Learning outcomes

• Describe in detail the various

methods by which human beings
acquire knowledge, including the
UNIT 1-
scientific method.

• Compare the scientific method with

CHAPTER 1-
other means of acquiring knowledge.

• Describe and contrast inductive or

THE
DIFFERENT
probabilistic reasoning, and deductive
reasoning.

• Discuss reductionism in social

science research.

• Compare and contrast social and

METHODS OF
natural sciences.

• Compare and contrast quantitative,

ACQUIRING
KNOWLEDGE
qualitative and mixed-methods
research.

• Analyse the relationship between

fact, theory, and observation.

• Describe the process of scientific Non-scientific methods of

research, in quantitative, qualitative
and mixed-methods research.
acquiring knowledge Pg 1-3
• Understand why ethics in research is
important. For children, the first source of knowledge is most often their

• Understand the underlying principles parents or caregivers. This method of authority is not only
of ethical research. used by children but very commonly by each of us when we rely
on the knowledge and 'wisdom' of prominent and significant people
• Understand particular ethical who are recognised as having a better grasp of their environment
guidelines.
than ordinary people. Thus, the statements of these 'qualified'
• Apply the appropriate ethical people are rarely questioned or challenged. On the contrary, the
guidelines to the process of research. knowledge imparted by them is usually accepted as absolute and a
certain amount of trust is placed in these authorities as sources of
knowledge.

A variation of this 'method of authority' is the mystical method, where the correctness of the
knowledge is assumed to reside in a supernatural source. In this case the 'knowledge producers' are
regarded as authorities due to their ability to transmit the truth or knowledge imparted to them by
supernatural forces.

The intuitive method is another way of making sense of the world. People sometimes make
judgements about the world based on what 'feels' right for them. They may not always be able to explain
their feelings in a way that makes sense to other people, but they have great faith in these feelings to help
guide them in their lives.

A variation of the intuitive method is what is referred to as conventional wisdom. Conventional

wisdom reflects so-called common-sense understandings of the world that are commonly accepted as
being true. However, the 'common wisdom' statements, being very general, without details on when they
are applicable, are often contradictory.

The rationalistic method is based on human reason. According to this approach, human beings have
the ability to think logically (or to reason), and thus to discover laws through purely intellectual processes.
The basis of knowledge is correct reasoning which enables one to know what must be true by principle.

The opposite of the rationalistic method is the empirical method, where facts observed in nature are
the foundation of knowledge. Objectivity of observation is emphasised and only what is observable, what
can be perceived by our senses, constitutes knowledge.

The scientific method and its properties Pg 4-7

The distinction between quantitative and qualitative research methods is mainly to address different
research aims. For example, the study of the efficiency of a certain training programme is quantitative
while the investigation of the attitudes and problems encountered by the participants of that programme is
qualitative.

In the case of quantitative research, the first step to knowing is a description of the object,
relationship, or situation. The object of the study must be accurately depicted. Here, evidently, the
empirical method of objective observation must be used. Thereafter, an explanation or statement of the
relationship between the described facts should be expressed, where possible in the form of a law. The
explanation is thus the result of a reasoning process using the rationalistic method and leads to the
formulation of a natural or social law ('law' is not used here in a legal sense). The stated explanation should
permit a prediction of future events under well-defined conditions. In other words, the explanation should
allow one to foretell the occurrence of some event. But to ensure that this explanation or law will enable
prediction, the correctness of the explanation must be tested. This is achieved by confronting it with reality
as perceived by the five human senses, using the empirical approach. Correct explanations leading to
ability to predict events should yield intelligent intervention, which enables changes occur that improve a
situation.

In the case of qualitative research, most frequently the researcher is concentrating on the observation
and recording of the events under study. Since very little is known about the situation under observation,
no a priori explanations are given and thus no expectations or predictions can be tested. However, it may
be expected that some explanations will be found at the end, on the basis of the data collection and
analysis.

SCIENCE ASSUMES THE FOLLOWING:

 The existence of natural and social laws. Science presumes order and regularity in natural and
social events. Without this assumption some of the main aims of scientific research, such as
explanation, prediction and the possibility to act, would not be attainable. These laws are assumed
to exist independently of the observer, and they describe the way phenomena interact or social
events occur.

 Laws can be discovered by human beings. Scientific research assumes that, although human
beings are part of nature and are themselves subject to its laws, they can discover those laws. We
obey biological laws determining the growth of our body, psychological laws determining the
development of personality, sociological laws in the organisation of society, as well as economic
and political laws. Again, it should be emphasised that these laws are not laws in a legal sense.
Instead they are the result of observations of certain regularities, which indicate that natural events
proceed according to our expectations based on our understanding of those events. For example, if
an infant did not grow but remained the same size for several months, this would be a violation of
our expectations and we would rightly conclude that some abnormality is present.

 Natural phenomena have natural causes. No supernatural powers are necessary to

grasp the cause of events and no unexplainable supernatural forces are needed to explain the way
nature functions.

 New knowledge is accumulated gradually and sequentially. Numerous

examples show how invention of new instruments of investigation and new approaches to
studying a problem can led to new advances in science. The new understandings yielded by these
advances often demonstrate the limitations of previous knowledge.

 Knowledge and truth are founded on evidence. There are specific and commonly
accepted rules that need to be adhered to for something to constitute evidence. Much of the time
observations that are made on the basis of the senses constitute evidence. Sophisticated
instruments have been developed to observe and measure social phenomena when observations
are in danger of being inaccurate or insufficient. In general, whenever a scientific claim is made, it
is necessary that the researcher provide evidence to support that claim. In the absence of evidence
(or proof that the claim is tenable), the claim remains at the level of conjecture or hypothesis.

 Scientific statements must be distinguished from common-sense

statements. Common sense is often characterised by contradictory statements. 'Absence makes
the heart grow fonder' and 'Out of sight, out of mind' are opposing statements of common wisdom.
Clearly, both are possible in different situations, and neither is appropriate to all situations.
Common-sense statements are the result of non-scientific observations in that they do not take into
consideration the different variables at stake. The weakness of the approach is that the
 preconditions for the validity of the statements are not specified due to lack of systematic
investigation.

 Scientific observation is objective. Objectivity here means that the result of the
observation is independent of a single observer. In other words, it corresponds to the description
made by anyone examining the same phenomenon. The more accurate a description is, stating
measurable properties, for instance, the greater the objectivity of the observation.

 Scientific observation is systematic. All possibilities are considered one at a time, in a

logical order. Natural sciences allow the best illustration of systematic observation.

Unfortunately, it is not always possible to observe every phenomenon or all of its aspects systematically.
Most often some characteristics remain unknown so that the explanation and, as a result, the prediction
may not be accurate. In fact, one can only predict that an event will take place if one knows all the
conditions and circumstances, which cause the event. In the social sciences in particular, one is rarely in
possession of all the information leading to the occurrence of a certain phenomenon and so can seldom
explain this phenomenon with certainty. In these cases, one has to content oneself with a probability
statement, asserting that if some given conditions are satisfied, the event will occur more often than if
some of these conditions are not met. Explanations that concede some uncertainties are called probabilistic
explanations.

Properties of scientific research Pg 8-9

 Scientific research is empirical since the aim is to know and understand reality. Each step is based
on observation, be it when collecting the basic facts or when testing an explanation or assessing
the value of a prediction or the result of an intervention. The assumption of scientific research is
that reality exists outside of the observer. Scientists attempt to understand the world beyond their
personal biases and assumptions about their particular research subject. Quantitative methods use
a variety of tools to help scientists be more detached and dispassionate about what they research.
Qualitative methods focus on acknowledging, understanding, and allowing for the biases in a
transparent manner. Whether or not this is attainable is often debated by social scientists.

 Scientific research is systematic and logical. Not only must the observation be done
systematically, but a certain logical order must also be followed. Logical predictions cannot be
made before a description has been given and an explanation of the observed phenomenon found.
One cannot prepare a questionnaire to gather all data necessary for a study, before having a clear
idea of the type of information required. An analysis of the different variables involved must be
undertaken prior to the formulation of the questions to be answered by the respondents.
  Scientific research is replicable and transmittable. Since the observation is objective and the
explanation logical, anyone placed in exactly the same circumstances can observe the same event
and, by reasoning, arrive at the same explanation and prediction. Moreover, it is possible to
communicate each step of the research and to transmit the acquired knowledge. The usual way of
doing this is to present the research and its results in a report or an article published in a journal. In
this sense, science is a very transparent process that invites criticism and scepticism. When
scientific ideas are able to withstand such challenges, they remain tenable. When evidence is
amassed that demonstrates a scientific idea to be false, then that idea must be abandoned.

 Scientific research is reductive. By grasping the main relationships between laws, the complexity
of reality is reduced. All details which are not essential, or which have little influence on the
process under investigation are omitted. In social sciences, by controlling for many variables, one
nearly always reduces the complexity of reality. One presupposes that the variables that have not
been taken into consideration have little effect on the issue in which one is interested.

Reductionism, or the method of considering only the essential and necessary properties, variables or
aspects of a problem, has to be used with great care as it can lead to significant bias in the research process.
If the selection of variables to be excluded is based on false assumptions, some important variables may
not be taken into consideration.

 5. A scientific claim, statement or theory must be falsifiable. This means that a scientific claim
must be stated in such a way that can be demonstrated to be false. For this to be the case the claim
must be testable, that is, it must withstand empirical scrutiny.
 CHAPTER 2- THE
SCIENTIFIC METHOD
APPLIED TO SOCIAL
REALITY
The distinction between natural and social sciences Pg 12
Research usually arises from some need, so that it has a particular purpose to fulfil. Natural sciences
investigate the properties and laws of natural phenomena. The development of natural sciences is at times
determined by the needs of production. commerce and industry.

On the other hand, social sciences investigate the properties and laws of social phenomena. The
development of social sciences has its roots in the need for understanding, management and manipulation
of social affairs.

The development of both types of science depends on the needs, values, aspirations and other
characteristics of the particular society in which it takes place. Thus, many countries in Africa show a
greater concern for developing techniques to reduce the spread of malaria than they do for research into
heart transplants. Social sciences, however, are much more affected by some aspects of the society than are
the natural sciences.
 Relationship between facts and theory Pg 13-15
A fact is an indisputable set of statements about reality.

A theory may be defined as a set of ideas or statements that explain a particular social phenomenon.

The scientific enterprise assumes that our fundamental knowledge is perceptual, that is, information
about the environment perceived through our senses. Only external, superficial relations between
information acquired by our senses can be known. Perception by itself is only one means of acquiring
knowledge. On its own, however, perception falls short of true knowledge. One may know that green fruits
are not ripe and should not be eaten, but what about watermelons? Rational knowledge, based on
judgment and logical thought, provides explanations or reasons for the relationships between observations.

Framework:
 Facts give rise to theory since they raise a need for explanations of the observed
phenomena. Observation of reality can lead to systematic research and the formulation of a
general explanation or theory. Facts, the cornerstone of knowledge, are not the product of random
observation, but should be both selective and meaningful.

 Theory serves as a basis for the gathering of facts since it specifies the facts to be
systematically observed. If a theory stipulates that the failure of cooperative unions is related to
the absence of particular managerial skills among the members, the observations will focus on the
existence or absence of these skills, rather than on the size of the union, the age and sex of its
members, or the climatic conditions under which the members live. Thus, observations are not
arbitrary or haphazard. They are guided by the underlying theory or concerns held by the
researcher.

 Facts allow a researcher to find support for, improve, disprove, or formulate a theory.
Facts lead to a redefinition or clarification of a theory. A scientific explanation or theory must be
subjected to empirical testing to determine whether it corresponds to observed reality. This means
in practice that a theory allows a researcher to predict some facts. If the predicted facts are not
observed during experiment or investigation, the theory is disproved and will need to be improved
or reformulated. Equally, if some observed phenomena do not coincide exactly with the
theoretical predictions, the theory must be reformulated.

 The process of theory reformulation is extremely important to the

advancement of science, and therefore of new knowledge. If theories were not
advanced, deeper understandings of social phenomena would not be achieved and knowledge
 would become stagnant. For the frontiers of knowledge to be pushed, theories need to be
continually refined and improved.

 Theories are typically not proved. When data corroborate a theory, this is taken as
evidence that the theory is tenable, that is, it remains a credible and viable explanation of the
observed data. However, there is always the possibility that, at some point in the future, some
observations or data might contradict the theory. Thus, social scientists usually accept that an
existing theory is tentatively the best explanation of the phenomena under study for the moment,
but that it may yet be falsified by further data. For this reason, it is uncommon to speak of theories
being proved. The notion of proof conveys a sense that a theory is incontrovertible, which is never
the case.

 Theory allows for a classification and conceptualisation of facts. It

summarises and provides explanations of how relevant facts relate to one another.

 Theory predicts facts.

Quantitative, qualitative and the mixed method

approaches Pg 15-16

Quantitative methods emerged from the philosophical belief that the world runs according to natural laws
and that the role of the scientist is to uncover or discover these pre-existing laws. Thus, truth is assumed to
be absolute and independent of the human beings that search for it. In comparison, qualitative methods
emerged out of more recent philosophical beliefs that truth is relative, and that knowledge is constructed
by human beings. In other words, our understanding of the world is a product of our personal assumptions,
biases, and prejudices.

The quantitative approach relies extensively on numbers and statistics in the analysis and
interpretation of findings that are generalised from the sample to the population. On the other hand, the
qualitative approach is often used when the problem has not been investigated before. Generally,
this method uses smaller samples, from which findings a better understanding of the phenomenon being
investigated is produced. Sometimes, circumstances or the problem under investigation demand that both
approaches be used in the same study in order to confirm or elaborate each other. This is called the
mixed-methods approach. For example, a mixed-methods study might combine:
 Structured questionnaires (quantitative)
 Open ended interviews (qualitative)
 Surveys (quantitative)
  Case studies (qualitative)
 Questionnaires (quantitative)
 Observations (qualitative).

In the quantitative approach, the researcher tends to follow what natural scientists do: that is collect
data (that is measurements or frequencies) according to a very specific set of steps, and in so doing attempt
to remain as objective and neutral as possible. Quantitative data is analysed using statistical procedures so
that findings can be generalised from a relatively small sample to the entire population. Alternately, other
researchers are more inclined to qualitative approaches in which the plan of the research is more
flexible and circular. The researcher investigates a problem from the respondents' point of view. The focus
of such a study is to determine what respondents think and feel about a particular phenomenon or issue. In
other words, reality is interpreted from the respondent's frame of reference.

Developing theory from facts Pg 18

This relationship is expressed through a succession of steps. Facts raise questions. These questions can in
turn be condensed into a problem that is then given a temporary solution, yet to be tested, which is called
the hypnosis.

Analogy is a correspondence between a phenomenon or event that has been studied already and another
phenomenon or event that resembles the first but has not yet been studied. Analogy permits one to draw
conclusions based on the similarities between objects and certain of their properties.

In model-building, one object or phenomenon, the well-known one, serves as the model. Here certain
properties of the object have been singled out, represented in their pure, simplified form and then studied
in the absence of the actual object. Ideal models are formed with the help of particular symbols.

The relationship between facts, problems, hypotheses,

models, and theory Pg 19
Analogy and model-building are therefore quite similar ways of discovering some properties of an object
or phenomenon by utilising the existing knowledge of another object or phenomenon. The advantage of
model-building is its reductive property, that is, only the main characteristics are considered. The
characteristics that do not affect the process under study are ignored and the properties of the model can
then be studied in the absence of the original object of study. A geographer can study the properties of a
particular region through the use of adequate maps without actually travelling to that region.

Of course, the utility or danger of analogy and model-building depends on adequacy of the analogy or
model. The similarity of two chosen objects or events might be too superficial. The model might be an
over-simplification of reality. Both flaws will lead to fallacious results.
 Steps in scientific research Pg 20-22
The quantitative research process generally involves the following steps:

1. Selection and formulation of the research problem- During this step the
researcher identifies an area of study and specifies one or more clear research questions.

2. Reviewing literature- The researcher will then acquire knowledge of current theory and
research in the field through the process of reviewing the existing literature on the subject. The
necessary variables will be identified and both conceptual and operational definitions will be
developed. Finally, the researcher will formulate testable hypotheses in relation to the stated
research questions.

3. Developing a research method- Having clarified the research question, variables and
hypotheses, the researcher will develop a research method. This entails choosing a particular
research design and sampling method.

4. Developing a data collection instrument- On the basis of the chosen research method,
an instrument is developed, and its appropriateness and quality (validity and reliability) are tested
through a pilot study, if possible.

5. Sampling- Using the identified sampling method, the actual sample is drawn from the general
population under study and a detailed description thereof is developed.

6. Data collection- The research instrument is then administered to the sample and data is
gathered or collected.

7. Data Analysis- Once data is collected it must be organised and checked for accuracy and
completeness. When this process has been completed the researcher will use a range of arithmetic
and statistical tests to describe the sample data and generalise from this data set to the population
from which the sample was drawn.

8. Interpretation of results- At this point the researcher considers what the results of the
statistical analysis can teach us about the research questions specified in step one. To what extent
has the research answered the questions? To what extent can we trust those answers?
 9. Conclusions and recommendations- Based on the interpretation of the results, the
researcher draws certain conclusions and makes considered recommendations. This may have to
do with social policy or action, or the recommendations may point to further research.

10. Dissemination of results- Finally, the results are presented in different forms such as
research reports, seminars and conference papers, books and, of course, peer-reviewed journal
articles.

The qualitative research process is more difficult to describe since the steps are generally less
linear. In fact, qualitative research often progresses in a circular fashion. Some stages may precede and
follow others on the basis of pragmatism.

1. Selection and formulation of the research problem- Qualitative research begins

with a research question in the same way as quantitative research does. However, the question is
typically more descriptive in nature, and while the different components of the question must be
defined at a conceptual level, operational definitions and hypotheses are seldom specified.

