Figure 5
Motifs in philanthropic funding and predictability. (a) We show the prevalence of 3-node and 4-node bipartite motifs compared to a bipartite configuration null model. The prevalence ratio versus the null model takes the value 1 when the motif appears the same number of times in the 2019 bipartite funding network and the bipartite configuration model (marked with a dashed line), while a value above 1 means the motif is overrepresented. We see that only the 4-node motif reflecting loops of length 4 is overrepresented. (b) The receiver-operator curve (ROC) for predictions using the AA index from 2018 on the network of grants over $10,000 to predict grants over $10,000 in 2019. The area under the curve (AUC) is 0.87. (c) Examples of resulting predictions. For four funders (in gold) we show the recipients for whom they were highest ranked (see SI note). On the left we show the resulting prediction network for the Sorenson Legacy Foundation and Tanner Foundation, both located in Utah. We highlight recipients in Utah with a purple border. On the right we show the recipients who ranked the Lozier and Daugherty foundations highest and highlighted with a black border those in Nebraska. We see that the network structure alone identified predictions consistent with the locality of grants. Figure generated with Cytoscape 3.10.
