The trade-off between child quantity and education is a crucial ingredient of unified growth models that explain the transition from Malthusian stagnation to modern growth. We present first evidence that such a trade-off indeed existed before the demographic transition, exploiting a unique census-based dataset of 334 Prussian counties in 1849. Estimating two separate instrumental-variable models that instrument education by landownership inequality and distance to Wittenberg and fertility by previous-generation fertility and sex-imbalance ratio, we find that causation between fertility and education runs both ways. Furthermore, education in 1849 predicts the fertility transition in 1880-1905.