Understanding Measurement in Social Science Research

Numbers shape almost every conclusion drawn in social science. When a researcher claims that contraceptive use has risen, that women’s autonomy varies across states, or that a particular health programme is working, those statements rest on something more fundamental than opinion. They rest on measurement. Yet measurement in the social sciences is not as straightforward as placing a thermometer in a glass of water. It involves translating fuzzy human realities, like attitudes, beliefs, and behaviours, into something countable. Understanding how this translation works is the first step toward producing research that holds up to scrutiny.

What measurement really means 🔗

At its simplest, measurement is the process of assigning numbers or symbols to characteristics of people, objects, or events according to a set of rules. The most influential definition was offered by the American psychologist S.S. Stevens, who described measurement in a 1946 article in the journal Science as the assignment of numerals to objects or events according to rules. The phrase “according to rules” is the part that often gets overlooked, but it is the heart of the definition. Without explicit rules, assigning a number is just guesswork.

Stevens’ contribution was revolutionary because it gave social scientists a defensible way to quantify phenomena that physicists and chemists once dismissed as unmeasurable. Before his framework, a committee of British scientists in the 1930s had argued that psychological attributes like loudness or anxiety could not be measured at all because they could not be physically added together the way lengths or weights can. Stevens responded by redefining measurement itself in broader terms, opening the door to systematic study of mental and social life.

The four levels of measurement 🔗

Along with his definition, Stevens proposed four levels at which measurement can occur. These levels remain the standard vocabulary in research methodology today.

Nominal scales simply place observations into categories that do not overlap. Religion, blood group, and marital status are nominal variables. The numbers attached to each category, if any, are just labels and carry no quantitative meaning.

Ordinal scales rank observations in order but do not assume equal distances between ranks. Socio-economic status categorised as low, middle, and high is ordinal. So is a Likert-style question that asks whether a respondent strongly agrees, agrees, is neutral, disagrees, or strongly disagrees.

Interval scales have equal distances between values but no true zero point. Temperature in Celsius is the textbook example; zero degrees does not mean an absence of temperature.

Ratio scales have both equal intervals and a meaningful zero. Age, income, number of children, and weight are ratio variables. Most demographic and health indicators sit at this level, which is why they permit the widest range of statistical operations.

Each higher level inherits the properties of the levels below it, and the appropriate statistical tools depend on the level at which a variable has been measured. Treating an ordinal variable as if it were a ratio variable is one of the most common errors in student research.

Why accuracy in measurement matters 🔗

Stevens emphasised that measurement only earns its name when the rules used are consistent and defensible. Two researchers studying the same phenomenon should arrive at comparable readings if their instruments and procedures are sound. This is what gives social science its claim to objectivity, however imperfect that claim sometimes is.

Accuracy has two technical companions: validity, which asks whether a tool measures what it is supposed to measure, and reliability, which asks whether the tool produces consistent results across repeated use. A questionnaire that captures financial anxiety but is labelled as a depression scale is reliable but not valid. A poorly worded question that respondents interpret differently on different days is neither.

The role of measurement in social sciences 🔗

Without measurement, social science would be limited to description and speculation. Measurement performs several essential functions in research.

Classification of phenomena 🔗

Measurement allows researchers to sort observations into mutually exclusive, non-overlapping categories. A respondent cannot be both literate and illiterate at the same time, and a household either has a toilet facility or it does not. Clean classification is a prerequisite for any meaningful comparison. The National Family Health Survey, for instance, uses precise definitions to classify households by drinking water source, sanitation type, and cooking fuel, so that estimates across districts and rounds remain comparable over time.

Enabling comparison and trend analysis 🔗

Once phenomena are measured consistently, researchers can compare groups, regions, and time periods. The fifth round of the National Family Health Survey, conducted in 2019 to 2021, allows policymakers to compare state-level indicators on fertility, maternal health, and nutrition with figures from earlier rounds. According to the official fact sheet, the survey expanded its biomarker testing to include waist and hip circumference measurements and broadened the age range for blood pressure and blood glucose readings. These additions reflect changing health priorities, including the rising burden of non-communicable diseases.

Theory building and testing 🔗

Measurement turns vague hypotheses into testable claims. The idea that women with more years of schooling tend to have fewer children is interesting, but it becomes a scientific proposition only when both education and fertility are measured precisely enough to test the relationship statistically. Whole bodies of demographic theory, including the demographic transition and the diffusion of contraceptive behaviour, have been built and refined through repeated measurement across countries and decades.

