Changing the System of Student Support in Norway: Intended and Unintended Effects on Students

Abstract

In 2002 the student finance system in Norway went through a major restructuring. The changes included an increase in student support and an introduction of progression‐dependent grants. Using two student welfare surveys conducted in 1998 and 2005, the paper analyses the effect of the changes on the students. The analysis compares the risk of study delays, the students’ weekly working hours, and the students’ concerns about student loan repayments before and after the changes among different groups of students. Contrary to the intended policy goals, the findings indicate no decline in study delays. However, the findings reveal an increase in the amount of time students spend in paid employment. This may indicate that students respond to short‐term economic incentives. Furthermore, the findings suggest increased social differences in the students’ concern for the student loan repayment after the changes.

Keywords:

• student finance
• education policy
• higher education
• social inequality

Notes

¹ The student grant is reduced by 60% of the income exceeding the income threshold. Thus, a student who earned NOK 110,000 in 2002 would have had the grant reduced by NOK 6,000; while a student who earned more than NOK 154,000 (approximately) would not receive any of the support as grants, only as student loans.

² The logit model is written as: Prob (P) = 1 / (1 + e–z), where z = a + b1×1 + b2×2…bnxn. The general linear regression equation can be written as: Y = a + bx + e (Lewis‐Beck, 1980).

³ Measuring the percent of variance explained in logistic regression analysis is more complicated than in ordinary least squares (OLS) linear regression analyses. There is no widely‐accepted direct analog to OLS regression’s R². Nagelkerke’s R² is the most‐reported of the pseudo R² estimates in logistic regression models. This measure is an adjusted version of the Cox and Snell R², which is based on the log likelihood for the fitted model compared with the log likelihood for the null model (with no predictors). While the Cox and Snell has a maximum value of less than 1, even for a perfect model, the Nagelkerke R² adjusts the scale of the statistic to cover the full range from 0 to 1 (Nagelkerke, 1991).

⁴ Students at private higher education institutions were only included in the 2005 survey.