Big picture
Before we can analyse an economic problem, we need data. Data can be collected first-hand for the purpose (primary data) or taken from sources where someone else has already collected them (secondary data).
Collecting data from every member of a population is often impossible, so we usually study a sample. How the sample is chosen, and how questions are asked, decide whether the results can be trusted. This chapter covers the tools and methods of collecting data and the errors to avoid.
Learning goals
Distinguish primary and secondary data and their sources.
Prepare a good questionnaire and explain the qualities of good questions.
Compare personal interviews, mail surveys and telephone interviews, and explain the pilot survey.
Distinguish a census from a sample survey.
Explain random and non-random sampling methods.
Explain sampling and non-sampling errors.
Name the important sources of secondary data in India.
Core ideas
1. Primary and secondary data
Basis | Primary data | Secondary data |
|---|---|---|
Meaning | Collected first-hand by the investigator for the purpose | Already collected by someone else, used again |
Originality | Original | Not original |
Cost and time | More costly and time-consuming | Cheaper and quicker |
Suitability | Closely suited to the purpose | May need adjustment to suit the purpose |
Example | A survey of students' study hours by a researcher | Population figures from the Census |
Precautions when using secondary data: check the reliability of the source, its suitability for the purpose, the adequacy of coverage, the definitions and units used, and the time period.
2. The questionnaire
A questionnaire is a list of questions about the subject of the survey, answered by the respondent. When questions are asked and answers recorded by an enumerator, it is called a schedule.
Qualities of a good questionnaire
Not too long: a manageable number of questions.
Logical order: questions move from general to specific.
Simple, clear language: easy to understand, without technical terms.
Not ambiguous: each question has one clear meaning.
No leading questions: questions shouldn't suggest an answer. "Don't you agree that prices are too high?" is leading; "What is your opinion about current prices?" is not.
No double questions: "Do you like tea and coffee?" mixes two questions.
No sensitive or personal questions that might offend, unless necessary.
Precise questions: avoid vague words such as "often" or "many".
Types of questions
Type | Meaning | Example |
|---|---|---|
Two-way (closed) question | Two possible answers | "Do you own a two-wheeler? Yes / No" |
Multiple choice question | Several answer options | "How do you travel to school? Walk / Bus / Cycle / Car" |
Open-ended question | Respondent answers freely | "What do you think should be done to reduce traffic?" |
Closed questions are easy to answer and analyse; open-ended questions give richer views but are harder to compare.
3. Modes of data collection
Mode | Merits | Demerits |
|---|---|---|
Personal interview (face to face) | Highest response rate; questions can be explained; non-verbal cues can be observed; complex questions possible | Expensive; time-consuming; interviewer may influence answers; needs trained interviewers |
Mail survey (questionnaire sent by post or email) | Least expensive; covers a wide area; no interviewer influence; respondents answer at their own pace | Low response rate; can't clarify questions; only literate respondents; long delays |
Telephone interview | Cheaper than personal interviews; quick; can clarify questions | Limited to people with phones; no observation of reactions; respondents may not cooperate |
Pilot survey (pre-testing): before the main survey, the questionnaire is tried on a small group. It helps find problems with questions, estimate the time and cost, and train the investigators.
4. Census and sample survey
Population (universe): all the items or individuals in the field of study.
Census (complete enumeration): data are collected from every member of the population. Example: the Census of India, held every ten years.
Sample survey: data are collected from a sample, a group selected from the population to represent it.
Basis | Census | Sample survey |
|---|---|---|
Coverage | Every unit | A selected part |
Cost and time | High | Low |
Accuracy | Very accurate if done well; no sampling error | Reasonably accurate; sampling error exists |
Suitability | Small or very important populations | Large populations, limited time and budget |
Detail | Fewer details per unit (because of scale) | More detailed information per unit possible |
5. Sampling methods
Random sampling (probability sampling): every individual has an equal chance of being selected, so the selection is free from the investigator's bias.
Lottery method: write each unit's name or number on slips, mix them, and draw the required number.
Random number tables: units are numbered and a sample is selected using published tables of random numbers (or a computer).
Example. To choose 10 students from 200 for a survey, number them 1 to 200 and pick 10 numbers by lottery.
Non-random sampling: not every individual has an equal chance; selection depends on the investigator's judgement or convenience.
Method | Meaning |
|---|---|
Judgement (purposive) sampling | The investigator chooses units they think are representative |
Convenience sampling | Units easiest to reach are chosen |
Quota sampling | The population is divided into groups, and a fixed number (quota) is chosen from each group by judgement |
Random sampling is preferred because it avoids bias and allows sampling error to be estimated.
6. Sampling and non-sampling errors
Sampling error is the difference between the value from a sample and the true value of the population. It arises because only a part of the population is studied. Example: the true average income of a town is ₹40,000, but a sample gives ₹38,500; the sampling error is ₹1,500. It can be reduced by taking a larger sample.
Non-sampling errors arise in collecting and processing data, and can occur in both a census and a sample survey:
Error | Cause |
|---|---|
Errors in data acquisition | Wrong recording, misunderstanding of questions, faulty measurement, deliberate wrong answers by respondents |
Non-response errors | Respondents can't be contacted or refuse to answer |
Sampling bias | Some members of the population are systematically left out of the sample |
Errors of processing | Mistakes in coding, tabulation or calculation |
Non-sampling errors are often more serious than sampling errors, and they don't necessarily fall with a larger sample.
7. Sources of secondary data in India
Source | What it provides |
|---|---|
Census of India (Office of the Registrar General and Census Commissioner) | Population, its size, density, sex ratio, literacy, migration, household amenities; held every ten years |
National Statistical Office (NSO) under the Ministry of Statistics and Programme Implementation (MoSPI) | Formed in 2019 by merging the Central Statistics Office and the National Sample Survey Office. It conducts national sample surveys on consumption, employment, health, education and more, and compiles national income estimates and indices such as the Index of Industrial Production |
Reserve Bank of India | Data on money, banking, credit and foreign exchange |
Labour Bureau | Data on wages, employment and consumer price indices for workers |
Ministries and departments | Data on agriculture, trade, health and other sectors |
International bodies | The World Bank, IMF and UN publish international data |
A reliable approach to questions
Decide whether data are collected first-hand (primary) or taken from published sources (secondary).
Check a question for leading words, double questions, ambiguity and sensitive content.
Recommend a census for a small or very important population, and a sample for a large population with limited time and money.
To decide if a sample is random, ask whether each unit had an equal chance of selection.
Classify errors: due to studying only a part (sampling error) or due to mistakes in collection and processing (non-sampling error).
Connections that matter
Introduction to statistics explained primary and secondary data and variables.
Organisation and presentation of data takes the collected data and arranges it.
Class 12 economics: national income relies on NSO data.
Common mistakes
Saying a census has sampling error. A census covers every unit, so it has no sampling error, though it can have non-sampling errors.
Calling convenience sampling random.
Writing leading or double questions when asked to frame a questionnaire.
Saying a larger sample removes non-sampling errors. It reduces sampling error only.
Revision checklist
I can distinguish primary and secondary data and list precautions for secondary data.
I can frame a good questionnaire and identify faulty questions.
I can compare personal, mail and telephone surveys and explain a pilot survey.
I can distinguish a census and a sample survey.
I can explain random and non-random sampling and the types of errors.
I can name India's main sources of secondary data.