Attention to COVID-19 –
A content analysis study of Swedish interim reports
Bachelor’s Thesis 15 hp
Department of Business Studies Uppsala University
Fall Semester of 2020
Date of Submission: 2021-01-15
Andreas Ström
Supervisor: Lars Frimanson
Abstract
The purpose of this study was to examine the attention to the COVID-19 pandemic displayed by top level management in companies listed on Nasdaq OMX Stockholm Large Cap, and how this afore- mentioned attention was affected by the board size and board gender diversity of respective company.
To accomplish this, a content analysis on a word-by-word level was conducted of all interim reports produced in 2020 by each company, and data regarding the board size and board gender diversity was gathered, for each company. The frequency with which each company mentioned select keywords concerning the COVID-19 pandemic was measured and then used as a comparative measure of the attention to the COVID-19 pandemic. In order to determine the magnitude of the impact of the independent variables a control variable, firm size, was introduced and linear fits were constructed of various combinations of variables. The resulting fits all clearly displayed an absence of correlation between either board size or board gender diversity and the attention paid to the COVID-19 pan- demic by top level management in large Swedish companies. Hence, this study suggest that there is no increase in board activity regarding daily operations in large Swedish companies during crises.
Keywords: Attention, COVID, Corona, Board Size, Board Gender Diversity, Board Structure, Con-
tent Analysis, Semantic Analysis
Contents
1 Introduction 1
2 Theory 3
2.1 Corporate attention . . . . 3 2.2 Board size . . . . 4 2.3 Board gender diversity . . . . 6
3 Hypothesis development 8
3.1 The impact of board size on corporate attention to crises . . . . 8 3.2 The impact of board gender diversity on corporate attention to crises . . . . 8 3.3 Analytical model . . . . 9
4 Method 10
4.1 General process . . . . 10 4.2 Literature review and Content analysis . . . . 12 4.3 Implementation and Statistical Model . . . . 14
5 Results and Analysis 16
5.1 Analysis omitting the control variable . . . . 17 5.2 Analysis considering the control variable . . . . 21
6 Discussion 25
7 Summary 27
7.1 Conclusion . . . . 27 7.2 Sources of Error and Future Work . . . . 28
References 38
Appendix 39
1 Introduction
At the time of writing, the COVID-19 pandemic has encompassed most of society, had a tremendous impact on many aspects of modern life, and brought with it previously unprecedented issues for hu- manity. Less than a year into the pandemic, there have been 50 million cases and a million deaths (European Centre for Disease Prevention and Control, 2020). The pandemic has unquestionably put a strain on the global economy and society in general, and the resulting situation is one of both new challenges and new possibilities. From an academic viewpoint, the pandemic is a unique possibility to study the effects of a great crisis, which will be the overall aim of this report, in order to further advance the understanding of the effects of crises on modern society. Major questions pertaining to the long-term effects of the pandemic on the economy still remain unanswered today, and most will likely not be definitively answered in the near future. This has, however, not deterred a large amount of research from being conducted as a start to mapping the pandemic’s immediate effects.
Throughout this pandemic it has become evident that attention to and management of crises can be a crucial trait for companies to possess, not only for their own sake, but also for society as a whole.
The modern literature pertaining to crisis management has studied a variety of dependencies, includ- ing board size and board gender diversity. The attention literature, pertaining to crises, focused on board size generally finds that a larger board size decreases risk taking, and thus generally increases the crisis preparedness (Gallud Cano et al., 2020; Huang and Wang, 2015; Nakano and Nguyen, 2012;
Pathan, 2009). The equivalent literature focused on board gender diversity is less uniform. Argu-
ments that board gender diversity improves firm performance during crises (Sun et al., 2015), due
to a more effective risk management strategy as a product of a larger amount of female directors
(Chen et al., 2016), are not uncommon. However, there are also studies arguing that this is a spurious
correlation, i.e. that board gender diversity has no impact on firm performance during crises (Engelen
et al., 2012). More general articles have also been written studying corporate attention of crises in
Greek companies after the Greek government-debt crisis (Vouzas and Nizamidou, 2018), and studies
discussing how stakeholder theory is superior to shareholder theory when it comes to proactive crisis
management (Alpaslan et al., 2009). However, it is clear that substantial board structure research
Building on this knowledge, this study aims to perform a content analysis on interim reports in order to explore whether there is a correlation between either the size or gender diversity of boards of directors at Swedish companies and the attention to the COVID-19 pandemic displayed by the top level management of respective company. This will contribute to current research by giving insight into the aforementioned relationship, as well as comparing the strength of the relationship to that of the control variables using regression analysis.
