The nature and purpose of social research

Introduction to Social Research Methodology

Author

Affiliation

Ben Stanley

Department of Social Sciences, SWPS University

Published

October 13, 2026

Objectives of social research

Social research serves several connected objectives, which can be thought of as a ladder of increasing ambition (Figure 2).

Figure 2: The objectives of social research as a ladder: each step builds on the one below.

Objective	What it involves
Understand and describe	Capturing social phenomena accurately — what is happening, to whom, and under what conditions
Explain	Identifying the causes and consequences of social events, moving beyond description to explanation
Predict	Using established patterns in existing data to anticipate likely future trends
Inform policy	Providing the evidence base on which governments, organisations, and communities can act

Description comes first because we cannot explain what we have not accurately characterised; explanation, prediction, and policy relevance build on that foundation. Not every study pursues all four objectives — a small exploratory study may aim only to describe — but together they capture the range of purposes that social research can serve. The ladder is also a tool for reading research critically: ask which step a study is standing on. A study that describes a correlation well may be quite unable to support the causal claim or the policy recommendation that someone tries to hang on it.

Types of social research

Textbooks often list four types of social research — qualitative, quantitative, descriptive and experimental — as if every study had to be sorted into one of four boxes. In fact the four belong to two distinct axes:

Axis	Question it asks	Types
Kind of data and reasoning	What kind of evidence, and how is it interpreted?	Qualitative — deep, subjective, interpretative; concerned with meaning and experience. Quantitative — numerical, statistical, generalisable; concerned with measurement and pattern.
Intervention	Does the researcher intervene, or only observe?	Descriptive — detailing events or situations as they are, without intervening in them. Experimental — testing specific hypotheses or theories by deliberately controlling and varying conditions.

Because these are different axes, the types combine rather than compete. Figure 3 crosses the two axes into a grid: the columns are the first axis, the rows the second, and every one of the four cells is a possible study. A large survey that maps how turnout varies with age is quantitative and descriptive. Sending a text-message reminder to a randomly chosen half of voters and comparing turnout across the two groups is quantitative and experimental. In-depth interviews about how first-time voters experienced an election are qualitative and descriptive. Trying out a new civic-education workshop in some schools and running focus groups to hear how pupils experienced it is qualitative and experimental. So when you classify a piece of research, do not ask “which one of the four is it?” but “where does it sit on each of the two axes?”

Figure 3: Types of social research: two axes, four combinations.

Qualitative and quantitative approaches

The contrast between qualitative and quantitative research is the most important single distinction in social methodology, and much of the rest of the course elaborates on it. It also reflects two different directions of reasoning.

Two logics: induction and deduction

Research can reason in two directions, and the direction it takes shapes the whole design. Deductive reasoning starts from theory: the researcher derives a specific hypothesis from a general proposition and then gathers data to test it, moving from the general to the specific. This logic is characteristic of quantitative work, where a clear expectation is specified in advance and then confronted with evidence. Inductive reasoning runs the other way: the researcher begins with observations, looks for patterns in them, and builds theory up from the ground, moving from the specific to the general. This logic is characteristic of qualitative work, where the aim is often to discover concepts and relationships rather than to test ones specified beforehand.

Figure 4 draws the two as halves of one cycle. Theory, the general, is at the top; observations, the specific, at the bottom. The deductive path runs down the left, from theory through a hypothesis to the observations that test it; the inductive path runs up the right, from observations through patterns to theory. In practice the two are complementary, and most research programmes travel round the whole cycle — an inductive insight in one study becomes a deductive hypothesis in the next.

Figure 4: Deductive and inductive logic as two halves of one cycle.

Qualitative research

Qualitative research focuses on human experiences, meanings, and interpretations — on how people make sense of their social world. It typically produces rich, detailed, contextual data drawn from relatively small numbers of participants, and it is especially valuable for exploring phenomena that are new, sensitive, or poorly understood, where the relevant questions are not yet clear enough to be turned into fixed measures.

Its characteristic methods, pictured in Figure 5, include:

In-depth interviews — exploring individual perspectives in detail, allowing unexpected themes to emerge.
Focus groups — observing how views are formed, contested, and revised through interaction between participants.
Ethnography — immersing the researcher in a social setting over time to observe behaviour in its natural context.

