Bell Curve The Law Talking Guy Raised by Republicans U.S. West
Well, he's kind of had it in for me ever since I accidentally ran over his dog. Actually, replace "accidentally" with "repeatedly," and replace "dog" with "son."

Monday, January 17, 2005

Here We Go Again

Hi Everyone,

The BBC is reporting that the New Yorker is reporting that US special forces are active inside Iran preparing for airstrikes against that country. The story also reports that US special forces are active inside as many as 10 Middle Eastern countries. This against a background of Bush administration statements that the election was an "accountability moment" and it produced an endorsement of the Middle East policy based on invasion, occupation and war.

Since this is Martin Luther King Day I think it would be appropriate to consider what that great American would think of these developments? Remember, King was not just a one trick pony. He was mostly remembered for his Civil Rights Movement work but he was also a peace activist and an advocate for the poor.

So what do you think?


US West said...

I think: Dahhhh. The press has been giving hints about our plans in Iran and Syria for months. The Washington Post in Dec had a long column split in two on its front page. The top article was on how Iranians were entering Iraq with plans to affect the Iraqi elections by voting for Sunni candidates. Under that was a second article about how Syria was harboring Baathists who were training resistances forces, what the press calls "insurgents". These are now commonly known stories and the US has gotten more vocal. And it is and open secret that the US would love to hit Iran and Syria. And with troop levels so low, we will either have to institute a draft or continue to use special forces who will train locals to raise rebellion, just as we did in Latin America. Why do you think all the old cronies from the Regan years are strategically placed in the Administration?

What would King say? He's say hit the streets and protest.

Dr. Strangelove said...

Since the week that the U.S. invaded Iraq, I've been saying that Bush's next step will be to link suicide attacks against US troops in Iraq to terrorist camps in Syria, Lebanon, or the West Bank--and then use this as a pretext to invade. I still stand by this prediction.

But I have been very skeptical that the U.S. would actually invade Iran. I think it's more likely that this sabre-rattling is intended to pressure the Iranians to stop meddling in Iraq and leave it entirely within the American sphere of influence, and perhaps also to dissuade them from continuing their nuclear program.

All-out invasion? I think not. But could the U.S. be preparing for airstrikes against Iranian targets, such as "terrorist" bases in Iran, or their nuclear sites? Yes, I can certainly see the U.S. taking that approach... and so can the Iranians, I'll bet. And that's the point, isn't it?

The Law Talking Guy said...

This is the sort of rumor I used to dismiss as alarmist talk. No longer. It is now all too clear that if Pres. Bush were to get on TV tomorrow and say that we had to attack Iran (citing ginned up evidence or whatever) the Red State Majority would immediately jump on board, unquestioningly supporting the government in whatever it does. Anyone who opposed it will be labeled a terrorist sympathizer. Watch. The danger of the Red State Republican is the absolute deference to authority and unthinking nationalism.

You only need to turn the clock back to the 1930s to see a time when it was considered normal for powerful countries to invade the weak, extract tribute, colonize, and even kill. Britain and the US did this easily (see Phillipines, China...). A return to the imperialist mindset, away from the liberalism of the Post-WWII period, can happen at any time. Indeed, the conquest of Iraq was little more than a projection of power for its own sake. The liberal consensus of the post-war period that supported, in principle, decolonization, human rights, and self-determination of peoples and nations, is under radical assault from Bush right now.

Raised By Republicans said...

I've often thought that its obvious that "Red State Republicans" or Evangelical conservatives would have such an automatic deference to authority. Think about their relationship to religion. They aren't Biblical literalists because they like complexity and nuance. They believe that because they are the sort of people who are desperate to be relieved of the responsibility of free will. They want to be told what to believe in a nice neat package with no exceptions to any rules.

I believe this is a point of discussion between the disguised King Henry V and his soldiers before Agincourt (according to Shakespeare's acocunt which I do not hold be be literally true).

US West said...

Don't bee surprised if boots end up on the ground in places like Syria and Lebanon. Consider that NPR reported two weeks ago that the new DOD budget seeks to end or cut spending on Navel and Airforce pet projects in order to redirect the money to the Army. This signals a big shift in DOD spending for the first time in a decade. The report also said that money allocated for "force realignment" (i.e. Rummy's big reform plans) has been redirected to the Army. Follow the money people. In addition, we have put up something like 20 permanent bases in Iraq. Do you really think we are coming home soon? We are closing bases at home and building new ones overseas. It is a repeat of the last days of Rome.

