This study extends previous research on teachers' self-efficacy by exploring reciprocal effects of teachers' self-efficacy and instructional quality in a longitudinal panel study. The study design combined a self-report measure of teacher self-efficacy with teacher and student ratings of instructional quality (assessing cognitive activation, classroom management, and individual learning support for students), and 2-level cross-lagged structural equation analyses were conducted. Data were collected from 155 German secondary mathematics teachers and 3,483 Grade 9 students at 2 measurement points. Although cross-sectional correlations between self-efficacy beliefs and characteristics of instruction were substantiated, the analyses only partially confirmed a causal effect of teachers' self-efficacy on later instructional quality. Instead, the analyses revealed a reverse effect of instructional quality on teachers' self-efficacy, with students' experience of cognitive activation and teachers' ratings of classroom management predicting teachers' subsequent self-efficacy. Our findings emphasize the importance of examining teachers' self-efficacy not only as a cause but also as a consequence of educational processes. Future research on teachers' self-efficacy should take a longitudinal perspective with varying time lags, identify possible mediator variables, and consider other aspects of teacher competence beyond self-efficacy when examining the effects of instructional quality.