This paper presents a new dataset on years of schooling across countries for the period 1960–2000, constructed from the OECD database on educational attainment and UNESCO surveys. The dataset improves upon previous data by using uniform classification systems over time and intensifying age-group information. These improvements allow the dataset to replace Barro and Lee's data in empirical research. In cross-country growth regressions, the dataset yields significant coefficients for schooling. In panel data estimates, the series remain significant even when accounting for physical capital accumulation. The estimated macro return is consistent with labor studies. These results differ from earlier literature due to reduced measurement error. The debate on human capital's role in economic growth has evolved over the past two decades. Initially, human capital was seen as a key driver of long-term growth. Later, a neo-classical view emerged, suggesting human capital was an ordinary input. A "revisionist" view later argued human capital's role was overstated. The paper argues that the debate's extremes stem from measurement issues, both conceptually and empirically. Conceptually, there is no clear definition of human capital, with years of schooling often used as a proxy. However, data shows that regions with rapid schooling growth started from low levels, making it unlikely that schooling increases significantly boost human capital. Mankiw, Romer, and Weil (1992) represented human capital indirectly, but this approach is rejected by data. Recent literature has turned to micro-approaches, such as the Mincerian model, which views human capital as an exponential function of schooling. This model suggests that the gap in human capital between rich and poor countries has remained constant over the last forty years. The second issue is data quality. Existing human capital data is unreliable, as shown by De la Fuente and Domenech (2002, 2006). The paper's contribution is a new dataset with improved quality, based on OECD data and UNESCO sources. The dataset uses age-group information and avoids inconsistent education classifications, reducing measurement error. These features lead to more accurate results.This paper presents a new dataset on years of schooling across countries for the period 1960–2000, constructed from the OECD database on educational attainment and UNESCO surveys. The dataset improves upon previous data by using uniform classification systems over time and intensifying age-group information. These improvements allow the dataset to replace Barro and Lee's data in empirical research. In cross-country growth regressions, the dataset yields significant coefficients for schooling. In panel data estimates, the series remain significant even when accounting for physical capital accumulation. The estimated macro return is consistent with labor studies. These results differ from earlier literature due to reduced measurement error. The debate on human capital's role in economic growth has evolved over the past two decades. Initially, human capital was seen as a key driver of long-term growth. Later, a neo-classical view emerged, suggesting human capital was an ordinary input. A "revisionist" view later argued human capital's role was overstated. The paper argues that the debate's extremes stem from measurement issues, both conceptually and empirically. Conceptually, there is no clear definition of human capital, with years of schooling often used as a proxy. However, data shows that regions with rapid schooling growth started from low levels, making it unlikely that schooling increases significantly boost human capital. Mankiw, Romer, and Weil (1992) represented human capital indirectly, but this approach is rejected by data. Recent literature has turned to micro-approaches, such as the Mincerian model, which views human capital as an exponential function of schooling. This model suggests that the gap in human capital between rich and poor countries has remained constant over the last forty years. The second issue is data quality. Existing human capital data is unreliable, as shown by De la Fuente and Domenech (2002, 2006). The paper's contribution is a new dataset with improved quality, based on OECD data and UNESCO sources. The dataset uses age-group information and avoids inconsistent education classifications, reducing measurement error. These features lead to more accurate results.