Independent Directors and Corporate Financial Performance – A Hong Kong Perspective (A DBA Dissertation Completed in August 2005)
Chapter 3 – Research Methodology
3.1 Research Question
Revisited
Possible influences of the
structure of company boards of Hong Kong listed companies and their
performances have not been researched adequately. This thesis pursues remedying
that deficiency in the literature by exploring relationships
between a firm's financial performance and the composition of its board of
directors, with a particular focus on the independence of the directors.
3.2 Research
Methodology in General
Consistent with the principles
underpinning research (Baker, 2001), this study into the possible relationships
between independent boards and corporate performance is to extend and improve
knowledge and understanding – to provide a sustainable explanation for those
observed phenomena relating to Hong Kong listed companies’ board structures and
performances. Thus, this study pursues an in-depth understanding which utilises
methods entailing “an organised, systematic, data-based, critical, objective,
scientific inquiry … investigation into a specific [perceived] problem …”
(Cavana, Delahaye & Sekaran, 2001, p. 5).
According to Cavana,
Delahaye and Sekaran (2001), there are three main philosophical bases in
business research:
- Positivist research – the use of precise,
objective measures usually attributed to quantitative data, beginning with
a theoretical position and moving towards concrete empirical evidence,
through a process of deductive reasoning in applying a general theory to
specific situations.
Arguably, that is a
rather narrow view of the positivist research domain. No research is purely
positivist. It is to be noted, however, that what is described here to be an
original theoretical position is itself the product of normative reasoning.
“Ideas precede facts”, data are thus subordinate to ideas (Roszak, 1986). The
data used in this thesis become relevant to the research problem being pursued
only in the context of the ideas underpinning the assertions that the presence
of independent directors will enhance CG and corporate performance.
- Interpretivist research – presents a rich
and complex description of how people think, react and feel under certain
contextually specific situations. The researcher identifies what is
meaningful to each individual being investigated and becomes involved with
these individual subjects. Inductive reasoning, by observing specific
phenomena to develop general theory, is adopted.
- Critical research – a concentration on
uncovering myths and revealing hidden meanings. Research findings are
presented in such a way so as to lead to transformation. Both deductive
and inductive reasoning are used.
Others, however, argue that there
are indeed only two broad approaches – positivistic and interpretivistic
(Baker, 2001; Amaratunga, Baldry, Sarshar & Newton, 2002). The positivist
research, with its emphasis on precise, objective measures, can be termed the quantitative
approach, while the interpretivist and critical researches are generally
referred to as the qualitative, or phenomenological, approach.
By definition,
quantitative research deals with numbers, corresponding to positivism, with a
research design highly structured that can readily be replicated. The research
plan is rigorous using research hypotheses. The researcher is said to be
independent from what is being researched, hence the findings are claimed to be
relatively value-free and unbiased. Yet, the origin of the ideas underpinning
the research problem and the selection of the specific problem to be researched
are influenced by subjective and value-oriented stances, arguably inevitably
influenced by the predilections of the researcher. Quantitative research aims
to explore causal relationship using static data, and to work towards
generalisations. Deductive reasoning is adopted while validity and reliability
can be readily measured (Cavana, Delahaye & Sekaran, 2001).
By contrast, much qualitative
research does not focus on numbers. It corresponds to phenomenology. Generally,
the research plan is flexible in nature; the researcher interacts with what is
being researched; the findings are relatively value-laden and biased; and
inductive reasoning is adopted (Cavana, Delahaye & Sekaran, 2001). The
objective of qualitative research is to understand a particular phenomenon rather
than discover any causal relationship. Though, of course, the greater the
understanding of a phenomenon, the greater the likelihood that potential causal
relationships will be exposed.
