TATT has been defined in various ways, and its measurement is not unequivocal (Patelli and Pedrini 2014). According to COSO (1992) TATT is seen as a relevant part of the control environment and is described as follows: “The tone set by top management is the most important factor contributing to the integrity of the financial reporting process. Notwithstanding an impressive set of written rules and procedures, if the tone set by management is lax, fraudulent financial reporting is more likely to occur.” Vaassen et al. (2009 p.66) consider TATT as “one of the most abused and misunderstood concepts in management control. It is about management giving the right example to their subordinates. By setting the proper TATT, management provides the most important contribution to shaping the organizational culture.” According to Vaassen et al. (2009, p.66), TATT can be defined as “the consistency between managers’ behaviour and their statements. The more consistency there is between managers’ behaviour and their statements, the stronger the organizational culture.” Since the COSO definition and that of Vaassen et al. (2009) largely overlap, they are used as a basis for this research.

According to Schein and Schein (2017), culture is a complex mixture of values, control systems, missions, structures and symbols that fit together in order to define what is appropriate for a particular organization. According to Hunt et al. (1989) organizations may have many values, for example, values that guide product and service quality, selection of distribution channels and treatment of customers. According to Sweeney et al. (2010), perceived ethical culture is an encouragement by top management to make ethical decisions and avoid unethical behaviour. The distinction between ‘ethical’ and ‘unethical’ behaviour as used in Sweeney et al. (2010) is also used in this article.

According to Lawton and Páez (2014) “the ethical dimension of leadership has, increasingly, been of interest, motivated partly by the corporate scandals that have involved the unethical behavior of top executives in leading organizations throughout the world”. One definition of ethical leadership often used is that of Brown et al. (2005, p.120). These authors define ethical leadership as “the demonstration of normatively appropriate conduct through personal actions and interrelationships, and the promotion of such conduct to followers through two-way communication, reinforcement, and decision-making”. Unethical leadership, as opposed to ethical leadership, is destructive and harmful. Brown and Mitchell (2010, p. 588) introduce a variety of unethical leadership styles, such as abusive supervision, supervisor undermining, toxic leadership, and tyrannical leadership. Research shows these leaders are oppressive, abusive, manipulative, and calculatingly undermining (Tepper et al. 2007).

Communication is an essential element for introducing the TATT (Swinkels 2003, p. 26). Communication of the right TATT can be effected by, amongst other, a code of ethics set by top management. Such a code aims to reduce ambiguity as well as increase consistent behaviour (Stevens 1999). Other available instruments for effecting TATT are compliance programs and leadership by example. In line with the first objective of this article, i.e. an attempt to measure TATT, the CEO letter is introduced as a proxy for TATT. In the next section an explanation is provided for the use of this proxy.

Companies provide a significant amount of information through their annual reports. Annual reports consist of two sections, financial statements and narrative sections (Balata and Breton 2005). Several studies evaluate the information content of the narrative disclosures, such as the Management Discussion and Analysis (MD&A), the officer’s comment in earnings press releases and CEO letters (Kearney and Liu 2014). Thus, qualitative information from management narrative disclosure can act as a complement to financial information. There is empirical evidence (e.g., Segars and Kohut 2001) that the CEO letter to shareholders communicates important aspects of TATT, thereby exerting important social influence over stakeholders. It is noteworthy that many prominent CEOs have publicly acknowledged their strong personal role in drafting CEO letters to shareholders (Jones 2005).

Until now, research on TATT in CEO letters has mainly been done in a US context (Amernic et al. 2010; Craig and Brennan 2012; Dikolli et al. 2019; Patelli and Pedrini 2014). This research however does not quantify the TATT. This article uses the CEO letters of AEX-listed companies to quantify TATT and bridges the gap by performing research on CEO letters within a Dutch corporate environment. Quantification of TATT provides insight in both the TATT-score per individual company and the (relative) TATT-score in comparison with other companies. How TATT is quantified is described in section 3.2.

Sections 2.1 and 2.2 focus on the proxy variable of this article, i.e. the “(un)ethical TATT”. This variable is based on analyzing CEO letters. In addition to the construct of this proxy variable, a list of possible explanatory variables is presented, offering possible explanations for a certain level of TATT. These possible explanatory variables are based on a review of (academic) resources (e.g., Amernic et al. 2010; Craig and Amernic 2011; Nakashima and Ziebart 2016).

For example, one of the identified variables is Internal audit environment. As is stated by Hansen, Stephens, Wood (2009) “we identify six potential ways the Internal Audit Function may increase its ability to improve the tone at the top”. Also Nakashima and Ziebart (2016), state “The results indicate a positive link between TATT and […] audit quality”. Therefore, it is assumed that there is a possible positive impact between a strong internal audit environment and TATT.

