Home Statistical Dictionary Longitudinal Survey
Longitudinal Survey
A survey design that collects data from the same (or a related) sample repeatedly over time, allowing direct measurement of individual-level change and a clearer temporal sequence between variables than a single-wave cross-sectional survey can provide.
In Plain English
A longitudinal survey follows people, or at least the same underlying units, across multiple points in time, rather than taking just one snapshot. This lets researchers directly observe how individuals change, does someone's health status improve after starting a new treatment, does someone's income rise after completing a training program, questions a cross-sectional survey simply cannot answer, since it never revisits the same people. The trade-off is cost and complexity: longitudinal surveys require tracking and re-contacting the same respondents over time, are vulnerable to attrition as respondents drop out along the way, and generally take much longer to field and analyze than a single cross-sectional survey.
Definition
A longitudinal survey collects data from the same sample units (a panel survey, tracking specific individuals) or from related, evolving cohorts (a cohort study in the epidemiological sense) at multiple points in time, in contrast to a cross-sectional survey's single-wave design. Longitudinal designs allow direct measurement of within-individual change over time and provide a clearer basis for establishing temporal sequence between variables (which changed first) than cross-sectional data can, strengthening, though not by itself proving, causal interpretation; the principal costs of longitudinal design are the greater time, expense, and logistical complexity of tracking and re-contacting the same units repeatedly, and vulnerability to attrition, participants lost to follow-up over successive waves, which can bias results if it is correlated with the outcome under study (differential attrition). Panel surveys and rotating panel designs are the two most common implementations of the longitudinal survey concept in practice.
Properties
- Longitudinal surveys provide a stronger basis than cross-sectional surveys for establishing which of two associated variables changed first, an essential (though not sufficient on its own) ingredient for causal inference, but they still cannot rule out confounding the way a randomized controlled trial can, an observed sequential relationship in longitudinal data can still be driven by an unmeasured third variable.
- Attrition across waves is one of the most consequential practical challenges in longitudinal survey design, if participants who drop out differ systematically from those who remain on the outcome under study, differential attrition, remaining longitudinal data can become biased in ways a single cross-sectional wave's nonresponse never could be, since the bias compounds and can correlate with change over time specifically.
- Panel surveys, tracking the exact same individuals across every wave, and rotating panel designs, tracking overlapping but not identical groups across waves, are the two dominant practical structures for implementing longitudinal data collection, each making a different trade-off between measuring pure individual-level change and managing respondent burden and long-run attrition.
At a Glance