Unit 21

Survey Design
Contents
21.1 lntroduction
2 1 .2 Preliminary Considerations
21.3 Stages IPhases in Survey Research
21.4 Formulation of Research Question
2 1 .5 Survey Research Designs
21 -6 Sampling Design
2 1.7 Conclusion

Learning Objectives
It i s expected that after reading Unit 12 you would be able to
Q Work out (a) the purpose of enquiry, (b) the unit on which the
survey is to focus and (c) the availability of resources
Q Divide the survey research into stages or phases
O Formulate the kinds of questions to be asked during the survey
*:* Decide the type of research design to employ
O Choose a sampling plan for survey.

21. I lntroduction
Design elements are crucial in understanding the complexity of undertaking
a survey. The survey method involves several aspects, which are crucial
for its execution in a systematic and scientific manner. Unit 21 provides
a description of the basic dimensions of survey in social research. The
topics discussed in this unit are: preliminary considerations for undertaking
survey research, different stageslphases i n conducting a survey,
formulation of research questions, types of research design, and details
of sampling design.

21.2 Preliminary Considerations

There are three preliminary considerations before any researcher decides to
undertake survey as a method of social investigation. The considerations are
i ) Purpose of enquiry,
ii) Population on which survey is to focus, and
iii) Resource availability.
Each of these considerations is essential for survey research.
The first prerequisite for undertaking survey research i s clear and well-
defined objectives, purposes or research questions. The researcher must
be clear that the survey method is the best method to study those
objectives (purposes or research questions). If there is a clear idea
about the objectives of a research study, it leads to the choice of the Survey Design

appropriate survey as a method of investigation.

The second prerequisite is specifying the population of study. This
influences the decisions about sampling and resources. The accessibility
of the popitlation i s an important element in developing survey design. It
judges the duration of study, the different layers of people to be included
and the size of the research staff as well as the sample.
The third important prerequisite in designing a survey i s to take stock of
the resources availability for study. Resources may be in the form of
financial costs, manpower requirement and the total time within which
the survey has to be completed. Surveys are usually labour intensive, the
single largest expenditure being the fieldwork costs which include the
interviewing time, travel time, and transport costs, etc. There are
additional costs involved i n survey research, namely, training and
supervision of interviewers, questionnaire construction, pilot testing the
questionnaire, printing, posting, coding, computer programming, etc.
Thus, the researcher must estimate the availability of resources before
starting the survey research.

21.3 Stages 1 Phases in Survey Research

Survey research calls for a systematic and comprehensive collection of
information about the attitudes, beliefs and behaviour of people. In a
survey, the researcher follows a deductive approach. He begins with a
theoretical or applied research problem and ends with empirical
measurement and data analysis. Once a researcher decides that survey
i s the best and appropriate method t o undertake his research, then
elaborate steps are taken in order t o make the application of the method
more scientific and systematic. Further, there i s no unanimity among
the researchers on the actual number of steps or their ordering. However,
here we deal with a few major steps in three phaseslstages that cover the
preparation, planning and execution of survey research (See Figure 21 .I).

1. Research Question

3. Tabulation and
Respondents and
3. Draw the Sample
4. Data Gathering 3. Supervision and

7. Finalising the Survey

Figure 21 .IMajor Steps in Three Stages of Survey Research + 7*I

0: .0
:
QuarrlitaLivs and The first phase covers design and planning for a survey. This is a very
Survey Met'itods
elaborate and Long phase in which the details of conducting a survey are
worked out. It is crucial as it sets the ground for the actual survey.
The important steps i n this phase are formulating a research question,
deciding the survey research design, working out the sampling procedure,
deciding the data gathering techniques, construction of a questionnaire,
conducting a pilot survey, modifying the survey instrument, training
field investigators/interviewers.
The second phase involves the execution of the main survey for data
collection. It includes locating and accessing the respondents and field
settings, conducting the survey, monitoring and supervision of the field
personnel.
In the third and final phase,-the researcher edits the data collected,
prepares codebook, and decides the tabulation for data analysis. Finally,
the researcher prepares the Write-up/Report of the study.
The listing of steps in a sequential form may give the impression that
once a step is covered there is no turning back. However, to the contrary,
it is often necessary to alter decisions made at an earlier point in the
light of developments during the later stages of research. For instance,
the problems encountered during the execution stage may call for
modification either of the sampling design or of the type of instrument
for collection or of the questionnaire. Further, it is also a mistake t o
assume that these steps can be followed one at a time without any
regard for subsequent steps. In reality, it is often important to keep in
mind the other steps while working out the details of one step. For
instance, it i s necessary to keep i n mind the intended data analysis
procedures while constructing a questionnaire.

