Risk-aversion has generally been found to decrease in income. This may lead one to expect that poor countries will be more risk-averse than rich countries. Recent comparative findings with students suggest the opposite, potentially giving rise to a risk-income paradox. Findings with students, however, may result from selection effects. We test whether a paradox indeed exists by measuring the risk preferences of over 500 household heads representative of the highlands of Ethiopia. We find high degrees of risk tolerance, consistent with the evidence obtained for students. We also find risk tolerance to increase in income proxies, thus completing the paradox. Using plausibly exogenous income proxies allows us to conclude that part of the causality must run from income to risk tolerance. It also avoids problems with measurement of income in developing countries that may have led to attenuation bias in some studies. Our findings further suggest that risk preferences cannot be blamed for the failure to adopt new technologies. Alternative explanations are discussed.