Research excellence and the O-ring theory

economics ▪ policy ▪

Yesterday afternoon, I had lunch with Roby Muhammad. For those of you who don't know who he is, Roby is a member of a small but growing number of quantitative sociologists. For the past few years, he has been doing cutting edge research on network behaviour with Duncan Watts (of six-degree fame) at Columbia University.

Anyway, after a brief chitchat, Roby told me he would like to see in Indonesia something akin to Stanford's Institute of Advanced Studies, to allow young, brilliant researchers to do basic (theoretical) research. He couldn't understand why no such institute existed in Indonesia. "It isn't that difficult. Three or four (likeminded) people can hang out and discuss together -- that would have been enough."

He asked my opinon on it. So, I told him of Michael Kremer's O-ring theory of development.

The O-ring is the malfunctioning seal ring that caused the explosion the the space shuttle Challenger in 1986. It didn't matter that the rest of shuttle was in perfect working order: one malfunctioning component was enough. It suggests that high-quality parts would be useless when paired with low-quality ones.

Michael Kremer used this example to illustrate a development problem: Brilliant workers (or scientists) prefer to work with other brilliant workers because by doing so, they are guaranteed the high payoffs of their high-quality skills.

There are two reasons why this is so. First, this arrangement averts highly skilled workers from O-ring-type accidents. Second, highly-skilled workers gained from their interactions with other similarly skilled workers -- the so-called "spillover effects". These gains are reflected in a collective payoff that are much higher than the sum of its parts.

As the result, there is what economists called "pooling equilibriums" where high-quality workers "agglomerate" with other high-quality workers, leaving low-quality workers together with similarly low-quality workers. Kremer used this to explain various stylized facts about rich and poor countries, and why incomes are diverging between the richest and poorest countries. But this theory can also be explain Roby's conundrum.

Roby took for granted the ease of which he could do theoretical modelling in the United States. I could imagine him and several of his colleagues, on a Friday afternoon, sitting in a coffee-shop scribbling out some new theory on a piece of napkin, say, of the network behaviour of an epidemic. The day ended well when a satisfactory model was developed, and the whole thing felt so easy. After all, what was needed was a pencil (or pen) and a piece of paper.

But Roby missed out on the fact that, right there and then, at Columbia University, he was probably sitting with people that are amongst the best in their fields. That's why getting to top graduate programs is so damn difficult. Administrators of those programs understand clearly this idea of positive spillovers amongst smart and highly-skilled people in an academic environment.

So, will dumping loads of money to create an institute that employs and draws the best of Indonesian social researchers solve the problem?

Well, not quite. You see, I believe that most of the times, Roby (and other researchers studying in the United States) managed to solve their problems on their own. But on those rare days when they got stuck -- say, in figuring out why a particular equation cannot be solved -- they can easily visit their colleagues elsewhere, say in the Math department, to help them out.

Similarly, when they need data to be collected or statistically analyzed, it's relatively easy to find highly-skilled data collectors (and, mind you, you need a special skill to collect data properly) and analysts at Columbia University. And that's only at Columbia -- there are still Harvard, MIT, Yale, etc. Meanwhile, I dare anyone to name just one graduate program in Indonesia that can claim it only admits the best and brightest students.

This goes back to the O-ring problem. One underappreciated fact about world-class researches is that most are actually results of collective enterprises, whether the principal authors realize them or not. Unless all of the high-quality parts are there, it's not possible to come up with a Challenger that can touch the moon.

So, does this mean that there is no hope for a world-class Indonesian "Institute of Advanced Studies"? Realistically, not in the next five (or even ten?) years. But now that we have identified the problems (and it isn't just about money!), we can start trying to think of ways to get around of them. If you have any idea, please do pitch in.

Update (9/10): This post on Greg Mankiw's blog suggests an implication of this phenomenon.