Proportional Stratified Sampling — Proportional Stratified Random Sampling
Proportional Stratified Random Sampling · Also known as: proportionate stratified sampling, proportional allocation stratified sampling, PSRS, proportionate stratified random sampling
Proportional stratified sampling divides the target population into non-overlapping strata (subgroups defined by a key characteristic such as age band, region, or gender) and then draws a simple random sample from each stratum so that each stratum's share of the total sample matches its share of the total population. Because each subgroup is represented in exact proportion to its population weight, the resulting sample mirrors the population structure closely without requiring post-hoc weighting adjustments.
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.
+10 more
When to use it
Use proportional stratified sampling when (1) a reliable sampling frame exists that includes the stratification variable for every unit; (2) the stratification variable is strongly correlated with the outcome, promising a variance reduction; (3) you need assured representation of all major subgroups for subgroup-level analysis; and (4) you have the resources to administer separate sampling operations per stratum. Avoid it when the sampling frame is incomplete or lacks stratum membership information, when the stratifying variable has no demonstrated relationship to the outcome (the variance gain disappears), when strata are very small (fewer than ~5 units per stratum), or when you need oversampling of rare subgroups — in that case, use disproportional stratified sampling instead.
Strengths & limitations
- Reduces sampling variance relative to simple random sampling of the same size when the stratifying variable correlates with the outcome.
- Guarantees that each stratum is represented in exact proportion to its population share, eliminating stratum under- or over-representation.
- Self-weighting design simplifies analysis — no differential weights are needed under full response.
- Supports reliable subgroup comparisons because each subgroup's sample size is determined by its population share.
- Transparent and easily auditable allocation formula — the proportional rule is verifiable by any stakeholder.
- Requires a complete sampling frame with stratum membership data for every population unit before sampling begins.
- Efficiency gain materialises only if the stratification variable genuinely predicts variation in the target outcome; a poorly chosen stratifying variable yields no improvement over simple random sampling.
- Small strata receive proportionally small samples, which may be inadequate for stratum-level estimates — disproportional allocation or oversampling is needed for rare subgroups.
- Administering separate random samples for each stratum adds logistical complexity compared to a single unstratified draw.
Frequently asked
How is proportional stratified sampling different from disproportional stratified sampling?
In proportional allocation every stratum receives a sample share equal to its population share (n_h / n = N_h / N), producing equal inclusion probabilities and a self-weighting sample. In disproportional allocation strata receive larger or smaller shares than their population weights — typically to oversample rare or highly variable strata. Disproportional designs require weighting during analysis to restore population-level estimates, but they can improve overall or subgroup precision when strata differ greatly in variability or size.
When does proportional stratification actually improve precision over simple random sampling?
Proportional stratification reduces variance only to the extent that units within strata are more homogeneous than the population as a whole. The variance reduction equals the between-stratum variance in the outcome. If the stratifying variable has no relationship to the outcome, within-stratum variance equals overall variance and stratification provides no efficiency gain. Rule of thumb: choose stratification variables that explain at least 10–20% of the outcome variance.
Do I still need weights if I use proportional stratified sampling?
Under perfect proportional allocation with complete response, every unit has the same inclusion probability (n / N) and the design is self-weighting — no differential weights are needed. In practice, however, non-response rates typically vary across strata, breaking the equal-probability property. In that case, non-response or calibration weights must be computed per stratum before making population-level inferences.
How do I handle a stratum that is too small to yield a reliable sample under proportional allocation?
If a stratum's proportional allocation yields fewer units than you need for reliable stratum-level estimates (commonly fewer than 20–30), switch to disproportional allocation: set a minimum floor for that stratum and apply post-stratification or inverse-probability weights to correct the imbalance when estimating population totals or means.
What software can I use to implement proportional stratified sampling?
R packages such as 'sampling' (Tillé & Matei) and 'survey' (Lumley) support stratified designs with proportional allocation and design-based variance estimation. In Python, the 'samply' and 'scikit-survey' libraries offer similar functionality. SPSS Complex Samples and SAS PROC SURVEYSELECT both implement proportional stratified selection with a straightforward syntax.
Sources
- Cochran, W. G. (1977). Sampling Techniques (3rd ed.). John Wiley & Sons. ISBN: 978-0471162407
- Kish, L. (1965). Survey Sampling. John Wiley & Sons. ISBN: 978-0471489009
How to cite this page
ScholarGate. (2026, June 3). Proportional Stratified Random Sampling. ScholarGate. https://scholargate.app/en/survey-methodology/proportional-stratified-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
- Disproportional Stratified SamplingSurvey Methodology↔ compare
- Multistage SamplingSurvey Methodology↔ compare
- Simple random samplingSurvey Methodology↔ compare
- Stratified SamplingSurvey Methodology↔ compare
- Systematic SamplingSurvey Methodology↔ compare