Establishing a respectful and well-functional program committee (PC) consisting of capable PC members is one of the most important tasks for conference organizers. However, little research has been done for automatic recommendation of PC candidates. PC member finding is a complex task, which could be influenced by many factors such as the candidates' research interests' match with conference topics, the candidates' social closeness with PC chairs, the candidates' authoritativeness, as well as the candidates' publication history in the conference. To examine the importance of each feature, we build a dataset that consists of papers from four conferences: KDD, SIGIR, JCDL and GIS (2007-2011) and split it into the training and testing subsets based on the temporal information. The results show that: i) the publication history is the strongest indicator of being PC members; ii) recommendations based on the social closeness also produce reasonable good results; iii) recommend high authority researchers as PC members fails to predict the real PC because there are a large proportion of PC members who actually only have low authority values (we use the PageRank value in coauthor networks to simulate researcher's authority); and iv) applying simple linear combination of different features can make reasonable improvements.