The quality of the OECD's Economic Outlook growth projections was last evaluated in-house at the peak of the previous business cycle, calling for a reassessment. This paper analyses the OECD's annual GDP growth projections for the G7 countries over the period 1991-2006 and compares them with the Consensus Economics forecasts. It shows that OECD growth projections display a number of desirable features: projections for the current year are unbiased and efficient; projection errors tend to shrink as the horizon shortens; and projections are directionally accurate most of the time. Like those produced elsewhere, the OECD projections also suffer from shortcomings: one-year-ahead projections display a positive bias, mainly reflecting a propensity to overpredict during slowdowns; spring one-year-ahead projections are far less informative than autumn ones; and turning points are poorly anticipated one year ahead. Regression tests suggest that the OECD and Consensus add value to naïve forecasts for spring current-year and autumn one-year-ahead projections.