This paper discusses the development of a social capital （SC） propagation model which affects local mutual aid after an earthquake disaster.　Mutual aid within community residences （local mutual aid） is important before the arrival of public help or volunteers.　Citizen networks or trusts are measured as SC, and SC is related to local mutual aid.　The case study cities Sendai and Hamamatsu, Japan were used for a web survey administrated to collect local mutual aid data.　A logit model was used to determine factors that affect local mutual aid.　Multi-Agent Simulation was used to develop the SC propagation model.　As a result, the model recognized that trust, interaction and exchange, community disaster recovery potential and neighborhood behavior affect local mutual aid behavior.　Experiencing a disaster encourages the creation of neighborhood networks to provide local mutual aid.　In the case of cities with a higher percentage of the younger generation, local mutual aid with neighborhood networks is effective during the safety confirmation stage and aftershock stage.　Moreover, when movers who are able to provide mutual aid assist the last 10% of the fixed persons who are unable to provide mutual aid, the fixed persons’ density affects the overall behavior of local mutual aid assistance. JEL Classification:Q54, R00, R58