Panel Spatial Autocorrelation
Panel Data Spatial Autocorrelation Analysis · Also known as: spatial autocorrelation in panel data, panel spatial dependence, spatio-temporal autocorrelation, cross-sectional dependence in panels
Panel Spatial Autocorrelation measures whether observations that are geographically close also tend to have similar values across repeated time periods. It extends classic cross-sectional spatial autocorrelation statistics such as Moran's I to panel data, enabling researchers to detect spatial dependence consistently over time and to diagnose whether a panel regression model requires a spatial component.
Read the full method
Sign in with a free account to read this section.
Method map
The neighbourhood of related methods — select a node to explore.
When to use it
Use Panel Spatial Autocorrelation when you have repeated observations across geographic units (countries, regions, census tracts) and want to confirm or rule out spatial dependence before fitting a panel regression. It is appropriate when the panel is balanced or near-balanced and units can be assigned a location or shared-border relationship. Do not use it as a standalone model for prediction; it is a diagnostic and exploratory tool. Avoid it when geographic proximity has no theoretical relevance to the outcome, or when the number of time periods is very small (T < 5), as period-level statistics become unreliable.
Strengths & limitations
- Directly tests whether spatial dependence is a stable feature of the panel, informing model selection (spatial lag vs. spatial error vs. aspatial panel).
- Period-by-period Moran's I values reveal whether spatial clustering intensifies, weakens, or shifts over time.
- Extends well-understood cross-sectional spatial statistics to the panel context without requiring a full structural model.
- Joint test statistics (e.g., Anselin's LM tests for panels) have good power against both spatial lag and spatial error alternatives.
- Compatible with standard fixed-effects and random-effects residuals, so it can be layered on top of standard panel estimators.
- Results depend heavily on the choice of spatial weights matrix W; there is no universally correct specification.
- The period-specific Moran's I values are not independent across time if there is serial correlation, complicating joint inference.
- With large N and small T the asymptotic approximations for the pooled statistic may be poor.
- Detects the presence of spatial dependence but does not identify its source or the correct spatial model to use.
- Edge effects and irregularly shaped or sized spatial units can distort the weights matrix and bias results.
Frequently asked
How does Panel Spatial Autocorrelation differ from ordinary Moran's I?
Ordinary Moran's I is computed for a single cross-section. Panel Spatial Autocorrelation applies Moran's I (or equivalent diagnostics) to each time period's residuals and then aggregates the results, providing an overall test that accounts for the repeated-measurement structure of the data.
Which spatial weights matrix should I use?
There is no universal answer. Queen or rook contiguity is common for administrative regions sharing borders. Inverse-distance or k-nearest-neighbour weights suit point data or highly irregular units. Sensitivity analysis across two or three plausible specifications is recommended; conclusions that change with W deserve caution.
What do I do if I find significant spatial autocorrelation?
The next step is model selection: compare the Lagrange Multiplier tests for the spatial lag panel model versus the spatial error panel model. If both are significant, the spatial Durbin panel model or the general nesting spatial model may be appropriate. Re-estimate with the preferred spatial specification and re-run the autocorrelation diagnostics on the new residuals.
Can I apply this with an unbalanced panel?
Yes, but with care. Period-specific Moran's I can still be computed as long as the same spatial units are observed in each period. Missing units complicate the construction of a consistent W matrix; listwise deletion or imputation may be needed before analysis.
Is Panel Spatial Autocorrelation the same as Pesaran's CD test?
They are related but not identical. Pesaran's cross-sectional dependence (CD) test is a general test for correlation among panel units without requiring a geographic structure, while Panel Spatial Autocorrelation specifically uses a spatial weights matrix. Pesaran's CD is more powerful when dependence is non-geographic; Moran-based panel tests are better when the dependence follows a spatial pattern.
Sources
- Anselin, L. (2013). Spatial Econometrics: Methods and Models. Springer Netherlands. (Originally published 1988.) ISBN: 978-9401577991
- Elhorst, J. P. (2014). Spatial Econometrics: From Cross-Sectional Data to Spatial Panels. Springer Berlin Heidelberg. ISBN: 978-3642403408
How to cite this page
ScholarGate. (2026, June 3). Panel Data Spatial Autocorrelation Analysis. ScholarGate. https://scholargate.app/en/spatial-analysis/panel-spatial-autocorrelation
Which method?
Set this method beside its closest kin and read them side by side — the library lays the books on the table; the choice is yours.
- Geographically Weighted RegressionSpatial analysis↔ compare
- Local Spatial AutocorrelationSpatial analysis↔ compare
- Moran's ISpatial analysis↔ compare
- Panel Spatial Error ModelSpatial analysis↔ compare
- Spatial AutocorrelationSpatial analysis↔ compare