Objective Circulating testosterone declines during male ageing, and low testosterone may predispose to ill health. We sought to determine whether greater participation in healthy behaviours predicted reduced risk of subsequent lower circulating testosterone in older men.Design Cross-sectional analysis of a population-based follow-up study.Participants A total of 3453 men aged 65–83 years.Measurements Lifestyle score, a tally of eight prudent health-related behaviours, was determined during 1996–99. Early morning sera collected in 2001–04 were assayed for total testosterone, SHBG and LH. Free testosterone was calculated using mass action equations.Results Mean (± SD) time between collection of lifestyle data and blood sampling was 5·7 ± 0·9 years. Lifestyle score correlated with subsequent total testosterone (r = 0·06, P <0·001) and SHBG (r = 0·07, P <0·001), but not free testosterone (r = 0·03, P = 0·08) or LH (r = –0·03, P = 0·12). In multivariate analyses, higher lifestyle scores (4 and above) predicted reduced risk of total testosterone and SHBG in the lowest quartile of values. For the highest category (≥ 7), odds ratio (95% CI) for total testosterone and SHBG in the lowest quartile were 0·37 (0·18–0·77) and 0·26 (0·13–0·54), respectively. Lower lifestyle scores including and excluding body mass index predicted higher risk of total testosterone and SHBG in the lowest quartiles.Conclusions In men > 65 years old, higher lifestyle score reflecting greater engagement in healthy behaviours predicts higher subsequent total testosterone and SHBG levels. This relationship appears cumulative and may reflect interaction between lifestyle and insulin sensitivity. Successfully promoting healthy behaviours in older men could ameliorate the age-related decline in circulating testosterone.