Inference with multi-indexed data: a review of recent advances and remaining challenges

Empirical researchers routinely have to deal with multi-indexed data. For instance, trade flows may be indexed by time, importer, exporter and type of good. These data naturally exhibit network-type dependence or multiway clustering along the different indices. Handling such dependence structures can prove challenging. In particular, it is often not obvious a priori whether clustering/dependence should be accounted for along only some dimensions or all at the same time. In the trade example, this boils down to the well-known practical question: “Should I cluster at the importer, exporter or importer-and-exporter level?”. These issues get even more complex when fixed effects get included in the picture. In this talk, I will present a series of recent research articles I have been working on that seek to give both a better theoretical understanding and simple practical guidelines to researchers working with multi-indexed data.