Are online social relationships more egalitarian than those take place offline? With the rise of online social networking sites like Twitter, users are able to choose freely with whom they will communicate. Prior researches on online social networking patterns have mainly concerned the implicit structural pressure or sociological irrationality, rather than underlying preference of users, in explaining the process of weak tie generations. In this study, we test whether online social relationships reflect offline features of users. This thesis consists of two parts - an individual-level analysis and a county-level analysis for US populations. In the individual-level approach using a logistic regression model with Twitter panel-data, we explore how residential county income affects the probability of a success of interaction on Twitter. Our model verified the tendency that people who live in wealthier counties are more welcomed as a new friend in Twitter. Also, the result shows that users in poorer counties care more about the opponent"s income than users in the richer counterpart. In the second part of this thesis, through a stochastic actor-based model for seven monthly observations, we evaluated the effect of median household income on the dynamics of the Twitter communication network between counties within each state.