Outline of Chapter 7: The Nature of Quantitative Research

1. What is quantitative research?

2. The main steps in quantitative research

3. Concepts and their measurement

4. Indicators — direct and indirect

5. Multiple-indicator measures and the Likert scale

6. Dimensions of concepts

7. Reliability and its three types

8. Validity and its five types

9. The four main preoccupations of quantitative researchers

10. Criticisms of quantitative research

11. Is it always like this? (gaps between ideal and practice)

Full Explanation of Chapter 7

1. What is Quantitative Research?

Quantitative research is a research strategy that collects and analyses numerical data. It follows a
broadly deductive approach — meaning it starts with a theory or idea, then collects data to test it. It
is rooted in positivism, meaning it tries to study the social world in an objective, scientific way, like
how a natural scientist would study nature. But the chapter makes clear that quantitative research is
more than just using numbers — it has a whole philosophy and set of practices behind it.

2. The Main Steps in Quantitative Research

The chapter presents quantitative research as a series of steps, though it notes that in real life these
steps are rarely followed this neatly. The steps are:

Step 1 — Theory: The researcher starts with a theory or a body of existing knowledge about a topic.

Step 2 — Hypothesis: From the theory, the researcher derives a hypothesis — a specific prediction
that can be tested. Not all quantitative research involves a formal hypothesis, but it is common,
especially in experiments.

Step 3 — Research design: The researcher decides on the overall plan for the study — for example, a
survey, an experiment, or a longitudinal study.
Step 4 — Devise measures of concepts: The researcher figures out how to measure the concepts
they are interested in. This is called operationalization — turning abstract ideas into measurable
things.

Step 5 — Select research site: The researcher chooses where the study will take place.

Step 6 — Select respondents or subjects: The researcher decides who to include in the study —
usually through a sampling process.

Step 7 — Collect data: The researcher administers questionnaires, conducts structured interviews, or
runs experiments to gather information.

Step 8 — Process data: The raw information collected is converted into data — usually numbers —
so it can be analysed, often using a computer.

Step 9 — Analyse data: Statistical techniques are used to look for patterns, relationships, and trends
in the data.

Step 10 — Findings and conclusions: The researcher interprets the results and draws conclusions.

Step 11 — Write up: The findings are written up and shared publicly through academic journals,
reports, or books. Once published, the findings feed back into the body of knowledge and can inform
future research, completing a loop back to Step 1.

3. Concepts and Their Measurement

A concept is simply a label we give to something we notice in the social world. Examples include
social class, job satisfaction, religious belief, academic achievement, and so on. Concepts are the
building blocks of theory.

In quantitative research, concepts must be measured. Once measured, a concept can be either an
independent variable (a cause or explanation) or a dependent variable (something being explained or
affected).

Why do we measure? The chapter gives three reasons. First, measurement helps us detect fine
differences between people that we might otherwise miss. Second, it gives us a consistent tool that
any researcher can use, making results comparable and reliable. Third, it allows us to precisely
estimate how strongly two concepts are related to each other.

4. Indicators — Direct and Indirect

Because many social concepts like job satisfaction or prejudice cannot be directly counted the way
income or age can, researchers use indicators — things that stand in for the concept and allow it to
be measured indirectly.

For example, you cannot directly see "job satisfaction," but you can ask people a series of questions
about how they feel about their work, and treat their answers as an indicator of it.
Indicators can come from many sources — survey questions, observations of behavior, official
statistics, or analysis of media content. The key point is that an indicator is something that represents
a concept even when the concept itself cannot be directly measured.

5. Multiple-Indicator Measures and the Likert Scale

Rather than relying on just one question or indicator, researchers often use multiple indicators of a
single concept. This is better because one question might be misunderstood, might only capture part
of the concept, or might miss important aspects.

For example, if you wanted to measure job satisfaction, asking only about pay would miss
satisfaction with the work itself, the working conditions, and relationships with colleagues. Multiple
questions together give a fuller and more accurate picture.

The most common tool for doing this is the Likert scale, named after Rensis Likert. This involves
presenting respondents with a series of statements and asking them to indicate how strongly they
agree or disagree, usually on a 5- or 7-point scale ranging from "strongly agree" to "strongly
disagree." Each person's scores on all the items are added together to give an overall score.

For the Likert scale to work well, the items must all be statements (not questions), they must all
relate to the same topic, and they should be interrelated — meaning people who score high on one
item should tend to score high on others too.

6. Dimensions of Concepts

Some concepts are complex and have multiple dimensions — different aspects or components. For
example, the concept of "alienation" was broken down into five dimensions: powerlessness,
meaninglessness, normlessness, isolation, and self-estrangement. The idea is that someone might
score high on one dimension but low on another, giving them a more detailed and accurate profile.
This kind of multi-dimensional thinking is more sophisticated than treating a concept as having just
one meaning.

7. Reliability and Its Three Types

Reliability refers to the consistency of a measure. There are three types.

