Exploring Gender Variability and Performances at the Extreme Tails of the Distribution in Reading

• In Event: Perspectives on PISA (Programme for International Student Assessment) Reading

Abstract

Purpose of the study
Gender differences in achievement are high-stakes issues since they may be associated with reduced career opportunities. Mean gender difference is one of the basic indicators published in every national or international survey on student achievement. In PISA reading, for example, the mean performance of girls has been significantly higher than that of boys in every administration since 2000: on average across OECD countries, and in every individual participating country, with two exceptions. However, in terms of educational policy, it is also vital to know if the extreme low or high tail of the score distribution shows a disadvantage for one gender or the other, even when there is little or no gender difference at the mean. It may help policy-makers to design well-targeted compensatory policies for low-skilled people and to set up policies aimed at increasing the number of highly skilled people. However, research on gender differences in variability and at the extreme tails of the distribution is quite rare (Nowel and Hedges, 1998). This study examined gender differences in PISA reading focusing on the variance and on the extremes of the distribution. Gender differences according to the reading medium (print versus digital) and to the reading aspects assessed will be presented, in addition to gender differences in the combined scale of reading.
Data and method
The main database was from the OECD-PISA 2009 assessment. All 67 participating countries were included. The OECD-PISA 2000 and the IEA-PIRLS 2006 databases were also used to make comparisons with some of the PISA 2009 results. Three types of statistics were computed. (1) The difference between males and females standard deviations.. (2) Difference in gender variability. (3) Effect sizes (i.e. the male mean minus the female mean divided by the pooled standard deviation) for the means and for percentiles 5, 10, 90 and 95 .
Results
In the vast majority of the participating countries, males variance was higher than female variance (measured either by the variance ratio, either by the gender difference in standard deviations). Moreover, males’ variability increased between PISA 2000 and PISA 2009. Gender variability seems also to increase with the age (15 year-old PISA students compared with PIRLS grade 4 students).
According to Cohen’s categorisation (1977), effects sizes for the mean gender differences are small but not negligible (0.45 on the combined reading scale). When examining the extreme tails of the distribution, effects size at the lower tail of the distribution are about twice as large as at the upper tail. This result suggests that the relative underachievement of boys compared to girls in reading is far more of an issue for the low achievers in reading.

Author

• Ariane Baye, Université de Liège