Demographics of Destruction — A bonus analysis

Just one extra piece of data analysis following my post on fixing the ARC Discovery Projects scheme. This little chunk ended up on the cutting room floor last night as I couldn’t fully make sense of it. But after 5 hours broken sleep, and some drawing on the shower window with my finger, I think I can explain it.

The analysis is the two dashed linear fits to a sub-set of the ARC’s data shown below.

ARC Tampered

The fits are to % proportion of all CIs in the 10-25 yrs post-PhD bands for male and female, and while I hate fitting a line to three data points the trend is unmistakable. Let’s try to unpack it a bit.

The rise from 0-5 yrs to 5-10 yrs makes sense — this is the next generation coming through into the fellowship stage — and it will be a large demographic fraction due to the ARC’s (worthwhile) recent focus on ECR support and our perverse use of Ph.D. students as a cheap labour force (for another post). This would then make the peak at 10-15 yrs and subsequent drop off attrition by the ‘game of musical chairs’ that happens first in the transition from DECRA to FT and then from DECRA/FT to tenured junior hire. Going forward, I predict this peak at 10 yrs post-PhD to shoot upwards, with the drop-off becoming shorter and sharper. This will essentially be the ‘Superdoc’ effect recently highlighted in Nature.

What is unusual is that this attrition doesn’t continue right through the dataset — if we’re serious about competition in science, shouldn’t we distill and distill so there’s only a few left at the end? Where are all these 25+ year applications coming from? How real is this thing we see in that graph?

Part of the reason I didn’t include this as an ‘appendix’ to the earlier post, is that I now need to start making assumptions to cover missing data — that earlier post is pure data analysis with no assumptions. The key here is to think about age rather than years post-PhD. Now I’m going to assume Ph.D. completion between 25 and 30, I know people will launch an attack on me about mature Ph.Ds, but if you work inside the system, you know those are typically down at <10% level, so bear with me. If you do this…

Table

…you get a column B like the above. What’s going on here is that 25+ years post-PhD ends up being 50+ age bracket, which is demographically broader than the other bands. We really want to compare apples with apples, so in Rows 10-14 I speculate about what that upper cohort probably really looks like. Retirements should kick in strongly from Row 11, and it’s consistent with many years of just ‘looking’ at the ARC outcomes list. Note that I’ve combined gender here, and have taken a gender-weighted success rate per cohort in order to get accurate numbers.

Let’s get back to graphs…

Model 1

Now that we’ve ‘unpacked’ the 25+ year cohort a bit things look more sensible. The green dashed line is ‘ramp-up’ from ECR programs, the red dashed line is a sensible trend for academic attrition due to the game of musical chairs and people finding other things to do. There’s only one place where the data doesn’t fit the trend and it’s in the 45-60 age bracket — I’ve highlighted it with a yellow triangle and will call it the Matthew zone. If you change the distributions in Rows 10-14 this effect doesn’t vanish, it just reshapes slightly (you need a lot of very old scientists getting grants to make it go away).

The glut of late career scientists is obvious, as is their disproportionately large access to available scientific resources (since that all starts with cash). Note once again, I’m purely using CI in any position statistics here and not lead CI or sole CI statistics. As discussed in my last post, this will only exacerbate massively what we’re seeing in the data I’m presenting.

Another way to see this is:

Model 2

where the blue dashed line is retirement attrition and the pink triangle is what I often call the ‘no-future fellowship’ or the ‘valley of the shadows’.

Probably not a lot more to say here unless the ARC is willing to release some sole CI and lead CI statistics so we can know the full story. I don’t know we’ll ever see that happen.

Otherwise, here’s yet more data pointing to ‘the scientific recession we will have to have’ in Australia (to quote Paul Keating), because the next generation are currently being starved at mid-career at the expense of the scientists near the end of their career.

Advertisements

2 thoughts on “Demographics of Destruction — A bonus analysis

  1. Pingback: Fixing ARC Discovery Projects | Fear and Loathing in Academia

  2. Pingback: Why it’s imperative for EMCRs to reframe ‘merit’ and ‘opportunity’. #SciPath16 | Fear and Loathing in Academia

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s