SVAR Identification with High-Frequency Macroeconomic Data

Starting from the theoretical observation that the identification problem of SVAR models is fundamentally a problem of variable aggregation, we propose a new identification method based on nowcasted macroeconomic data, i.e. on macroeconomic series reconstructed at high-frequency. The idea is that contemporaneous relations between macroeconomic quantities tend to vanish as the sampling frequency increases. We show, both theoretically and empirically, that by increasing the sampling frequency of the variables is possible to fully identify all the structural shocks with no further information. We also discuss how to parsimoniously account for the long-lasting persistence of high-frequency variables when estimating the structural relations through a high-frequency heterogeneous VAR model