THE COPPERBELT UNIVERSITY

SCHOOL OF BUSINESS
(LECTURE NOTES)

COURSE CODE: BS 360

COURSE NAME: MARKETING RESEARCH

LECTURER: PROFFESSOR J. TEMBO

ACKNOWLEDGEMENT
These notes are based on and developed from materials in the following texts:

1. Gilbert Churchill and Dawn Lacobucci, Marketing Research-Methodological

Foundations, Ohio: South-Western, 2002
2. Del I. Hawkins and Donald S. Tull, Essentials of Marketing Research, New
York: McMillan, 1994
 UNIT ONE

TOPIC ONE: UNDERSTANDING MARKETING

RESEARCH

Nature, role and function of Marketing Research

The principle function of marketing research is to provide information to assist marketing
managers make better decisions. To help us better appreciate marketing research‟s role in
decision making, let me visit the concept of decision making and context in which a
marketing manager makes decisions.

What is decision making?

Most of our activities involve decision making of one kind or another. We make
decisions, for example, of whether or not we will get out of bed or continue lying in bed;
eat breakfast or skip it; go to work or not; buy a car or something else; when we should
marry; what career in life to pursue and so on. Any decision we make has five (5) salient
features:
i. There must be a decision which a decision maker wishes to attain;
ii. There are alternative courses of action of attaining the objective;
iii. There are payoffs or penalties associated with each course of action;
iv. There is uncertainty about which course of action to take; and
v. The decision is the choice of the optimal course of action.

Marketing decisions are made with some objectives in mind. Many marketing objectives
derive from corporate objectives. A typical marketing objective might be to increase
market share, which might be derived from a corporate objective of growth. Assuming
that our marketing manager‟s objective is to increase market share, he/she must then turn
to the task of identifying the various ways of attaining increasing market share. Possible
alternatives might be to offer his products at a lower price than his competitors, to
improve the quality of his product or to carry out increased advertising of his product. He
then will attach some benefits or costs to each of these alternatives. He is then faced with
the dilemma as to which options to go for. This dilemma is rooted in his ignorance or
uncertainty about the efficacy of any of the alternatives. His selection among the choices
will constitute a decision, and it will be based on some rationalization ad judgment as to
which of the options is optimal. Implicit in decision making is that the decision maker
makes an informed decision given the uncertainty associated with each alternative course
of action.

Nature and role of Marketing Research

The basic mission of marketing management is to satisfy a customer‟s wants and needs.
The problem for the marketing manager is that he does not precisely know what these
wants and needs are and how to satisfy them. To overcome this lack of knowledge or
uncertainty, he/she must develop a system of availing himself information about
consumer wants and needs and how consumers satisfy their needs and wants. Thus all the
information-related tasks to help the marketing manager get to know the needs and wants
of his market are the purview of marketing research. The role of marketing research,
then, is assisting a marketing manager make informed decisions in the marketing of
goods and services. This role can be better appreciated by recasting it in the context of
the task of marketing management by using a framework developed by McCarthy and
Perreault.

The McCarthy and Perreault model, illustrated in figure 1, holds that the customer is at
the centre of marketing activity. The first challenge a marketing manager faces, however,
is that
Figure 1: task of marketing management

Political/ Legal

Price
Competitive/
Customer
Cultural/ Social Place Product Technological
Promotion

Economic

SOURCE: E. J McCarthy and W. Perreault, Basic Marketing: A Managerial

Approach, 8th ed

a customer‟s needs and wants are not precisely known. Logically, if marketing effort is to
be effective, it must be targeted at known needs and wants. Much decision making in
marketing is consequently devoted to the task of identifying customer‟s needs and wants.
The role marketing research is to provide marketing managers with information that
enables them make informed decisions in identifying more accurately the needs and
wants of their customers or prospects. Much research is devoted to determining
customers‟ needs/ wants. Such research has to do with opportunity identification: is there
a need out there in the market which is not satisfied? Is there a better way of doing what
we are doing? What competitive advantage do other firms have? Will our product be
acceptable in the market place?
Another area of uncertainty in need identification relates to consumer preferences. While
needs are basic and easily predictable, wants are much more difficulty to ascertain
because they are conditioned by an individual‟s preference. Preference, in turn, depend
on a number of factors, such as, income, taste, background, social standing etc. how each
of these factors, or a combination of them, influence individual choice is not a matter that
can be easily ascertained. Managers need information that will help understand not only
what buyers need but how they go about exercising their choice. Thus a miller of maize
does not make the kind of decisions that a manager of the restaurant has to make, because
the demand for maize meal relates to satisfying a need which is largely homogenous and
one which is satisfied all too often in the same way. In contrast, a patronage of the
restaurant is relatively more complicated: the habit of eating at the restaurant is not as
widely spread as eating maize meal. In our tradition a Zambians, eating at a restaurant
may be alien and is probably an adopted habit arising out of social standing in an urban
setting or job demands. And even if a Zambian should find his or her way into a
restaurant, what they choose to eat there may not be as easy to predict as what they eat in
their homes. Thus, for a restaurant manager, areas of uncertainty stem from lack of
knowledge about demographic, ethnic, or economic profile of his customers. Customer‟s
individual tastes, price and service levels preferred by customers, location and
appearance of the restaurant, image, and reputation, etc. the point is that preferences are
determined by forces that are not easily discernible to marketers, and hence the need for
research aimed at assisting managers make proper decisions as to who needs what.

The McCarty and Perreault framework further holds that the marketing manager
manipulates elements of the marketing mix, namely Price, Product, Place and Promotion,
in order to put into effect a marketing program. The marketing manager faces a problem
in that he does not know or is uncertain about the impact of manipulating any one of or
all the element of the marketing mix. For example, while general economic theory holds
that price is inversely related to demand, the concept of price inelasticity holds that in
some instances demand may not be responsive to price changes as predicted by the
general economic theory. Indeed, customers react in different ways to price changes. For
the marketing manager, there is consequently some uncertainty surrounding the
effectiveness of manipulating price to induce a desired demand levels. Similarly, there is
no guarantee that programmed product, distribution or promotion strategies will lead to
desired outcomes. Since the manager is faced with uncertainty as regards the
consequences of his manipulation of any, or all elements of the marketing mix, marketing
research plays the role of reducing uncertainty by providing such information as will
enable a manager more accurately predict the effect of his marketing program.

A marketing manager must finally deal with a third area of uncertainty occasioned with
uncontrollable factors obtaining in the environment in which a firm operates, and whose
impact on the firm‟s marketing activities cannot be determined with certainty. These
factors comprise technological, political, economical, competitive, legal, and culture and
social developments. Because these factors are external to the firm, their impact, whether
positive or negative cannot be easily ascertained, and it is therefore the role of marketing
research to provide information which will hopefully make the marketing manager not
only aware of them but also make informed decision regarding their impact on marketing
operations. Consider, for example, how the opportunities and threats created by the
following external events:
 In the pre-liberalization era, many companies operated as monopolies. Because
there was no threat of competition, such companies became somewhat
complacent in the area of product development and enhanced product quality.
With the advent of liberalization, there was need for some re-orientation to cope
with the threat posed by competition. Many local manufacturers and service
providers that had enjoyed protection from competition in pre-liberalization era
has to contend with a customer who had now a wider choice of goods or services
and prudently sought better quality product or price bargains. Companies that
could not meet customer expectation lost ground in market share and profits. The
collapse of many companies in the post liberalization era was due, in large
measure, to their lack of foresight in anticipating the threats to their survival
which loomed into the horizon as a country inched its way towards liberalization
and attendant economic development.
  The enactment of the competition and fair trading act of 1995 further reinforced
the spirit of competition that was implicit in the policy of liberalization. Through
this piece of legislation, competition was legally encouraged, or behavior that
undermined competition was prohibited. Specifically, no firm could enjoy a
dominant position such that it could have power to arbitrary determine price and
control production or distribution; firms were prohibited from such collusion as
would otherwise enable share amongst themselves the domestic market, fix
prices, or reduce production; mergers or acquisition intended to kill, restrict or
distort competition were forbidden and; selective distribution which resulted in
restricting access to a commodity was forbidden. Further, in order to protect and
promote consumer welfare, misleading or deceptive conduct and disclaiming
liability for defective goods were prohibited.
 The rise of consumerism and corporate social responsibility in Zambia demands
that, in addition to profit consideration, firms just their decisions from a moral and
ethical perspective. Increase in the price of energy and utilities are constantly
being subjected to public scrutiny in spite of the economic rationalizations from
the service providers

Churchill and Lacobucci (2002:8) propose that an alternative way of looking at the role
of marketing research is to examine how management might use it. They suggest that
when marketing research is used for planning, it serves to help managers identify
opportunities that can be taken advantaged of and threats which can be avoided; for
instance, in identifying unexploited markets or alerting management to unfavorable
trends in the market place. When it is used for problem solving, it focuses on the short or
long term strategies in the application of the four elements of the marketing mix. The
issues which marketing research can be used to address include: how should a product be
packaged, what price should be packaged for a product, where, and by whom, should the
product be sold: or how much should be spent on advertising and promotion? Finally,
control oriented marketing research assists managers to become aware of dangers and be
informed about their operations. Questions that might be asked include: what is our
market share by region or by customer type? Are customers satisfied with our service?
How is our company, product or employees perceived?

Definition of Marketing Research

The basic role of marketing research is one of reducing the level of uncertainty associated
with marketing decision making by providing pertinent information to the decision
maker. The American Marketing Association has defined marketing research as:

“the function which links the consumer, customer, and the public to the marketer
through information-information which is used to identify and define the
marketing opportunities and problems; generate, refine and evaluate marketing
actions; monitor marketing performance; and improve our understanding of
marketing as a process.” (Churchill et al; 2002:6)

Of note in this definition is the centrality of information and its application to decision
making

INCREASING IMPORTANCCE OF MARKETING RESEARCH

Marketing research owes its existence to the shift from the production-oriented to a
consumption-oriented economy that is sweeping across the globe. As firms realize that
they cannot sell all that they are able to produce, they are beginning to draw on marketing
research to help them determine consumer needs, and prudently match production to
consumption. The more specific trends which increase the importance of marketing
research include the following:

Shift from local to national to international marketing

The need for information gets more pressing as a firm moves fro m a smaller, familiar
and simpler local market to the larger, less known and more complex national, regional,
or international market. This is particularly so when it comes to understanding external
variables beyond one‟s environment. For instance, as firms seek new markets beyond
their national borders, they ordinarily carry out feasibility studies in the foreign markets
to help managers decide o whether or not they should venture into the foreign market.
Such studies might include an assessment of the economic, political, cultural
consideration which might have a bearing on operations. With the onset of globalization,
firms will need marketing research information to increase understanding of the possible
impact of such factors.

Transition from consumer needs to consumer wants

As societies become more affluent, there is a tendency to shift from basic needs to wants.
With the rise in discretionary income, consumers not only fulfill basic needs but
additionally indulge in wants. A rich and poor man must both satisfy hunger pangs, but
the former has a wider choice of food than a poor man. Economic prosperity is mirrored
in more income, which in turn fuels demand for more and diverse consumption.
Marketing research plays a vital role on providing information on choices made by
consumers as they spend their discretionary income.

Shift from price to non price competition

In undeveloped and simple economies, competition is on the basis of price. As markets
develop and competition increases, marketers resort to non price competition such as
product differentiation, advertising and sales promotion in order to stimulate demand.

Changing market structure

In the last few years, there has been, globally, a shift from state driven to market driven
economic systems. In state driven economic systems, major economic policies and
decisions are centrally determined, living little rooms for firms to make production and
marketing decisions. In the case of Zambia, the state-dominated era was characterized by
the prevalence of state-owned firms in all major industries, mining and commercial
endeavors. It was part of the avowed policy of government to pursue of socialist
programme. The state-owned companies were by design the agents of government
economic socialist policy. With the advent of liberalization and privatization of state-
owned enterprises from 1991 onwards, there was a deliberate and market shift towards a
market driven economy premised on forces of supply and demand in the market place,
with little or no government intervention. Firms have now to rely on their own initiative
by understanding the economic and market forces which have a bearing on their survival.
This new era requires firms to make decisions regarding what product line they should
produce, at what price they should sell their products and services, what is the appropriate
level of advertising and promotion, what is the firm‟s market and how this market evolve
in future? Marketing research will play a vital role in providing answers to these
questions.

INHIBITING FACTORS TO GROWTH IN MARKETING RESEARCH

IN ZAMBIA
Even though marketing research is increasingly being recognized as a significant
business activity, its practice varies considerably. Some of the factors which inhibit the
growth of marketing are:

Emergence of international companies with a highly developed

marketing research system in their own countries
Many international companies that operate in developing countries have highly
developed market research systems in their home countries on which they continue to
rely and use as they continue to explore lesser known, foreign markets. Most research is
therefore transferred to, rather than grown in, developing countries. This has a tendency
of reducing research activities in developing countries. Consider, for instance, the fact
that Coca-cola advertisements, other than for the language factor, tend to be standardized
across many nations. This implies advertising research is done centrally, thereby
depriving target markets of conducting their own individualized research.
Marketing concept is still not widely acknowledged and practiced
Consumer discretion is the necessary tenet for the growth of marketing research. But
because most developing countries still face constraints in production, such as poor and
backwards technologies and lack of resources, they continue to experience low levels of
consumption and, consequently, consumer discretion has yet to be realized in full, if at
all, these countries must continue to grapple with the more immediate need of increasing
production. The marketing concept, which holds that the consumer is at the centre of
marketing activity, seems to be hopelessly out of place.

Convergence of tastes and preferences

It is held in most economic circles that the impetus to globalization lies in the increasing
convergence of needs and wants among different nations and cultures. Products and
services conceived in one part of the globe are easily finding market acceptance in other
part of the globe. Discoveries of science and technology have wider and common usage
among peoples of the world. This is particularly so in the case of manufactured goods,
information technology, electronics, tourism, banking, sports, health and education.
There is more common ground and greater understanding among nations than divergent
and isolationism, thus affording everyone more opportunities for access to, and sharing
of, knowledge and information.

The marketing profession is still in its infancy in many developing

countries.
Many developing countries adopted the socialism soon after attaining their political
independence and are therefore recent converts to non-socialist ideology. However, even
in spite of this shift to capitalism, many inhabitants of these countries still experience
high levels of poverty, unemployment and little or no income. The guiding philosophy of
most businesses is still to sell what is produced rather than to produce what to be sold.
Under these circumstances, the challenge of decision making regarding how best to
satisfy customers is diluted to the level of simply selling.
Who uses marketing research?
1. manufacturers are the principal users of marketing research, especially those dealing
in consumer goods

2. service organizations, both profit and non-profit

3. Retailers, especially the large supermarkets/department stores. The type of research

may cover:
 store location research
 Effectiveness of their promotional programs.

4. the government through is agencies such as Central Statistical Office

ORGANIZATION OF MARKETING RESREARCH

Marketing research activities may be organized by way of in-house research, whereby
the firm conducts its own research or the firm may hire an outside organization to
conduct research on its behalf. The choice between the two options depends on a number
of considerations. The first one is the economy. It is cheaper to conduct in-house research
than to contract research to an outside firm. It is uncommon for an outside agency to
provide information more economically. Thus, small firms or those severely constrained
by lack of resources all too often opt to conduct their own research. A second
consideration is available of expertise and/or equipment. If the firm does not have a kind
and number of expertise or equipment required to conduct a particular research project, it
may have to hire outside consultants to do carry out the research. Thirdly, there may be
instances when it is politically incorrect, imprudent or morally objectionable to entrust
access to certain information to insiders or outsiders. A company that is anxious to
protect its secrets is more likely to establish its own units of research and development
than risk a rank of outsider have access to company secrets. At the other end, in-fighting
among insiders may prompt a firm to hire an outsider to do a job that insiders cannot do
in a prescribed time.

As observed earlier, marketing research information is intended to assist managers make

better decisions in the marketing of goods and services. Marketing research activities will
in this regard be located in the marketing department to provide such information as the
marketing manager may require. Depending on the nature, scope and complexity of
research activities, a firm may place research activities under a research manager, who
reports to the marketing manager, as shown in figure 2; alternatively, the research activity
could be a functional responsibility of a Sales, Distribution or Advertising Manager.

Figure 2: In-house research organization of research

Marketing Manager

Research Sales/Distribution Advertising

Manager Manager Manager

There are a number of organizations that provide research to a firm. These comprise:

 custom research firms

These refer to institutions that are invited by the client firm and thereon work with
the client to develop and implement a complete research project. They may range
 from one-person consultancy establishment to well established firms such as audit
firms that also provide management consultancy services.

 Syndicated services
These refer to firms that routinely collect information that they provide to firms
subscribing to their services.

 Universities
Universities and research institutions usually provide research and consultancy
services in a variety of disciplines.

 Government agencies ( Central Statistical Office)

 UNIT TWO

MARKETING RESEARCH AND MARKETING

INFORMATION SYSTEM

As observed earlier, the principal function of marketing research is to provide marketing

managers with information that will assist them make better decisions pertaining to the
marketing of goods and services. In order to effectively carry out their responsibility,
marketing managers need information to address the array of issues falling within the
ambit of their responsibility ,such as, changing customer needs and preferences, the
impact of product design, price, promotion pr distribution strategies on demand, the
effect and so on and so forth. It is the role of marketing research to generate, assemble,
transmit and interpret all such information. It not only provides the link between the firm
and its environment, but more importantly, facilitates the evaluation of the success of the
marketing programme.

There are three approaches to the process of providing marketing intelligence: a

marketing research, a marketing information system and a marketing decision support
system. A marketing research system is undertaken to solve a specific problem faced by
the marketing manager. It is issue and time specific in the sense that information is
assembled with a known and specific problem in mind, and consequently information is
collected on a regular, continuous basis in anticipation of a problem a marketing manager
might encounter. The design of the marketing information system involves analyzing the
decision-maker and the realm of the decision-making responsibilities, and then analyzing,
identifying and specifying the type of information the decision-maker will need on a
regular basis or periodically. In a marketing information system, the information is pre-
formatted and standardized for use if and when need arises. As shown in figure 3,
marketing research is but a component of, and feeds into, a marketing information
system. Because marketing research information must relate directly to an identified
problem, there is usually guidance as to what research information is needed. The
information needs of the marketing information system however are much broader. Such
information is generally of three types:

Recurrent information- this refers to information provided on a periodic basis on

such variables as:
 Market share by product line, region
 Customer awareness of firm‟s advertising and promotion campaigns
 Activities of leading competitors
 Customer satisfaction with firm‟s product
Monitoring information- this refers to information collected on regular scanning of
selected sources, such as:
 A summary of reports on competition
 Government reports
Requested information-this refers to information developed in response to a specific
request by a marketing manager. This information would not flow without requests and
may very well not exist in the system.

Figure 3: Nature of Marketing Information System

Environment Data Marketing

Research

Guidance
Decision
Info
Request
Recurrent Info Marketing
Marketing Monitoring Info Information
Managers Requested Info System
Feedback
In short, the principal focus in MIS is to anticipate
 What decision each decision maker will regularly make
 What information is needed to make these decisions
 How often this information should be received.

THE WEAKNESSES OF MARKETING INFORMATION SYSTEM

The advent and development of marketing decision support system arose because of
certain shortcomings of marketing information systems. One shortcoming of an
information system is that it is premised on a standardized or structured information need
package which a typical manager need to make decisions. Most problems encountered
are, however, not structured, requiring structured solutions. There are thus in practice
very few report format that are optimal for different users. Decisions are compromises
requiring flexibility, and decision making is unstructured, with different managers using
different factors and in different ways to arrive at the decision. The second weakness of
the marketing information system is that there is often resistance to changes which are
implicit and inevitable in analyzing, designing, developing and implementing a dynamic
marketing information system. Managers may be reluctant, or simply not know, what
factors they ought to take into account in making a decision about a particular issue.
Thirdly, given a dynamic nature of business and the marketing environment, coupled
with the time required to develop an effective marketing information system, a marketing
information system may not respond in time to issues of the day.