This report is structured as follows. In section 2, the underlying theory regarding board size and
gender diversity is presented, which the hypothesis development (section 3) summarizes and con-
cretizes. The content analysis methodology and regression model are discussed in section 4, as well
as results presented in section 5, and discussed in section 6. Concluding statements are presented in
section 7.
2 Theory
Within this section previous research regarding corporate attention board size, and board gender diversity is discussed, as well as its usefulness to this study.
2.1 Corporate attention
The expression “attention” is in itself quite broad. Within the world of corporate research the concept
“attention-based view” has come to be quite popular, which is a method of analysis that has proven
useful a number of times (Barnett, 2008). Attention is primarily, within this field, what someone or
something is focused on. Sullivan (2010) demonstrates quite concisely how an attention-based view
can be utilized to study corporations. She does this through examining primarily which property
decides what problem the attention of management teams is focused on, finding that urgency is a
dominant reason. A critical factor that may limit the generalization of her study is however that the
study is focused on a government agency. This finding is still another reason to study the attention
of top level management during COVID-19, as one can assume that pandemic-related problems hold
a substantial sense of urgency. Similarly, Greve (2008) discusses the matter of sequential attention
and in which order management prioritizes which problem in conjunction with what goals holds their
attention. There is also research (primarily based on quantitative data, which could be problematic
considering the following conclusions) suggesting that the attention companies pay to risk shifts with
company size and age (Wennberg et al., 2016), and that an inability to detect and remedy issues
within an organization, due to a disparity in attention, could be detrimental to a company (Rerup,
2009). Another shift in managerial attention is, however, found to be due to deregulation, which is
not in conflict with the theory that a shift due to size or age exists, but may instead be another real-
ization of the same effects (Cho and Hambrick, 2006). Other related research regarding the structures
of attention have considered how new issues are introduced into an organization’s already existing
hierarchy of problems with various levels of urgency (Joseph and Wilson, 2017), or how international
firm performance is correlated to top level executives time devoted to understanding the global mar-
ket (Bouquet et al., 2009). An issue in these latter studies does arise, as they are primarily focused
on either data gathered from a questionnaire, or a singular case study. Both articles discuss these issues, but the effect of a limited generalization of the results remain. It is also hypothesised that varying types of management control affects the way attention manifests itself in a company (Ocasio and Wohlgezogen, 2010). The collection of research presented here does, hence, imply that when such a considerable issue as COVID-19 appear, most companies attention should shift towards this issue.
There is, however, a few principal critiques that should be brought to light concerning prior research utilizing an attention-based view. A substantial fraction of previous research, including many of the aforementioned articles, neglects the impact of diverging interests withing their respective subjects of study, which in turn could skew both results and the hypothesised explanations for these, an effect that is hinted at in e.g. the article by Wennberg et al. (2016). Secondly, the absence of differentiation between the origin and the consequences of a shift in attention, which can have a similarly skewing influence of the results. An example of both aforementioned effects can be shown in the article by Rerup (2009) (Joseph and Wilson, 2017; Rerup and Salvato, 2012). Within this study the first of these critiques will not be taken into account, whilst the second is minimized naturally due to the nature of the methodology of the study.
2.2 Board size
There exists a large amount of research that uses board size as an independent variable. Within the
field the causal relationship between board size and firm performance is often studied (Dwivedi and
Jain, 2005), and a majority of studies used in this review find a robust negative relationship between
the board size and various economic indicators of firm performance. In Boone et al.’s (2007) “The
determinants of corporate board size and composition: An empirical analysis” the notion that an
increased board size leads to a trade off in firm-specific benefits (earlier proposed by Raheja (2005))
is rigorously supported. Similarly Bennedsen et al. (2008) show a negative correlation between board
size and firm performance in larger firms, but provide inconclusive results for small and medium size
firms. Eisenberg et al. (1998) do, however, provide conclusive evidence that this notion also holds
true for Finnish small and medium size firms. This is further supported for companies in Singapore
and Malaysia by Mak and Kusnadi (2005), for Thai banks by Pathan et al. (2007), and for companies in India by Garg (2007). Conyon and Peck (1998) also shows that board size in general is negatively correlated to firm performance for European companies. However, this negative correlation is not always the case as show by both Kyereboah-Coleman and Biekpe (2008) whom did not find a signif- icant correlation, and Mohamed (2009), whom examined a large number of banks and their Tobin’s Q ratio (a measure of assets value first proposed by Kaldor (1966)), and found a positive correlation between board size and firm performance.
