We design a new field experiment to test pro-social behaviour: will a household return a letter that has been incorrectly addressed? On average, we find that half of all letters were returned. Return rates do not vary significantly according to the gender, race or ethnicity of the fictitious...