You are on page 1of 8

Research design

This section looks more generally at the design of sociological research. Oberg (1999) suggests
there are four interconnected stages of research design:

1 Planning is where the researcher decides on the strategy - such as what to research and how to
research it and formulates research questions or hypotheses (assumptions, suppositions, guesses).

2 Information gathering involves identifying a sample (groups of people) to study, conducting an


initial pilot study and applying research methods to collect data.

3 Information processing relates to the idea that once data has been gathered, its meaning must
be analysed and interpreted.

4 Evaluation involves both an internal analysis that asks questions about how the research was
conducted (for example, whether the research method was appropriate) and an external analysis
whereby conclusions are reported to a wider public audience for their analysis and criticism.

A. The research problem

This is the initial stage, when the sociologist decides things such as the general topic to study and
then develops more specific ideas about what aspect to study. This may be accompanied by a
review of previous research (also known as Literature Review) in the area under consideration. A
review like this may generate ideas about what to study, whether to replicate previous research
and how to avoid errors made in previous research.

B. Research hypothesis or question

You can start a research study based on two different grounds:

> If a hypothesis/assumption is used as starting point- Ginn and Arber's (2002) analysis of how
motherhood affects the lives of graduate women was based on the hypothesis 'The effect of
motherhood on full-time employment is minimal for graduate women' - it must be tested and this
means using research methods suitable for this purpose.

> If a research question is used as starting point- Conway's (1997) examination of parental
choice in secondary education was based on the question 'Does parental choice help to
strengthen the advantage of the middle classes over the working class?'- the research method
used must be capable of generating high levels of descriptive data.

In other words, you can either start your research by basing yourself on an
assumption/hypothesis or by focusing yourself on a question; then you conduct your study.
C. Collecting data

Before data can be collected, the researcher needs to identity the people - or respondents - who
will be the subject/sample of the research. Although it would be ideal to select and study
everyone in a particular group (the target population), this is not always possible. For example, if
the target population was 'doctors in India', the size and geographic distribution of such a
population would make it impossible to observe or question everyone personally. This is where
sampling enters the research process.

A sample is a relatively small proportion of people who belong to the target population. In the
example above, the researcher might choose 1,000 doctors and by studying their behaviour, the
researcher may draw certain conclusions about all doctors. However, this only works if the
sample is representative (demonstrative) of the target population. Representativeness may be
more significant than sample size because it relates to whether the characteristics of the sample
accurately reflect those of the target population. If 60% of the target population are male, then
60% of the sample should be male; it cannot be the contrary, or else it will be unrepresentative
(unreliable, misleading, illusory).

If the sample is really representative, then anything discovered can be generalised to the target
population. A researcher can make statements about a group they have not studied (the target
population) based on the behaviour of a group they have studied (the sample).

Sampling Frame

Sampling: a sample involves a small number of people drawn from a much larger (target)
population.

Constructing a representative sample often requires a sampling frame. This is a list of everyone
in a target population (such as an electoral or school register).

In order for a researcher to contact people in his/her sample, to interview them for example, the
researcher must logically know who the people are. However, simply because a sampling frame
exists does not mean a researcher will automatically have access to it. Access to respondents may
be denied for such reasons:

>Legality: names cannot be revealed by law

>Confidentiality: for example, a business may deny access to its payroll records

>Privacy: some groups do not want to be studied.

There are a number of sampling methods used by sociologists to try gaining a representative
sample, which will be discussed below.
1. Simple random sampling

This is based on the probability that the random selection of names from a
sampling frame will produce a representative sample. In other words, random
samples are based on chance distributions. For the sample to be truly random,
everyone in the target population must have an equal chance of being chosen. A simple
random sample, therefore, is similar to Lotto:

 Everyone in the target population who has played the Lotto is identified on a sampling
frame. All these people have equal chance of winning the Lotto. It is a question of luck!
 Then, on TV the balls are selected by random/ or by chance until the sample is complete
(that is, a set a six balls). It is only then that we would get to know who has been chosen
as the winner of the game.

Another example: A 30% sample of a target population of 100 people would involve the random
selection of 30 people.

The problem with this method is that, purely by chance, the random sample may not be
representative of the survey population. For example, there may be too many people of one sex,
of one age group, or who live in the same area.

2. Systematic sampling

This is a variation on simple random sampling that is often used when the target population is
very large. It involves taking sample/people directly from a sampling frame. For a 25% sample
of a target population containing 100 names, every fourth name would be chosen.

This technique is not truly random- for example, the fifth name on the list could never be
included in the sample; so not everyone has an equal chance of inclusion. However, it is random
enough for most samples.

3. Stratified random sampling

Stratified/divided random sampling is a way of attempting to avoid the possible errors caused by
simple random sampling. In this case, the sampling frame is divided into strata (layers) or sub-
groups relevant to the hypothesis/assumption being investigated (such as groups of a similar age,
sex, ethnic group, or social class) and a random sample is then taken from each sub-group. In
other words, each sub-group is treated as a separate random sample in its own right.

For example, in a survey of doctors, we may know from previous research that 8% of all doctors
are Asian, and so the sociologist must make sure 8% of the sample also are Asian.
As such, stratified random sampling has the advantage of being much more representative than
simple random sampling, because all the characteristics of the survey population are more certain
to be represented in the sample.

