Working Paper: NBER ID: w9153
Authors: Edward L. Glaeser; Bruce I. Sacerdote; José A. Scheinkman
Abstract: In many cases, aggregate data is used to make inferences about individual level behavior. If there are social interactions in which one person's actions influence his neighbor's incentives or information, then these inferences are inappropriate. The presence of positive social interactions, or strategic complementarities, implies the existence of a social multiplier where aggregate relationships will overstate individual elasticities. We present a brief model and then estimate the size of the social multiplier in three areas: the impact of education on wages, the impact of demographics on crime and group membership among Dartmouth roommates. In all three areas there appears to be a significant social multiplier.
Keywords: social multiplier; social interactions; aggregate data; individual behavior
JEL Codes: I20; J24; R23
Edges that are evidenced by causal inference methods are in orange, and the rest are in light blue.
| Cause | Effect |
|---|---|
| Positive social interactions (Z13) | social multiplier effect (C31) |
| Neighbors' average outcome (D79) | individual's outcome (I14) |
| Aggregate relationships (C43) | individual elasticities (D11) |
| Social multiplier (groups of eight floors) (C39) | 1.4 (Y20) |
| Social multiplier (groups of 28 dorms) (C92) | 2.2 (Y20) |
| Social multiplier (county level) (C31) | 1.7 (Y20) |
| Social multiplier (state level) (C31) | 2.8 (Y20) |
| Social multiplier (national level) (E16) | 8.2 (Y20) |
| Social multiplier (puma level) (C39) | 1.67 (C29) |
| Social multiplier (state level) (C31) | 2.17 (human capital spillovers) (J24) |