Stability means that if you measure the same thing twice, you should get similar results. The test-
retest method checks this by administering the same measure to the same people at two different
points in time and seeing if the results are consistent. If results change a lot, the measure may be
unreliable. The difficulty is that you cannot always tell whether a change in results is because the
measure is unreliable or because something genuinely changed for the respondents in between the
two tests.
Internal reliability applies to multiple-indicator measures. It asks whether all the individual items in a
scale are actually measuring the same thing. If people score high on some items but low on others in
a seemingly random way, the items may not be measuring the same concept and the scale lacks
internal reliability. The most common way to test this is Cronbach's alpha, a statistical test that gives
a number between 0 and 1. A score of 0.80 or above is generally considered acceptable.

Inter-observer consistency refers to situations where more than one person is recording or
categorizing data. If different observers categorize the same behavior differently, there is a reliability
problem. This is particularly relevant in structured observation and content analysis.

8. Validity and Its Five Types

Validity asks whether your measure is actually measuring what you think it is measuring. Even if a
measure is reliable (consistent), it might not be valid (accurate). Reliability is necessary but not
sufficient for validity.

Face validity is the most basic level. It simply asks whether the measure appears, on the face of it, to
be measuring the right thing. Experts in the field might be asked whether the measure looks
appropriate. This is an intuitive judgment rather than a statistical test.

Concurrent validity checks whether the measure relates appropriately to another measure taken at
the same time. For example, a measure of job satisfaction could be tested by seeing whether people
who score low on job satisfaction also tend to be absent from work more often. If there is no
relationship, the validity of the job satisfaction measure is called into question.

Predictive validity is similar but uses a future criterion rather than a simultaneous one. For example,
does a measure of job satisfaction taken now predict absenteeism in the future? If it does, this
supports the predictive validity of the measure.

Construct validity involves testing hypotheses derived from theory. For instance, if theory says that
unsatisfied workers tend to do more routine jobs, you could test whether your measure of job
satisfaction actually shows this expected pattern.

Convergent validity involves comparing your measure with a completely different way of measuring
the same thing. If both measures give similar results, this supports the validity of both. A famous real-
world example involves British crime statistics compared with the British Crime Survey, which
measured crime differently by asking people about their experiences as victims. The two measures
gave very different pictures, showing a lack of convergent validity — and raising serious questions
about which one was more accurate.

The chapter also notes honestly that in practice many researchers do not formally test reliability and
validity. Most measurement in sociology is done by what Cicourel called "measurement by fiat" —
meaning measures are simply assumed to be valid without proper testing.

9. The Four Main Preoccupations of Quantitative Researchers

Causality is about explanation. Quantitative researchers do not just want to describe what is
happening — they want to know why. They think in terms of independent variables (causes) and
dependent variables (effects). For example, does authoritarianism cause racial prejudice? The
challenge is that in most survey research, data on all variables are collected at the same time, making
it difficult to prove that one variable caused the other rather than the other way around.
Experiments handle this better because they can control and manipulate variables.

Generalization means wanting the findings from a study to apply beyond the specific people studied.
If you survey 1,000 people in one city, can you say the findings apply to the whole country? To do
this, researchers try to use probability sampling — selecting respondents randomly so the sample is
representative of the wider population. However, strictly speaking, you can only generalize to the
population from which the sample was drawn, not beyond it, though many researchers push their
generalizations further than they should.

Replication means that research should be designed so that other researchers could repeat it and
check whether they get the same results. This is important for checking objectivity and reducing bias.
However, in practice replication is rare in social science because it is considered unexciting, it is hard
to reproduce the exact same conditions, and journals are less likely to publish replications.

10. Criticisms of Quantitative Research

Qualitative researchers have raised four major criticisms of quantitative research.

First, quantitative research treats people the same as the objects of natural science — as if humans
were like molecules or atoms that simply react to external forces. Critics argue that humans are
fundamentally different because they give meaning to their actions and interpret the world around
them.

Second, measurement gives only a fake sense of precision. The connection between a measure and
the concept it is supposed to represent is often just assumed, not proven. Also, people may interpret
the same question differently, yet quantitative research treats their answers as if they all mean the
same thing.

Third, the use of standardized instruments like questionnaires disconnects research from real life.
When people answer survey questions, we do not know how much their answers actually reflect
their real-world behavior and experiences.

Fourth, focusing on relationships between variables creates a static, frozen picture of social life. It
ignores the process by which people actually interpret and create meaning in their daily lives.

11. Is it Always Like This?

Reverse operationism happens when researchers collect data and then develop their concepts from
the data, rather than specifying concepts first. This is essentially an inductive approach sneaking into
what is supposed to be a deductive process.

Reliability and validity testing is often skipped in practice because it costs too much time and
money. Studies show that the vast majority of published quantitative research does not properly test
the stability or validity of its measures.

Sampling is another area where ideal practice is often not followed. Proper probability sampling is
expensive and time-consuming, so many researchers use non-probability samples — which then
limits the statistical analyses they can validly perform.

The chapter's honest conclusion is that the model in Figure 7.1 represents good practice and a useful
guide, not a description of what always actually happens.