A growing body of literature underlines the reluctance of respondents to surveys to express certain attitudes openly. The standard mechanism attributed to strategic masking of sentiment is social desirability pressure, which results in a bias such that respondents match their responses to expectations of what the social context expects. Therefore, social desirability bias (SDB) is the resulting gap between overt and covert expression. However, little is known about how social desirability pressure changes over time. This is difficult to observe as there are two sources of change for some subgroups (e.g., political party supporters). First, the extent to which sentiment is masked could increase or decrease for group members, which reflects attitudinal change. Second, new entrants and exits could change the overall pattern of masking, reflecting compositional change. A further complication is that the best approach to measure SDB, the list experiment, can only estimate group-level covert and overt attitudes. This work proposes a new method to decompose the observed change in SDB into compositional and attitudinal change. Using longitudinal data from Norway, this work offers insight into how observed change in SDB for some political subgroups (e.g., Labour [AP] and Progress [FrP]) converge and diverge only when attitudinal and compositional change can be differentiated.