Appendix A

View the most recent version.

Education definitions in the IALS 1994, IALSS 2003, and LFS

In this appendix, we examine the different questions concerning educational attainment in the IALS 1994 and the IALSS 2003, comparing them to responses in the Labour Force Survey (LFS) in 1994 and 2003.

The IALS 1994 and IALSS 2003 have similar but not identical education questions. Both ask the respondent about the number of years they have gone to school, with zero years being an option. The IALSS 2003 then has a question asking whether the respondent completed high school, whether directly or by a high school equivalence test. IALS 1994 does not include this latter question. Both surveys include the question, "What is the highest level of schooling that you have ever completed?" However, they differ in the possible categories of response. The IALS 1994 has more possible categories at the sub-high school completion level but IALSS 2003 has more categories between the high school and university levels (5 compared to 1 in the IALS 1994). The latter difference raises the possibility of differences in how people who obtain trade certificates and apprenticeships are treated. These categories are explicitly listed in the IALSS 2003 question, giving even high school drop outs who went on to get other certificates the chance to record a higher level of education. With the shorter list in the IALS 1994, where the relevant category is the very general sounding "completed non-university Post-Secondary, one might imagine that respondents might neglect to mention apprenticeships, for example.

We have experimented with several education category definitions using the IALSS 2003 questions. We focus our attention on trying to match the IALS 1994 with the IALSS 2003 rather than the other way around because the education questions and categories are more extensive in the IALSS 2003. The choice of our categorization has little impact on the results we present in the body of the paper. We settled on a categorization in which we classify individuals who respond that they have not completed high school (to the direct question on that point) and state that their highest level of schooling is either less than high school, high school diploma, or some post-secondary education that did not result in a certificate or degree as having less than high school education. Our assumption in this is that these non-high school graduates would not have responded that their highest level of schooling was "completed non-university Post-Secondary" in the 1994 questionnaire. We classify individuals who give the same responses to the highest level completed question but state they have graduated from high school in response to the earlier question as high school graduates. Our college category includes all categories in the highest level of schooling question that are above the not-completed post-secondary category but below the bachelor's degree category. We gather all university degrees at the BA level and above together in the university category.

The resulting categorization yields the following comparison between the two datasets, where each entry in the IALS 1994 and IALSS 2003 columns shows the proportion in the corresponding category in the given year.

The results show a strong decline in the lowest education category and strong increases in the two post-secondary categories.

As a point of comparison, we also formed the proportion in corresponding categories from the Labour Force Survey for 1994 and 2003 as follows:

The broad patterns are similar in terms of the declines in the proportion who have less than a high school degree and the increases in the top two categories. The proportions in the university categories are also very similar across the two types of datasets in each year. However, the LFS has much higher proportions in the some post-secondary category, particularly in the first period, and correspondingly lower proportions in the bottom two categories. From this, we draw two conclusions. First, it is not clear which type of dataset (the literacy surveys or the LFS) should be viewed as more reliable. It is worth noting, in particular, that the LFS includes responses about the target individual's education by other household members if the target individual cannot be contacted. It is not known how many observations are affected by this "proxy response" and the potential measurement errors it introduces. In the literacy surveys, on the other hand proxy response is not an issue since the target individual must necessarily be contacted directly in order for them to take the test. On the other hand, this may itself raise issues of representativeness relative to the LFS since the LFS at least has some information on people who are hard to contact directly. Second, the broad trends are similar across datasets and thus the IALS datasets seem likely to provide an accurate picture of the impact of changes in the educational composition of the population over time.