Kevin is currently professor of applied statistics at the Open University and his report, The Future of Academic Statistics, looked in detail at various aspects in the area. This included current research funding, the organisation and make-up of statistics departments and an analysis of student numbers and masters courses available in the subject. He also looked at the state of the job market for students graduating with statistical qualifications.
To follow up on his report, we asked Kevin to comment on the issues he raised and where the problems he identified stand today.
In your report you looked at how funding from research councils is a complicated mix of core and applied statistics. Going forward, how do you think this should be monitored more closely and are there any emerging pressures that may tip the balance in the wrong direction?
I think we (as the RSS and the statistical community generally) have to take on board that any such monitoring needs to be done by us – the research councils have their own priorities, statistics does fit into their priorities but in complicated ways that differ across the research councils. So, if anyone is going to spot issues over the balance between different areas of statistical research, it will be us statisticians, not the research councils. There are ways and means of bringing influence to bear on individual research councils, but to do this effectively requires evidence. Currently I’d say we aren’t doing as good a job as we might in monitoring where research funding is going, but I know from the work I did for my report that it is time consuming and complicated to try to obtain a clear picture, over all the research councils, of where the money is going, and without that it’s very difficult to say if areas are being neglected (or indeed over-funded, though I never heard anyone complain about that in their own area of research).
What counts as ‘tipping the balance in the wrong direction’ in this context is not particularly clear either. Most academic statisticians would support the need for appropriate research on methodology, on generally applicable ideas and on some things that do not have an absolutely clear immediate application. Most academic statisticians would also point out the importance of statistical involvement in applied research in the natural and social sciences, medicine, and many other disciplines. But I’m not sure that we have a clear consensus on the appropriate balance between these aspects, as long as it doesn’t approach 100% in one direction and 0% in the other.
Personally my main concern would be with a change in the balance against statistical thinking and research, and in favour of other (qualitative or quantitative) approaches to knowledge generation, rather than with a change in the balance between ‘purer’ and ‘more applied’ statistical research. Statisticians seem always to have been pretty inventive in working out how to get their research funded from available sources, as long as there remains a reasonably clear understanding of the value of statistical methods and statistical thought – it’s these last values that we have to continue to promote.
You also discussed the level of statistics masters courses available in the report, is the number of these courses rising or falling and how are they evolving?
I didn’t find any clear evidence in my report that the number of masters courses in statistics was declining – indeed there seemed to have been a considerable increase in the numbers of courses in the years before the research on which my report was based (that is, up to 2011). It’s true that a few well-known courses had disappeared, but overall there had been an increase. There was, however, evidence of a decline in numbers of students undertaking taught postgraduate studies in statistics, particularly those domiciled in the UK, and although I haven’t been able to look at the data since those for the 2010-11 academic year, it would appear that the decline has continued, to some extent at least.
The current position still seems to be that, if a student has the resources to do a masters course without needing funding, there is a wide choice of courses available. The availability of funding is a very big if, though. But even when I wrote my report, several MSc courses were managing to attract good numbers of good students with little or no funding on offer.
In some statistics-related areas at least, new courses are still appearing rapidly. As an example, there are MSc courses in Data Science, Data Analytics, Big Data, and variations on those titles at universities thoughout the country. One has to bear in mind that most of these courses are run by computer science departments, and some may not have much or any involvement from statisticians or include much of what we would recognise as statistics. But some of them certainly have substantial statistics content, and some of these new courses have considerable amounts of student funding available.
Overall, it looks to me as if the availability of statistics-related masters courses is changing rapidly, rather than declining.
Do you believe that academia is attracting a good level of quality and diverse statisticians in the face of competition from better paid sectors?
The data I have, in relation to this particular question largely consist of no more than anecdote, so I’ll have to be circumspect. Colleagues from some universities tell me that they are having considerable difficulty in attracting good applicants to academic posts in statistics, particularly those at higher levels. However, other colleagues from other universities tell me that this is far from the case with them. I can’t spot a very clear pattern in which places say they are doing well for applicants, and which aren’t.
It seems clear that there is considerable competition from the financial sector, and perhaps increasingly so from data science/Big Data. I’d say the picture is mixed. But statistics is far from being the only academic subject where there is major competition from better-paid areas outside academia. Perhaps, if we’re having difficulty recruiting, we could share ideas with some of those other subjects. And there are still reasons to work in academia that go beyond the level of pay, in our subject and in others.
How do you think the whole academic sector can go about measuring the introduction of quantitative methods in a standardised way given that it traverses so many departments?
There’s a view that all students in higher education, no matter what their main subject is, should be studying some appropriate quantitative methods – and it isn’t only statisticians who put this view forward. But measuring progress is really challenging. I don’t think there’s even a clear definition of quantitative methods that runs across all subjects, and in any case, quantitative methods that make sense within, say, a degree in linguistics are not going to be exactly the same quantitative methods that would be appropriate in a degree in sports psychology.
I think that something fairly crude, but with a bit of tailoring to different subjects, could sensibly be done – perhaps defining some basic and some more advanced statistical skills within each of a set of broad subject groups, and recording (on some sample basis) to what extent these are taught to undergraduates and how that changes over time. But subject curricula change over time, and indeed rather more rapidly in many subjects than in the mathematical sciences, so even the definitions of what is basic and what is advanced might have to be updated in a way that will make the interpretation challenging.
With all the excitement currently surrounding big data, do you think the UK is educating enough skilled statisticians/data analysts to meet this future need?
This is very difficult to predict. We know for instance, from the major changes over time in the numbers of computing students in the UK and the US, that the availability of people with appropriate skills at the graduate level can change fairly quickly in response to actual or perceived changes in demand. But that that doesn’t necessarily deal with the availability of more senior staff, whose skills take longer to develop, or those with more advanced skills or experience.
There is also an issue that applies to mathematical subjects, including statistics, to a greater extent than many other fields – for most students, what’s studied at university level depends pretty heavily on mathematical and quantitative skills acquired at school level, and so potential blockages in the quantitative pipeline at school level can cause real difficulties. My gut feeling is that we can educate enough statisticians and data analysts, but only if we keep pressing the importance of this to those who have the resources. That would include government, but also, crucially, those who hold the purse strings within universities. That links back to the previous question – who is going to persuade our Vice Chancellors and Deans that we need to teach more quantitative methods and statistics?