21.4 Formulation of the Research Question

The first step in a survey design is the formulation of a research question.
Research question here refers to the major goal problem or objective of
the study. In fact, the formulation of a question should precede the selection
of survey method for data collection. Though survey has a wide range of
applications in the social sciences, it may at times turn out that the
other methods of data collection (such as in-depth interviewing or
participant observation, etc.) may be more appropriate or suitable to
the problem chosen. Thus, it i s necessary to think of the appropriateness
of the survey method for investigating a particular research problem.
Further, it may be kept in mind that there i s a need to formulate a
research question at the beginning of research. But, this emphasis does
not mean that the initial research question is the final one (see Box
21.1). Often research questions arise unexpectedly in the middle of a
conversation with informants. These need not be ignored. Their addition
or modification of questions decided earlier is always a useful excercise.
 Survey Design
Box 21. Refine Your Research Questions
According to David de Vaus (2002), it is important to know what we are Looking
for, it is a mistake to let the initial focus blind us to other unanticipated questions
which may perhaps be more interesting, important or manageable than the initial
one. Research questions can be refined and new issues emerge in the course of
survey research.

How does one arrive at a research question? Often, the researcher's

curiosity about a particular social phenomenon guides him to choose a
particular research question. However, the researcher can formulate his
research question in more than one way. First, the starting point could
be a theory or a model. The choice of theoretical framework or model
will have implications for the kind of questions that are to be formulated
and the kind of data that are to be collected. Second, significant social
policy implications may guide researchers in formulating their research
problem. In other words, the problem chosen should be relevant to the
issues of social improvement, social change, or social action. 'Third, the
goal of producing social criticism may lead some researchers to formulate
research questions suitable for that objective. Some research questions
may meet several or some of these criteria, but it is rare that a research
question will meet all of them.
Further, it is important that the survey's general purpose be translated
into a specific central problem/aim/objective. It is not enough t o say,
for example, 'I am interested in getting some answers about inequality'.
One should be clear about the kinds of questions one wants to answer.
What answers to what questions? Does he want to know the extent of
inequality, i t s distribution, i t s causes, its effects, or what? What sort of
inequality one is interested in - social, political, economic, etc? Over
what period? Where?
The process is of narrowing down from a general area of interest to a
level that is specific enough for concrete empirical research to be carried
out. After specifying the primary objective of the survey, the next step
involves identification and itemising of the subsidiary topics that relate
to the central purpose. For example, in a study of student attitudes to
the quality of undergraduate education in a college, the subsidiary issues
may include the quality of teachers, the infrastructural facilities, the
admission procedures, the type of courses offered, the content of the
courses, the interaction between teachers and students, the facilities
for remedial study, the facilities and scope for extra-curricular activities,
etc. Once the subsidiary items have been identified, the researcher
looks for the specific information required relating t o each of those
items. For example, with respect to the type of courses required, detailed
information i s needed on the duration of courses, the status of the
courses (credited or non-credited), orientation of the courses (theoretical
or practical), etc.
A thorough review of the existing literature (books and journals) on the
Quantitative and topic is helpful while narrowing the specific topic, identifying the subsidiary
Survey Methods
topics, itemising the subsidiary topics and spelling out the kind of
information required on these topics for data collection. A review of
literature is helpful because it reveals to us the scope and coverage of
research that has already been conducted and helps us in identifying
important gaps so that an attempt could be made to f i l l those gaps
through survey research.
There are other ways too that can help a researcher in formulating his
research question. One i s the exploratory study, which is a tentative and
relatively unstructured investigation of a few people who are similar in
many respects t o those we intend t o study as part of the main
investigation. For instance, i f we intend t o work on the quality of
undergraduate education i n a college, one may have t o visit an
undergraduate college, interview a few students, and identify the specific
research questions for the study. The most important goal of such an
exploratory study is that it stimulates the thinking of the researcher and
helps him identify specific issues that may be worthy of inclusion in the
study. Sometimes, it may help us in understanding the way the concepts
are understood differently by different people. For instance, in our earlier
example, different students may view differently the notion of 'quality'
in undergraduate education. For some, it may mean good teachers1
teaching. For others, it may be good infrastructure or a range of relevant
courses of study, etc.
Another aspect of the research question formulation is that of consultation
with those who are knowledgeable about the topic and who have done
related research. An encounter with these experts can help researchers
in anticipating the pitfalls of the proposed research study.

