Abstract
The Granular Instrumental Variables (GIV) methodology exploits panels withfactor error structures to construct instruments to estimate structural timeseries models with endogeneity even after controlling for latent factors. Weextend the GIV methodology in several dimensions. First, we extend theidentification procedure to a large $N$ and large $T$ framework, which dependson the asymptotic Herfindahl index of the size distribution of $N$cross-sectional units. Second, we treat both the factors and loadings asunknown and show that the sampling error in the estimated instrument andfactors is negligible when considering the limiting distribution of thestructural parameters. Third, we show that the sampling error in thehigh-dimensional precision matrix is negligible in our estimation algorithm.Fourth, we overidentify the structural parameters with additional constructedinstruments, which leads to efficiency gains. Monte Carlo evidence is presentedto support our asymptotic theory and application to the global crude oil marketleads to new results.