Key takeaways
- Primary data are first-hand; secondary data are reused from other sources.
- A pilot survey tests a questionnaire before the main enquiry.
- A census covers every unit; a sample survey covers a part and carries sampling error.
- Random sampling lets the error be estimated; non-random sampling does not.
- Non-sampling errors such as bias and non-response can occur even in a census.
This is Chapter 2 of the Statistics for Economics chapter list for Class 11 Economics. It follows Chapter 1 and leads into organising what is collected.
2.1Sources of Data: Primary and Secondary
Primary data are collected for the first time by the investigator for a specific purpose, so they are original and fresh. Secondary data were collected earlier by someone else and are reused. Primary data cost more time and money, while secondary data are cheaper but may not fit the present purpose.
A researcher studying household water use in her ward can walk door to door and ask; the answers are primary data. If she instead uses the Census or a survey report from an earlier year, the figures are secondary data. A set of primary data in one person’s hands becomes secondary for the next person who uses it.
| Basis | Primary data | Secondary data |
|---|---|---|
| Meaning | Collected first-hand for the present purpose | Already collected for another purpose |
| Cost and time | High | Low |
| Fit with the problem | Exact | May need adjustment |
| Reliability check | Investigator controls quality | Depends on the original source |
| Examples | Own survey of 60 households | Census tables, Economic Survey, NSS reports |
Before using secondary data, check who collected them, when, how, and whether the definitions match yours.
Choosing between primary and secondary data
The choice depends on the question, the time and the budget. A researcher asking how many students in her school use public transport must collect primary data, since no existing source has it. A researcher studying the literacy rate of districts can use Census tables, which are secondary data covering the whole country.
Often both are combined: secondary data give the background and primary data fill gaps. Whichever is used, the researcher should record the source, the date, the definitions and the method, so that readers can judge the reliability of the findings.
2.2How Primary Data Are Collected
Primary data are usually collected with a questionnaire, a set of ordered questions answered by respondents, through personal interviews, postal or online forms or telephone calls. A good questionnaire has short, clear, neutral questions, and is tested on a small group in a pilot survey before the main enquiry begins.
Designing a questionnaire
- Keep questions short, simple and in a logical order; ask easy ones first.
- Avoid leading words. “Do you agree that the new bus service is excellent?” pushes the respondent; “How would you rate the new bus service?” does not.
- Ask one thing at a time; avoid double questions.
- Give answer options that do not overlap and cover all cases.
Modes of collection
| Mode | Strength | Weakness |
|---|---|---|
| Personal interview | High response; can explain questions | Costly and slow; interviewer may influence answers |
| Postal or online form | Cheap; reaches many | Low response; cannot clarify questions |
| Telephone call | Quick and low cost | Only reaches those with access; short answers |
The pilot survey
A pilot survey tries the questionnaire on a few respondents. It exposes confusing wording, missing options and the time needed, so the faults are fixed before the main survey is run.
A pilot survey in action
Suppose a student plans a questionnaire on household water use and tests it with five neighbours. Two of them are unsure whether “use” includes water for gardens, a third cannot estimate litres but can count buckets, and the form takes twenty minutes, which is too long. The student now defines the term, switches to buckets with a conversion, and cuts questions.
Such changes save time and improve accuracy in the main survey. The pilot also shows how many households are likely to be unavailable, so that the sample can be planned with extra units. Without it, errors would be discovered only after all data were collected, when they could not be corrected.
2.3Census and Sample Surveys
A census studies every unit of the population, while a sample survey studies only a selected part. A census gives complete and exact information but costs more and takes longer. A sample, if chosen so that it represents the population, gives reliable results quickly and cheaply, and is the only option when testing destroys the unit.
The population in statistics is the whole group under study, such as all farm households in a block. A sample is a part of it. A representative sample mirrors the population in the features that matter.
| Basis | Census | Sample survey |
|---|---|---|
| Coverage | Every unit | A selected part |
| Cost and time | High | Lower |
| Accuracy of coverage | Complete | Subject to sampling error |
| When suitable | Small population or need for complete detail | Large population; testing destroys the unit |
Random and non-random sampling
In random (probability) sampling, each unit has a known chance of selection, which allows the error to be measured. Common forms are simple random, systematic (every kth unit) and stratified sampling (sample taken from each sub-group). In non-random sampling selection depends on convenience or judgement; it is quick but cannot show how far off the result may be. Exit polls rely on samples of voters.
