE_BENCHM*FAMILY							0.080
							(0.063)
Controls		Yes	Yes	Yes	Yes	Yes	Yes
N		699	699	699	699	699	699
Adj. R²		0.168	0.168	0.157	0.159	0.195	0.154

Notes: This table presents regression results for the dependent variable PAGE with the different ETRCOND variables in columns (1) to (6), including an indicator for family-run firms and interactions with the ETRCOND variables. All control variables from the main model (including industry- and year-fixed effects) are included but not reported. The models are estimated with OLS regressions. All variables are defined in Table 2. Standard errors clustered by firm are presented in parentheses. ***, **, and * indicate significance at 1%, 5%, and 10% levels respectively, in two-tailed tests.

Table 7: ETR Decrease Level and Degree

Panel A: Level of Previous Year ETR
	M_REPORT (1)	M_REPORT (2)	M_REPORT (3)	M_REPORT (4)	PAGE (5)	PAGE (6)	PAGE (7)	PAGE (8)
DECRLEVEL_75-100	0.252				−0.141
	(0.987)				(0.106)
DECRLEVEL_50-75		−0.253				−0.098
		(0.481)				(0.062)
DECRLEVEL_25-50			0.610**				0.059***
			(0.253)				(0.021)
DECRLEVEL_0-25				−0.122				−0.029
				(0.319)				(0.032)
Controls	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
N	699	699	699	699	699	699	699	699
Adj. R²					0.153	0.154	0.162	0.152
Pseudo R²	0.195	0.195	0.205	0.195

Notes: This table presents regression results for the dependent variable M_REPORT (Logit regressions) in columns (1) to (4) and the dependent variable PAGE (OLS regressions) in columns (5) to (8). The DECRLEVEL variables are variations of DECR1, taking the value one if the ETR decreases in the current relative to the previous year and the ETR of the previous year is in the range included in the variable name, zero otherwise. All control variables from the main model (including industry- and year-fixed effects) are included but not reported. All variables are defined in Table 2. Standard errors clustered by firm are presented in parentheses. ***, **, and * indicate significance at 1%, 5%, and 10% levels respectively, in two-tailed tests.

Table

Panel B: Degree of ETR Decrease
	M_REPORT (1)	M_REPORT (2)	M_REPORT (3)	M_REPORT (4)	PAGE (5)	PAGE (6)	PAGE (7)	PAGE (8)
DECRDEGREE_0-5	0.659***				0.037*
	(0.233)				(0.019)
DECRDEGREE_5-10		0.147				0.025
		(0.335)				(0.031)
DECRDEGREE_10-20			0.248				0.052
			(0.381)				(0.045)
DECRDEGREE_>20				−1.000*				−0.125**
				(0.528)				(0.053)
Controls	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
N	699	699	699	699	699	699	699	699
Adj. R²					0.156	0.152	0.153	0.158
Pseudo R²	0.205	0.195	0.195	0.200

Notes: This table presents regression results for the dependent variable M_REPORT (Logit regressions) in columns (1) to (4) and the dependent variable PAGE (OLS regressions) in columns (5) to (8). The DECRDEGREE variables are variations of DECR1, taking the value one if the ETR decreases in the current relative to the previous year and the percentage points of the decrease are in the range included in the variable name, zero otherwise. All control variables from the main model (including industry- and year-fixed effects) are included but not reported. All variables are defined in Table 2. Standard errors clustered by firm are presented in parentheses. ***, **, and * indicate significance at 1%, 5%, and 10% levels respectively, in two-tailed tests.