Coffee consumption has been frequently reported for its protective association with incident dementia. However, this association has mostly been reported in studies with short follow-up periods, and it remains unclear to what extent reverse causality influences this association. Studying the long-term effect of coffee consumption on dementia with stratified follow-up time may help resolve this issue. In the population-based Rotterdam Study, coffee consumption was assessed in 1989–1991 (N = 5,408), and reassessed in 1997–1999 (N = 4,368). Follow-up for dementia was complete until 2011. We investigated the association of coffee consumption and incident dementia for the two examination rounds separately using flexible parametric survival models. We studied the entire follow-up period as well as stratified follow-up time at 4 years. For both examination rounds, we did not find an association between coffee consumption and dementia over the entire follow-up. In contrast, for both examination rounds, a protective association was observed only in the follow-up stratum of 0–4 years. Our data suggest that coffee consumption is not associated with incident dementia during long-term. The protective association observed in the short-term might be driven by reverse causality.