2. Reviewing literature- The purpose of reviewing literature is to learn first-hand what has
been studied on the specific question and thereby increase the researcher's understanding of the
concept under investigation so that he or she will ask more relevant questions. Sometimes,
literature study takes place at the same time as data collection. In fact, unlike in the quantitative
research, a comprehensive review of available literature may be deliberately postponed until the
data collection stage.

3. Developing a research method- At this point, the researcher determines the most
appropriate way of gathering data that will shed light on the research questions.

4. Developing a data collection instrument- In qualitative research, the most commonly

used data collection tools are interviews and focus group discussions. These are used in the
collection of spoken and written words. If the researcher wishes to collect data by means of an
interview, then he or she first needs to design an interview guide.

5. Sampling- Since it may not be possible to study the whole population, the investigator may
need to sample the population. In qualitative research, the main aim is to understand the
phenomenon under investigation very deeply and therefore representative samples are not a
 requirement as they are in quantitative studies. The researcher needs to find a sample of
participants that carries most of the characteristics associated with the phenomenon under
investigation.

6. Data collection and analysis- In qualitative research, data collection and analysis often
take place at the same time or as alternating processes in a cycle. Qualitative data includes words
(written and spoken), artifacts, pictures and video. In the case of interviews, data collection may
include the use of audio tape recorders and video cameras. Data collection and analysis may in
turn lead to more literature study, making the process ever more flexible and cyclic.

7. Interpretation of results- During this phase, the researcher studies the results of the data
analysis in order to gain some overview. To what extent have the research questions been
addressed and answered? To what extent can these answers be trusted?

8. Conclusions and recommendations- As is the case in quantitative research, based on

the interpretation of the results, the researcher draws certain conclusions and makes considered
recommendations. These may have to do with social policy or action, or the recommendations
may relate to further research.

9. Dissemination of results- Ultimately, as for quantitative research, the findings of the

research must be shared with others. This may take the form of a report, a seminar or conference
presentation, a book or a peer-reviewed journal article.

Finally, mixed-methods research is likely to follow either the quantitative or qualitative research process
outlined above, depending upon which kind of data collection is dominant in the research design. Some
mixed-methods research has quantitative data at its core and this data is then elaborated or explained with
qualitative data. In this case a quantitative research process is likely to be followed. Other mixed methods
designs have qualitative methods at the core and thus follow a more qualitative research process. Of
importance to mixed-methods designs is the point at which quantitative and qualitative data are mixed.

CHAPTER 3- RESEARCH
ETHICS
 Researchers' social responsibility in developing
countries Pg 27-28
Scientific research has the ability to generate accurate information which is an important resource. Think
how much politicians spend on social research to determine how the population is feeling about key social
issues. Thus, social research is a source of power which can be used for social benefit or abused for
personal gain. For almost every new research finding, there are people who stand to gain from it, and there
are people who might lose because of it.

Four issues are relevant:

1. Highest quality practice. Internationally, the quality of social research is extremely

variable. Unfortunately, poorly performed social research devalues all social research. It is
therefore important that social researchers are properly trained, supported and supervised while
they are still inexperienced. It is very important that researchers critically review their own values
and biases. The requirement that methodologies, results and conclusions are recorded in detail and
made available to peer review is most important. Only if one knows exactly how a piece of
research was conducted can the results be evaluated.

2. Building capacity of all sectors of the community. In some cases, social research
methodology is seen as being beyond the reach of the majority of people. Although it is true that
some sophisticated techniques require a degree of specialised knowledge, virtually all problems
can be usefully tackled using the methods explained in this book. It is important that a wider range
of people learn how to conduct useful research so that they can benefit directly from these
techniques and judge more accurately the value of research in which they have an interest.

3. Relevance. The question of the relevance of research is a controversial and political one in
many circles. However, since the resources needed to support social research are somewhat
scarce, it is safe to say that researchers should aim to make a useful contribution to society. While
it is true that people disagree on what is useful and what is not, all social researchers should ask
themselves what contribution their work is making to society.

4. Promulgation of results. Too often important research findings are lost because of the
manner in which findings are presented. It is part of the researcher's responsibility to make the
results available in a form that is usable by the people who can benefit from them. In some cases
publication in a journal is sufficient to achieve this. In other cases, the people who are most likely
to benefit from the study are unlikely ever to read a scientific journal. In these cases it is required
that researchers find other ways of making their findings known. If promulgation of results is not
handled properly, then the research is unlikely to make a difference to society and the previous
point about relevance becomes meaningless. Many Africans are greatly angered by researchers
from the developed world who study our problems and our people (often in pursuit of degrees and
academic recognition), but then never make their results and conclusions available to the people
whom they have studied in a way that can make a difference to their lives.
 What are research ethics? Pg 28-29
The study of research ethics helps to prevent research abuses and assists investigators in understanding
their responsibilities as ethical scholars. Research ethics places an emphasis on the humane and sensitive
treatment of research participants, who may be placed at varying degrees of risk by research procedures. It
is always the researcher's responsibility to ensure that his or her research is ethically conducted. In fact,
before a single participant is contacted, the researcher must ensure that the research plan can pass an
ethical evaluation. To this extent, ethical standards attempt to strike a balance between supporting freedom
of scientific inquiry on the one hand and protecting the welfare of participants on the other. In other words,
researchers have the right to search for truth and knowledge, but not at the expense of the rights of other
individuals in society. Similarly, participants have basic rights when they elect to participate in a research
study, chiefly rights to privacy and to protection from physical and psychological harm. The goal of
research ethics is to minimise the risk to participants. Many academic disciplines, including disciplines in
the social sciences, have professional bodies that have published guidelines to help researchers think
through ethical issues as they pertain to their research. Most published ethics guidelines have identified
common principles of research ethics.

Principles of ethical research Pg 29-31

Non-maleficence
The most basic principle of research is that participants must not be harmed by participating in the research
project. It is important to note that harm may occur intentionally or unintentionally during the course of a
research study, and thus the researcher must be aware of the various possible adverse events that are likely
to occur throughout the duration of a project and beyond. Thus, research should never injure or harm
participants. One consequence of this principle is that the use of placebos might be unethical since, by
withholding effective treatment, a researcher might worsen the health of a research participant. If a
particular research procedure produces unpleasant effects for participants, the researcher should have the
firmest scientific grounds for conducting it.

Beneficence
It is important that research not only does no harm, but also potentially contributes to the well-being of
others. At times this might place a researcher in a difficult position. What is beneficial to one group may
not be so to another. Consider the case where, despite government efforts to promote new crops,
commercial farmers have been slow to plant the new crops. A researcher is called in to find ways of
ensuring that the farmers plant the crops required by the government. Whose interests are being served in
this case? How might the research be adapted to serve the farmers' interests? Part of the principle of
beneficence is the question of competence. People who are professionally trained by accredited and
acknowledged institutions are much more likely to be competent at what they do. Thus, the qualifications
of researchers are of considerable importance. The principle of beneficence requires social and behavioural
researchers to conduct research that is effective and significant in promoting the welfare of people.
 Autonomy
The principle of autonomy incorporates the freedom of individuals' actions and choices to decide whether
or not to participate in research. No person should be forced, either overtly or covertly, to participate in
research. At the core of the principle of autonomy is the right to participate voluntarily in social research or
decline to participate. In the past, prisoner populations have been used for medical and psychological
research. Very commonly, these prisoners were unable to refuse participation. To this end, the principle of
informed consent is of paramount importance. Participants need to be informed about the nature of the
research project in which they are being asked to participate, so that they can make an informed decision
about their participation. Thus, participants must be given clear and sufficient information on which to
base their decision. Informed consent requires that all participants be fully informed of the nature of the
research, as well as the risks, benefits, expected outcomes and alternatives before they agree to participate.
The principle of autonomy can, however, affect the generalisability of results. People who agree to
participate in research may differ from the many others who do not volunteer. This is known as the
volunteer effect and is discussed in greater detail in Chapter 11 that deals with issues of sampling.

Justice
The principle of justice is based on the belief that all people should be treated equally. Thus, people should
not be discriminated against in research on the basis of race, gender, disability, income level or any other
characteristic.

Fidelity
The principle of fidelity implies faithfulness and keeping promises or agreements, specifically between the
researcher and the participant. Thus, engaging in deception or breaching confidentiality is an ethical
violation that infringes on a participant's rights.

Respect for participants' rights and dignity

As human beings, all participants have legal and human rights. No research project should in any way
violate these rights when participants are recruited. It is necessary therefore to ensure that the dignity and
self-respect of participants is always preserved. An important part of protecting people's dignity is
understanding and respecting their culture. For example, in many cultures it is considered highly
inappropriate to discuss certain topics with anyone other than one's closest confidants. A researcher
concerned with women's health and contraception must therefore find ways to gather information about
this important topic without offending the research participants.

Sometimes well-meaning researchers fail to respect clients' rights. Consider for example the case of a non-
governmental organisation that offers counselling and legal assistance to rape survivors. To measure the
impact of their work, they contract a researcher who telephones a random sample of past clients in order to
find out how beneficial those clients consider the service to have been. Have the clients' rights to privacy
and confidentiality been compromised? How might the study have been designed to better protect the
clients' rights?
 Important ethical guidelines Pg 31-36
Often, the ethics of conducting a research project are not very clear-cut, but these overarching principles
can serve to guide ethical decision-making by researchers. In general, when the ethical issues are not clear,
it is important to consult. In addition to the overarching ethical principles described above, the following
are important ethical requirements to which researchers must adhere.

Ethical review
Before a researcher proceeds with a particular study, he or she should submit a detailed proposal of the
intended research to a process of peer review. This proposal should anticipate what ethical concerns might
be raised during the implementation of the project and should explain how these concerns will be handled.
Ethical review is handled differently in different countries, and unfortunately in some African countries the
mechanisms of ethical review are poorly developed. It is essential that researchers abide by the laws of
their countries, the ethical codes of their professions and the policies of their institutions. Where formal
mechanisms of ethical review are not adequately developed, it is the responsibility of the researcher to find
suitably knowledgeable and experienced peers to conduct a formal ethical review of the proposed research.

Ethical review is the most important tool by which professional researchers ensure that the ethical
standards of our work are maintained. It is essential that all ethical concerns are resolved before the
research project begins.

Informed consent and voluntary participation

As mentioned previously, participants have a right to know what the research is about, how it will affect
them, the risks and benefits of participation, and the fact that they have the right to decline to participate or
to discontinue their participation at any time during the process if they choose to do so. Usually, a
researcher or an assistant will take some time to explain to participants what the study entails and what is
required of them in terms of participation. Each participant may then be asked to sign an informed consent
form, which is an indication that they indeed understand what has been explained to them. Typically,
participants receive a copy of the form for their own records.

Confidentiality
Confidentiality is an ethical requirement in most research. Information provided by participants,
particularly sensitive and personal information, should be protected and not made available to anyone other
than the researchers. Thus, data collected from participants should at all times be kept under secure
conditions.

Anonymity
The principle of anonymity is linked with confidentiality. A participant's data must never be associated
immediately and obviously with his or her name or any other identifier. Usually, researchers assign a
number to a participant's data to ensure that the data remain anonymous. Anonymity applies to all aspects
of the research process, from the first time that the researcher makes contact with a potential research
participant, to the publication of reports and findings.

Appropriate referral
Sometimes it may be the case that research participants' well-being (psychological, emotional, physical
and/or social) may be compromised as a result of participation in research, for example, after a traumatic
experience such as war, social turbulence or displacement. The possibility of this happening must always
be anticipated by the researcher and processes need to be put in place to manage any negative
consequences should they occur. This is most commonly done by arranging appropriate referral to a
counselling centre, social work agency or medical facility.

Discontinuance
Participants must be given every assurance that that they are free to discontinue their participation at any
time without being required to offer an explanation. Thus, at any time in the conduct of a research project,
if a participant decides for any reason that he or she would like to stop participation, this wish should be
respected. Participants should also be informed that stopping participation will not prejudice their
receiving services, if the project is conducted at a clinic, government department, school or counselling
centre.

This can be important, as some participants may feel that their social benefits may be affected if they do
not agree to participate or if they discontinue their participation. In rare cases, there may be unavoidable
negative consequences for the participant who withdraws from a study, such as developing drug resistance
following withdrawal from tuberculosis treatment. In these cases, it is the responsibility of the researcher
to ensure that the participant fully understands the consequences of his or her decision. The researcher
should also take the necessary steps to get further assistance for the participant leaving the study (see
Appropriate referral above).

Research with vulnerable populations

Vulnerable persons are those who may not have the necessary degree of understanding to give informed
consent to participate in research. Examples of vulnerable persons include children, mentally ill persons,
disabled persons, unemployed persons, refugees, asylum seekers and homeless persons. Whenever
research involves vulnerable persons, the researcher must be sensitive to their needs and the fact that they
may not fully appreciate the implications of requests to participate in research. At the same time, it also
inappropriate to be patronising or condescending to people and make assumptions about their competence
that may be unfounded. In some cases, when it is not possible or appropriate to obtain informed consent
from a participant (for example, a child).it may be necessary to obtain permission from a third party (for
example, a parent). The following case example illustrates some of the challenges of working with
vulnerable populations.
 Deception
In deception, the researcher hides the true nature of the study from the participant. This is done to prevent
the participants from altering their natural behaviour as a result of knowing that it is being observed.

Deception causes many ethical problems. If used, safeguards must be employed.

The most common safeguards are:

 The researcher needs to ensure that the deception poses no serious or long-term risks.
 The researcher needs to explain the true nature of the deception to the participant in a debriefing,
which occurs once all the data has been gathered.
 The debriefing should counter any lingering misconceptions, possible discomfort or risk that may
have been generated by the deception.
 Debriefing should be done with care to ensure that participants are not left with bad feelings or
doubts about themselves based on their performance in the study.

Ethics in analysis and reporting

Researchers are not allowed to change their data or observations. The fabrication or falsification of data is
a very serious ethical transgression. There have been some examples in the history of social research where
data have been fabricated or falsified. When reporting the results of a study, the researcher should report
on technical shortcomings, failures, limits of the study, negative findings, and methodological constraints.
The final ethical responsibility always rests with the researcher.

In general, the researcher must always judge the research in terms of:

 its value to science

 the amount of risk it poses to the participants
 whether the potential benefits outweigh the risks, and
 whether adequate safeguards have been included to minimise the risks.

Reporting back to research participants

It is the duty of the researcher to ensure that the rights of participants to be informed of the results of the
study are protected. This means that the researcher must present the results of every study to the research
participants in a form that is easily understandable. Factors such as participants' home language, culture
and education level must be considered. Foreign researchers have a duty to submit a report of their
research to the appropriate authority of the host country.

Obtaining access to research participants by means of gatekeepers

Sometimes it may be necessary for a researcher first to approach a gatekeeper before approaching
participants directly to participate in a study. For example, if you wish to research children attending
school, it will be necessary to contact the school and even the relevant national or provincial educational
authorities. The same applies to other institutions such as hospitals, prisons, private companies or
commuters using the transport system, if participants are to be approached on the premises of these
institutions.
 Relationships with organisations
Research is often made easier by organisations or institutions that assist researchers in gaining access to
particular groups of people. Many organisations will be interested in learning the results of the research
that they have supported. They may want to use the findings to inform their own practice and policies. For
this reason, it is important that researchers make sure that they return to the organisations that have assisted
them in data collection in order to pass on their results and conclusions. This is one way in which research
can directly make a difference in the lives of people. Research is not only about knowing more about our
world. It is about making informed decisions about what needs to be changed so that peoples' lives can be
improved. Feedback should be provided in a form that can be easily understood by the audience. Not all
organisations will be able to digest a scientific report. In such cases a meeting and discussion might be
more appropriate.

Publication
Publication of research results is important, as it is a way of communicating the results to the scientific
community. When research results are published, it is important that participants not be identified by name
or in any other way that would make it possible for them to be identified. Publication credit must be given
to all persons who have contributed to the research, either in the form of authorship or in the form of an
acknowledgement. Plagiarism has become a serious problem at many institutions. Plagiarism is when a
writer presents material or research findings as his or her own work when it is in fact the work of someone
else. Plagiarism is essentially academic theft, and for this reason it is a serious offence.

Occasionally, publication presents some difficult problems to researchers. Consider the following case: An
international agency concerned about education conducts research to see how a particular country's high
school syllabi are changing over time. The results demonstrate that, in one country, the mathematics
syllabus has been gradually simplified over many years. The education department demands that the
agency not publish the research report because it will undermine confidence in the department.
 UNIT 2- CHAPTER 3-
PROBLEM CONCEPTION
AND BACKGROUND
INFORMATION
Sources and identification of research problems Pg 41
Although social reality provides innumerable questions, selecting a research problem or question is a
delicate task. Out of the great variety of queries arising from the environment, one has to sort out those that
are appropriate for scientific investigation. Only general guidelines on how to choose and formulate a
research problem can be stated, as well as some criteria that should be satisfied for a successful research
project. The difficulty of providing precise directives for the selection of research questions and problems
arises from the diversity of possible topics and their sources.

Sources of research topics Pg 42-43

 Observation of reality- The most obvious source of research topics is contact with, and
observation of, the external world. Many of the world's great scientific advances have been the
result of nothing more than observation.
  Theory- Theory is another source of research topics. If a theory is correct, one expects that
events will unfold in pretty much the same way as predicted by that theory. Research is necessary
to verify the correctness of the assertion that the particular situation will occur as spelled out in a
theory.

 Previous research- Previous investigations often inspire new studies because of

contradictions between results or concerns about the procedures used. Alternately ambiguous
results may require further exploration. Finally, it might be useful to repeat earlier studies with
different groups of participants.

 Practical concerns- Very often, research stems from concrete problems encountered in the
everyday life of a certain community.

 Personal interest- Sometimes, research projects are undertaken as a result of the personal
interest of the researchers. In this case, the relevance of the research to the broader community is
sometimes of less importance. Research out of personal interest may simply be to satisfy one's
curiosity or generate more knowledge.

