Wednesday, September 5, 2012

My approach to people I disagree with

Brad DeLong leaves a comment that should not be immediately dismissed or immediately endorsed, and which I would differ on a wee bit. My difference on this goes a ways in explaining why I put more time into talking with libertarians and Austrians than some of my readers like. Brad writes:

"1. Re: "Steve understands the problems around inference from non-experimental data, so why does he promote the graph as demonstrating the failure of the stimulus?"

Could it be because Steve Horwitz is not in the education business but the propaganda business?

2. Re: "'Krugman and people like him don't understand that interest rates coordinate intertemporal choice.' I was surprised the first time I heard [Steve Horwitz] say this, but then he revisited it several times. Does anyone honestly believe Krugman and people who agree with Krugman don't understand that the interest rate is a price signal?"

See my answer to (1).

There comes a point when you have to divide people into those who are trying to understand what is going on and those who are trying to pull the wool over their (and your) readers' eyes.

I see no evidence that Horwitz has not long ago crossed that boundary and left honest discourse behind--and here today we have two more examples

No one with a PhD in economics (no one with a BA in economics, for that matter) should be publicly displaying the Romer-Bernstein graph and found not saying one of two things:

1. I think the forecast was great, and thus the stimulus was a terrible failure, or
2. I think the forecast was too rosy, so inference from this graph will give you an underestimate of the impact of the stimulus.

I'm no forecasting expert, so I am happy to hear someone defend #1. My concern was that Steve seemed to be sampling from both. No one with a PhD in economics should do that, particularly because the public will buy it if they do.

And yet we observe people with PhDs in economics doing that.


It could be the propaganda business, this is true. But let's put that a little differently. I think Steve firmly believes from a scientific perspective, and not an ideological or political perspective, that fiscal policy cannot work. We never hold scientific perspectives based on one piece of evidence or theoretical exposition. It's usually a whole constellation of arguments that gets us to our position.

The flip side of that is that no single piece of evidence or theoretical exposition is going to change someone's mind on a scientific question. Nor should it. Science is too messy for that - no one would ever hold a viewpoint for more than five minutes if that were how we staked out positions.

So rather than being into propaganda Steve might just be presenting a synthesized position and making a bad claim about what the Romer-Bernstein graph says as a means of streamlining the presentation of his argument.

This actually isn't that bad of a decision if you think the scientific question has important and immediate practical consequences. You are, after all, just treating noisy data like noisy data (from your perspetive). That's not particularly dishonest if that's what you're doing.

The point is, Brad and I don't think that's noise. We think it's signal.

So the point I'd make to Brad is that you'd observe the same behavior (w.r.t. the Romer-Bernstein graph) from Steve if he were (1.) a propagandist, or (2.) an honest scientist dismissing what he perceives to be noise as noise.


Our job is to convince him it's not noise - it's meaningful and he should adjust his priors. If he is an honest scientist who think's it's noise, we may improve the situation by convincing him of that. At the very least, pestering him might get him to choose his words more carefully so he doesn't mislead the public. At the very least, he'll think of us every single time he brings up the graph in the future.

And if he's a propagandist, what do we lose by treating him like an honest scientist? We expose the error regardless. Ignoring his argument is only going to spread more misconcptions to the public, after all.

It seems to me the optimal strategy whether he is a propagandist or an honest scientist is to treat him like an honest scientist. At the very least it will increase the likelihood that he'll treat us like honest scientists in the future.

That's my view at least.


Two pleas for the comment thread of this one:

- Please, please, please review your comments for an ironic lack of self-awareness before posting. If I don't find Brad's formula convincing, you're not going to get anywhere by writing "why do you even listen to Brad - he's such a propagandist!" I don't take that sort of thing very seriously, and Brad has the added advantage that I think he's right on a lot of the scientific questions so the "propagandist" line will get even less traction there than it did with Steve.

- I know this is a long, navel-gazing type of post. Call me a traditionalist, but I think the way we conduct social interactions is important. And if that's not enough, it's my damn blog. If you think a long post on social interactions isn't worth your time, then commenting on it certainly isn't worth your time.


  1. Bivariate correlations between gdp growth and government spending is the much better MIT-method of "proof". Horwitz should try that!

    1. And I've hit Krugman many times on that one.

      Actually, the fact that he does that hurts the case for stimulus. That should bias the multiplier estimate down too.

  2. a firm reminder.

    Per Munger, you should disregard both people with incentive caused bias and economists with only a hammer.

    libertarians and Austrians and Horwitz meet both tests

    1. The third sentence paints with far broader a brush than I'm willing to use.

      Let's explore the second sentence a little. The economist with only a hammer is ultimately an empirical question. Are all problems nails? I have my doubts about that. So I'm willing to agree with the rule even though it's an empirical question. Apply with care, though. As for incentive-based bias, I think this needs to be explored.

      Why would incentive seek out scientific cover? Because scientific cover offers greater defense in deliberations than mere incentive. How do we deal with this? Deny them scientific cover. Letting them proceed on the pretext of having scientific cover is no solution to the problem at all. So I disagree with Munger on that (assuming you're accurately representing him).

      Again - regardless of whether the person you're disagreeing with is doing it out of honest science or propaganda or incentive it's always best to respond to bad science with good science. It is optimal no matter what you're dealing with.

  3. No one knows better about the propaganda business than DeLong. End of discussion.

    1. FYI to readers - the link provided in this commenters profile implies that this is probably not Steve.

  4. "It seems to me the optimal strategy whether he is a propagandist or an honest scientist is to treat him like an honest scientist. At the very least it will increase the likelihood that he'll treat us like honest scientists in the future."

    I agree.

  5. Re: "Our job is to convince him it's not noise - it's meaningful and he should adjust his priors." But he won't. Anybody who is in the mark-your-beliefs-to-market business at all has changed his view of the world over the past five years. Steve Horwitz hasn't changed his view of the world at all. You are setting yourself an impossible task.

    "And if he's a propagandist, what do we lose by treating him like an honest scientist?" There are lots of smart people who are trying to understand the world--who are not mere propagandists--who could use more notice and who you could productively engage with. Spending your time in the bootless exercise of trying to convince Horwitz to adjust his view of the world when 2007-2012 has not done so has an opportunity cost. You should factor that in.


    Brad DeLong


All anonymous comments will be deleted. Consistent pseudonyms are fine.