Executive summary

There is considerable interest among members of the policy and academic communities in measuring academic performance through standardized tests. This is because learning is important to function in everyday life (for example, reading instructions for home electronics, following directions on maps), to be engaged in society (for example, following the news, voting), and for skills acquisition (for example, obtaining educational credentials). Researchers have even linked increased literacy skills to higher productivity (Coulombe and Tremblay 2006). Labour economists are also interested in the learning process. Although it is well established that more education is causally linked to higher earnings (Card 1999), the precise mechanism is not so well understood. Education may act as a signal in the labour market, allowing potential employers to screen in 'good' candidates based on how well they have done in a formal school setting. Whether they have learned anything that is useful for the job refers to the human capital aspect of education. In general, empirical findings can be better explained by signalling models than by human capital theory, although the literature finds evidence that both factors play important roles in wage determination (Weiss 1995).

What determines academic performance? One possible factor is innate abilities that are present at birth. Since it is difficult to confirm this hypothesis without further investigation, no attempts at doing so will be made in this study. A more feasible and, perhaps, more policy relevant goal would be to focus on environmental factors. Broadly speaking, these can be categorized into exposure to schooling and exposure to other aspects of life (for example, being reared by one's parents, spending time with friends, watching television or reading newspapers). If schooling contributes toward academic performance, we should expect standardized test scores to improve with more years of schooling. This would be relatively simple to test if similar assessments were administered to students in different school grades, which is often not the case. Moreover, the interpretation of the results would not be clear, even if similar assessments were administered. As students age by (say) one year, they are exposed to both an additional year of schooling and an additional year of life in general. A lot can happen outside of the school system over the course of one year that can influence learning. Some students may benefit from an additional year of encouragement to succeed from their parents, while others may have to deal with one more year of parental neglect. Some students may benefit from talking with their friends about career aspirations and how to achieve them, while others may be lured by their friends to partake in illicit activities.

The objectives of the study are twofold. First, I will assess the extent to which academic performance improves with an additional year of schooling. In particular, which areas improve the most: reading, mathematics, or science? Second, does an additional year of schooling confer the same academic benefits onto different groups of students? Specifically, do girls and boys benefit equally from more schooling? What about youth from higher and lower income families? The answers to these questions are particularly important, since large gaps in academic performance have been identified by sex and parental income. Moreover, the gaps in academic performance have been linked to large gaps in university attendance.

Top of Page

Identifying the returns to schooling can be problematic, since individuals who have chosen more schooling may do so because they might have higher abilities. Clearly, an exogenous variation in schooling is required for identification. The approach used in this study takes advantage of a setting whereby large samples of students of roughly similar age wrote the same standardized tests, but the students were in different school grades because of school-entry laws, thus creating a sharp discontinuity in school grades. In some cases, students who were one-day apart in age were in adjacent school grades, but wrote the same tests. In other words, one additional year of schooling is associated with as little as one additional day of life in general in this setting.

The Programme for International Student Assessment (PISA) offers us this unique opportunity. Rather than focusing on students in a particular grade, as is commonly done in standardized testing, the PISA tests in reading, mathematics and science were administered to students who were 15 years old on December 31, 1999. The actual assessments were completed in April or May 2000, which means that most students were in Grade 10, although a substantial portion of students were in Grade 9 at the time. In most jurisdictions, being in Grade 9 at the beginning of the year in which students turned 16 is only possible if students failed a grade, or started school a year late. In either case, comparing test scores of Grade 9 students with those of students in Grade 10 would yield little insight, since there is an obvious selection process distinguishing the two groups of students. However, in two Canadian provinces—Nova Scotia and Quebec—school laws for determining entry into the school system are different than those in the rest of the country. While most provinces base school entry on the student's age as of December 31, students in Quebec are allocated to different school grades based on their age as of September 30. The cut-off date for Nova Scotian students is October 1. This means that in these two provinces, students born between January 1 and September 30 (or October 1) would normally be in Grade 10 at the time of the assessment, while those born later in the year would normally be in Grade 9. In other words, students in Nova Scotia and Quebec are allocated to different grades based on potentially very small differences in age, yet they are assessed using the same instrument. I apply a simple regression discontinuity design to examine the differences in reading, mathematics and science scores that may exist around those cut-off dates.

The findings suggest that one additional year of high school (Grade 10) is associated with a large improvement in overall reading and mathematics performance and a smaller improvement in science performance. However, improvements are not equally distributed: mathematics scores improve more for boys than for girls, and reading and science scores improve more for lower than for higher income youth. Most importantly, I find no evidence that girls or higher income youth benefit more from an additional year of high school in any test area.

So, what is behind the gender gap in reading performance and the broader income gap in academic performance? The findings suggest that high school factors may fail to provide much insight. Candidate explanations that cannot be ruled out are those related to earlier school experiences, influences in the home or even factors present at birth. As a result, more work is needed in this area. For example, it would be useful to investigate the role of earlier school years on learning, especially in view of understanding gender differences in academic performance. More detailed data on classroom strategies might be useful in this case. Also, the role of the teacher's gender may be important, as suggested by a recent American study (Dee 2007). It may also be useful to estimate the role of parental resources in shaping the income gap in academic performance.

View the publication The Returns to Schooling on Academic Performance: Evidence from Large Samples Around School Entry Cut-off Dates in PDF format.