Global economic transformations have never been as dramatic as in the twentieth century. Most countries have experienced radical changes in the standards of income per capita, technology, fertility, mortality, income inequality and the extent of democracy in the course of the past century. It is the goal of many disciplines--economics, history, demography, sociology, political science--to comment these transformations, assess their causes and describe their consequences. But one major obstacle hinders the analysis of such long term processes: the lack of data. In particular, there does not exist any data spanning over the whole century that describes one fundamental aspect of economic development: education, the knowledge of nations. In this paper, the authors make a contribution by building consistent series of average years of schooling in 74 countries for the period 1870-2010. This has never been achieved before probably because of the huge amount of data that needed to be treated adequately to ensure comparability across countries and time. This involves about 30 000 figures. Their series derive from two data sets. The first one spans over 1870-1960 and is original, the second describes the period 1960-2010 and has been constructed by Cohen and Soto (2007), quoted hereafter as Cohen-Soto. Several assumptions were needed to complete a consistent data set describing educational attainment over such a long period. Thus, it is important to gauge to what extent the authors' series are influenced by these assumptions. A large part of this paper is devoted to a discussion of this issue. In particular, the authors find that missing data can generate sizeable measurement errors at the beginning of the period, especially for less advanced countries. They show that in most European countries schooling is estimated accurately as soon as 1870, and that 1900 constitutes a good start date for other countries. Also, comparing their figures with Cohen-Soto in 1960, the authors find a high correlation of 0.96. As the two methodologies are completely different, one relying on perpetual inventory of enrolment at school, the other mainly on surveys, this proves to be an excellent result. Besides, the comparison between the two databases in 1960 motivates a statistical framework that corrects time-persistent measurement errors in the authors' historical data set, as well as systematic ones in Cohen-Soto. Indeed, a third of their data relies on surveys conducted in the 1990s, which were used to infer average schooling in 1960. However, the latter authors neglected differential mortality across educational groups. As a result, they "overestimated" average schooling in 1960, or equivalently, "underestimated" the growth of schooling between 1960 and 1970. As a result, the data reflects an unprecedented global development of education that has accelerated after the Second World War. From that perspective, the twentieth century has clearly been the "Century of Education". Importantly, the authors show that their global distribution of years of schooling has widened since 1870. They also find that the two globalization periods have witnessed a convergence in average years of schooling for all countries with average schooling above a minimal threshold of 2 years--about 30% of literate people. This convergence has been rapid during the former globalization era and much more modest since 1980. Annexed are: (1) Additional tables and figures; (2) Annex to the Century of Education (Christian Morrisson and Fabrice Murtin); and (3) Data. (Contains 4 tables, 19 figures and 34 footnotes.)