4. Stratified Quota Sampling

Although a sampling frame is always useful, it is not strictly necessary. It is enough just to know
the characteristics of the respondents in order to construct a sample. Quota sampling is a method
in which interviewers are told to go and select people who fit into certain categories, according
to their proportion in the survey population as a whole.

For example, an interviewer may be asked to question thirty men and women over the age of 45.
Yet, the choice of the actual individuals who have been chosen is left to the honesty of the
interviewer (unlike other sampling methods in which actual named individuals are identified).
Now, once the thirty people have agreed to participate in the study, the quota for the sample is
complete and no further people can be selected. As such, this technique is not truly random
because not everyone in the target population has an equal chance of being selected.
5. Non-representative sampling

Researchers generally find representative sample useful, but there are times when a non-
representative sample serves the purpose. For some types of research, the sociologist might not
want to make generalisations about a very large group based only on a small sample. They might
simply be interested in the behaviour of the group itself rather than what it represents.

Case studies, for example, involve studying the behaviour of a particular group (or case) in great
detail and can be used to illustrate how a non-representative sample work. The 'sample' is the
target population. In Ward's research on rave participants and Venkatesh's gang study, the fact
that each small group was only representative of itself was not viewed as an issue, since neither
of these two sociologists wanted to generalise their findings; they simply wanted to understand a
particular group in depth.

6. Opportunity sampling

In some circumstances it may not be possible to create a representative sample. Here, the
researcher may be forced to settle for opportunity sampling, a general type of sampling - with
two main sub-divisions:

(i) Best opportunity sampling involves deliberately choosing a sample that gives the best
possible opportunity to test a hypothesis. If the hypothesis is false for this group, it will probably
be false for other similar groups. Goldthorpe et at. (1968), for example, wanted to test the claim
that the working class in the UK was becoming almost identical to the middle class. Their best
opportunity sample consisted of highly paid car-assembly workers in Luton. This group was
chosen because if any working-class group was likely to show lifestyles similar to their middle-
class peers, it would be these 'affluent workers'.

(ii) Snowball samples work on the principle of 'rolling up' more and more people to include in
the sample over time. The researcher would identify someone in the target population who was
willing to participate in their research. This person then suggests more people who are also
willing to participate. These then suggest further possible participants, until the researcher has a
usable sample. Although this technique is unrepresentative, it may be the only option in certain
situations:

 Wallis (1977) used snowball sampling to contact (ex-)members of the Church of


Scientology when his request to interview current members was rejected.
 In Sappleton et al.'s (2006) research into gender segregation, 'respondents were enlisted
through personal appointments, prior contacts and cold calls'.

Opportunity sampling can be a useful technique when no sampling frame is available and when
the researcher knows little or nothing about the characteristics of their target population.
7. Pilot study

Before embarking on a full-scale study, many researchers choose to run a pilot study to test the
various elements of their research design. Pilot studies are a research tool normally used for one
of two reasons:

(i) As a 'mini version' of a full-scale study designed to test the feasibility of carrying out such a
study. In other words, before embarking on a study that may take up large amounts of time,
money and effort, a researcher may conduct a smaller study to identify any problems, such as
access to respondents, that may occur in a larger study. A pilot study is also helpful in
determining the resources (such as staffing and finance) needed for a study. The results of a pilot
study can be used to demonstrate to funding bodies that a full study would be feasible and
worthwhile.

(ii) To pre-test a research method, such as a questionnaire. This might involve testing different
types of question, examining and analysing the data it produces to ensure the questions will
produce the data required, as well as identifying and eliminating possible sources of bias or
unreliability (such as misleading or ambiguous questions).
D. Data Analysis

Foucault (1970) argues that data can never speak for itself; it has to be analysed and interpreted
(by bringing together and categorising related ideas). Analysis and interpretation takes place on
three levels:

1. Private/internal analysis involves using concepts like reliability and validity to ensure data is
logical and consistent.

2. Practical analysis relates to the purpose of data collection- the idea of doing something with
the data. Wilkinson and Pickett (2009), for example, used comparative data drawn from
secondary sources to make extensive correlations between social inequality and crime. These
suggested a possible causation: the more unequal a society is, the higher its relative level of
crime.

3. Public/external analysis relates to the idea that all research represents the outcome of a process
of social construction. Not everything the researcher saw, heard or recorded is presented for
public consumption. This is partly because it would result in lengthy reports and partly because
some data may be considered irrelevant to the research objectives.

E. Presenting completed research

Glaser and Strauss (1967) suggest that the final stage of the design process involves four related
elements:

1. analysing related research to discover common themes and trends in the data

2. reflecting on the research itself; does it, for example, support or contradict the hypothesis?

3. is it possible to discover patterns in the data?

4. does the research suggest ways the data can be linked to create an overall theory?

Once the data has been analysed and interpreted it can be presented in terms of its:

 findings
 conclusions about the hypothesis (has it been supported or contradicted?)
 limitations, which might include discussion of various research problems that may have
affected the study
 suggestions for further research
 improvements to the research design.
Summary view of Research Design

You might also like