Partial Interventions in Networks

Ryan Kor,Junjie Zhou
DOI: https://doi.org/10.2139/ssrn.3868878
2021-01-01
Abstract:We study the optimal welfare maximising intervention in a multiple activity network model taking into account both within-activity network spillovers and cross-activity interdependences. We analyse the effects of a restriction in the intervention space, by limiting the activities which the planner can intervene in, and show that the impact of a restriction greatly differs depending on the interdependency between activities. When the planner's budget is large, the percentage loss in welfare is significant when activities are complements, while there is asymptotically no loss in welfare when the activities are substitutes and the planner can intervene in at least two activities. When the budget is small, there are instead diminishing returns in the dimensions of intervention. For each case, we obtain approximations of the optimal interventions in terms of the eigenspaces of the network matrix, and provide a simple choice of intervention for the planner that achieves an asymptotically equal welfare.
What problem does this paper attempt to address?