Understanding Scales and Their Techniques in Social Science Research

Researchers in population and family health often need to measure things that don’t fit into a measuring tape or a weighing scale. How do you quantify a mother’s satisfaction with an Anganwadi centre? Or a community’s trust in the public health system? This is where scales step in. Scales are systematic tools that convert abstract feelings, opinions, and behaviours into numbers that can be analysed, compared, and used to shape policy.

What are scales in social science research? 🔗

A scale is a measurement instrument used to capture latent or non-observable variables such as attitudes, opinions, perceptions, motivation, and behavioural intentions. Unlike a thermometer that directly reads temperature, scales rely on a set of carefully designed items, usually statements or questions, whose responses are combined to produce a meaningful score.

The need for scaling arose because early researchers realised that simply assigning arbitrary numbers to opinion questions did not guarantee accurate measurement. As one academic resource notes, early opinion measurement involved questionnaires whose numerical values were assigned arbitrarily, with nothing to guarantee that one item tapped the same attitude as another. From the 1930s onwards, sociologists and psychologists developed hundreds of standardised scales to fix this problem.

In population and family health studies, scales help researchers measure things like contraceptive acceptance, attitudes towards immunisation, mental health status of adolescents, stigma around leprosy or HIV, and satisfaction with public services. The field of psychometrics has expanded to cover knowledge, skills, aptitudes, beliefs, motivation, and health-related issues, all of which depend on well-constructed scales.

Why do we need scales? 🔗

Direct questions like “Do you trust the health system?” give limited information. A scale, on the other hand, breaks the construct into multiple items, each tapping a different aspect. The combined score gives a more reliable and nuanced picture. Scales also allow comparisons across regions, demographic groups, or time periods, which is essential when evaluating programmes like the National Health Mission or Beti Bachao Beti Padhao.

Levels of measurement behind every scale 🔗

Before choosing a scaling technique, researchers must understand the four levels of measurement, because they decide what statistical analyses are possible later.

Nominal scales only classify items into categories with no order, such as religion, marital status, or blood group. Ordinal scales rank items but do not specify the distance between ranks, for example, the order of preference for three family planning methods. Interval scales have equal intervals between values but no true zero, like temperature in Celsius. Ratio scales have equal intervals and a true zero, allowing all mathematical operations. As explained in a recent overview of research scaling, ratio scales such as height, weight, age, and income are the most informative because zero genuinely means “none”, making ratios meaningful.

Most attitude scales operate at the ordinal or interval level. Recognising which level you are working with prevents the common mistake of running advanced statistics on data that cannot support them.

Comparative scaling techniques 🔗

Scaling techniques are broadly divided into two families: comparative and non-comparative. Comparative scaling involves a direct comparison between two or more objects, and the responses are interpreted only in relation to each other. The data generated are mostly ordinal in nature.

Paired comparison scaling 🔗

Here, respondents are presented with two objects at a time and asked to choose one based on a stated criterion. If a researcher wants to know which immunisation awareness poster is more effective, the respondent is shown two posters and asked to pick the better one. With more options, every possible pair is shown. This method is simple and forces a clear choice, but the number of pairs grows quickly as items increase, which can tire respondents.

Rank order scaling 🔗

In this technique, respondents are given a list of objects and asked to rank them from most to least preferred. For example, asking ASHA workers to rank five issues, malnutrition, anaemia, neonatal mortality, adolescent pregnancy, and substance abuse, by the urgency of intervention in their village. Rank order scaling is faster than paired comparison and mirrors real-life decision-making, but it does not show how much one item is preferred over another.

Constant sum scaling 🔗

Respondents distribute a fixed number of points, usually 100, among several items to reflect the relative importance of each. A community member evaluating a Panchayat’s spending might be asked to allocate 100 points across sanitation, primary healthcare, education, and roads. The result captures relative weights well, but respondents may struggle if there are too many items to balance.

Q-sort and Bogardus social distance scales 🔗

The Q-sort method is used when the number of items is large; respondents sort statements into piles ranging from most to least agreed with. The Bogardus social distance scale measures the closeness or distance people are willing to maintain with members of different social groups, useful for studying caste, religious, or migrant-related attitudes in Indian society.

Non-comparative scaling techniques 🔗

Non-comparative scales evaluate each item independently, without reference to other items. They generally produce interval-level data and are by far the most widely used in social and health research. As a reference summary of scaling techniques explains, noncomparative scales include continuous rating, Likert scales, the semantic differential, Stapel scaling, Thurstone scales, and multidimensional scaling.

Likert scale 🔗

The Likert scale, developed by Rensis Likert in 1932, is the most popular attitude measurement tool. Respondents indicate their level of agreement with each statement, usually on a five-point or seven-point continuum ranging from “strongly disagree” to “strongly agree”. A study assessing perceptions of the Janani Suraksha Yojana could include statements like “The cash incentive motivated me to deliver in a hospital”, with responses summed across items to produce a total attitude score.

