Diversity December №1: Causal Mechanisms!

The causal mechanism for my holding a bunch of fire is striking a lighter. Or is it my debilitating need for attention from strangers?

Welcome to the first entry in the Diversity December diversity and social science extravaganza! We’re kicking off by talking about one of the most important concepts in social science: causal mechanisms.

A causal mechanism is the thing that causes a thing to happen. I know, I’m getting technical already. The idea is that we know about a lot of empirical associations (for example, that funds managed by racially and/or gender diverse teams tend to outperform those managed by homogeneous teams), and we have a lot of ideas about other likely associations (for example, that toxic cultures lead to less diversity), even if they’re hard to pin down empirically.

When we see empirical relationships between two variables in the world — e.g., team diversity and performance — we may be tempted to use this discovery to motivate a new policy, or expand recruitment efforts, and so on. But there are two big challenges when it comes to actually understanding why two things we care about seem to “go together.” And if we don’t address them, we’re likely to make incorrect (yet still data-driven) decisions.

The first challenge is directionality — if A is associated with B, does that mean A is causing B, or B is causing A? Or, perhaps, is there a third variable C that is driving both A and B? For example, if diversity is associated with better performance, it might be because:

1. Diversity leads to better performance (A → B)
2. Better performance leads to diversity (B → A)
3. Some third factor leads to both diversity and performance (C → A & B)

The second thing to look for is the mechanism itself. For each of the supposed relationships above, we can ask: what’s the thing that’s driving the outcome we see? Let’s see how this plays out in the same example:

1. Diversity → better performance: One possible causal mechanism is that gender and racial diversity are proxies (e.g., visible stand-ins) for cognitive diversity. Thus, diverse teams might outperform non-diverse teams perhaps because they can generate more creative ideas or solutions.
2. Better performance → diversity: It could be that companies with better performance have more monetary and attention resources to spend on initiatives to increase diversity.
3. Something else → both diversity and performance: Perhaps a healthy, inclusive culture where people enjoy coming to work leads to both better performance and more diversity. I know — sounds crazy.

Most companies I worked with assume the first relationship: that more diversity → better performance, and they do not further interrogate why that might be the case. This means they do things like:

• Recruit more women and/or POC under the assumption this will automatically lead to better performance on anything, anywhere, regardless of what team they’re on.
• Not do a blessed thing to interrogate whether the company culture or institutions are harmful or exclusionary of women and/or POC, and maybe that’s one reason there were so few to begin with.
• Ignore, shut down, or penalize these new, diverse employees when they point out problems. Or, if they have good ideas, take the credit, knowing that new hires are less likely to stand up for themselves, and even if they do, you can punish them for being difficult. This is even more insane given the likely causal mechanism behind diversity and better performance of new ideas that challenge the status quo.
• Make sure that, even as you recruit more women and POC, they are still from the same top-tier universities, because you believe they are the only ones that can produce valuable employees, rather than also a likely separating equilibrium for class.
• Get frustrated when women and/or POC leave the company, and then come up with and never question your own ideas about why they’re leaving (e.g., “black people aren’t interested in this industry,” or “women are leaving to have children”).
• Hire a social scientist to explore the likely causal mechanisms, and then ignore what she says anyway. Ok, now I’m just venting.

So what should we do about it? Discovering the actual casual mechanism is pretty close to impossible most of the time, but one method (some might even call it a scientific method!) for getting closer is:

1. When you notice an empirical relationship you’re interested in, before you jump to a conclusion, think about what the causal direction and mechanism(s) might be
2. Ask: If I am right, what else would I see in the world?
3. Go find out if you see it. If you do, you cannot rule this out as a likely mechanism. If you don’t, you’re probably wrong.
4. Be open to being wrong! Then ask: What else could it be?
5. Repeat

Empirical associations are the first step towards better understanding what causes what. But this is just the observation stage of your inquiry. Before you leap to a bunch of policies, ask: Why might we be seeing this? What is the likely causal direction and mechanism at work here? (Hint: there may be more than one.)

Good luck!

And now: A Circus Story!

The second time I ever used the fire fans depicted in the picture for this article, I was performing at Cirque le Soir Shanghai (where else?) and one of the 14 wicks flew off mid-performance. It wasn’t just the cotton that’s wrapped around the pole that you drench in lighter fluid and set on fire, it was the whole damn pole. It flew across the very crowded nightclub and my colleague and friend Panjun, who was dressed as a clown at the time, happened to catch it. The audience thought it was an amazing trick and they cheered and cheered, while we performers all panicked because I almost killed everyone.