Tuesday, February 20, 2007

Lets Ride that Iraq Train!

In Law and Economics today I learned about the Drowning Man case. In this case, a man took a canoe into the middle of the lake. The canoe tipped over, and he could not swim. The Defendant saw him flailing in the water, calling for help... and did nothing. He just stood there for thirty minutes until the Plaintiff eventually drowned. The Plaintiff's estate sued... and lost!

That's right, as Americans, you are within your rights to stand by the side of a lake and watch a man drown, even if you have a jetpack and a rope in your possession. (So long as you didn't CAUSE him to be in that position).

One of the major reasons for this is that we don't want to incentivize others to do stupid things because they know they'll be saved. In other words, you have a duty to take care of your own safety. If you can't swim, don't go canoeing! The Park Service has seen the results of this, as lots of moron hikers have been heading into snowstorms because they know the Park Service will bail them out.

And yet, the Democrats responsibility to essentially Bail Out Iraq has become a stick to bludgeon them with. Democrats are supposed to "man up" to our mess in Iraq. We can't leave now, because the Republicans have created a situation where "early withdrawal would produce 'massive civilian casualties.'"

It's refreshing to play a make-believe game where, like a randomized war-strategy game, all the pieces were placed willy-nilly and the players have to do the best they can with the results. From this perspective, it seems bizaare to be in the middle of this Middle Eastern civil war. Who rolled those dice? Get out of there and pursue our actual financial and strategic interests.

But I guess Democrats need to take responsibility for the man in the lake. Because we have a duty to save them


Aaron said...

okay but what happens when you replace "Democrats" in this formulation with "Americans"?

Tommaso Sciortino said...

More accurately, let's say that you believe that you yourself don't know how to swim back and you'll run a 90% chance of dying while attempting a rescue?

Your hypothetical would be interesting Kevin, if we believed that we could fix the situation in Iraq. If - like me and most Americans - you don't think that the problem of Iraq becomes reasonably simple.

Aaron said...

Actually, the jury's still out on what the majority of Americans think:


Tommaso Sciortino said...

What about this?

In a USA TODAY/Gallup Poll, 63 percent of Americans want to set a timetable for U.S. troops to return home by 2008;

Paul said...

How about the fact that Drudge's poll is from a openly partisan polling outfit?


Drudge, of course, omits that fact, but since when did we start taking him as anything resembling a reliable source for these kinds of things?

Rebecca C. Brown said...

I'm also not sure who the man in the lake is supposed to represent. Iraqi civilians? American troops?

If it's the Iraqi people, I'm with Tom: we don't seem to be able to actually help them, and we're getting drowned in the process of figuring it out. Also, in the Man In The Lake analogy, if we were certain we couldn't swim and save him, we'd call a rescue crew who could save him. Who do we call to "save" the Iraqis?

If the man represents the American troops, the choices are to pull him out of the lake (send 'em home) or give him another non-swimming buddy he can work with to try to get out of the lake (surge).

And what is the fact that the man is drowning supposed to represent? If the man represents Iraqi people, does his inability to swim represent a lack of democracy? If so, there are fifty lakes with fifty drowning men that America is "obligated" to save. If the drowning represents the lack of infrastructure and stability that the American presence has encouraged, then I guess we'd be the jackasses who broke the poor guy's canoe to begin with.

Anyway, yeah, I don't see how the Democrats are responsible for any of these scenarios. The Democrats of 2002 are co-responsible at worst.

Kevin said...

Drowning man = Republicans. I guess it could mean Iraqis, but that's not what I had in mind

Aaron said...

I mean, really the question is the methodology of this outfit, not their bias. The fact that the head of the Lancet is a prominent fixture at anti-war rallies and has referred to America and England as an "axis of Anglo-American imperialism" doesn't necessarily discredit the most recent Johns Hopkins study -- what matters is whether the methodology of the study itself is suspect (which it is).

Paul said...

Well, yes, I agree that even a cursory reading of the Drudge article makes it pretty clear that the methodology was flawed. You (presumably) had already read that, though, and still thought it was credible, so I thought I'd just point out that the whole operation is dedicated to "finding" numbers that are good for the Republican party. If you have the actual data and models used, please share. In lieu of that, I'll settle for knowing that the very purpose of a poll like that is to make people overestimate public support of the war to make the pro-war position seem more respectable. They wanted somebody to read their numbers from somebody like Drudge, and then mention them in a conversation about the merits of pursuing the Republican party's official line on the war. Mission accomplished.

Aaron said...

Huh? No, Paul, what you've exposed is the possibility of bias in the survey. I'll agree that's potential here, but that's something that you have to, you know, establish by examining the data. Just as it would be irresponsible for us to dismiss the latest Johns Hopkins survey out-of-hand because their publisher says silly things in public, it's equally irresponsible to merely assume insistently that the data here must be cherry-picked.

It would also be an especially irresponsible flourish to do so while insisting after the fact that it's in fact up to your debating opponent to provide evidence that you have shown no interest in investigating while making strong claims about the legitimacy of the survey.

Paul said...

Aaron, have you even read that Drudge flash? Those questions are obviously leading. 57% support "finishing the job"? We're supposed to think that's a meaningful number? And "I don't really care what happens in Iraq after the U.S. leaves"? Exactly how loaded would the question have to be before you stopped and thought, "Hey, this smells fishy?"