 6. Researching topics of interest to another party- In many cases social researchers

earn their money by taking on contract research which is paid for by another party. This other
party might, for instance, be a business which requires some market research about a new project,
a government department that wishes to test the feasibility of a new social policy, or a civil society
organisation that needs evidence about a particular problem being experienced by the community.

Identifying a research problem Pg 45-48

In the identification of a research problem, one can distinguish three steps in narrowing the range of
interest. These are: the selection of a topic area, the selection of a general problem, and the reduction of the
general problem to one or more specific, precise and well-delimited questions. In general, the sources of a
research problem are to be found in a combination of direct observations and experiences, theory, previous
investigations, and practical concerns.

The choice of the topic area may be dictated by various factors: intellectual or academic (the social
organisation of cockroaches - topic: comparative social psychology), practical interest (the queuing at bus-
stops - topic: urban public transport), or personal interest (why do I forget so quickly what is taught in
lectures? Topic: learning, motivation, and memory). Nowadays, topics of research are strongly influenced
by social conditions. It may be 'fashionable' to conduct research into a particular area during a particular
period.
Within a topic area, a general problem must be chosen which may be mainly of scientific and intellectual
interest, or revolve around a practical concern. In either case, the objective might be for more information
about a particular issue or it might be to explain the relationship between existing facts.

Another important constraint to be borne in mind when choosing a research problem is that, although the
problem relates directly to empirical reality, the feasibility of data collection may be doubtful or
impossible.

Based on the above, a well-chosen Problems. or question should satisfy as many possible of the following
criteria:

 Be timely.

 Relate to a practical problem or question.

 Relate to a wider population.

 Relate to an influential or critical population.

 Fill a research gap.

 Permit generalization to broader principles of social interaction or general theory, or enhance our
understanding of a particular situation.

 Sharpen the definition of an important concept or relationship, or shed more light on a concept or
relationship.

 Create or improve an instrument for observing and analysing data.

 Provide possibilities for a fruitful exploration with known techniques.

 Have implications for a wider range of practical problems.

Finally, before going ahead with the actual study, a researcher should check whether the following more
general criteria are satisfied.
  Empirical testability. The question must refer to empirical facts and be answerable through
the observation of reality. The question should not be based on moral and ethical judgements, or
cultural beliefs. Scientific research cannot answer such questions.
 Feasibility. This refers to whether or not the proposed study is manageable, taking into account
the available time and literature, financial means, the size of the sample, and the available methods
or instruments for collecting data.
 Critical mass. This refers to the breadth, or scope of the research. Is the problem so narrow,
specific or trivial that it is not worthwhile pursuing?
 Interest. This refers to the motivation of the researcher to carry out the proposed research. Since
conducting a research project is a long and demanding task, strong motivation is essential.
Motivation that is only external (like the need to carry out a research project as part of a course on
research methodology) will often lead to a deficient, biased research outcome. If, on the other
hand, the choice of topic is dictated by a deep-seated interest, the quality of the research is likely
to be positively influenced by such interest. When the research process becomes challenging, it is
interest more than anything that sustains the investigator to carry the project through.
 Theoretical value. The importance of the theoretical relevance of research has already been
reflected in many of the criteria given for a well-chosen problem. The general idea can be
summarised by the question of whether or not the research will contribute to the advancement of
knowledge in a particular field of research and how useful this knowledge will be for the further
development of that field.

 Social relevance and practical value. This refers to the relevance of the research
results to society. What changes will the results effect in the actual life situation of those studied?
How can the study results be adapted to the present situation so as to improve it? Are practitioners
likely to be interested in the result? There has been a great deal of discussion and debate about the
obligation of research projects conducted in Africa to be plainly pertinent to the needs of African
societies.

Literature review Pg 49
Literature review involves a search and study of current writings on the problem under investigation. In
order to conceive the research topic in a way that permits a clear formulation of the problem and the
hypothesis, some background information is necessary. This is obtained mainly by reading whatever has
been published that appears relevant to the research topic. This process is called the literature review.

Although acquaintance with different theories and models, as well as research results, takes place, by
necessity, before a clear statement of the problem can be formulated, a literature review is an ongoing
process.
 The purpose of the review Pg 49-50
 To sharpen and deepen the theoretical framework of the research. That is, to study the different
theories related to the topic, taking an interdisciplinary perspective where possible.

 To familiarise the researcher with the latest developments in the area of research, as well as in
related areas. In particular, the researcher should become acquainted with the problems, questions,
hypotheses and results obtained by other researchers in order not to duplicate existing efforts but
to widen and deepen them. Previous results are a starting point for new research.

 To identify gaps in knowledge, as well as weaknesses in previous studies. That is, to determine
what has already been done and what is yet to be studied or improved. In many fields of study,
theories developed in one context are assumed to hold in other contexts too.

 To discover connections, contradictions or other relations between different research results by

comparing various investigations.

 To identify variables that must be considered in the research, as well as those that might prove to
be irrelevant. This finding is often a result of the comparison of different investigations, done in
different contexts.

 To study the definitions used in previous works as well as the characteristics of the populations
investigated, with the aim of adopting them for the new research. Often some definitions are found
to be correct and unbiased so that they can be adopted for the new investigation along with other
basic characteristics of the population. By using the same definitions we facilitate comparisons
between our research and the work of others.

 To study the advantages and disadvantages of the research methods used by others, in order to
adopt or improve on them in one's own research. It is unwise to impose research methods on a
particular community without due regard to the culture of that community. Some research topics
are easier in some cultures and much more difficult in others. Anthropologists speak about cultural
relativism, the recognition that different theoretical frameworks and methodological approaches
have different utility within different cultures.
 CHAPTER 5- TYPES OF
RESEARCH
Ways of classifying research
There are several ways of classifying research studies. One of the most important focuses on the
methodology used. Quantitative research relies on measurement to compare and analyse different
variables. By contrast, qualitative research uses words or descriptions to record aspects of the world.
Mixed-methods research uses both measurements and descriptions in a complementary fashion to
deepen the researcher's understanding of the research topic. The differences between these types of
research are discussed in greater detail and illustrated below.

A second way of classifying research arises from the reasons for which the research is being conducted,
that is to say, the research aim. Studies that primarily seek to increase human understanding of a particular
aspect of society are often referred to as basicsocial research Pg 59. By contrast, studies that
primarily aim to solve a particular problem confronting a group of people are often referred to as applied
social research.

Participatory research, action research, community needs assessments , as well as

monitoring and evaluation studies tend to be more applied in nature and are discussed further in
later chapters.

A third classification of research is between primary and secondary research. Primary research occurs
when the researcher collects data with the specific intent of answering a particular research question.
Secondary research occurs when a researcher uses data that has been collected for some other reason
to answer a new research question.

A final, and more traditional way of classifying research is based on the demands of the research question.
In cases where very little is known about the research topic, we speak of exploratory research Pg 60.
The purpose of exploratory research is to determine the breadth and scope of a particular topic, and to
generate initial questions or hypotheses to guide more searching research.

Where the researcher is interested in describing a phenomenon, the research is called descriptive
research Pg 61. Descriptive research may be conducted from a quantitative or qualitative perspective.
Patterns and trends in social phenomena such as the population characteristics, or the human development
index of a country, can be described in figures and statistics. Descriptions can also take the form of verbal
narratives and descriptions derived from interviews, essays, novels, and even poems and songs.

When the research question requires an understanding of the relationship between variables, the research is
called correlational research Pg 61-62. Both quantitative and qualitative methodologies can be used
to determine the existence of a relationship between variables. Quantitative methods do this through
statistical approaches (see Chapter 14) while qualitative methods do this through documenting the co-
occurrence of particular themes in the data (see Chapter 15).

Finally, when the research question demands that the researcher explain the relationship between variables
and demonstrate that change in one variable causes change in another variable, the research is called
explanatory research Pg 62. Explanatory research is used to test theories of how social scientists
believe the world works. This is the most challenging form of knowledge production in the social sciences
for several reasons. Firstly, it is philosophically challenging to prove that change in one variable causes
change in another. Secondly, very few phenomena in the social sciences have a single cause, and causal
pathways are seldom one-directional. Social scientists find that they are frequently studying complex
systems with multiple causes and effects, often relating to one another in a cyclic or circular fashion.
Thirdly, it is very difficult to filter out all the other factors that might confuse or obscure the particular
causal relationships in which the researcher is interested.
 Quantitative, qualitative and mixed-methods research
Quantitative research methodology relies upon measurement. counting, and the use of various
scales. These are discussed further in Chapter 12. Numbers form a coding system by which different cases
and different variables may be compared. Systematic changes in 'scores' are interpreted or given meaning
in terms of the actual world that they represent.

However, there are some kinds of information that cannot be recorded adequately using quantitative data.
These words might come from recorded interviews or focus groups, written responses to open-ended
questions, diaries, letters, stories and other forms of literature, or from the field notes of a diligent observer
of social phenomena. Research using this kind of data is called qualitative research.

At the same time, there are also ways in which text can be coded to produce numbers, and mixed-
methods research attempts to combine the advantages of quantitative and qualitative methods and to
avoid their disadvantages. Researchers using mixed methods will collect both quantitative and qualitative
data and then find ways of combining, or mixing, the two types.

Basic and applied research

Sometimes the researcher's primary motivation is to contribute to human knowledge and understanding
about a particular phenomenon. This is usually achieved by gathering more facts and information that may
challenge existing theories and allow new ones to be developed. The actual utility or application of this
newly acquired knowledge is of little concern to the researcher. This kind of research is called basic
research.
At other times the researcher's primary motivation is to assist in solving a particular problem facing a
particular community. This is referred to as applied research and is often achieved by applying basic
research findings to a particular community's challenges. In this way applied research may assist that
community to overcome their problem or at least design interventions to address it.

Criteria for selection of research type

It should be clear that the choice of the type of research, whether exploratory, descriptive, correlational or
explanatory, cannot be arbitrary. It depends on the following factors:

 The object of research. Does the researcher have enough information to establish a
relationship between variables? Can the different variables be manipulated or controlled in order
to permit a study of the effect of one variable on another?
 The aim of the research. Does it benefit the researcher to establish a causal relationship, or
is correlational or descriptive research more appropriate?
  The nature of the data to be collected. How sophisticated are the available techniques
that will be used for data collection? Do they require further development before they can be used
in more elaborate research? How many participants does the researcher have access to?

Research and the internet

The growth of the internet over the past 20 years has produced many new opportunities (as well as some
dangers) for social researchers. Africa has the lowest internet penetration of all the continents. Only about
6% of Africans have access to the internet, compared to about 35% of people globally. This is sometimes
referred to as the 'digital divide'. However, advances in cellular phone technology are giving more and
more African people access to the internet.

 Data collection can be quicker and more affordable.

 The researcher can reach respondents over a very broad geographic area, such as the whole of
Africa, or even the whole world.
 Respondents can participate in data collection from the privacy and comfort of their own homes.

CHAPTER 6- RESEARCH
QUESTIONS AND
VARIABLES
 Formulation of the research question
A research question is usually expressed as a general question about the relationship between two or more
concepts or variables.

It is important to decide exactly which question it is that you are working on before you begin. The
research question itself always includes various concepts or variables. It is very helpful to identify and
define these carefully. How to achieve this is discussed in the next sections.

Concepts, variables and constants

The word 'concept' refers to a particular idea that can be described in words. As such it is more closely
associated with qualitative research. The word 'variable' has a similar function in quantitative research.
'Variable' means that something varies and that its variation can be measured.

Since a variable is something that varies from one observation to the next, it needs to be contrasted with a
constant, which does not vary. If you recruit a sample of people into a study, it is likely that the
members of the sample will vary in age and therefore age is a variable. They may also vary in terms of
intelligence, socio-economic status, linguistic group or gender. All of these are variables.

Identification of the variables

Some variables may be the ones influencing other variables, determining the values of these affected
variables. These are the independent variables (indicated by IV). Other variables may be subject to
other causes so that their values are influenced by the values of other variables. These are the dependent
variables (indicated by DV). Independent and dependent variables are the two most important types of
variable. As the names indicate, both are tied to each other by a certain relationship: the variations in one
(the dependent variable) are a function, or a consequence, of the variations in the other (the independent
variable). Put another way, changes in the IV cause changes in the DV.

Independent and dependent variables are defined more precisely as follows:

The independent variable is that factor which is measured, manipulated or selected by the researcher to
determine its relationship to an observed phenomenon that constitutes the dependent variable.

The dependent variable is that factor which is observed and measured to determine the effect on it of the
independent variable, that is, it is that factor that appears, disappears, diminishes or amplifies, in short,
varies as the experimenter introduces, removes or varies the independent variable.
Another kind of variable is the moderator variable. In the simple example of the relationship between the
age and height of children, a researcher could argue that other factors also influence the height of a child.
These include the child's sex, the height of his or her parents and the quality of the child's diet. These
additional variables affecting the dependent variable are called moderators. In other words, the moderator
variable is that factor which is measured, manipulated or selected by the researcher to discover whether or
not it modifies the relationship between the independent and dependent variables. Since moderator
variables share some of the characteristics of independent variables (in that they are measured or
manipulated by the researcher to observe the effect on the dependent variable), they can be thought of as
'secondary independent variables'

Independent and dependent variables are defined more precisely as follows:

 The independent variable is that factor which is measured, manipulated or selected by the
researcher to determine its relationship to an observed phenomenon that constitutes the dependent
variable.
 The dependent variable is that factor which is observed and measured to determine the
effect on it of the independent variable, that is, it is that factor that appears, disappears, diminishes
or amplifies, in short, varies as the experimenter introduces, removes or varies the independent
variable.

Another kind of variable is the moderator variable. In the simple example of the relationship
between the age and height of children, a researcher could argue that other factors also influence the height
of a child. These include the child's sex, the height of his or her parents and the quality of the child's diet.
These additional variables affecting the dependent variable are called moderators. In other words, the
moderator variable is that factor which is measured, manipulated or selected by the researcher to discover
whether or not it modifies the relationship between the independent and dependent variables. Since
moderator variables share some of the characteristics of independent variables (in that they are measured
or manipulated by the researcher to observe the effect on the dependent variable), they can be thought of as
'secondary independent variables'

Control variables are those factors that are controlled by the researcher to out or neutralise any effect
they may otherwise have on the observed phenomenon.

The antecedent variable, as indicated by its name, appears before the independent variable and
determines it.

The intervening variable is in some ways the opposite. It is a consequence of the independent
variable and it determines the variation of the dependent variable. It falls between the independent and
dependent variables, so that, if it is held constant, the latter will not vary.

An extraneous variable is a variable that influences both the independent and dependent variables,
giving the false impression of a relationship between them.
 Conceptual and operational definitions
A Conceptual definition is the definition of a concept in terms of of set of other concepts.

A conceptual definition cannot be true or false, but it may or may not be useful for communication. Below
are some properties that conceptual definitions should have in order to be useful.

 A conceptual definition must denote the distinctive characteristics of that which is defined. It must
include all things belonging to this class, but only these things.

 A conceptual definition should not be circular. In other words, it should not describe something by
using the same concept.

 A conceptual definition should be stated positively, expressing the properties shared by the objects
and not the properties they lack.

 A conceptual definition should be stated in clear and unequivocal terms to avoid different
interpretations.

Even a very well-formulated conceptual definition does not fulfil the need of a researcher who wants to
assess the existence of some empirical phenomenon. This is the function of an operational
definition. Thus an operational definition not only gives precise indications as to what the fundamental
characteristics of a concept are, but it also gives precise indications about how to observe, describe or
measure the characteristics under study. Stated in another way: an operational definition is based on
the observable characteristics of an object or phenomenon and indicates what to do or what to observe in
order to identify these characteristics.

Hypothesis formulation
Essentially, research problems are questions about relationships among variables, and hypotheses are
tentative, concrete and testable answers to such questions. In other words, a hypothesis, which is a
suggested answer to a question, has to be tested empirically before it can be accepted and incorporated into
a theory. If a hypothesis is not supported by empirical evidence, it must be rejected and the researcher is
obliged to suggest another one. In this sense, the role of hypotheses is not only to suggest explanations for
certain facts or problems but also to guide the investigation. While it is possible to develop hypotheses in
qualitative research, the formal statement and testing of hypotheses is more commonly associated with
quantitative research. This is because qualitative research is often more exploratory and descriptive in
nature.
The following are the main characteristics of usable hypotheses.

 A hypothesis must be conceptually clear. All variables identified must be clearly

described, using operational definitions.
 A hypothesis should have empirical referents. This property is an essential feature
of a scientific approach to problems. It is fulfilled as soon as operational definitions have been
found for all the concepts appearing in the statement of the hypothesis. Particular care should be
taken to avoid moral judgements, attitudes and values. Expressions like 'good', 'bad', 'ought to',
'should', and the like are not scientific.
 A hypothesis must be specific. This property reflects the fact that the range of the
problem must be narrow enough to allow a precise, well-delimited investigation. If the problem is
too wide, the hypothesis will be too general and thus not testable. Again, good operational
definitions will safeguard against this pitfall. A hypothesis such as 'peasants are more prejudiced
than workers' may sound interesting, but, unless 'peasants', 'workers' and, above all, 'prejudiced'
have been given operational definitions, the hypothesis is not testable. 'Prejudiced' as such,
without specifying that the prejudice is directed toward other racial groups, nationalities or foreign
foods, may be beyond measurement. Moreover, a specific hypothesis makes mention not only of
the independent and dependent variables, but also of the moderator and all control variables. As
mentioned above, the effect of working conditions on nurses' bedside manner can only be reliably
analysed if all other factors which could also influence bedside manner, such as nurses' training
and the patient's social class, have been taken into account. Thus, the conditions under which
relationships between variables are examined must be stated explicitly.
 A hypothesis must be testable with available techniques. Although some
investigations may be relevant and specific, it may not be possible to carry them out because of a
lack of instruments or methods to measure or describe one or other of the variables. One can easily
measure the temperature of a sick child, but how does one measure a lack of affection that affects
the child's health? Thus, when formulating a hypothesis, it is very important to ensure that
adequate techniques of observation and measurement are available. One way to go about this is to
study carefully the operational definitions and be sure that they clearly indicate the methods of
measurement or description.

