**2. Data**

In the present work, we consider a universe of 300 liquid stocks from the NASDAQ exchange between 3 January 2017 and 25 September 2020. High frequency price data are obtained from LOBSTER [20] and sampled to obtain non-overlapping one-minute returns. This frequency was also adopted in [10] and other works in the literature for the detection of price jumps, as it is understood that below this limit, microstructural noise becomes relevant and can impact the validity of the method.
