Global identification of dynamic panel models with interactive effects

We investigate the problem of global identification in dynamic panel models with interactive effects, in the large-N, fixed-T setting.

While local identification, typically established via the Jacobian matrix, is well understood, global identification has remained a more elusive and challenging issue.

It is commonly believed to be unachievable in this context.

However, we demonstrate that the model is, in fact, globally identified for almost all configurations of the factors.

Our analysis also covers models with additive fixed effects, including unit-root cases in which previous studies have reported non-identification from differenced moments.

We show that, even in these settings, the level covariance structure delivers global identification.