Amidst all of the studies of public attitudes, there are relatively few studies that look at how we learn about inequality – yet if we know how people learn about inequality, then we have ideas about how people’s attitudes can be changed. So I was really interested to hear a presentation by LSE/Harvard’s Jonathan Mijs, looking at how the nature of universities influences how we learn about inequality. In this post I explain Mijs’s study (which I liked), and also his policy recommendations (which I didn’t)…
How do we investigate learning about inequality?
Mijs’ study has a simple idea at its heart: racial inequality is central to wider inequalities in the US (even more than the UK), and universities are a key place where different races1 can learn about each other and the scale of the (dis)advantage they face. Given that most US universities randomly allocate students to roommates, he looks at whether being randomly allocated a roommate of a different race makes us more aware of wider inequalities. (He uses an existing long-running survey called the College Freshman Survey and College Senior Survey, across 436 US universities).
The answer, in short, is ‘yes’. His slides haven’t been made available on the LSE website unfortunately, but you can see some of the results in a post at The Conversation, including this figure, which shows how people assigned a different-race roommate were noticeably more likely to explain inequality through structural factors:
I really like the use of randomized roommate assignment here – it’s a powerful way of getting at causal effects (in an area in which I don’t really trust most simple regressions). It’s worth adding that we’re assuming that roommates really are assigned at random, rather than people being matched to roommates that seem to be similar than them. However, there’s only a few things that university administrators would know about first-year students on-paper, and these are things that Mijs already controls for (and he does a sensitivity analysis restricted to first-year students), so this seems convincing to me. (Though I wasn’t as convinced about the analyses of student attitude change because of roommates after the first year for this reason).
While I liked the study, there are a couple of things that are worth bearing in mind here:
- There’s no published paper on this yet – I think he has a working paper available and it featured in his as-yet unpublished thesis.
- In Mijs’ talk, he actually presented a slightly different set of results from the same analysis (on students’ agreement with ‘Through hard work, everybody can succeed in America’ and ‘Racial discrimination is no longer a major problem in America’). He found that the effects were statistically significant but quite small (about 0.04-0.10 of a standard deviation).
- Mijs also went on to show a lot complex interaction effects in his talk – that is, that the effects of having a roommate of a different race varies depending on your own race (and therefore your roommate’s race), and also the context of your institution. This was all quite reasonable – the impact of having a different-race roommate is greater at less diverse institutions, for example – but also slightly ad-hoc, which is why I don’t focus on them here.
What I really wanted, though, was to push this a bit further to look at exactly what is going on here. We can’t be completely sure that the effects are due to the race of the roommate – different-race roommates may also be different in other respects (e.g. socioeconomic class), and it might be this other aspect of difference that matters. Or plausibly the impact of a different-race roommate depends on what they’re like and how you get on with them – maybe having a super-smart different-race roommate leads you to think differently than having a different-race roommate that gets lower grades than you do. So I hope he continues to deepen this analysis even further in future!
So what should we do in response?
Where I parted ways from Mijs, though, was in his policy recommendations – and this comes to an issue that I’m going to return to on the blog, because Mijs followed a pattern than I often see among academic researchers. Mijs logic went something like this: I’ve found out that different-race roommates make us more aware of the inequality that exists (which is a good thing) ⇒ universities should introduce a policy of deliberately assigning people different-race roommates. But I think this is a flawed logic.
The point is that the effect of all policies depends on their meaning, and their meaning is changed by doing something deliberately as opposed to accidentally. Say that you currently had a different-race roommate who you didn’t like. As it stands, this is just the luck of the draw: some people get roommates they like, some don’t, but it’s nobody’s fault. But now imagine that the university had a policy of deliberately assigning different-race roommates. You may now blame your university for having a roommate you didn’t like, and moreover, this might create resentment that makes people less inequality-aware rather than more. I don’t know how far this would happen – but it’s clear that there is a gap between the effect-in-the-research and the likely-effect-of-the-policy.
More broadly, if we’re going to recommend policies, we need to think about them more deeply – about how they would be delivered in practice; about how this context differs from the context we studied; about what their unintended consequences might be; about the ethics as well as the effectiveness of what we’re proposing. Yet too often, social policy studies find that something matters, and then go straight to recommending that the thing is changed in a particular way, without really thinking this through or developing expertise on the policy scenario in question. This is arguably a collective issue rather than a matter of individual blame, and I have doubtlessly done the same myself; this is just what our journal papers, theses etc are expected to look like…
That said, I do think that Mijs’ theoretical approach has much value for policy. He argues that people mostly look at macro-level factors (culture/ideology, as I’ve discussed previously here) or micro-level factors (individual characteristics), but that this meso-level – our social context – is really important because it shapes the inequality that we see around us: if we’re surrounded by people like ourselves, then we will tend to underestimate the scope of wider inequalities. So aside from this paper on universities, he also has another interesting paper on school diversity (DOI), and an as-yet unpublished survey experiment getting people to think about diverse neighbourhoods, with hopefully more to come in future years.
Despite our disagreements, it’s an interesting line of work, and interesting food for thought.
1 I use the term ‘race’ following Mijs, but obviously I mean this in the sense of a culturally-constructed group, rather than something with an essential biological meaning.