Figure 5: Qualitative research: its focus, three methods, and the data it produces.

Quantitative research

Quantitative research seeks to quantify variables — to measure how they are distributed and to identify patterns and relationships among them. It produces standardised data that can be summarised, compared across cases, and tested for statistical significance. With appropriate sampling, its findings can be generalised from the sample studied to the larger population from which the sample was drawn. The cost is depth: a survey can show that turnout among the young has fallen, and by how much, but it is far less good at showing what voting means to a young person.

Its characteristic methods, pictured in Figure 6, include:

Surveys administered to large samples, using standardised questions.
Experiments that manipulate one or more variables under controlled conditions, comparing a group that receives the treatment with a control group that does not.
Statistical analyses of existing numerical datasets.

Figure 6: Quantitative research: its focus, three methods, and the data it produces.

Qualitative and quantitative compared

Table 1 sets the two approaches side by side. Each row is a dimension on which they differ, and the strength of each is the mirror image of the other’s weakness.

Table 1: Qualitative and quantitative research compared.

	Qualitative	Quantitative
Data	Words, images, meanings	Numbers, measurements
Logic	Inductive — builds theory	Deductive — tests theory
Samples	Small, purposively chosen	Large, ideally representative
Methods	Interviews, focus groups, ethnography	Surveys, experiments, statistics
Strength	Depth, context, discovery	Breadth, comparison, generalisation

Mixed methods

The two approaches are best seen as complementary rather than opposed. Qualitative work is strong on depth, meaning, and discovery; quantitative work is strong on breadth, measurement, and generalisation. Many of the most convincing studies combine them in a mixed-methods design. Figure 7 shows the two simplest combinations. In the first row, qualitative interviews come first, to understand a phenomenon well enough to design a good survey — the inductive phase feeds the deductive one. In the second row, a large survey comes first, establishing the pattern, and its results are used to pick a handful of especially interesting cases to study in depth. Session 9 names these the exploratory sequential and the explanatory sequential design.

Figure 7: Two ways of combining qualitative and quantitative research in one study.

The research process

Whatever the approach, social research moves through a recognisable sequence of stages, drawn as a chain in Figure 8: choosing a topic, designing the study, sampling, collecting data, analysing and interpreting it, presenting the findings, and reflecting on the study’s limits. The links are not independent tasks to be done in any order. Each decision constrains the ones that follow, so the chain is only as strong as its weakest link: a brilliant analysis cannot rescue a biased sample, and a perfect sample cannot rescue a poorly worded question.

Figure 8: The stages of a study as a chain of seven links, with the running example threaded through them.

The lecture followed one question through every stage, shown as the orange thread in Figure 8 and in orange on each figure below: why do young people vote less than older people? The sections below follow the same sequence, from the choice of topic to the acknowledgement of limitations, each ending with what that stage means for the running example.

Choosing a research topic

A good research topic satisfies three criteria simultaneously, and Figure 9 draws them as three overlapping circles, with a good topic in the centre where all three meet:

Relevance — it speaks to a current societal issue or to a genuine gap in existing knowledge.
Feasibility — it is achievable within the resources, time, expertise, and access actually available to the researcher.
Ethics — its potential benefits outweigh any risk of harm to the people it involves.

Figure 9: A good research topic meets three criteria at once; the running example passes all three.

Choosing a topic is therefore an exercise in balance: the most interesting question is of little use if it cannot be answered with the means at hand or cannot be pursued responsibly. Beginning researchers most often fail on feasibility, choosing a sweeping question — “what causes political apathy?” — that no single study could answer.

Running example. “Why do young people vote less?” passes all three tests (the orange ticks in Figure 9): it is relevant, because turnout among the young is falling in many democracies; feasible, because survey data on turnout and age already exist; and ethically low-risk, because it asks people about their voting rather than exposing them to harm.

Research design and planning

The research design is the blueprint that connects the research question to the evidence needed to answer it. Figure 10 draws it that way: the topic enters on the left, the evidence needed to answer it leaves on the right, and the design in between consists of three tasks:

Crafting clear research questions or hypotheses that the study can realistically answer.
Defining the scope of the study — what is being studied, who is included, and when and where the research takes place.
Deciding on the methods and data-collection tools appropriate to those questions, whether qualitative, quantitative, or mixed.