However, a central concept many of the aforementioned articles, concerned with the relationship of board size and firm performance, neglects is the endogeneity within the subjects of their research.
A large number of these studies finds a concrete causal relationship, but fail to consider the possibil- ity of their data being endogenous, i.e. that there exists a possibility of the causality being fully or partially reversed, meaning that previous firm performance actually would affect board size (Wintoki et al., 2012). Interestingly, other research papers also claim that a large amount of previous research within the area rather is wrongly interpreted and, in fact, does not show this clear negative correlation.
Notable are Bennedsen et al.’s (2008) “The causal effect of board size in the performance of small and medium-sized firms”, and Garc´ıa-Ramos and D´ıaz’s (2020) “Board of directors structure and firm financial performance: A qualitative comparative analysis”, where the possibility of misinterpretation and various conflicting theoretical models are discussed.
It has also been hypothesised that this perceived negative correlation is due to the varying deci-
sive power of differently sized boards. Cheng (2008) finds that larger boards tend to make more
conservative decisions, and speculate as to whether or not this has a significant impact on board per-
formance. Huang and Wang (2015) do, however, find that smaller boards are more conservative when
it comes to decisions regarding debt financing. Within this field of study the concept of accountability
is central, often it is found that accountability decreases with board size, which one can assumes leads
to a lowered attention of crises (Omar et al., 2013). It needs to be noted that this assumption mainly
is founded in quantitative data, meaning that it is not founded on any interpretative component and
should be taken as such. All in all, the practice of using board size as an independent variable is widespread. This is positive as it can provide in-depth guidance on the subject if necessary.
2.3 Board gender diversity
The literature on gender diversity within boards as an independent variable is more heterogeneous than the literature examined in the section above, even if its effect on firm performance is a recurring subject. Within this subject area the impact of gender diversity on the probability of stock crash or bankruptcy is also popular. Adusei and Obeng (2019) examines whether bankruptcy becomes more or less probable, and finds that the latter holds true for boards showing greater gender diversity.
Qayyum et al. (2020) displays that gender diversity has a negative correlation to the stock price crash risk on the Asia-Pacific market, and Jebran et al. (2020) supports this further for specifically China. However, Shoham et al. (2020) shows that gender diversity has a negative impact on the phe- nomenon of company cross-listing. Goodstein et al. (1994) further shows that board diversity may be a significant constraint on strategic change in turbulent times, and Adusei (2019) shows that gender diversity may have a negative impact on “technical efficiency”. These latter notions are, however, disputed, e.g. Ahmadi et al. (2018), Merve and Cemil (2016), Ozdemir (2020), and Song et al. (2020) all show a positive correlation between gender diversity and firm performance, throughout the world.
Low et al. (2015) also supports this, but further brings to light the finding that these positive effects seem to be diminished in countries with higher female economic participation and empowerment.
They reason that this most likely is due to tokenism diminishing firm performance. Similarly, Liu
et al. (2014) support that gender diversity has a positive impact on firm performance, but only in so
called legal person-controlled firms, and not in state-controlled firms (in China). Li and Chen (2018)
show that the same is true if, and only if, the firm’s value does not surpass some critical value. A
study in the Netherlands and Denmark did not find a significant correlation between the two at all
(Marinova et al., 2016), and other studies found that gender diversity can be directly harmful by
impeding strategic change, and hence diminishing firm performance (Kouaib and Almulhim, 2019;
Triana et al., 2014). Despite the varying conclusions made in previous research, the practice of using gender diversity within boards as an independent variable is widespread.
