What is Probability Proportional to Size (PPS)?
Probability Proportional to Size (PPS) is a sampling method where units (usually clusters like villages or schools) are selected with a probability proportional to a known size measure (for example population or number of households). It helps ensure larger clusters are more likely to be chosen when cluster sizes vary widely.
PPS is used mainly in multi-stage or cluster sampling. Instead of giving every cluster the same chance of selection, each cluster’s chance is tied to a size measure (e.g., population, number of households, school enrollment). Larger clusters therefore have a higher probability of being picked. That makes it easier to approximate a representative sample of individuals without needing to list every person beforehand.
Typical steps: choose a size measure for each cluster, compute the selection probabilities proportional to those sizes, select clusters using those probabilities, then sample a fixed number of elements (people/households) within each chosen cluster. If you sample the same number of people inside each selected cluster, PPS can produce a self-weighting design (each person in the whole population has roughly the same chance of selection), simplifying analysis.
Usage example
A council wants a household survey across 100 neighbourhoods that vary from 50 to 5,000 residents. They use neighbourhood population as the size measure, select 30 neighbourhoods with PPS (so bigger neighbourhoods are more likely to be picked), then interview 10 households chosen randomly within each selected neighbourhood.
Practical application
Why it matters: PPS makes field surveys more efficient and more likely to reflect the true population when clusters differ greatly in size. It reduces the administrative burden of listing every individual before sampling, lowers travel and data-collection costs by concentrating work in selected clusters, and—when combined with equal within-cluster samples—can produce a design where individual selection probabilities are equal (self-weighting). That simplifies analysis and reporting for teams who aren’t survey-methods specialists.
Caveats: PPS depends on a reliable size measure; if that measure is wrong the sample can be biased. If you don’t sample the same number of people inside each selected cluster, you must apply weights afterward to get unbiased estimates.
FAQ
How is PPS different from simple random sampling (SRS)?
SRS gives every individual or cluster the same chance of being chosen. PPS gives clusters different chances based on size. PPS is useful when clusters vary a lot in size and you want larger clusters to be more likely chosen without listing every individual first.
What should I use as the "size" for PPS?
Common size measures are population, number of households, registered voters, or school enrollment—whatever best approximates how many people you want represented. Choose a measure that’s current, accurate and closely linked to the outcome you’re measuring.
Do I need to weight responses after a PPS sample?
Sometimes. If you select the same number of respondents within each chosen cluster, PPS can be self-weighting and no further weighting is needed. If within-cluster sample sizes vary or if selection probabilities aren’t exactly proportional to true sizes, you’ll need to apply design weights to correct estimates.
When should I avoid PPS?
Avoid PPS when you lack a reliable size measure, when clusters are similar in size (SRS may be simpler then), or when your study requires equal probability selection of clusters for other design reasons. Also be cautious if small clusters contain important subpopulations you don’t want underrepresented.