To understand population dynamics it is necessary to understand vital rates, which may be affected by a wide range of factors including environmental variables such as weather. Weather conditions can affect birds’ vital rates directly through increased mortality due to impaired conditions, or indirectly via changes in body condition and/or behaviour. Most understanding of direct and indirect effects of weather comes from studies of breeding birds, whereas the situation in non-breeding periods is less clear. Here, we analysed annual survival of non-breeding Mallard Anas platyrhynchos, the most hunted waterfowl species in Europe, and assessed whether survival is related directly to winter harshness and/or indirectly via changes in winter recovery distributions. Recovery data on Mallards, initially marked in southeast Sweden, were analysed with an information-theoretic approach using program mark. Over 10 000 Mallards were marked in two time periods, 1964–1982 and 2002–2008, of which 13.3 and 4.7%, respectively, were later recovered. Mallards had lower annual survival in the early trapping period (0.58–0.63) than in the later period (0.69–0.71), with no clear effects of sex, age or year. Within each study period, winter harshness did not directly correlate with survival. However, milder winters may have contributed indirectly to higher survival in the second period, as winter harshness data were correlated with the distances to recovery positions for females, and also because winter recovery areas have shifted northeast during the past decades, possibly indicating a shortened migratory distance. Migration is costly, and there is therefore a likely linkage between migration behaviour and survival of dabbling ducks, in which direct as well as indirect effects of winter harshness may play a role. Other factors, such as hunting pressure, are also likely to have changed in the past decades, and may also have contributed to improved survival of Mallards in northwest Europe.