This paper reappraises Tachibanaki and Yokoyama (2008) -an empirical analysis that indicates no apparent backward shifting of employer social insurance contributions- by modifying their empirical strategy. First, we attempt to control for a spurious positive correlation between wages and employer's contribution rates by trend variables. Second, we exclude two industries from our sample that have small numbers of workers and establishments to remove sampling errors in wages. Our results imply that the social insurance burden shifts back on to employees to a certain extent, contrary to Tachibanaki and Yokoyama (2008). Our finding is consistent with other existing studies.