MARKETING DECISION SUPPORT SYSTEM

A marketing decision support system is intended to overcome these short coming. It
differs from marketing information system in that t allows the decision maker to access
data and accords him the opportunity to manipulate the data to suit his decision
requirements. Bennett (1995) defined a marketing decision support system as;
“a coordinated collection of data, system tools, and techniques with supporting software
and hardware by which an organization gathers and interprets relevant information from
business and the environment and turns it into the basis for marketing decisions”

The essential features of the marketing decision support system are that the information
gathered fro business and the environment forms the basis for marketing actions, and that
decision makers or managers are allowed to interact directly with the data and models. In
this regard, its components are:

The Data System

 The data system comprises processes used for capturing data from both internal
and external sources, and methods for sorting data.
 A typical data system will contain information on customers, competitors, general
economic and industry trends and demographic variables.
 Marketing research would supply some of the information input.
 Data system will capture relevant data in sufficient detail and put it in accessible
form
Figure 4: Marketing Decision Support System
Interface Manager

Data Base Manager A

Inquiries

Computer
System Manager B

Responses
Model Manager C
System
Model System
 This refers to routines that allow a decision-maker to manipulate the data into a
form that will be useful to decision-making.
 The routines for manipulating the data are usually in the form of the statistical or
mathematical statements and may range fro basic operations to complex analysis
of the presumed relationship between two or more variables

Interface
 This refers to the software that allows the manager to interact with the data and
the model.
 It is the language system which permits managers with little or no programming
skills to access and explore the database and, using the model system, enable
them to produce information I formats that meets their information needs.
 It is this singular feature which distinguishes an information system from a
decision support system.

Managers
 The managers, who must make decisions by using information, represent the most
important aspect of the marketing decision support system. If they fail to use it
(properly), the system will be of no value
 Among other things, managers must understand and accept the models involved,
and how to do the relevant analyses.
 It is also preferable that they are involved in all aspects of the design of the
system.

AN OVERVIEW OF THE MARRKETING RESEARCH PROCESS

A research process is the series of activities that are carried out in conducting a research
project. These activities comprises of essentially of the following:
1. Problem Formulation
A research undertaking must begin with a clear understanding and specification
on the problem to be solved. Since marketing research information is intended to
facilitate decision making, a useful approach towards understanding of a problem
is to ask: what is it that necessitates that information be collected before action is
taken? The need to collect information is premised n the lack of knowledge, or
uncertainty, about an anticipated course of action. A major step in this area is to
define the objective of the specific project at hand.

2. Research Design
A research design is a framework, blueprint, or plan for the collection and
analysis of data. A research design is anchored on how much is known about the
problem being studied. There are three research designs: exploratory, descriptive
and experimental or causal research design. An exploratory research design is
used when there is little or no information about the problem being studied. Its
thrust then is to gain some insight into the problem. In contrast, descriptive
research designs are used when there is some knowledge about the problems and
the variables which relate to it. The thrust of the descriptive research design is to
provide an accurate and detailed description of the research problem. In a causal
or experimental research design, the thrust is to examine the functional
relationship of the variables in the problem model. Most research designs under
this category examine a causal relationship between variables; for example, the
relationship between sales and advertising.

3. Determination of Data Collection Method

The next step in the marketing research process is to determine how information
is to be collected. Information can be collected by questioning those deemed to
have answers or by observing the phenomenon of interest. When questioning is
used as a method of collecting data, several ancillary questions arise: what is the
nature of the questions to be asked? How should the questions be worded? How
shod questions be administered-is it in personal interview, over the telephone,
 through the mail, or by fax? If information is to be obtained by observation,
decisions must be made about what specific behaviors or events should be
observed and recorded and the means of observing and recording them.

4. Sample Design
Ideally, information should be collected by a census, that is, from the entire
population of interest. In reality, this is not possible. A researcher is then
compelled to collect information from only a part of the target population. The
process of collecting information from a part of, rather than the entire, universe is
known as sampling. To base decision on sample data requires validation of the
sample information. Such validation is done by addressing concerns about the
method used to select the chosen sample and adequacy of the size of the sample.

5. Data Collection
The actual collection of data addresses a number of questions. Not only must the
researcher administer matters to do with recruitment, training and deployment of
staff in the field, he/she must also deal with the greater problem of how to
mitigate errors which can potentially emanate from fieldwork. Such errors include
missing data occasioned by respondents who may be unavailable or unwilling to
answer questions; errors arising from the interviewer-interviewee interaction bias;
errors in recording responses; and outright cheating by interviewers.

6. Analysis and Interpretation of Data

Once the raw data has been collected, it must be processed with the view of
drawing meaning from it in line with the objectives of the study. This process
begins with editing for omissions, inconsistencies, misspellings and errors to
ensure that the collected data is of acceptable standards. This is then followed by
coding which is the arrangement of raw data into categories of responses. The
coding of closed questions is obvious and simple since the categorization of
responses will have been pre-determined in the construction of the questionnaire.
Coding of open questions is however necessary because of possible diversity of
 responses. The analysis of data then follows and interpretation of the finding is
made.

7. The Research Report

Because marketing research information is project-based, the research report is
the form in which information s provided to decision-makers. It is what the
decision-maker sees of the project. Its usefulness therefore, depends on the extent
to which the information contained therein is used as a basis for decision-making.
To this end, the report must be complete, accurate, clear, and concise and be of an
acceptable format.

Conclusion
The marketing of goods and services necessitates decision-making which enhances
customer satisfaction o a sustainable basis. The uncertainty and complexity of decision
making requires that marketing managers avails themselves of information that will
optimize their decision-making. This is the principal function and role of marketing
research. Marketing managers need information to make critical decisions in relation to
what need and wants to satisfy, how best to satisfy the identified needs and wants, what
new opportunities and threats are emerging in the external environment and what should
be the strategic response to identified environmental trends.
 UNIT THREE

TOPIC TWO: PROBLEM FORMULATION

Marketing research information is intended to facilitate better decision-making. By it

nature, decision making presumes the existence of the problem. It follows then that
marketing research is to play an intended role, the problem which necessitates marketing
research information must be clearly understood and specified. It is widely acknowledged
that a problem can only be solved if it is well defined

A decision problem can be said to exist when a situation obtains which represents a
problem or an opportunity such that a decision maker is required to take some action.
Churchill and Lacobucci (2202) have argued that research problems can arise from three
sources.

The first source is the internal and external environment in which a firm operates. The
external environment, involving technological, economic, political, legal and social
developments, may present opportunities which can be exploited, or problems which may
require some remedial action. New technologies in information and communication, for
example, have created immerse opportunities for better and more efficient ways of
conducting business. Political change from a one party system to multiparty democracy
heralded economic liberalization and privatization, with attendant opportunities and
problems. Economic liberalization, for instance, created the opportunity for expanded
markets but concomitantly presented problems of competition for local businesses which
hence forth ad to compete not only among themselves but also against foreign firms and
imports. The atmosphere of competition was further boosted by the enactment of the
Zambia Fair Trading and Competition Act in 1995. The principal aim of this act was to
encourage competition or to discourage behaviors which restricted, killed or otherwise
undermined competition. To this end, the Act, among other things, seeks to curb
monopolistic tendencies by prohibiting firms from enjoying a dominant position that
would give them power to determine prices or control production and distribution of
goods; from collusion that would enable them to share the domestic market; and from
effecting mergers and acquisition that would restrict, kill or distort competition

The second source of opportunity or problem requiring marketing decision making is

planned change. Strategic marketing decisions derive naturally from corporate strategic
decisions. If a firm seeks growth as its corporate vision, this will induce commensurate
marketing strategic response of, say, searching for new markets. The role of marketing
research in this instance will be to carry out feasibility studies that will enable
management to decide whether or not to go ahead with the planned change of seeking
growth.

A third source of opportunity and problem is a myriad of ideas which come to the
attention of the decision-maker by chance and in a random manner. Such ideas include
complaints from customers who are dissatisfied, suggestions for improvements, reports
by sale people, and deliberation and exchange of ideas at business conferences.

A decision maker‟s problem, which is our emphasis, must be distinguished from a

research problem. Typically, the former arises when the decision-maker perceives a
variance between a status quo and an idealized or expected situation. For example, I a
case of a marketing manager, if the sales are low and the market share is shrinking, this is
a problem because the desirable state of affairs should be for sales and market share to be
increasing. Thus, from a managerial perspective, the problem can be stated as one of low
sales, or one of change from low to high sales. The task of marketing research is to
provide information that will assist the marketing manager take some action to mitigate
the decline in sales or increasing market share. The research problem translate into an
attempt to answer the questions: why are sales falling? How can market share be
increased? The difference between the two types of problems is that the decision-maker‟s
problem is action-oriented, that is, what to do about the undesirable state. In contrast, a
research problem focuses on what information is needed in order to take some action.
It is important at this stage to clear certain myths and misunderstanding about the role of
marketing research I the context of decision making. The first is that not very decision
requires marketing research information. It is the prerogative of the decision maker
whether or not decision will be made with or without the benefit of the research
information. The decision maker needs to determine for himself whether or not he will
need information before taking action, and what information he will need if he decides
that he needs information. In general, a decision maker will be inclined towards
marketing research the less certain he is about the consequences of the action he might
take. It is therefore, necessary for the decision maker and the researcher to have a
common understanding of the benefits envisaged in undertaking any research. As a
corollary, a researcher must desist simply going through the motion of providing the
requested information. Rather, both researcher and decision maker must be candid with
each other and work together toward enriching the decision that will have to be made.

The context of a decision

In order to shed further light on the relationship between research information and
decision making, it is appropriate to illustrate the nature of decision situation and how
research information fits into the decision-making process. A decision situation
comprises of the following elements:

a) An objective function
A problem exists when that which is desired cannot be realized. The objective of the
marketing manager is to maximize sales, and when it cannot be realized, the manager
experiences a problem. To illustrate the point, assume that the management of
FRESCO, a soft drink company, wishes to introduce a new brand of a soft drink as
part of its growth strategy. The objective function of management therefore is to seek
more sales by introducing a new brand. It can be reasonably assumed that the firm is
not entirely satisfied with its current level of sales and consequently wishes to take
some action about this state of affairs.
b) Alternative Courses of Action
A second feature of a decision situation is that there are at least two courses of action
a decision maker can follow, but a decision maker is not certain about which course
of action to follow. In our example, the options open to management are to introduce
the new brand or not to introduce the new brand. Research can assist management by
clarifying the uncertainty about which option to follow.

c) Consequences of Alternative Courses of Action

A third element of a decision situation is to assess the consequences of each
alternative course of action. Researchers are primarily responsible for this. However,
before research is undertaken, management must ask themselves the following
questions:
1. Do we have to do research, or can we make a decision without undertaking
additional research?
2. If we do research, what are the areas of uncertainty which we wish to have
clarified by this research?
3. How much should we be willing to spend on the research?
4. If a number of research projects are under consideration, which one or ones
should be selected?

To answer the first question, management will be guided by their knowledge and
experience of the probable consequences of introducing or not introducing the new
brand. These consequences are known as “payoffs”. To go back to our example of
Fresco, let us assume that, on the basis of their knowledge and experience,
management feels that the introduction of the new brand will led to any of the three
possible outcomes; there could be a highly receptive to the introduction; there could
be an average reception; and there could be poor reaction. These possible reactions
from the market have attendant implications of realizable market share. They are
called states of nature because they are beyond the control of management.
Management must live with the risk of any of them occurring. Let us assume further
that management believes that if the market is highly receptive to the new brand, the
firm‟s market share will increase by more than 5 percent and profit will increase by
K5Million. If the new brand reception will be average, management believes that
market share will be between 1 and 5 percent and profits will rise by K3 million. If
the reception to the new brand will be poor, management believes that market share
will be less than 1 percent and the firm will incur a loss of K5 million. Because
market reduction to the new brand is beyond the control of the firm, management has
estimated the likelihood (probability) of any of these outcomes occurring. The
payoff table shows the respective consequences of each course of action given each
state of nature. The table shows that the likelihood of highly receptive, average and
poor consumer reactions are 0.30, 0.50 and 0.20 respectively.

d) Decision Without Research Information

Table 3.1 is known as the payoff table. It shows the decision options to-introduce and
not to introduce; the likely outcomes- highly receptive, average and poor reaction to
the introduction of the new brand; the probabilities of the probabilities; and the
rewards and penalties associated with each decision option. The information in table
3.1 can be used to make decisions whether or not to introduce a new brand. Note that
this information represents the best information available to management and is not
research information. The decisions will accordingly be made without the benefit of
research information. The decision will be based on management‟s knowledge,
experience or even gut feeling. The expected value of each decision alternative can be
calculated by multiplying outcome probability by the outcome reward or penalty and
summing. Management can compare the two expected values and make a decision on
which alternative to use. In our example, the optimal decision is to introduce the new
brand because the expected value of the decision to introduce the new brand is higher
than the expected value of not introducing the new brand.
Table 3.1: Payoffs of decision options

States of Nature
Highly Receptive Average Poor
(+5%) (1-5%) (<1%)
Decision Option Prob=0.30 Prob=0.50 Prob=0.20
Introduce K500,000 K300,000 K-500,000
Do not Introduce 0 0 0

EV (Introducing): (500,000x0.3) + (300,000x0.5) + (-500,000x0.2) = 200,000

EV (not introducing) (0x 0.3) + (0 x 0.5) + (0x 0.2) =0

If management chooses not to carry out research, the optimal decision is to introduce the
new brand.

Decision under perfect information

It should be noted that this analysis is based on estimates provided by management of
Fresco, represents management‟s best judgement without the benefit of research. If
management is satisfied that this information is correct, then there is no need for further
research and they can go ahead and introduce the new brand. If, however, management
has doubts about making a decision either way, then it must decide on which areas of
uncertainty research information will address. In our example, the area of uncertainty
relates to how consumers will react to the new product.

If management has perfect information, that is, if all uncertainty could be eliminated, the
correct decision could be made depending upon which state of nature existed. Thus, as
indicated in table 3.2, the option to introduce the new brand would be selected if
management was certain of a highly receptive or average reaction against the decision not
to introduce the new brand since the latter option yield nothing. On the other hand, if
management was certain of the poor reaction to the introduction of the new brand, the
selected option would be not to introduce the new brand because of the potential loss of
K5 million.
Table 3.2: Decision under uncertainty
States of nature
Pay off Decision
Highly receptive K500, 000 Introduce
Average K300, 000 Introduce
Poor -K500, 000 Do not introduce

The expected value of a decision under certainty can be computed by multiplying the
reward or penalty by the probability of being told for certain a particular consumer
reaction would obtain and summing:

EV (decision under certainty): (500, 000x0.3) + (300, 000x0.5) + (0x0.2) = 300, 000

The expected value of perfect information is the difference between the expected value of
decision under certainty and the expected value of optimal decision under uncertainty:

EV of perfect information: 300, 000 – 200, 000 = 100, 000

The significance of the value of perfect information is that it identifies an upper limit on
the amount that might be spent on research. In other words, the cost of anticipated
research should not exceed the expected value of perfect information.

The role and significant of research information

It was observed earlier that the uncertainty facing management in our example is how
consumers will react to the new brand. Management is not in a position to determine
precisely how consumers will react to the new brand. At best, management could only
estimate the likelihood of consumer reaction: that the probability was 0.3 that the
consumers would be highly receptive to the new brand, that probability was 0.5 that the
reception would be average and that the probability was 0.2 that the reception would be
poor. The principal role of marketing research is to provide information which shed more
light on the probabilities associated with consumer reaction. This is done by using sample
information to revise the earlier probabilities.

Assume that there are two possible samples of information that indicate the likelihood of
how consumers are likely to react to the new brand. These are shown in tables 3.4 and
3.5. They indicate conditional probabilities of getting a particular test market result given
each state of nature. For instance, in the case of project A, the table is read as follows: if
the true state of nature is high reception for the new brand, the probability that the
research finding (test market result) will indicate a test market result of over 5% or highly
receptive reception is 0.60; if on the other hand the true state of nature are average
reception or poor reception, the respective probabilities of getting a test market result of
over 5% market share are 0.20 and 0.10. in considering whether or not to use sample
information, management will want to estimate how the new information will change the
new probabilities. For example, if the research project shows that the new brand will
attain a 4% market share, how should the prior probabilities of 0.30, 0.50 and 0.20 for
highly receptive, average and poor reaction respectively change? The premise here is
that the research finding will approximate the true state of consumer reaction. That
is, if the test market result shows a market share of over 5%, the true consumer
reaction to the new brand is most probably highly receptive.

Table 3.3: conditional probabilities of getting each test market result

given each state of nature

PROJECT A
States of Nature
Test Result of Highly Receptive Average Reception Poor Reception
market share
Over 5 percent 0.60 0.20 0.10
1-5 percent 0.30 0.50 0.20
Below 1 percent 0.10 0.30 0.70
 PROJECT B
States of Nature
Test Result of Highly Receptive Average Reception Poor Reception
market share
Over 5 percent 0.70 0.15 0.05
1-5 percent 0.25 0.60 0.15
Below 1 percent 0.05 0.25 0.80

REVISING PRIOR PROBABILITIES GIVEN A PARTICULAR TEST

MARKET RESULT IN THE CASE OF PROJECT A
The procedure for calculating a revised probability given a conditional probability and a
particular test market result is shown in tables 3.4, 3.5 and 3.6 in the case of project A,
and in Tables 3.9, 3.10 and 3.11 in case of project B. using the example of project A, the
analysis consists of the following steps:

Establishing prior probabilities

Prior probabilities are the probabilities management assigns to the possible existence of
each state of nature. In our example, these states of nature refer to the consumer reaction
to the introduction of a new brand. The probabilities of 0.30, 0.50, and 0.20 were
assigned to consumer reaction of highly receptive, average reception and poor reception
respectively. They are called prior probabilities because they exist prior to obtaining
additional information. As stated earlier, they are subjectively determined by
management using best judgement and constitute management‟s best estimate of a
likelihood that a particular reaction is the one which actually exists in the real world.

Determining conditional probabilities for different market test outcomes

If the project is undertaken, the result can show a market share of over 5%, a market
share of between 1-5%, or a market share of less than 1%. As explained earlier, what
table 3.3 indicate is that, in the case of project A, if the true state of nature is that there
will be a poor reception for the new brand, the probability is 0.70 that the test market will
show this result (below 1%)

Management and the researcher are required to estimate the conditional probabilities.

