Sunday 20 March 2016

Wishing X caused Y

Imagine some topic, X. 

Most people agree that X is desirable for its own sake - a good thing. 

Many people want to believe that good things automatically result in good outcomes in other areas we can measure, like Y. They want it to be true that good thing X also means improvements in Y.

They want it so badly that they start to think anyone who doesn't believe in the link between X and Y must not really like X in the first place.  

There will be consultancy reports claiming to find links between X and Y. Like places (or people) with more X have more Y, ignoring that there are many other potential differences between the kinds of places (or people) that wind up with lots of X and the kinds of places (or people) without much X. And that those covariates could be what's driving the perceived relationships between X and Y.

Or, flip it around. Suppose some commonly disparaged thing R is commonly believed to cause bad outcome S. The consultancy reports show a link. All the right-thinking people believe the link. If you question the link between R and S, even if you hate R, maybe it's because you secretly like R. 

It isn't hard to wind up in spots where you can't question the link between X and Y, or between R and S, without being believed to be a bad person - and especially if some who question the link really do hate X or like R (or just like investigating contrarian hypotheses, and contrarians are always bad). And then you're in a preference falsification equilibrium, where everyone either lies or shuts up about the evidence on X and Y, or R and S. 

In a world of opportunity costs where you have to choose where to direct your attentions, and there are whole vectors of X Y, and RS, all over the place, what's the point of even heading into areas where you will not only fail to make any progress, but also look like a jerk while failing to make any progress? There are plenty of areas where Treasury hasn't yet decided to play the short "let's be popular" game instead of the long "let's stick with the evidence" one, closing the Overton Window on any kind of reality-based assessment. Windows remain open elsewhere.

No comments:

Post a Comment