Similarly to the research concerned with board size, the issue of endogeneity arises within this field
too (Wintoki et al., 2012). A large fraction of the literature within this field concerned with firm
performance and its relation to board gender diversity overlooks the impact earlier firm performance
has on the board of directors of a company. Although this issue is not as grave as when considering
board size, it can still skew eventual results substantially. Furthermore, the generalized results within
this field are few, as many studies are limited to e.g. certain geographical areas, which reasonably has
a large impact on the results due to a large variance in political motifs such as quotation legislation
(Kagzi and Guha, 2017). This notion is also supported by the heterogeneity in results between studies
(see (Adusei, 2019; Adusei and Obeng, 2019; Jebran et al., 2020; Qayyum et al., 2020; Shoham et al.,
2020)). Criticism regarding the rigidity and possible manipulation of financial data has also been
brought forward, suggesting that data based purely on financial records may not be entirely reliable
(Dezs˝ o and Ross, 2011), a consideration occasionally omitted. Despite aforementioned limitations
within earlier research, in-depth guidance on the subject can still be obtained from earlier research.
3 Hypothesis development
This section concretizes the purpose of this study, by summarizing the aforementioned literature and developing two hypotheses.
3.1 The impact of board size on corporate attention to crises
The literature pertaining to board size and attention to crises is, at times, somewhat conflicting. On the one hand smaller boards tend to make more conservative decisions when it comes to debt financing (Huang and Wang, 2015) and display a higher accountability (Omar et al., 2013), but on the other hand larger boards tend to, in general, make more conservative decisions (Cheng, 2008). When it comes to risk taking however, within the field of risk management and crisis attention, there exists a clear majority of articles advocating for the moderateness of larger boards (Gallud Cano et al., 2020;
Huang and Wang, 2015; Mary et al., 2014; Nakano and Nguyen, 2012). This uniformity in previous research leads to the first hypothesis of the study:
H1: An increased board size positively impacts the attention paid to the COVID-19 pandemic by top level management.
3.2 The impact of board gender diversity on corporate attention to crises
The subject of board gender diversity and its impact on risk management and crisis attention is
less homogeneous than it’s equivalent regarding board size. This research field displays articles on
both sides of the spectra. One side argues the point that board gender diversity significantly impedes
strategic change in turbulent times (Goodstein et al., 1994; Kouaib and Almulhim, 2019; Triana et al.,
2014). Whereas the other end of the spectra argues that the probability of bankruptcy and stock crash
significantly decreases with an increased board gender diversity (Adusei and Obeng, 2019; Jebran et
al., 2020; Qayyum et al., 2020). Between these two there also exist articles who found no correlation
of the two (Engelen et al., 2012; Marinova et al., 2016). Furthermore, the positive effects of board
gender diversity seem to be diminished in countries displaying higher female economic participation
and empowerment, which is the case in Sweden (Ortiz-Ospina, 2018), and is reasoned to be due to to- kenism (Low et al., 2015). However, all in all, there seems to exist a majority of research arguing that increased board gender diversity improves firm performance during crises (Sun et al., 2015), due to an improved crisis management (Chen et al., 2016). This produces the second hypothesis of this study:
H2: An increased board gender diversity positively impacts the attention paid to the COVID-19 pandemic by top level management.
3.3 Analytical model
Figure 1 shows the analytical model of the study, hypothesising that board size and gender diversity are positively correlated to the attention to COVID-19 the top level management of a company displays.
Board size
Gender diversity
Covid-19 Attention to
Figure 1: Analytical model of the problem, displaying independent variables (to the left) and their
impact on the dependent variable attention (to the right).
4 Method
This section begins by presenting, and advocating for, the general methodology used in this study, followed by the specifics of how the literature review was conducted and helpful material regarding the content analysis. The final part discusses explicitly how the data was gathered and processed.
