ISS 2016 Statistics Paper-2 Solution: Question 31 (NSSO Consumer Expenditure Survey Sampling Design)

Continuing our question-by-question walk through the ISS Statistics Paper-2 archive, we now arrive at Question 31 of the 2016 paper. This one moves away from formula-heavy estimation and into the practical world of how India's large sample surveys are actually designed and conducted on the ground.

Quick Summary

• Topic: Sampling Designs used in large-scale government surveys (Applied Statistics / Sample Surveys)

• Question Reference: ISS 2016, Statistics Paper-2, Question 31

• Correct Answer: (c) Stratified Multi-stage Random Sampling

The Question As It Appeared in the Paper

Which one of the following sampling designs is used in conducting the Consumer Expenditure Survey by the National Sample Survey Office (NSSO)?

• (a) Simple Random Sampling

• (b) Quota Sampling

• (c) Stratified Multi-stage Random Sampling

• (d) Judgement Sampling

Step 1: Understand What the Question is Really Testing

This is a conceptual question about survey methodology rather than a numerical one, so there is no formula to plug values into. Instead, you are being asked to recall a real design choice made by a real national agency, the NSSO, when it runs the Consumer Expenditure Survey (CES) across the entire country. To answer correctly, you need to know two things: what each listed sampling design actually means, and which one is practically feasible when you are trying to survey households spread across every state, district, and village in India.

Step 2: Go Through Each Option and Test It Against Reality

Let us examine every option one at a time, the way you would eliminate wrong choices in an exam hall.

Option (a) Simple Random Sampling (SRS):In SRS, every unit in the population has an equal and independent chance of being chosen, and selection is typically done from a single, complete list of all units. For a country-wide survey, this would mean having one giant list of every single household in India and drawing names from it directly. That is administratively impossible to maintain and far too expensive to execute, so SRS alone is ruled out for this survey.

Option (b) Quota Sampling:Quota sampling is a non-random method where the investigator simply fills pre-decided quotas (say, a certain number of respondents from each income group) without following any probability-based selection rule. Because it does not give every unit a known, non-zero chance of selection, it is not a probability sampling method, and government sample surveys like the CES are built entirely on probability sampling so that reliable, unbiased estimates with measurable sampling error can be produced. This rules out (b).

Option (d) Judgement Sampling:Here the investigator personally decides, based on their own judgement, which units "represent" the population best. Again, this has no probabilistic basis, no way to compute standard errors, and is prone to the surveyor's personal bias. A survey as important as the CES, whose results feed into national poverty and inflation estimates, cannot be built on personal judgement. This rules out (d).

Option (c) Stratified Multi-stage Random Sampling:This is the one method built specifically for exactly this kind of national-scale problem, and we unpack why in the next step.

Step 3: Understand Why Stratification is Needed First

India is enormously diverse — a household's spending pattern in rural Bihar looks nothing like one in urban Mumbai. If you sampled completely at random from the whole country in one shot, purely by chance you could end up with a sample overloaded with, say, urban Maharashtra and almost nothing from rural Northeast India. To prevent this, the population is first divided into strata that are internally similar — commonly by State/Union Territory and by sector (rural vs. urban). Sampling is then carried out separately within each stratum. This guarantees that every region and every sector gets fair, proportionate representation in the final sample, and it also reduces the sampling variance compared to one large unstratified draw.

Step 4: Understand Why "Multi-stage" Comes Next

Even after stratification, you still cannot list every single household in, say, rural Uttar Pradesh and sample directly from it — that list simply does not exist in a usable form, and even if it did, the selected households would be scattered across thousands of villages, making field visits absurdly expensive. The solution is to sample in stages:

• Stage 1:Within each stratum, first randomly select a number of villages (in rural areas) or urban blocks (in urban areas). These are called the First Stage Units (FSUs).

• Stage 2:Within each selected village/block, randomly select a smaller number of households. These are the Second Stage Units (SSUs).

This "narrowing funnel" approach is what makes the word multi-stage apply. It keeps fieldwork geographically clustered and manageable for investigators, while random selection at every stage still preserves the probability-sampling property needed for valid statistical inference.

