Data collection methods in research: types and examples

Data collection is the planned way you gather the information that answers your research question. The plan matters as much as the questions. Ask the same women the same thing aloud in Hindi, or through a link in English, and you will get different answers.

Every method falls into one of two groups, depending on who collected the data first.

Shalini's study would need both: her own data on how the women used their loans, and published figures to show where her villages stood against the state and the country.

Different guides list the methods in different ways, but the same five keep coming back: surveys, interviews, focus groups, observation and experiments. Records and documents make a sixth, usually secondary data.

Quantitative data collection methods give you numbers you can count and compare: how many, how much, how often. Qualitative data collection methods give you words, reasons and meanings: why, how, what it felt like. Many studies use both. That is called mixed methods, and our guide on how to write a research methodology shows how to justify it in one sentence.

Shalini's question had two halves. What the women spent their loans on needed a count. Why they chose it needed their own words. So she needed a method from each side.

A survey asks many people the same questions, so their answers can be added up. In India, research methods textbooks such as C. R. Kothari's Research Methodology: Methods and Techniques separate two kinds of survey form, and the difference decided Shalini's whole plan.

An online form is a questionnaire. It works well for college students or teachers, who can read it and reply on their phones. It works badly where many people cannot read the form or have no phone of their own, because those who reply stop being a fair picture of the group. Our guide on sampling methods in research explains why self-selected replies lean one way.

Shalini turned her form into a schedule. She translated it into simple Hindi, cut it from 40 questions to 22, mostly with fixed options, and planned to visit each woman at home and read every question aloud.

A survey tells you what. An interview tells you why. Interviews come in three kinds, depending on how fixed the questions are.

• Structured: the same questions, in the same order, for everyone.
• Semi-structured: a short guide of main questions, with follow-ups that change from person to person. The most common kind in student research.
• Unstructured: a conversation around a theme, with almost no set questions.

A focus group brings a small group together to discuss a topic, with the researcher as moderator. In their practical guide to qualitative research, Darshini Ayton, Tess Tsindos and Danielle Berkovic suggest 6 to 10 participants. With too many, they note, the discussion can become chaotic; with too few, there may not be enough interaction. Focus groups suit how a group thinks and decides, not private matters: few people discuss debts or illness in front of their neighbours.

Shalini planned semi-structured interviews with women who had spent their loans on different things, and one focus group per group, after the monthly meeting, on how they decided who borrowed first.

People do not always do what they say they do. Observation means watching and recording behaviour as it happens. In participant observation the researcher joins in the life of the group; in non-participant observation the researcher watches from the side. It can be structured, with a checklist, or unstructured, with detailed field notes.

Shalini asked two groups if she could sit in on their monthly meetings. She sat at the back with a one-page checklist: who spoke, who asked for a loan, how the amount was agreed, and whether anyone was refused. In the first meeting she saw what no survey would have shown: two women asked for loans, and the group leader spoke for both.

An experiment is different. The researcher changes one thing on purpose, keeps everything else the same, and measures what happens, usually against a comparison group that did not get the change. A soil science student testing two fertiliser doses on matched plots is running one. Experiments are strong at showing cause and effect, but need control over the setting. Shalini could not decide who received a loan, so an experiment did not fit, and she said so in her methodology.

For Indian students, some large sources of secondary data are open to all.

Secondary data saves months, but it was built for someone else's question. Before you use it, find out who collected it, when, how and from whom, and read the definitions. A survey's "household" may not mean what you mean.

Shalini used two kinds. The NFHS fact sheets give the share of women with a bank or savings account they themselves use, so she could compare her villages with the state and the country. And with each group's permission, she copied a year of entries from its loan books, with names replaced by codes. The books gave exact amounts and dates that no one could be expected to remember.

There is no best method, only the one that fits. If the question itself is still loose, sharpen it first with our guide on how to write a good research question. Then work through these, in order.

1. What does your question ask?How many, how much or how often points to a survey, an experiment or records. Why, how or what it is like points to interviews, focus groups or observation.
2. Does the data already exist?Look for a reliable secondary source before collecting your own. If it answers part of your question, use it and collect only the rest.
3. Who are your participants?Think about language, reading, phones and working hours. A method that suits college students may not suit farmers.
4. What can you manage?Count the days, the travel and the cost. A small study done properly beats a large one done in a hurry.
5. Is it ethical?Some methods need consent forms, recordings or entry into homes. Check what your ethics committee expects.

Using more than one method on the same question is called triangulation. When the methods agree, you can trust the finding more. When they disagree, you have found something worth explaining.

Shalini's final plan: a schedule for every woman in the sampled groups, interviews with a few, a focus group and observed meetings in each group, the loan books, and NFHS figures for context.

Never take a new questionnaire, schedule or interview guide straight to the field. Test it first on a few people like your participants. A research methods chapter on INFLIBNET's e-books platform calls a pilot study "a small-scale implementation of a larger study or a part of a larger study", and describes pre-testing as a way to find questions that respondents cannot understand.

Shalini tried her schedule with eight women in a village outside her sample. One question failed. She had asked for the "rate of interest per annum", and every woman answered it differently. She rewrote it the way the groups talked: "For every 100 rupees you borrow, how much extra do you pay back each month?" The answers became easy to compare.

When you write your methodology, report your data collection so that someone else could repeat it:

• each method, and why it suits your question
• the tools: the questionnaire, schedule, interview guide or checklist, with the full version in an appendix
• the language, and who translated the tools
• where, when and by whom it was collected
• the pilot test and what you changed after it
• how consent was taken, and how the data was recorded and stored
• for secondary data: the source, the year and any changes you made

Dr. Minz read the chapter and asked one thing. "And the Google Form?" Shalini smiled. She had kept it for a short follow-up with the group leaders, the one group who really could reply on their phones.