What the evidence says about mentoring and befriending
What the evidence actually says
Mentoring produces a small but real average benefit for young people, on the order of 0.2 of a standard deviation across the outcomes usually measured, according to meta-analysis of youth mentoring programmes by DuBois and colleagues in 2011.
A coordinator who has to write an outcomes section in a funding bid, and a volunteer or family who wants to know whether any of the claims made for a one-to-one scheme are actually true.
The credibility of the bid. A reviewer who finds one unsupportable claim reads every other claim in the document differently.
A scheme promises a transformation it cannot produce, is funded on that promise, is measured against it, and closes having done real but unremarkable good that nobody counted.
An effect of that size means most mentored young people look much like most unmentored young people, and the difference shows up in the average of a group rather than in the life of any one child. The more useful finding underneath it is that programme practice moves the effect more than mentoring itself does: matching, training, supervision and how long the match lasts account for more variation in results than the presence or absence of a mentor.
The claims made for mentoring, tested one at a time
The claims made for mentoring and befriending fall into four bands of evidential strength, strong, moderate, weak and not established, and a bid is safest when it makes claims only from the top two, strong and moderate. The table separates what is consistently found from what is asserted in sector literature and does not survive a reviewer.
| Claim | What the research supports | Strength |
|---|---|---|
| Mentoring improves academic, behavioural, social and emotional outcomes on average | Yes, but the average effect is small, around 0.2 of a standard deviation | Strong |
| How a programme is run changes results more than mentoring does | Yes. Matching, training, supervision and match length are the active ingredients | Strong |
| Matches that end before three months are associated with outcomes worse than no match at all | Yes, this is the association reported by Grossman and Rhodes in 2002. The study establishes the association rather than the cause, and it is why the sector asks for a year | Strong |
| Befriending reduces loneliness | Modest benefit on loneliness measures | Moderate |
| Befriending reduces social isolation | Not shown. A scheme can move how a person feels without changing how many contacts they have | Weak |
| Befriending reduces depression or use of health services | Weaker and less consistent than the loneliness finding | Weak |
| Mentoring cuts reoffending or raises attainment by a stated amount | No usable figure. This site puts no number on it | Not established |
| Every pound spent returns a stated sum of social value | No. Figures of this kind circulate in sector literature and none withstands scrutiny | Not established |
| Volunteers benefit from mentoring too | Reported consistently, but almost entirely by volunteers who stayed | Weak |
A standard deviation is a measure of spread, not a percentage. Converting 0.2 of one into a claim such as “improves attainment by a fifth” is the single most common misuse of this literature and a reviewer will catch it.
Why a small effect is still worth funding
A small effect is worth funding for three reasons, none of which requires overstating it. First, the cost is bounded and can be stated in advance, at a working range of £700 to £2,500 per active match per year; where a scheme sits in that band depends on how complex its referrals are, whether the coordinator is full time, and whether volunteers travel. Second, small effects delivered across many people are how population level change usually happens. Third, the counterfactual for many referred young people and isolated older adults is not a better service, it is nothing at all. What a bid must not do is convert any of that into a ratio: there is no defensible social return on investment figure for volunteer mentoring or befriending in the UK, and no published national outcome benchmark.
Mentoring, mentorship and coaching are not interchangeable words
Mentoring and mentorship describe the same activity from two angles: mentoring is what happens in the room, mentorship is the arrangement considered as a whole, and in UK practice the two carry no difference in meaning that any scheme need act on. Coaching is a different method with a different evidence base. A coach works on defined performance goals over a short, structured series of sessions. A mentor works on a relationship first and the goals follow from it, which is why the mentoring literature treats duration as an active ingredient and the coaching literature does not. Training is different again: it transfers a defined body of knowledge and succeeds or fails on whether the learner can then do the thing. A workplace mentoring programme that is really a training programme with a friendly label will be measured as training, and will look like a poor one.
What the mentor gets, and why that evidence is weak
Mentors report gains in confidence, communication skill, understanding of a community they did not know, and for younger volunteers, employability. Those reports are consistent and sincerely given. They are also weak evidence, for a structural reason: they are collected from volunteers who were still there to be asked, so those who found the match distressing or left at month four are largely absent. Any bid citing volunteer benefit should call it reported benefit and should not price it.
What befriending is claiming, and it is a different claim
Befriending claims something narrower than mentoring and should be assessed against its own measures. The reason it attracts funding is loneliness: the Office for National Statistics reports that around 6% of adults in Great Britain feel lonely often or always. The evidence shows modest benefit on loneliness measures, and is weaker and less consistent on depression and on use of health services. Loneliness and social isolation are different things measured in different ways, and a scheme can shift the first without touching the second. A reference that blurs those two is not being generous, it is being useless.
The mentor or befriender
- Is told before agreeing that the average effect is small, and that turning up for a year is the part that carries it.
- Is never recruited on a promise that they will change a life.
- Is told what the scheme will measure, and that their own reported benefit is not the funded outcome.
The young person or the person visited
- Is told what the scheme is for in terms they can check against their own experience, not in outcome framework language.
- Is not described in a bid or case study in a way they have not agreed to.
- Is never the evidence for a claim the scheme cannot support.
The scheme signs the bid, so the scheme carries the claim. It must be able to show a reviewer where each stated outcome comes from, how it is measured, and what it does when the measure does not move. A scheme that cannot answer the third question has an evaluation, not an evidence base.
A funder asks for a return on investment figure and the scheme does not have one
The coordinator says so in writing, and offers instead the measures the scheme actually collects: match length, match survival at twelve months, and change on a validated loneliness or wellbeing measure. Inventing a ratio to fill the box follows the scheme into every future application, and eventually meets someone who checks it.