It is often assumed that low levels of teacher self-efficacy (TSE) leads to negative outcomes, including burnout; however, the temporal order of the construct predictions has rarely been examined. We used an autoregressive cross-lagged panel design to examine whether TSE and burnout are concurrently associated with each other, whether TSE predicts future burnout levels, and/or whether burnout predicts future TSE levels. An initial sample of 3002 Croatian teachers (82% female) from across three educational levels (i.e., elementary, middle, and secondary schools) with varying years of teaching experiences (M=15.28, SD=10.50) completed questionnaires on their levels of TSE and burnout (exhaustion and disengagement) at three time points (at approximately six-month intervals). We found that burnout has a more prominent role in predicting future levels of TSE than TSE does in predicting future levels of burnout. These findings challenge the theoretical and empirical conceptualizations assuming that TSE is a predictor of burnout. Policies and interventions that focus on decreasing teacher burnout rather than increasing TSE levels may be best.