Calculating joint probabilities

As indicated in tables 3.4, 3.5 and 3.6, a joint probability is the conditional probability
multiplied by the prior probability. It is derived by combining the probabilities that a
given research outcome will be obtained if a given state of nature exists with the
probability that the specified state of nature does in fact exist. In reference to table 3.4,
the figures in the second column, under the heading of prior probability, refer to
management‟s estimate of the probability of consumer reaction. As stated earlier,
management assigned probabilities of 0.30, 0.50 and 0.20 to consumer reaction being
highly receptive, average and poor respectively. The figures under the column headed
conditional probability are taken from table 3.3. These are the probabilities of getting a
particular research outcome given that a particular state of nature actually exists. The
joint probability that a particular state of nature is the true state of nature and that a
particular research outcome will occur is presented under the column headed joint
probability and is calculated by multiplying the prior probability by conditional
probability. For instance, the joint probability that consumer reaction is highly receptive
is indeed the true state of nature and that a research outcome showing a market share of
over 5% will occur will equal to 0.30 x 0.60= 0.18 that a highly receptive consumer
reaction exists and that a research outcome showing an increase in market share of over
5% will result if research is undertaken. Similarly, there is a probability of 0.04 that a
poor consumer reaction exists and that a research outcome showing an increase in market
share of between 1 and 5% will result if research is undertaken (see table 3.5)
Table 3.4: Revised probability if research outcome is over 5%
State of Nature Prior Conditional Joint Revised
Probability Probability Probability Probability
Highly receptive 0.30 0.60 0.18 0.60
Average 0.50 0.20 0.10 0.33
Poor 0.20 0.10 0.02 0.07
0.30 1.00

EV (introducing): (500,000x0.6) + (300,000x0.33) + (-500,000x0.07) = 364,000

EV (not introducing): (0x0.6) + (0x0.33) + (0x0.07) =0

Optimal decision: Introduce

Table 3.5: Revised probability if research outcome is 1-5%

State of Nature Prior Conditional Joint Revised
Probability Probability Probability Probability
Highly receptive 0.30 0.30 0.09 0.24
Average 0.50 0.50 0.25 0.66
Poor 0.20 0.20 0.04 0.10
0.38 1.00

EV (introducing): (500,000x0.24) + (300,000x0.66) + (-500,000x0.10) = 268, 000

EV (not introducing): (0x0.24) + (0x0.66) + (0x0.10) =0

Optimal decision: Introduce

Table 3.6: Revises probability if research outcome is below 1%
State of Nature Prior Conditional Joint Revised
Probability Probability Probability Probability
Highly receptive 0.30 0.10 0.03 0.09
Average 0.50 0.30 0.15 0.47
Poor 0.20 0.70 0.14 0.44
0.32 1.00

EV (introducing): (500,000x0.09) + (300,000x0.47) + (-500,000x0.44) = -34, 000

EV (not introducing): (0x0.24) + (0x0.66) + (0x0.10) =0

Optimal decision: Do not introduce

Estimating the occurrence of research outcomes (marginal probabilities)

Joint probabilities calculated under each research outcome are used to estimate the
probability that the research outcome will occur. Thus, the probability, (also known as the
marginal probability), that a research outcome showing less than 1% market share (see
table 3.6) will occur can be calculate by summing the joint probabilities associated with
this research outcome as follows:

HIGHLY RECEPTIVE AND LESS THAN 1% = 0.03

OCCURRING TOGETHER
AVERAGE REACTION AND LESS THAN 1% = 0.15
OCCURRING TOGETHER
POOR REACTION AND LESS THAN 1% OCCURRING = 0.14
TOGETHER
0.32

Similarly, the probability of research outcomes of 5% and 1-5% market share occurring
are 0.30 and 0.38 respectively.
Revising prior probabilities
Management estimated prior probabilities of 0.30, 0.50 and 0.20 for the existence of the
three consumer reaction-highly receptive, average and poor respectively. After research
results are obtained, management will use them to revise these prior probabilities. For
example, given tat management had assigned a probability of 0.30 for highly receptive
consumer reaction, how might this initial probability change if the research outcome
shows that the consumer reaction resulted in a market share increase of more than 5%, 1-
5% or a decrease in market share to below 1%? The revised probability indicates the
probability of the true state of nature (consumer reaction) in the light of research
outcome. Table 3.5, for example is read as follows: if the research outcome is the
consumer reaction of between 1 and 5% market share, the probability that the true state of
nature is average reaction is 0.66, that the true state of nature is poor is 0.10 and that the
true state of nature is highly receptive is 0.24. These revised probabilities are obtained by
dividing the probability of occurrence of a research outcome by the joint probability. This
calculation is illustrated in table 3.7

Table 3.7: probability of a given state of nature being the true state if a
given research outcome is obtained
POSSIBLE RESEARCH OUTCOMES
STATE OF NATURE OVER 5% 1-5% LESS THAN 1%
HIGHLY RECEPTIVE 0.18÷ 0.30=0.60 0.09 ÷ 0.38=0.24 0.03 ÷ 0.32=0.09
AVERAGE RECEPTION 0.10 ÷ 0.30=0.33 0.25 ÷ 0.38=0.66 0.15 ÷ 0.32=0.47
POOR RECEPTION 0.02 ÷ 0.30=0.07 0.04 ÷ 0.38=0.10 0.14 ÷0.32=0.44

Determining of the expected value of decision after research

The final step in the analysis answers the question, “what is the value of information
which will be generated by the research?” to arrive at the value of research information,
the value of the decision with the benefit of research is computed as shown in figure 3.8
by summing the values of optimal decisions under each research outcome. Thus, if
project A was used, the expected value of the decision would be K211, 040. It will be
recalled, however, that expected value of the decision without research was earlier
computed to be K200, 000. the value of research information is therefore the difference
between the expected value of the decision with research (K211, 040) and the expected
value of the decision without research (K200, 000), that is, K11, 040

Value of research information: 211, 040 – 200, 000 = 11, 040

Table 3.8: expected value of decision after test marketing

Test Result Decision EV of Decision Probability EV of
of Decision Decision
Over 5 percent Introduce 364, 000 0.30 109, 200
1-5percent Introduce 268, 000 0.38 101, 840
Below 1 percent Do not introduce 0 0.32 0
EV of Decision after Test 211, 040

If a number of research projects are under consideration, which one

should be selected?
Since in our example there are two projects, a decision is required regarding which of the
two projects should be selected. Clearly, the project which adds more value to the
decision should be the one to be selected. It is therefore necessary to compare the relative
expected value of the research information between the two projects. The analysis which
follows corresponds to the one which has been done in the case of project A. following
through with the analysis, it can be observed that, in the case of project B, the expected
value of the decision with the benefit of research information is K241, 40 (refer to table
3.12). To arrive at the expected value of the information generated by project B, the
expected value of decision without research is deducted from the expected value of the
decision with research. Hence, the expected value of research information under project
B is:

K241, 240 – K200, 000 = K41, 240

At this stage, project B would be selected since the expected value of its information is
greater than that of project A. if the cost of undertaking a particular project is taken into
account, this must be deducted from the expected value of research to yield the net worth
of the project. Tables 3.13 indicate that the cost of project A and project B are K10, 000
and K20, 000 respectively. A comparison is then made of the value of the net worth
between the two projects. Tables 3.13 illustrate these comparisons and shows that project
B has a larger net worth (K21, 240) than that of project A (K1, 040) and should therefore
be selected.

THE EXPECTED VALUE OF RESEARCH UNDER PROJECT B

Table 3.9: Revised probability if research outcome is over 5%
State of Nature Prior Conditional Joint Revised
Probability Probability Probability Probability
Highly receptive 0.30 0.70 0.21 0.70
Average 0.50 0.15 0.08 0.27
Poor 0.20 0.05 0.01 0.03
0.30 1.00

EV (introducing): (500,000x0.70) + (300,000x0.27) + (-500,000x0.03) = 416, 000

EV (not introducing): (0x0.70) + (0x0.27) + (0x0.03) =0

Optimal decision: Introduce

Table 3.10: Revised Probability if research outcome is 1-5%

State of Nature Prior Conditional Joint Revised
Probability Probability Probability Probability
Highly receptive 0.30 0.25 0.08 0.20
Average 0.50 0.60 0.30 0.73
Poor 0.20 0.15 0.03 0.07
0.41 1.00

EV (introducing): (500,000x0.20) + (300,000x0.73) + (-500,000x0.07) = 284, 000

EV (not introducing): (0x0.20) + (0x0.73) + (0x0.07) =0

Optimal decision: Introduce

Table 3.11: Revised probability if research outcome is below 1%

State of Nature Prior Conditional Joint Revised
Probability Probability Probability Probability
Highly receptive 0.30 0.05 0.02 0.06
Average 0.50 0.25 0.13 0.42
Poor 0.20 0.80 0.16 0.52
0.31 1.00

EV (introducing): (500,000x0.06) + (300,000x0.42) + (-500,000x0.52) = -104, 000

EV (not introducing): (0x0.06) + (0x0.42) + (0x0.52) =0

Optimal decision: Do not introduce

Table 3.12: expected value of decision after test marketing

Test Result Decision EV of Decision Probability EV of
of Decision Decision
Over 5 percent Introduce 416, 000 0.30 124, 800
1-5percent Introduce 284, 000 0.41 116, 440
Below 1 percent Do not introduce 0 0.31 0
EV of Decision after Test 241, 240

Value of research information: 241, 240 – 200, 000 = 41, 240

Net Value of research: 41, 240 – 20, 000 = 21, 240

 THE OPTIMAL PROJECT IS DETERMINED BY CALCULATING AND COMPARING
THE NET WORTH OF RESEARCH UNDER EACH PROJECT, THUS;

Table 3.13: Summary of value of proposed research

Research EV of Value of Cost of Net Value of
Project Introduction Research Research Research
A 211, 040 11, 040 10, 000 1, 040
B 241, 240 41, 240 20, 000 21, 240

Select project B because it has a higher net value.

 UNIT FOUR

TOPIC: RESEARCH DESIGN

Once a research problem is specified and understood, the next step in t research process is
the formulation of the plan of study for the collection and analysis of data. This plan of
study is known as the research design. It is intended to guide research efforts to ensure
that study is relevant and carried out as cheaply as possible.

A useful way to classify research design is to focus on the principal objectives of the
research. A research design can be exploratory, descriptive or causal.

An exploratory research design

An exploratory research design is used when there is little or no knowledge about the
problem. It is therefore intended to shed light or offer new insight or ideas about the
problem at hand. Specifically, it is used (a) to formulate a problem for more precise
investigation or for developing hypotheses. This may particularly apply in situations
where the problem is too broadly or vaguely stated; (b) to help in establishing an order of
priorities, such as in instances where there are several explanatory variables to a
phenomenon; (c) to clarify concepts which are vague or carry multiple meanings; (d) to
clear problems which stand in a way I carrying out a study and; (e) to familiarize oneself
about the problems to be studied.

Because the dimensions of the research study are not precisely known and because
knowledge about the problem is lacking, an exploratory study is characterized with
flexibility. The researcher must sniff around for information, relying on imagination,
ingenuity, judgement or mother luck. In this regard, there are three primary sources of
information. One source of information involves literature search comprising conceptual
literature, trade literature, published data and previous published studies. Information can
also be obtained from key informants, also known as experience surveys. These are
individuals deemed to have some knowledge about the problem at hand. Finally, one may
have recourse to an intensive analysis of a few selected cases.

A Descriptive Research Design

A descriptive study is a more common type of a marketing study. It is used to provide an
accurate description to the problem model. More specifically, it may be used (a) to
describe characteristics of certain groups. For instance, one may be interested in knowing
who is a “typical” traveler with respect to age, sex, tribe, race, education etc; (b) to
estimate the proportion of people who possess a particular attribute or who behave in a
certain way. One could be interested, for example in a number of people who prefer a
particular brand against other brands; (c) to make specific predictions. For instance, one
could be interested in a number of people who will purchase a product in, say, three
years.

In contrast to an exploratory study, there will be some knowledge about the problem at
hand in a descriptive study. The thrust of the study is to get more detailed description of
the problem variables. The format for the collection and the analysis of data tends to be
structured and more rigid than in an exploratory research design. As an illustration, we
might e interested in how the purchase of the products varies among people with the
certain level of education, and we might develop the following framework, called a
dummy table, to study the possible relationship between newspaper readership and the
level of education. Such a profile would reveal, for example, which newspaper is
preferred by people of a particular level of education. Of note in such a framework is that
there is some idea of the kind and structure of information required. What a descriptive
research design attempts to do is to provide more detailed information.
Table 4.1: Readership of Newspaper by Education
Level of Education Times Post Mail Total
Up to Grade VII 73 77 75 225
Grade VII to Grade XII 90 20 40 150
University Graduate 40 50 35 125
203 147 150 500

There are three source of information in descriptive studies. One source is the use of
panels. A panel is a select group of people from whom information is sought on a
continuous basis. A major advantage of panel data is that organizational costs are
minimized as information may be sought from the same group of people, albeit over a
number of issues. A second advantage is that one is able to achieve consistency and make
comparisons over time. However, panel interviews suffer from not being representative.
There may also be costs of trying to keep the panel together.

A second source of information in descriptive studies is to carry out a cross-section (or

survey) study whereby a sample of elements is selected from the population of interest
and a study made at single point in time. Many marketing research studies are usually of
the survey type. They are preferred because marketers wish as much as possible to
capture the mood of their typical market. Survey studies offer the advantage of being
wide, extensive and representative in scope.

A third source of information is through an in-depth study of few selected cases. This is
known as case study approach. In contrast to a survey, a case study is narrower, intensive
and clearly not representative. However, its advantage over a survey approach is that it
offers in-depth knowledge about the subject by narrowing the scope to a few cases. In a
case study, one attempts to identify factors which are common, or features which are
common to a subgroup, or feature which are unique to a specific case. Sharp contrast can
be revealed better in a case study than in a survey study. A survey study, on the other
hand, aims at providing a general idea of some phenomenon.
A Causal Research Design
A major limitation of the descriptive studies is that they cannot satisfactorily establish
causality, in the sense that X can be said to be the probable cause of Y. causal or
experimental designs are intended to explore cause and effect relationships.

It is apt that the common sense notion of the concept of causality is distinguished from
scientific notion. The following three distinctions are made:

i. In scientific investigation, we speak of X being on e off the possible causal factors

rather than single causal factor of Y. this is a recognition that other factors may
lead to the occurrence of Y. the mere fact that they are unknown is not reason
enough to ignore that they exist
.
ii. In scientific investigations, the causal relationship is the probabilistic rather than a
deterministic one
.
iii. In scientific investigations, a causal relationship is inferred rather than proven.

A major implication arising from the scientific notion of causality is not error-free and a
causal relationship should not be stated as if the relationship has been exhausted, is
certain or proven beyond doubt. Causal relationship cannot therefore be conclusively
demonstrated. Nevertheless, inferences of causality can be made on the basis of three
types of evidence:

(a) Concomitant Variation

Concomitant variation is said to occur when the causal factor X and the effect
factor Y, occur or vary as predicted by the hypothesis. For example, according to
the marketing theory, sales are dependent on a number of outlets. Table 3.2 below
is an illustration of concomitant variation in that regions that had more sales
outlets had relatively more sales.
 Table 3.2: Sales by number of outlets
Sales Outlets Sales
Region Number % Units %
Lusaka 1, 000 57 K10, 000 56
Copperbelt 500 29 K6, 000 34
Livingstone 82 05 K200 01
170 10 K1, 500 09
1, 752 100 K17, 700 100

(b) Time Order of Occurrence of X and Y

This concept holds that if X is the cause of Y, then the occurrence of X must
precede the occurrence of Y. advertising and sales promotion are intended to
increase sales. According to the principle of time order of occurrence, sales must
occur after advertising or sales promotion. The validity of this principle depends
on the time lag between the occurrence of X and Y. in general, the shorter the
time lag between the occurrence of X and Y, the greater our confidence in
inferring causality between the two variables. If the time lag is considerable, then
causality could be due to other extraneous factors.

(c) Elimination of Other Possible Causal Factors

In some situations, there may be many professed factors alleged to cause a
phenomenon to occur. Our interest then is to establish which ones of these many
factors can be said to be the cause. This is done by eliminating the likelihood of
other factors being the causal factors. For example, dissatisfaction with the
product can be caused by poor quality, high price, poor service etc. but not all of
these may be the cause of dissatisfaction at a time of interest. If dissatisfaction is
caused by poor quality, then one must show by elimination that dissatisfaction has
not been caused by high price, or poor service.

Experimentation
Experiments are mechanisms for providing evidence of causality. They allow one to
check for all the three types evidence of causality.
The essence of an experiment is for the researcher to manipulate one variable, called the
independent or causal variable, in order to observe the effect of the manipulation on
another variable, called the dependent or effect variable.

Experiments can be of two types. A laboratory experiment is were the setting of the
experimental situation is contrived, or determined by the researcher. A field experiment is
where the situation is a natural one. Consider a situation in which a researcher wishes to
determine whether or not the demand for ice cream is influenced by flavour. In a
laboratory experiment, the researcher could gather, say, children and ask each one of
them to pick a flavour of ice cream they preferred. He could then determine the most
popular flavour from the selections made by the children and infer causality from the
observations made. The researcher contrives the setting for this experiment. In contrast,
in the field experiment, would not have control of, for instance the test units or the timing
of the experiment. Any individual could conceivably buy ice cream as they went on
doing their normal shopping. The purchase of ice cream would not be dictated by the
researcher but by natural events. Causality would thus be inferred under natural
conditions.

Validity- External and Internal

The major advantage of the laboratory experiment is that it offers internal validity, to the
extent that the experimental situation can be manipulated to allow the researcher to
control for the effect of other possible causal factors. That is, evidence of causality is
demonstrated by the researcher‟s ability to attribute the effect of the experiment to the
manipulation of the independent variable and not to other factors. On the other hand, a
field experiment offers external validity, in the sense that observed results arising from
the experimental situation can be expected to occur in other situations.

In our ice cream example, if the experiment were confined to children by refusing adults
to take part, the researcher would be confident in making a statement that inferred that a
particular flavour of ice cream is most popular among children who took part in the
experiment. He would not be correct to extend his conclusion to beyond the spheres of
the experiment. On the other hand, if every one who came along could buy ice cream of
their choice in a natural course of events, it could be reasonably inferred that the flavour
that was found to be popular in the experiment was equally popular beyond the confines
of the experiment. Nevertheless, in a field experiment, there is a possibility that other
factors, unknown to, or beyond the control of the researcher, may have contributed to the
observed effect. Such factors, if known to the researcher, can be controlled for in a
laboratory experiment but not in a field experiment.

Extraneous Factors
Extraneous factors are events that cannot be manipulated by the research, which
nevertheless have an effect on the variable to be observed. The following are some
common examples of extraneous factors.

History
History refers to events outside the spheres of the experiment but occurring at the same
time as the experimental variable. Sales at Christmas cannot be attributed to advertising
alone because there might be other factors which might spur demand, such as a Christmas
bonus or the festive nature of the occasion.

Maturation
Maturation refers to the changes that occur within the test units during an experiment due
to the passage of time. Where test units are people, maturation refers to the fact that
people get old, get tired and perhaps feel hungry. Responses to the experimental variable
may be affected, for example by hunger or fatigue.

Testing
Testing refers to the changes induced by the very process of experimentation. The testing
effect can be of two types. The first type, known as the main testing effect, is when the
earlier response to an experimental variable conditions subsequent responses. This can
happen, for instance, when earlier responses are used as reference points in responding to
an experimental variable. An example might be a panel of housewives requested by the
company to record monthly expenditure on an item. In their desire to be consistent,
housewives might be influenced in their subsequent answers by the answers they gave
earlier. A second type, known as the interactive testing effect, is where a prior response
affects a subsequent response to an experimental variable. For instance, if some one is
introduced to some product, they are likely to be sensitized and will become more aware
of subsequent advertisement of the product because of the initial alertness that was
created. The reaction to the advertisement is amplified by the prior exposure.

Instrument Variation
Refers to changes that are induced by variation in the measurement instrument or in the
administration of the instrument. Variation in responses may occur, for instance, when
two interviewers are used instead of one.

Statistical Regression
Refers to the changes brought about by the tendency of polar points to move towards te
average during the course of an experiment. In a panel interview, for instance,
interviewers may start off with an initial position on an issue and shift from the initial
position to another position in the course of the interview.

Selection Bias
Most experimental designs make use of an experimental group and a control group.
Selection bias is said to occur when there is no way of certifying that the test unit selected
and assigned to groups were equal prior to the selection and assignment to groups.
Inherent inequality among test units may lead to bias in a way test units responds to the
experimental variable. To minimize the incidence of selection bias, there should be
random selection and assignment to groups.
Experimental Mortality
Refers to effect caused by loss of test units during the course of an experiment. If the
experiment involves one group, mortality occurs when respondents that remain in the
study differ in their responses to the experimental variable from those who withdraw. The
question then is; would the answers have been the same if test unit had remained the
same?

If there are two groups or more, mortality refers to errors caused by differences in
responses due to the loss of respondents from the various groups in the experiment.

Reactive Error
Changes caused by the artificiality of the experimental situation ad/ or the behaviour of
the experimenter. Reactive error manifests itself by way of dampening, emphasizing or
altering the effects of the experimental variables on the test units. Reactive error arises
principally because of the tendency of human treatments to anticipate what the
experimental situation is all about and the attempt to produce the anticipated behaviour.
An example of reactive error is the exaggerated behaviour one might notice from models
at a fashion show.

Measurement Timing Error

This occurs when pre-measurements or post-measurements are made at an inappropriate
time. It is normal, for instance, to expect the effects of the independent variable to occur
immediately. But consider the case of advertising: it is difficult to determine the suitable
time to measure the effects of advertising.

Surrogate Situation Error

This error occurs when the environment, the population sampled or treatments
administered are different from those that will be encountered in the actual situation. If a
product is intended for a foreign market, should product acceptance be undertaken in the
domestic market?
 EXPERIMETAL DESIGNS
As indicated earlier, experimental designs are intended to capture a causal relationship. In
experimental design, a researcher deals with two variables: one is the independent (or
causal) factor and the other is the dependent (or effect) factor. In order to analyze a causal
relationship, the independent factor is manipulated and the effect of the manipulation
observed. In the following discussion of experimental designs, the following symbols
will be used to illustrate causality:

X will designate an experimental (or causal) variable

O will designate an observation/measurement

a) “After Only” Design

This design involves manipulating the independent variable and following this with a
post-measurement as shown below.

