Using data from the Household, Income and Labour Dynamics in Australia survey (waves 2001-2014), we construct measures of unexpected health shocks for each year, using information on respondents' views about the expected and past evolution of their health status. By definition, unexpected health shocks are immune to the problem of reverse causality (running from health condition to retirement). Our findings indicate that retirement increases the likelihood of positive health shocks and decreases the probability of negative shocks for men, with no clear results for women. These shocks are mirrored by variations in life satisfaction of the same nature (e.g. increased life satisfaction in case of unexpected positive health shocks). Other indicators of mental and physical health taken from the SF-36 vary in the same way, i.e. improve unexpectedly after retirement for men. These findings suggest that, at least in the case of men, people's desire to retire may not be based on perfectly correct expectations about the impact of this move, but is aligned with its actual consequence: retirement exerts a positive causal impact on health.