Capture-Recapture Population Estimation
Also known as: Mark-Recapture, Tag-Recapture, Mark-Release-Recapture, İşaretle-Yeniden Yakala
Capture-recapture (also known as mark-recapture) is a statistical method for estimating the size of an unknown population by sampling it twice and tracking which individuals appear in both samples. Formally systematized for closed animal populations by Otis, Burnham, White, and Anderson in their landmark 1978 Wildlife Monographs paper, the method extends naturally to human populations, epidemiology, and incomplete administrative records.
Key highlights
- Estimates total population size without requiring complete enumeration, making it practical for elusive or geographically dispersed populations.
- The Chapman correction renders the basic estimator nearly unbiased even with moderate sample sizes, and closed-form confidence intervals are readily computed.
- Extends naturally to three or more samples via loglinear or Huggins models, accommodating heterogeneous capture probabilities.
- Widely applicable beyond ecology: used in epidemiology, census undercount estimation, software bug counting, and multiple-register linkage studies.
Intuition
This section is available to Pro members. Upgrade to Pro
How it works
This section is available to Pro members. Upgrade to Pro
When to use it
Use capture-recapture when you need to estimate a population size that cannot be directly enumerated and at least two independent sampling occasions are feasible. Core assumptions are a closed population (no births, deaths, immigration, or emigration between samples), equal catchability for all individuals, independence of samples, and reliable identification of recaptured individuals. Violations of closure or heterogeneous capture probabilities can bias estimates substantially. When capture probability varies by individual, stratified or model-based variants (e.g., loglinear models across multiple lists) are preferable.
Strengths & limitations
- Estimates total population size without requiring complete enumeration, making it practical for elusive or geographically dispersed populations.
- The Chapman correction renders the basic estimator nearly unbiased even with moderate sample sizes, and closed-form confidence intervals are readily computed.
- Extends naturally to three or more samples via loglinear or Huggins models, accommodating heterogeneous capture probabilities.
- Widely applicable beyond ecology: used in epidemiology, census undercount estimation, software bug counting, and multiple-register linkage studies.
- The closed-population assumption is often unrealistic for human or wildlife studies conducted over extended periods with births, deaths, or movement.
- Equal catchability is rarely met; individual heterogeneity in capture probability inflates estimates and can be difficult to detect or model without additional data.
- Small recapture counts (m2 close to zero) produce highly variable and potentially infinite estimates, making the method unreliable for very sparse samples.
- Matching individuals across samples requires reliable identification; misidentification or mark loss directly biases the population estimate.
Common pitfalls
This section is available to Pro members. Upgrade to Pro
Applications
This section is available to Pro members. Upgrade to Pro
Frequently asked
What is the minimum number of recaptures needed for a reliable estimate?
A common practical rule of thumb requires at least 7–10 recaptures (m2) for the Chapman estimator to have acceptable bias and for confidence intervals to be informative. Below this threshold, the estimate becomes highly unstable and the upper confidence bound can be extremely large or even undefined under some formulations.
How do I test whether the closed-population assumption is met?
For wildlife data with multiple trapping occasions, closure tests implemented in programs like CAPTURE or MARK (e.g., Stanley and Burnham's test) assess whether capture rates are consistent with a closed population. For administrative data linkage, expert judgment about the study window and population dynamics is typically required, as formal tests are rarely available.
Can capture-recapture be used with more than two samples?
Yes. With three or more samples or lists, loglinear models fit to the contingency table of capture histories allow estimation of the unobserved cell (individuals missed by all samples) while modeling dependence between lists. This multi-list approach is standard in epidemiological and census applications and is more robust to list dependence than the two-sample case.
Sources
- 1.Otis, D. L., Burnham, K. P., White, G. C., & Anderson, D. R. (1978). Statistical inference from capture data on closed animal populations. Wildlife Monographs, 62, 3–135.
You have read it. What now?
Cite this page
ScholarGate. (2026, June 2). Capture-Recapture. ScholarGate. https://scholargate.app/survey-methodology/capture-recapture