..........................
Reflection and Action 21.1
Suppose you have to carry out a survey of waste disposal methods, and you need
to formulate research questions. Work out the major goal problem or objective
of your study. Write down answers t o the following questions on a sheet of
paper.
Questions
9 In the light of the objective of your study, do you find the survey method a
useful and appropriate way of data collection? Provide specific reasons of
accepting or rejecting survey method as a method of data collection.
O How have you arrived at each of the research questions? State clearly the
choice of your theoretical model.
+:* Have you looked at the social policy with regard to waste disposal? Elaborate
the implications of the policy in terms of it guiding your research problem.
9 Is your study going to generate social criticism of an issue of importance for
the general public? If yes, you need to work out how your study is going to
lead other researchers t o formulate further research in the same area.
O Have you transformed the general objective of your study into a specific
central problem? It is important that you do this exercise at this stage and
state the narrowing down of the problem from general to specific level for
concrete research to be carried out.
 9 Have you consulted the experts in the area of your research to find out ' Survey Design
I what to expect in your field of inquiry and what are the likely pitfalls to I
I anticipate and t o be careful about? I
I Write your answers to above questions with the help of the relevant text in I
IL --,--.-------------------
Section 21.4.
--A
I

Types of research questions

There is no simple way to define research questions. However, according
to David de Vaus, we can delineate different kinds of questions and
provide guidelines to help focus research -- descriptive research questions
or explanatory research questions or both.
a) Descriptive research questions
It is difficult to focus a descriptive question but five questions can help.
*: What i s the time-frame of our interest?
9 What i s the geographic Location of our interest?
9 patterns for sub-groups?
What aspect of the topic are we interested in?
How abstract is our interest?
b) Explanatory research question
The first step in formulating explanatory research question i s to decide
whether we are looking for causes or consequences. We must list possible
causes and consequences and then collect relevant data. There are a
number of ways of coming up with such a List, namely, previous research,
the facts, our own hunches, an interaction with the informants.
Four questions might help formulating explanatory research question.
What am I trying to explain?
*: What are the possible causes?
*: Which causes will I explore?
*: What are the possible mechanisms?
Thus, there are varieties of issues that need to be kept in mind while
formulating the research question that proposes to undertake survey as
a method of data collection. It actually triggers the mind of the researcher
as how to opcrationalise the research,

2 1.5 Survey Research Designs

After the formulation of the research question, we must decide the type
of design to employ. Research designs for surveys are mainly of two types,
namely, experimental and descriptive. Experilmental designs use
environmental arrangements and rely on two clr more groups of a
participants or observations. When a randomly ccmstltuted group of a
hundred children i s compared three times, the silrvey design i s called
experimental. 'The descriptive design produces information on groups and
phenomena that already exist. No new groups are created. Descriptional
designs are also called 'observational' or 'explanatory' designs. +1110
Quantitative and Experimental designs
Survey Methods
According to Arlene Fink (1995), experimental designs are characterised
by comparing two or more groups, at least one of which i s experimental.
The other i s a control (or comparison) group. An experimental groups is
given a new or untested, innovative programme, intervention, or
treatment. The control is given an alternative (e.g. the traditional
programme or no programme at all). A group is any collective unit.
Sometimes, the unit is made up of individuals with a common experience,
such as men who have participated in a war or those who have had
undergone treatment at a hospital or those who have attended a particular
school. At other times, the unit i s naturally occurring: a classroom,
business, hospital, or a prison.
There are different types of controls. Firstly, there are concurrent controls
in which participants are not randomly assigned to groups. 'Concurrent'
means that each group i s assembled at the same time. For example,
when 10 of 20 schools are randomly assigned to an experimental group
while, at the same time, 10 are assigned to a control group, you have a
randomised trial or true experiment. Secondly, there are concurrent
controls in which participants are not randomly assigned to groups. These
are called non-randomised controlled trials, quasi-experiments, or non-
equivalent controls. Thirdly, in 'self-controls', a group i s surveyed at two
different times. These require pre-measure, post-measures, and are
called 'longitudinal' or 'before-after' designs. Fourthly, 'Historical Controls'
which make use of data coklected for participants in other surveys.
Finally, 'combinations' of all consist of concurrent controls with or without
pre or post-measures. Of all the experimental designs, longitudinal designs
are used quite often in social research.
Longitudinal designs
In a longitudinalB design, data are systematically collected over a period
of months or years in such a way that it is possible to observe trends in
attitudes or behaviours over the specified period. Consider the following
example. If a researcher is interested in studying the changes in the
spending behaviour of the white-collar employees, she or he will have to
conduct the study at different points of time in a year or periodically. A
study of this kind of a problem involves longitudinal (or time-series)
designs.
Trend studies and panel studies
There are two types sf longitudinal designs: Panel Study and Trend Study.
In a panel study, the same respondents (people) are interviewed two or
more times. In a trend study, two or more different samples of people
(respondents) are drawn at different times from the same population
(for- examples see Box 21.2). You may have noticed that the television
and newspaper media draw samples from time to time to study trends.
 Survey Design
11 Box 21.2 Examples of Panel Study and Trend Study II
Suppose we wish t o understand the changing support for the leading political
parties in the general elections to Parliament, two months before the election.
For this study, we draw a sample of adults in the voting age group and 60 percent
support, let us say, Political Party I. One month before the election, we re-
interview the people who were a part of our sample. Perhaps, 55 percent of
them now support Political Party I. This would be a panel study. On the other
hand, i f we draw a new sample of voters, one month before, and find that 55
n t of the new voters preferred Political Party I, this would be a trend

