Agricultural labor accounts for the largest share of child labor worldwide. Yet, measurement of farm labor statistics is challenging due to its inherent seasonality, variable and irregular work schedules, and the varying saliences of individuals' work activities. The problem is further complicated by the presence of widespread gender stratification of work and social lives. This study reports the findings of three randomized survey design interventions conducted over the agricultural coffee calendar in rural Ethiopia to address whether response by proxy rather than self-report has effects on the measurement of child labor statistics within and across seasons. While the estimates do not report differences for boys across all seasons, the analysis shows sizable self/proxy discrepancies in child labor statistics for girls. Overall, the results highlight concerns on the use of survey proxy respondents in agricultural labor, particularly for girls. The main findings have important implications for policymakers about data collection in rural areas in developing countries.