The socio-economic gradient in health remains a controversial topic in economics and other social sciences. In this paper we develop a new duration model that allows for unobserved persistent individual-specific health shocks and provides new evidence on the roles of socio-economic characteristics in determining length of life using 19-years of high-quality panel data from the German Socio-Economic Panel. We also contribute to the rapidly growing literature on life satisfaction by testing if more satisfied people live longer. Our results clearly confirm the importance of income, education and marriage as important factors in determining longevity. For example, a one-log point increase in real household monthly income leads to a 12% decline in the probability of death. We find a large role for unobserved health shocks, with 5-years of shocks explaining the same amount of the variation in length of life as all the other observed individual and socio-economic characteristics (with the exception of age) combined. Individuals with a high level of life satisfaction when initially interviewed live significantly longer, but this effect is completely due to the fact that less satisfied individuals are typically less healthy. We are also able to confirm the findings of previous studies that self-assessed health status has significant explanatory power in predicting future mortality and is therefore a useful measure of morbidity. Finally, we suggest that the duration model developed in this paper is a useful tool when analyzing a wide-range of single-spell durations where individual-specific shocks are likely to be important.