Nowcasting from disaggregates in the face of location shifts

Given a need for nowcasting, we consider how nowcasts can best be achieved, the use and timing of information, including disaggregation over variables and common features, and the role of automatic model selection for nowcasting missing disaggregates. We focus on the impact of location shifts on nowcast failure and nowcasting during breaks, using impulse saturation, its relation to intercept correction, and to robust methods to avoid systematic nowcast failure. We propose a nowcasting strategy, building models of all N disaggregate series by automatic methods, forecasting every variable each period, then testing for shifts in available measures, switching to robust forecasts of missing series when breaks are detected.