The null hypothesis is a hypothesis the states that the variables in question are not in fact related.
 UNIT 3- CHAPTER 7-
PARTICIPATORY AND
ACTION RESEARCH
Participation in knowledge generation

Paulo Freire, the famous Brazilian teacher, philosopher and critical thinker, argued that:

... apart from inquiry, apart from the praxis, individuals cannot be truly
human. Knowledge emerges only through invention and re-invention, through
the restless, impatient, continuing, hopeful inquiry human beings pursue in the
world, with the world, and with each other.

(Pedagogy of the Oppressed, 1970)

Freire believed that what makes us truly human is the way we constantly create and recreate meaning
about our world. He believed that knowledge is created through people struggling, debating, observing,
analysing and arguing with each other about how to make the world a better place in which to live. When
one group creates meaning about another group, the former wields power over the latter.

In contrast to the other kinds of research previously discussed, participatory research is

distinguished by two characteristics: the relationship between the people involved in the research, and the
use of research as a tool for social change, as well as for increasing human knowledge. Thus, the
conventional roles of researcher (the expert) and participants (naive objects unaware of the research
hypotheses) are changed. Participatory research encourages the active participation of the people whom the
research is intended to assist. In this way, it empowers people to be involved in all aspects of a project,
including the planning and implementation of the research, and any solutions that emerge from the
research. Everybody involved in the research project works together as a team.

Participatory research techniques focus on particular problems facing communities and attempt to use
research (and the resulting action) as a tool to bring about social change. As such, the nature of
participatory research is predominantly applied. Such social change is achieved through the democratic
collaboration of social researchers, community members and various other parties. Together, and as equal
players, the participants investigate the problem and its underlying causes (including socio-economic,
political and cultural factors) and then take collective action in order to bring about long-term solutions to
that problem.

Participatory research uses all the conventional tools of social research. However, tools that acknowledge
the value of the opinions and thoughts of all people, such as focus groups, indepth interviews and
participant observation, tend to be more popular than structured interviews, questionnaires and simple
observation. As a result, there is no one way of doing participatory research.

A form of participatory research that is growing in popularity is action research, an approach that suggests
a particular complementary relationship between action and research.

Steps in action research

Participatory research is not necessarily action research, although action research 15 always participatory.
Beyond the characteristics of participatory research introduced above, action research demands that the
social scientist and the community work as equal partners in the planning and implementation of a project.
Each brings valuable resources to the project. Furthermore, action and research take place alternatively in
an ongoing learning process for everyone involved. Because of this dynamic process between researcher
and community, and between research and action, there is no general formula for doing action research.
One can, however, sketch a very broad framework that describes how action-research projects often
proceed.

1. Request for assistance. Ideally the initiative for an action-research project comes from members of a
community who find themselves in difficulty. The idea should never come from social researchers who
believe (rightly or wrongly) that they know what is best for a particular community and set out to
demonstrate to the community the best way to proceed. Very often requests for assistance come to social
scientists via social workers and others in the front line of service delivery. This is because community
members themselves are often not aware that the services of social scientists are available to them. They
are, however, familiar with people who work within the community and it is to these people that problems
are first referred. In this way, frontline workers play a very important role in bringing the challenges facing
communities to the attention of social scientists. They can also help social scientists achieve a better
understanding of what is happening on the ground.

At all times, however, the social scientist must be wary of disempowering a community by dictating what
problems that community is facing and how such problems ought to be solved.

2. Negotiation. Negotiation most often occurs between the social scientist and representatives of the
community. This raises an important ethical issue for the social scientist. In some cases, community
representatives may not be truly representative of the community. It may be necessary for the social
scientist to check whether the situation, as presented by the community representatives, accurately reflects
the will of the majority of people within that community. This is called an 'accountability gap'. Also,
the social scientist and the community representatives should make sure that every person or organisation
that may have something important to contribute to the project is present to discuss the problem. This list
may include religious, labour, political, youth and other kinds of groups, as well as government, non-
government and funding organisations.

At this point all the participants should establish a broad ethical framework based on mutual trust, within
which they can work together. In other words, if the representatives from any group feel that they will not
be able to work usefully with people from another group (for whatever reason), this should be discussed
during this phase. In a community where rival gangs exist in violent conflict, one gang may refuse to
participate in the project if the other gang is present. In this case a process of mediation is needed before
the project can begin. Opposing religious groups and political parties may experience similar problems.

Once a comfortable working relationship has been established, it is time to consider the goals of all the
groups that will be involved in the project. In most cases these goals will be different. Consider an action-
research project involving only two parties or co-researchers: the community and a team of social
scientists.

The community desires a solution to its particular problem. The social scientists hope to identify a more
general solution adaptable to a range of similar problems that may be experienced by other communities.
For the social scientists' goal to be met, the community may have to participate in a far more complex
project than its original one. Similarly, if the community's goal is to be achieved, the social scientists
cannot stray too far from the specific details of the problem at hand. Thus both the social scientists and the
community must reach a compromise on the goals of the project. The complexity and importance of
negotiation increase when more than two parties are involved in the project.

In deciding upon the various roles of the co-researchers it is important to constantly think critically about
power. Who will the decision makers be? Who will be paid for their time on the project, and who will be
expected to donate their time for free? Many action-research projects are managed by a committee of
people including both the social researcher and representatives of the community. Sometimes the social
researcher, as well as community members who are part of the data collection and analysis processes, are
paid for their time and expertise. Whatever the arrangements, as soon as such a contract has been agreed
upon, the participants are ready to enter the next phase.

3. Planning. In the third phase of an action-research project, all the groups involved must work together to
find a way of solving the community's problem and of meeting all the goals set out in the action-research
contract. Essentially, this phase involves three distinct tasks. First, the co-researchers must find a way of
defining the problem that is clear and acceptable to everyone. Very often this is not as straightforward as it
sounds. The co-researchers may have different ways of explaining the world. Reaching a point of shared
understanding might require long and patient discussion. Secondly, the co-researchers must determine
exactly what information is needed in order to find a solution to the problem, and how that information is
to be collected. This is the essence of the action-research project because at this point the co-researchers
decide what their specific aims are and how these aims are to be achieved. Also at this point it is necessary
for the participants to decide how the data will be collected. Finally, the co-researchers must break the
action-research project down into manageable tasks and distribute responsibility for each of these tasks
amongst the participants.

It is during the negotiation and planning phases that the participants build up a rapport, or sense of
cooperation. A good relationship between everyone involved is essential for the success of the action and
research involved in the final stages.

4. Implementation. Once the co-researchers have made a plan about how to proceed with the project,
implementation can begin. It is during this phase that action and research take place. Very often
implementation begins with a period of research where the resources and needs of the community are
systematically assessed and the necessary information to guide appropriate action is gathered. This process
then informs some kind of action which the action-research partners undertake together. Thereafter, the
results of the action are assessed, and a further period of research (although this time of an evaluative
nature) is initiated. Depending on the results of this research it may be necessary to develop or completely
redesign the original action undertaken. In this way, action and research continue as alternate processes in
the search for solutions to the community's problems.

Using action research in developing countries

Action researchers are often criticised by social scientists who are accustomed to conventional research
methods.

In an action-research project extraneous variables are difficult, often impossible, to control. It is therefore
very difficult to be sure positive results are due to the action taken by the co-researchers.

Action researchers usually respond to these criticisms by acknowledging that action research is not the
only form of research available and that for some kinds of problems action research is not appropriate.
However, in developing countries, where communities are in great need of immediate solutions to vital
survival problems, where social researchers are few and far between and often quite ignorant of the reality
to be addressed, and where financial resources are scarce, action research is a particularly valuable tool.
Action research,
1. is concerned with solving particular problems facing communities.

2. helps individuals, organisations and communities to learn skills and access resources so that they can
function more effectively in future.

3. provides a way of spreading the understanding gained through research to people and communities who
can benefit from those findings.

4. attempts to understand the individual and the community within a broader social context.

5. facilitates communication between social researchers and communities in need of assistance.

5. shakes the 'ivory tower' of many social scientists and makes their work directly beneficial to
society.
 CHAPTER 8- RESEARCH
AND COMMUNITY
DEVELOPMENT
Research and community projects

Applied social research has an important role to play in the planning, management and implementation of
community projects. In many parts of Africa the needs of communities are great and the resources
available to address those needs are few. For this reason it is imperative that the available resources are
used to the greatest possible effect. We cannot afford to waste precious resources on community
development projects that do not achieve the desired goals. Social researchers have an important role to
play in this regard. We can help to identify which community needs should be given priority so that when
addressed, they produce the greatest positive results for the greatest number of people. We can also design
research to identify the critical steps upon which a project depends, as well as the pitfalls that must be
avoided at all cost. In this way researchers are able to predict the likelihood of success under different
conditions, thereby supporting developers in achieving their goals. Finally, researchers are able to analyse
the actual implementation of small, pilot projects, thereby identifying further problems or opportunities
which could not be recognised before the project actually started, information that is extremely important
when it comes to multiplying the size of the project or bringing it up to scale so that it achieves its planned
results.

Unfortunately, good research is expensive and takes time. All too often development agencies and
governments are not prepared to spend the time and money necessary Rather than doing the research, they
focus on those goals they believe ought to be prioritised, they use the methods that they think will work,
and they go directly to the full-sized project without taking the time to test their ideas. Nobody living in
Africa has to look very hard to find well-intentioned community development projects which failed to
deliver the promised results. Whether one investigates the areas of housing development, job creation,
safety and security, healthcare, or any number of others, you will always find good ideas that were not
properly researched and so were ultimately unsuccessful. It is important to know that the money and time
spent in doing rigorous research is usually well worth the money and time saved on failed or partially
successful projects. This is one area in which development agencies and government could learn a lesson
from the private sector. Large retail corporations never launch new products or open new stores without
extensive, time-consuming and expensive research. They want to be sure that their idea will work before
they invest large amounts of money.

Needs assessment

The first step of any community intervention is a detailed analysis of the particular challenges facing that
community, and the available resources with which those challenges may be met.

Usually community-based organisations, local leaders and other representative structures believe that they
have a fairly good idea of the challenges facing their community. It is important to assess the extent to
which these role players actually do understand the concerns and needs of all people within the
community. (Refer back to the discussion of accountability gaps in Chapter 7.) It is important to examine
the extent to which community representatives have analysed the causes of the underlying community's
problems. It is essential that scarce resources are put to the best possible use so that the community derives
maximum long-term benefit.

Of course, different community members have different needs. When resources are scarce, these needs
might exist in competition with one another. Often young people would like money to be spent on sport
and recreation facilities, while their mothers are more concerned with better public transport to get their
children to school. Shopkeepers, on the other hand, might want to spend public funds on better lighting
outside their shops so that they can safely stay open after dark. Difficult decisions must often be made
about the best way to use limited resources. Good choices can only be made when decision-makers are in
possession of all the relevant facts. With the correct information it is possible to prioritise the various
needs and to tackle those correct information it is possible to prioritize the various needs and to tackle
those that affect the community the most profoundly first.

The needs assessment survey is a survey of various facets of the community in order to define the
different concerns of all the community's members. The first step in a needs assessment survey is to
negotiate entry into the community itself. Most communities have people who control access to the
members of that community. These people are sometimes referred to as gatekeepers. In rural
communities the gatekeepers might be traditional leaders or a council of elders, in urban communities they
might be a group of elected councillors or party officials. Access to a school will be controlled by the
principal or parent-teacher association, and access to a state hospital might require permission from a
senior official in the country's Department of Health. While negotiating entry, the researcher will gain
valuable insight into how the community runs, who has the most power and who has very little power,
what tensions exist between community groups, and what political alliances are in operation. This
understanding is important for the researcher to ensure that all groups are represented in the needs
assessment. It is especially important that those groups whose voices are unlikely to be heard at
community meetings or through structures that claim to represent all the people of the community, are
heard. Groups that are not sufficiently represented often include women, children, the elderly and the
disabled. In some communities people from particular ethnic backgrounds or who follow particular
religious beliefs find themselves without a voice in the usual structures. These people should also have a
chance to discuss the problems facing their community.

Pilot projects

A sound needs assessment provides the foundation for a community project. The results of the needs
assessment identify the priority needs or problems within the community and the root causes thereof.
Having done this work, it is possible to plan a pilot project to solve the problems.

A pilot project is a project that is run on a limited scale, for example, with a single small community
over a short period of time, in order to test the effectiveness and community support for the proposed
solution. Since pilot projects are trials of a possible solution to the community's problems, it is essential
that they involve a large research component.

The development of a pilot project depends upon the community developers being able to explain a theory
of change. A theory of change describes in a step-by-step fashion how the particular community
project activities are intended to produce the positive outcomes or changes hoped for by the community.
Often a theory of change is presented as a flow chart. We illustrate this by returning to the example of the
drought in west Sudan.

Feasibility studies

Even after completing a thorough needs assessment and a successful pilot study, it is difficult to predict
what will happen when the project is scaled up. Scaling up refers to the process of expanding a small
pilot project until its effects are felt across a whole district, province or country.

Feasibility studies try to identify further problems and obstacles that might arise when a large and
expensive project is implemented. Again, a process of checking assumptions must be followed. Can we
assume that the conditions that held for a pilot project will continue to hold in the same way on a large-
scale project? What new problems will arise as a result of the expanded scale of this project, and how do
we solve them? What opportunities might arise as a result of the scale and how can we make the most of
these? We illustrate this process by continuing our discussion of the example from west Sudan.

These questions are all concerned with the problems and opportunities associated with increasing the scale
of a project.

A feasibility study must do two things:

 It must help project planners to identify each of the assumptions underlying the scaled up project
plan. In other words, what have the planners assumed will happen as a result of the project?

 It must estimate the likelihood of each of these assumptions being met. Where the assumptions
have a good chance of being met, the project can go ahead as planned. But, where the planners
have made risky assumptions, it may be better to rethink the project or at least develop
contingency plans.

CHAPTER 9- PROJECT
MONITORING AND
EVALUATION
Monitoring is a process whereby key indicators relating to the activities and outputs of a project are
constantly measured. By keeping a watchful eye on such indicators, project managers can quickly tell
when a project is, or is not, reaching its goals, where the project might be failing, and possibly how to fix
the problems. Evaluation studies are used to assess the impact that a particular project has had, analyse
why it did or did not achieve its goals, or suggest ways in which impacts might be increased or even
multiplied in future. Monitoring is typically conducted by someone working within a project and thus has
an internal perspective. Evaluation may happen from an internal or external perspective, and reflects on the
overall success of the project. Both these research activities are intended to increase the effectiveness and
efficiency of interventions. Moreover, monitoring and evaluation work may also generate knowledge that
is transferable to other projects and contexts.
 Project indicators
…

Selection of indicators

The terms 'monitoring' and 'evaluation' are often confused. They are actually different enterprises, which
rely on different methods and have different objectives. However, both monitoring and evaluation require
the identification of indicators. Indicators are dependent variables that provide information about the
progress or impact of a particular project or intervention. Thus, for example, the quantity of produce,
measured in tons, reaped from fields would be an indicator of the effectiveness of different irrigation
systems used in those fields. If the tonnage is high, the irrigation must have been good. If the tonnage is
low, the irrigation system was probably somewhat ineffective.

By measuring indicators during the course of a project, project managers are able to monitor day-to-day
progress, and quickly tell when something has gone wrong. By comparing measurements of indicators
taken before a project with measurements of the same indicators taken after the project, a researcher is able
to evaluate the overall impact of the project. (This is a pre-test/post-test design that will be discussed
further in Chapter 10.)

Indicators should be identified for every step in the theory of change. Monitoring research is concerned
with indicators that reflect on day-to-day activities and intermediate milestones or outputs of the project.
Evaluation research is concerned with the indicators of overall impact. So, to fulfil both the monitoring and
evaluation functions properly, it is essential that indicators are found to help keep track of each action and
each anticipated outcome in the theory of change.

Usually monitors and evaluators have a choice of many different indicators for each step of the process.
The selection of the best indicators for a particular project can be a very sensitive matter. One needs to
consider the accuracy of the indicator as well as the cost and time needed to collect the data. Very
inaccurate indicators will be of little help in drawing precise conclusions about the effectiveness of a
project, but it is equally unhelpful to create a situation in which it takes more time and money to monitor a
project than it does to actually implement it. Remember that monitoring and evaluation are tools to
facilitate good community development work, but they should not drain resources away from that work.

Researchers commonly use the acronym SMART to assist them in selecting appropriate indicators.
SMART stands for:

 Specific
 Measurable
  Accurate
 Realistic
 Time-bound.

Specific means that an indicator should measure only that aspect of the theory of change that it is
intended to measure. Indicators are used to identify points at which the project is failing (in the case of
monitoring) or has failed (in the case of evaluation). An indicator that is not specific, that is, which
measures more than one aspect of the theory of change, will not help the researcher identify exactly where
the problem lies.

Measurable means that an indicator can be expressed quantitatively, in other words, that is it possible
to collect the necessary data.

Accurate means that the indicator is sufficiently sensitive to consistently pick up changes in the project.
Some indicators might produce different results when being measured by different people, or may miss
subtle changes in the aspect of the project being monitored.

Realistic means that the data can be collected given the time, finance, number of monitoring personnel
and technologies available to the researcher. There is no value in selecting an indicator that takes so much
time to apply or costs so much that the project timetable and budget is overwhelmed by the monitoring
process. This means that researchers must be very practical when planning a monitoring or evaluation
framework.

Time-bound means that an indicator should be collected according to a certain timetable. In some
projects the indicator might be measured on a daily basis, although for bigger community projects,
indicators are usually measured on a monthly, quarterly or even annual basis.

