This paper provides comprehensive and detailed empirical regression analyses of the sources of wage persistence. Exploring a rich matched employer-employee data set and the estimation of a dynamic panel wage equation with high-dimensional fixed effects, our empirical results show that permanent unobserved heterogeneity plays a key role in driving wage dynamics. The decomposition of the omitted variable bias indicates that the most important source of bias is the persistence of worker characteristics, followed by the heterogeneity of firms' wage policy and last by the job-match quality. We highlight the importance of the incidental parameter problem, which induces a severe downward bias in the autoregressive parameter estimate, through both an in-depth Monte Carlo study and an empirical analysis. Using three alternative bias correction methods (the split-panel Jackknife (Dhaene and Jochmans, 2015), an analytical expression (Hahn and Kuersteiner, 2002), and a residual based bootstrap approach (Everaert and Pozzi, 2007, Gonçalves and Kaffo, 2015)), we observe that up to one-third of the reduction of the autoregressive parameter estimates induced by the control of permanent heterogeneity (high dimensional fixed effects) may not be justified.