Stubborn States refers to the fact that it is far more difficult to change the constitution of a substantial workforce by any quality such as a Protected Characteristic, (eg, race or gender) than is generally realised. A good example of this is given by BT’s extreme example for their Non-Openreach workforce to March 2025 – but even modest changes are far more difficult than may be imagined. Let’s look at a simple model.
Say a large investment company has a staff of 30,000, of which 30% are female, evenly distributed by age. Let’s also assume that new entrants all start at age 21, and everyone leaves at age 51 (the work is very stressful). Ignoring mortality, we would have 1,000 staff at each age.
The CEO decides to set a gender diversity target of 50% female staff, to roughly correspond with the UK population. How should he go about achieving this?
The first thing to look at is the NET RECRUITMENTS for each year.
This is simply gross recruits less exits. We need to look at this for both women and men.
The simpler one is exits. The total is 1,000, of which 300 are women and 700 are men.
Let’s assume next year we recruit 500 women, and 500 men, according to the CEO’s assessment. The new recruits are clearly 50% female.
How close to 50% are we after one year? Not very far.
Before, we had 9,000 women (30,000 x 30%), and 21,000 men.
Now we lost 300 women by retirement, and gained 500, so we now have
9,000 – 300 + 500 = 9,200.
We also lost 700 men by retirement, and gained 500, so we now have
21,000 – 700 + 500 = 20,800 men.
Total still 30,000. Percentage women = 9,200/30,000 = 30.667%
An increase of just 2/3 of 1%! But we need to change by 20% (50% – 30%) to hit target.
So, as we change by 2% after three years, we need ten times that to hit our target, ie
10x 2% = 20% – and 10 x three years = 30 years.
So to get to 50% women would take 30 years!
Try telling the Chief Exec that! (Especially as the turnover rate for FTSE100 Chief Execs is about 30%! He or she would want to be there to take credit for the results!)
As our exit rates are fixed, we can only change our recruits. Let’s go all out, and recruit 100% women and no men. How long would that take then?
Because of the way the maths works, the answer is not 30/2 = 15 years, as you might expect, but about 8 ½ years – that’s because of something called gearing, which I won’t go into here.
However, you now hit legal problems – it’s difficult to claim your recruitment is meritocratic if you have 100% female recruitment – see the sections on the “Curve of DEIth”.
If you are recruiting from top graduates, my assessment is that the average female recruit material in this class would be 58%, and I don’t think you could get away with more than 10% above this. Using 68% female recruits gets you there in a bit under 16 years.
No plan I have seen for changing the gender makeup of a workforce is longer than ten years… and there’s only one that long.
Now, you may be thinking “That model is oversimplified – our firm isn’t like that”. Let’s see what arguments here might stand up – or not. There are four.
First – your recruitment may be at an older age than 21 – well, that won’t help you a lot, especially as then the retirement age is likely to be a similar period later (or they won’t earn a decent pension over their working life).
Second – exits. “People leave, and are replaced”. OK. If they are replaced by someone of the same age and gender – no difference. However, the combination of “not many leavers” (most firms are unhappy with more than 10%), and the likely similarity of age and gender of the replacement rather stymie that argument. Again, the Curve of DEIth effect stops an “all women” replacement policy.
Third – “We are not static, but growing (or even shrinking)”. If the latter, your recruits are likely to come down – making targets even harder to reach. But even if we have 20% annual growth in recruits – quite a challenge – then the time taken to hit our target falls – but only from just under 16 years to just under nine years (using the 68% female recruitment model).
Fourth – the “Tie Breaker” argument. This says that if two candidates are meritwise identical, you can pick the black/gay/female one alone, based on that characteristic. That will only work very rarely, unless it’s a small group. Imagine you are picking 100 recruits from 1,000 candidates. Then if, say, the 51st and 52nd candidates are the same meritwise, tiebreaks are irrelevant – they both get offers. If numbers 145 and 146 are the same, ditto – neither gets an offer. It’s only around the 100th one where it might matter.
Finally – many companies – even major ones – have never studied why certain groups are not attracted to working for them – or have relatively high exit rates. Despite this, they still choose to publish extravagant recruitment and change targets and achievements. Instinctively, they see that highly slanted recruitment stats could encourage legal and media attack. They think that by publishing only entire workforce change, this behaviour is hidden. However, Stubborn States technology, as part of the Diversity Maths suite, quickly pulls the veil aside. We can quickly perform 100,000 or more assessments, using company specific statistics, to check the credibility of any claim, as you’ll see when we look at Stubborn States in detail.