This article quantifies the importance of demographic and food environment variables in determining the convergence rates for obesity incidence across the U.S. Using a spatial autoregressive model with county level data for 2004-2012, a period of rapid spread of obesity in the U.S., to estimate 𝛽- and 𝜎convergence rates; then we estimate a probit model to assess their determinants. Empirical results show 𝛽-convergence and 𝜎-convergence occurred in the US and its four regions in 2004-2012. The Northeast and the West have the highest speed of convergence in years 2008-2012. It was also found that 𝛽convergence doesn’t occur in counties in metro area in 2004-2012 and 2008-2012. The convergence rate is the largest in completely rural counties and consistently higher for men compared to women. A second-step probit regression shows that states with higher Hispanic proportion, higher availability of fruit and vegetables stores and full-service restaurants are less likely to have 𝛽-convergence, while states with higher poverty rates and sex ratio are more likely to have 𝛽-convergence. It also shows 𝜎-convergence is more likely to happen in states with a higher proportion of Hispanics and higher availability of fruit and vegetables.