The Likert technique assumes equal weights for all items and is therefore called a summated scale. As noted in an open social science textbook, items with reversed meaning from the original direction of the construct must be reverse coded before summating, otherwise the total score becomes meaningless.

Semantic differential scale 🔗

Developed by Charles Osgood, this scale asks respondents to rate an object on a series of seven-point bipolar adjectives such as “good-bad”, “useful-useless”, or “trustworthy-untrustworthy”. It is widely used to assess image and perception, for instance, how citizens perceive a particular government hospital or a public health campaign.

Stapel scale 🔗

A variation of the semantic differential, the Stapel scale uses a single adjective with a numerical range from +5 to -5 (without zero). It is helpful when designing bipolar opposites is difficult, but it can confuse respondents who are not familiar with negative numbers.

Thurstone scale 🔗

Thurstone’s method, called the method of equal-appearing intervals, was the first systematic attempt to give attitude measurement scientific rigour. A panel of judges rates a large pool of statements on an 11-point favourability scale, and statements with high agreement are retained. Respondents then indicate which statements they agree with, and the average scale value of endorsed statements becomes their attitude score. According to a teaching summary from San José State University, Thurstone scales include statements with numerical values indicating favorable or unfavorable attitudes, and the average of agreed-with statements determines the individual’s attitude. It is rigorous but labour-intensive, which is why it is less common today.

Guttman scale 🔗

The Guttman scale, also called a cumulative scale, arranges items in an increasing order of intensity, so agreement with a stronger item implies agreement with all weaker ones. For example, a scale of women’s empowerment might begin with “I can decide what to cook today” and progress to “I can decide to start my own business without permission”. Agreement with the last item assumes agreement with all earlier ones. This cumulative property makes it useful for measuring the strength of attitudes, although designing a truly unidimensional Guttman scale is challenging.

Applications in population and family health research 🔗

Scales are deeply embedded in evaluations of social and health programmes. Researchers studying gender attitudes in India have used Likert-style statements to capture views such as “A woman’s most important role is being a good homemaker”. A report from the Abdul Latif Jameel Poverty Action Lab notes that direct survey questions on social attitudes can be affected by social desirability bias, which the Marlowe-Crowne scale of social desirability helps detect.

In disease-specific research, scales have been built for very specific contexts. A team in Tamil Nadu developed and validated a culturally appropriate scale to measure attitudes of healthcare providers towards persons affected by leprosy, guided by the Affective, Behavioural and Cognitive model. Similarly, a systematic review found that several internationally used instruments such as the General Health Questionnaire, Patient Health Questionnaire, Edinburgh Postnatal Depression Scale, and the Perceived Stress Scale have been translated and adapted into multiple Indian languages for clinical and community research.

These adaptations matter because a scale developed in one cultural setting cannot simply be translated word-for-word. It must be linguistically validated, culturally adjusted, and statistically tested for reliability and validity before being applied to a new population.

Choosing the right scale 🔗

The choice of scaling technique depends on the construct being studied, the level of measurement needed, the literacy and cognitive load of respondents, and the planned statistical analysis. Likert scales work well for most attitude studies. Semantic differentials suit image and perception research. Paired comparisons and rank ordering work when comparing a small set of well-defined options, such as different methods of contraception or alternative welfare schemes.

Whichever scale is used, three qualities are non-negotiable: validity (does the scale measure what it claims to measure?), reliability (does it produce consistent results across time and respondents?), and cultural appropriateness for the population being studied. A scale that ignores any of these risks producing data that looks scientific on paper but fails to reflect reality.

What do you think? If you were designing a scale to measure citizens’ satisfaction with a flagship health programme like Ayushman Bharat, which technique would you choose and why? And how would you handle the risk that respondents might give socially desirable answers instead of their honest opinions?

References

1. https://ebooks.inflibnet.ac.in/socp3/chapter/scaling-and-measurement/
2. https://pmc.ncbi.nlm.nih.gov/articles/PMC3146211/
3. https://scienceinsights.org/what-is-scaling-in-research-types-and-techniques/
4. https://www.sciencedirect.com/topics/social-sciences/scaling-technique
5. https://socialsci.libretexts.org/Courses/Orange_Coast_College/SOC_200:_Introduction_to_Sociology_Research_Methods_(Ridnor)/05:_Indexes_and_Scales/5.03:_Scale_Construction
6. https://www.sjsu.edu/people/fred.prochaska/courses/ScWk242/s0/Thurston-Guttman-and-Likert-scales.pdf
7. https://www.questionpro.com/blog/thurstone-vs-guttman-scale/
8. https://www.povertyactionlab.org/blog/3-13-20/measuring-social-attitudes-and-norms-impact-evaluations-examples-india
9. https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6177202/
10. https://pubmed.ncbi.nlm.nih.gov/39502591/