4.1 General process
In order to study the produced hypotheses three components were needed. These are measures of the
board size, the board gender diversity, as well as the attention of COVID-19 of top level management,
for each company studied. The set of companies that was used in this study consisted of the compa-
nies of the OMXS Large Cap list. Specifically this list was chosen due to convenience and simplicity,
as these companies can be assumed to follow standard financial reporting procedures, be considered
as legitimate, and operate with a board structure more familiar to the author (compared to a typical,
say, American board structure). The first two components needed to study these companies, board
size and gender diversity, are the simpler ones. These have been manually collected for each company
by visiting their web pages and recording the total number of board members and the number of
female board members. The board gender diversity was then quantified as the fraction of female
board members divided by the total amount of board members. The board size was normalized by
dividing each individual company’s board size by the maximum board size found, in order to simplify
the calculations. The quantification of top level managements COVID-19 attention can be considered
to be the most vague measurement out of the three, in this study it will be considered to be the
frequency with which a company mentions a set of keywords in their interim reports, i.e. the number
of times a keyword is mentioned divided by the total amount of words written in each report. The
set of keywords used consists of “covid”, “corona”, “epidemic”, and “pandemic”, and all of the afore-
mentioned keywords various inflections. This can be considered to be one of the more basic variations
of content analysis, as it only analyses the reports on a word-by-word level, as opposed to a sentence
or paragraph level, further discussed in subsection 4.2. However, this form of content analysis is still
a useful tool and has been successfully used numerous times (Slattery, 2014). The reasoning behind
quantifying the attention to the COVID-19 pandemic of top level management as this aforementioned frequency arises from the field of sentiment analysis, where the frequency of keywords mentioned is interpreted as indicators of the author’s underlying opinions or beliefs. This form of analysis has become more popular with the rise of artificial intelligence and machine learning technology, and has been widely used to study corporate financial reports (H´ ajek and Olej, 2013; Pagliarussi et al., 2016;
Ren et al., 2013). The practicalities of recording the frequencies used in this study are discussed below, in subsection 4.3.
Furthermore, throughout this study the term “top level management” is used. This has the primary purpose of encompassing both the board of directors and senior management within the companies.
The necessity for a term such as “top level management” arises when one takes a closer look at the
purpose of this study, which is to examine the attention to COVID-19 of top level management,
based on the frequency of mentioned keywords in the interim reports of respective company, and
see how this varies depending on the board of directors of respective company. Within this purpose
there seems to be a logical fallacy arising, as the board of directors do not actively have a hand in
producing the interim reports. There are, however, reasons to question this fallacy. It is true that
the board of directors do not have a hand in the day-to-day, and are not immediately responsible for
the interim reports of a company, but they are responsible for choosing a CEO and are ultimately
responsible for his or her decisions. This creates an indirect coupling between the board of directors
and the interim reports of each company, through the CEO and senior management. It is also true
that the board of directors should have a more long-term plan in sight, as opposed to the CEO’s more
direct operational planning and decision making. This has been, and continues to be, the way that
the board of directors interacts with senior management. During crises the boards of directors field of
practice does, however, expand. It is still the CEO’s responsibility to handle the day-to-day, but the
board needs to be one of action, in order for the company to perform well. Boards have an unique
possibility to engage in a supportive role to senior management, which is more crucial than ever
during this pandemic. Allowing the senior managers to focus on immediate threats and opportunities
with the confidence that a competent set of board members, focused on immediate fundamentals and
dealing with secondary threats, can keep the company aimed at its long-term goals (Deloitte, 2020;
von Post and Pozen, 2020). This may be an unorthodox practice, but, as put by Konigsburg et al.
(2019):
In a crisis, when the stakes are high and scrutiny is intense, the board has a unique role.
Stepping in may be uncomfortable, but stepping aside is not an option.
Perhaps a more direct research question would have been focused on the structure of the senior management team, as opposed to the board structure, but considering that the scope of this study is not wide enough to encompass both, and this study has a focus on corporate governance, the decision to focus on the board of directors was made.
4.2 Literature review and Content analysis
To structure the literature review a few limitations were set. Due to the large sample size of pre- vious research, it was determined that studies conducted within the last 30 years were to be used.
To review the use of board size as an independent variable the set of selected keywords chosen were
“board”, “size” and “corporate”. To study earlier literature pertaining to board gender diversity as independent variable the keywords “board”, “gender”, “diversity”, and “corporate” were used. The keywords “crisis”, “attention”, “management”, “corporate”, and “risk” were used to conduct the lit- erature review of corporate attention. The review of previous content analysis literature is not limited to the last 30 years, due to the smaller sample size. The set of chosen keywords were here “content”,
“analysis”, “annual”, “quarterly”, “interim”, “financial”, and “report”.