Measuring indicators

Once indicators have been selected, the researcher must develop some way of collecting data on each of
them. Some indicators are extremely simple and collecting data on them may be quick and easy. For
example, to monitor attendance at meetings one merely needs a record of the meeting noting who was
present, and to monitor the delivery of textbooks to schools, one merely needs to collect delivery notes
signed by the school principal. These are pieces of data that already exist in the normal administration of
day-to-day work.
 Project monitoring

Project monitoring is a tool for managing the ongoing implementation of a project. I. is important
that a monitoring framework be established during the planning phase of a project. To be effective, it must
be in place before the implementation phase begins The following steps outline a method for establishing a
monitoring framework.

 Identify at least one indicator for each activity or outcome in the theory of change. If resources
allow, two or more indicators should be identified, including a mix of quantitative and qualitative
options.

 Design an instrument to measure each indicator on a regular basis. Such instruments should not
be overly complex or unwieldy.

 Set clear targets for each indicator to show when an activity has been completed, or an outcome
reached.

 During implementation, measure each outcome regularly and compare the with the targets so as to
be constantly aware of which aspects of the project have been successfully completed, which
aspects are progressing smoothly and where any problems might lie.

 Produce regular brief reports to keep all participants and stakeholders up to date progress of the
project.

Project evaluation

The methods of social science can be used to assess the design, implementation and usefulness of social
interventions. This type of research is called evaluation research. Social interventions are most often
thought of as sophisticated programmes (such as drug rehabilitation programmes, campaigns against
malnutrition, skills training courses or the introduction of appropriate technologies to promote small-scale
industry). Actually, any attempt to change the conditions under which people live (no matter how simple,
or who is responsible for it) can be thought of as a social intervention. As has already been shown, one of
the central concerns of social research is action. Evaluation research aims to test interventions to see how
effective they are and therefore represents an important means of linking action and research in a
constructive manner. Social interventions may benefit from evaluation research in a number of ways.
Three of the most important ways are listed below.
  Evaluation research may be used to identify neglected areas of need, neglected target groups and
problems within organisations and programmes. This is referred to as diagnostic evaluation.

 A comparison of a programme's progress with its original aims is another of the functions of
evaluation research. This may serve to adjust the programme to the particular needs and resources
of the community within which it is situated. Evaluation designed to promote the future
effectiveness of a programme is called formative evaluation.

 Finally, evaluation research can furnish evidence of the usefulness of a programme. In this way a
programme may gain credibility with funding organisations, as well as the community within
which it is operating. This is known as summative evaluation.

Although different, each of these three types of evaluation is complementary and most evaluators are
expected to think about all of them simultaneously.

Diagnostic evaluation

Diagnostic evaluations are designed to inform researchers and project managers about the present
situations within communities, highlighting current problems, trends, forces and resources, as well as the
possible consequences of various types of intervention. Thus, diagnostic evaluation is a technique for
gathering data which is crucial to the planning of a new project. As such, it is important that this research
is carried out before a project is designed. The different ways in which diagnostic evaluations assist
organisations and communities are explained further below.

Firstly, it often happens that an organisation or community is aware that something is going wrong, but is
unable to identify exactly what the problem is. In other words, organisations and communities often find it
difficult to define their problems in a precise manner.

Secondly, diagnostic evaluation can help communities and organisations in societies that are under going
rapid social change. Such organisations or communities might desire change, but may not fully understand
how such change will affect them. Diagnostic evaluations can serve to map out the full range of probable
outcomes of any project.

Thirdly, during the initial planning stages of a project, it is useful for the project manager to specify the
project's broad aims in a couple of points. These broad aims may be usefully broken down into several
more manageable objectives that should be stated explicitly and in detail. This would allow them to be
evaluated (using other evaluative research techniques) at a later date. The specification of both the broad
aims and the various objectives must be guided by carefully conducted diagnostic research.

Finally, the researcher and the project manager should investigate problems that may arise during the
course of the project. Where such problems cannot be avoided, contingency plans should be established to
overcome them, so that the project may achieve its objectives.

Formative evaluation

Formative evaluation relates to the development and implementation of a programme. Its aim is to shape
the programme so that it will have the greatest beneficial impact upon the target community. Formative
evaluation is an evaluation of the programme in order to improve it. Longer term interventions use
formative evaluations at regular intervals during the life of the programme to ensure that it adapts to
changes in social reality and thus continues to produce the greatest possible benefit. The questions that the
social scientist must consider are of both a theoretical and a practical nature. On a theoretical level, the
social scientist must consider whether the content of the programme has been adequately adapted to the
social reality and whether conceptual definitions have been adequately operationalised. This is especially
important when the particular body of social theory has its roots in the United States of America, Europe or
parts of the world where the social reality is very different to that of Africa. All too often social scientists
wrongly assume that conditions and people are the same all over the world, and that theories developed in
far-off places are just as valid locally. On a more practical level, the social scientist must think about
problems such as the availability of resources for the project, the most effective ways of using those
resources, potential areas of difficulty in the programme, and so on.

There are many different ways of carrying out formative evaluations and the method depends largely upon
the project being assessed. Very often, however, the evaluation relies heavily upon the social scientist's
experience of similar programmes, understanding of social reality and theoretical knowledge. In some
cases, social scientists may not have all the available information needed for a satisfactory formative
evaluation.

One very useful method that does not depend on the social scientist's expertise in the area of the
programme is the reputability study. This technique involves the researcher identifying experts from
within the community, from academic institutions, from government and non-governmental organisations,
and elsewhere. Note that 'experts' are not only people with academic qualifications or important positions.
People from the community in which the programme is to be run (who may have no qualifications or titles)
are experts on conditions prevailing in the community, the nature and extent of community problems, the
community's likely reaction and many other aspects which might affect the programme. The choice of
experts is a sensitive issue and one should be careful to ensure that a wide range of people are represented
so as not to have bias in the final results. The researcher interviews the experts (either individually or as a
group) and presents an overview of the programme. The experts are asked for their opinions on the way the
programme has been planned and the method of implementation. Their comments and criticisms form the
base of the formative evaluation. Formative evaluations and reputability studies are almost always
qualitative in nature.

Summative evaluation

Another broad aim of evaluation research is summative evaluation. Summative evaluations set out to
determine the extent to which programmes meet their specified aims and objectives. This information is
used to gain credibility with various groups of people, particularly potential funders and future target
communities. Moreover successful programmes may be replicated in other communities. If the designers
of a programme can demonstrate scientifically that their programme has had certain positive effects, then
people are likely to be much more enthusiastic about the programme being implemented in their
community. They are also more likely to receive funding for similar programmes. Although summative
evaluations ought to happen at the end of a programme, they are often carried out at regular intervals
during the life of long-term programmes as well. The process of summative evaluation research generally
occurs according to the following five steps.

1. The identification of the aims and objectives of the programme. The

programme is thought of as the 'treatment', that is, it is one level of the independent variable (the
other being 'no treatment'). Summative evaluation compares the 'treatment' group with a 'no
treatment' group to see whether the 'treatment' has caused any positive change in the former. To
assess the change, the researcher must know what the 'treatment' is designed to achieve. The
identification of the programme's aims and objectives requires close cooperation between the
programme developers and the social scientist carrying out the evaluation. Unless the social
scientist knows exactly what the programme hopes to achieve, the evaluation cannot proceed to
the next step. The question of whether or not the aims and objectives were adequately chosen falls
in the domain of formative evaluation and is not considered here.

2. The formulation of the aims and objectives in measurable terms. At this point
it is important for the researcher to translate the aims and objectives into observable changes that
can be measured in the target community. The variable that is expected to change will become the
dependent variable of the evaluation research. In other words, the conceptual definitions of the
programme designers must be translated into operational definitions so that they can be studied
through the methods of social science.

3. The construction of the instrument of measurement. An instrument must be found

or designed which is capable of accurately measuring the dependent variable chosen in the
previous step. A wide variety of such instruments is available to social scientists, including
 carefully designed psychometric and sociometric tests, interviews, questionnaires and observation
techniques.

4. Designing the evaluation study and data collection. Designs used for summative
evaluation are discussed in detail in the next chapter. Note that in order to carry out a summative
evaluation, it is necessary for the researcher to compare the group that received the 'treatment'
with a similar group that did not receive the 'treatment'. These two groups represent the two levels
of the independent variable discussed in step one. The groups must be similar to start off with.
Thus, if the group receiving the 'treatment' changes positively, the researcher can conclude that the
'treatment' has been successful. Summative evaluations are almost always quantitative in nature
because inferential statistics (see Chapter 14) allow the researcher to estimate whether differences
are real or merely due to chance factors.

5. Reporting back. Once the evaluation has been completed, the researcher should present the
findings to those responsible for the intervention, the participants and any other interested parties.
This must be done in such a way that the methods, results and conclusions of the evaluation can be
easily understood, even by people with little experience of social research.

UNIT 4- CHAPTER 10-

RESEARCH PLANNING
AND DESIGN
 What is a research design?

A research design relates directly to the answering of a research question. Because research is a
project that takes place over an extended period of time, it is unthinkable to embark on such an exercise
without a clear plan or design, a sort of blue print. A design in quantitative research is a detailed outline for
the testing of a hypothesis, spelled out in clear and definite terms. It is a specification of the most
appropriate operations which need to be performed in order to test a specific hypothesis under given
conditions. This design is not, however, to be confused with research management which is a plan
to guide the researcher through the research process. In quantitative research, the vital question facing the
scientist is: What steps should be taken in order to demonstrate that a particular hypothesis is true and that
all other possible explanations have been rejected? On the other hand, the concept of research design in
qualitative research is more flexible, sometimes cyclical and much less detailed than it would be in the case
in quantitative research. In the qualitative case, as the many steps of literature review, theory building,
sampling and data collection do not constitute separate activities but are continuously interrelated (see
Chapter 2), the planning and design of the entire process must remain dynamic.

The purpose of research design is to ensure high internal validity. In quantitative research, internal
validity is concerned with the question, 'Do the observed changes in the dependent variable actually relate
to changes in the independent variable?' In other words, internal validity examines the extent to which a
particular research design has excluded all other possible hypotheses that could explain the variation of the
dependent variable. We call this eliminating plausible rival explanations or isolating the dependent
variable. In order to achieve high internal validity in quantitative research, a research design must control
as many extraneous variables as possible. In qualitative research, internal validity is sometimes referred to
as credibility (see Chapter 13) and is concerned with whether the researcher's method of data collection
and analysis (see Chapters 12 and 15) addresses the research question adequately. Will the researcher be
able to develop a convincing argument in anser to the research question from this study design?

In order to achieve the objectives of social research, the scientist requires a carefully thought-out strategy.
The first steps in constructing solid research planning and design require the researcher to answer several
fundamental questions about the research.

These questions relate to the focus, the unit of analysis and the time dimension of the problem at hand.
These concepts are explored below.

The focus of research

Social research can be used to explore almost any topic in the social world, although with differing degrees
of accuracy, certainty, confidence and success. However, it is useful to try to classify the most common of
these topics, since they sometimes require different types of research design. The focus of research may be
understood in terms of three different categories: conditions, orientations, and actions.

1. Conditions are studied when the researcher wishes to explore the current state of the respondents of the
research. For example, a researcher who measures the unemployment rate in 20 West African cities is
interested in the current condition of the labour markets in those cities. Similarly, a researcher interested in
the health of elderly people in rural areas might measure their heart rate after light exercise. This kind of
study would be located within the quantitative paradigm, since variables such as unemployment rate and
heart rate presuppose numeric measurement. The same holds true if the present conditions related to
unemployment (or health of a community) are investigated qualitatively through the use of focus groups
composed of members of the community under investigation.

2. Orientations are concerned with participants' attitudes, beliefs and lifestyles. Researchers interested in
religious and political views would, for example, be interested in orientations. Attitudes, beliefs and
lifestyles can only be quantified and therefore measured numerically with difficulty. Thus, they remain
qualitative in nature. However, an association may be found between respondents' attitudes, beliefs and
lifestyles that might be correlational and therefore quantitative. In studying orientations the researcher is
provided fertile ground in which to conduct a mixed-method study using both qualitative and quantitative
dimensions.

3. Actions are also very often the focus of research. These actions may be observed directly or may be
reported by the actor or others who observed the actor. For example, a study of how people travel to work
in urban areas might involve extensive observations of road, taxi and bus routes as well as trains (direct
observation), or alternatively it might rely on interviews with workers or employers (an indirect method).

In most cases, conditions, orientations, and actions are not mutually exclusive and the social researcher
must be sensitive to all of them simultaneously. For example, the relationship between unemployment and
depression involves all three. Experiences of being without work are likely to depend upon the length of
time participants have been unemployed and their current level of savings (conditions), their beliefs about
themselves, their futures, and their families (orientations), as well as the amount of energy they spend each
day looking for, or trying to create, work (actions).

The unit of analysis

The second important factor that the researcher must consider when planning an appropriate research
design is that of the unit of analysis. This becomes particularly important when the researcher begins
to draw a sample (see Chapter 11) with which to work. The unit of analysis is the person or object from
whom the social researcher collects data. The unit of analysis may be at a micro (individual, family,
group), mezzo (organisation, community) or macro (national) level. One of the aspects that clearly
distinguishes quantitative from qualitative studies is the procedure involved in selecting the units of
analysis (for more details, see Chapter 11). In quantitative studies, where the underlying purpose is
generalisation of the findings to the entire population, selection of participants is done largely through
probability samples. In the case of explanatory studies the principle of random assignment must also be
observed. In situations where participants have not been randomly selected, the researcher should exercise
caution before generalising findings to the entire population: the data from such a unit can only describe
that unit. However, when combined with similar data collected from a group of similar units, the data
provides an accurate picture of the group to which that unit belongs.

In qualitative research, the selection of the unit might change in the course of the research. For instance, in
the previous example, the researcher might have planned to interview only the members of the families
affected by unemployment. Other types of units, such as teachers and nurses, might have to be added in the
course of the research.

Although many people believe that social scientists always focus exclusively upon the phenomenological
experiences of individual people, this is not always the case. There are several possible units of analysis
which fall into broad categories.

 Individuals are the most common unit of analysis. In this case, the researcher investigates the
conditions, orientations, actions, attitudes and beliefs of a group of individual people. When the
individual is used as the unit of analysis, the people that actually take part in the study are often
selected because they belong to a particular group, such as police officers, women, people with
cancer, and so forth.

 Groups of people are also sometimes studied. Some examples of research where the unit of
analysis is more than one person are studies of siblings and identical twins, marital relationships,
family functioning and small-group functioning. In this case an entire group (and not each of its
members) constitutes one unit and can be compared to another group (another unit).

 Organisations with formal structures constitute a particular kind of group that is often used as the
unit of analysis for social research. In this case, questions of interest might relate to proportions of
employees from different social groups, organisational structure, profit and communication
channels.
  Occasionally, the unit of analysis is a period of time. For example, a researcher may wish to
determine whether there is a systematic change in infant mortality in a given community over a
20-year period (each unit is one year), or how much rain falls each month over a year (each unit is
one month), or how some cultural beliefs or values evolve through time.

 Finally, a common unit of analysis is a particular social artifact. Social artifacts are the products
of social beings and can be anything from poems and letters to automobiles and farming
implements. A systematic analysis of such artifacts may provide valuable information about the
individuals and groups that created or used them.

It is very important that the researcher keeps the units of analysis clearly in mind throughout the research
process. It is often very tempting to draw conclusions about one unit of analysis when, in actual fact, the
research is focused on a different one. This is called the ecological fallacy.

The time dimension

A third fundamental aspect of any research is the manner in which it deals with time. Observations may all
take place at a particular time or may be deliberately stretched over a long period.

When all data is collected at the same time, the research design is cross-sectional. The discussion of
correlational research in Chapter 5 is relevant here. The researcher using this design attempts to understand
a topic by collecting a cross-section of information relevant to that topic. For example, a survey of
nutrition patterns and infant mortality across a variety of living conditions may reveal differences between
peoples' nutrition and the conditions under which they are living, as well as a relationship between these
factors and infant mortality. The inherent difficulty with cross-sectional designs is that, because they do
not allow the researcher to measure change over time (since all the data is collected at once), it is very
difficult to demonstrate causality. In the above example, the researcher may be able to demonstrate that
infant mortality and nutrition are related, but will not be able to show that poor nutrition is a cause of infant
mortality. (Refer to the requirements of causality discussed in Chapter 5.)

Nevertheless, the immediate nature of cross-sectional designs, as well as the relative ease of data
collection, makes these designs the most common choice for social scientists.

Longitudinal designs spread data collection out over time. For example, a researcher interested in
unemployment in Nairobi between 2000 and 2009 conducts a broad yearly survey in the city in order to
find out what proportion of the population is employed at any one time. Data collection occurs several
times, in 2000, 2001, etc. (Note that the survey must take place during the same period of each year in
order for the results to be comparable.) Once all the data has been collected, the researcher may be able to
demonstrate that there is a predictable trend in the level of unemployment in Nairobi, information that may
be useful to job-creation programmes, town planners and many other people.

Evaluation research (discussed in Chapter 9) may be cross-sectional or longitudinal in nature. Some

longitudinal designs are introduced below.

 Cohort studies use a type of longitudinal design that tracks particular age group/s over time. For
example, the 'Birth to Twenty' project running in South Africa (Richter et al ., 2007) is recording
the development of hundreds of South African children over the first 20 years of their lives. These
researchers record a wide range of physical, emotional, social and educational variables relating to
children's development.

 Tracer studies are used to 'trace' people, or to follow their lives over a period of time. In most
cases, the data is collected at only one point, which is cross sectional (perhaps 10 years after
graduation or release from prison). As such, the tracer study is not strictly longitudinal, but
produces data that simulates a longitudinal design.

Types of research design

Every project requires a research design that is carefully tailored to the exact needs of the problem. Of the
many different designs that have been developed by social scientists over the years, only the few that are
most frequently used are presented here. More complex designs can be studied in specialised books.

Research designs have two essential components. The first is observation. At some point in every research
project, the researcher must observe and record differences in the different variables involved in the
research. Some research designs require that the researcher observe more than one group, at more than one
time. The second essential component of research design is the analysis of the relationships between the
variables. This may be done in quantitative research by manipulating certain variables in order to observe
the effect on other variables, or by observing corresponding changes in more than one variable. In
qualitative research, this analysis is done by comparing information and then selecting further participants
to help to clarify the relationship.

