Literacy rates; random sampling; marriage certificates; human capital.

N01, N33, J24

This paper quantitatively assesses to what extent signatures in marriage certificates can inform about literacy rates in pre-industrial states. The direct estimates are based on a novel and balanced random sample of marriage certificates for pre-unification Italy in 1815. Such figures are compared to all alternative sources available for close years, including direct and indirect approaches who focused on selected areas of Italy. The new empirical methodology is as important as the results found with it. Two main findings emerge. First, marriage certificates can accurately measure literacy rates in preindustrial Italy, but only when accompanied by rigorous sampling procedures. Indeed, the proposed empirical approach allows to go from local to aggregate estimates that generally are in line with other estimates for the period, and can be scaled up or applied in other contexts. Second, North vs Centre-South divides in 1815 are lower than previously suggested. This supports the hypothesis that French reforms and later restoration governments triggered a process of widening North-South literacy divides.