We develop and apply a procedure to test the welfare implications of a beauty and non-beauty contest based on survey forecasts of interest rates and yields in a large country sample over an extended period of time. In most countries, interest-rate forecasts are unbiased and consistent with both models, but are rarely supported by yield forecasts. In half of the countries, a higher precision of public information regarding interest rates increases welfare. During forward guidance, public information is less precise than private information.