How large should a sample be?
There is no single answer. A larger sample reduces sampling error, but the gain becomes smaller as the sample grows, while cost and time keep rising. The sample size depends on how varied the population is, how precise the answer must be, and the resources available. A population in which all units are alike needs only a small sample, while a varied one needs more.
Stratified sampling can improve accuracy without a large increase in size, since it ensures that each important group appears in the sample. The method of selection matters more than size: a large biased sample can be less reliable than a small random one.
Cost and detail compared
A census can give detailed figures for small areas, such as a single village or ward, which a sample cannot. A sample survey gives results quickly for the country or a state at lower cost. This is why India uses both: the Census every ten years for a full count and detailed local data, and sample surveys for regular updates on subjects such as consumption, jobs and health between censuses.
2.4Sampling and Non-Sampling Errors
Sampling error is the gap between a sample estimate and the true population value that arises because only part of the population is studied; it falls as the sample grows. Non-sampling errors come from faulty design, non-response and recording mistakes, and can occur even in a census.
Take eight farm households with monthly incomes (in ₹ thousand) of 12, 14, 15, 18, 20, 22, 25 and 34. Their mean is 20. Five different samples of three households give means of 16.33, 19.67, 24.67, 17.67 and 20.33. Each differs from 20; the difference is the sampling error.
Non-sampling errors
- Sampling bias: a method that favours some units, such as sampling only households near the main road.
- Non-response: selected people cannot be found or refuse to answer.
- Errors in data acquisition: wrong recording, careless answers, or an unclear question.
- These errors are not cured by taking a larger sample; better design and supervision are needed.
Reducing non-sampling errors in practice
Several practical steps help. Test the questionnaire on a few respondents before the main survey. Train the interviewers and explain terms in the local language. Make repeated visits to households that are not found at first, to reduce non-response. Check completed forms for impossible values, such as a child recorded as a worker, and verify a sample of interviews.
Surveys also protect respondents’ privacy, because people answer honestly only when they trust that their answers will not be misused. Official agencies, such as the Census and the National Statistics Office, follow these rules, and their reports describe how errors were checked.
2.5Census of India and the National Sample Survey
The Census of India counts the whole population about every ten years and has run since 1881. The National Sample Survey, begun in 1950 and now run by the National Statistics Office, collects sample data on consumption, employment and other topics. Together they are major sources of secondary data.
The Census records population size, age, sex, literacy, occupation and housing. It is carried out by the Office of the Registrar General and Census Commissioner, under the Census Act, 1948. The NCERT book gives the 2011 count as 121.09 crore.
The National Sample Survey (NSS) was started in 1950. Its rounds cover household consumption expenditure, employment, health, education and enterprises, and are published by the National Statistics Office (NSO) under MoSPI. Because they are sample-based, their results come with estimated errors.
| Feature | Census | NSS |
|---|---|---|
| Method | Complete count | Sample survey |
| Frequency | About every ten years | Regular rounds on different subjects |
| Conducted by | Registrar General and Census Commissioner | National Statistics Office, MoSPI |
| Typical use | Population size and structure | Consumption, employment, poverty estimates |
Worked Examples
Example 1: Sampling error
- Given
- Population (8 households) mean income = ₹20 thousand. Sample C has incomes 18, 22 and 34.
- Formula
- Sampling error = sample estimate − population value
- Steps
- Sample mean = (18 + 22 + 34) ÷ 3 = 24.67. Error = 24.67 − 20 = +4.67.
- Answer
- The sample overstates the population mean by ₹4.67 thousand.
Example 2: Proportional stratified sample
- Given
- A school has 500 students in Class 9, 300 in Class 10 and 200 in Class 11. A sample of 100 is to be drawn using stratified sampling in proportion to size.
- Formula
- Stratum sample = (stratum size ÷ population) × total sample
- Steps
- Class 9: 500 ÷ 1,000 × 100 = 50. Class 10: 300 ÷ 1,000 × 100 = 30. Class 11: 200 ÷ 1,000 × 100 = 20.
- Answer
- Draw 50, 30 and 20 students from the three classes.