Pro Tip: The easiest way to remember this for the exam is to separate the two ideas cleanly in your head — "stratified" answers the question "how do I make sure every type of area is represented?", while "multi-stage" answers the question "how do I physically go about selecting units without a full population list?" Any large government survey (CES, Employment-Unemployment Survey, NSS rounds in general) almost always combines both ideas, so whenever you see NSSO in a question stem, stratified multi-stage random sampling should be the first design that comes to mind. This is exactly the kind of pattern-recognition habit that gets drilled repeatedly in a structured classroom setting — the way it's usually taught at a place like Sunrise Classes, students are made to link "NSSO / large-scale government survey" with "stratified multi-stage" so firmly that the association becomes automatic, rather than something to reconstruct from scratch under exam pressure.

Step 5: Confirm the Final Answer

Putting it together: SRS is impractical for a population this size and diversity, quota and judgement sampling are non-probabilistic and unsuitable for official statistics, leaving stratified multi-stage random sampling as the design actually used by the NSSO for the Consumer Expenditure Survey. This matches option (c).

Checking against the official answer key for ISS 2016 Statistics Paper-2: the key lists the answer to Question 31 as (c). Our reasoning and the official key agree —the correct answer is (c) Stratified Multi-stage Random Sampling.

Why This Question Matters

Sample survey methodology questions like this one show up repeatedly across ISS papers because the exam is, at its core, testing whether a future officer of the Indian Statistical Service understands how India's real statistical machinery actually works. Knowing the theoretical formulas for stratified and multi-stage sampling is necessary, but knowing which real survey uses which design — and why — is what separates a candidate who has only memorised formulas from one who understands applied official statistics. This particular fact (NSSO CES uses stratified multi-stage random sampling) reappears in various forms across NSS-related questions, GS papers, and even interview rounds, so it is worth locking in permanently rather than treating as a one-off fact.

Frequently Asked Questions

What is the difference between multi-stage sampling and multi-phase sampling?

In multi-stage sampling, you select smaller and smaller sampling units at each stage (villages, then households) but always collect the actual survey data only from the final-stage units. In multi-phase sampling, you collect some information from a large sample in an initial phase, and then collect additional, more detailed information from a smaller sub-sample drawn from that same first-phase sample. The two ideas solve different problems and should not be confused with each other.

Why can't the NSSO just use simple random sampling directly?

Simple random sampling needs one complete, accessible list of every unit in the population to draw from. No such single list of every household in India exists or could realistically be maintained and updated, so direct SRS across the whole country is not administratively feasible for a survey of this scale.

What exactly are First Stage Units and Second Stage Units in the CES?

The First Stage Units (FSUs) are the villages (in rural areas) or urban frame survey blocks (in urban areas) that get randomly selected first within each stratum. The Second Stage Units (SSUs) are the individual households randomly selected from within each chosen FSU. Data is ultimately collected only from these final selected households.

Is stratified multi-stage sampling always used for every NSS survey round?

Stratified multi-stage random sampling is the standard broad approach for most large-scale NSS surveys, including the Consumer Expenditure Survey and the Periodic Labour Force Survey, because it balances representativeness with field feasibility. Exact details of stratification variables and number of stages can vary slightly across specific survey rounds and years.

Does stratification reduce or increase the sampling error compared to an unstratified design?

When strata are formed sensibly (so that units within a stratum are more similar to each other than to units in other strata), stratified sampling generally reduces the overall sampling variance compared to an equivalent-sized unstratified random sample, because it removes between-stratum variation from the error term.

How should I prepare for this type of conceptual, non-numerical MCQ?

Rather than only solving numerical problems, build a short mental note-sheet of "which real survey uses which sampling design and why" — CES, PLFS, Census, and similar well-known surveys are common exam favourites. Revisiting a handful of past-year questions on survey methodology, like this one, is usually enough to cover the recurring patterns.

If any step above felt unclear, drop your doubt in the comments and we will work through it together — and if this helped, do share it with a fellow ISS aspirant who is also working through the Paper-2 archive question by question.