X O

An example of an after-only design might be a situation where a company invites

people to a function where a new product is to be introduced. Subsequent purchases
of the new product by those who attended the function might then be used to infer a
causal relationship: that the purchase where brought about by the introduction of the
new product at the function.

After-only designs do not adequately control for the effects of other extraneous
factors such as history, maturation, selection and mortality.

b) “Before-After” Design
In this design, a prior measurement is made before the experimental variable is
administered. After a prior measurement is taken, an experimental variable is
introduced and a post-measurement is taken. The impact of the experimental variable
is taken to be the difference between the prior measurement and post-measurement.
 01 : Pre-measurement or prior observation
X : Experimental variable introduced
02 : Post-measurement or after observation
02 - 01 : Effect of experimental variable

Although this design is an improvement on the after-only design, it does not control
for the effect of extraneous factors of history, maturation, testing, mortality or
instrument variation.

c) Before-After with Control Group

Two features distinguish this experimental design from the previous two designs.
These are the introduction of the control group and the randomization of the selection
and assignments of the respondents to the two groups. Thus, in this design, there are
two groups, one designated as an experimental group and the other designated as a
control group. The experimental variable is administered to the experimental group
but not to the control group. Randomization means that respondents are randomly
selected and randomly assigned to the two groups.

Experimental Group Control Group

01 02 : Before measurement
X
03 04 : After measurement

The effect of the experimental variable is then said to be:

(03 – 01) - (04 - 02);

Or put another way, the analysis of the composition of the difference between the two
groups shows that the effects of the experimental variable is contained in the
difference between the pre-measurement and post-measurement in the experimental
group. However, as observed in the previous designs, this difference will also contain
other extraneous factors but, now, not selection bias because of randomization. The
difference between the pre-measurement and post-measurement in the control group
reflect extraneous factors only as this group was not exposed to the experimental
variable. Note however, that because of randomization in the process of selection and
assignment to groups, it can reasonably be assumed that extraneous factors affected
both groups equally. Hence, the difference between the two groups is the impact of
the experimental variable as shown below:

03 – 01 = Experimental variable X + Extraneous factors

04 – 02 = Extraneous factors

(03 – 01) - (04 – 02) = Experimental Variable

The only extraneous factor not controlled for in this design is the interactive testing
unit. This is because of the prior measurement taken in the experimental group. The
composition of the difference can now be refined to read as follows:

03 – 01 = Experimental Variable X+ Extraneous factors+ Interactive

04 – 02 = Extraneous factors

(03 – 01) – (04 – 02) : Experimental variable + Interactive

d) Four- Group Six- Study Design
The major objective here is to control for the interactive testing effect. The features of
this are as follows:
 The sample of respondents is randomly selected
 There are four groups comprising two experimental groups and two control
groups. The groups are designated as experimental group 1, control group 1,
experimental group 2 and control group 2
 Respondents are randomly assigned to the four groups
 Pre-measurements are taken for the two groups designated as experimental
group 1 and control group 1.
 The experimental variable is introduced in the two groups designated
experimental groups
 Post-measurements are then taken four all the four groups

This can be depicted as follows:

EG1 CG1 EG2 CG2

01 02

X X

03 04 05 06

The impact of the experimental variable is determined by comparing the second

experimental and control groups as shown in the table below. Note also that the
analysis enables one to determine (a) the magnitude of the interactive testing effect by
comparing the two experimental groups and (b) the size of the extraneous variables
by each of the control groups.
Composition of Differences:
EG1 03 – 01 = Experimental Variable + Extraneous variable +
Interactive Testing effect
CG1 04 – 02 = Extraneous Variables
EG2 05 – ½ = Experimental variable + Extraneous variables
( 01 + 02 )
CG2 06 – ½ = Extraneous variables
( 01 + 02 )
Effects of Experimental = EG2 – CG2
Variable

e) Post measurement Only with Control Group Design

A careful examination of the four-group six-study design will reveal that to determine
the impact of the experimental variable one only needs to take the second
experimental and control groups, that is, EG2 and CG2. The after only with control
group design can be illustrated as follows:

EG2 CG2

Before measurement
Experimental Variable X
After measurement 05 - 06
Effect of experimental variable 05 - 06

The extraneous variables are controlled for by randomizing and assuming therefore
that they equally affect the two groups and cancel each other. The interactive testing
effect is controlled for by not taking pre-measurements
f) Time Series Experiment
This type of design can be illustrated as follows:

01 02 03 X 04 05 06
As can be illustrated, the time series experiment requires that the researcher has
access to the same test units. What is envisaged is that a series of measurement are
taken prior to the introduction of an experimental variable. These initial
measurements are intended to indicate a pattern in the behaviour being measured.
Then, an experimental variable is introduced and therefore a series of measurements
are again taken. The difference in patterns between the two series of measurement is
ten said to have been due to the introduction of the experimental variable.

An example of the time series experiment might be a situation in which a decline in

sales is observed. Then, X is observed and the decline in sales is arrested and sales
begin to increase. This pattern might appear as indicated below.

Figure 4.1: An illustration of a time series experimental design

Sales

X
Jan Feb Mar Apr May Jun Jul

Obviously, if the pattern changes before the introduction of X, or does not change
despite the introduction of X, we can reasonably infer tat X has no relationship to the
pattern of sales.
A weakness of time-series experimentation is that it cannot control for extraneous
factors such as history.
 UNIT FIVE

TOPIC FOUR: TYPES AND SOURCES OF

RESEARCH INFORMATION

We have thus far examined two topics, namely, problem formulation and research
designs. We now turn to discussion of (a) types and sources of research information
and (b) methods of collecting research information.

TYPES OF DATA
There are basically two types of data. The first type is secondary data and the second
type is known as primary data. Secondary data refers to information collected for
some other purpose other than the problem at hand. Primary data on the other hand
refers to information collected specifically for the study at hand. Given a research
problem, it is tempting to rush into survey where one has to generate primary data.
However, it is advisable to check first whether or not information already exists on
the problem to be studied.

Advantages of Secondary Data

 Economy
It is cheaper to use secondary data than to generate information. Use of primary
data usually entails designing and printing data-collection forms, hiring and
training field workers, transporting them to field areas and maintaining them, and
editing and tabulating data.

 Speed
Secondary data can be obtained far more quickly than primary data. If data
already exists, in published form and/or in a central place, it can be obtained at the
fraction of the time one would spend in generating fresh information.
  Accessibility
Some secondary data sources provide information which could not be obtained by
a typical organization. For instance, information provided by government sources
is not easily obtainable if a non government source were to solicit the information.

Disadvantages of Secondary Data

There are two potential drawbacks to secondary data. These are possible unsuitability and
the likelihood of inaccuracy.
 Unsuitability
Unsuitability of secondary data may arise because of a number of factors:

 Difference in the units of measurement

Our currency is in Zambian kwacha and yet we may come across studies in
which the currency used is in pound sterling or US dollar. Consider an
extreme case where trading is conducted by barter! Other examples where
secondary data are expressed in units different from what we might consider
appropriate relate to size of an establishment (profit, gross sales, number of
employees, or physical space); consumer income ( at individual, family,
household or spending unit).

 Different Class Definition

There may be lack of a generally acceptable understanding of where one
draws the boundary for certain characteristics. Thus, while we use the terms
such as poor and rich, or low, medium and high income, there are wide
variations in the understanding of who is poor, who is of low income as
opposed to those of medium income. A person categorized as poor in the
United Kingdom may not be poor by Zambian standards.
  Recency
Research information becomes less valuable the older it gets. Secondary
information is by definition historical rather than current and may therefore be
unsuitable simply because it is somewhat out of date.

 Inaccuracy
Secondary data may not be accurate. Again, this arises principally because the data
was collected for some other purpose and may have been processed by several
intermediaries to suit their interests. The more intermediaries there are the greater the
likelihood of data manipulation. It is therefore to use a primary source rather than a
secondary source. Since secondary data are derived from some other source, it
follows then that there is less likelihood of error with a primary source than with a
secondary source. If data are collected from a source which originally gathered them,
one is afforded an appreciation and explanation of the context or the explanation of
how the data were collected. In contrast, secondary sources do not afford one the
opportunity of such an explanation. Accuracy of secondary data may also be
compromised by the purpose for which it was collected. One should be wary, for
instance, of data published for propaganda purposes, of information contained in an
advertisement or promotional campaign, data published anonymously or in
controversial circumstances, or information used in advancing the interests of an
industrial or commercial or any other group.

Finally, one may assess the accuracy of secondary data by examining the quality of
the data itself. A useful guideline is to look at the ability and reputation of the source
that generated the data and the methodology used in generating the data.

Sources of Secondary Data

Sources of secondary data can be distinguished by whether or not they are within the
organization commissioning the research
Internal Sources of Secondary Data
Accounting and financial records are an important source of secondary data. Such
records include a sales invoice, budgets, annual reports, and reports by sales people.
Typically, a sales invoice will contain information relating customer identity and
location, nature and quantities of items shipped to customers, the terms of payments, and
discount offered to the customer. Information gleaned from a sales invoice can facilitate
understanding of the nature and pattern of the company‟s sales and profitability of market
segments. Budgets and financial plans indicate what the company intends to accomplish
and, read together with management accounts, provide insight into how well the company
is doing.

Accounting and financial reports can be supplemented by reports from other departments.
The marketing department might provide some useful information on advertising
expenditure and the effectiveness of advertising in generating sales; the human resource
department might provide pertinent information on employees; and production might
have records on production outputs and levels of productivity.

External Sources of Secondary Data

The following are some of the important sources of secondary data:

1. Computerized Databases
These are basically information banks consisting of data and/or textual
information which are made available in computer reader form for electronic
distribution. Of particular interest to marketing studies are bibliographic and
numerical databases which appear in form of articles or reports. Numeric
databases contain numeric data on topics such as sales, population, etc.

The information in the databases is accessible to be used for a fee. Universities

and commercial organizations can supply such information.
2. Association
Associations frequently publish or maintain detailed information on industry
sales, growth patterns, operating characteristics etc. such associations may include
chambers of commerce, the Zambia National Farmers Union, Zambia Association
of Manufacturers, NGOs

3. Government Agencies
Examples here include the Central Statistical Office, Government printers,
Government Gazette and occasional publication by ministries. These sources
carry information o population, income, housing, agriculture, industrial and
commercial sales, employment, etc.

4. Syndicated Services
A wide array of data on both consumer and industrial markets is collected and
sold by commercial organizations. The information sought comprises retail sales,
household purchases, media usage, attitudes, knowledge and behaviours.

This information is collected by means of surveys, audits (store, product or retail

distribution outlet panels.)

5. Directories
Directories are a useful source of information on topics of study. One could
conceivably have a “Directory of Manufacturers in Zambia”, a “Directory of
Telephone Subscribers”, or the “Business Telephone Directory” – yellow pages.
Directories are commonly found in libraries.

6. Other Published Sources

Included in this categories are periodicals, books, dissertations, special reports,
newspapers and the like that may contain information relevant to marketing
decisions.
 A library is a useful and perhaps a necessary source.

7. External Experts
These are individuals outside the research firm whose job provides them with
expertise on the subject being studied. These include university researchers,
government officials associated with trade, commerce and industry, financial
analysts, editors and writers for trade and business publications and distributors.

8. Databases
There are many international databases to which one may get linked to access
information outside one‟s country.

9. Foreign Government Sources

This is usually available at embassies of foreign governments. The US and UK
stands out in this regard. Zambia through trade missions and embassies provides
information on investment opportunity in Zambia.

10. International Political Organizations

Here one might include the UN and its various agencies, the World Bank,
COMESA and SADC.

PRIMARY DATA
For the reasons that we discussed earlier, it may not be possible to use secondary data
for the study at hand, in which case we must generate information for the study at
hand. This information can be of the following types:

A. Demographic/ socioeconomic characteristics

Marketers are often interested in the socioeconomic characteristics such as
age, education, sex, income, marital status, occupation, or social class. We
are often interested in how these characteristics impact on the marketing
 phenomenon. For instance, we might be interested in how consumption of
a product is influenced by age, education, marital status and so forth.

B. Attitudes/ Opinions
Marketing managers not only want to know the level of consumer
awareness of a firm or its products but also the attitudes or opinions held
about them. Attitudes or opinions are dispositions towards something. It is
often assumed that a favourable attitude (disposition) will lead to an actual
purchase. It is principally for this reasons that marketers are interested in
collecting information about attitudes/ opinions.

C. Purchasing Intentions
Marketers are interested in people‟s anticipated or planned future
behaviour. A typical response model could be categorized as follows: will
definitely buy, will probably buy, not sure or undecided; and will
definitely not buy. A word of caution is in order here. Intentions do not
always translate themselves into actual purchase behaviour.

D. Activities or Behaviour
Many primary data-gathering studies are concerned with finding out what
consumers are doing. How much is spent in the retail supermarket? Who
does the shopping? When is the shopping done? What brand of soft drink
is bought?

E. Knowledge
We also might be interested in what respondents know or do not know
about an object or phenomenon. For instance, the effectiveness of an ad is
usually measured by how much can be recalled about the ad or its
message.
 F. Motivation
This refers to a drive or an urge to do something. It is, in marketing
circles, the “why” of consumer behaviour.

It is an impulse or an inner state that directs behaviour towards goals.

Marketers are interested in determining why people behave as they do. If
we understand a person behind a behaviour, we understand behaviour and
in turn influence future behaviour

Methods of Obtaining Primary Data

There are basically two ways of obtaining information. One is to ask information and the
other is to observe the phenomenon of interest. The list of questions we ask is known as a
questionnaire.

QUESTIONNAIRE DESIGN
The primary concerns in the design of the questionnaire are completeness and accuracy.
A good questionnaire must collect all the intended data and must do so as accurately as
possible. While there are no rules about a good questionnaire construction, some
guidelines are worth bearing in mind when designing a questionnaire.

1. Brief
The brief summarizes the boundaries and constraints of the proposed study. All
the people concerned with the study must be brought together and be made to
appreciate the following:

 the purpose of the study;

 the sponsors of the researcher and their specific interests;
 the areas to be investigated;
 the intended completion date for the study;
  to whom the completion of the exercise is to be entrusted, that is, by in-
house, outside consultants, a trade associate or a combination of
researchers;
 whether or not a study is to be a quick “broad brush” on an in-depth
investigation

2. Decision about Question Content

Need for the data asked by a question
A useful guide here is for the researcher to ask himself how the data generated by
a question is going to be used. This necessarily means that the questions and the
responses they elicit must be within the context of the objectives of the study and
that the recommendations will address the concerns of the sponsors of the study.

Ability of the Question to provide Data

Is the question on its own sufficient to generate the information required, or is it
necessary to have two or more questions? An example of lack of sufficiency often
occurs in studies involving attitudes. Marketers are not only interested in whether
or not customers have a favourable or neutral attitude, but are additionally
interested in the intensity or degree of customer‟s attitudes.

Ability of Respondents to answer accurately

Inaccurate answers to questions arise when respondents are uninformed, forgetful,
or inarticulate.

Uninformed respondents are likely to leave a question unanswered, or worse,

falsely report knowledge of an issue. It is equally advisable to avoid questions that
tax the memory of a respondent. It might be useful in this regard for a respondent
to have a record they can refer to. A question such as “what groceries did you buy
last month”? can only be answered accurately if a record was maintained of
groceries purchased. Respondents must also be sufficiently articulate, or they
should not be required to answer questions which are difficult and complex.
 Willingness of the Respondents to Answer
A person may have the ability to answer a question accurately but be unwilling to
answer it. Refusal may arise either because the information sought is of personal
nature, embarrassing, sensitive or reflects on prestige. For instance, generally,
people tend to be unwilling to provide answers to questions about their alcohol
consumption, sex habits, or private and socially unacceptable indulgencies.

Some of the ways of overcoming the problem of non-co-operation include the

following:

 Hiding a sensitive question in a group of other, more innocuous, questions.

 Assuring the would-be respondent that the behaviour or attitude in question is

not unusual before asking the specific question. For instance, those who might
not wish to reveal their drinking habits might feel assured by a statement to
the effect that “most studies indicate that one of every four Zambians drink
alcohol”.

 Phrasing the question in terms of others and how they might feel or act. For
example, “do you think that most people drink in Zambia?”

 Stating the responses in terms of a number of categories that the respondents

may simply check. In the case of respondents unwilling to reveal their exact
age, one might simply ask the respondent to tick in which of the following age
category they fall: 20-29 30-39 40-49 50-59 60+

3. Decisions about Question Format

Questions can take two forms: they can either be of the open ended type or the
fixed alternative type.
 Open ended questions leave a respondent free to offer any replies that seem
appropriate. No indication of the required reply is given. For example:

 Why do you buy this brand of soap?

____________________________
____________________________
____________________________

 What do you think about beer?

____________________________
____________________________
____________________________

The fixed alternative questions indicate the choice of response that is anticipated.
This format essentially restricts a respondent to a predetermined set of possible
responses. Respondents are asked to choose the alternative that most closely
corresponds to their position on the subject. For example:

 Why do you buy this brand of soap?

It is affordable

It is well packaged

It has a nice fragrance

Advantages Disadvantages
Open ended  Permit a wide range of  Difficult to analyze in
Questions replies cases where large
  Enable questions to be quantities of data are
answered in amassed
considerable length  Require high skilled
 May solicit information interviewers
unknown to the  Seldom appropriate if the
researcher computer is to be used in
the analysis of data
Fixed –alternative  Easier to administer  Require special skills to
Question and respond to develop all possible
 More amenable to use responses to a question
of a computer  Allow little detail to be
 Easier to analyze and given in the replies
interpret  May introduce bias by
exposing to all possible
responses
 May not capture
respondent‟s true position

Open ended response questions can be successfully used when a study involves a
small number of people. Closed response questions are most appropriate when a
large population has to be studied. They can also be used in screening respondents for
more detailed questioning.

4. Question Phrasing
Question phrasing is about conveying the researcher‟s meaning in a way that will
the respondents to capture the same meaning. Useful guidelines include:
 Use simple words in a language that will be readily understood by the
target respondents.
 Avoid using words with double meaning. For example, the word,
“plant” in the question, “who is in charge of the plant?” could be
interpreted in two ways: a plant could be understood to mean a factory
or vegetation!
 Avoid using words which are ambiguous or imprecise, such as “often”,
“sometimes”, “regularly”, “never” or “occasionally”
  Avoid using leading questions. These are questions which suggest
what the answer should be. A question stated “Many people prefer
Roan Antelope mealie-meal. Which one do you prefer?” will give the
impression that the minority position is out of line and may lead to a
respondent to side with the majority investigation
 Avoid double barreled questions. These are questions which call for
two or more responses and tend therefore to create confusion for the
respondent. An example might be: “How do you assess Shoprite on
price and quality?” or “How many times in a week do you shop at
shoprite and what items do you buy?”

5. Question Sequence
The order in which questions are asked is important. Questions should be asked in
a logical flow and in a way that allows the respondents to retain interest through
out the interview. It is therefore useful to understand the broad categories of
questions that comprise a questionnaire. The first types are introductory questions.
These attempt to define the respondents as far as practical or relevant and may
include personal characteristics such as:

 Respondent‟s name
 Respondent‟s sex
 Respondent‟s marital status
 Respondent‟s age group
 Respondent‟s profession
 Respondent‟s address
 Company name and address
 Company‟s major products
 Type of ownership
 Number of employees, etc
Introductory questions additionally serve to “warm up” or establish rapport with
the respondent. They should therefore not antagonize the respondent. In this
regard, in the case of sensitive and difficult questions, they may be placed at the
end of the questionnaire to avoid situations where the respondents are turned off
and become uncooperative.

The second type of question deal with the subject of the study as indicated in the
brief. Generally, these should be placed after the introductory questions. Where
there is potential danger of respondents becoming uncooperative because of the
nature of any introductory question, it is prudent to ask questions that deal with
the subject of the study prior to asking classification questions. The following are
useful tips in maintaining logical flow of questions:

 Use a funnel approach. Start with broad questions and progressively narrow
down the scope. For example:
 What improvements are needed in service policy in XYZ Ltd?
 How do you like the quality of service in XYZ Ltd?