However, as noted by Denscombe
(2003), and consistent with the noting of Roszak’s observations, the two
approaches are not mutually exclusive in practice, for few researchers rely on
one approach to the exclusion of the other, and indeed few research problems
can be explored completely using solely a quantitative or solely a qualitative
approach. The distinction between those approaches is frequently presented in
too simplistic a manner, as the assumptions associated with the two approaches
frequently overlap. Further, in a strict sense, the distinction between
“quantitative” and “qualitative” relates to the treatment of data and the
mechanisms by which the data or other information are analysed, rather than the
research methods as such.
Even if the distinction
is not clear cut, each of the research processes have to be undertaken in a
scientific manner for the results to be serviceable. A common misconception is
that scientific means measurable and quantifiable (Cavana, Delahaye &
Sekaran, 2001). The label scientific method is likewise a misnomer. It
falsely implies that true scientists pursue their research in only one
fashion. In more informed settings it is recognised that physical scientists
employ the entire range of techniques from following hunches, guesstimates, to
counting observations of actual phenomena (see Goldberg (1989) for example, for
the manner in which Aberdeen University scientists sought and discovered a
natural opiate in the brain). True science does not proceed in the absolutely
orderly manner characterising scientific papers. It does not motivate Noble prize
winner Medawar to declare that “Is the Scientific Paper a Fraud” (Medawar,
1991). Instead, scientific research refers to the need for an honest, truthful,
accurate and complete investigation (Cavana, Delahaye & Sekaran, 2001). The
term scientific research applies to both basic and applied research, thereby
denoting they can respectively be purely for generating a body of knowledge, or
be applied to solve a specific issue.
The characteristics of
good scientific research, according to Cavana, Delahaye and Sekaran (2001),
include:
- Purposiveness – the research needs a
definite aim or purpose.
- Rigour – the research needs a theoretical
base and a sound methodological design.
- Testability – research questions or
hypotheses developed can be tested with data collected.
- Replicability – the results should be
supported when the same type of research is repeated in other similar
circumstances.
- Accuracy – people can put faith or
confidence in the research findings.
- Objectivity – the conclusions drawn should
be based on the facts derived from the findings of actual data.
- Generalisability – the research findings
in one setting can be applied to other settings.
- Parsimony – using simple rather than
complex research frameworks.
Similarly,
Denscombe (2002) considers the following ten points as the foundations for
social research:
·
Purpose
·
Relevance
·
Resources
·
Originality
·
Accuracy
·
Accountability
·
Generalisations
·
Objectivity
·
Ethics
·
Proof
Though the
two frameworks differ slightly in their foci, both regard purposiveness,
accuracy, objectivity and generalisability as essential elements of good
research.
It must be noted that
there are limitations to applying scientific research in business. Unlike the
physical sciences, it is not always possible to conduct investigations that are
entirely value-free in the management and behavioural areas (Cavana, Delahaye
& Sekaran, 2001). The results in business research may not be exact and
error-free. In reality there are always difficulties, for instance, in
measurement and collection of data, or in obtaining representative samples.
Further, almost all data selected for investigation in business research are
subject not only to the range of variables used in the enquiry, but also to
myriad other influences, many of which are unknown to the researcher. The ceteris
paribus assumption always applies, though we might expect that in few
instances are “other things being equal”.
Basic research methods generally
include one or more of the following (Baker, 2002b):
- observation;
- survey;
- experiment; and
- secondary data study
Whereas observation is
usually the first physical step in the scientific method in identifying a
problem (Baker, 2001), contemplation is frequently the first scientific
process. It is to be noted that such identification draws upon a preconditioned
mind as to what is expected, normal or usual. Survey covers interview and
questionnaire, while experiment includes the laboratory activities of
undertaking controlled testing, and field experiment. Secondary data study
simply refers to studying data already gathered by others. No matter which
approach is adopted, it is advisable to first undertake qualitative research
before trying to quantify the direction and extent of any hypothesised
relationships (Baker, 2001).