Translating abstract concepts into indicators 🔗

Many of the most important ideas in social science, such as empowerment, social cohesion, or wellbeing, cannot be observed directly. Researchers approach them through indicators that can be measured. Women’s empowerment, for example, is often captured through indicators like participation in household decisions, freedom of movement, and ownership of assets. The NFHS includes questions on each of these areas, which together build a measurable picture of an otherwise abstract concept.

Measurement in practice: examples from research 🔗

Theoretical definitions become much clearer once we see how measurement is actually used in the field.

Measuring attitudes and beliefs 🔗

Attitudes are mental positions that cannot be observed directly, so researchers measure them through structured questions. A respondent might be asked whether they agree, disagree, or feel neutral about a statement such as “a working mother can establish just as warm a relationship with her children as a mother who does not work.” The pattern of responses across a population reveals prevailing beliefs about gender roles. NFHS-5 included a battery of items on attitudes towards gender roles, HIV/AIDS, and lifestyle, providing data that researchers and government agencies use to design awareness campaigns and policy interventions.

Measuring behaviours 🔗

Behaviour is, in principle, easier to measure than attitude because it leaves traces. Researchers can ask whether a woman gave birth in a health facility, whether a child was fully immunised by the age of one, or whether a household member smokes tobacco. Self-reported behaviour, however, is not always accurate; respondents may underreport stigmatised behaviours and overreport socially desirable ones. To address this, surveys often combine self-reports with biomarker data. The clinical, anthropometric, and biochemical component of NFHS-5 measures haemoglobin, blood pressure, and blood glucose directly, reducing reliance on what respondents remember or are willing to share.

Polling and election prediction 🔗

Election polling is one of the most visible applications of social measurement. Polling agencies measure voting intentions, candidate preferences, and the salience of issues across carefully selected samples. Done well, polling can capture the views of citizens who would otherwise be invisible in policy debates. The Pew Research Center notes that public opinion polling helps elected leaders understand how non-voters feel about issues and provides a counterweight to powerful organised interests. Polling failures, however, also illustrate how vulnerable measurement is to sampling errors, biased question wording, and shifting respondent behaviour.

Measuring public opinion for policy 🔗

Beyond elections, public opinion measurement informs policy formation across health, education, and welfare. Surveys on attitudes towards childhood vaccination, family planning methods, or menstrual hygiene practices guide the design of communication strategies and service delivery models. NFHS-5 added new measurement areas on bathing practices during menstruation and methods and reasons for abortion, recognising that these previously underexplored areas affect health outcomes and require targeted interventions.

The challenges that remain 🔗

Measurement in the social sciences is powerful, but it has limits that researchers must take seriously. Much of what is measured in social research sits at the nominal or ordinal level, which restricts the statistical operations that can legitimately be performed. Cultural and linguistic differences complicate the translation of questionnaires across regions, which is a real challenge in a multilingual country. Self-reporting biases, recall errors, and the social desirability of certain answers can systematically distort findings. And even Stevens’ framework has been critiqued by contemporary methodologists who argue that the four-scale typology is too rigid for the diverse data social scientists actually collect.

These limitations are not reasons to abandon measurement. They are reasons to do it more carefully, with clearer definitions, better-tested instruments, and honest acknowledgement of what numbers can and cannot tell us.

What do you think? If you were asked to design a measurement tool for “family wellbeing” in your own community, what indicators would you choose, and which abstract aspects do you think would be hardest to capture in numbers?

References

1. https://methods.sagepub.com/reference/encyc-of-research-design/n292.xml
2. https://en.wikipedia.org/wiki/Level_of_measurement
3. https://courses.lumenlearning.com/suny-hccc-research-methods/chapter/chapter-6-measurement-of-constructs/
4. https://www.dataforindia.com/nfhs-explainer/
5. https://www.dhsprogram.com/pubs/pdf/OF43/India_National_Fact_Sheet.pdf
6. https://www.nfhsiips.in/nfhsuser/nfhs5.php
7. https://microdata.worldbank.org/index.php/catalog/4482
8. https://www.pewresearch.org/fact-sheet/topic-why-public-opinion-matters-and-how-to-measure-it/
9. https://pmc.ncbi.nlm.nih.gov/articles/PMC10657051/
10. https://www.researchgate.net/publication/342098870_Mathematization_Not_Measurement_A_Critique_of_Stevens'_Scales_of_Measurement