A Comparison of Two Ways of Measuring Level of Education in Norway: Questions in Survey Interviews versus Register Data
English title: A Comparison of Two Ways of Measuring Level of Education in Norway: Questions in Survey Interviews versus Register Data
Author(s): Ø. Kleven - K. Ringdal -
Type: Conference paper/poster
It is well known that there are discrepancies between information in interview about individual behaviour and the actual behaviour. In some instances responses in interviews may be checked against register data: driving license, library card, and the act of voting in national elections. The latter has been examined in Norway and the evidence indicates that data from the interviews results in higher estimates of turnout than actual the turnout. In several Norwegian surveys we have experienced that the distribution in level of education based on the survey shows a larger proportion of the highest levels compared to official statistics on education. In this study we will compare measures of the level of education in Norway based on interview questions with aggregated and individual level estimates based on register data. We will estimate the amount of discrepancies in the two ways of measuring level of education and also try to explain the differences. We will test three hypotheses on why such differences may be observed. The non-response hypothesis implies that there is a positive correlation between levels of education and the willingness to participate in surveys. If this is so, the marginal distribution of levels of education obtained from the interviews will be more positively skewed than the one based on register data. The artefact hypothesis implies that the observed discrepancies are mainly due to artefacts, especially errors in the registers. The social desirability hypothesis simply states that social desirable activities or characteristics will be over-reported, and undesirable ones will be underreported. This will create systematic measurement error. In modern society high education is generally regarded as desirable, therefore some exaggeration of the actual level of education will result from survey interviews. Furthermore, since we expect rare characteristics to be less often over-reported than common ones, we expect the effect of social desirability on over-reporting university level education will to be stronger for younger than for old people. On the other hand, memory problems will be positively correlated with age. This may result in more random errors for older than younger respondents.
Conference name: European Conference on Quality in Survey Statistics
Start date: Apr 24, 2006