The broad research approach then
leads to sampling selection and design which can be probability-based or
non-probability-based (Baker, 1999), the key being to have a representative
sample of the population that does not contain any systematic bias. Expectedly,
validity and reliability are two essential elements of good research. Validity
refers to the accuracy and specificity of the measuring process (Baker, 2002a),
and is analogous to whether a measure is measuring the concept which the
researchers think is being measured (Baker, 1999). It can be generally split
into internal validity and external validity: the former attesting to the
confidence that can be placed on the causal relationship found in research
design, and the latter to the extent of the generalisability of the results to
other field settings (Cavana, Delahaye & Sekaran, 2001). Reliability, on
the other hand, reflects the level of internal temporal consistency and
stability of the measuring instrument (Cavana, Delahaye & Sekaran, 2001),
the precision, sensitivity and consistency of measurements (Baker, 2002a), and
is thus akin to whether the measure will produce similar outcomes on different
occasions (Baker, 1999).
In quantitative
research, hypotheses are tested by carefully analysing the data, usually
through the application of statistical packages (Cavana, Delahaye &
Sekaran, 2001). Hypotheses in this context can be stated as a proposition about
the relationship between two or more variables expressed in the form of a
testable statement (Baker, 2001). In this study the primary hypotheses reflect
the perceived relationships between the presence of independent directors on
company boards and those companies’ financial performances with a view to
determining whether there are any grounds for presuming a causal relationship.
3.3 Research
Method and Sample Selection
As noted earlier,
quantitative and qualitative research are not mutually exclusive. They are
complementary, and preferably should be used collectively in research
(Amaratunga et al., 2002). Quantitative research, for example, is strong
in allowing large-scale data collection and analysis at a reasonable cost and
effort, but weak in ascertaining deeper underlying meanings and explanations
(Amaratunga et al., 2002).
In contrast,
qualitative research has complementary strengths and weaknesses. Creswell
(1994) argues that the concept of triangulation, combining methodologies in the
study of the same phenomenon, was based on the assumption that any bias
inherent in methodology would be cancelled out when used in conjunction with
other methodologies. This can be done by drawing together multiple types of
evidence gathered from different sources, using different methods of data
collection (Baker, 1999). Despite the obvious advantages from using a mixed
methods approach, the advocated use of a principally single methodology is
often based solely on pragmatic considerations, such as time and cost
constraints, and the need to limit the scope of the study (Amaratunga et al.,
2002).
These pragmatic
considerations also form the basis for selecting the research method for this
study. As a result, a single methodology is chosen.
While recognising the
normative characteristics of the ideas underpinning the call for increased CG
and the role that independent directors are presumed to play in that process,
the quantitative approach is considered more appropriate to the current study
in order to obtain an answer to the research question and test the hypotheses
in a reliable and orderly manner. The population consists of all companies
listed on both the Main Board and the GEM of the HKEx. With the purpose of
adding validity, the sample is the same as the population.
In broad terms, the
research process entails: first, collation of a complete list of all Hong Kong
companies listed on the HKEx, followed by the collection of secondary data in
relation to outside board composition and company financial performance from Yahoo
Finance, local newspapers, company annual reports, and information provided
by the HKEx. Subsequently, linear regression tests are to be run through SPSS
using outside board composition as the independent variable and financial
performance indicator as the dependent variable. With a confidence interval
pre-set at 95%, common statistical tools including R2
(coefficient of determination), p-value, r (correlation
coefficient) and Fisher’s r-to-z transformation, where appropriate, will
be employed to identify whether any significant relationship exists. In order
to add reliability, different definitions of outside board composition
and financial performance indicator (including accounting ratios and
market-based returns) will be employed. Possible distortions due to different
operationalizations used (Daily, Johnson & Dalton, 1999), will be minimised
as far as practicable.
