Thank you for your interest in our data! If you use this data, please cite:

Serina Y Chang*, Emma Pierson*, Pang Wei Koh*, Jaline Gerardin, Beth Redbird, David Grusky, and Jure Leskovec. "Mobility network models of COVID-19 explain inequities and inform reopening". Nature, 2020.

Currently, we provide the hourly mobility networks from March 1 to May 2, 2020, that we used in our paper, so that other researchers can reproduce our work and build on it. In future releases, we will also include a separate subdirectory containing networks with greater coverage from January 2019 to present.

Our mobility networks capture, for each hour, the number of people who live in census block group (CBG) c_i visiting point-of-interest (POI) p_j. We used SafeGraph Patterns data to estimate the number of hourly visitors at each POI, SafeGraph Social Distancing Metrics to estimate the number of people in each CBG who left their homes to visit POIs, and apply the iterative proportional fitting procedure (IPFP) to estimate the matrix from these row and column marginals. Please see the Methods and Supplementary Methods sections in our paper for details.

We provide the mobility networks for the 10 metropolitan statistical areas we study in our paper. For each metro area, we include:

• *_2020-03-01_to_2020-05-02.pkl.gz: the hourly mobility networks. This is a pickle file that, when loaded in Python, produces a list of matrices, one per hour. The shape of each matrix is M x N, where M is the number of POIs and N is the number of CBGs. There should be 1,512 matrices in total, representing each hour from March 1, 2020, 12am to May 2, 2020, 11pm, in the time zone of that metro area.
• *_poi_ids.csv: the POI ids corresponding to the M rows in each hourly matrix.
• *_cbg_ids.csv: the CBG ids corresponding to the N columns in each hourly matrix.