This article investigate the role of social influence for the expectation formation of economic agents. Using self-organizing Kohonen maps the repeated cross-section data set of the University of Michigan consumer survey is transformed into a pseudo panel allowing to monitor the expectation formation of cohorts with regard to business confidence over the whole available time span (January 1978 - June 2013). Subsequently the information theoretic concept of transfer entropy is used to reveal the role of social influence on the expectation formation as well as the underlying network structure. It is shown that social influence strongly depends on socio-demographic characteristics and also coincides with a high degree of connectivity. The social network estimated in this way follows a power-law and thus exhibits similar structure as networks observed in other contexts.