3.3.1 Research
Design
The Apply Daily,
one of the widely circulated newspapers in Hong Kong, shows the year-end dates
of listed companies in its daily stock quotations. Over 60% of the companies
have their year-end on the 31 December. The only other widely used year-end
date, 31 March, is adopted by some 30% of the companies. Other year-end dates
make up the remaining 10% of the companies. Therefore, 31 December 2003, being
the most representative and also the most recent when research work was
started, is chosen as a reference date. An official list of all listed
companies as of that date was obtained from the HKEx. This list shows 861
companies listed on the Main Board and 187 on the GEM.
It was explained in
Chapter 2 that in addition to the purely Hong Kong companies, there are
two groups of mainland-Chinese-related companies listed
on the HKEx: the “H-Share”, or Chinese enterprises (CEs), and the “Red Chips”,
or China-affiliated companies (CAs). A common characteristic discriminating the
CEs and the CAs from other Hong Kong listed companies is that they are
predominantly controlled by mainland Chinese interests. The main difference per
se between the CEs and the CAs is that the CEs are PRC-registered whereas
the CAs are not. Of the 2003 official list, there were 64 CEs and 73 CAs on the
Main Board, as well as 27 CEs and 1 CA on the GEM.
Board composition information
will be drawn from company annual reports. That information includes the number
of executive directors, NEDs and INEDs. As elaborated in Chapter 2, under the
Listing Rules of the HKEx, NEDs are members of the board who do not hold any
office in the company and have no management responsibility. Those NEDs who,
also, do not have any interest in the company are designated INEDs.
Financial data will be extracted
from Yahoo Finance – book value of equity as well as accounting
performance indicators: ROE and ROA. In relation to accounting ratios, ROE is commonly used to indicate a company’s performance as it is
generally considered to provide information on how effectively managers are
employing funds invested by the shareholders to generate returns. ROA, on the other hand,
essentially shows how much profit a company is earning “on the assets” used in
its business. Of course, both ROE and ROA are subject to the vicissitudes of
asset valuation. With that caveat in mind, ROA eliminates the distortions of
varying financing structures and displays a usable measure of the return in
relation to a company’s size.
For companies with
year-end dates other than 31 December 2003, the latest available accounting
ratios will be those as extracted on 30 October 2004, the cut-off day for data
extraction. The year-end dates generally range from as early as 31 July 2003,
to as current as 30 June 2004. In order to prevent the usage of out-dated data,
those companies with their latest published financial data earlier than 31 July
2003, signifying these companies are way behind schedule in disclosing
operating results, will be discarded from further analysis.
In relation to market
data, 24 September 2004 has been chosen as a reference date. While it allows
close to nine months to elapse for the majority of companies to reflect on the
impact of independent directors on company financial performance, it falls
ahead of the 31 October 2004 deadline imposed by the HKEx for listed companies
to increase their number of INEDs from two to three. In addition, it is not a
month-end date that might attract possible manipulation because of month-end or
quarter-end reporting. Two indicators, Price-to-Earning ratio (P/E) and Market
Value of Equity to Book Value of Equity (MV/BV), will be employed as measures
of market performance. Conventionally, both ratios are taken to be indicative
of market perceptions of company value on a unitary basis: the former compares
with historical earnings while the latter matches up with historical equity
value. P/Es and MVs are drawn from the Hong Kong Economic Times, a
premier financial newspaper in Hong Kong, while BVs are drawn from Yahoo
Finance.
In the case of
obviously questionable data, such as ROA exceptionally high or unreasonably
low, references will be made to the annual reports of the companies concerned
to clear any possible irregularities.
3.4 Hypotheses
and Sampling Design
3.4.1 Hypothesis
1
HR1a suggests that there exists a relationship between the percentage of
independent directors in the board of directors and corporate financial
performance.
All companies on the
Main Board and the GEM will be used in running a series of simple regression
analyses. The Percentage of NEDs, an independent variable, will be used
to predict variation in ROE, the dependent variable, to see if any
significant relationship exists.
