Proportional Multistage Sampling
Also known as: proportional PPS multistage sampling, multistage probability proportional to size sampling, proportionate multistage cluster sampling, PPS multistage sampling
Proportional multistage sampling is a probability sampling design that selects units across two or more hierarchical stages — for example, regions, then districts, then households — where the number of units drawn at each stage is proportional to the size of each higher-level unit. By weighting selection probabilities to match cluster size, it produces self-weighting samples that closely mirror the population structure and simplify variance estimation.
Read the full method
Sign in with a free account to read this section.
Method map
The neighbourhood of related methods — select a node to explore.
When to use it
Use proportional multistage sampling when the target population is geographically dispersed or hierarchically organized and a complete individual-level sampling frame is unavailable or prohibitively expensive. It is the design of choice for national health surveys, education assessments (e.g., PISA), and large-scale household surveys. The method requires a reliable size measure for each PSU (e.g., census counts). Do not use it when clusters are so small or homogeneous that within-cluster variance is negligible — in that case simple random or stratified sampling will be more efficient. Avoid it when cluster boundaries are unstable, contested, or unknown.
Strengths & limitations
- Eliminates the need for a complete individual-level sampling frame at the outset, reducing cost and logistics.
- Proportional allocation produces approximately self-weighting samples, simplifying analysis.
- Scales efficiently to very large or nationally representative surveys covering millions of elements.
- Aligns selection probabilities with cluster size, preventing over-representation of small clusters.
- Widely accepted by funding bodies and journals for national and cross-national surveys.
- Design effect (DEFF) is typically greater than one due to within-cluster homogeneity, reducing effective sample size relative to simple random sampling.
- Requires an accurate and up-to-date size measure for each PSU; outdated census counts can introduce selection bias.
- More complex to implement and to explain than single-stage designs; requires statistical software that accounts for the complex survey structure.
- Variance estimation is more complicated than for simple random samples and requires specialized routines (e.g., Stata svy, R survey package).
Frequently asked
What is probability proportional to size (PPS) and why is it used?
PPS is a selection mechanism in which each unit's probability of being chosen equals its size divided by the total size of all units at that stage. It is used in multistage designs because it ensures that elements in large clusters are not under-represented relative to elements in small clusters. When combined with a fixed number of SSUs per selected PSU, PPS produces equal overall selection probabilities for all population elements, yielding a self-weighting sample.
How many stages should I use?
Two to three stages is most common. Two stages (e.g., regions → individuals) are simpler to manage; three stages (e.g., regions → districts → households) are used when no adequate secondary sampling frame exists at an intermediate level. Adding stages increases field efficiency but also increases variance. The design effect typically rises with each additional stage, reducing statistical power.
Do I need to apply survey weights in my analysis?
Under a perfectly proportional PPS design the sample is self-weighting and unweighted estimators are unbiased. In practice, non-response, post-stratification adjustments, or imprecise size measures often mean design weights are needed. Always use survey-adjusted estimation routines (e.g., R survey package, Stata svy commands) to account for clustering and stratification in standard error calculations, regardless of weighting.
How does proportional multistage sampling differ from proportional stratified sampling?
Proportional stratified sampling divides the population into mutually exclusive strata and draws a sample from each stratum in proportion to its size — all within a single stage. Proportional multistage sampling operates across successive hierarchical levels, selecting clusters first and then elements within clusters. Both exploit proportionality to achieve representativeness, but they differ in structure: stratified sampling requires a complete frame; multistage sampling does not.
What sample size is recommended for the number of PSUs?
Standard guidance (e.g., DHS methodology) recommends selecting at least 25–30 PSUs to obtain stable variance estimates. Fewer PSUs yield unreliable design-effect corrections even if the total element count is large. When budget forces a trade-off, prefer more PSUs with fewer SSUs per PSU over fewer PSUs with many SSUs.
Sources
- Kish, L. (1965). Survey Sampling. John Wiley & Sons. (Chapters 6–7 on multistage and PPS designs.) ISBN: 978-0471489009
- Cochran, W. G. (1977). Sampling Techniques (3rd ed.). John Wiley & Sons. (Chapter 11 on multistage sampling with proportional allocation.) ISBN: 978-0471162407
How to cite this page
ScholarGate. (2026, June 3). Proportional Multistage Sampling. ScholarGate. https://scholargate.app/en/survey-methodology/proportional-multistage-sampling
Which method?
Set this method beside its closest kin and read them side by side — the library lays the books on the table; the choice is yours.
- Cluster SamplingSurvey Methodology↔ compare
- Multistage SamplingSurvey Methodology↔ compare
- Proportional Cluster SamplingSurvey Methodology↔ compare
- Proportional Stratified SamplingSurvey Methodology↔ compare
- Stratified SamplingSurvey Methodology↔ compare
- Systematic SamplingSurvey Methodology↔ compare