 Using the funnel approach, the latter question should precede the former
question because it is more general than the former question.
 Place difficult or sensitive questions late in the questionnaire. If respondents
feel threatened or find it difficult to figure out the meaning of the question,
they are unlikely to proceed answering the rest of the questions.
 Use the branch questions diligently. This approach is used to direct
respondents to different places in the questionnaires, based on the response to
the question at hand. For example:

Yes No
i. Do you wear a tie or a cravat?
(If “no”, skip question iii)
 ii. Indicate how frequently you wear the tie

At least once a week

At least once a month

At least once in three months

iii. Are you likely to wear tie in the foreseeable

future?

6. physical characteristics of the questionnaire

The physical appearance of the questionnaire is important because, among other
reasons:
 it affect the acceptance of the questionnaire
 it affects the accuracy of the replies that are obtained
 It facilitates the processing of the replies.

An important aspect of the physical characteristic of the questionnaire is the cover

letter. The following are useful points to bear in mind:
 use personal communication: “dear Mr. Simasiku” rather than “dear
Sir”
 Make it clear at the very beginning that you are asking for a favour.
“Will you do us a favour?”
 Stress importance of the research project and its purpose.
 Appeal to the vanity of recipient by stating his importance.
 Indicate how the recipient may benefit from the research.
 Emphasize that completing the questionnaire will take only a short
time.
 Provide a stamped reply envelope (for mail questionnaire).
 Guarantee anonymity and confidentiality of answers.
  Offer to send reports on the results of survey.
 Indicate your appreciation.
 Supply your credentials or authority behind the project.
 Enclose a token of appreciation
 Be brief.

 Space the questions from each other and avoid question crowding.
 Number the questions, especially when branching questions are employed.
 It is advisable to have questions on one side of the paper and answers on the other
side of the page; for example:

Question Tick as appropriate

o What is your company‟s vision? ___________________________

___________________________
___________________________

Yes No
o Does your company have a formal plan?

Tick as appropriate
o Indicate which of these factors are crucial
in the provision of service

Order processing

Price

Delivery

Quality of Product
  Use good quality paper.
 If possible, indicate the address to which the questionnaire has to be
returned. This should be positioned at the end of the questionnaire
 The respondent should be thanked for his help in the study.
 If there are rewards for this cooperation, this should presented promptly to
avoid any possible ill feeling
 If it is necessary or practical, advise respondents when the results of the
study will be ready and offer to avail them to the respondent.

7. the Pre- test

It should not be assumed that a questionnaire will be appropriate for final
administration. A pre test serves to identify and correct unforeseen errors or
problems in the construction of the questionnaires.

METHODS OF ADMINISTERING A QUESTIONNAIRE

Once a questionnaire has been constructed, a decision must be made about how it is
going to be administered. The following are the major methods of administering a
questionnaire:

1) personal Interview
This involves face-to –face communication between the researcher (interviewer) and
the respondent (interviewee). Its relative advantage and shortcomings in comparisons
to other methods are:

Advantages:
 It is accurate because it affords the researcher the opportunity to explain or
elaborate points of difficult to the interviewee.
 It enables the researcher to collect as much information as possible, provided,
of course, if he is able to meet the expenses of interviewing the number of
people he has targeted.
  It enables the researcher to secure the highest response rate as he is personally
present to secure the response.
 It is quite flexible as it permits both the researcher and the respondent the
opportunity to adjust the pace and tempo of the interview.
 It affords the researcher some control over the sample and the interview.

Disadvantages:
 It can be quite expensive if research assistants have to be used; expense
related to training, travel, subsistence, lodging and renumeration can be
substantial.
 It is rather slow.
 Anonymity is rather compromised by the face-to-face interaction and is
therefore unsuitable for colleting sensitive information.

2) Telephone Interview
This involves asking and responding to questions over the telephone
Advantages:
 It is quite fast.
 It is in the intermediate range with regard to cost, accuracy and anonymity.

Disadvantage:
 Coverage may be affected because not every one has a phone.
 Response rate may be adversely affected by the “unsolicited” nature of the
phone call.
 Amount of data that can be collected will probably be on the low side

Table: an overview of methods of administering questionnaires

Method
Criteria Personal Telephone Mail Interview On-line
Interview Interview Interview
Cost
Accuracy
Speed
Amount of data
colleted
Response rate
Flexibility
Control
Perceived
anonymity
Obtaining
sensitive
information
Potential for
interviewee
bias

3. Mail Survey
In this method, the interviewer neither come face-to-face with, nor speaks to, the
respondent; rather a questionnaire is posted to the respondent and would normally be
mailed back to the interviewer upon completion.

Advantages:
 It is least expensive
 Appropriate for ensuring anonymity.

Disadvantages:
 Least control of sample and interview
 Response rate tend to be slow

4. On-line Interview
This involves conducting interviews electronically.
Advantages:
 Quite fast
 It is also flexible
 Disadvantages:
 Quite expensive at least initially
 Confidentiality is often compromised,
 There is limited control of sample.

OBSERVATION
A second method of observing primary data is by observing the phenomenon of interest
according to prescribed rules given the objectives of the study.
Condition under which Observation method is appropriate
 The behaviour being observed must be overt or external. It cannot be applied to
measure inert or internal behaviour, such as motives, attitudes, beliefs or
knowledge.
 The behaviour must be repetitive, frequent or predictable. It is not appropriate to
apply when the phenomenon to be observed occur rarely or the occurrence is
unpredictable.
 The behaviour must cover a relatively short time. Decision-making process do not
lend themselves to observation because their time span is longer rather than short

Types of observation methods

 Natural versus Contrived Observation
Natural observation occurs in a natural setting. In contrast, contrived observation
is induced by human beings. For example, a natural observation might involve
observing shopping on randomly chosen day: who is buying and what products
they are buying; how much of the product is being bought and at what frequency;
and whether or not shoppers look at price tag before selecting an item. The
observation will be contrived if the researcher determines the experimental
situation by, for instance, organizing and making a group of products available for
purchase; selecting who will take part in an experiment; or creating an artificial
shop. In general, a contrived observation occurs when there is self-consciousness
of the experimental situation.
 Open versus Disguised Observation
In open observation, the respondent is aware of being observed; in disguised
observation, the respondent is not made aware of being observed. When a store
detective openly watches shoppers and makes no secrets of watching shoppers as
they go about their shopping, this is construed as open observation. On the other
hand, if he is positioned in a place here shoppers cannot see him, or hidden
cameras are used to record shopping, this can be regarded a disguised observation.

 Structured versus Unstructured Observation

In a structured observation, what is to be observed is predetermined and
anticipated; in an unstructured observation, what is to be observed is not
programmed nor can it be anticipated. An exploratory study is an example of an
unstructured observation because it is premised on collecting data about an
unknown phenomenon. When one observes a phenomenon under prescribed
guidelines, this can be construed as structured observation. For instance, at job
interviews, a structured observation is when panelists are given criteria against
which to assess a candidate, such as confidence, personality, etc.

 Direct versus Indirect Observation ( Trace analysis)

Direct observation refers to the observation of current behaviour. In contrast, pat
behaviour can be observed by examining trace of it. In the case of food, direct
observation entails observing what is being eaten; indirect observation would
refer to checking physical traces of food that was eaten in the past, by looking at
or thrown away food containers or packages.

 Human versus Mechanical Observation

Mechanical observation is when mechanical devices are used as opposed to
human beings making the observation. To detect pilfering in a store, human
observation is when a tore detective is deployed, and mechanical observation is
when video cameras are used to film shoppers.
 UNIT SIX: SAMPLING PROCEDURES

Sampling is a common practice in our daily lives. All of us, at one time or another, draw
conclusion about a large group on the basis of the sample: we “taste” Munkoyo before we
purchase a cupful of the drink; we try on a shirt or dress before actual purchase; or sip a
cup of coffee. In all these instances and many others; an attempt is made at studying a
subgroup and using the results so obtained to make conclusions about the large group.

Marketing information may also be obtained through sampling procedures. Thus, instead
of collecting information from a universe, information is always obtained from the
subgroup of the universe and inferences made about the population on the basis of sample
information.

1. WHAT IS SAMPLING?
Sampling is the process of collecting information from a subset of the population of
interest and making conclusion about the population on the basis of sample
information

2. REASONS FOR SAMPLING

A census involves counting, measuring or obtaining information from every element
of a designated universe. A sample entails counting, measuring, or collecting
information from a part of the target population. While a census should be a preferred
way of collecting data, it may not be feasible or suitable in every situation. The
following are some of the reasons a sample, rather than a census, may be preferred:
a) Time Constraint
In many marketing situations, information may be produced urgently. For
instance, competitive pressures and shifting customer demands may require a
company to respond quickly and decisively. In such instances, collecting
information through a census would take so long as to undermine a firm‟s
ability to react in a timely manner to counter the threats to its existence. Thus,
a sample would be preferred to a universe.

b) Financial Constraint
Limited resources can force a firm to deal with a subgroup of a universe
instead of obtaining information from all elements in a universe. More
resources would be spent in dealing with a universe than a part of the
universe. For instance, if the population of interest is dispersed over a large
geographical area, requiring a researcher to travel over long distances,
sampling the universe is a cheaper way of obtaining information.

c) Universe Size
Even where there is no time or financial constraint, the sheer size of the
universe would make it impractical or impossible to reach every individual or
unit. Some people may be located in inaccessible places, while others would
be difficult o trace. A sample makes accessibility easier than a universe.

d) Destructive Nature of Sampling

Where sampling involves the actual destruction of an item, sampling is
preferred to a census. For instance, in quality control situations involving
food, on cannot sample every item. Rather, a sample is tasted to establish
whether the food is of acceptable standards instead of eating all the food.
 e) Sufficiency of an Approximation
In some situations, an exact value of the population parameter may not be
necessary. Sample information may suffice to enable management to make a
decision. For instance, in deciding on what production levels should be used,
it is not necessary to know the precise amounts which will be consumed. On
the contrary, production levels are based on estimates of consumption, in
which case an exact description of a population characteristic or an exact
value of a population parameter may not be necessary.

f) Accuracy
In some cases, sample results tend to be more accurate than those obtained
from a census. One is more likely to make errors with large populations than
with a sample. In the interest of greater accuracy, it may be prudent to deal
with a sample rather than a census.

3. SAMPLING PROCESS
The sampling process consists of the following steps:
 Defining the population
 Specifying the sampling frame
 Specifying the sampling unit
 Specifying the sampling method
 Determining the sample size

3.1 Defining the Target Population

A population is defined as the entire group of persons, events, or objects of interest to
the researcher. To be complete, a population must be defined in terms of the
following dimensions:
 Element  A unit of the population, such as, an individual
student at CBU
Sampling unit  The means of representing the elements of the
population, for example, hostels, schools.
Extent  Bought laptop computer
Time  During the 2007-2008 academic year.

3.2 Specifying the Sampling Frame

A sampling frame is a list of items in the universe from which a sample is selected. If,
for instance, three students are to be chosen to represent the CBU student body on an
overseas study tour, a sampling frame is a list of all CBU students from which the
three students will be selected. Such a list is useful in the selection of a sample
because it facilitates an objective and unbiased approach to the selection of sample
items. In the absence of an exhaustive list of items eligible for selection, one cannot
be entirely sure whether certain segments of the population were given a chance to be
selected during the selection process. A sampling procedure which consistently
ignores certain parts of the universe cannot be said to be representative. A sampling
frame ensures that all items in the universe have a known chance of being selected‟

3.3 Specifying the Sampling Unit

The sampling unit is a basic unit containing the elements of a population to be
sampled. It is the initial contact point that contains the elements to be sampled. If the
element of the population can be contacted directly, a sampling unit is identical to the
element. When the element of the population cannot be identified, the sampling unit
comes into play. For instance, it may be difficult to identify individual students other
than by going to their hostels; in such an instance, the hostel become an initial contact
point and is therefore a sampling unit. Another example is that of housewives, where
a sampling unit might be the households.
3.4 Selection of Sampling Method
Having determined the sampling frame, the next point of consideration is how items
in the universe are selected for inclusion in the sample. Probability methods of
sample selection are those in which every item in the universe has a known chance of
being selected for the sample. In other words, the selection of sample items is left to
chance and therefore independent of possible manipulation by the person making the
study. Non-probability sampling methods are those which do not provide every item
in the universe with a known chance of being included in the sample. That is, there is
an element of subjectivity since the selection of sample items is not independent of
the person making the study.

3.4.1 Types of Samples

Consider a population of 1000 students from which a researcher wishes to pick 100
students. Probability sampling is when the researcher does not in any way influence
how the 100 students are selected. Such a situation can occur when each of the 1000
students has a good chance of being selected as the next student. On the other hand,
when some students have systematically a higher or lower probability of being
selected than other students, the selection is said to be researcher-determined and the
selection procedure is said to be non-probability selection method because population
elements are not selected on the basis of chance.

Probability Samples Non-probability Sample

Simple random sample Convenience sample
Stratified random sample Quota sample
Cluster sample Judgement sample
Systematic sample

3.4.1.1 NON- PROBABILITY SAMPLING PROCEDURES

In probability sampling procedures, the basic premise is that all the elements in the
universe have an equal chance of being selected. In practice, probability sampling
procedures are rarely used. For one, in most situations, it is impractical, difficult or
impossible to have a precise list of all universe items. Even if a list were available,
some universe items could be unavailable or unwilling to participate in a study. In
many such situations, a researcher‟s practical option is to select a “typical” segment
for study. When a sample is selected by means which do not give every element an
equal chance of being selected, we have non-probability or researcher controlled
sampling. This is done in the following ways:

3.4.1.1.1 Convenience Sampling

Convenience sampling is when selection of items is determined by consideration of
personal convenience to the researcher such as selecting an item as opposed to
another is easier or less troublesome; is suitable in respect to the time or venue; or
serves the researchers interest or comfort. An example of convenience sampling
would be a selection of friends, neighbours, acquaintances; or conducting interviews
in safe or convenience neighbourhoods.

Despite these shortcomings, convenient sampling is suitable where a representative

sample may not be entirely necessary such as in exploratory research or in
questionnaire testing.

3.4.1.1.2 Judgement Sampling

Judgement sampling is based on the researcher‟s opinion of what constitute a
representative sample. It involves a researcher exercising his good judgement as to
who in the universe can constitute a sample element. A good example of judgement
sampling is the process of selection used in picking a group of “experts” to speak on
an issue.

3.4.1.1.3 Quota Sampling

Quota sampling involves, first, determining how many cells or segments the
population should have. Having determined how the population is to be divided, the
next step is to decide what percentage or proportion of sample items will be in any
one cell or segment. Finally, an arbitrary selection of items for each cell or segment is
made.

The diagram below illustrates these steps. The first step involves dividing the
population by age (18-25 years, 26-49 years and 50 years and over) and sex (male and
female). Thus, if a researcher wanted a sample comprising 300 people

Table 6.1: Illustration of Quota Sampling

Age
Sex 18-25 years 26-49 years 50+years TOTAL

Male 12% 21% 16% 49%

Female 13% 21% 17% 51%

TOTAL 25% 42% 33% 100%

Age
Sex 18-25 years 26-49 years 50+years TOTAL

Male 36 63 48 147

Female 39 63 61 153

TOTAL 75 126 109 300

He might specify that 25% should be in the 18-25 age category, 42% in the 26-49 age
category and 33% in the 50 plus age category. Assume that there are three research
assistants. Each of these research assistants is assigned a quota of respondents to
interview. If research assistant X was given the 18-25 age category, he would be required
to interview 75 respondents, i.e. 25% of 300 respondents. He would further be required to
interview 36 males and 39 females. However, which particular males and females he
selected would be left entirely up to the research assistant.

3.4.1.2. PROBABILITY SAMPLING PROCEDURE

Simple Random Sample
Simple random sampling rests on the premise that every item in the universe has a
known and equal chance of being selected for the sample. This assumes that all
universe items are known and can be individually and specifically listed as members
of the universe.

To select a sample from a universe, let us assume a universe of four-obviously an

oversimplified case- from which we wish to select a sample comprising any two
elements selected at random. One way of selecting such a sample is to have all
possible samples of size n=2 identified on chips and a draw made of one chip, that is:

Sample Chip Number

AB 1
BC 2
CD 3
AC 4
BD 5
AD 6

Thus, the four elements in the universe might be designated by the letters A, B, C and
D. the possible sample comprising two elements are designated as AB, BC, CD, AC,
BD and numbered 1 to 6 respectively. A draw of the chips would be made, and any
chip drawn would constitute the sample.

Note that this way of drawing a sample is premised on a complete and known list of
all population elements. Secondly, all items in the universe, or all possible samples,
have a known and equal chance of being selected. Lastly, the chances of selection
were not researcher determined. We can see from this simple example that the items
for the sample were randomly selected.

Another and, perhaps, more practical, way of selecting a simple random sample is
through the use of the table of random numbers. Again, the principal of complete and
known list of population elements must be obtained to enable one to assign numbers
to each population element. Thus, given the total number of items in the universe is
known, N, each item in the universe numbered serially from 1 to N. the net step is to
determine the size of the sample, n. then, beginning at any arbitrary point on the table
of random numbers, one begins to select those numbers between 1 and n in their order
of occurrence until a sample of size n is obtained.

The principal of randomness is satisfied because the digits in the table of random
numbers are not arranged, nor do they appear, in any systematic or discernible
manner. Assuming the starting point is chosen arbitrary, there is no way of
predetermining which numbers will be selected from the table. It is as if numbers
appear in the table by chance!

To illustrate the use of a table of random umbers, let us assume a population of 100
students from which we wish to select a sample of 30 students. We would number
each student from 1 to 100. a sample of 30 students would be selected by picking the
first thirty numbers in the table from 1 to 100. The entry point in the table would be
arbitrary, by entering the table at the randomly selected row and column. Having
established the starting point, we could next arbitrary choose in which direction to go,
either along some row or column- upwards, downwards, or horizontally. The first
 thirty numbers to be randomly selected would constitute the random sample. Note
that were the number appears more than once, it is only picked the first time.

Stratified Random Sampling

Simple random sampling is based on the premise that the population from which the
sample is to be drawn is homogenous. On the contrary, in many practical situations,
populations of interest tend to be heterogeneous rather than homogenous. For
example, a population of housewives can be differentiated on the basis of age,
income, social standing, religion, education and so on. Where it is considered that
such socio-demographic characteristics have a bearing on the variable being studied,
the stratified random sample is designed in such a way as to include sampling units
that posses the characteristics.

A stratified random sample involves the following steps:

1) A population to be sampled is divided (or stratified) into groups which are
mutually exclusive but all inclusive.

2) A simple random sample is then chosen independently from each stratum.

Although a simple random sample presupposes that every population element
has an equal chance of being selected, this does not necessary guarantee that
all characteristics will be equally represented in the final sample. A stratified
random sample ensures representation, though not equally, of all
characteristics obtaining in the population.

A stratifies sample bears some similarity to the quota sample only to the extent that the
population is divided into cells representing desired characteristics and that the sample is
premised to have those characteristics. The difference between the stratified and quota
sample is a material one, and this is in a way the sample elements are selected. A
stratified sample is the probability sample and selection of sample element is accordingly
not researcher determined as is the case with quota sampling.
Proportion versus disproportion stratified sample
A distinction can be drawn between a proportionate stratified sample and a
disproportionate stratified sample. A proportionate stratified sample contains elements in
the stratum in proportion to the relative number in the population, while a
disproportionate stratified sample contains elements in the stratum in disproportionate to
the number in the universe. For instance, in table X below, a population of 50, 000
persons has 10,000, 15,000 and 25,000 persons in the high income, middle income and
low income categories respectively.

The size of the proportionate sample for each category is relative to the size in the
universe, while a disproportionate sample for each category bears no relationship to the
size in the universe. Thus, 10,000 high income earners constitute 20% of the population.
Consequently, in a sample of 5,000, they will correspondingly constitute 20% of the
sample.