The outcome of this analysis is presented in Table 1.

However, in view of the explorative nature of this article, Table 1 can be regarded as an initial attempt to identify variables that may have impact on the proxy variable TATT. When discussing the out-of-range CEO letters (see section 4.4), this list of variables is used.

Table 1.Download as CSV 

Explanatory variables and respective impact on TATT.

Explanatory variable	Impact on proxy variable TATT
Code of conduct	+
Compensation plans	-
Corporate Governance	+
Difficult operating circumstances	-
Internal audit environment	+
Internal control environment	+
Investments and disinvestments	-
Risk taking culture	-
Supervisory Board	+
Compliance with law and regulations	+

Based on the previous sections a theoretical framework is constructed in Figure 1.

Figure 1.

Theoretical framework.

TATT is considered to be the proxy variable in the study. TATT is operationalized by analyzing CEO letters with the use of text analysis. This quantitative analysis results in measuring and visualizing possible outliers of “(un)ethical TATT” (see section 4).

In this framework, the term TATT is introduced twice. In the upper box, the representation of TATT is of a qualitative nature; TATT represented in the lower box is the quantitative representation after analyzing CEO letters, i.e., the score, measured by the textual-analysis program.

The applicable out-of-range CEO letters, i.e. a score more than 2 standard deviations from the mean resulting from text analysis, are further analyzed by close reading, taking into account the scores from text analysis as well as the identification of possible explanatory variables. The combination of both the analytical power of text analysis and close reading is in line with previous research, such as Craig and Amernic (2016), Murphy (2013) and Amernic et al. (2010). Close reading complements text analysis, because that is where “the nuances of a message are to be found” (Murphy 2013, p. 57).

The sample selection consists of companies listed at the AEX index from 2007 till 2016. The AEX index is composed of 25 of the most frequently traded securities on Euronext Amsterdam. Table 2 contains the data set.

Table 2.Download as CSV 

Sample selection.

Population, based on total number of AEX companies	251*
Minus unavailable annual reports	-/- 10
Minus unavailable CEO letters	-/- 23
N in analysis	218
% available	87%

* In 2013, 26 companies were part of the AEX index.

For the analysis of CEO letters, as an indicator for TATT, a text analysis program named DICTION is used. DICTION has a dictionary-based approach and uses a built-in database consisting of 50,000 previously analysed texts (DICTION manual). DICTION uses 31 standard dictionaries (word lists) to search a given text. DICTION processes the text, looking for an exact match of the words contained in the 31 standard dictionaries (Diction manual). The 31 dictionary measures have labels such as ‘numerical terms’, ‘ambivalence’, ‘self-reference’, and ‘tenacity’. In addition, based on the text processing described above, DICTION produces measures of 4 calculated variables (‘insistence’, ‘embellishment’, ‘variety’ and ‘complexity’) by linguistically based methods of calculation.

The above-mentioned DICTION-related 31 dictionary measures, as well as the 4 calculated variables (35 in total) are used to calculate the 5 master variables of DICTION (certainty, activity, optimism, realism and commonality).

The definitions of these 5 master variables (DICTION manual) are mentioned below:

• Certainty – Language indicating resoluteness, inflexibility, completeness and a tendency to speak ex cathedra;
• Activity – Language featuring movement, change, the implementation of ideas and the avoidance of inertia;
• Optimism – Language endorsing some person, group, concept or event, or highlighting their positive entailments;
• Realism – Language describing tangible, immediate, recognizable matters that affect people’s everyday lives;
• Commonality – Language highlighting the agreed-upon values of a group and rejecting idiosyncratic modes of engagement.

These 5 master variables are used as building blocks for quantifying TATT.

After importing the CEO letters into DICTION, the program calculates the master variables, derived from the 35 underlying variables. To be able to operationalize the concept of TATT, an all-inclusive TATT score is calculated. The construct of the TATT score is based on DICTION’S 5 master variables. Some variables have a positive relationship with TATT, whilst others have a negative connotation. The formula for the TATT score is:

(2 ×Realism + 2 ×Commonality - 1 ×Activity - 1 ×Optimism - 2 ×Certainty) ÷ 8

Both the use of a positive/negative connotation as well as the use of different weighting factors, per master variable, are based on Patelli and Pedrini’s research (2014, p. 14, Table 6). In their research, they found evidence that three out of the five thematic indicators were statistically associated with financial reporting aggressiveness, an artefact of a negative TATT. Therefore, in this research, the thematic indicators of realism, commonality and certainty are multiplied by 2. Remaining variables have a weighting factor of 1.⁴

To determine outliers in TATT scores in this analysis, a lower and upper bound range are constructed by subtracting and adding 2 standard deviations from the mean value per master variable. The choice for applying 2 standard deviations is in line with research of Amernic et al. (2010, p. 109 and p. 131). In their research, also based on the exploration of CEO letters in combination with DICTION, the authors applied 2 standard deviations from the mean for every DICTION variable, in order to define out-of-range CEO letters. By using the 2 standard deviations from the mean value, a 95% interval is implied.

