![Sample our Education journals, sign in here to start your access, latest two full volumes FREE to you for 14 days]({"type":"image" , "src" : "/sda/5932/TOC-Subject-Sample-2021-Education.jpg"})
Abstract
This paper uses regression analysis to test if the universities performing less well according to Shanghai Jiao Tong University’s world universities league tables are able to catch up with the top performers, and to identify national and institutional factors that could affect this catching up process. We have constructed a dataset of 461 universities across 41 countries. We found consistent evidence of a moderate degree of catching up, especially amongst non-US universities. Larger universities as well as universities located in English speaking countries not only perform better on average, but also catch up more over 2003–2009. Universities located in lower income countries are also catching up more. The performance of private universities, as compared to that of public universities, varies substantially between the US and the other countries.
Notes
1. .Billaut, Bouyssou, and Vincke (Citation2009) criticise the ARWU ranking as having irrelevant criteria and faulty aggregation. Brooks (Citation2005) points out, many quantitative indicators used by university rankings have poor link to teaching and research quality and have serious methodological drawbacks.
2. For instance, as one of the major players in the international education market, Australia’s education sector constitutes the country’s third largest export item, contributing A$15.5 billion to its total exports in the year 2008 (Australian Education International, Citation2009).
3. In 2009, France had three top 100 universities but no top 20 universities in the ARWU.
4. The per capita academic performance indicator is the weighted scores of the other five indicators divided by the number of full-time equivalent academic staff. It receives a weighting of 10% in the ARWU. The function of including this indicator is to control for the size effect.
5. The use of logarithmic transformation is to allow for non-linearity in the effect of the explanatory variable. Such transformation typically is not applied to variables measured in percentage or rate because it will easily lead to confusion in interpreting the coefficient.
6. We can also use the number of students to measure university size. However, since we also include student to academic ratio in the model, including student number will be redundant and cause multi-collinearity problem.
7. See Endnote (4).
8. Since Oxbridge represents a very minor (2 out of 461) portion of the entire sample, it is very unlikely that the measurement errors of their faculty sizes will bias the overall results, even if they were truly different from the others in this aspect.
9. It is measured in purchasing power parity, constant 2000 $US terms.
10. It is only proxy measure for two reasons. Firstly, it does not include private expenditure. Secondly, it includes expenditure on primary and secondary education.
11. The 10 countries are: Australia, Canada, Hong Kong, India, Ireland, New Zealand, Singapore, South Africa, the UK and the US.
12. The variables in the dataset are not normally distributed according to skewness and kurtosis test for normality.
13. The difference is significant at the 0.05 level.
14. Founded in 1096, the University of Oxford is the oldest university in our dataset, while the University of Duisburg-Essen in Germany, founded in 2003, is the youngest.
15. Due to the fact that the correlation between the independent variables in general is very low, the estimation results are very similar to those obtained from basic OLS regressions.
16. Since adding more explanatory variables will inevitably improve the explanatory power of the model, the adjusted R-squared statistics are used to account for this by ‘penalizing’ the use of more explanatory variables.
17. The 5% confidence intervals of the two coefficients overlap with each other.