Trend studies provide information on net changes and panel studies provide
information on both net and gross changes. Suppose at Time 1, sixty out
of every 100 people say they would vote for Political Party A . Between
Time 1 and Time 2, 10 of the Political Party A voters might change their
opinion and vote for Political Party B and five of the Political Party B
might vote for Political Party A. Thus, between Time 1 and Time 2,
many people have changed their vote (opinion) and this is a gross change.
The net result of these individual changes i s that at Time 2, 55 of every
100 people would vote for Political Party A. Thus, i f we use a panel
design we can count the number of people in our sample who change
their opinion between Time 1 and Time 2. If we use a trend design, we
can describe the net change in the population, but we cannot count the
number of individual change. If the net change i s 5 percent (from 60
percent to 55 percent supporting Political Party A), the gross change
might be anywhere between 5 percent and 85 percent. Trend studies
describe how the distribution of a variable i s changing for the population
studied. Panel studies describe how individual members of a population
are changing (see Box 21.3 for an example of panel design).

Box 21.3 An Example of Panel Design

Lazarsfeld, et.al, find a classic example of panel design in "The People's Choice".
(1944). The researchers interviewed a panel b f 600 residents of Erie County, Ohio
(USA), once a month between May and November of 1940 with respect to how they
intended to vote in the presidential election of 1940. Of particular interest t o these
investigators was the process by which panel members decided to vote the way
they eventually did.