Three categories of research design can be distinguished for quantitative research according to the level of
scientific rigour involved in proving the causal relationship, or the aims of the qualitative research.
  Pre-experimental (exploratory and descriptive) designs (see also Chapter 5) are
essentially the methods of qualitative research. They satisfy the aim of the researcher to describe
and understand a phenomenon. As instruments for quantitative research, pre-experimental designs
are the least adequate in terms of scientific rigour and thus are least likely to establish a clear
causal relationship between the independent and dependent variables. In cases where quasi
experimental or experimental designs are not possible, the quantitative researcher is forced to use
a pre-experimental design, as these have far fewer requirements than the other designs. Three pre-
experimental designs are presented below: one-shot case study pre-test/post-test design intact
group comparison design.

 Quasi-experimental designs are designs which do not meet the exacting criteria of
experimental designs, but which manage to approximate experimental conditions. Although these
designs have fewer requirements than experimental designs, they can achieve a similar level of
scientific rigour. They often provide correlational relationships between variables. Four quasi-
experimental designs are presented:
1. contrasted groups design
2. post-test-only cohort design
3. pre-test/post-test cohort design time-series design.

 Experimental designs are the most rigorous of all the designs and have strict requirements.
They provide explanatory relationships between variables. Three experimental designs are
presented:
1. pre-test/post-test control group design
2. post-test-only control group design
3. factorial design.

Although the designs are presented here in order of increasing scientific rigour, it should be noted that
experimental designs were developed before quasi-experimental designs. The latter were developed to
accommodate the constraints of social reality that could not be accounted for using the former. Thus quasi-
experimental designs appear as alternatives to stricter experimental designs.

Pre-experimental designs

Pre-experimental designs are largely qualitative and often use small, non-probability samples (see Chapter
1). Descriptive and exploratory research respond to the need of qualitative researchers to gain a deeper
understanding of a particular social phenomenon Quantitative researchers sometimes use exploratory
designs to find out whether a full. detailed and rigorous investigation is possible. This is a way of
generating hypotheses which can be tested later with more sophisticated research designs.
 One-shot case study

In quantitative research, a one-shot case study is most often used to determine whether an event (or
intervention) has any effect upon a group of participants. The dependent variable is measured after the
event (post-test) and conclusions are drawn.

Qualitative research uses this method extensively: the most common use being case studies and focus
groups.

In quantitative research, the lack of an initial measure of functioning, often referred to as the baseline,
makes it very difficult to convincingly demonstrate change resulting from the event and therefore does not
exclude plausible alternative explanations as mentioned in Chapter 5. Thus the one-shot case study is
generally regarded as uninterpretable and, when used, is vulnerable to many criticisms. The following
design is used to overcome this problem.

Pre-test/post-test design

In the pre-test/post-test design, the quantitative researcher measures the dependent variable before (pre-test
or baseline) and after (post-test) the treatment or event that is expected to bring about a change. As a result,
the scores for the dependent variable can be compared at two points in time, and the difference between the
before and after scores may be due to the event that occurred between them. The same can be achieved in
qualitative research, when the perceptions of a situation before and after some event are compared, for
instance, by comparing the attitude of young people towards alcoholism before and after a public
campaign against drunk driving.

In quantitative research, the pre-test/post-test design is the most common one for impact studies or
summative evaluations (discussed in Chapter 9). In this case the event is the intervention that is being
evaluated. However, from a quantitative viewpoint, even this approach raises some concerns. It is not
impossible that other changes occurred at the same time as the event (the installation of the monitoring
system) and those changes, not the monitoring system, might have been responsible for the change in the
dependent variable. This is particularly true when a long period of time has elapsed between the pre- and
post-tests. These changes threaten the validity of the study's conclusions that the changes observed
between the pre- and post-tests occur as a result of the intervention, with a plausible alternative explanation
that the changes might be due to a spurious variable (see Chapter 5).

One of the problems with a single group design lies in the impossibility of determining what might have
happened if the intervention had not been made. This is called a counterfactual condition. These
other changes that might be confounding the study are of two different types: those that occur within the
environment and those that occur within the participants.

Events that arise within the environment are referred to as history. In the absenteeism example it is
possible that the workers' attendance was influenced by a fear of losing their jobs due to increasing
unemployment, changes in weather conditions which made it easier to get to work, or the end to a flu
outbreak in the area. Any of these reasons could be used to explain the change in the absentee rate and the
manager therefore has still not demonstrated conclusively that the device works.

Events that arise within the participants are referred to as maturation. It is also possible that the
workers' attitude to their job has changed and, as a result, they are staying away from work less often.
Again, this confounds any change that might have occurred due to the installation of the new monitoring
device and the manager cannot show that it actually works.

In cases where participants must be tested (such as tests of memory or driving skill) a bias called the test
effect may arise. Participants may become bored with the test procedure, or they may become practised
and thus improve, or they may become fatigued. In other words, this bias results from a change in
participants' responses to the test instrument with repeated usage.

Instrumentation refers to changes in the measuring device or procedure over the course of the study.
So, for example, having used one person to conduct the assessment interview at pre-test, and another
person to do it at post-test, might account for changes in post-test scores. If tests are subjectively scored,
this may account for variations in scores that are therefore not due to the intervention.

When participants drop out of a study, the scores that remain at post-test may not be representative of the
original sample that was recruited for the study. This is referred to as attrition or mortality.

A final source of error in the pre-test/post-test design is that of regression towards the mean. All
variables fluctuate somewhat due to chance. It is possible that the initial measurement (pre-test) was not an
accurate reflection of the dependent variable. As a result, the second measurement (post-test) is likely to be
different but closer to the actual level of the dependent variable, regardless of the effect of the event. Using
the previous example, it may have happened that on the day that the initial measurement was taken, an
unusually high number of people were absent from work. As a result, it is likely that the second
measurement will be lower, even if the new monitoring device has no effect.
 Intact group comparison design

Instead of using only a single group, the intact group comparison design uses two groups, one of which is
affected by the event or treatment, while the other is not. Instead of comparing the performance of one
group before and after the event or treatment, this design compares the scores of two groups, only one of
which was affected by the event or treatment. Since both performances are measured at the same time, the
intact group comparison design overcomes the difficulties of history, maturation and regression towards
the mean.

Quasi-experimental designs and Quasi-experimental rigour than is possible designs in allow pre-
experimental the researcher designs, to maintain even a higher when the level demands of control of
experimental designs cannot be met.

One of the hallmarks of a true experimental design is random assignment of participants to treatments.
Random assignment allows the researcher to control many of the threats to internal validity. A true
experimental design always includes random assignment of participants to conditions, manipulation of the
independent variable and comparisons between groups. However, this may be too expensive, too difficult
to conduct or even unethical, and is therefore seldom possible to achieve! It may be simpler to evaluate
naturally occurring differences in treatment settings. Thus, quasi-experimental designs may be used to
provide preliminary support for verifying potentially important treatments. The major identifying
characteristic of a quasi-experimental design is the lack of random assignment of respondents to
conditions.

Quasi-experimental designs

Quasi-experimental designs allow the researcher to maintain a higher level of control and rigour than is
possible in pre-experimental designs, even when the demands of experimental designs cannot be met.

One of the hallmarks of a true experimental design is random assignment of participants to treatments.
Random assignment allows the researcher to control many of the threats to internal validity. A true
experimental design always includes random assignment of participants to conditions, manipulation of the
independent variable and comparisons between groups. However, this may be too expensive, too difficult
to conduct or even unethical, and is therefore seldom possible to achieve! It may be simpler to evaluate
naturally occurring differences in treatment settings. Thus, quasi-experimental designs may be used to
provide preliminary support for verifying potentially important treatments. The major identifying
characteristic of a quasi-experimental design is the lack of random assignment of respondents to
conditions.
 Contrasted group design

One solution (but not the best) to the problems of the intact groups design is to use groups that clearly
contrast. In other words, the researcher's goal has changed from finding similar groups to finding groups
that are essentially dissimilar or contrasting in their main characteristic. If the researcher knows that groups
are only different in terms of one aspect (the independent variable) and records a difference between the
groups in terms of another aspect (the dependent variable), then she can conclude that the differences in the
dependent variable are related to the differences between the two groups. Note that the important
difference between this design and the others discussed so far is that this design does not allow for an
independent treatment, but is based on differences that already exists between the two groups.

Cohort designs

Cohorts are likely to be similar to each other since their environment is the same except for the treatment
variable. For example, the Grade 6 class at a school in 2005 is likely to have similar demographic
characteristics to the Grade 6 class at the same school in 2006. Cohort designs are strengthened when the
researcher can argue that the two cohorts share similar environments, except for the interventions that they
receive.

This design is similar to the one-shot case study design except that the post-test cohort tests age groups
after the passing of time. These testing occur at different points in time because the cohorts follow each
other through the system such as a school system, a university system or workers undergoing a training
programme.

Pre-test/post-test cohort design

A more sophisticated cohort design is the pre-test/post-test cohort design. The main advantage of
this design is that the researcher is more confident that the groups were similar at pre-test, that is, prior to
treatment.

Time-series design
Time-series designs represent an improvement over the one group pre-test/post-test design of the pre-
experimental group. Rather than relying on a single measurement before and after the event or treatment,
several measurements all before (pre 1, pre 2 and pre 3) and after (post 1, post 2 and post 3) the event or
treatment are made. This allows one to observe the effects of history and of maturation, test effects and
regression towards the mean. This is achieved by comparing each measurement with the measurements
that were taken before and after it. Differences between those measurements taken before the event (for
example, pre 3-pre 2) and between those taken after the event (for example, post 2-post I), but not between
the before and after measurements, must be due to such variables as regression towards the mean, history,
maturation and test effect. The biases can be taken into account when differences due to the treatment are
examined and interpreted. The most important difference is thus the one between the before and after
measurements (that is, post 1-pre 3) since this must be due to the event or treatment. Putting all this
information together allows the researcher to draw conclusions about the effect of the event or treatment,
taking into account the effect of the various confounding variables.

The limitation of this design lies in the difficulty of obtaining a series of repeated measures. It is often
difficult to test the same group of people six to eight times. Further, over time some group members may
no longer be available (attrition), which changes the composition of the group. This bias is called
experimental mortality and is discussed in greater detail later in this chapter.

Experimental designs

Most experimental designs that meet the objectives of explanatory research are characterised by their use
of randomisation to create two or more equivalent groups. The use of randomisation negates the
difficulties of ensuring that the groups are identical. (The symbol 'R' is used in the following diagrams to
identify groups that have been generated using random procedures.)

Randomisation

Randomisation requires that every participant involved in the study has an equal chance of being
assigned to any of the groups of the study. This can be achieved by first identifying the entire group of
participants, then randomly dividing this group into two or more subgroups (depending on the chosen
design) through the use of random number tables, coin flipping or various other randomisation techniques.

It is important that some systematic random technique of determining which subject falls into which group
is utilised. Arbitrary assignment to groups is not necessarily random. (Techniques for generating random
numbers are introduced in Chapter 11.) It is important here to distinguish between random selection and
random assignment. Random selection means that each member of the population has an equal
chance of being selected into the sample. Random assignment means that each member of the
sample has an equal chance of being assigned to any of the conditions or groups in the study. The terms
'random assignment' and 'randomisation' mean the same thing.

When two groups are created using randomisation they are most likely to be equivalent. This is particularly
true when the researcher is working with large groups. The smaller the groups, the less sure the researcher
can be that they are actually equivalent. The advantage of randomly generated groups is that the researcher
starts the experiment with two (or more) equivalent groups. If only one group is subjected to the treatment,
the researcher can be reasonably sure that any difference between the groups thereafter is due to the effects
of the treatment and nothing else. The group that does not receive the treatment is called the control
group, while the group that receives the treatment is called the experimental group. The purpose of
a control group is to compare treated participants with non-treated participants. In this way the effect of
treatment versus no treatment can be examined.

Placebo control group

The placebo control group is another type of control group. Respondents or participants in a placebo
control group are led to believe that they are receiving a viable treatment, although in reality what is given
to them is supposed to be ineffective.

So, in a study of the effectiveness of counselling to reduce anxiety among cancer patients receiving
surgery, the placebo control group might participate in a discussion group with no active group
counselling. The reason for doing this is that it enables the separation of the specific effects of a treatment
from the effects due to client expectations, receiving attention from an experimenter or a group leader or
therapist, and other non-specific factors.

The Hawthorne effect is relevant here in accounting for the differences between preand post-test
scores. This effect was discovered during a study of worker productivity in the United States. Workers in a
factory were assigned to either a treatment group or a no treatment group. The treatment group received an
intervention designed to increase productivity, such as better lighting and various other things. The control
group received nothing, but was simply studied by the researchers. They received only the attention of the
researchers, and without receiving any active intervention, the participants in the control group also
increased their productivity. This change in post-test score for the control group was explained as the effect
of receiving attention from the researcher. It should be noted that the data obtained in the original
Hawthorne study has been criticised as being flawed. Nonetheless, it is generally accepted in research that
just by receiving attention from a researcher, participants' post-test scores might change.
 Matching

Of course, it is impossible to randomly assign participants to groups when the independent variable cannot
be manipulated. An obvious example of a variable that cannot be manipulated is sex. Almost all people are
either male or female and the researcher can do nothing about it. The same is true of religious affiliation,
cultural background, income, age, place of residence and many other characteristics. When one of these
factors is the independent variable, the researcher must use other techniques to establish equivalent groups.
One procedure is matching the elements of the group. In this case, the control for the equivalence of the
two groups is based on having knowledge of the main characteristics of the elements, persons, events or
objects to be investigated. The researcher forms pairs of members having identical characteristics
considered relevant to the research. The members of each pair must differ in terms of the independent
variable only. For the example of absenteeism, the workers could be matched on such properties as age,
sex, skill, health, family situation and work experience. Each member of each pair is then randomly
assigned to a different group. In this way, the two groups so constituted will have equivalent properties and
they are called dependant groups (see Chapter 11 for dependent samples and Chapter 14 for statistical
tests).

Another method of matching, which has the advantage of being more efficient especially in the case of
large groups, is aimed at forming groups that have the same global characteristics. For instance, the two
groups must have the same number of males and females, the same average height, and the same number
of skilled workers. Of course, the matching is not as precise as in the first case, but it is essential that the
matching is done for all relevant factors. Where important factors are neglected in the matching process,
the groups are no longer equivalent for the purposes of the research, and the results are compromised.

The choice between randomisation and matching often does not exist. Matching procedures can only be
used if enough variables relevant to the research are known. If, for instance, age is an important factor, but
the experimenter is not provided with the ages of participants, no matching can be done on the basis of age.
Moreover, if the number of variables to control is large, it will become increasingly difficult to find enough
individuals with the same combination of characteristics. Furthermore, in very large groups, the matching
procedure becomes very time-consuming, tedious and costly.

It is only through the creation of equivalent groups by randomisation or matching that the following
experimental designs are possible.

Pre-test/post-test control group design

The pre-test/post-test control group design requires participants to be randomly assigned to two groups,
one of which becomes the experimental group while the other is the control group. Note that a control
group is similar to the comparison group of the intact group's design, except that it is arrived at through
random assignment. Both groups are measured at the beginning of the study. Thereafter, the experimental
group is subjected to the event or treatment.

Because of randomisation, the two groups can be expected to be equivalent at pre-test. However, it is
possible that they will differ, in which case the difference in the pretests is taken into account when
comparing the post-test results. Note that this design also allows the researcher to measure the effects of
history, maturation and regression towards the mean at the same time as providing a powerful method for
impact studies or summative evaluations. It is the design of choice where the possibility of random
assignment exists.

Weaknesses of the design

It is possible that giving a test twice - a pre-test and post-test - might sensitise participants to the material,
and this may make post-test scores different to what they might have been without a pre-test. However, if
the two groups are truly randomly assigned, then this effect will be the same for the control group and so
therefore no real threat to internal validity.

On the other hand, it may be a threat to external validity (see more detail at the end of this chapter and in
Chapter 13). The treatment may or may not have had an effect on post-test scores but the fact that
participants had a pre-test might have affected the way in which they responded to the post-test.

It is also possible that treatment in itself of might the not pre-test have and an impact the treatment on post-
test that scores, results but in substantially there may be higher an interactive score on effect the post-test.
The effect of this might be that the researcher believes that a programme is excellent and therefore
promotes it. However, when the practitioners implement the workshop without the pre-test, they find a
much weaker treatment effect than they were expecting. This is a serious problem, and it is what is meant
by a threat to external validity.

Post-test-only control group design

An experimental design that has virtually all the experimental rigour of the pre-test/ post-test control group
design is the post-test-only control group design. Randomisation aims at ensuring that the experimental
and control groups are identical except for the fact that only the experimental group receives the treatment
or event. Therefore the pre-test/post-test comparison incorporated into the previous design may be
superfluous. The simpler design represented by Figure 10.8 does away with the pre-testing.
Since all the data is collected at one time, the problems of maturation, history, test effects and regression
towards the mean do not arise. Bear in mind that it is impossible for the researcher to be certain that the
two groups are indeed equivalent to start with, but when a random assignment technique is used and there
are enough participants in the study (at least 30 in each group), the researcher can safely use the post-test-
only control group study.

Factorial designs

Factorial designs can be thought of as generalisations of either the pre-test/post-test control group design or
the post-test-only control group design. The important difference, however, is that factorial designs can
incorporate two or more independent variables, whereas the previous designs only allow for a single
independent variable. A factorial design with two independent variables, each having two levels, will be
studied here. Call the variables x and y, and their levels x, x2, y, and y ,.

Developing a research design

Very often research is hampered by constraints of resources, participants and time. Rarely can any of the
designs discussed in this chapter be imposed directly onto actual research projects. Rather, these designs
explain the logic of social research and should serve as a foundation for good research design. The
researcher's work is complicated by many sources of bias and error which must each be dealt with as
effectively as possible in order to ensure the highest quality of research.

