Cluster Sampling Calculator

Cluster sampling is a probability sampling method where the population is divided into groups (clusters), a random selection of clusters is chosen, and then individuals within those clusters are surveyed. It is especially useful when the population is geographically spread out and a complete list of individuals is unavailable. See also our calculate Probability P(p̂), Mean of Sampling Distribution (μp̂) & Standard Error (σp̂) — Sampling Distribution of Proportion.

The design effect (DEFF) measures how much the variance of an estimate increases due to the cluster sampling design compared to simple random sampling of the same size. A DEFF greater than 1 means clustering reduces statistical efficiency, which is why you need a larger sample. It is calculated as DEFF = 1 + (b − 1) × ρ, where b is the cluster size and ρ is the intracluster correlation.

Intracluster correlation (ICC or ρ) measures how similar individuals within the same cluster are relative to individuals from different clusters. A higher ICC means members of a cluster share more similar characteristics, which reduces the information gained from sampling multiple people in the same cluster and increases the required sample size. Typical values range from 0.01 to 0.20 in social research.

Not everyone you contact will respond (response rate), not everyone who responds will qualify (eligibility rate), and your sampling frame may not cover the entire population (coverage rate). Each of these factors means you must contact more people to achieve your effective completed-interview target, so the final sample size is inflated accordingly. You might also find our calculate Normal Probability for Sampling Probability P useful.

If you have no prior information about the prevalence of the characteristic you are measuring, use 50% (0.5). This is the most conservative assumption and yields the largest — and safest — sample size estimate, ensuring you do not underpower your study.

A higher confidence level requires a larger sample size. For example, moving from a 95% to a 99% confidence level increases the z-score from 1.96 to 2.576, which grows your required sample proportionally. Balancing precision with available resources is key to choosing the right confidence level.

In one-stage cluster sampling, all individuals within selected clusters are surveyed. In two-stage cluster sampling, a random subset of individuals is selected from within each chosen cluster. This calculator uses the two-stage approach, where you specify the number of interviews per cluster, which affects the design effect and total number of clusters needed.

Once you have the cluster-adjusted sample size (SRS size multiplied by the design effect, then inflated by adjustment rates), divide that final sample size by the number of interviews per cluster. The result gives you the minimum number of clusters to randomly select from the population.