To add reliability,
other performance indicators will also be used. Three more replicate tests will
substitute ROA, P/E, MV/BV, respectively, for ROE.
In order to check
whether INEDs have greater impact on company performance than NEDs, another set
of regression tests of the four performance measures will be repeated using Percentage
of INEDs instead of Percentage of NEDs. The two sets of tests will
then be compared to reveal whether HR1a holds.
HR1b states that there exists a relationship between the
number of independent directors in the board of directors and a firm's
financial performance.
As in
assessing HR1a, two sets of regression tests, using the four
performance measures, will be undertaken. The only differences will entail
moving from relative to absolute measures, replacing Percentage of NEDs
with Number of NEDs in the first set, and Percentage of INEDs
with Number of INEDs in the second.
Companies on
the Main Board vary widely in size. Most are small to medium by global
standards. In terms of market capitalisation, excluding CEs and CAs, about 10
companies in the sample might be considered large – market capitalisation
exceeding HK$100,000 million, while more than 200 are extremely small companies
– less than HK$200 million. To isolate the effect of company size, for the
purpose of this study, companies of comparable market capitalisation are
classified into large (over HK$10,000 million), medium (between HK$10,000
million and HK$1,000 million) and small (less than HK$1,000 million). The
aforementioned regression tests will be repeated for each of the three size
groups.
3.4.2 Hypothesis
2
HR2 presumes that the relationship between board composition and
corporate financial performance is stronger in growth-oriented companies than
in non-growth-oriented companies.
Data relating to all
the companies on the GEM and the Main Board, excluding CEs and CAs, will be
used in the testing of Hypothesis 2. Separate linear regression tests
will be undertaken for the GEM companies and Main Board companies. Similar to
the manner of testing Hypothesis 1, the two groups of variables used will
be – Percentage of NEDs, Percentage of INEDs, Number of NEDs,
Number of INEDs; and ROE, ROA, P/E, MV/BV.
By applying Fisher’s r-to-z transformation, the correlation coefficients
of the two independent groups of companies will be transformed into z-scores
and compared to assess whether the difference is statistically significant.
In addition, companies with high
MV/BV can also be interpreted as being perceived by the market to have high
growth potential. As a sensitivity analysis, companies from the Main Board
group with high MV/BVs will be extracted for separate regression tests against
those Main Board companies with low MV/BVs, to evaluate whether the results are
consistent with those of the GEM group against the Main Board group.
3.4.3 Hypothesis
3
HR3 states that the relationship between board composition and
corporate financial performance is stronger in companies majority-owned by
non-mainland Chinese than in companies majority-owned by mainland Chinese.
As there are insufficient CEs and CAs on the
GEM to draw any representative sample, those on the GEM can be grouped
effectively with those sourced from the Main Board in the testing of Hypothesis
3 – with separate regression
tests for the CE group and CA group, using the same two groups of variables – Percentage
of NEDs, Percentage of INEDs, Number of NEDs, Number of
INEDs; and ROE, ROA, P/E, MV/BV. The results
for CEs and CAs will then be compared with those for non-CE and non-CA
companies to isolate any noticeable differences in correlation.
3.5 Key
Assumptions in Research Design
The research design
focuses solely on board independence, but ignores the human behaviour and the
social context in which it is manifested. It was noted earlier that social
relationships, friendships, for example, and other forms of possible conflict-settings which are not readily
detectible, though might compromise independence, will not be taken into
consideration (see Section 1.6).
It is also to be noted
that this study is based on data relating to companies listed in Hong Kong
under its current regulatory framework. Although similar results can be
predicted for companies with similar characteristics and regulated under
similar framework, they might not be generalised to drastically different
companies, for example, large American companies under much tighter regulatory
control, or small companies in developing countries with poor regulatory
control. Further, the study is focused on shorter-term financial performance.
Longitudinal studies might reveal further information on the effect of board
composition in the long run.
No comments:
Post a Comment