Subsequently, the out-of-range CEO letters in this research are studied in a qualitative manner by means of close reading. Close reading is one of the available qualitative methods for assessing the TATT of a company. This is in line with the research design of Amernic et al. (2010), in which close reading as a qualitative method is used. As presented in the theoretical framework (section 2.4), ‘close reading’ takes place based on pre-determined explanatory variables (section 2.3). It is important to mention that this research aims to explore the presence of possible explanatory variables (as included in the theoretical framework) in the CEO letters, not to validate them, since this would require additional data from other sources than CEO letters and additional statistical analysis thereon.

The full data set contains 218 individual observations. The time span is 10 years (2007–2016). In this period 38 companies were part of the AEX index and under examination for this article. Some companies are only present for one year, while others are represented in the AEX index for the full 10 year period. As discussed in section 3.2, 35 variables are used to calculate the 5 master variables. Please see Table 3 for a break-down of the 5 master variables for the full data set.

Applying descriptive statistics (e.g., standard deviation) to the data set implies a normal distribution of the underlying data points. In order to interpret Table 3, the concept of a normal distribution is relevant. In order to test whether a normal distribution is applicable to the data set, the Shapiro-Wilk test is conducted in Stata. The resulting p-value is higher than 5%, indicating that the null hypothesis of normality is not rejected.

Table 3.Download as CSV 

Descriptive statistics full data set (N = 218).

	Certainty	Activity	Optimism	Realism	Commonality
	(-/-)	(-/-)	(-/-)	(+)	(+)
Median	49.420	50.505	54.360	53.615	49.330
Modus	50.470	50.080	55.630	52.330	49.590
Average	48.836	50.654	54.423	53.382	49.326
Maximum	54.720	58.230	62.590	60.260	56.830
Minimum	37.510	46.110	47.020	44.420	44.550
ST.DV	2.708	2.152	2.853	2.669	1.933
2x st.dv	5.417	4.303	5.705	5.337	3.867
OUTLIER (-2x st. dv)	43.420	46.350	48.717	48.045	45.459
OUTLIER (+2x st.dv)	54.253	54.957	60.128	58.719	53.192

In line with other research (e.g., Amernic et al. 2010), an out-of-range CEO letter is based on a deviation from the mean of at least 2 times the standard deviation on a two-tail basis. For example, a score for a positive-related master variable on the left side of this threshold implies a less than average positive impact on the TATT score. A score for a positive-related master variable on the right side of this threshold implies a more than average positive impact on the TATT score.

As introduced in section 3.2 the TATT score is constructed based on the five master variables.

Subsequently, the TATT for the full data set is calculated. Table 4 shows the results.

Table 4.Download as CSV 

TATT scores based on full data set (N = 218).

Item	TATT
Median	0.328
Modus	0.675
AVERAGE	0.333
ST.DV	1.143
2x st.dv	2.286
OUTLIER (- 2x st. dv)	-1.953
OUTLIER (+2x st.dv)	2.619

Figure 2A shows the development of the average TATT, calculated per year for AEX companies, over a 10 year period. In addition, also the average scores for each master variable per year are presented.

In addition, Figure 2B shows a relatively stable, positive average TATT score. No average TATT score for a single year during this 10 year period deviates substantially from other years.

Figure 2.

TATT development over time.

In order to answer the CRQ, the total data set (N = 218) is used to identify the outliers c.q. the out-of-range CEO letters. Table 5 shows the outliers. This table shows 11 out-of-range CEO letters. This is 5% of the total data set (N = 218). In line with the research design in section 3, this quantitative analysis is complemented by a qualitative review (close reading) of these 11 CEO letters.

Table 5.Download as CSV 

Outliers.

