Cross-Lagged Meta-Analysis Services
Does X lead to Y, does Y lead to X, or both? Cross-lagged meta-analysis synthesises the reciprocal, over-time relationships between two variables across many longitudinal studies—settling questions of direction and precedence that no single panel study can answer alone.
Cross-lagged meta-analysis synthesises the cross-lagged (over-time) effects between two variables from multiple longitudinal panel studies. By pooling both directions—X predicting later Y and Y predicting later X—while accounting for their stability over time, it estimates the reciprocal relationship across a literature and clarifies which direction dominates.
What cross-lagged meta-analysis does
A recurring question in the social sciences is one of direction: when two variables are related and change together over time, which drives which? Does job satisfaction raise performance, does performance raise satisfaction, or do they reinforce each other? A single cross-lagged panel study offers evidence, but individual studies are underpowered and often disagree. Cross-lagged meta-analysis pools their over-time effects to give a clearer, more powerful answer.
The cross-lagged panel design measures both variables at two (or more) time points and estimates four kinds of path: each variable’s stability over time (its autoregressive effect), and the two cross-lagged effects—earlier X predicting later Y, and earlier Y predicting later X—each controlling for the outcome’s own prior level. Cross-lagged meta-analysis synthesises those cross-lagged effects across studies, in both directions at once, so the relative size of the two paths reveals which direction of influence dominates across the literature.
When to use it
Use cross-lagged meta-analysis when your question is about the reciprocal, directional relationship between two variables over time, and there is a body of longitudinal panel studies reporting the relevant correlations. It is the tool for “which comes first” and “is the relationship mutual” questions—precedence, reciprocity, and the relative strength of two competing causal stories.
It is closely related to its neighbours: it is a specialised meta-analytic SEM focused on the cross-lagged structure, it shares the longitudinal concerns of longitudinal meta-analysis, and—critically—when the pooled studies use different time lags, the honest route is CoTiMA, because cross-lagged coefficients depend on the lag. Plain cross-lagged meta-analysis is most defensible when the studies’ time intervals are reasonably comparable.
What cross-lagged meta-analysis adds
| Bivariate meta-analysis | Cross-lagged meta-analysis | |
|---|---|---|
| What it pools | A single correlation X–Y | Both over-time directional paths |
| Direction | None (symmetric) | X→Y and Y→X separately |
| Controls prior levels? | No | Yes (autoregressive stability) |
| Answers | Are they related? | Which direction dominates? Is it mutual? |
| Watch out for | — | Different time lags → use CoTiMA |
Doing it credibly—and its limits
Cross-lagged meta-analysis is typically implemented as a meta-analytic SEM: the correlations among the variables across time points are pooled across studies (stage one), and the cross-lagged panel model is fitted to the pooled matrix (stage two). Doing this properly means using the two-stage approach so the standard errors and fit statistics reflect the evidence, testing whether the correlations are homogeneous enough to pool, and modelling the autoregressive (stability) paths—without them, a “cross-lagged” effect is just a lagged correlation, not a directional one.
Two cautions matter most. The time-lag problem is central: because cross-lagged coefficients depend on the interval between waves, pooling studies with very different lags is misleading—where lags vary substantially, CoTiMA is the rigorous alternative. And the traditional cross-lagged panel model has known interpretive limits—it does not by itself separate stable between-person differences from within-person change, and newer variants address this—so directional findings are reported as evidence of temporal precedence, not proof of causation.
Temporal precedence is strong evidence—but not proof of cause. A dominant cross-lagged path across a literature is compelling, yet it still rests on observational panel data. We report direction and reciprocity honestly, and flag the time-lag issue that decides whether plain pooling or CoTiMA is appropriate.
What we deliver in
We implement cross-lagged meta-analysis via two-stage meta-analytic SEM in R (metaSEM), and—where lags differ—continuous-time methods (ctsem / CoTiMA), with reproducible code, the pooled matrix, and the fitted model.
How we deliver a cross-lagged meta-analysis
Cross-lagged meta-analysis draws on both our meta-analysis and SEM practices, run on a full systematic-review workflow—so the directional model is estimated on a sound, reproducible synthesis.
We begin with a registered protocol and a specified cross-lagged model, a comprehensive search, and careful extraction of the cross-time correlations and the exact time lags. We pool the correlations (stage one), check homogeneity and the comparability of lags, and fit the cross-lagged model (stage two)—switching to continuous-time methods when the lags demand it.
Reporting follows PRISMA standards with SEM conventions, reporting both cross-lagged paths, the stability paths, and the lag handling transparently.
You receive the pooled cross-lagged estimates in both directions with their uncertainty, the stability paths, a clear statement of which direction dominates (and whether the relationship is reciprocal), heterogeneity and lag diagnostics, and reproducible code and data. The result is a defensible, literature-wide answer to a directional question.
Cross-lagged meta-analysis across Management & Allied Studies
“Which comes first” questions run through the whole of the social sciences—and we apply cross-lagged synthesis across the disciplines we serve.
Management & Organisational Studies
Reciprocal relationships such as satisfaction and performance, or engagement and turnover intention, resolved across the literature.
Applied & Organisational Psychology
A core setting: directional and reciprocal effects among attitudes, well-being, and behaviour over time.
Marketing & Consumer Research
Whether attitude drives behaviour, behaviour drives attitude, or both, across longitudinal consumer studies.
Economics & Public Policy
Directional questions between paired indicators measured repeatedly across evaluations.
Education & Learning Sciences
Reciprocal effects such as motivation and achievement—which leads which, synthesised across studies.
Health & Behavioural Science
Directional relationships between behaviour and outcome measured in longitudinal panels.
Cross-lagged meta-analysis: common questions
A “which comes first” question?
If you need to establish direction and reciprocity between two variables across a body of longitudinal studies, cross-lagged meta-analysis is the tool. We design and deliver it—via meta-analytic SEM, or CoTiMA when the lags demand it.