Figure 10: Research design as the blueprint between a topic and the evidence.

A weak design cannot be rescued by sophisticated analysis later; decisions taken at this stage shape everything that follows.

Running example. The design stage sharpens the broad topic into a testable question: not “why do young people vote less?” but does political interest explain the age gap in turnout? That question has a clear structure: it names what must be measured and suggests how we will know whether it has been answered.

Variables and operationalisation

Designing a study also means deciding what, exactly, will be measured. A variable is any characteristic that varies across the cases under study — age, income, turnout, political interest. In explanatory research it is common to distinguish the independent variable (the presumed cause) from the dependent variable (the presumed effect); the research question typically asks whether, and how, changes in the former are associated with changes in the latter. The top half of Figure 11 shows the two roles, joined by an arrow.

Most of the concepts social scientists care about — engagement, trust, prejudice, wellbeing — are abstract and cannot be observed directly. Operationalisation is the process of turning such a concept into something concrete and measurable. The bottom half of Figure 11 shows an example: “political engagement” operationalised through three indicators — whether a person voted, how often they discuss politics, and whether they belong to a party. How faithfully those indicators capture the underlying concept is precisely the question of validity, marked by the red bracket and discussed below. Good operationalisation is what makes the leap from an interesting idea to a researchable study.

Figure 11: Variables, their roles, and operationalisation.

Running example. Political interest is the independent variable and turnout the dependent variable. We operationalise interest as a self-rating on a 0–10 scale and voting as turnout in the most recent national election. Writing those definitions down turns the idea into a researchable study — and exposes it to the question of whether the measures really measure what we think they do.

Sampling techniques

Because researchers can rarely study an entire population, they study a sample — a subset chosen to stand in for the whole. How that sample is selected determines what can legitimately be claimed on the basis of it.

Sampling type	Principle	Example	Trade-off
Probability	Every member of the population has a known, non-zero chance of selection	Simple random sampling	Supports generalisation and statistical inference, but demands a complete sampling frame
Non-probability	Participants are selected by specific criteria or convenience	Convenience, quota, or snowball sampling	Faster and cheaper, but carries a higher risk of bias

Figure 12 shows why the difference matters. The population on the left mixes younger voters (dark dots) and older voters (light dots), and the younger ones are concentrated in one corner — the part of the population that is easiest to reach. The probability sample (blue rings) is scattered across the whole population, and its make-up mirrors the population’s: three younger voters in ten, against one in four overall. The non-probability sample (red rings) is drawn entirely from the convenient corner, and seven of its nine members are younger voters: it is skewed towards whoever is easiest to reach.

Figure 12: Probability and non-probability sampling from the same population.

The overriding concern is representativeness: how faithfully the sample reflects the population it is meant to describe. Probability methods make representativeness assessable; non-probability methods may still be appropriate, especially in qualitative work, but their limits must be acknowledged.

Running example. The population is all eligible voters. A random sample of 1,500 of them lets us compare turnout across age groups with confidence that the differences reflect the population rather than a quirk of who ended up in the sample. A convenience sample of our own students would tell us almost nothing about voters in general, however easy it would be to collect.

Data collection methods

The choice of data-collection method should fit the research question, the population, and the practical constraints of the study.

Method	What it gathers
Surveys	Standardised information from many respondents through structured questionnaires
Observations	Systematic records of behaviours or events as they occur, with or without participation
Interviews	Direct responses from participants, ranging from tightly structured to open and conversational
Archival sources	Pre-existing records, documents, and datasets, used rather than newly created

No method is best in the abstract. Figure 13 gives each method a characteristic strength (in green) and limitation (in red): surveys excel at breadth and comparability but are weak on depth and context; observation captures what people actually do rather than what they say, but is labour-intensive and hard to generalise from; interviews offer depth and nuance but are time-consuming and shaped by the relationship between interviewer and interviewee; and archival work is unobtrusive, suiting the past and sensitive topics, but is limited to the records that happen to exist.

Figure 13: Four methods of data collection, each with a strength and a limitation.

Running example. A structured survey (outlined in orange in Figure 13) is the natural fit: it reaches a large, representative sample and asks everyone the same questions about their voting and political interest, producing the standardised, comparable data the analysis needs.

