Time trends in the effect of socio-economic factors on low birth weight, stillbirth, perinatal deaths and deaths up to the age of one were studied using a medical birth registry linked to census information from 1975, 1980, and 1985. For each census year, delivery outcome the following year was studied. Two socio-economically different groups of women were studied, defined by occupation/education, cohabitation, and citizenship--one privileged and one under-privileged group. A difference in birth weight distribution was found between the two groups which is only partly explainable by different smoking habits in early pregnancy and did not substantially change during the ten year observation period. In 1976, there was virtually no difference in infant mortality between the two groups. In 1981 and 1986, infant mortality had decreased in both groups but more strongly so in the privileged group, and a difference between the groups therefore appeared.