#	Company	Significantly rated above average TATT vs Significantly rated below average TATT
1	ASML 2007	Significantly rated above average TATT
2	BAM 2009	Significantly rated below average TATT
3	Unilever 2009	Significantly rated below average TATT
4	Wolters Kluwer 2009	Significantly rated below average TATT
5	Unibail Rodamco 2010	Significantly rated below average TATT
6	Unibail Rodamco 2011	Significantly rated below average TATT
7	Philips 2012	Significantly rated below average TATT
8	OCI 2014	Significantly rated below average TATT
9	TNT 2014	Significantly rated below average TATT
10	RELX 2007	Significantly rated below average TATT
11	Arcelor Mittal 2010	Significantly rated above average TATT

Two important remarks have to be made about this table:

1. The above-mentioned table needs additional explanation. For example, the out-of-range TATT-score of the CEO letter TNT 2014 can be partly explained by the out-of-range score on one of the master variables, i.e. “Certainty”. A number of selected sentences in this CEO letter can be regarded as an indication of a high score on this master variable, i.e. “Certainty”. For each out-of-range TATT-score in Table 5, this qualitative analysis has been executed.
2. In this table, the outcome “Significantly rated above average TATT” indicates a TATT-score more than two standard deviations above the average TATT-score; consequently the outcome “Significantly rated below average TATT” indicates a TATT-score more than two standard deviations below the average TATT-score.

In the aforementioned section, 11 outliers are determined. These letters qualify for the next step of analysis: close reading. Exceptional sentences in each out-of-range CEO letter, indicating an out-of-range score on one or more master variables, are determined. The explanatory variables checklist is used to read and analyse each letter and an initial attempt is made to link one or more potential explanatory variables (section 2.3.) to each CEO letter.

The following steps for each out-of-range CEO letter are executed:

1. Determining exceptional sentences, based on close reading;
2. Comparing close reading findings with DICTION scores; and
3. Identifying relevant explanatory variables.

In the close reading exercise of section 4.4, one or more explanatory variables per out-of-range CEO-letter are identified. Table 6 shows the identified explanatory variables including the frequency of each explanatory variable, as identified in the 11 out-of-range CEO letters.

Table 6.Download as CSV 

Frequency of identified explanatory variables, based on close reading.

#	Explanatory variables	Frequency total	Frequency ethical	Frequency unethical
1	Difficult operating circumstances (-)	8	1	7
2	Investments and disinvestments (-)	4	0	4
3	Supervisory Board (+)	3	1	2
4	Corporate Governance (+)	1	0	1
5	Compliance with law and regulations (+)	1	0	1
6	Internal control environment (+)	0	0	0
7	Internal audit environment (+)	0	0	0
8	Risk taking culture (-)	0	0	0
9	Code of Conduct (+)	0	0	0
10	Compensation plans (-)	0	0	0

The use of the colour red indicates that the finding does not correspond with what is presented in Table 1.

The following findings can be derived from Table 6.

1. In eight CEO letters, the explanatory variable ‘difficult operating circumstances’ was identified. Based on the theoretical framework, this could have a negative impact on TATT. Seven out of eight findings correspond with the theoretical framework.
2. In four CEO letters, the explanatory variable ‘investments and disinvestments’ was identified. Based on the theoretical framework, this could have a negative impact on TATT. All findings correspond with the theoretical framework.
3. The explanatory variable ‘Supervisory Board’ is mentioned in three CEO letters. In the theoretical framework, a possible positive impact was defined between a change in the composition of the Supervisory Board and TATT. However, this possible positive impact only holds for one of the three CEO letters, as shown in Table 6.
4. The explanatory variables “Corporate governance” and “Compliance with law and regulations” are mentioned once in two separate CEO letters. The possible positive impact of both variables on TATT, based on the theoretical framework, does not appear in our analysis.
5. Five explanatory variables are not mentioned in any of the CEO letters under examination. This could be due to the fact that a CEO letter is a non-regulated document. Therefore, their impact on TATT does not appear in our analysis.

The CRQ of this study is:

‘To what extent is it possible to quantify Tone at the Top at AEX-listed companies from 2007 to 2016 and consequently, what findings and trends can be derived from the results?’

Based on the research conducted, the answer to the research question is twofold:

1. Quantification of TATT is possible by using CEO letters as a proxy. These CEO letters are analyzed in DICTION, resulting in 5 master variables. Based on a literature review, e.g., Patelli and Pedrini (2015); Schrand and Zechman (2012) and Hooghiemstra (2000) – these master variables are used to construct (relative) “ethical” or “unethical TATT”.
2. The main finding is that the average TATT is relatively stable for AEX companies over time. However, on closer examination, relevant deviations from the mean TATT are identified for a number of AEX-listed companies. The resulting 11 outliers are further analyzed with close reading. This analysis generates a first step to determine explanatory variables which are associated with TATT: difficult operating circumstances, investments and disinvestments, changes in composition of Supervisory Boards, corporate governance and compliance with law and regulations. This is a first step towards validating our theoretical framework.

This article contributes to academic research and also has a few relevant practical implications. However, the research also comes with some caveats, grouped in 4 categories: