Most panel inference procedures discussed so far assume that apart from the possible presence of individual invariant but period varying time-specific effects, the effects of omitted variables are independently distributed across cross-sectional units. Often economic theory predicts that agents take actions that lead to interdependence among themselves. For example, the prediction that risk-averse agents will make insurance contracts allowing them to smooth idiosyncratic shocks implies dependence in consumption across individuals. Cross-sectional units could also be affected by common omitted factors. The presence of cross-sectional dependence can substantially complicate statistical inference for a panel data model. However, properly exploiting the dependence across cross-sectional units in panel data not only can improve the efficiency of parameter estimates, but it may also simplify statistical inference than the situation where only cross-sectional data are available. In Section 9.1 we discuss issues of ignoring cross-sectional dependence. Sections 9.2 and 9.3 discuss spatial and factor approaches for modeling cross-sectional dependence. Section 9.4 discusses cross-sectional mean augmented approach for controlling the impact of cross-sectional dependency. Section 9.5 discusses procedures for testing cross-sectional dependence. Section 9.6 demonstrates that when panel data are cross-sectionally dependent, sometimes it may considerably simplify statistical analysis compared to the case of when only cross-sectional data are available by considering the measurement of treat effects.

ISSUES OF CROSS-SECTIONAL DEPENDENCE

Standard two-way effects models (e.g. (3.6.8)) imply observations across individuals are equal correlated. However, the impacts of common omitted factors could be different for different individuals.