Psychologists and sociologists usually interpret answers to happiness surveys as cardinal and comparableacross respondents (Kahneman et al. 1999). As a result, these social scientists run OLS regressionson happiness and changes in happiness. Economists, on the other hand, usually only assume ordinalcomparability and have mainly used ordered latent response models. As a consequence, economists haveby and large not taken satisfactory account of fixed individual traits. We address this latter problemby developing a conditional estimator for the fixed-effect ordered logit model. The empirical findingspresented show that it makes virtually no difference whether one assumes ordinality or cardinality ofhappiness answers, whilst allowing for fixed-effects does change results substantially. This leads us toadvocate allowing for and endogenising the persistent personality traits that make up these fixed-effects.