How teachers spend their working time is crucial for the quality of teaching and teachers’ well-being. For this reason, teacher surveys generally measure the number of hours worked. By comparing estimates from survey questions and time-diaries, we argue that commonly used survey methods are prone to bias. We use results of a unique, large-scale study where 7,486 teachers kept a diary for seven days resulting in 1,250,000 hours of registered activities. Large differences in the accuracy of estimating working time exist in sub-activities (e.g., school organisation) and job characteristics (e.g., part-time versus full-time). Implications for research and policy are discussed.