Table 6.2: Proportionate and Disproportionate sampling

Proportionate Disproportionate
Stratum Number Sampling Sampling

High income 10,000 1,000 500

Middle income 15,000 1,500 2,500

Low income 25,000 2,500 2,000

50,000 5,000 5,000

elements in proportionate sample. Middle income earners constitute 30% of universe

elements. Finally, low income earners constitute 50% of ample elements. In contrast, in
disproportionate sampling, we find that high income earners constitute only 10% of the
sample even thought they constitute 20% of the universe; middle income earners
comprise as much as 50% of the sample while they are only 30% of the population; and
low income earners comprise 40% 0f the sample while they are half low income earners
in the universe.

Disproportionate sampling on the higher side, such as for the middle income earners is
used when there is greater variability among items in the universe which comprise a
selected stratum. On the other hand, less than disproportionate sampling is used when
there is homogeneity among items.

Cluster Sampling
Cluster sampling involves dividing the population into mutually and exhaustive sub
group and then taking a sample of sub groups. If all the elements in the subgroup so
selected are studied, this is called one stage-cluster sampling; f a random sample of
elements is made from the selected sub groups, then this is called two-step cluster
sampling.

The diagram below illustrates the difference between one-stage and two-stage cluster
sampling. Suppose we have a universe comprising sixteen individuals from which we
wish to select four individuals. Under cluster sampling we would divide the population
into mutually and exclusive sub groups. In our example, these groups are denoted A, B. C
and D and each subgroup consist of four individuals. We would then proceed to select at
random a sample of sub group to study. Since we wish to study four individuals and each
sub group consists of four individuals, under one-stage cluster sampling, this would
involve selection of any one of the sub groups. This way, all the elements in the sub
group would be studied.

Under two-stage luster sampling, not only is a random sample of sub groups made but a
further selection of the four individuals must be made from the sub groups. Thus, we
might proceed by selecting randomly any two sub groups, and then randomly select two
individuals from each sub group.
Area Sampling
Area sampling is a type of cluster sampling. It is applied in instances where a detailed list
of universe items is not known, e.g. in a geographical area. Thus, blocks of households or
streets on which houses are located would be used as clusters. A random sample would
be selected by randomly selecting a street(s) and studying all households in the selected
street(s) of houses (one-stage area sampling) or, having selected a street(s) of houses then
selecting a sample of households from each pre-selected street(s) of houses (two-stage
area sampling). An illustration of area sampling is shown on page 32.

Table 6.3: Illustration of Area sampling

One-stage Two-stage Selected
Street Households Area sampling Area Sampling Sample in
Two
Stage

1st street 1234

nd
2 street 5678 5678 5678 5 6

3rd street 9 10 11 12

4th street 13 14 15 16 13 14 15 16 14 16 14 16

One-stage Two-stage

1. A population is divided into four streets 1. a population is divided into four streets

2. On each street are households 2. On each street are households

(Number may not be precisely known) (Number may not be precisely known)
3. A street is selected from the four streets 3. A Street is selected from the four streets
And all households on the selected but not all households are studied from
Streets are studied. the selected street. Instead, another
selection is made of two households from
Households on the selected street.

Systematic sampling
The five steps in systematic sampling are to ensure that a list of all universe items is
available. Second, the universe items must be arranged in some kind of order. To select a
sample of size n, one begins to select sample items at any interval. For instance, if there
were 26 households and we wish to –

A B C D E F
G H I J K L
M N O P Q R
S T U V W X
Y Z

Select a sample of 8 households by systematic sampling; we could arbitrary pick every

3rd house, as indicated in the example above. One can see that if rows represented streets
on which the households were located, a bias would (unintentionally) be created by
selecting four houses which are positioned at the corner. This may not be wrong in itself,
unless if there were some common characteristics among people positioned at the corner.

3.5 Determining the Sample Size

Since sample values are used to estimate sample parameters, the question of how accurate
and reliable sample values are becomes important. It is in this regard that sample size
becomes critical in sampling procedures. Generally, accuracy is greater with larger
samples but, larger samples are greater associated with high costs. In practice, reliability
and accuracy may be compromised so long as the sample is considered as representative
of the population in question. Thus, among methods that are used to determine the sample
size, it is only the traditional statistic method, for reasons that we shall see, that offer
some confidence in the ensuing sample estimate.

The following are some of the ways the sample size may be determined:

3.5.1 Arbitrary determination

There are no explicit consideration given either to the precision of the sample results or
the cost of obtaining the results. This method is exemplified by a statement such as “I
want a sample of 50, 100 or whatever….”

3.5.2 All-You-Can-Afford
This method is based on cost consideration, in particular the budget for the project. The
focus is on the cost of information to the exclusion of concern about its value in terms of
accuracy.

3.5.3 Historical evidence or precedence

This method of determining sample size is based on sample size that might have been
used for similar studies in the past.

3.5.4 Required size per cell

This involves basing the sample size on the sampling of analytical procedure. For
instance, in chi-square analyses, it is desirable that the counts per cell be no less than five
(5). In the case of quota or purposive sampling, the researcher might be guided by what
good judgement dictates should be the number of respondents either as a whole or in
strata of a sample.

3.5.5 Traditional Statistic method

The traditional statistical method for determining the size of the size of a probability
sample is given by the formula
 n = σ2Z2
A2

Where n = the sample size

σ = an estimate of the variance in population from which the sample is to

be drawn

Z = the desired level of confidence in the result

A = the desired level of accuracy, or the error from sampling that the
the research will allow.
This formula is based on the concept of a sampling distribution of a statistic. Assume the
statistic in question is the mean. The sampling distribution of the mean is the relative
frequency distribution of the means of all possible samples of size N. to illustrate the
sampling distribution of the mean, let us assume we have population of size N=4 and we
wish to draw all possible samples of size n=2 from this population. Assume that our
population consists of four individuals whose annual income, in million of kwacha, is as
indicated in figure 1

Figure 1: Income of A, B, C and D

Individual Annual Income

A 60
B 80
C 40
D 20
We readily see that the average income of these four individuals is K50million, that is,

M = 60+80+40+20
4

= 50
Assume we wish to draw a sample of size n=2 individuals and compute the average
income of all possible samples of size n=2 drawn from this population. The relative
frequency distribution of the means of all possible samples of size n=2 drawn from this
population would be as shown in figure 2
Figure 2: Income of A, B, C and D

Sample Annual Income

AB 60+80 =70
2

AC 60+40 =50
2

AD 60+20 =40
2

BC 80+40 = 60
2

BD 80+20 = 50
2

CD 40+20 = 30
2
2

1
30 40 50 60 70

A relative frequency of sample means shows the number of time a sample mean
occurred. It I thus a measure of the probability of occurrence of the sample mean value.
The relative frequencies are shown in the histogram in the lower part of figure 2. A
normal curve is shown in the same figure. It can be seen that the distribution is normally
distributed. The normal curve in the lower part of figure 2 is the sampling distribution of
the sample means.

In general, a sampling distribution of the means of simple random samples that are large
(that is, more than 30) has:

i. A normal distribution
ii. A mean equal to the population mean, that is:

70+50+40+60+50+30
6

= 50, 000
iii. A standard deviation called the standard error of the mean, that is equal to the
population standard deviation(s) divided by the square root of the sample size
(√n):

σx = σ
√n

It is important to note that the standard error of the mean is simply the standard
deviation of the sampling distribution of the mean. The only reason that the standard
error of the mean is called that instead of the standard deviation is to indicate the it
implies to the distribution of sample means and not only to a single sample or a
population.

A basic characteristic of a sampling distribution is that the area under it between any
two points can be calculated as long as each point is defined by the number of
standard errors it I away from the mean.

The number of standard errors a point is away from the mean is referred to as the Z
value for that point. For example, the area under one side of area under one side of
the curve between the mean and point that have Z values of 1.0, 2.0, and 3.0,
respectively, are as follows:

Z value (number of standard error) Area under the curve from the mean
to the point defined by the Z value.

1.0 0.3413

2.0 0.4772

3.0 0.4986
The area between to points under a sampling distribution is a probability. This, mean that,
the probability of the sample with a mean falling within the range of one standard error is
0.6826 (0.3412x 2) or 68 percent.

A visual examination of the sampling distribution of the mean in figure 2 shows that the
mean of a sample of size 2 is likely to be quite close to the actual population mean. The
shape of the curve indicates that most of the sample means are clustered near the centre
of the distribution, which also happens to be the true mean. But the mean of the particular
sample of size 2 could be any of the sample means in the distribution, with some of them
being a substantial distance from the true mean and the true value of the mean is the
sampling error. If we examine the formula for the standard error of the mean, σ x = σ
√n
, we can see that the standard error of the Mean becomes smaller as the sample size
increases. In other words, increasing sample size will reduce the potential sampling error
because as the sample increases, the sampling distribution becomes clustered more
closely around the true population value.

A point estimate of a mean is an estimate involving only a single value. If we wished to

estimate the average income of the four individuals in our example and used a random
sample of two (2) individuals to estimate the true (population) average income, the
sample mean computed from the two individuals would be a point estimate. It is quite
clear that point estimates based on sample mean are seldom correct as estimates of the
true mean. An interval estimate is thus more reliable as it consists of the range of values
with the probability that the interval will enclose the true value of the mean. This
probability is called the confidence coefficient and the interval is called the confidence
interval.

Here is how an interval estimate of a mean is applied. A sample is taken and a sample
mean is calculated. We know that this sample mean falls some where within the sampling
distribution, but we do not know at what location. We do know, however, that there is a
probability of 0.3413 (34.13%) that it lies within 1 standard error above and a probability
of 0.3413 (34.13%) that it lies within 1 standard error below the actual population mean.
We may, therefore, make an interval estimate that allows us to be 68.26% (34.13 + 34.13)
confident that the population mean (m) lies within the interval formed by the sample
mean ( x ) plus 1 standard error (σ x ) and the sample means minus 1 standard error, that
is:
x- 1 σ x ≤ M ≤ x + 1 σ x
the 68.26 percent is the confidence coefficient of the estimate

Example:
Assume that we wish to determine the average expenditure per month spent on food.
Assume further that the expenditure for food is normally distributed. Now suppose a
random sample of 30 individuals is selected and is found that the average amount spent
by these 30 individuals is K180, 000 per month, with a standard deviation of K14, 000.
What then is the true monthly expenditure on food?

Solution

 We want to estimate a population mean (that is a true average expenditure) that

we do not know from the sample mean that we know (k180, 000).
 The K180, 000 is a point estimate as it involves a single value. It can be used as
the best guess, or estimate, of the true average expenditure.
 From our discussion of the sampling distribution, we know that sample means are
clustered around the true mean. We would feel assured if this particular sample
mean of K180, 000 was one of those sample means clustered around the true
mean. But it could be any of the means located at a distance from the true mean.
This distance is the sampling error.
 An interval estimate will provide us a range of values with the probability that the
interval will enclose the true mean. This probability represents the level of
confidence we have in the sample estimate, and is known as a confidence
coefficient.
  If, for example, we wish to be 95.44 percent confident that the interval estimate
will enclose the true mean expenditure, then, the range of values will be specified
as follows: x- zσ ≤M≤x+zσ
√n √n
 Note that the expression z σ is the allowable error in our estimate
√n
 x is the sample mean i.e. K180, 000
 z is the number of standard errors specified by the confidence coefficient. If you
go to the table of areas under the normal curve, the z value is 1.96 at the 95.44
confidence coefficient.
 s is given as the standard deviation
 the interval estimate is thus given as:

180, 000 – 1.96 14, 000 ≤M ≤ 180, 000 + 14, 000

√30 √30

174, 992 ≤M ≤ 185, 005

 We can be 95.44 percent confident that the true average monthly expenditure lies
between K174, 992 and K185, 008

Computing the probability sample size

You will recall that the formula for determining the probability sample size was given as:

n = σ2 Z2
A2
In our discussion of the concept of a sampling distribution, an estimate of the variance of
the population from which the sample is to be drawn was given by the standard error of
the distribution of the statistic: in the case of the sampling distribution of a mean, that is,
 σ x=σ
√n
The z value represents the number of standard errors from the mean as specified by the
confidence coefficient. The expression zσ is in reality the specification of the
√n
level of accuracy. If A is used to denote accuracy desired, then, A = zσ
√n
and solving for n, we have

n = σ2Z2
A2

Example:
The average advertising expenditure per year of the random sample of 50 firms is
K12million with a standard deviation of K1million.
How many firms must be surveyed to ensure that the estimate is not more than
K0.2million out at the 99% confidence level?
Solution
Sample size is given by n = σ2Z2
A2

Given that σ= 1 million

Z = 2.58
A = 0.2 million

Then n = (12) (2.582)

(0.22)

= 166
Calculation of sample size in calculation problems involving proportions
Marketers are sometimes concerned about the proportion of people who behave in a
certain way. For instance, an advertiser of a product may be interested about the
percentage of people who read his advertisement. In such a case, marketing research will
be dealing with the proportion and, consequently, with the sampling distribution of the
proportion.

The sampling distribution of the proportion is the relative frequency distribution of the
proportion (p) of all possible samples of size n taken from a population of size N. The
same basic reasoning used to determine thee sampling distribution of the mean applies to
the sampling distribution of the proportion for a sample random, thus has:

i. A normal distribution
ii. A mean equal to the population proportion (P)
iii. A standard error (σp) equal to σp = P(1-P)
n
We can now turn to an illustration of an interval estimate of a proportion, and to the
determination of a sample size.

Example
In order to evaluate the success of the new product, a company market tested the product
among 200 housewives of whom 60 favoured the new product.

a) How correct does this estimate the proportion of housewives who favour the
product at the 95% confidence level?
b) How many housewives must be interviewed in order to establish this proportion
within 2% accuracy at the 95% confidence level?

Solution to (a)
Estimating the interval estimate of the proportion of housewives who favour the new
product is given by the interval:
 p - z pq ≤ P ≤ p + z pq
n n

Where P = population proportion

P = sample proportion (60/200 = 0.3)
q = 1 – p (0.7)
z = 1.96
n = 200

Hence: 30 – 1.96 (30) (70) ≤ P ≤ 30 + 1.96 (30) (70)

200 200

24 ≤ P ≤ 36

Solution to (b)
Estimating the sample size is given by: n = z2 pq
A2

= (1.96)2 (30) (70)

22
= 2016

It will be observed that the formula for determining the sample size is derived from the
expression z pq, which specifies the level of accuracy in estimating the true proportion.
n
Hence, accuracy can be specified by the expression, z pq , and solving for n gives us the
n
formula for determining the sample size.
 TOPIC FOUR: FIELD PROCEDURE AND NON

SAMPLING ERROR

NON SAMPLING ERROR

In our consideration of sample size determination, we referred to sampling error as the
difference between a sample estimate and the universe value (or long-run average value
in repeated measurements). We noted further that this error arises purely from the process
of sampling. However, a difference between a sample estimate and a population
parameter may occur due to other reasons other than sampling. Such errors are known as
non sampling errors and they may occur because of errors in conception, logic,
interpretation of replies, statistical or arithmetical computation or simply mistakes in
reporting the results of the study.

TYPES OF NON SAMPLING ERROR

1. Non Observation Errors
These results from the failure to obtain data from parts of the survey and are of two types:
Non coverage errors occur when part of the universe was not included. This is
essentially a sampling frame problem which may occur in the manner of administering
the questioning method. Non-probability sampling as opposed to the probability
sampling, will not guarantee that every one is given an equal opportunity of being
interviewed. Typically, the researcher determines from whom information is colleted
when convenience, judgement or quota sampling are used. The method of administering
the questionnaire may also cause a non-observation error. Surveys by telephone
invariably exclude those who are not listed in the telephone directory, and mail surveys
may exclude those who are not listed on the mailing list. This type of error can be
minimized or eliminated by improving the quality of the sampling frame.
 Non-response errors arise where no response is secured from respondents included in
the survey. Examples of non-response errors are:

 Not-at-home- these refers to non responses because respondents were not

available at the time the interview were taking place. This type can be minimized
by making a prior appointment or calling back later.

 Refusals- these refers to those who refuses to respond to interview. The rate of
refusal depends on:
o Nature of respondent
o Auspices of the research
o Circumstances surrounding the contact.
o The nature of the subject, or
o The interviewer

Empirical results on reasons given for refusal to participate include:

o No time/ in a hurry
o Inopportune time/ timing inappropriate
o Not interested/ does not apply to me
o Do not want to be bothered/ invasion of privacy
o Believed it was a sales pitch
o A waste of time
o Questions were too personal

Possible remedial measures:

o Appropriate choice of interview in the case of nature of the respondent
o Sell the value of the research to the respondents
o Anonymity of both the sponsor and the respondent
o In the particular case of mail interview, remedial measures might include:
 Follow-up on the contacts
 Making a (passionate) appeal
  Provision of return stamped envelope
 Personalization of addressing respondents
 Providing incentives
 Ensuring that the questionnaire is not too long

2. Observation (field) Errors

Occur when those who have agreed to participate in the study refuse to cooperate or
give inaccurate information, or when there are errors in the processing of the data or
in reporting the findings. Because these errors occur during the course of an
interview, they are referred to as response errors. The Kahn and Cannel
interviewer-interviewee interaction model illustrates how response errors may occur.
The model postulates the following:

 Each person brings certain background and psychological predispositions to the

interview. Background characteristics include age, education, race, religion, sex,
socioeconomic status, e.t.c. these background factors affect reported responses.
 When the interviewer and the interviewee are of the same background, it
facilitates better communication and more information is more likely to be
obtained than if the backgrounds are different. This is easy to see when the
interviewer and the interviewee are, say, of different races.
 Thus, it is productive to match the background characteristics of interviewer and
interviewee as far as possible in order to ensure an effective interview.
 Background factors conditions psychological factors, such as perception,
attitudes, expectations, motives, e.t.c. in turn; psychological factors affect the
responses interviewers receive. For instance, the extent to which an interviewer
will probe a respondent, and the extent to which the respondent will respond to
probing, will depend on the attitude or perceptions of either or both parties. It is
therefore, important that these factors are not allowed to affect the responses.
This can be done by structuring the questionnaire and the manner it should be
conducted so that there is little room for attitude on the part of the interviewer or
respondent.
  Background and attitudes, perceptions, motives, e.t.c induces certain behaviours
in the interviewer and respondent. Interviewer behaviours which may lead to
response bias include:
o Errors in asking questions and more so in probing for more information.
o Errors in recording the responses.
o Errors occasioned by cheating.
 On the part of respondents, the impact of Behavioural factors may manifest itself
in the way respondents respond to questions: are the responses, for example,
inadequate or inaccurate?

3. Office errors:
Non sampling errors are not restricted to data collection. They may also be caused by
lapses in editing, coding, tabulating, and analysis of data.

Figure 6.1: The Kahn and Cannel Model of Bias in the Interview

Interviewer Interviewee
Background Background
Characteristics Characteristics

Psychological Psychological
Factors Factors

Behavioural Behavioural
Factors Factors

Source: Robert L. Kahn and Charles Cannell, The Dynamics of Interviewing (New York:
John Wiley & Sons, 1957) in Gilbert Churchill and Dawn Lacobucci, op.cit. pg 542
 TOPIC EIGHT: ANALYSIS OF DATA

Introduction
The raw data that is contained in a questionnaire is in an unorganized form. It must be
drawn out and be presented in a form which facilitates understanding as per objectives of
research. The purpose of data analysis is, therefore, to draw meaning from the raw data.
This process starts with editing.

1. EDITING
Editing involves the inspection and, where necessary, the correction of each
questionnaire or observation form. The basic purpose of editing is to impose some
minimum standard on the information collected. This involves checking for the
following:

1.1Completeness
The check for completeness involves checking for any missing data. This can be
manifested by a blank to a specific question. This could be due to an oversight or lack
of knowledge about the issue the respondent was asked about. The immediate
corrective measure is to seek clarification from the field worker. Other ways of
handling missing data comprise:

1.1.1 The Individual Approach- this involves predicting the respondent‟s

probable response. For instance, age can be predicted from the physical
inspection of the respondent, or from other information provided by the
respondent. If the respondent has indicated elsewhere that he/she has
retired from employment and has grandchildren, it can be assumed that the
respondent is at least 50years old since the retirement age in Zambia is
55years.
 1.1.2 The group Average Approach- this involves taking the average of those
who answered and substituting it for a respondent who did not answer.
The caution here must be that the non responses be few in order not to
distort the group average approach.