By studying the literature relating to previous content analyses, and how these are used to determine
the dependent variable, a deeper understanding of the methodology can be obtained. An example of
content analysis applied to annual reports, similar to the methodology of this study, is the study “The
Importance of Information Technology: An Empirical and Longitudinal Study of the Annual Reports
of the 50 Largest Companies in the United States” by Peslak (2005). In this study content analysis is
used to find information technology keywords, and the author finds that there exists a decline in the
use of these terms. Lajili and Z´ eghal (2005) uses the same methodology to examine risk information disclosures in Canadian TSE 300 annual reports, their findings show a high degree of risk disclosure intensity, though the authors still suggest a more formalized procedure for risk disclosure. The study
“Content Analysis of Annual Reports for Corporate Strategy and Risk” display the inherent power in digitalised content analysis, while examining corporate strategy’s influence on risk and return in three industries (Bowman, 1984). Corporate responsibility was the main concern of a report written by Everaert et al. (2007), which also utilised content analysis of annual reports. The four aforementioned reports all provide examples of rigorous content analysis of annual reports, which are assumed to be closely enough related to interim reports for the examples to hold their validity. Other studies have used content analysis to examine other forms of media, Furrer et al. (2008) examines management studies, Soyoen and Jisu (2010) look at corporations online blogs, and Landrum and Ohsowski (2018) study corporate sustainability reports. In combination with previously mentioned studies these ex- pand the exemplary content analyses and will assist in the content analysis of this study.
A point of criticism that content analysis faces, especially at a word-by-word level, is the issue of
attempting to infer meaning from an inherently abstract set of data. The aforementioned effect is
further amplified by the inherent reductive properties of this form of content analysis and, hence, is a
central criticism that needs to be taken into account when analyzing the results of this study (Mukher-
jee and Bhattacharyya, 2013). However, this study does avoid another set of issues often associated
with modern content analysis, issues arising from the use of convenience sampling. As technology has
advanced and data is readily available in modern times, the automatic sampling of said data, within
expansive content analysis studies, is often based on some common feature. This feature could be
e.g. a common hashtag defining a set of tweets or a common reference defining a set of articles. The
primary issue with this method of data collection is that it can be nearly impossible do determine
whether the resulting set of data is representative of the research question. As this study rather uti-
lizes a purposive sampling (one based on some logic pertaining to the research question (Riffe et al.,
2019), i.e. a set of interim reports directly connected to a well defined set of companies), the study
avoids these issues. But instead, it runs into another issue, the matter of limited generalizability due
to a set of data that is not representative of a larger group (Lacy et al., 2015). At times some or all of these points of criticism are skimmed over when content analysis is used as a primary methodology of data collection (Everaert et al., 2007; Furrer et al., 2008; Pagliarussi et al., 2016), which is something this study attempts to keep in mind, if not avoid entirely.
4.3 Implementation and Statistical Model
In order to process the larger amount of data that was used in this study a more quantitative method- ology was used. Apart from being able to easily process larger amounts of data, this way of working can also be seen as advantageous as it removes the possible element of human bias (Basias and Pollalis, 2018). To determine the frequencies corresponding to the COVID-19 attention of companies Python (van Rossum, 1995) was used. The necessity to create a new program able to accomplish this arises as the underlying nuances of the interim reports are large, and as there is a substantial number of them to process, two factors many existing text readers are not able to deal with simultaneously at an adequate pace. The script written, seen in Listing 1, accomplishes this and was used in part to, for each company studied, count the total number of words in an arbitrary interim report, determine how many times a keyword was mentioned in said report, and sort the resulting values based on company and financial reporting period. This is the purpose of the first method, word count().
It is not uncommon to see the variable firm size being used in the field of empirical corporate fi-
nance, which is not surprising as it can be seen as a fundamental characteristic of a firm. This
practice is typical in part due to the “measurement effect” in “size effect” (Dang et al., 2017). In
order to account for this and see if the size of a company significantly affected the results the chosen
control variable was chosen as the number of employees. A considerable con of using the number of
employees as a control variable is that it does not take part time workers in consideration, which can
skew the results. However, as previous research has concluded that number of employees is one of
the preferred measures of firm size when studying corporate governance (Hashmi et al., 2020), and
that it is an easily accessible statistic, it was determined to be used as a control variable. The data
was gathered partially by utilizing web scraping, but needed to be complemented manually due to
a lacking data set. In the method control scrape() (Listing 1) the automatic part of this process can be seen, which gathers data from TradingView, chosen due to its easy access to a quite large set of data. The use of sites such as “TradingView” is not common practice in previous research on corporate governance, but as it was deemed to be a legitimate site with an easily accessible database, it was used to gather most of the data to the control variable. Manually complemented data was gathered from the site of respective company.
