This paper explores the relationship between policy settings and extreme positive and negative growth events, what we call GDP tail risks, using quantile regression methods. Conditioning on several country characteristics such as the size, stage of development and openness to trade as well as macroeconomic policies, the following findings for a panel of mostly OECD countries emerge: First, countries with stronger banking supervision and capital market development, better quality of governance, higher foreign reserves and several labour market characteristics such as higher unemployment benefits and greater spending in active labour market policies tend to experience less severe negative growth shocks (negative tail risk). Second, greater use of macro-prudential tools is generally associated with less extreme positive growth shocks (positive tail risk) and lower average growth. Third, larger automatic stabilisers are associated with both less severe negative and positive growth shocks but also lower average growth.