Must every unit in a selected area be surveyed?

No. A further probability sample can be selected, provided its selection process is reflected in the weights.

Can remote sensing replace all fieldwork?

It can support frame construction and measurement, but whether it measures the target accurately depends on the survey's purpose and validation.

Does area sampling require equal-sized areas?

No. Designs can accommodate different sizes, with appropriate selection and estimation methods.

AI Glossary / Statistics and probability / Intermediate

Area sampling

Definition

Area sampling is survey sampling in which geographical areas serve as sampling units at one or more stages. Researchers select areas and then observe all relevant units within them or sample units within the selected areas. A probability-based design uses known selection probabilities to support estimation for the defined target population.

What it means

A complete list of people, businesses or farms may be unavailable, while maps can provide a usable geographical starting point. Area sampling lets researchers concentrate listing and fieldwork in selected places. It is the sampling method, whereas an area sampling frame is the organized geographical foundation from which selections are made. Selecting convenient neighborhoods is not automatically a probability sample merely because those neighborhoods have boundaries.

How it works

Define the population, reference period and geographical coverage. Partition the covered territory into identifiable sampling units, possibly grouped into strata. Select areas using a documented probability design. In selected areas, enumerate the relevant units or create a current list and select a further sample. Overall inclusion probabilities combine the applicable selection stages. Estimation uses weights and a variance procedure appropriate to stratification, clustering and other design features. Operational rules must specify how to handle units crossing boundaries, inaccessible locations and changes between frame construction and fieldwork.

Three practical examples

Household research: select enumeration areas, list occupied dwellings within them and sample households for interviews.
Agricultural statistics: select land segments and record crop areas or other agricultural activities using clear rules for holdings that cross segment boundaries.
Local business research: select blocks and enumerate establishments, helping include businesses missing from an existing directory while documenting coverage limitations.

Think of it this way

Think of surveying selected pages of a map before listing every address in a country. The map narrows where detailed work is needed. It does not remove the need to select places fairly, find the units inside them and account for their selection chances when drawing conclusions.

Worked example

Suppose a town is completely divided into 100 nonoverlapping blocks. Select 10 blocks by simple random sampling without replacement and count every eligible shop in each selected block. The observed block counts are 8, 12, 6, 14, 10, 9, 11, 7, 13 and 10. They sum to 100, so the sample mean is 10 shops per block. Each block has inclusion probability 10/100 = 0.1, giving a total estimate of 100/0.1 = 1,000 shops. The squared deviations of the counts from 10 sum to 60, so the sample variance is 60/9, approximately 6.667. Under this simple design, the estimated variance of the total is 100 squared × (1 - 10/100) × 6.667/10 = 6,000. The estimated standard error is approximately 77.46 shops. This calculation uses blocks as sampled units; treating the 100 observed shops as 100 independently sampled blocks would be wrong. It assumes complete, accurate enumeration within sampled blocks.

Illustrative case study

Fictional case: Lantern Market Research wants to estimate the number of small repair businesses in a city. An online directory misses informal premises, so the team builds a block-based probability sample and conducts field listing. Investigators distinguish active businesses from vacant addresses and document businesses operating across multiple premises. Analysts estimate totals using the area design and report remaining coverage concerns, including home-based operations that field listing may miss. The case illustrates survey planning; it does not claim that geographical sampling automatically discovers every business.

Common misconceptions and mistakes

Area sampling is not the same as small-area estimation, which concerns estimates for geographical or other domains and may use model-based techniques. Equal selection probabilities for areas do not always imply equal probabilities for people, especially when subsampling a fixed number from areas of different sizes. Nearby units often resemble each other, so clustering can affect precision. A larger number of interviews in a few selected areas may provide less information than the same number spread across more areas. Maps, weights and random selection do not by themselves correct nonresponse or missing target units.

Business use

Businesses can use area sampling for market assessment, service coverage research and independent checks on customer or establishment lists. Compare the cost of dispersed travel with the precision consequences of concentrating observations. Decide whether the unit of interest is a person, household, business, premise or parcel of land; these are not interchangeable. Preserve area identifiers and design information so analysts can estimate uncertainty properly. Budget for listing and quality checks, not just interviews, because incomplete identification within selected areas can undermine an otherwise sound design.

Check your understanding

Q: Why is the worked example's expansion factor ten?

Answer: A: Each block has a one-in-ten selection chance, and all eligible shops within a selected block are counted. Each observed block total therefore represents ten blocks under the stated design.

Related terms

Area sampling frame
Cluster Sampling
Multistage Sampling
Stratification
Coverage error
inclusion probability
survey weight
enumeration area and design effect

Learn next

extend the worked example by sampling shops within selected blocks and calculate the combined inclusion probabilities. Then compare designs that visit more blocks with fewer interviews per block against designs concentrated in fewer blocks, using realistic costs and within-area similarity.

Sources

Sources: OECD Glossary of Statistical Terms, area sampling terminology: https://www.oecd.org/content/dam/oecd/en/publications/reports/2008/09/oecd-glossary-of-statistical-terms_g1gh9ad7/9789264055087-en.pdf ; USDA NASS, Area Frame Samples: https://www.nass.usda.gov/Education_and_Outreach/Understanding_Statistics/Foundation_of_Estimates/Area_Frame_Samples/ ; FAO, guidance on probability area frames and identifiable, nonoverlapping units: https://www.fao.org/4/w0022e/w0022e04.htm .

Previous: ARD kernel Next: Area sampling frame