Home Statistical Dictionary Design Effect

Design Effect

The ratio of a survey estimator's actual variance under a complex sampling design to the variance it would have had under simple random sampling of the same size -- a single summary number for how much a design's clustering, stratification, or weighting inflates or reduces precision.

Sampling Theory & Survey MethodsIntermediateFree

In Plain English

The design effect (often abbreviated DEFF) is a single number that answers the question, 'compared to a simple random sample of the same size, how much did our actual sampling design help or hurt our precision?' A design effect of 1 means the complex design performed exactly like SRS. A design effect above 1, the typical case for cluster sampling, means the design inflated the variance, so the effective sample size is smaller than the nominal one. A design effect below 1, which stratification can achieve, means the design actually improved precision beyond what SRS would give. It's the standard summary statistic survey statisticians use to translate a messy real-world design back into an equivalent, easy-to-interpret SRS sample size.

Definition

The design effect is defined as $DEFF = \frac{Var _{design} ( θ ^ )}{Var _{SRS} ( θ ^ )}$ , the ratio of an estimator's actual sampling variance under the survey's complex design to the variance it would have under simple random sampling of the same nominal sample size $n$ . For cluster sampling, DEFF is well-approximated by $1 + (\overset{n}{ˉ}_{c} - 1) ρ$ (Kish's formula), where $\overset{n}{ˉ}_{c}$ is the average cluster size and $ρ$ is the intraclass correlation; DEFF exceeds 1 whenever clustering introduces within-cluster homogeneity (the typical case), and can fall below 1 when effective stratification removes between-stratum variance. The design effect converts a design's nominal sample size into an effective sample size, $n_{eff} = n / DEFF$ , the sample size a simple random sample would need to achieve the same precision, a concept used both in survey analysis (to correctly inflate standard errors computed by software that assumes SRS by default) and in sample size planning (to determine how much larger a clustered sample must be to match the precision of a planned SRS).

Formula

DEFF = \frac{Var _{design} ( θ ^ )}{Var _{SRS} ( θ ^ )}, n_{eff} = \frac{n}{DEFF}

Notation

Var_{design}

—Actual sampling variance under the survey's real design

Var_{SRS}

—Variance under simple random sampling of the same size

n_{eff} = n / DEFF

—Effective sample size, in SRS-equivalent terms

Properties

Software that computes standard errors assuming simple random sampling (the default in many basic statistical packages) will understate the true standard errors of a clustered survey design unless the design effect is explicitly applied, ignoring DEFF is one of the most common analytical errors made with complex survey data.
The design effect compounds across multiple stages and sources of design complexity in a real survey, clustering, stratification, and unequal weighting each contribute their own component to the overall DEFF, which is why national survey organizations typically report an overall DEFF for key estimates rather than assuming it is 1.
Sample size planning for a clustered survey must inflate the target sample size by the anticipated design effect to achieve a desired precision, if a planner wants the precision of an SRS of 1,000 but expects DEFF = 2 from clustering, the clustered survey must sample roughly 2,000 units to compensate.

At a Glance

DEFF = 1the complex design performs identically to simple random sampling

DEFF > 1typical for cluster sampling; the design inflates variance

DEFF < 1achievable with effective stratification; the design improves precision

Used forcorrectly inflating standard errors and planning clustered-sample sizes

Last updated September 8, 2026← Back to the Dictionary