Monday, February 24, 2014

Anti-Reaction FAQ Spot Check and Epistemic Principles

Matt Simpson can say things I can't predict.

So I felt this was worth checking up on.

The suicide rate is full of confounders, so I skipped that one. See also Nassim Taleb about measuring distributions by sampling their extremes. (What we want to measure is happiness or satisfaction, not suicide per se.)
1.2: Is everyone falling further and further into debt?
But to me, the new graph looks like gradual decrease in debt since World War II up until Reagan’s big military buildup, followed by a gradual retreat from that military buildup. My God, won’t somebody stop Progressivism before it’s too late?!?!
My counter-hypothesis is that Alexander is cherry-picking his data.

For a start, 2008 is an awfully convenient end point.

Simpson had this cogent rebuttal:
However, it also supports the hypothesis that government debt has an equilibrium level. It used to be around a fairly sensible 35%, but has been steadily increasing since 1981.

Thus far, the data is not conclusive in either direction.

I hope Alexander didn't look at post-2008 data. Hopefully you can safely trust that I didn't look at post-2009 data before the above logic.

United States gross debt passed 100% in 2012 and currently stands at 106%. If you knew Reagan had a military buildup and knew nothing of the present but the debt graph, you would think America was at war.

Conclusion: Alexander is not predictive. Reaction is predictive.

To check my work: the IMF itself, the source of half these numbers, thinks public debt is on the reactionary trajectory.


Can't agree. Baseline goes up. Still looks like America's starting a war.

I don't think Alexander is intentionally cherry picking his data. I think he has some epistemic diseases.

First, I don't actually trust any of these numbers.
"Since you are a citizen of a repressive society, you should be extremely skeptical of all the information you get from schools, the media, and popular books on any topic related to the areas where active repression is occurring. That means at least politics, history, economics, race, and gender."

For comparison I looked up Canada's debt and found that everyone is lying about it, and I have no idea how the IMF calculated its numbers. It is not science: it is not verifiable.

The US numbers aren't verifiable either. They're probably being cooked or massaged down the way Financial Post massaged the number up. No disinterested observer would conclude the US government is really only a third of the economy, for example. Add in NGOs and regulatory costs borne by the private sector. Add in decisions 50% controlled by regulation at 50% government action. That's just off the cuff.

Alexander trusts the numbers.

Second, Alexander thinks that numbers can settle questions of interpretation, which, as I have hinted above, they cannot.

Either you think in straight lines or you don't, but in the context of an internet debate, the only way to check is to compare to a known straight-line thinker. Which you can only identify correctly if you think in straight lines in the first place. There's no point in trying to check.

Certainly, if you trust that neither Alexander nor I looked at post-2009 numbers, we can simulate prediction and thus do the ultimate straight-line check. However, that position is immediately vulnerable to any troll who wants to claim they don't trust, as they are indistinguishable from someone who legitimately does not trust.

For example, each time I gather more data, it fits with my previous hypothesis. You could make the argument I'm misinterpreting the data, but that reflects just as badly on Scott Alexander's charts. If they don't support my argument, they don't support his either.

But, if you trust the numbers, and trust they can settle interpretation, then you're not cherry picking: one instance of solid numbers is plenty. Scott Alexander is reasonably good at picking solid number sets.

Third, Alexander has a terrible case of missing the forest.
Does it really fucking matter what the debt is? Even on reactionary blogs, isn't this a side show? (I hope it's a side show.) My point is not that debt is going up, my point is that Alexander's reasoning doesn't support his conclusion, suggesting he has a habit of supporting conclusions badly.

The point of the debt argument is democratic(descriptively) government is unsustainable. More broadly, Carlyle was right. As it (apparently) happens, this is manifesting in unsustainable rises in debt.

But it's also manifesting in unsustainable rises in private debt. It's a general culture of imprudence.

Of course, Carlyle is unscientific too. He published his conclusions far more than his reasoning. Were we to try to reconstruct it, we have no way to check we've reconstructed it correctly. But every time he correctly predicts, we should worry a little more about the predictions that haven't quite come true yet for us.

Good luck finding a think-tank publishing an imprudence ratio. Or, even worse, trying to measure how many individuals have destroyed their lives by taking on debt. How many negative utilons, exactly, does 'destroyed' mean?

However, there's no reason in principle both governments and people couldn't have an unprincipled exception and maintain pristine debt records.

The reactionary prediction is that progressive governments will destroy themselves and grievously harm everyone nearby on the way down. That progressivism makes you more imprudent that you would otherwise have been.

The progressive prediction is that they will not only survive, but thrive, and benefit everyone on the way up. That, (for example) we're becoming imprudent because it is no longer necessary to be so prudent. We can chill and everything will be fine.

You cannot correctly distinguish between these predictions based on suicide rate variations on the order of 1%. It doesn't matter how much debt governments hold unless and until they, all of a sudden, are holding too much.

The most important things cannot be measured at all, even in principle.

1 comment:

Anonymous said...

Brilliant post.