There are both advantages and disadvantages to panel designs. Most of

the advantages relate to being able to interview the same respondents
repeatedly over a period of months or years to see i f any changes occur in
their thinking and attitudes. If we find shifts in attitudes or reported
behaviour, we are in a better position to argue that there has been a
real change in the population.
Another advantage of panel design i s that they permit much more
,information to be collected about each respondent than i s feasible through
Quantitative and other designs. Panel designs also avoid a heavy dependence on the memory
Survey Methods
of respondents for information about his time covered by the panel.
However, they may s t i l l rely on memory for questions about the
respondent's past.
'The primary disadvantage of panel study is sample mortal.ity; that is,
there will be a loss of panel members owing to lack of cooperation,
death, or change of residence, etc. Another disadvantage is that it
takes a long time to collect data and hence the costs involved may be
more. Further, repeated surveys on the same population may cause
respondents t o become overly sensitized to their role as "selective"
respondents.
Descriptive designs
There are two types of desuiptive designs, namely, cross-sectional designs
and cohort designs.
Cross-sectional designs
In a cross-sectional@design, the data is collected at one point in time.
This is the most frequently used and most fundamental design in social
survey research and is sometimes called 'single-shot survey'. This design
is most appropriate for making inferences about the characteristics of
the population from which the sample is draw&;and Inferences about
the relationships between variables at that point in time at which data is
collected. 'The interviewing for such studies is sometimes carried out in
less than a week, but more typically requires a few weeks and in some
cases a few months. For example, a survey of current people's choices of
a pol,itical party is a cross-sectiranal survey.
A researcher might use the data either to describe a sample on one variable
or a number of variables or to demonstrate the association between these
variables. Consider the following example. A researcher is interested in
exploring the attitudes of the parents on the education of their children.
He selects a sample of households with school going children in a community
(Rural/Urban) and interviews the parents in those select households. After
the interviews are completed, the researcher analyses the data and draws
conclusions. The research design involved here is a cross.sectional design.
One of the advantages of this design is that a researcher can classify a '
sample into many quite different sub-groups t o explore the separate
dimensions of the research topic. A researcher may also use this design
in causal analysis. Interestingly, by asking questions about the past it is
possible to study certain longitudinal (overtime) social phenomena such
as social mobility using a cross-sectional design.
However, the cross-sectional designs have two important l'imitations: (1)
it is difficult to establish the time order of the variables, (2) it is difficult
to analyse change over time. To illustrate the first point, suppose in a
study, we find that the organisational affiliation is associated with political
 Survey Design
arises as to whether our respondents first joined organisations and later
came to feel efficacious because of their experiences in that organisation,
or whether they felt efficacious first, which led them to join a particular
organisation. Alternatively, whether both causal sequences are operating,
either for different people or for the same people. If we gathered data
at one point in time, it i s difficult to know when differences on each
variable developed. Since causes come before effects, it is difficult to
know in a cross-sectional design which variable should be taken as
independent, causal agent.
A second limitation of the cross-sectional design is that it is difficult to
use it for analysing changes over time. Consider the analysis of voting
behaviour. At any given point during an election, we can ask respondents
how they would vote 'if the elections were held today'. However, voting
decisions might change during a campaign and different voters may exhibit
different patterns of change. Some know their choice from the beginning
and never vary; some switch back and forth between candidates; some
are undecided until Election Day. Moreover, people who decide early in a
campaign and never waver may have different characteristics and may
experience different pressures than people who decide late. In sum,
data collected at one tlme make it difficult to understand social processes
that occur over tlme.
11) Cohort deslgns
These forward-looking, or prospective, designs provide data about changes
in a specific population. Suppose a survey of the aspirations of athletes
particlpatin in the 1996 Olympics i s given in 1996, 2000, and 2004. This
8
i s a Cohort design and the Cohort i s 1996 Olympics.
Cohort designs can also be retrospective, or look back over time (a
historical Cohort) i f the events being studied actually occurred before
the onset of the survey. For example, suppose a group of persons was
diagnosed ten years ago with disease X. If you survey their medical
records to the present tlme, you are using a retrospective Cohort.

I Reflection and Action 21.2 I

I In the case of your study of waste disposal methods, you would have formulated
( research questions and then also refined them. For maklng clear the type of research
I questions and the type of research design you have formulated and adopted,
answer the following questions on a sheet of paper.
I Questions
I + Are the questions formulated by you a mix of different types mentioned in
I the text or of one speciflc type only?
I O What type of survey design have you selected for the survey in your research
I on waste disposal methods?
I Write in detail about the type of desiun you have selected and alsa give reasons I
I why you have selected this one and not any of the other designs mentioned in I
I the text. I
I I
L,,--, -,. -----------,,---,--J
Quantitative and
Survey Methods 21.6 Sampling Design
After survey design has been selected, a sampling plan must be chosen.
Many details about sampling have been discussed earlier in the units of
Block 4 and Block 5, and therefore I am including only those that are new.
Why sampling?
Because the researchers may not include the whole of the population in
survey research for reasons of costs and manpower, a part of the
populationi s selected for understanding the characteristics of a population.
Sampling, therefore, i s a process of systematically selecting cases for
inclusion in a research project. A researcher gets a set of cases (or a
sample) that is more manageable and cost effective to work with than
the pool of all cases. For example, it would be much less costly and time
consuming to measure variables on 200 than 20,000 people. The
researcher is not just interested in a small subset of cases. Instead, he
would like to generalise about the entire pool. If well done, sampling lets
the researcher measure variables on the smaller set of cases but generalise
results accurately to all cases. For instance, i f sampling is well conducted,
a researcher can measure variables with, say, 2000 cases, and'generalise
to 200 million. Survey researchers argue strongly that there would only be
2 - 4 percent variation if all the 200 million cases were used for the study.
The next question is how it is possible to generalize from a handful of
cases. It i s not based on any magic, but on logical statistical reasoning
that has been tested repeatedly with empirical evidence. Moreover, a
researcher cannot use just any sample for the purposes of generalization.
There are well laid down sampling procedures that require rigorous
exercises in order to increase the level of precision in a study.
Basic terms i n sampling
i ) Sampling element
A sampling element is the unit of analysis (or case) in a population. It
can be a person, a group, an organisation, a written document, or
symbolic message, or even a social action (a divorce, a fight, etc.) that
i s being measured.
ii) Population
It is the large pool of elements from which a sample i s drawn. Sometimes,
the term 'universe' is used interchangeably with 'population'. To define
the universe or population, a researcher specifies the unit being sampled,
the geographical location, and the temporal boundaries of populations.
The term 'target population' refers to the specific pool of cases the
researcher wants to study.
iii) Sampling ratio .
The ratio of the size of the sample to the size of the target population is
the 'sampling ratio'. For example, a college has 2000 students, and a
researcher draws a sample of 200 from it. Thus the sampling ratio is
20012000 = 0.1, or 10 percent.
iv) Sampling frame Survey Design

