Studies have found that loneliness is as bad as smoking or obesity for mortality risk, and the prevalence of loneliness is predicted to increase with ageing populations, more people living alone, and with chronic health conditions. Despite the substantial literature on loneliness, there is little detailed research on the extent of economic gradients. In this paper we provide this evidence using a sample of around 400,000 respondents (aged 40-70) from the UK Biobank. We focus on differences in loneliness across educational attainment, household income and neighbourhood deprivation, as well as recent major life events including financial difficulties. Using two statistical approaches, we find a substantially higher probability of experiencing loneliness, but also social isolation and a lack of social support, for men and women with low socioeconomic status, even when comparing those residing in the same postcode district. Additionally, the recent experience of financial stress is strongly associated with worse social health. Our results are robust to a panel analysis that accounts for intercorrelations between loneliness, social isolation and lack of social support, and controls for sample attrition.