Hierarchical forecast reconciliation on Italian covid-19 data

Due to the spread of the Covid-19 pandemic, many different actions to limit personal freedom have been decided according to the trend of the epidemi- ological situation. Often these decisions were made at the national level or, in a second phase, at the regional one. In order to ensure the implementation of actions and behaviours more in line with the real epidemiological situation, it would be useful to decide and give specific information at a more granular territorial level (provinces) even if one of the main problem is the quality of the data. This paper presents how the use of forecast reconciliation techniques can help forecasts at a detailed territorial level, while exploiting the more robust information available at the aggregate level.