1.1.3 Arbitrary Substitution- this involves randomly assigning a value in place

of the missing value.

1.1.4 Leave the question blank and record the answer as a non response- this
must be taken into account when interpreting data and making inferences
about the population.

1.2 Consistency
Inconsistency is said to obtain when any two or more answers contradict each other
when they are supposed to move in one direction. For instance, if a respondent
indicated that he prefers Colgate on one part of the questionnaire and later indicated
that he has never used it, the researcher would find it a problem to attach any
meaning to what the respondent meant. Such an inconsistency may indicate errors
in the collection or recording of a response, or ambiguity in the mind of the
respondent regarding the meaning of the questions.

1.3 Uniformity
Lack of uniformity is when responses are not recorded in the same units. For
instance, if the study is aimed at finding the size of an establishment in terms of
sales, and the respondents indicate the size of the firm in terms of profit, the
response base is not uniform.
1.4 Legibility
Written responses, including symbols or abbreviations used, must e legible or
decipherable. Misspellings must be corrected at this stage.

1.5 Comprehensibility
It is important that responses are understandable to all concerned. One must avoid
situations where only the field worker comprehends a response.

2 CODING
This is a procedure for assigning symbols to responses in order to facilitate
categorization of the responses. The first step in coding is to specify the category of
responses. For instance, sex can be categorized into male and female: education can
be categorized into Grade VII, Grade XII, Graduate; economic status can be
categorized into rich and poor and so on. The second step in coding is to assign a
symbol to each response category. A symbol is usually a numeral. In our example,
the coding could look like this:

An example of Coding
Response Response Symbol
Variable Category
Sex Male 01
Female 02
Education Grade VII 11
Grade XII 12
Graduate 13
Economic Status Poor 21
Rich 22

Coding is relatively easy for closed questions, but relatively difficult for open-
ended questions. In the closed question format, responses are by design categorized.
In the open-ended questions, respondents are free to express an answer in any way
they prefer and this makes it difficulty to anticipate what kind of answers will be
given.
3 TABULATION
Tabulation is a frequency distribution of a variable. Thus, tabulation involves
categorizing responses and counting the number of cases which fall into each
category. The tabulation may take the form of simple (one way) tabulation or a
cross tabulation.

Simple or one-way tabulation

Involves counting or studying a single variable is studied. Table 6.1 is simple
illustration of a simple tabulation. It is simply a frequency distribution of the
consumption of bread by students.

Table 6.1 Bread consumption

Number of loaves Number of Students
0 2
1 3
2 5

Other uses of one-way tabulation are:

 To show the degree of response to an item- once responses are tabulated,
one can readily determine the degree of response or non response to an item.
 To locate errors, such as when the number of responses do not tally with the
sample size.
 To locate outliers unusual, out of the ordinary cases.
 To determine the empirical distribution of the variable under study. For
instance, the following can be derived from a simple tabulation:
 Graphical presentation of the phenomenon of interest
 Histogram
 Cumulative distribution function

 Facilitate computation of summary measures

(a) Percentages- these show the relative magnitude of items.

Loaves consumed number percentage

0 2 20%

1 3 30%

2 5 50%

10 100%

It is advisable not to use percentages if the base is small. Base should preferably be
at least 50

(b) Averages-
These are measures of central tendency and can be in the form of an arithmetic
mean, a median or a mode. Caution is called for in the application of averages,
particularly in the following situations:

Slanted statistics- all the above measures of central tendency qualify as

“averages” but may in face be different if the distribution is skewed. Because of
the different averages involved, there is always a danger of slanting the statistics
in the given direction.

The non existent average- there are some unusual distributions which do not have
averages, such as a mean or median. For instance, per capita income distorts the
slant in income distribution between the rich and the poor. Per capita income as a
 measure of average income tends to obscure significance differences in market
segments.
Nobody is average- the average represents a single point on a line, and a point has
no area. The probability of any person being average is therefore zero. The point
is significant in practical situations. What does average imply? In our example,
nobody consumes 1.3 loaves of bread: they either consume 2, 1 or nothing!

Spurious precision- because of the way averages are computed, one may come up
with too precise a figure which has little practical significance. For instance, an
average consumption of 1.3 loaves of bread per student is arithmetically correct
but it is spurious and has no practical significance.

(c) Measures of dispersion.

These measures focus on the distribution of items, or the spread of items in the
universe. The focus is thus on differences among items in a universe. Such
measures comprise:

Range- this refers to the difference between the highest and the smallest values,
such as the difference between the highest and the lowest paid.

Inter-quartile- range- refers to the range of the middle 50 percent

Average deviation- this refers to the average of the deviations of the various items
in the group from the mean. Absolute numbers are used to eliminate negative
values.

∑ xi – u

Standard deviation- this is the same as the average deviation except that to
eliminate negative values, we use the square of the deviations of the individual
values from their arithmetic mean.
 The importance of standard deviation is that it is one of the parameters of the
normal curve. In practice, this means that if we know that the population is
normally distributed and we know the standard deviation, we can determine what
percentage of the population falls between selected intervals (as measured in
standard deviations)-viz:

68% of the population will lie within one standard deviation of the mean.
95% of the population will lie within two standard deviation of the mean.
99% of the population will lie within three standard deviation of the mean.

Cross-tabulation
Cross-tabulation is used to examine the relationship between two or more variables. It
involves studying two or more variable simultaneously. In many marketing situations, we
are often forced with decisions that require an understanding of how variables relate to
each other. For instance, the relationship between demand and price has implications for
pricing decisions; the relationship between demand and advertising raises questions about
how much should go into advertising, to which target group advertising should be
directed;; how advertising should be conducted; and when advertising should be carried
out. Cross tabulation shows how two or more variables might be related.
Table 8.1 shows the raw data on the purchase of electronic equipment by 15 students.
The students are shown by age and background. The following are some of the ways this
data may be used.

Table 8.1: Purchase of Electronic Equipment among 15 Students

STUDENT AGE BACKGROUND MODE OF PURCHASE
A 16 Urban Yes
B 21 Urban No
C 18 Rural Yes
D 16 Rural No
E 20 Urban Yes
F 19 Rural No
G 17 Urban No
H 16 Urban Yes
I 22 Urban Yes
J 21 Rural No
K 20 Urban No
L 22 Rural No
M 19 Urban Yes
N 18 Rural Yes
O 20 Urban No

Table 8.2: purchase by age

Purchase
Age Yes No TOTAL
Below 20 5 (63%) 3 (37%) 8 (100%)
20 and above 2 (29%) 5 (71%) 7 (100%)

Table 8.2 shows the relationship between age and purchase. Sixty three (63%) percent of
the students below the age of 20 bought the equipment, while thirty-seven (37%) percent
did not purchase the equipment. Among the older students, that is, the students twenty
years and older, twenty-nine (29%) percent bought the equipment, while seventy-one did
not purchase the equipment. The data suggest that the younger students are more likely to
buy than the older students.

Table 6.2: Purchase by background

Purchase
Background yes No TOTAL
Rural 2 (33%) 4 (67%) 6 (100%)
Urban 5 (56%) 4 (44%) 9 (100%)

Table 6.3: Purchase by age and background

Background
rural urban
Age Yes No Total Yes No Total
Below 20 2 (50%) 2 (50%) 4 (100%) 3 (75%) 1 (25%) 4 (100%)
20 and above 0 (0%) 2 (100%) 2 (100%) 2 (40%) 3 (60%) 5 (100%)
Table 6.2 presents data showing the relationship between background of the student and
purchase. Sixty-seven (67%) percent of the students with the rural background did not
buy the equipment, while fifty six percent with an urban background did purchase the
equipment. The data suggests that the students with an urban background are more likely
to purchase the equipment than students of a rural background.

Table 6.3 shows the interaction of age and background in the purchase of equipment.
Among the students below 20 years, seventy-five percent of those in the urban areas
bought the equipment, while their rural counterparts were evenly divided: fifty percent
purchases and fifty percent did not purchase. Among the older students, 100% of rural
students did not purchase, while sixty percent of urban students did not purchase. The
data thus, seem to suggest purchase is related to age and background, and that, more
specifically, the older students with rural backgrounds are not likely purchasers, while the
younger students of urban backgrounds are likely purchasers.

Possible scenario of introducing a third variable

Human behaviour is influenced by many factors, some more influential than others. Thus,
in an examination of how two variables are related, one may not get the full picture of the
marketing phenomenon. In our example, the initial relationship examined showed the
relationship between age and purchase behaviour. But with the introduction of the third
variable-background- the relationship was amplified and qualified: it is the younger
people with urban backgrounds who are the best prospects.

The initial relationship has therefore been retained with the introduction of a third
variable. In some situations, the initial relationship may turn out to be spurious with the
introduction of a third variable. For example, if e confined our analysis to students who
are below 20 years and have a rural background, we notice that the initial relationship
between age and purchase disappears: that is, there is no difference in purchase behaviour
among students with a rural background in this age group. In still other situations, a
relationship may be established with the introduction of a third variable where none
initially existed.
ESTIMATION AND HYPOTHESIS TESTING
There are clearly many ways of analyzing data. What specific meaning we wish to draw
out will depend on the research objectives. There are nevertheless three broad categories
of data analysis.

1. Estimating a Parameter of Interest

Problems of estimation involve estimating a variable of interest from sample data. For
instance, we might be interested in determining the average income of a group of people,
or in establishing the proportion of people who buy a particular product. In estimation,
we use sample data to form an opinion about the magnitude of a parameter. That is, in
trying to determine the average income of the population in question, we would draw a
sample from that population and find out the average income of the sample elements. It is
this ample statistic from which we would base our opinion about the actual income of the
population. Thus, a parameter is unknown and it magnitude is inferred from sample data.

EXERCISES
1.1Estimating a Population Mean:
Assume we wish to estimate how much money was spent on buying Christmas gifts by
Kitwe residents. A random sample of 50 Kitwe residents is selected and it is found
that the average amount spent on buying Christmas presents was K3.5 million with a
standard deviation of K0.27 million.

We thus wish to estimate the average expenditure during the festive season by
Riverside/ Kitwe residents. Let us assume we desire to estimate the true average
expenditure at the 0.95 confidence level.

Solution:

x+z σ
√n
 3.5 + 1.96 0.27
√ 5 0

3.5 + 1.96 (0.04)

3.5 + 0.08

3.42 ≤ µ ≤ 3 . 5 8

We can thus be 95% confident that the true average expenditure is between 3.42 and
3.58

1.2Estimating a Population Proportion

Another variant of estimation might be a situation where we wish to estimate the
proportion of people who are interested in watching the Africa soccer competition.

ZNBC is interested in determining the proportion of viewers interested in watching a

television broadcast of the Africa competition. A random sample of 300 viewers is
taken and only 36 viewers appeared to show interest.
What is the true proportion of disinterested viewers at the 95% confidence interval?

Solution:
p + z pq
n

0.12 + 1.96 0.12 x 0.88

300

0.12 + 1.96 (0.02)

0.12 + 0.04
 8% ≤ p ≤ 16%

We can be 95% confident that the true proportion of disinterested viewers lies
between 8% and 16%

2 Analysis involving examination of a difference between the

magnitude of a parameter and a sample value.
In estimation, we used sample data to form an opinion about the magnitude of a
parameter. Recall that the parameter was unknown and the objective then was to
estimate this unknown quantity.

In some situations, the parameter may be known but we may wish to know, whether
due to some development, this parameter has changed or has remained the same. In
examining whether or not the magnitude of the parameter has changed, we start off
with the hypothesizes value of the magnitude of the parameter, and the analyze
sample data to see whether or not they support the hypothesis of there being a change.

2.1 Example 1: Test involving a population Mean

The current monthly allowance to student is based on an average monthly
expenditure of K175, 000 which was computed in 2005. the standard deviation was
then K36, 000.

To justify their request for an upward adjustment in their monthly allowance, students
have taken a sample of 100 students and have found that this sample of 100 students
spent an average of K184, 000.

Is there justification for an upward adjustment of the allowance? Use a significance

level of ά = 0.05
Solution

(a) Observe a difference-

Which can be said to have an implication on the decision whether or not to adjust
the allowance. In other words, have the expenditure demands actually risen from
the monthly of K175, 000 to K184, 000? Is this difference statistically
significance?
Old average expenditure: K175, 000
New (observed) expenditure: K184, 000

(b) Specify the Null and Alternative hypotheses-

The Null hypothesis is to assume that the average expenditure has remained the
same, that is, it is still at K175, 000 and that the difference we have observed is
not significant. The alternative hypothesis is to assume that there has been a
change in average expenditure, and the observed difference reflects a real
difference between the parameter (k175, 000) and the sample statistic (K184,000).

Null hypothesis: Ho: µ= 175, 000

Alternative hypothesis: Ha: µ≠ 175, 000

(c) Specify the significance level

ά = 0.05
if our null hypothesis is true, the sampling distribution of means from simple
random samples of 100 students will be normally distributed around a population
of K175, 000 with a standard deviation of 36, 000 S
√ 100 √

Our selected sample of mean x = K184, 000 is one of the values of means in this
sampling distribution, provided that the population parameter is K175, 000 that is:
 x±z σ
√n

Or 175, 000 ± 1.96 36, 000

√100

175, 000 ± 7, 056

167, 944 ≤ µ ≤ 182, 056

In terms of normal distribution of sample means, this is illustrated in figure 8.1.

The probability is 0.95 that our forthcoming sample mean will lie in the region
from K167, 944 to K182, 056 when the average is K175, 000. This is regarded as
the acceptance region. A value outside this region will imply that the difference
between K175, 000 and K184, 000 is real and significant and we would therefore,
on the balance of probability, reject the null hypothesis.

Figure 8.1: An Illustration of the distribution of sample means

167, 944 175, 000 182, 056

(d) Identify the test statistic and its sampling distribution.

In general, the test statistic will assume the following form:

Test statistic (Z) = sample statistic – hypothesized value of parameter

Standard error of the statistic
 =x-µ
σ

√ n

= 184, 000 – 175, 000

36, 000
100

= 2.5

(e) Compare the test statistic to the specified Z value

Given the significance level, ά = 0.05, for a two tailed test, the Z value is 1.96.
The computed Z value is 2.5

Figure 8.2: comparison of Z values in the case of a two-tailed test

-2.5 -1.96 1.96 2.5

If the computed Z value is greater than the specified Z value, reject the null
hypothesis otherwise accept null hypothesis. In this case, reject null hypothesis
and conclude that the average monthly expenditure is no longer K175, 000.
Example 2: An example of a one-tailed test
Assume that GRZ is willing to review student upkeep allowance but only if the
average monthly expenditure is indeed greater than K145, 000. A random sample
of 225 students shows a monthly expenditure of K147, 000, with a standard
deviation of K10, 000.

Should the allowance be revised on the basis of this information? Use ά= 0.05

Solution:

1. Observe a difference
GRZ criterion: K145, 000

Sample result: K147, 000

2. State the null and alternative hypothesis:

Null: Ho: µ ≤ K145, 000

Ha: µ > K145, 000

3. Specify the significance level

Let us assume ά = 0.05. This is one-tailed test and the specified Z value is
therefore 1.64 as shown in the figure below.

0.04500

0.05
0 Z = 1.64

4. Identify the test statistic and its sampling distribution

Z = sample statistic – hypothesized value of parameter
Standard error of statistic
 = 147, 000 – 145, 000
10, 000
223

= 2000
667

= 3.0

5. Compare the computed Z value to the specified value

1.64 3.0
Since computed Z value is greater than specified Z value, we reject null
hypothesis and conclude that there is justification for the review of the allowance.

OR:
If the monthly expenditure is K145, 000, we would expect the sample result to tie
with the interval shown in the diagram below:
µ±Z σ
√n

145, 000 + 1.64 10, 000

√ 225
 145, 000 + 1093

146, 093

145, 000 146, 093 147, 500

Since the observed value of K147, 500 is outside the acceptance region, we reject
the null hypothesis.

2.2 Example 2: Test involving a proportion

Sound investment are considering opening a retail outlet at CBU but will only go
ahead if at least 20% of the students express interest in music. A random sample
of 625 is taken and 140 students indicate interest in music.

Should sound investment go ahead? Use a significance level of 0.05.

Solution:
Step 1:
The observed difference which can have important implication for the investment
decision is whether the proportion of students who are interested in music is
significant to justify the opening of an outlet at CBU. The proportion of students
 interested in music is 22.4% compared to the criterion proportion of less than
20%.

Step 2:
Specification of the null and alternative hypothesis
Null hypothesis: the proportion is less than 20%, that is, p < 20%
Alternative hypothesis: the proportion is 20% or more, that is, p ≥ 20%

Step 3:
Specify the Z value at the given level of significance. Since this is one-tailed test, Z
value is 1.64

Step 4:
Identify the test statistic and its sampling distribution. This is a sampling distribution
of a proportion and a test statistic is given by:

Z = sample statistic – hypothesized value of a parameter

Standard error of the statistic

= 22.4 – 20
(20)(80)
625

= 1.5

Step 5:
Decision Rule: reject null hypothesis if the computed Z value is greater than specified
Z value, otherwise, accept null hypothesis since the computed Z value (1.5) is less
than the computed Z value (1.64). We fail to reject the null hypothesis. The observed
 proportion of students interested in music is not significant enough to justify opening
a retail outlet at CBU

2.3 Example of the test of the difference between two sample mean
A marketer is seeking to promote his product by exhibiting it at a commercial show.
Before a show, he takes a random sample of 32 and finds that the average amount
spent on his product was K50 million with a standard deviation of K2 million. After
the show, he takes a random sample of 36 and spends that the average amount spent
on his product is K54 million with a standard deviation of K3 million.

Was there an increase in the amount spent on his product? Use 0.01 significant level.

Solution:
1. The observed difference between the two patterns of expenditure is as follows:

Before the show after the show

Sample size (n) 32 36

Average expenditure K50m K54m
Standard deviation K2 m K3 m

The question is whether the K54 million is significant enough to indicate a real
difference from the K50 million.

2. Specify the null and alternative hypotheses

Ho: µ1 = µ2 or µ1 - µ2 = 0
Ha: µ1 ≠ µ2 or µ1 - µ2 ≠ 0

3. Specify the Z value at the 0.01 significant level.

This is Z = 2.58
 4. Specify the test statistic and its sampling distribution
Z = (x1 – x2) – (µ1 - µ2)
σ21 σ22
n1 n2

= (50 -54) – (0)

22 32
32 36

= 6.5

5. Decision
Since the computed Z value (6.5) is greater than the specified Z value (2.58), we
reject the null hypotheses. The expenditure before and after the show are not the
same.

Another example of a test of a difference between two sample means.

A company is promoting a new stock feed. A flock of 100 birds is split into two groups of
50 each. One group designated as a experimental group, is subjected to the new stock
feed, while the other designated as a control group, is fed on the old stock feed. At the
end of the experiment, the following information is obtained.

Experimental group Control group

Average weight 15.2kg Average weight 14.9kg
Standard deviation 2.1kg standard deviation 2.0kg
Sample n = 50 Sample n = 50
Is the new stock feed effective?
1. Observed difference
Difference in average weight: that is, 15.2 kg versus 14.9 kg

2. Specify null and alternative hypothesis

Ho: M1 = M2 or M1 – M2 = 0
Ha: M1 ≠ M2 or M1 – M2 ≠ 0
3. Specify the level of sampling error (or level of significance)
ά= 0.05 ; z = 1.96

4. identify the test statistic and its sampling distribution

Test statistic = the difference between two populations (difference between two
sample means

Z = sample statistic – hypothesized value of a parameter

Standard error of the statistics

= 0.73

5. compare the critical value z = 1.96 to the computed value z = 0.73

-1.96 0.73 1.96

Conclusion
We fail to reject the null hypothesis and conclude that there is no real difference between
the two populations. That is, the new stock feed is not effective.

2.4 Testing the difference between two sample proportions

Assume we are interested in determining whether the proportion who shop at Xmas
is the same or different between the rural and urban region. Assume that in the
 random sample of 225 rural residents, 54 of them were found to have shopped at
Xmas.

Use a significance level of 0.05

Region Sample size number of shoppers at Xmas

Rural 225 54
Urban 175 52

Solution:
1. The observed proportions are:
24% for the rural region and 30% for the urban region. The question is whether
the two proportions are significantly different from each other.

