Robust structural analysis when theoretical information is scarce

• Applicant: Professor Dr. Helmut Herwartz
• Subject Area: Statistics and Econometrics
• Term: since 2022
• Project identifier: Deutsche Forschungsgemeinschaft (DFG) - Project number 504720211

During the first year of the project we have contributed the approach of kernel based maximum likelihood (KML) estimation to the literature on statistical identification of structural models (mainly structural VARs, SVARs). Going beyond the case of homoscedastic shocks we have also developed a heteroskedasticity-robust variant of KML estimation that applies in the empirically relevant case of co-heteroskedasticity which crucially challenges established data-based approaches to identification in SVARs (both heteroskedastiicty-based and independence-based identification). Moreover, we have delivered important contributions to the empirical analysis of global oil markets (and the structural determinants of CO2 emissions) and the (heterogenous) transmission of monetary policy of the European Central Bank across member states of the European monetary union. From our empirical exercises and the related empirical (and theoretical) literature the assumption of time invariant structural i.e. causal relations might be subject to debate such that more flexible approaches to statistical identification deserve attention. For purposes of time varying structural modeling we suggest two alternative model classes, Markov switching (MS) approaches and so-called functional coefficient models (FCMs). While MS- and FCM-based identifications have their own merits, we argue that both approaches also promise cross confirming insights, for instance, if the latent states in MS models correspond to the business cycle while key gaps in the macroeconomy (e.g., output or inflation gaps) are of effective performance in shaping the structural parameters of interest in the framework of FCMs. Issues of dimension reduction and software provision remain high on our research agenda.

• DFG Programme: Research Grants