A population i s an abstract concept. Except for specific small populations,

one can never truly freeze a population to measure it. For example, in a
city at any given moment, some people are dying, some are traveling
outside the city, some have gone on temporary migration t o other cities,
and some are taking birth. The researcher must decide exactly who t o
count. Should he count a city resident who happens to be on a holiday or
outside the city when the time i s fixed for the study?Therefore, the notion
of 'population' i s abstract and it exists in the mind, but i s impossible t o
pinpoint concretely. Since it i s an abstract concept, except for small
populations, a researcher needs t o estimate the population. Thus, it
requires an operational definition.
A researcher operationalises a population by developing a specific l i s t
that closely approximates all the elements in the population. This l i s t is a
'sampling frame'. The researcher may choose from many types of sampling
frames: telephone directories, driving licences, ration cards, membership
of a club, students' registration in a university, etc. A good sampling frame
i s crucial to good sampling. A mismatch between the sampling frame and
the conceptually defined population can be a m a ~ o sourcer of error. For
instance, i f you select the telephone directory as a sampling frame, it
constitutes only 5-10 percent of the population in a city, say, Delhi. The
directory does not l i s t those who do not have a telephone connection. It
also does not update quickly the frequent shifts of residences and changes
in telephone numbers in a city.
v) Parameter and statistic
Any characteristic of a population i s a 'population parameter'. For
instance, students from the science stream of children in the age group '
6-1 1 years, etc. It i s the true characteristic of the population. Parameters
are determined when all elements in a population are measured. The
parameter i s never known with absolute accuracy for large populations,
so researchers must estimate it as based on samples.
Researchers use information from the sample, called a 'statistic', t o
estimate popu!ation parameters.
Sampling errors
If many samples are taken from the same population, it i s unlikely that
they will all have characteristics identical either with each other or with
the population. In brief, there will be what i s called 'sampl,ing error'.
Sampling error is not necessarily the result of mistakes made in the
sampling procedure. Rather, variations occur due to the chance selection
of different individuals.
Types of sampling designs
Sampling designs are of two types. The first i s called probability-sampling
design and the second i s non-probability sampling design. In Unit 15 of
Block 5 our focus was on how to carry out calculations while here we are
discussing each type of sampling design in terms of i t s essential character
Quantitative and
S~rrveyMethods
and procedure. All the same you are likely to find some repetition of
subject matter which has been retained to reinforce the learning of the
same.
i) Probability sampling design: Probability sampling provides a statistical
basis for stating that a sample is representative of the 'target population'.
In probability sample, every element i n the population has a known
chance of being included in the sample. 'That means, every member of
the target population has a non-zero probability of being included in the
studylsample. This allows for estimates of the accuracy of sample findings
in approximating what we woul,d find out if we had conducted a census
of the total population.
ii) Non-probability sampling design: In non-probability sampling designs,
we do not know whether an element of the population has an equal
chance of being selected. Its probability of selection cannot be
determined, as is that with probability sampling where each element has
a 50 percent chance of being selected and 50 percent chance of not
being included in the sample. The non-probability samples are drawn
based on judgment regarding the characteristics of the target population
and the needs of survey. With non-probability sampling, some members
of the eligible target population have a chance of being chosen and
others do not. Thus, the statistical estimates of precision cannot be
made with this sample. The non-probabllity sampling designs are preferred
when there is no possibil,ity of probabil,ity sampl,ing. Whenever feasible,
probability-sampling designs are preferred.
Probability sampling procedures
1) Simple random sampling
In simple random sampling, each member of the population under study
has an equal chance of being selected. The method involves selecting at
random from a list of the population (a sampling frame) the required
number of subjects for the sample. Because of the probability and chance,
the sample should contain subjects with characteristics similar to the
population as a whole. For example, some old, some young, some tall,
some short, some rich, some poor, etc. One essential requisite for this
kind of sampllng i s that a complete list of the population (sampling
frame) is needed.
The biggest advantage of simple random sampling is that a researcher
can get an unbiased sample without much technical dlfficulty. For instance,
once a member (or element) is selected, he (or it) is not eligible for a
second chance and is not returned t o the pool. 'This is what makes a
simple random sample relatively unbiased. The typical ways of selecting
a simple random sample are mainly through a lottery or through a table
of random numbers or now through computer generated random
numbers. The Lottery method is adopted for smaller populations or sampling
frame. For bigger sampling frames, the computer-generated numbers
are selected.
 It may be kept in mind that simple random sampling may not guarantee Survey Design