In developing a design for a particular research problem, there are several questions to be considered.

1. Is a cross-sectional or longitudinal design more appropriate? Studies involving change over time
virtually always use longitudinal designs. Cross-sectional designs are most useful for describing
populations, and differences between populations, at a particular moment in time.

2. If a longitudinal design is used, how many repeated observations will be made? How will the
problems of history, maturation, test effect and regression to the mean be addressed? Is there any
danger of experimental mortality affecting the usefulness of the design?

3. Is it possible to create equivalent groups either through randomisation or matching? If this is

possible, then experimental designs are the most appropriate. If not, the researcher is forced to use
a quasi-experimental design.
It is important for social scientists to be very critical of their own and their colleagues' research designs.
The following section explores the critical evaluation of research designs by summarising the common
sources of bias in designs, as well as the criteria against which designs should be evaluated.

Summary of sources of bias in research design

Because every research project has its own particular set of problems, a complete list of the possible
sources of bias in research does not exist. The quality of both quantitative and qualitative research can be
affected to different extents and in different ways by these factors. There are, of course, many other
sources of bias that relate more specifically to methods of data collection, subject selection, data
presentation and analysis. These are discussed in other chapters and only those relating to research
planning and design are dealt with here.

History and maturation

History refers to changes which occur in the world, other than those intended by the researcher, and which
might affect the results. Maturation refers to changes that occur within participants and thus confound the
researcher's design. These effects are particularly problematic to designs that require that data be collected
at more than one time. One common solution is to collect data at a single point in time only. Post-test-only
designs are advantageous in this respect. A second solution is pre-test/post-test control group and time-
series designs, where repeated measurements allow the researcher to estimate the effects of history and
maturation. These effects are then taken into account in the interpretation of the results. Clearly, when
qualitative data are collected, the same issues and solutions are valid.

Regression towards the mean

Regression towards the mean arises when researchers base their conclusions upon single measurements. If
a pre-test score is unusually high due to chance factors, the post-test score is likely to be lower, regardless
of any changes due to the experimental variables. The most effective solution to this problem is the
repeated testing adopted by time-series designs. In the case of qualitative research, the freedom of selecting
the participants as the research develops, gives the opportunity to compensate for this tendency.

Test effect

In quantitative research, prior exposure to a test or measurement technique can bias a person's responses.
This is particularly problematic when participants are tested and retested, as in all pre-test/post-test and
time-series designs. Particular test effects are boredom (when exactly the same test is repeated), practice
(as participants learn how to respond to the test) and fatigue (particularly evident when the test procedure
is lengthy). To counter these effects, the researcher should reduce the number of times that participants are
tested, vary the test slightly so as to reduce boredom and practice effects, and use shorter tests to reduce
fatigue. Alternative instruments are technically very difficult to construct and are also draining on
resources. The same is true for qualitative research.

Instrumentation

Some of the many difficulties of developing appropriate instruments are discussed in a later chapter.
However, instruments present a problem for design as well, particularly when different instruments are
used to test the same concept. The researcher must be sure that the instruments are equally sensitive and
accurate, otherwise changes between the two measurements might be due to differences in the instruments
and not due to any differences within the participants. Also, instruments constructed for use in developing
countries should be sensitive to people's lack of experience with advanced technology, as well as to
differences in culture. Where possible, instruments should be pre-tested with a pilot study in order to check
their appropriateness, accuracy and, where more than one instrument is used, their equivalence.

Sometimes, when measurement instruments are used in inappropriate ways, very powerful reactive effects
can be produced. For example, it is not uncommon for potential employees undergoing a selection
procedure to be subjected to instruments that were designed for the measurement of mental illness. In this
case, potential employees will be extremely sensitive to being turned down for reasons of mental health,
and are unlikely to respond truthfully to the instrument. This defeats the purpose of the selection
assessment. Although data collection in qualitative studies rely more on natural methods, such as
interviews and focus groups, the way in which they are conducted should be in accordance with cultural
and other social and environmental factors.

Experimental mortality

Often participants drop out of a research project during the data-collection procedures.

This is particularly true of research that takes place over an extended period of time. Time-series designs in
particular are vulnerable to this source of error. Although experimental mortality produces difficulties for
data analysis, the most important problem relates to the design of the research. It is possible, if not likely,
that the people who drop out of a study are systematically different from those who remain with the study
until the end. This may result in biased findings. The researcher should design the research so that it is
convenient for people to participate until the end, and should impress upon the participants the importance
of their continuing cooperation.
 Reactive effects

When participants are aware of being observed, they behave in unnatural ways, that is, they 'react' to being
observed. One example of a reactive effect is test anxiety. The measuring instrument may increase the
arousal levels of some participants and thus influence their scores. Similarly, some participants try to
please the experimenter and provide those results that they believe are desired. Others will do just the
opposite and try to confound the study in order to find out how the researcher will react.

A well-known reactive effect is the Hawthorne effect. The Hawthorne effect occurs when participants of a
study feel that their selection makes them a privileged group and, as a result, they perform differently to
how they would otherwise.

The most effective way of countering this source of bias is to use unobtrusive techniques of data
collection. In other words, if the participants are not aware that they are being observed, there is no reason
to expect them to act unusually. Of course, this is not always possible and then the researcher should
attempt to collect data in a way that causes the least disturbance to participants' lives. Practically, this may
mean collecting data in participants' usual environments, using techniques that do not require special skills
or unusual apparatus.

Selection bias

Since many studies incorporate more than one group of participants, it is important that the quantitative
researcher be sure that these groups are equivalent to each other in all respects except for the independent
variable of the study. When the independent variable is beyond the researcher's control and a quasi-
experimental design has to be used, there is little that can be done. When the researcher is in control of
which participants fall into which group, there are two techniques that can be used to ensure that the
groups are similar. The best technique for a large group of participants is random assignment, which was
discussed earlier in this chapter. This entails assigning the participants to different groups on the basis of a
randomisation procedure. When the researcher has only a few participants and has reason to believe that
randomisation might not result in equivalent groups, the option of matching is available. This issue affects
the qualitative researcher far less, as she can select participants in the course of her research and therefore
has much more flexibility in order to avoid this bias.

Note that questions of history, maturation, test effects, reactive effects, instrumentation and attrition apply
for both quantitative and qualitative research. Regression towards the mean is not a problem that
qualitative researchers need to think about.
 Relationship between internal andexternal validity

As explained at the beginning of this chapter, good research design leads to high internal validity,
where the researcher is able to answer the research question convincingly. The potential of a design to
achieve this aim is referred to as the validity of the design, and it is measured in terms of two separate but
related dimensions: internal and external validity.

External validity is concerned with the question, 'To what extent do the results obtained in this study
apply to the population being studied and to contexts different to those of this specific study?' In qualitative
research, external validity is often referred to as transferability. External validity examines the extent to
which the results of the study can be generalised. The researcher must consider two factors in order to
achieve high external validity. Firstly, the sample must reflect the experiences of the population as fully as
possible: in quantitative research the demand is for a representative sample, while qualitative researchers
use a range of other methods to ensure that their study includes the full range and depth of respondents'
experiences. These topics are discussed in greater detail in Chapter 11.

Secondly, the researcher must ensure that the study simulates reality as closely as possible. The conditions
and situation must be seen as normal, depicting the usual reality of the participants. This means that the
tests and tasks that are required of the participants must be planned so as to minimise the whole range of
reactive effects. When people behave differently due to their participation in a research project, the
findings are immediately less valid than they would have been had the participants behaved as they would
every other day of their lives. Techniques for ensuring less reactivity include making data collection as
unobtrusive as possible and testing people within their usual surroundings.

Unfortunately, internal and external validity tend to be inversely related. That is, studies with high internal
validity often have low external validity and vice versa. Studies that take place in a particular social
context have high external validity but cannot control the wide variety of real-world variables that could
interfere with the effect of the independent variable on the dependent variable. Thus these studies have low
Internal validity. Similarly, studies that take place within more controlled environments have high internal
validity, but they may be so far removed from everyday reality that their external validity is low. Very
seldom does a design achieve high levels of both Internal and external validity. In most cases social
scientists tend to design studies that have either high internal or external validity and then compare their
study with other studies in the same field that have high validity on the other dimension.

Sometimes, in intervention research, the distinction is made between efficacy and effectiveness research.
Efficacy studies test the success of one treatment over another under tightly controlled lab-like
conditions. Thus, in an efficacy study, the researcher tries to control and isolate the variable of interest.
One is only interested in whether the treatment works under ideal conditions. The researcher might provide
transport and childcare for participants, and even compensate them to participate in the study. This will
ensure compliance with the treatment and reduce attrition rates. Studies of this nature have very good
internal validity, but less external validity. In the real world it is often not possible to give people transport,
childcare and pay them to come for treatment. In fact, it is supposed to be the other way round. In the real
world we conduct effectiveness studies, to see if there is actually uptake of the service and what its
effect is if people have competing demands on their time. Do they still come for treatment, do they come
regularly, on time, for the full duration of the treatment course, or not? Under these circumstances, is the
treatment effective? This is the main difference between efficacy studies and effectiveness studies.

CHAPTER 11- SAMPLING

The purpose and types of sampling

How does one judge if the customer service of Bank A is better than that of Bank B, that Shop C is better
equipped than its competitor D, or that a shirt of Make X is of a higher quality than a shirt of Make Y?
Judgement is usually based on experience, either personal or reported by others. One might observe how
many times and for how long customers have to wait in a bank before being attended to, or ask the bank
customers to express their opinion about the personnel working in the bank. Equally one could investigate
how often an item is found in Shop D when it was not available in Shop C, and so on. On the basis of few
observations one can generalise, or infer, properties of the whole and draw conclusions.

Sampling theory is the scientific foundation of this everyday practice. It is a technical accounting device to
rationalise the collection of information, to choose an appropriate way in which to restrict the set of
objects, persons or events from which the actual information will be drawn.

Without doubt, if one wants to collect accurate information about a group of persons or objects, the best
strategy is to examine every single member or element of the group.

But it is also possible to reach accurate conclusions by examining only a portion of the total group. This
assessment of only a portion of the group is commonly used in both the social and natural sciences.
Medical assessments of blood characteristics of a patient are done after analysis of just a few drops of
blood. Similarly, the interviewing of just a few students can allow an inference about the attitudes of the
whole class towards a particular lecturer. Admittedly, many factors might reduce the accuracy of the
results based on a small section of the whole group, but these are not of immediate concern and will be
dealt with later.

The entire set of objects or people that is the focus of a research project and about which the researcher
wants to determine some characteristics is called the population. It could be a population of all the cars
assembled at a particular factory during 2005, all the houses in a town, or all the primary school teachers in
the country. The subset of the whole population, which is actually investigated by a researcher and in the
case of quantitative research whose characteristics will be generalised to the entire population, is called the
sample. A sample might consist of every tenth car produced in that factory, every fiftieth house in that
town or 100 primary school teachers selected from a list of trade union members. In each sample, a car, a
house or one primary school teacher constitutes an element or unit of analysis. Specific values or
quantities that relate to the population as a whole, such as the average age of all primary school teachers,
are called population parameters. When the corresponding values or quantities are drawn from the
sample, such as the average age of the 100 teachers constituting the sample, they are called sample
statistics or simply statistics (see Chapter 14). Statistics are therefore estimates of population
parameters.

Sampling theory
As explained in Chapter 5, quantitative and qualitative studies do not have the same aim, nor do they use
the same methods. Most often, quantitative research aims at testing a hypothesis on a representative
sample, so that the results can be generalised to the whole population. It is thus deductive and tends to
simplify the complexity of reality. On the other hand, qualitative research aims at acquiring a deeper
insight into a complex phenomenon that might be quite specific and unique, and which appears in different
ways in the various units of the population. Qualitative research is concerned with studying people in their
natural context. It is thus inductive and holistic. It does not aim at generalising the results to the population
but aspires towards some aspects of the new knowledge and understanding being transferrable to other
units of the population.

Quantitative sampling theory is the study of the relationship between a population and the samples drawn
from it. Since the aim of research is to determine some characteristics of a certain population, one of the
objectives of sampling is to draw inferences about the unknown population parameters from the known
sample statistics. These are obtained by collecting data from the sample. The process of generalising from
findings based on the sample to the population is called statistical inference. Sampling means
abandoning certainty in favour of probability. Because a large part of the population has not been
investigated, statements made about the population on the basis of what has been found to be true for the
sample are of necessity probability statements. One is aware of introducing possible error by asserting that
the units that had not actually been studied would produce the same results as those studied. Here
probability is understood as the likelihood of an event happening, an element being found or a statement
being true. An example of a probability statement is: If the average age of a sample of primary school
teachers is 34 years, the average age of the whole population of teachers probably lies between 32 and 36
years. In many instances, if the sampling or selection of a sample is correctly and carefully carried out, the
margin of error can be accurately calculated.

The main advantages of sampling, as compared to the collection of data on the whole population, are the
following:

1. Gathering data on a sample is less time consuming. If a student population numbers 4 000 it would
take a researcher at least 4 000 hours to conduct a one-hour interview with each student. It would
take only 200 hours, or 5% of the time, to interview a sample of 200 students.

2. Gathering data on a sample is less costly since the costs of research are proportional to the number
of hours spent on data collection. Moreover, a large population may be spread over a large
geographical area, involving large travel expenses. Such expenses are likely to be decreased by
reducing the number of respondents to be studied. Other expenses, such as the cost of reproducing
data collection instruments like questionnaires, will also be reduced.

3. Sometimes, sampling may be the only practical method of data collection. This Is the case in
studies where the property under investigation necessitates the destruction of the object. When
 testing the resistance of an object to wear and tear, such as testing the lifespan of an electric light
bulb, one must wait till the bulb is 'dead'. It would be inadvisable to test an entire population, for
example, the annual output of a light bulb plant, as the plant would have no bulbs left to sell. This
is called destructive sampling and in this case testing can only be realistically conducted on a
sample.

4. Sampling is a practical way of collecting data when the population is infinite or extremely large,
thus making a study of all its elements impossible.

Main sampling concepts

Good sampling must satisfy different criteria for quantitative and qualitative research.

1. Quantitative research. A sample is considered to be adequate if it enables the obtained

results to be generalised to the whole population. It is also based on an estimate of how
representative the sample is of the whole population, that is, how well, in terms of probability, the
sample statistics conform to the unknown population parameters. This last topic will be introduced
here but elaborated upon in Chapter 14. The sample is selected before starting data collection and
analysis.

2. Qualitative research. A sample is considered to be adequate if it allows all possibilities or

aspects of the researched phenomenon to be identified. In other words, when the researcher
reaches the conclusion that collecting more data and increasing the sample does not bring any new
information, data saturation has been achieved. It is based on an estimate that further expansion of
the sample will not bring any further element to the research. Because of this approach, the sample
is partly selected in an ongoing way during the process of data collection and analysis.

In both cases, one has to have a well-defined population and an adequate sample.

A well-defined population

As mentioned above, a population, sometimes referred to as a 'target population', is the set of elements that
the research focuses upon. In the case of quantitative research, the results obtained by testing the sample
should be generalised to the entire population. In the case of qualitative research, the results obtained are
exhaustive to all researched aspects found in the population. It is essential to describe the target population
accurately. This can be done most effectively by clearly defining the properties to be analysed. In
quantitative research it is done by providing an operational definition, thus allowing measurement (see
Chapter 6). Once this is done it should be possible to compile a list of all elements of the population under
investigation or at least to determine whether or not an element belongs to that population. In this way the
population will be well defined.

Once an operational definition is given, boundary conditions can be established which

makeiteasytoascertain whether or not an element belongs to that population. A sample can then be easily
selected from the population.

Although social scientists are often involved in constructing samples from populations of human beings,
there are many other possibilities as well. Consider the economist who is interested in economic indicators
over an extended period of time. She will sample from a population of dates. A historian interested in a
particular person's life story might use a sample drawn from the population of all of that person's written
correspondence. A geographer might work with a sample of water catchment areas while a sociologist
might look at the attendance at a sample of different churches. Whatever the unit that comprises a
particular population, the same basic sampling methods apply.

The sample

Because of the very different aims and methods of quantitative and qualitative research, the discussion of
samples and sampling for each is presented separately. tavbsalt

Quantitative sampling: Although a sample is, by definition, a subset of the population, to be useful it must
have all the properties of the population so as to be representative of the whole. To follow up the example
of the students, selecting a sample of only very dull or very brilliant students would be wrong, because this
group of students would not represent the whole of the student body. Thus one of the major issues in
sampling is to determine samples that best represent a population so as to allow for an accurate
generalisation of results. Such a group is called a representative sample.

The first means of ensuring a representative sample is the use of a complete and correct sampling frame.
This is the list of all units from which the sample is to be drawn. To use a telephone directory for selecting
a sample would be wrong as it would include only people owning a phone. Such a sample could never be
representative of an urban population in a developing country. It might at best represent high-income
urbanites, but usually not even that. An inadequate sampling frame that discards parts of the target
population has been the cause of many poor research results. Particular care must be taken to avoid this
pitfall.

It must therefore be stressed that an adequate sampling frame should not exclude any element of the
population under investigation. An even stricter requirement would be that all elements of the population
should have the same chance of being drawn into the sample, or at least that the probability of this
happening can be specified. In consequence, sampling theory distinguishes between probability or random
sampling, and non-probability sampling.

Probability or random sampling is when the probability of including each element of the
population can be determined. It is thus possible to estimate the extent to which the findings based on the
sample are likely to differ from what would have been found by studying the whole population. In other
words, the researcher can estimate the accuracy of the generalisation from sample to population. A sample
is randomly selected when each member of the population has a known chance of being selected for the
sample.

Non-probability sampling is when the probability of including each element of the population in a
sample is unknown. In this case it is not possible to determine the likelihood of the inclusion of all
representative elements of the population into the sample. Some elements might even have no chance of
being included. It is thus difficult to estimate how well the sample represents the population and this makes
generalisation highly questionable.

