Earnings inequality among men and the returns to higher education have increased substantially in most OECD countries since the 1980s (Gottschalk and Smeeding, 1997). A number of explanations for this development have been proposed, but no consensus has yet been established. For example, international trade and globalisation, decline in minimum earnings and in unionization, and increased use of performance pay have been suggested as possible explanations. A number of papers argue that the return to (unobserved) skills is increasing because of a growth in the demand for skills caused by skill-biased technological change (see e.g. Juhn, Murphy and Pierce, 1993). However, the fact that many European countries and the US have experienced rather different trends in earnings inequality is often argued to undermine this explanation since these countries should have experienced similar technological changes (Gottschalk and Smeeding, 1997).Two important stylized facts for the US are that since the beginning of the 1980s and continuing through the 1990s the measured return to additional years of schooling has increased, and this has been accompanied by a rise in the dispersion of earnings within groups with the same years of schooling (see e.g. Card and Lemieux, 2001 and Katz and Autor, 1999). An important part of the change in earnings inequality in the US has been linked to the lasting growth in the college/high school earnings premium, but this does not explain the increased earnings dispersion among workers with the same educational level. The basic assumption underlying much of the research that attempts to explain these trends is that the number of years of schooling an individual receives consistently proxies for skill. However, years of schooling is a biased measure of skill if the composition of individuals with the same years of schooling differs along an unobserved dimension determining their earnings. By the same token, changes in this unobserved dimension may affect the observed trends in the returns to schooling as well as the within-group dispersion of earnings. The use of years of education as a proxy for skills disregards on the one hand the large variation in the returns to fields of study over time and across countries, and on the other hand, the substantial variation in the composition of fields of study over time and across countries. For example, cross-country differences in the composition of fields of study over time might explain cross-country differences in the trends in inequality and return to education. Further, the influence of skill biased technological change on choice of field of study may in itself vary between countries. In particular, higher tuition fees and more dispersed earnings might cause larger shifts towards fields with high returns in some countries than in others, and thus bigger increases in the returns to college and in earnings inequality. This may explain why we observe different trends in earnings inequality and return to education across countries which experienced similar technological changes. The objective of this study is to examine to what extent changes in the field of study composition is driving the rise in the return to college and earnings dispersion among college educated men in Norway and the US. To do so, we use a generalized version of the decomposition method introduced by Oaxaca (1973) and Blinder (1973). The method applies counterfactual measures such as “how much would a worker, with the mean characteristics of the Norwegian workforce, have been paid in the US?”, or “what would the variance of earnings have been if returns to age and education were as in 1990, but the composition of the workforce’s age and education was as in 1980?”. These counterfactuals are compared to the actual or other counterfactual measures, to give an estimate of the significance of the field of study compositions to these measures. We start by documenting the trends in earnings inequality in Norway and the US between 1970 and 2001, both among all working men, and within educational groups. Moreover, we describe the time trends in the returns to higher education as well as the compositional changes with respect to field of study and age in the male workforce. Next, we examine the effect of differences in the field of study composition on the mean log earnings among male college graduates - as a measure of the college premium - in Norway between various years from 1970 to 2001. Next, we investigate to what extent cross-country differences between Norway and the US in field of study composition can explain the large differences in college premium between these countries. We also explore the effect of field of study on the variance of log earnings among Norwegian male college graduates from 1970 to 2001. In addition, we explore the effect of changes in the age composition on the mean and variance of log earnings. For example, mean earnings are likely to increase when the workforce grows older, because older workers will tend to have more experience. Earnings dispersion will also increase if the variation in earnings is higher among older than younger men. Changes in the age composition might also affect the estimated field of study effect, if the returns to fields differ by age. Our calculations are produced with Stata, SAS and Excel. We find that differences in field of study composition have little impact on the differences in mean earnings, both across time and countries. Differences in age composition, however, affect the mean earnings remarkably. In particular, 40 percent of the mean earnings increase among Norwegian male college graduates in the 1980s could be accounted for by changes in the age composition. Further, we find that the variance of log earnings among working male college graduates fell drastically in the 1970s and increased in the two succeeding decades. We also show that the driving force behind this development was changes to the residual variance, rather than changes to the composition or return to field of study or age. In the 1970s, changes to the returns to field of study counteracted the decrease, while changes to the age composition seem to have contributed significantly to the decrease. Further, changes to the return to experience counteracted the increase in the 1980s, while changes in the return to field of study contributed significantly to the increase in the variance in the 1990s. Changes to the field of study composition, however, do not seem to be an important factor behind the evolution of the log earnings variance.