a perfect representation of the population. In other words, it may be

wise to say that most random samples are close to the population most
of the time but may not perfectly match the entire population.
ii) Systematic sampling
This is a modified version of simple random sampling. .It involves selecting
cases/elements from a population l i s t in a systematic rather than random
fashion. Here, the researcher calculates a 'sampling interval' rather
than using a list of random numbers. The interval becomes his quasi-
random selection method. Thus, in a systematic sample, every 'n'th
member has a chance to be included in the sample. For instance, i f we
are to select 10 cases out of a total of 100, every tenth will have a
chance to be selected. In this case, 'n' is any number between 1 and 10.
Thus, the starting point in the systematic sampling is chosen at random.
In most cases, a simple random sample and a systematic sample yield
virtually equivalent results. One important situation in which systematic
sampling cannot be substituted for simple random sampling occurs when
the elements in a sample are organised in a cycle or pattern. For example,
if sampling is organised of married couples with the male first and
female second, such a pattern gives the researcher an unrepresentative
sample i f a systematic sample i s a used. His systematic sample can be
non-representative and include only wives or only husbands, depending
upon the manner in which cases are organised or patterned.
II iii) Stratified sampling
Stratified sampling involves dividing the population into homogenous groups
i (sub-groups or strata) each group containing subjects with similar
1 characteristics. For example, in the earlier example, Group A might contain
only males and Group B only females. After dividing the population into
strata or sub-groups, the researcher draws either a simple random sample
or systematic sample or both from each of the sub-groups.
How does one decide on sub-groups?The strata or sub-groups are chosen
I because evidence is available that they are related to the outcome. The
1 justification for the selection of the strata can come from literature and
expert opinion.

I In stratified sampling, the researcher controls the relative size of each

1 stratum, rather than letting random processes control it. This guarantees
I representation of different strata within a sample. However, one condition

i is that the stratified sampling procedures produce samples that are

more representative of the population than simple random sample or
systematic sample i f the stratum information is accurate.

1 iv) Cluster sampling

I The most widely employed probability sample design in survey research
is cluster sampling. It addresses two problems - first is the Lack of good
t sampling frame, and second the costs involved i n reaching a sampled
It element or a case. For example, there is no single list of undergraduates
Quantitative and in colleges of a city. Even i f one gets an accurate sampling frame, it
Survey Methods
would cost too much to reach many of the undergraduates as the colleges
are spread out geographically in the city. In this case, instead of using a
single sampling frame, researchers use a sarr~plingdesign that involves
'clusters'. In this case, the cluster will be the college.
A cluster i s a naturally occurring unit (e.g., a school, which has many
classrooms, students, teachers; a city with zones, namely, Each, West,
South, Central, Iqorth, etc., states, etc.). The clusters are selected
randomly and all members of the selected clusters are included in the
sample or simple random or systematic or stratified samples are taken
out of each cluster. Cluster sampling is used in large surveys. It differs
from stratified sampling in that with cluster sampling one starts with a
naturally occurring constituency. The researcher selects from among the
clusters and either surveys all members of the selection or randomly
selects from among them. The resulting sample may not be
representatives of areas not covered by the cluster, nor does one cluster
necessarily represent another.
v) Stage sampling
Stage sampling, also known as multi-stage sampling, is an extension of
cluster sampling. It involves selection of a sample in several stages. That
is, taking samples from samples. Suppose we want to survey children's
academic performance in schools from a large city. One type of stage
sampling might be to select a number of schools; at random, and from
each of these schools select a number of classes and select children from
within these classes. Another type of stage sampling could be to select
one school, in terms of either the geographic region of the school in the
community/city or in terms of the type of school (publ.ic, private aided,
private unaided, etc.) and take their simple random sampling, systematic
sampling, or stratified sampling or even mixture of all these.
Non-probability sampling procedures
i ) Accidentai or convenience sampling
This i s also called 'man-on-the-street' survey and it involves choosing
the individuals/cases that are readily available on the street, at a market
place, at a school, or at a cinema theatre, etc., until the required
sample size is obtained. This kind of survey can produce ineffective,
highly unrepresentative samples and i s not recommended. When a
researcher haphazardly selects a sample that is convenient, he can easily
get a sample that seriously misrepresents the population. Such samples
are cheap and quick, although biases and errors are there in plenty. An
example of this kind of survey i s the kind of interviews television
programmes conduct on the street.
ii) Quota sampling
Quota sampling has often been described as the non-probability equivalent
of stratified sampiing. In quota sampling, the researcher divides the
population into sub-groups or categories such as men and women, reserved
castes and non-reserved castes, arts and commerce, younger and older, Survey Design

