A bigger problem than the variability of marking is that universities are almost wholly ignorant of its extent and sources. I find as an external examiner in many universities that there is little analysis of the mark profiles across modules, even on the same course, and these can, and usually do, vary enormously.
I construct my own frequency distributions and do chi-square tests and usually never fail to show that the mark distributions are highly significantly different. There can be reasons for this, but a basic issue is that typically no one knows about it. All that is ever examined, if anything, is the distribution of degree classes.
Much of the more extreme variation would be controlled if results were analysed and if more consistent assessment practices were adopted across modules (these can vary enormously, too).
Universities are accused of massaging degree classes upwards, but my experience of attending exam boards is that most universities have not got a clue what the distribution of classes is. But who am I to talk, I'm only a statistician.
Eamonn Judge, Leeds Metropolitan University.