When we examine the changes in performance over time, as can be seen in table A.5.3, there is a statistically significant effect from P-score on the changes in the number of grade 5 departments between 1992 and 1996. The coefficient is 0.17 (t-statistic = 3.30) and the level of statistical significance is 1 percent. However, when we move to the next time period of 1996-2001 in table A.5.4, the P-score has a significant effect only prior to inclusion of the finance variable. This is also the case, in table A.5.5, for years 1992-2001, where P-score is significant at the 5 percent level until income is added into the regression, at which point P-score becomes weaker.
The impact of organizational revenue is likely explained by size. As pointed out above, in the case of Manchester University, one of the largest in the United Kingdom, 80 percent of departments scored somewhere in the 5s in 2001. It is, therefore, to be expected that institutional revenue will be more important in the tables reporting results for all 5 grades where the numbers of RAE submissions are considerably higher although the standards being reached are, on the whole, lower.
In the style of Granger causality, when I run a regression using, as a dependent variable, performance in all grade 5 departments in 1992 with vice chancellors P-scores in 2001, there is a positive gradient but not a statistically significant relationship. The coefficient is 0.035 (t = 0.82).