The method var gathering() (Listing 1) was written purely to simplify the data transfer from Python
to Matlab (MATLAB (R2020a), 2020), which was used to perform the statistical calculations and
plotting (see Listing 2). As is customary when performing correlational research the modeling was
done by first checking the data sets for heteroscedasticity (Basias and Pollalis, 2018), in order to de-
termine which statistical model was appropriate for the data sets. Whether or not heteroscedasticity
existed in either data set was determined using Engle’s ARCH test (Engle, 1982). Furthermore, the
variables need to be checked for multicollinearity, this was done using Belsleys collinearity diagnostics
and seeing what the resulting condition indices were. If they were greater than 10, extra care needed
to be taken when performing the regression (Belsley, 1991). As heterosckedascity was not present
and the multicollinearity of the variables was low (lower than seven for all variables), the method
used for the linear regression was linear least squares (Carroll, 1982). However, if heteroscedasticity
would have been present or the data sets had displayed considerable multicollinearity a more robust
regression model would have had to be used, in this case either general or robust least squares would
have been adequate. (Johnston and DiNardo, 1997)
5 Results and Analysis
The results take the form of plots of the collected data and accompanying regression, with tables of corresponding values and errors. This section will begin by presenting the single variate regres- sion analysis only considering board size or gender diversity, followed by an analysis considering the combination of the two variables. After these the control variable, firm size measured as the natural logarithm of the number of employees of each company, is introduced to the analysis. At first by considering its impact on board size and gender diversity separately, and concluding by considering its impact on the two variables combined.
The four data sets that were gathered, and some of their properties, are displayed in Table 1. Here we can see that the independent variable, attention, displays a mean value of 0.2741%, and a com- paratively low variance. The set describing board size does however show a large variance, where the minimum board size found was 2, and the maximum 17. Similarly, board gender diversity also shows a large variance (258.68%), with a mean of 32.740%. This indicates that neither of these data sets can be approximated using a normal distribution (or at least is very badly approximated for a 100 data points). The control variable does, however, display a mean of 8.3503 and a variance of 4.7845.
Table 1: Properties of the used data sets
Variable Mean Median Variance Range
Attention [%] 0.2741 0.2472 0.0260 (0, 0.63)
Size [#] 8.7800 8 13.305 (2, 17)
Gender Diversity [%] 32.740 33.333 258.68 (0, 66.67)
Employees [ln(#)] 8.3503 8.8467 4.7845 (2.71, 12.82)
5.1 Analysis omitting the control variable
The focus of Figure 2 is to display how the attention of top level management is related to board size. Here each blue dot represents a company listed on Nasdaq OMX Stockholm Large Cap and its recorded attention to the COVID-19 pandemic (ranging from 0 to 1) and board size (ranging from 2 to 17). A linear fit to this data is represented by a red line, whose specifics are described in Table 2. Figure 2 displays the lack of a significant correlation between board size and the attention to COVID-19 of top level management. There may seem to be a slight increase in attention with an increase in board size, but by studying the errors of the fit (in Table 2) one can see that there is a large residual sum of squares (RSS = 3.917) and a coefficient of determination close to zero (R
2≈ 0.01), implicating that this correlation does not exist, and indicating a future rejection of H1 .
2 5.8 9.5 13.3 17
Size (#) 0
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Attention
Fit of size
data fitted curve
Figure 2: The data where the attention to the COVID-19 pandemic of the top level managements of a company is plotted over board size (blue dots), and the resulting fit only considering board size (red line).
Figure 3 displays how the attention of top level management is related to board gender diversity. Here
each blue dot represents a company listed on Nasdaq OMX Stockholm Large Cap and its recorded
attention to the COVID-19 pandemic (ranging from 0 to 1) and board gender diversity (ranging from
0% to 67%). A linear fit to this data is represented by a red line, whose specifics are described in Table 2. Similarly to Figure 2, Figure 3 does not show the existence of a correlation between the attention to COVID-19 of top level management and board gender diversity. The specifics of the fit, seen in Table 2 further supports this notion of a lack of a correlation by displaying a R
2< 0.01 and a RSS = 3.9541, indicating both a non-existent correlation and a large deviance from the fit, suggesting a future rejection of H2 .
0 16.7 33.3 50 66.7
Gender Diversity (%) 0
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Attention
Fit of gender diversity
data fitted curve