When a candidate's campaign releases internal polling numbers, reasonable people don't assume that those numbers accurately reflect the actual state of affairs in the world. Instead, they assume that the numbers have been massaged to benefit the candidate one way or the other. That's not a bug of the polling, it's a feature.

Here we have not only evidence of partisan bias, but also concrete evidence of bad methodology, just from Drudge.

Rebecca C. Brown said...

I'll throw my "academic and real-world experience with data collection and survey design" weight around: Paul is right.

Aaron said...

First of all, Paul, you really should quote the questions in full if you're going to evaluate them honestly, lest you be accused of, well, cherry-picking. You're right, "finishing the job" is exceedingly ambiguous and vague and would be a horrible question. But that's not what the survey asks: it investigates whether or not Americans "support finishing the job in Iraq, that is, keeping the troops there until the Iraqi government can maintain control and provide security and the Iraqi war is a key part of the global war on terrorism." That's a bit more specific. And so it goes for the other question that you've abbreviated: there's a world of difference between just asking somebody whether or not they broadly "care what happens in Iraq after the U.S. leaves" and what the survey is actually asking: whether or not people endorse the trade-off between what may happen in Iraq after we leave and having troops home immediately.

I was actually going to raise the analogy to campaign polls as well. But just as we don't take polls with specific political agendas as gospel, neither do we dismiss them outright if they ask meritorious questions. And some of the questions here are pretty suggestive:

* By a 53 percent - 46 percent margin, respondents surveyed said that Democrats are going too far, too fast in pressing the President to withdraw troops from Iraq.

* When asked which statement best describes their position on the Iraq War, voters are evenly divided (50 percent - 49 percent) between positions of "doing whatever it takes to restore order until the Iraqis can govern and provide security to their country," and positions that call for immediate withdrawal or a strict timetable.

* The survey also found that voters thought it would hurt American prestige more to pull out of Iraq immediately (59 percent) than it would to stay there for the long term (35 percent).

Now there could well be reason to dismiss the methodology of this survey, but like I said, you haven't really provided reason to think so.

In any case, this is all also fairly peripheral to Kevin's main point.

Paul said...

It's pretty meaningless to say, "well, part of that question wasn't loaded". Yes, I pulled out the loaded phrases, but that's because those are the phrases that load the question. And let's not be silly: the problem with "finish the job" isn't that it's ambiguous or vague, it's that it's defining a very specific course of action - not coincidentally, the official preference of the Republican party - as our responsibility or duty.

Of course nobody dismisses campaign polls out of hand, but they are assumed to be biased in favor of the campaign, even if the questions are straightforward. (E.g., "Who do you support for the nomination?") Yes, loading questions is one way to get the poll results you want. This POS survey clearly does that. But another way is just to manipulate the answers you get to generate desirable results. You can make peculiar demographics assumptions, for instance, and say that some group or other was over- or under-represented in your sample. The point is that we assume the pollsters are doing that if it's clear where their allegiances lie.

I'd say the "too far, too fast" question is nearly as loaded as the other two I mentioned earlier - it treats the Democratic position as one of immoderation. It would be sort of like asking, "Are Republicans being too aggressive, too quickly in their prosecution of the war?" Some of the other questions aren't so clearly loaded, but those also give less glaringly pro-war results.

Nobody doubts that a sizable number of people are in favor of the surge, or the war, or whatever. But why is there any doubt that POS is overestimating that number?

Tommaso Sciortino said...


this poll is being ridiculed even by by a GOP pollster who supported the war.

Well, I suppose that's what you get when you value Drudge more than TPM.

Thinker said...

Political Arithmetik posted the best discussion of the POS poll that I've seen. ( href="http://politicalarithmetik.blogspot.com/2007/02/polling-on-iraq-sometimes-wording.html">http://politicalarithmetik.blogspot.com/2007/02/polling-on-iraq-sometimes-wording.html)

His bottom line:

Despite question wording that seems extreme compared to the other pollsters here, POS got at most a 5 point increase in support for keeping troops in Iraq until the country is stable. And the POS question produced very little difference from the average results for immediate or timed withdrawal. So critics who have jumped on the question wording bandwagon may be right about the wording, but they are substantially wrong about the effect.

One of the most annoying aspects of question wording effects is that sometimes they are large when you don't expect it, and sometimes they are small even when you are sure they should be large.

It is similarly ironic that those who delighted in the POS results were oblivious to the fact that there was not really any new news here. Apparently other pollsters from the "drive-by media" and elsewhere had been reporting substantially the same results for a while now.

Finally, the results of this question are: 49% withdraw immediately or by a deadline, and 50% stay until Iraq is stable. I'd say a 49-50 split isn't a strong indication of overwhelming support for either side. (There are other questions in the survey that address different issues-- some are more supportive of the war and some are less. Towards the high end is "I support finishing the job in Iraq, that is, keeping the troops there until the Iraqi government can maintain control and provide security for its people" 57% agree. But also: "Iraq will never become a stable democracy", 60% agree. Picking which results you like while ignoring the ones you don't may be good politics but it is bad polling analysis.)

If we phrase the question differently do we get different answers? Sometimes. I could illustrate that with other questions about Iraq easily. Opinion about the war is complex, with shifting coalitions of supporters and opponents depending on how questions slice the policy options. And sometimes choice of wording plays a substantial role in that. But not this time.