Data analysis and interpretation

Once data have been gathered, they must be analysed and interpreted. The techniques differ by approach, and Figure 15 shows the two routes:

Qualitative analysis identifies themes and patterns of meaning, through methods such as thematic analysis and content analysis.
Quantitative analysis summarises and models data using descriptive statistics, regression, t-tests, and related techniques.

Figure 15: Two routes from raw data to an answer to the research question.

Both routes end in the same place: an answer to the research question. But, as the red warning in Figure 15 says, an answer is only as trustworthy as the measures behind it. Sophisticated analysis can dress up a bad measurement in respectable clothing; it cannot make it true.

Running example. The analysis has two steps. First, compare turnout across age bands — descriptive statistics that establish the pattern. Then use regression to ask whether political interest accounts for the gap: once interest is taken into account, how much of the age difference in turnout remains? That is the move from describing a pattern to explaining it, the second step of the ladder in Figure 2.

Validity and reliability

Two qualities of measurement matter regardless of approach, and they are easily confused:

Quality	Question it answers	Plain meaning
Validity	Are we measuring what we actually intend to measure?	Accuracy
Reliability	Would the same procedure produce the same result again?	Consistency

Figure 16 pictures the difference as an archer’s target, with each dot an arrow. On the first target the arrows are tightly clustered but off the bullseye: the measure is reliable but not valid — consistently wrong. On the second they are scattered around the bullseye, none of them on it: centred on the right place on average, but not reliable. Only on the third, tightly clustered on the bullseye, is the measure both reliable and valid.

Figure 16: Validity and reliability pictured as an archer’s target.

A measure can be reliable without being valid — consistently measuring the wrong thing — so both must be established. If a 0–10 “political interest” scale actually captured how much someone enjoys arguing, it could give consistent results every time and still measure the wrong thing.

Presenting research findings

Research has little value if its findings cannot be understood by those who need them. Good presentation rests on three habits:

Clear communication — explaining results in plain language and avoiding unnecessary jargon.
Visual aids — using well-designed charts, graphs, and tables to make patterns easier to grasp.
Relating findings back to the original research questions or hypotheses, so the audience can see what has actually been learned.

Figure 17 shows the same finding — turnout by age group, using illustrative figures rather than real survey results — presented two ways. The left-hand chart breaks all three habits: its title is jargon, its six colours mean nothing and have to be decoded from a legend, its gridlines bury the bars, and nothing says what the chart means. The right-hand chart, which is the running example, follows them: a plain-language headline (1), a clean chart with one colour, the youngest group highlighted and the values written on the bars (2), and one sentence relating the result back to the research question (3).

Figure 17: The same finding presented two ways (illustrative data).

The goal is to make the evidence accessible to its intended audience, whether academic, policy-making, or public — and the same result may need presenting differently to each.

Limitations and critiques

Every study has weaknesses, and acknowledging them openly is a mark of good research rather than a failure. Sound practice involves the three habits in the top row of Figure 18:

Honestly acknowledging the weaknesses of the study’s design, data, or analysis.
Being open to feedback and critique from peers, reviewers, and the wider community.
Using identified limitations to clarify what future research should address.

Figure 18: Acknowledging limitations, and how they move knowledge forward.

This candour is not a formality. It is precisely what allows knowledge to accumulate, as the middle row of Figure 18 shows: each study’s acknowledged limits become the next study’s questions.

Running example. Self-reported turnout is usually over-stated, because voting is socially approved and people are reluctant to admit they did not vote. That is a real weakness of our survey, to be flagged plainly — and flagging it immediately suggests the next study, which validates self-reports against official voting records.

Conclusion

Methodology matters because the quality of our conclusions depends directly on the quality of our methods. Social research is also an evolving practice — new data sources, tools, and analytical techniques continually reshape what is possible and what counts as good evidence. Above all, the stages reviewed here — choosing a topic, designing the study, sampling, collecting data, analysing it, presenting the results, and acknowledging limitations — are not isolated steps but a connected chain of consequential choices. A decision taken early constrains what can be claimed at the end. The running example made the same journey: from a vague question about young people and voting, through a testable hypothesis, operationalised measures, a representative sample, a survey and an analysis, to an honest admission of its limits. Holding the whole process in view is what it means to think methodologically about the social world.

What is social research?