2. Specify the null and alternative hypothesis

Ho: p1 = p2 or p1 – p2 = 0
Ha: p1 ≠ p2 or p1 – p2 ≠ 0

3. The specified z value at the 0.05 level of significance is Z = 1.96

4. Specify the test statistic and its sampling distribution.

This is a sampling distribution between two sample means, and its test statistic is
given by:

Z = (p1 – p2) – (P1 – P2)

p1q1 p2q2
n1 n2

= (30 – 24) – 0
(30)(70) (24)(76)
175 225

= 1.3
5. Decision:
Since the computed Z value (1.3) is less than the specified Z value (1.96), we fail
to reject the null hypothesis. The true proportions are the same for the urban and
rural sectors.

THE CHI-SQUARE ANALYSIS

 It is used to examine differences between two sets of data, one an observed
frequency and the other, called an expected frequency. Conditions for use of
the chi-square are:
 There must be two sets of data
 The sample size for the two sets of data must be the same
 Each cell in the data must contain at least five counts.

 The tabled value of X2 is based on:

 The specified significance level of ά
 The number of degrees of freedom- which is one less than the number of
categories, v = k-1

Example 1: Goodness of fit

A researcher wishes to determine whether or not the three brands of mealie-meal are
equally preferred. A random sample of 150 individuals is asked to indicate which brand
of mealie-meal they prefer. They indicate as follows:
Brand No. preferring brand
Brand A 35
Brand B 55
Brand C 60
150

ά = 0.05
 1. observe difference in frequency distribution between observed and expected
pattern
Observed Expected
Brand A 35 50
Brand B 55 50
Brand C 60 50

2. specify the null and alternative hypotheses

Ho: All brands are equally preferred
Ha: The three brands are not equally preferred.

3. specify the level of significance and the tabled X2:

ά = 0.05
Degree of freedom, (k-1): (3-1) = 2
X2 = 5.99

4. compute the test statistic, X2

X2 = (observed – expected)2
Expected

Brand Observed Expected (O – E )2

E
A 35 50 4.5
B 55 50 0.5
C 60 50 2.0
X2 = 7.0

5. Decision: Reject Ho if computed X2 > tabled X2 since computed X2 (7) is greater

than (5.99), we reject null hypothesis and conclude that the three brands are not
equally preferred.
Example 2: Test of independence
A company wishes to determine whether or not females are different from males in the
pattern of consumption of their product. A sample of 1000 females and 1000 males is
taken and shows the following consumption pattern:
Degree of use Female Male
Heavy 220 250
Moderate 270 300
Light 230 200
Non users 280 250
1000 1000

ά = 0.05

1. observe difference in pattern

2. specify the null and alternative hypotheses:

Ho: there is no difference between M and F
Ha: there is difference between M and F

3. specify significance level and X2

ά = 0.05
X2 = d.f = (k-1) = (4.1) = 3
X2 = 7.81

4. compute X2
Observed Expected
(Female) (Male) (O – E)2
E
Heavy 220 250 3.6
Moderate 270 300 3.0
Light 230 200 4.5
Non users 280 250 3.6
X2 = 14.7
 5. Decision: since computed X2 > tabled X2, we reject null hypothesis, and conclude
that there is a difference in consumption patterns between males and females.

Example 3: Test of independence

Are sex and preference for soft drink independent?
Preference
Sex Coke Fanta Sprite TOTAL
Male 106 93 215 414
Female 6 39 73 118
112 132 288 532

1. observe difference

Expected Frequencies
Preference
Sex Coke Fanta Sprite TOTAL
Male 106 93 215 414
Female 6 39 73 118
112 132 288 532

2. specify the null and alternative hypotheses

Ho: Sex and preference are independent
Ha: Sex and preference are not independent

3. significance level and table X2

ά = 0.05
d.f= (c-1) (r-1) = (3-1) (2-1) = 2
X2 = 5.99
 4. compute the test statistic, X2
Observed Expected (O – E)2
E
106 87 4.1
93 103 1.0
215 224 0.4
6 25 14.4
39 29 3.4
73 64 1.3
X2 = 24.6

5. Decision: Reject null hypothesis and conclude that sex and preference are not
independent.

Example:
A manufacturer‟s past record shows that 42% of housewives on the Copperbelt use his
product. A survey of 100 housewives in each of the five Copperbelt towns shows that the
numbers using his product are 50, 41, 35, 47 and 39.

Observed (n = 100) Expected (O – E)2

E
50 42 1.5
41 42 0.02
35 42 1.2
47 42 0.6
39 42 0.2
X2 = 3.52

ά = 0.01
d. f = (5 -1) = 4
X2 = 13.28

Decision: we fail to reject null hypothesis and conclude that the product is consumed by
42%- there is no regional variations.
MEASUREMENT OF ATTITUDE
1. Concept of attitude
An attitude can be defined as a predisposition to respond to something. It has the
following components:

a A cognitive component: refers to the person‟s beliefs, knowledge or

information about an object.

b An affective component: refers to a persons feeling of like or dislike

concerning an object.

c A Behavioural component: refers to action tendencies, or predispositions

towards an object

2. The importance of attitude in marketing

Attitude is important in marketing because it influences on purchase behaviour.
This is perhaps best illustrated by what happened to the pork industry in the
U.S.A in the 1980s at that time, pork producers noticed a decline in the per capita
sales of pork. The reason adduced then for the decline was that Americans had
become more concerned about health and diet. They were interested in the
youthful, healthy and leaner appearance. In contrast, pork was associated with fat,
and led to negative attitude to it among consumers.

The Pork producers Council felt this attitude was unjustified as pork had some
desirable characters. They commissioned a study to uncover beliefs and attitudes
about pork. Among other things, the study revealed that:

 White meat was preferred to red meat.

 Only 9% of respondents thought of pork as white meat
 Females with college or technical education (career women) were most
negative towards red meat.
  The preference for white meat was primarily because it was perceived to
be leaner.
 Test was very important when selecting a cut of meat
Based on these things, the pork producer council developed and tested an
advertising campaign to improve customer‟s attitudes towards pork by:
(i) Repositioning pork as a white meat
(ii) Changing beliefs about its key attributes

They mounted an advertising campaign in which pork was emphasized as white

meat, and highlighted the following attributes of pork:
 That it is tasty
 That it is appropriate for light meals, and
 That it is easy and quick to prepare

Following this campaign, the sales of pork picked up, ostensibly because of
change of attitude towards pork accessioned by the pork!

3. The concept of measurement

Measurement of attitude refers to the assignment of numbers to reflect an attribute
towards an object. In this regard, the following are some of the meanings that can be
attached to a number:
 To identify an object (nominal scaling), for instance
o A personal identification number
o A car registration number

 To obtain the relative standing of an object to an attribute (ordinal scaling). That

is, a number is used to indicate “more or less” of an attribute e.g.
o Year of study: 1st, 2nd, 3rd, 4th.
o Ranking of preference
1 2 3
  To indicate or compare differences in attribute of an object (interval scaling) e.g.
A B C D E
20 40 60 80 100
The difference between C and E, for instance, is twice the difference between A and
B.
 To indicate absolute differences in attributes of an object (ratio scaling). In this
case, there is usually a zero starting point e.g. sales

4. Techniques for measuring attitudes

the Thurstone- equal-appearing interval-

The essential features of this sale are as follows:
 A large number of statements is generated reflecting different aspect of attitude of
the object in question. An example of such statements might be
 These statements can be described as favourable, unfavourable or neutral.
 Each statement is assigned a number, on scale ranging from 1 to 11, of how it s
independently perceived as unfavourable, neutral or favourable.
 If the respondent is in agreement with an unfavourable statement, he/she is said to
have a negative attitude towards Shoprite; if a respondent agrees with a statement
that carries a number that corresponds with a favourable aspect, he/she is deemed
to have a favourable attitude, and so on.

Procedures for constructing the instrument/questionnaire

1. generate a large number of statement reflecting positive, negative, and neutral
attributes about an object, e.g.
Shoprite attracts both shoppers and non-shoppers

Shoprite constantly changes its prices

Shoprite offers a good service

2. A group of judges are then asked to rate each statement on the scale ranging from
1 to 11, which reflect degree of favourableness.

Unfavourable Neutral Favourable

1 2 3 4 5 6 7 8 9 10 11

Judges should not reflect their opinion, but rather whether the statement in
itself is Unfavourable, neutral or favourable.

3. The researcher then determines or computes the average rating (scale value) for
each statement. The procedure for determining the scale value for a statement is:

Statement: Shoprite offers a good service

Unfavourable neutral favourable
1 2 3 4 5 6 7 8 9 10 11
No. of judges 0 0 0 0 0 0 3 2 4 1 0
Proportion of judges 0 0 0 0 0 0 0.3 0.2 0.4 0.1 0
Cumulative proportion0 0 0 0 0 0 0.3 0.5 0.9 1.0 1.0

The scale value is given by the median or 50th centile

V50 = l + C – ∑pb i, where

pw

V50 = the value of the 50th centile

l = the value of the lower boundary of the interval in which the centile fall

C = centile in question
 ∑p = the um of the proportion below the interval in which the centile falls

pw = the proportion within the interval in which the centile falls

i = the width of the interval

V50 = 7.5 + 0.5 – 0.3 1

0.2

= 8.5

4. Having determined the scale value of each statement- which imply that a
statement can have a value in the unfavourable, neutral or favourable range, a
respondent is asked to indicate which statement they agree with.
5. The attitude score is taken to be the average score of the scale value of the
statements agreed with.

Statement Scale value Attitude score

1 7.5 - 7.5 + 6.0 = 6.75
2

Weaknesses of the Thurstone Technique:

(a) It is possible for people showing agreement with widely different scored
statements to end up with the same arithmetical mean score- hence assessed to be
the same.

A: 4, 6, 8 6
B: 6 6
 For instance, when A agrees with the statements whose scale value are 4,6 and 8;
representing a neutral disposition of 6, and B agrees with only one statement
whose scale value is 6; both A and B are said to have the same attitude.

(b) It does not measure intensity of feeling; it merely categorizes into type of attitude.
Judges are not generally willing to spend time going through the many statements.

(c) The rating by judges could be influenced by their own attitude; that is, another
group of judges could score differently

(d) It is somewhat laborious

Use of Interquartile Range in Refining Instrument

 On the practical scale, a researcher is unlikely to use all the statements he
generated and for which judges have indicated scale values.
 It is pragmatic for the researcher to select statements which best reflect the
attitude score.
 Suppose we wish to select one statement for slot 8 from among three statements
whose scale values have been computed as indicated below.

Statement Scale value

A 8.2
B 8.3
C 8.2
 Pick statement with the least dispersion because it indicate greater consensus
among the judges. Dispersion is measured by the interquartile range whose
formula is given by
Q = V75 – V25, where
Q = Interquartile Range
V75 = 75th centile
V25 = 25th centile
 V75 = 8.5 + 0.75 – 0.5 1
0.4
= 9.1

V25 = 6.5 + 0.25 –0 1

0.3
= 7.3

V75 – V25 = 9.1 – 7.3

= 1.8
If this statement had to compete for a slot, then the other statements would have to
have an interquartile range value, Q, for less than 1.8 to be selected.

LIKERT SUMMATED SCALE

1) The thrust of this scale is to measure intensity;
2) Select a series of responses which represent various degrees of agreement or
disagreement
Strongly Neither Strongly
Agree Agree A nor D Disagree Disagree
5 4 3 2 1
Shoprite is expensive

Shoprite is far

Shoprite staffs are polite

Shoprite offers variety

Shoprite is congested

Shoprite is poorly ventilated

 3) Administer instrument to respondents.
4) Attitude score is the sum total of scores on each statement. This must be
compared to sum standard, e.g. the maximum possible score or in comparison
to other group(s).

THE SEMANTIC DIFFERENTIAL SCALE

This scale is used to show the profile of attitudes. Thus, given an object, bipolar
objectives or words are generated reflecting the attitudes towards an object in
question. Let us assume we wish to profile attitude toward apples of a particular
brand. A semantic differential scale ccan be used:
 To show the attitude of all respondents towards this particular brand of
apples
 To measure and compare attitudes of various groups towards this
particular brand
 To compare the attitude toward this brand with the attitude of another
brand

APPLES
Fresh ……………………… Stale
Affordable ………………... Not affordable
Accessible ………………... Inaccessible

RATING SCALES
Rating scales require a respondent to tick at an appropriate place on the scale to
indicate his or her attitude towards an object. The scale is often in the form of a
continuum on particular attributes of an object. The following are some of the
ways rating scales are applied:
 To indicate overall attitude towards objects.
 Tinned beef
Like Do not like
Price
Quality

 To indicate degree to which an object contains a particular attribute

Very sweet Bitter

 To indicate importance attached to an attribute

Very important Not important

Courtesy

Cleanliness

Convenience

Music

4.4.1. Type of rating scale

Comparative Vs Non comparative
Comparative rating scale provides an explicit standard against which the object of
interest is to be compared.
For example,
How do you like Fanta compared to Coca-cola?
Like it very Like it more The same Like it less Like it much
much less
- - - - -
How do you like the taste of Fanta?
Like it very Like it more Does not Dislike it Dislike it
much matter very much
- - - - -

Rank order scale

This requires the respondent to rank a set of objects to specified criteria. Rank
Fanta, Coca-cola, in terms of preference from 1 to 3, with 1 being the most
preferred and three being the least preferred.

Fanta 3

Sprite 2

Coca-cola 1

The Constant sum behaviour

This requires the respondent to divide a constant sum (say 10 to 100) among two
or more object in a way which reflects the relative standing of the object on the
attribute.
For example, in terms of sweetness, divide 10 points between Fanta and Coca-
cola.

Fanta 7

Coca-cola 3

Observation of behaviour
Under this technique, it is presumed that observed behaviour is indicative of a
particular attitude. For instance, there are certain type of behaviour which are
indicative of interest in a phenomenon: window shopping signals interest in goods
on display; attendance of commercial shows is indicative of interest; questions
asked about a product may be a prelude to actual buying; and time spent a product
or checking a price tag is a step away from a decision whether or not to purchase
an item.

Performance of a task
How well or badly a task is performed reflects one‟s attitude towards the
particular task. The saying „you have a bad attitude toward work‟ suggests that
one‟s performance is unsatisfactory.

Indirect technique or reaction to stimuli

How we react to stimuli may also indicate our predisposition towards an object.
The words in reaction to an object all too often mirror our attitude toward the
object. We for instance, tend to speak favourably about people, events or products
we like.

Physiological reaction
How our bodies involuntarily and physiologically react to object may also
indicate our attitude towards an object. Thus, whether we have tears of joy or
grief entirely depends on whether or not we like what has prompted us to shed
tears. We salivate over food we like!

RESEARCH REPORT
The results of a research report may be reported in written or oral format, or both.
While many readers of a report will not be able to ascertain the quality of a
research design, questionnaires or intricacies of the experiment or field work, they
will, however, easily recognize the quality level of a report. The quality of the
report is all too often used as a major indicator of the research itself.

The guidelines are intended to assist you in writing a good research report. It is
important to note here that “good” is used from the perspective of the person who
is going to read the report- whether it is a decision maker, a scholar or any other
interested party in the issue that was studied. Thus, it should be obvious in writing
a research report that a researcher should keep in mind the audience.

FOCUS ON THE OBJECTIVE OF THE STUDY

As observed at the onset of the course, research is initiated and undertaken to help
make a decision. The report should, therefore, be built around the decision and
know the information generated by the researcher is relevant to the decision.

USE TERMINOLOGY THAT MATCHES THE

VOCABULARLY OF THE READERS
Minimize the use of technical terms. As technical terms are considered to be of
use to some readers, they should be placed in the appendix.

AVOID ERRORS IN CONSTRUCTION, GRAMMAR AND

SPELLING
The seriousness and credibility of the researcher‟s work can be enhanced by the
effort a researcher makes in avoiding incorrect sentences or misspellings. Word
processors with spelling and grammar check can be used to minimize such errors
but, most important, the researcher should take time to edit the report.

LOGICAL FLOW
A good research report must reflect a clear and organized thinking. A person
reading a report must see how any section of the report relate to the whole. This
requires that sections of the report logically follow each other.
 FORMAT OF THE REPORT
There is no best format for all accessions. Considerable variations in the
organization of the report should be dictated by the nature of the audience and the
topic of the report.
The following guidelines are never the less generally applied

(a) Title page

The title page should identify the topic, which captures the essence of the
research study. It should also bear the researcher‟s name and for whom the
report is prepared.

(b) Declaration
A declaration is a statement to the effect that the work was completed by the
researcher and the works of others have been duly acknowledged

(c) Table of contents

This is a list of chapters, sections, subsections and parts of the subsections
arranged in the sequence in which they appear, including page numbers where
each one starts. Appendices, tables or figures may be listed unless they are
numerous, in which case it is advisable to list them on a separate page(s)
immediately following the table of contents.

(d) Abstract
This is one-page summary of the study. It should include a brief statement of
the problem, the research objectives, how and where the research was
conducted, sample description, and a brief description of results and
recommendations.
(e) Introduction
This is the first chapter or section of the report. It should give a background to
the study and may include such questions as why you think the issue you are
investigating is important, problematic, controversial or necessary. The
background is then followed with the precise statement of the problem. It is
necessary to formulate a working statement of the problem e.g.
 Why X and not Y
 Why then and not now
 Why this and not that

It is important that the researcher documents the existence of the problem. The
statement of the problem then follows in which the researcher will specify
what it is he/she wants to accomplish. These objectives must be specific,
workable and measurable. Conclude the introduction by pointing out the
potential significant of the study. For example, will it add to existing theory or
will it test a theory? Will it correct previous research, resolve inconsistencies
in the existing literature, or solve a practical problem.

(f) Literature review and Theoretical Framework

Review what others have done and discuss how this fits into what you are
doing. Use an organization scheme to guide you in your review. For instance,
in attempting to arrive at a workable definition of a concept you are dealing
with, you may wish to point out how various scholars have defined the
concept, the various strength and weaknesses of each definition and why you
have defined the concept the way you have. One scheme includes historical
analysis, problem-cause solution relationship, or the known to the unknown.
A theoretical framework relates to the relevant theory on the subject. Discuss
and develop statements that have an analytical framework for explaining or
describing a phenomenon. Spell and define the concepts basic to the theory
and assess the assumptions being made, if at all, as to whether they are
reasonable or defensible. Do not forget to reference your work!!!
(g) Conceptual Framework
Use the literature review and theoretical framework to develop your argument
or thesis. What does previous research reveal about the different aspects of the
problem and what new dimension are you preparing to add? What research
questions and hypothesis have emerged? A conceptual framework, then, is the
researcher‟s own perspective of how he/she intends to deal with the research
problem. This is the “essence” of the research because it is at this stage that
scholarship will manifest itself.

(h) Research Methodology

This is a summary of the methodology used to meet the objectives of the
research project. Specifically, it addresses the following questions:
 What data are needed?
 What are the sources of the data?
 How are the data to be collected?
 How are the variables to be operationalised?

(i) Research Finding and analysis

This is where detailed finding are presented and analyzed. The finding should
be organized around the objectives of the study. They must consist of a
description of the results and show in summary tables, chart, graphs and the
like. Detailed finding, especially those involving computations, should be
presented and analyzed. The finding should be organized around the
objectives of the study. They must consist of the description of the results and
show how the hypotheses were tested. Findings can be shown in summary
tables, charts, graphs, and the like. Detailed findings, especially those
involving computation, should be presented in appendices.
(j) Conclusion and Recommendations
In the conclusions, summarizes the findings but more importantly refers to
hypotheses, objectives and questions. Have these been addressed and to what
extent? What do the results mean and what are the implications of the study?
What were the limitations of the study?

Recommendations are suggestions of how the problems can be resolved or

improved. Also, indicate marketing implication, such as how the findings
might relate to the practice of marketing. Finally, suggests directions for
future research.

(k) References
On the separate page under the heading references, list the sources refered to
and cited in the report. The sources should be listed alphabetically by author‟s
last name.

(l) Appendices
Place in the appendix items that will appeal to only a few readers or that may
be needed only for occasional reference. Such items include a copy of the
questionnaire used, detailed statistical tables or special reports.

-The end-