Xavier Sala-i-Martin discusses the classical approach to convergence analysis, focusing on σ-convergence and β-convergence. The paper applies these concepts to various data sets, including 110 countries, OECD countries, U.S. states, Japanese prefectures, and European regions. It finds strong evidence of σ-convergence and absolute β-convergence in most data sets, except for the cross-section of countries, which shows σ-divergence and conditional β-convergence. The speed of conditional convergence is close to 2% per year. The paper defines σ-convergence as a decrease in the dispersion of real per capita GDP levels over time, while β-convergence refers to faster growth of poorer economies. These concepts are related but distinct. σ-convergence is about the distribution of income across countries, while β-convergence is about the growth rates of individual economies. The paper analyzes cross-country evidence, finding that the cross-section of 110 countries does not show β-convergence, suggesting that poor countries do not grow faster than rich ones. However, when conditioning on variables that proxy for the steady state, the data shows conditional β-convergence, with a convergence speed close to 2% per year. The paper also examines regional convergence within countries, finding evidence of both absolute and conditional β-convergence, as well as σ-convergence, in various regions. The speed of convergence is consistently close to 2% per year. The paper concludes that the classical convergence literature shows that there is no σ-convergence or absolute β-convergence in the world economy between 1960 and 1990. However, when conditioning on variables that proxy for the steady state, the data shows conditional β-convergence. The estimated speed of conditional convergence is close to 2% per year. The paper also notes that the process of σ-convergence seemed to stop for about a decade in the mid-1970s. The estimated capital share implied by the convergence speed is higher than traditional estimates, suggesting that the neoclassical model may predict too high a speed of convergence. The paper emphasizes the importance of conditional convergence in understanding economic growth and convergence.Xavier Sala-i-Martin discusses the classical approach to convergence analysis, focusing on σ-convergence and β-convergence. The paper applies these concepts to various data sets, including 110 countries, OECD countries, U.S. states, Japanese prefectures, and European regions. It finds strong evidence of σ-convergence and absolute β-convergence in most data sets, except for the cross-section of countries, which shows σ-divergence and conditional β-convergence. The speed of conditional convergence is close to 2% per year. The paper defines σ-convergence as a decrease in the dispersion of real per capita GDP levels over time, while β-convergence refers to faster growth of poorer economies. These concepts are related but distinct. σ-convergence is about the distribution of income across countries, while β-convergence is about the growth rates of individual economies. The paper analyzes cross-country evidence, finding that the cross-section of 110 countries does not show β-convergence, suggesting that poor countries do not grow faster than rich ones. However, when conditioning on variables that proxy for the steady state, the data shows conditional β-convergence, with a convergence speed close to 2% per year. The paper also examines regional convergence within countries, finding evidence of both absolute and conditional β-convergence, as well as σ-convergence, in various regions. The speed of convergence is consistently close to 2% per year. The paper concludes that the classical convergence literature shows that there is no σ-convergence or absolute β-convergence in the world economy between 1960 and 1990. However, when conditioning on variables that proxy for the steady state, the data shows conditional β-convergence. The estimated speed of conditional convergence is close to 2% per year. The paper also notes that the process of σ-convergence seemed to stop for about a decade in the mid-1970s. The estimated capital share implied by the convergence speed is higher than traditional estimates, suggesting that the neoclassical model may predict too high a speed of convergence. The paper emphasizes the importance of conditional convergence in understanding economic growth and convergence.