How sure are we about the things we think we know?

The latest issue of the British Journal of Sociology has some interesting content, well worth reading. My attention was caught by a piece by Paul Wakeling & Daniel Laurison called 'Are postgraduate qualifications the 'new frontier of social mobility' in which they advance the entirely plausible argument that obtaining more than an u/g degree is an important gateway to occupational success and that social class background has something to do with the likelihood that you will obtain a p/g certificate. This is not an absurd argument and it certainly fits the anecdotal evidence I have to hand. It's good that somebody has tried to look at this in a more systematic way.

Being a Devil in the detail sort of person though I was struck by one of their tables (Table 2 for those that can get behind the paywall). This gives, among other things, for both sexes and for  10 year birth cohorts the percentage obtaining at least an undergraduate degree. Naturally I was interested in my own birth cohort, which in 2014 - the year the data pertain to - was the 53 to 62 year olds. 

According to Wakeling & Laurison in that cohort 22% of  UK women & 25% of UK men had at least an u/g degree (the standard error for both figures is less than 1%). That brought me up sharp. Can this really be true? It certainly doesn't chime with my subjective experience. A back of the fag packet calculation suggested to me that in 1979 about 6% of my own school cohort transitioned at age 18 into some form of higher education (university, polytechnic or degree awarding college of higher education). Now one should be very wary of generalizing from one's personal experience and my secondary school was not noted for its academic prowess, but could my experience have been so atypical?

That sent me off in search of some facts. Now the cohort aged 53-62 in 2014 were born  between 1952 and 1961 and would have first had the opportunity to enter university (at least an English and Welsh university) between 1970 and 1979. Single year cohort specific participation rates are difficult to come by and for various reasons are subject to a good bit of approximation. The best source I could quickly find was the 1997 Dearing Report which, for good or ill, set the course on which UK HE has since been travelling. If there is an authoritative source this is it. Dearing in Table 1.1 gives the following HE participation rates for 'the standard 18 year old cohort': 1970  8% and 1980 12%.  Let's assume these numbers are broadly accurate. How can we reconcile them with the numbers that Wakeling & Laurison report for the same cohorts in 2014? The LFS numbers they rely on are roughly  twice as large. This is a truly massive difference.

There are a number of possibilities. Let's start with the least likely. It could be that net migration massively favours graduates over non graduates. Perhaps we import smart people and export the less educationally accomplished.  There could be something in this but it is unlikely to account for a difference of this order of magnitude.

Much more likely is that a proportion of the 18 year old cohort enters higher education later on in life. They take Open University courses, get sponsored by their employer or return to education after having children. This is plausible, in fact it is more or less certain that the Dearing numbers must be a lower bound.  But can it be true that something like this doubled the proportion of graduates in a birth cohort? If it did then this is a truly big story and we should be hearing much more about the success of the UK's 'alternative routes'. But again I'm sceptical. Fantastic as the OU and access courses are I just don't believe they doubled the proportion of a birth cohort with a university degree.

So have Wakeling & Laurison got it wrong? No, or not entirely. However their numbers are, I think, not quite what they seem. It's easy to go to the 2014 Labour Force Surveys (their source) and look at the numbers oneself.  The LFS contains a number of sources of information on 'university degrees'.  For example the variable  QUAL_1 tells us whether a respondent has a 'Degree level qualification including foundation degrees, graduate membership of a professional institute, PGCE, or higher'. If we look at the cohort specific rates of obtaining this level of qualification it matches pretty closely Wakeling & Laurison's numbers.

QUAL_1 is a very generous definition of what counts as an undergraduate degree. Foundation level degrees are emphatically not Bachelor levels degrees. They are, by design vocational and offered by all sorts of providers including McDonalds (I'm not joking). The requirements for graduate membership of a professional institute are  rather flexible and it is entirely possible for someone to obtain such an honour without ever darkening the doors of an higher education institution. 

It also worries me that this indicator is likely to be misleading when applied to a historical sequence of birth cohorts. When I left school in 1979  common destinations were nursing training or a non degree level training course for primary school teaching. You didn't need a degree to enter either of these professions. Now things have changed and nursing and teaching are all graduate  professions. I wonder how the LFS deals with this? Is a nursing qualification acquired in 1981 retrospectively awarded degree level status?

Fortunately there are other indicators in the LFS that more reliably establish whether someone has obtained a genuine u/g degree. DEGCLS7 for instance records the class of degree awarded and on the assumption that 'Does not apply' indicates that the respondent doesn't have an u/g degree this gives an estimate of 18% for the 1961 birth cohort. This still looks high to me, but it moves the number in the right direction. Other indicators give roughly similar numbers.

That then leaves us with the et alors? It's possible that overestimating the proportion with an  u/g doesn't do much damage to Wakeling & Laurison's argument. It's also possible that it muddies the interpretation of the social class background coefficients.  Part of the argument relies on conditioning on holding an undergraduate degree and then showing that there are still differences in the likelihood of acquiring a postgraduate qualification or a higher social class position  by social class background. If the selected subset of apparent u/g  holders actually contains a high proportion of people who don't have a genuine u/g degree and these latter are more likely to be people of working class origin then the disadvantage this reveals is not about transiting from an u/g to a p/g degree but about obtaining entry to a university in the first place.

As I said at the beginning. The Devil is in the details.