Probability samples are of a much higher quality because, when properly constructed, they are
representative of the population. Although it is difficult to determine the extent to which non-probability
samples are representative of the population, they have some practical advantages. When the necessary
population lists are not available, non-probability sampling remains the only possibility for the researcher.
Also, non-probability sampling is almost always cheaper, faster and often adequate for homogenous
populations. Finally, it should be noted that to some extent, the disadvantages of non-probability sampling
could be reduced by enlarging the sample. Non-probability sampling is thus frequently used in the social
sciences.

In the following section, the most common sampling procedures will be examined.

These are:

1.Probability sampling, which includes:

 simple random sampling

 interval or systematic sampling
 stratified random sampling
 multistage probability sampling.

2.non-probability sampling, which includes:

 convenience or availability sampling

 purposive or judgemental sampling
 quota sampling
 cluster sampling.

Probability sampling for quantitative research

Simple random sampling

Firstly, it is important to clarify what is meant by 'random' since, in everyday language, it is often used to
mean 'accidental'. It is not random to choose the first ten students who enter a lecture theatre, or students
sitting in the front row. Neither is it random to choose ten students with average test results. In all these
cases some underlying criteria were used to select the samples, which might have properties peculiar to
them. For example, the first students arriving for the lecture may be the most dedicated ones. The ones
sitting in the front row may have poor sight or may want to attract the attention of the lecturer. Even the
sample of 'average' students does not represent the heterogeneity of the student population. Random, in the
scientific sense, expresses the idea of chance being the only criterion for selection. Thus, the selection of
an element from a population is called random when the chance, likelihood or probability of being
included in the sample can be calculated for each element of the population.

Accordingly, simple random sampling is a sampling procedure that provides equal opportunity of selection
for each element in a population. There are various techniques for selecting randomly. The most common
are the lottery techniques where a symbol for each unit of the population is placed in a container, mixed
well and then the 'lucky numbers' drawn that constitute the sample. The symbol for each unit of the
population could be the names of participants written on identical pieces of paper, or a number assigned to
each participant. Of course, sample size must be established beforehand. otherwise one would not know
how many 'symbols' to draw.

A more sophisticated method, particularly useful for large populations, is the use of random number
tables or electronic random number generators.

Interval or systematic sampling

This type of sampling is very similar to the previous one. This technique, instead of relying on random
numbers, is based on the selection of elements at equal intervals, starting with randomly selected element
on the population list.

Unfortunately this method also has constraints. Like simple random sampling it relies on the availability of
a complete, unbiased population list. Moreover, this list must not have any cyclical or periodic
characteristics. This can be clarified using the previous example. If, at regular intervals along the
predetermined street, there are houses with peculiarities, such as corner houses containing shops, one
should compare the length of this interval with k. If the two intervals coincide, the consequence may be
that too many corner houses with shops, or none at all, are selected for the sample. In both cases, the
sample will not accurately represent the population, but give a distorted image of it. An even more trivial
mistake would be to adopt a six-day observation cycle when comparing the volume of sales in different
supermarkets: Mondays in shop A, Tuesdays in shop B, Saturdays in shop F. Surely, the large difference in
sales between shops A and F can be attributed more to the Saturday shopping habits of people than to
intrinsic properties of the shops?

Thus, when a peculiar regularity is detected in a list and the peculiarity coincides with the size of the
interval k, the list should either be rewritten to avoid this regularity, or another method of sampling should
be adopted.

Stratified random sampling

As mentioned previously, simple random sampling, and even systematic sampling in its purest form, are
seldom used in social sciences research because they are cumbersome for large populations. But they
become useful tools when used as part of other random sampling techniques, such as stratified and
multistage sampling. More about multistage sampling later.

The principle of stratified random sampling is to divide a population into different groups, called strata, so
that each element of the population belongs to one, and only one, stratum. Then, within each stratum
random sampling is performed using either the simple or the interval sampling method. Although many
samplings are performed, each is done for a relatively small population only. This increases the availability
of adequate lists and facilitates selection of a simple random sample without decreasing the quality of the
sample in any way.

To clarify the advantage of stratifying a population, an extreme case is given. Suppose that, using the
previous example, it is suspected that men are strongly opposed to cooperatives while women vigorously
support them. Suppose also that the population is very heterogeneous and that due to urbanisation, men
constitute only 10% of the population. In other words, out of the population of 8 000 people, 800 are men
and 7 200 are women. When selecting a sample of, for instance, 100 out of this population, using simple
random sampling, chance factors may strongly distort the proportions of men and women. For such a small
sample it is possible that only four men (4%) will be selected out of the 800 men, and 96 women out of the
7 200 women. This gross underrepresentation of men will heavily distort overall research results and the
finding might be that the total adult population is strongly in favour of cooperatives, which does not reflect
reality. The distortion is caused by the sampling method, in this instance, simple random sampling. (Note
that only for very large samples can one be assured that the composition of the population will be reflected
in the sample when simple random sampling is used.)
Stratified sampling, on the other hand, by preserving proportions even of very small samples, will allow
for any small minority to be properly represented. Of course, if the population is very homogenous with no
marked differences of opinion between sexes, or if both sexes are fairly equally represented (48% vs 52%,
for instance), simple random and stratified sampling will lead to similar results.

Even stratified random sampling presupposes a definite population with a known composition, and the
correct choice of the criteria for stratification is crucial. However, it offers great accuracy, even for small
samples, and reduces costs of sampling considerably.

Multistage probability sampling

One of the major constraints of simple random, systematic and stratified sampling is the availability of
complete lists of elements or units. Often such complete lists do not exist. For example, each school may
have an up-to-date list of its pupils, but no national list may be available. Similarly, a village headman
might be able to help in compiling a list of all widows or orphans in his village, but such information is
very seldom found in a birth-and-death register at national level. At grass roots level many small income
generating groups that are more or less structured and more or less temporary may have been formed, but
few may be formally registered as cooperatives or as members of a union. Such groups, in the main,
remain informal and are known only within their direct environment. For these cases another type of
sampling must be used. The principle underlying multistage sampling is to start by sampling a population
which is much more general than the final one. In a second stage, on the basis of the first sample, a new
population is considered, one that is less general than the first one, and a new sample is subsequently
determined. The procedure is continued until the population to be investigated is reached and a final
sample is drawn.

While not as accurate as simple random sampling, the advantage of multistage sampling become
apparently when other sampling methods fail due to a lack of complete lists of elements for a population
under investigation. Even if a complete list of populations can be compiled directly, multistage sampling
can cut down expenses by reducing the cost of compiling long lists and by reducing the travel expenditure
necessary when respondents are spread over a large area. In fact, if a complete list of orphans was available
and a simple random sample of 500 was selected, these 500 orphans would be spread all over the country.
This would mean high travel costs and a great deal of travel time. Multistage sampling, by selecting only a
few geographical areas (five districts), reduces travel costs and time, and permits careful planning of the
data-collection process.

Non-probability sampling for quantitative research

Convenience or availability sampling
This sampling method, the most rudimentary one, consists of taking all cases on hand until the sample
reaches the desired size. The interviewer will choose, for instance, a convenient place where he or she is
assured of finding many people: a supermarket, a bus stop or a bar. Obviously, this can introduce serious
biases, as men will be overrepresented in bars, and old and wealthy people will be under-represented at bus
stops. Taxi drivers, waiting at a stand for customers, would welcome a chat with the interviewer but might
not be representative of a population. Students are the usual 'guinea-pigs' of social scientists though still
not representative of the whole population. This makes generalisation based on such samples extremely
risky, although the samples so chosen are convenient for researchers in terms of time and money.

Purposive or judgemental sampling

This sampling method is based on the judgement of a researcher regarding the characteristics of a
representative sample. A sample is chosen on the basis of what the researcher considers to be typical units.
The strategy is to select units that are judged to be the most common in the population under investigation.
For instance, a typical school pupil may be thought of as being '12 years old, male, Catholic and having
parents in a clerical profession'. Only students meeting these criteria will be chosen for the sample. The
great danger in this type of sampling is that it relies more heavily on the subjective considerations of the
researcher than on objective criteria. Although it has some value, especially if used by an expert who
knows the population under study, this technique often leads to non-representative samples.

Quota sampling

This sampling method is the non-probability equivalent of stratified sampling. The purpose here is to draw
a sample that has the same proportions of characteristics as the population. However, the sampling
procedure, instead of relying on random selection, relies on convenience.

Although much less accurate than stratified sampling this method is often more convenient and
economical. No lists need be compiled and all data can be collected at an arbitrary location.

Cluster sampling

Finally, the very useful technique of multistage quantitative sampling has an equivalent in non-probability
sampling, which is called cluster sampling. The process is the same as described above except that the use
of simple random sampling is replaced by a non-probabilistic selection method, such as the availability of
the elements or the convenience of reaching them.
 Other sampling possibilities
Independent versus related/dependent samples

As already mentioned in Chapter 10 when dealing with randomisation of groups, two or more groups or
samples might be selected in such a way as to make them independent. Each unit is drawn randomly from
the population and is also randomly assigned to one or other group. Groups formed in this way constitute
independent samples. Alternatively, groups or samples can be related, usually when their elements have
been matched by specific properties. In such cases, they are called related or dependent samples.

When researchers draw up samples in this way, every member of the sample has a partner somewhere else
in the study and so the samples can no longer be thought of as independent of each other. They are related
or dependent.

Sample size: How large should a sample be?

Students and new researchers involved in quantitative research often ask, 'How large does my sample have
to be?' The best answer is, 'It depends.' It depends on the kind of data analysis the researcher plans, on how
accurate the sample has to be for the researcher's purposes, and on population characteristics.

Generally speaking, a large sample is more representative but very costly. A small sample, on the other
hand, is much less accurate but more convenient. A census, which is a survey of the whole population, will
be more accurate than a survey using a restricted sample, but it will be very expensive and its results may
take so long to analyse that they become outdated.

However, a large sample size alone does not guarantee a representative sample. A large sample without
random sampling or with a poor sampling frame is less representative than a smaller one with random
sampling and an excellent sampling frame.

The size of the sample is an important determinant of the statistical power of the research . Statistical
power refers to the likelihood that inferential statistical tests (see Chapter 14) will be able to test the
research hypotheses adequately. Smaller samples have less statistical power than larger samples. The
major criterion to use when deciding on sample size is the extent to which the sample is representative of
the population. This can be expressed in terms of probability. One usually needs to have a 95% chance that
the sample is distributed in the same way as the population. Formulae exist for determining a sample size
that satisfies such a given level of probability, but they fall beyond the scope of this book. One should,
however, note that the more heterogeneous a population is, the larger the sample must be to represent the
characteristics of the population correctly. The rule of thumb for choosing a sample size is that is should be
at least 5% of the population, but this remains quite an inaccurate guideline, though it is certainly usable
when precise formulae are not available.

A researcher's decision about the best sample size depends on three things:

 1.the degree of accuracy required

 2.the degree of variability or diversity in the population
 3.the number of different variables to be examined simultaneously in the data analysis.

All else being equal, larger samples are necessary if one wants a high degree of accuracy, if the population
has a great deal of variability or heterogeneity, or if one wants to examine many variables in the data
analysis simultaneously. Smaller samples are sufficient when less accuracy is acceptable, when the
population is homogeneous, or when only a few variables are examined at a time. The analysis of data on
subgroups also affects are searcher's decision about sample size. If the researcher wants to analyse
subgroups in the population, he or she needs a larger sample.

Sampling errors and related problems

As has been repeatedly mentioned, the purpose of sampling theory is to select samples, which reproduce as
closely as possible the characteristics of a population. This aim is never completely achieved due to two
types of error.

The first type of sampling error is due to chance factors. It may happen that, in a particular sample,
one element and not another has been included. This type of error is the inevitable result of sampling and
can never be completely eliminated, but the sampling error can be calculated statistically.

The second type of sampling error is due to bias in selection, arising primarily from faulty technique.
These biases are frequently avoidable. They may or may not be deliberate. For example, an interviewer
may fail to take into account one criterion, such as the age of respondents, or the respondents themselves
may give incorrect information about their age. Some strata of a population may be over- or under-
represented in a sample. For example, a sample that contains 50% rural and 50% urban respondents would
under-represent the rural population and over-represent the urban population of virtually all less
industrialised countries today.

Even when a representative sample has been drawn, a very important source of bias is non-response
error. This type of error comes about when an element of the sample does not respond to a measurement
instrument or is not available for some unknown reason. As a consequence such elements are excluded
from the group, which changes the constitution, and thus the representativeness, of the sample. There are
many reasons for non-response. Firstly, it may not be possible to interview or test a person because of
illness, language differences or other factors. Secondly, it may be that the chosen respondents cannot be
found because of changes in residence or name, or because of death. Thirdly, the selected person may be
absent whenever the interviewer calls, purely by chance. Lastly, the person can refuse to collaborate and
not answer questions or give information.

Sampling for qualitative research

Coming back to the aim of qualitative research, to investigate in depth a certain phenomenon, it becomes
clear that probability sampling methods are too rigid and do not allow seldom occurring elements to be
sufficiently represented and studied. Therefore, since qualitative research does not concentrate on the
'average' person (representative of the population at large) but rather on the diversity of cases,
nonprobability sampling is used.

The three types of non-probability sampling introduced under quantitative sampling, convenience or
accidental, purposive or judgement, and quota sampling are the most frequently used. In qualitative
research these techniques do not have the disadvantages mentioned for quantitative research since
generalisation to the target population is not the criterion for sample quality. In addition, one should
mention the very useful snowball sampling, a sub-category of accidental sampling. Some sub-
categories of purposive sampling, such as maximum variations sampling, extreme/deviant case sampling
and critical case sampling, as well as theoretical sampling, will be described below.

Convenience or availability sampling

As described under quantitative sampling, this is a very handy sampling procedure based on the
availability of the units of the target population.

Snowball sampling (also called chain or referral sampling)

A more sophisticated sampling technique, particularly useful for identifying people not listed or difficult to
find, is to identify a few participants and then to rely on each participant to guide you to the next one, etc.
Through this method one can increase the size of the sample but also develop a network, for instance, of
specialists: one astrologiat or palm reading practitioner might introduce the researcher to a colleague,
facilitating, access to that person.
 Qualitative approach to purposive or judgemental sampling

As the name indicates, this type of sampling rests on the assumption that the researcher knows what type
of participant is needed. In quantitative research, the most "representative' element will be looked for. In
qualitative research, on the contrary, the element that is the most complex and rich in information will be
the most valuable. The researcher will thus purposefully choose participants on the basis of some specific
criteria that are judged to be essential. There are many ways of reaching a deeper knowledge and
understanding through the choice of more sophisticated purposive sampling, some of which are described
below.

Extreme or deviant case sampling

This method concentrates on getting information from the most extreme cases, the 'exceptions to the rule'.
Highly unusual manifestations of the researched phenomenon, obviously not representative of the target
population, may sometimes be the most revealing cases. Here the assumption is that comparing these
extreme cases might lead to a better in-depth understanding of the underlying factors of a problem.
Studying learning processes in children, one might sample the strongest and the weakest children in each
class. But one might also be interested in studying the few children who react negatively to the
introduction of an otherwise very successful learning programme. Researching the importance of some
managerial skills, one might choose to study very successful and very poorly performing businesses.

Critical case sampling

In some instances, one can assume that some cases have key position. They are particularly useful and can
assist in deciding on the importance or the prevalence of some explanations. For instance, a teacher will
assess the level of difficulty of an exam paper by stating: 'If student A can pass this test, then all others
can!' Here studying the case of A allows generalising and applying the results to the entire class. This
method has the advantage of reducing the sample to the few cases most likely to give a lot of essential,
relevant information.

Maximum variations sampling

In this case the researcher aspires to having the most heterogeneous sample. For instance, studying causes
or triggers for divorce, the search for a great diversity in the sample will include a wide range of
professions without taking into account their frequency in the total population. Thus uncommon
professions (animal breeders and astronomers) will be given the same importance as common professions
(coal miners and receptionists). In general, maximum variation samples will thus contain extreme or
deviant cases, critical cases and typical cases in which nothing is unusual. The importance here is to have
the widest possible range of cases. This will yield a rich variety of variation for the phenomenon under
study. However, for such samples to be all-inclusive, they will also be relatively large and costly.

Quota sampling

Proportional quota sampling has been described under quantitative sampling. To this method one can add a
less restrictive one, non-proportional quota sampling. The principles are the same, but here the researcher
does not make a concerted effort to reach the numbers or quota corresponding to the proportions in the
population. What is important is to have all categories represented in the sample. One is satisfied if the
information given by the participants of a certain quota has reached saturation.

Combination or mixed purposeful sampling

Often a combination of various sampling methods is chosen to arrive at the most appropriate sample. This
flexibility is particularly helpful when dealing with triangulation (see Chapter 13). Most important, always,
is that the sampling strategy focuses on and fits the aim of the research, the method and instruments used,
as well as the available resources.

Sampling process and sample size: when is enough, enough?

As evidenced in the examples above, an important difference between quantitative and qualitative
sampling is that, in the first case, the sample and the sample size are determined before data collection,
whereas, in the second, once the type of sampling has been decided upon, the units constituting the sample
are identified parallel and concomitant to the data collection. Moreover, in the ideal case, the data
collection will continue as long as new elements or facts are found and will stop when no new details are
available (also called samplingto redundancy). Thus the size of the sample is not determined in
advance but becomes a consequence of the exhaustion of new discovery or data saturation, as defined
earlier. Unfortunately, in reality, time and money sometimes demand some compromise.

The importance of good sampling

The quality of research, whether quantitative or qualitative, is directly related to the sampling procedures,
the adequacy of the technique chosen, as well as the professionalism of implementation and the
appropriacy of the sample size. Even though the aims and procedures for different types of research can be
quite different, in the end the importance of the quality of sampling will reappear when evaluating the
validity and trustworthiness of the findings in Chapter 13.