etc., and then decides the proportion of individual in each category or

sub-group. Thus, the number of respondents i s fixed in various categories
of the sample.
Quota sampling i s an improvement because the researcher can ensure
that some population differences are in the sample. However, it i s not
to argue that it i s fully representative. For instance, once quota categories
are selected in quota sampling, the actual number of individuals in each
category i s selected based on the convenience or accidental sampling.
Further, the researcher bias might also creep into the selection of actual
cases of individuals who are included in the study.
iii) Purposive or judgmental sampling
In purposive sampling, researchers handpick the cases t o be included in
the sample based on their judgment of their typicality. That means, the
judgment of the researcher i s used i n selecting cases with a specific
purpose in mind.
Purposive sampling i s appropriate in the following three situations. First,
a researcher uses it t o select unique cases that are especially informative.
Second, a researcher may use purposive sampling to select members of
a difficult-to-reach, specialised population. Third, the purposive sampling
i s preferred when a researcher wants t o identify particular types of
cases for in-depth investigation.
iv) Dimensional sampling
This i s a further refinement of quota sampling. I t involves jdentifying
various factors of interest in a population and obtaining at least one
respondent of every combination of those factors.
v) Snowball sampling
It i s also called a network or chain referral or reputational sampling and
it i s a method for identifying and sampling (or selecting) the cases in a
network. Snowball sampling i s based on an analogy to a snowball which
begins small but becomes larger as it rolls on wet snow and picks up
'
additional snow. Friendship networks are the most important sources of
this kind of sampling. Further, professional associations, chat groups,
etc. may also provide bases for such sampling.
Sample size
A question often asked i s af the size of the sample. How large should it
be? The best answer could be 'it depends'. That means there i s no clear-
cut answer to this question. However, the literature on survey sampling
does give a general advice to prospective survey researchers. For some
people, a sample of thirty i s the bare minimum i f the researcher plans
to some form of statistical analysis of the data. However, the size of the
sample depends on the plans of the researcher and how accurate the
sample has to be for his purposes, and on population characteristics. It
may be stated here that a large sample alone does not guarantee a
 representative sample. Further, when we increase the sample size, we
increase costs. Larger samples mean increased costs. A researcher may
keep in mind all these considerations while judging the size of the sample.
In Section 15.3 of Unit 15, you have already learnt how to calculate
sample size, when estimating mean and when sampling for proportion.

r-------------------------- 1
Reflection and Action 21.3
I
Discuss in a note of about five hundred words the sampling type and design of
I
I the survey you need to carry out in your research on waste disposal methods.
I
I While writing the note, take help,of the relevant text in the unit. I
L-,-,-,,,,,,,------------,-J

2 1.7 Conclusion
'This unit has attempted to discuss the initial concerns of the survey
' researcher. I t spelled out the essential prerequisites for undertaking a
survey and the steps t o be followed in survey research in order t o make
it more systematic. The unit provided a detailed account of research
question formulation, survey research designs, and the sampling designs.
Thus, it tried to equip the student with the basic terms and concepts '

and the process with suitable examples.

Further ~ e a d i n g g
Aldridge, Alan. And Levine, Ken. 2001. Surveying the Social World -
Principles and Practice in Survey Research. open University Press:
Buckingham
Fink, Ar1en.e. 1995. The Survey Hand Book. Sage Publications: Thousand
Oaks
Cohen, Louis and Manion, Lawrence. 1994. Research Methods in
Education. Routledge: London and New York
Moser, C. A. and G. Kalton 1973. Survey Methods in Social Investigation.
The English Language Book Society: London
de Vaus, David. 2002. Surveys in Social Research. Routledge: London