How Reliable are Register Data for Studies of the Return on Schooling? An examination of Swedish data.

Some important studies have recently used Statistics Sweden's register data on educational levels and annual earnings to estimate the economic return on schooling. This is a powerful way of getting large samples and of adding information to data sets with other useful information. By using detailed information in the Level of Living Surveys, we investigate two potential flaws with this approach. First, the register data on education do not pick up some partial studies at high school and college which do not lead to formal degrees. Our results show that such studies have a significant positive economic return. Nonetheless, the use of this information as independent variables in regression models does not change the original estimates. So omission of partial studies does not flaw the results based on registers only. Second, also using the Level of Living Surveys, we compare estimates based on annual earnings as the outcome measure with those using hourly earnings. We find a marked difference, particularly for higher education levels. However, by eliminating observations with low earnings one gets results for annual earnings that are similar to those for hourly earnings. [ABSTRACT FROM AUTHOR]