Example 3: Systematic sampling
- Given
- A list has 240 shops; a sample of 20 is needed. The first shop is chosen at random as number 5.
- Formula
- Interval k = N ÷ n; the sample is 5, 5 + k, 5 + 2k, …
- Steps
- k = 240 ÷ 20 = 12. The shops are 5, 17, 29, …; the 20th is 5 + 19 × 12 = 233.
- Answer
- Select every 12th shop starting from number 5, ending at 233.
Common Mistakes
- Believing a census has no errors. It can suffer from non-response and recording errors.
- Believing a bigger sample cures all error. It reduces sampling error, not bias or wording faults.
- Treating secondary data as always weaker. They are the right choice when time, money or access is short.
- Writing leading questions and then reporting the answers as neutral findings.
10 Important Questions for Class 11 Board and School Examinations
Questions are written in original words. Marks are shown beside each question.
Q1Data collected for the first time by an investigator are called (a) secondary (b) primary (c) published (d) derived 1 mark · MCQ
The correct option is (b) primary data.
Q2Which error can occur even in a census? (a) sampling error (b) non-sampling error (c) both (d) neither 1 mark · MCQ
The correct option is (b) non-sampling error.
A census covers every unit, so there is no sampling error, but recording mistakes and non-response can still occur.
Q3Assertion (A): A pilot survey is run before the main survey. Reason (R): It helps find faults in the questionnaire. Which is correct? 1 mark · Assertion-reason
Both A and R are true, and R explains A.
Q4Distinguish between primary and secondary data. 3 marks
Primary data are collected first-hand for the present purpose, while secondary data were collected earlier by others.
- Primary data cost more time and money but fit the problem exactly.
- Secondary data are cheaper and quicker but may not match the purpose.
- Examples: a household survey you conduct is primary; Census tables or the Economic Survey are secondary.
- A set of primary data becomes secondary for the next user.
Q5What is a representative sample? Why is it important? 3 marks
A representative sample reflects the features of the population from which it is drawn.
It is important because conclusions about the whole population rest on it. If the sample is biased, for example if only households near the main road are asked, the results will be wrong however large the sample is. Random selection, in which each unit has a known chance of being chosen, and stratified sampling, which includes every group, help to make a sample representative.
Q6Compare a census and a sample survey on any four points. 4 marks
A census studies every unit of the population, while a sample survey studies only a selected part.
- Coverage: complete in a census, partial in a sample.
- Cost and time: high for a census, lower for a sample.
- Error: a census has no sampling error, while a sample estimate has some.
- Use: a census suits small populations or needs for complete detail; samples suit large populations or cases where testing destroys the unit.
The Census of India is held roughly every ten years.
Q7Explain any two sources of non-sampling error. 4 marks
Non-sampling errors arise from faulty design or recording and can occur even in a census; two sources are non-response and errors in acquiring data.
- Non-response: selected households cannot be found or refuse, so their views are missing and the results are biased.
- Errors in data acquisition: respondents misunderstand a question, or enumerators record wrongly, so the data are inaccurate.
A larger sample does not cure these errors. They need a better questionnaire, trained interviewers and careful supervision, and a pilot survey can expose many faults in advance.
Q8Explain the steps and cautions in designing a questionnaire. 6 marks
A questionnaire should be short, clear, ordered and neutral, and it should be tested before use.
- State the aim of the survey and decide what information is needed.
- Frame short, simple and unbiased questions; avoid leading questions and questions that ask two things at once.
- Put easy questions first and sensitive ones later.
- Give answer options that cover all cases and do not overlap.
- Run a pilot survey on a few respondents to find confusing questions.
- Add a covering letter if the form is sent by post, explaining its purpose and promising confidentiality.
Following these steps raises both the response rate and the accuracy of the data.
A well-designed questionnaire reduces non-response and the errors that arise from misunderstood questions, so the data collected can be trusted.
Q9Describe random and non-random sampling with examples of each. 6 marks
In random sampling every unit has a known chance of selection, while in non-random sampling selection depends on convenience or judgement.
- Random: simple random sampling by lottery of roll numbers; systematic sampling of every 12th shop on a list; stratified sampling, taking students from each class in proportion.
- Non-random: convenience sampling, asking whoever is nearby; judgement sampling, choosing units the investigator thinks typical.
