Using an original dataset collected among motorcyclists in New Delhi (2011), this paper compares three different survey measures of risk attitudes: self-assessment, hypothetical lotteries and income prospect choices. While previous research on risk aversion measurement methods in developing countries mainly looked at specific groups such as rural farmers or students, the dataset I use covers a large and heterogeneous urban population. I first show that all measurements are positively and highly correlated with one another, this being even more the case within methodologies and within domains. Subsequently, I investigate the predictive power of these different individual risk-aversion measurements on occupation choices and health decisions. Most of my elicited risk preferences appear to predict risky health behaviors well. Puzzling results are found with the lotteries and may be interpreted either as evidence of risk-compensation between domains or as an incapacity to capture the desired characteristic. Finally, thanks to information on religious beliefs and practices, I am able to verify that cultural background does not impact on the relationship between risk preferences and risky conducts. Overall, this analysis highlights that elicitation of risk-aversion measurements through surveys in a developing country like India thus appears possible.