A method, system and computer program product for identifying and quantifying sentiment and promotion bias. A set of sentiment as a measure of a group opinion is received, where a sentiment is expressed as a weight associated with an edge between two vertices in a graph. Balanced graphs associated with spanning trees corresponding to multiple views of a signed graph are constructed using the set of sentiments. Scores are assigned for each of the multiple views of the signed graph to determine an influence one group has over another group while maintaining agreement. An inequitable ratio of the assigned scores over agreeable sets of vertices of the signed graph is obtained. Based on the inequitable ratio of the assigned scores over agreeable sets of vertices of the signed graph, a more accurate sentiment bias is identified and quantified in contrast to currently existing sentiment analysis tools.