- Random methods let the sampling error be estimated.
- Non-random methods are quick and cheap but may be biased.
Exit polls rely on samples of voters; their accuracy depends on how the sample is chosen.
In a stratified sample the population is divided into groups, such as classes in a school, and a sample is taken from each in proportion to its size. This ensures that every group is represented and often gives more accurate estimates than simple random sampling for the same cost.
Q10A population of 8 farms has yields 12, 14, 15, 18, 20, 22, 25 and 34 quintals. A sample of three farms gives 14, 22 and 25. Find the sampling error in the mean. 4 marks · Numerical
The sampling error is +0.33 quintal.
Given: population mean = 160 ÷ 8 = 20; sample 14, 22, 25.
Formula: error = sample mean − population mean.
Steps: sample mean = 61 ÷ 3 = 20.33. Error = 20.33 − 20 = 0.33.
Final answer: the sample overstates the mean by 0.33 quintal.
10 Points for State PSC and UPSC Aspirants
- India’s first synchronous Census was held in 1881; the 2011 Census was the 15th and the seventh since Independence.Why it matters: a common factual question on the Census series.Source: Office of the Registrar General and Census Commissioner
- The Census is conducted under the Census Act, 1948 by the Registrar General and Census Commissioner, Ministry of Home Affairs.Why it matters: who runs the Census and under which law.Source: India Code; Census of India
- The National Sample Survey began in 1950; its work is now done by the National Statistics Office under MoSPI.Why it matters: NSS rounds are cited across economics papers.Source: MoSPI
- The Periodic Labour Force Survey (PLFS), started in 2017, gives employment and unemployment estimates from a sample of households.Why it matters: the standard source for labour market data.Source: MoSPI
- The Household Consumption Expenditure Survey gives spending patterns used to update consumer price weights and poverty estimates.Why it matters: links Chapter 2 to Chapter 7 on index numbers.Source: MoSPI
- The Agriculture Census is conducted every five years by the Department of Agriculture and Farmers Welfare and records landholdings.Why it matters: a census of holdings, not of people.Source: Ministry of Agriculture and Farmers Welfare
- The Sample Registration System of the Registrar General estimates birth rate, death rate and infant mortality from a large sample.Why it matters: a sample-based demographic source between Censuses.Source: Office of the Registrar General and Census Commissioner
- The National Family Health Survey collects health and nutrition data from households and is run by the International Institute for Population Sciences for the Health Ministry.Why it matters: frequently cited for child and maternal health indicators.Source: Ministry of Health and Family Welfare
- The Collection of Statistics Act, 2008 requires respondents to give correct information to official surveys and protects its confidentiality.Why it matters: the legal backing for non-response issues in official surveys.Source: India Code
- Large NSS surveys use stratified multi-stage sampling, with villages or urban blocks as first-stage units.Why it matters: shows how sample design works in practice.Source: MoSPI
These are revision notes, not official predictions of exam content.
Quick Revision
- Primary = first-hand; secondary = already collected.
- Census = every unit; sample = a part.
- Random sampling: simple, systematic, stratified.
- Sampling error falls as the sample grows; non-sampling error does not.
- Non-sampling errors: bias, non-response, recording errors.
- Pilot survey tests the questionnaire.
- Census since 1881; NSS since 1950.
Related Glossary Terms and Chapters
Glossary:
Primary dataSecondary dataCensusSample surveySampling error
Related chapters: Introduction, which explains why data are needed; Organisation of Data, which shows what to do with raw figures once collected.
Sources
- NCERT textbooks: Statistics for Economics, Class 11, National Council of Educational Research and Training. Syllabus reference. Accessed 23 October 2026.
- Office of the Registrar General and Census Commissioner, India, Ministry of Home Affairs. Accessed 23 October 2026.
- Ministry of Statistics and Programme Implementation (MoSPI), Government of India. Accessed 23 October 2026.
- India Code: Central Acts, Legislative Department, Government of India. Accessed 23 October 2026.
- Press Information Bureau, Government of India. Accessed 23 October 2026.
About this page
Drafted with AI assistance from the NCERT chapter structure, in original words. Content is checked and approved by Dr. Anamika Shrivastava, PhD Economics, before publication. Read how this site works or send a correction.
Syllabus year 2026-27 · Last updated