Sixty years of the same people, followed through their whole working lives.
Three nationally representative UK datasets, more than 190,000 observations, three different methods, and one finding that holds across all of them.
001
Their childhood did not happen after their career
( Why this data is different )
versus snapshot
Most workforce research is a snapshot. It surveys a group once and finds correlations, and a correlation in a snapshot cannot tell you which thing came first, so it cannot tell you what to do.
These three studies follow the same individuals across decades. One begins with babies born in a single week of 1958 and tracks them to age 55. When you know what a household looked like before someone could read, then watch what happens to their salary at 23, 33, 42 and 55, the direction of the effect is not in question.
That is what makes the findings usable in a boardroom. They are not associations that might run the other way.
002
What background is worth, per month
( British Household Panel Survey )
18 waves
What it asked.If you hold education, age, health, marital status and wealth constant, does the background someone came from still move their pay? It should not, if pay tracks contribution.
What it found.It does, and by a lot.
Just under twenty thousand pounds a year, between people the model has already made equivalent on education, age and health.
The part that matters commercially.The effect is not spread evenly. Separating above average from below average earners, background mattered three to four times more at the top of the distribution than across the sample as a whole. Whatever background is doing, it does most of it in the pay bands where your largest salaries sit, which is also where discretion is widest and documentation is thinnest.
When it bites.The effect peaks between thirty and forty, the years in which people are selected for the roles that determine the rest of their career. It remains significant throughout working life.
003
One week of births, thirty-two years of pay
( National Child Development Study )
ages 23 to 55
What it asked.Follow one cohort from birth to late career. Does the gap between them narrow as their own choices accumulate, or widen?
What it found.It widens, and keeps widening until people are in their late forties.
Where the effect comes from.By adding controls in sequence the study separated the channels. Roughly a fifth runs through childhood environment and early skill formation. A small share runs through visible discrimination. Around an eighth runs through physical characteristics employers respond to. The remainder, close to two thirds, runs through networks and cultural capital: knowing how the institution works, knowing who to ask, and being comfortable in rooms other people find intimidating.
That decomposition is directly actionable. It says sponsorship and structured access to senior people will move more than bias training will, and it says how much more.
People from advantaged backgrounds were more than twenty per cent less likely to be made redundant.
Background does not only raise pay. It buys protection when the organisation contracts, which means your last restructure probably moved your equity position backwards without anybody intending it.
Stated precisely.Within a model explaining 57 per cent of variance in lifetime income, background remains a significant independent predictor. Moving from the tenth to the ninetieth percentile of background corresponds to a 25 to 31 per cent difference in income, falling to around 10 per cent once measured ability is controlled for. We state it this way because the shorter version, that background explains 57 per cent of income, is a different and false claim.
004
The modern picture, and the number that sells
( Understanding Society, 2009 to 2024 )
40,847 individuals
What it asked.Take the largest contemporary UK household panel and decompose the effect properly. How much runs through education, and how much does not?
Finding one, the headline number
Between a typical employee from a professional background and a typical employee from a working-class background. Not the extremes, the ordinary middle of your workforce. Roughly a quarter of salary, sitting inside a pay structure everyone involved believes is meritocratic.
Six independent estimation methods were run against the same data and all six agreed on direction and magnitude. We lead with the interquartile figure rather than the full range comparison, because the full range compares two people who barely exist in any real workforce.
Finding two, the one that changes strategy
Degree attainment is almost identical across background bands. In the raw data, 50.9 per cent of the working-class band held degrees against 51.3 per cent of the professional band. The credential is not the mechanism.
The strategic implication is uncomfortable and worth stating plainly. If your entire equity programme is aimed at access to education and qualifications, and most are, it is aimed at four per cent of the problem.
Finding three, the glass floor
Background protects against low pay roughly twice as strongly as it accelerates high pay. It is a floor before it is a ladder.
For an employer this reframes the whole question. The cost is not only that capable people from working-class backgrounds are underpaid. It is that underperformance among people from advantaged backgrounds is being quietly absorbed, at senior salary levels, by a mechanism nobody designed and nobody can currently see.
Finding four, it compounds with tenure
Each additional year of age adds to the premium. The gap roughly doubles across a working life, which means your most senior population is your least equal population, and retention is compounding the problem rather than solving it.
Finding five, the same degree buys different odds
Identical credentials, 1.87 times the chance. If your organisation recruits on degree class and treats that as a level playing field, this is the number that says otherwise. Put differently, background substitutes for roughly twelve years of education in the outcomes it produces.
005
What this means for a firm of two thousand people
( Translation )
magnitude
The research is population level. The following is arithmetic rather than a client result, and your own figure will differ. It is here because the order of magnitude is usually the thing that gets the meeting.
| Employees below the workforce mean on background | Typically around half |
| Typical premium between quartiles | 23.3% |
| Effect on redundancy exposure at equal performance | 20%+ |
| Effect on reaching the top quartile at equal credentials | 1.87× |
| Direction of travel with tenure | Widening |
Against a mid sized professional services payroll, a gap of this shape is normally a seven figure annual exposure before anybody has argued about fairness. That is the number a finance director engages with, and it is what a readiness assessment estimates first.
How the exposure is quantified→006
What this evidence does not establish
( Limits )
Every vendor publishes findings. Fewer publish the boundaries around them.
Association, not individual causation
The models identify systematic patterns across a population. They do not establish that any particular pay decision was caused by any particular person's background, and no PRISM output should be described that way.
Estimators can disagree
The panel study finds the effect strongest among higher earners. The contemporary study finds the protective floor strongest at the bottom. These are different methods answering slightly different questions, and we present both rather than choosing the more convenient one.
Self report has limits
Background data is reported by the person it belongs to, and recall is imperfect, particularly for older respondents.
Benchmarks are indicative
Population benchmarks come from nationally representative samples. Your sector, region and occupational mix differ from the national picture, so comparison is a guide rather than a verdict.
Leavers are invisible
Findings describe the workforce present at the point of measurement. People who already left are not observed unless leaver data is inside the scope.