What is really dangerous about Bush is that he has no need to be re-elected now and the Democrats are such a mess that he has no reason to believe that Neo-cons won't do well on 2008. So there is little reason for him to hold back. Remember, he talked of a bold agenda.

LTG mentions the 1930s. I also remember something a good German friend once told me. “Complacency among the people is dangerous. That is what gave Germany Hitler.” I also remember that the Germans never thought a thing like Hitler could happen there. But a democratic system was overturned by one man and one party. Guess, what? Many people think it can’t happen here. I used to have this unwavering faith in our Constitution and our system. But I see now how fragile it all is. A 1930’s style leadership can ( and I think is happening) happen anywhere. People are sheep. They are easily manipulated. Americans are hardly an exception that rule.

Raised By Republicans said...


What we are seeing is a failure of the institutional/multi-partisan approach. As much as I support that approach over others, I must admit that our system of "checks and balances" is not functioning as it normally does.

It is exceedingly unlikely that a bare 51% majority could simultaneously gain at least nominal control of the House, the Senate, the Presidency and the Supreme Court. I doubt its happened since just after the Civil War (a period now recognized as one of corruption and intellecutal poverty). Even then I suspect the actual majorities were bigger.

The founding fathers intentional had each of those institutions chosen by different constituencies for different terms of office to PREVENT what we see now. Their intent was that only massive majorities of opinion could result in such broad control.

I can only hope that my expectations that the effects of increased urbanization will lead to more liberal voting prove justified.

US West said...

The Founding Fathers did not count on there being a tyranny of political parties in this country. We need more parties. And they have to have some kind of power parity. I'd say, off the top of my head 3-5 is optimal. 2 is too few. And this is the problem. We are now dictated to by the head of RNC. And proof of this is how important the choice of DNC head is becoming. For the first time that I remember, it is on the front page.
That tells you where the real power lies.

US West said...
This comment has been removed by a blog administrator.
Dr. Strangelove said...

***We interrupt this blog for a side-note on statistics.

RbR writes, "It is exceedingly unlikely that a bare 51% majority could simultaneously gain at least nominal control of the House, the Senate, the Presidency and the Supreme Court."

Well, let's do a few calculations...

If the nation is evenly divided and we assume this means there is an independent, 50-50 chance that any of the 4 institutions RbR mentioned would dominated by a given party, that leads to a 6.25% chance that all 4 would belong to a single party. But even this crude calculation ignores the correlations between these 4 institutions. For example, having the Presidency makes it more likely you will also control the Supreme Court (in time), and winning the Presidency tends to give you "coat tails" in both Houses.

So there's probably better than a 6.25% chance--maybe 10% or so? Just throwing out a guess. Anyhow, neither figure is "exceedingly unlikely" in my book. A minor point, perhaps, but I'm just saying that this kind of thing is bound to happen every so often and it doesn't mean the system is broken, as RbR fears.

***We now return you to your regularly scheduled blog.

Raised By Republicans said...

Dr. Strangelove,

You have a reasonable starting point of 6.25% chance of all 4 institutions being controlled by the same party. But you make some critical omissions and errors in your assumptions to make that number get bigger....

First, the coat tail effect is not universal. In fact, in the vast majority of "off year elections" (that is elections that select 100% of the House and 1/3 of the Senate but not the President), the President's party loses support in both houses of Congress. Clinton and Bush have both managed to make gains but these remain rare events. So your assumption here is just not supported by historical evidence. This isn't my main point of disagreement.

Second, the Supreme Court is chosen gradually over time. LTG can tell us what the average rate of turnover is but I suspect that not all sitting Presidents even get to chose a single Supreme Court justice and for their choice to shift the majority would require that the judges retiring be of the opposing political affiliation and retire at the right time. So that is another conditional probability to worry about that your analysis doesn't account for.

Third, the House, the Senate and the Presidency all have different constituencies (House seats are local, Senate seats are state wide, Presidency is national). So geographical distribution of the hypothetical 51% majority has to be just right to result in majorities for one party in all three institutions. Concentration of that majority in the South might give you the Senate but be a disadvantage in the House. Conversely concentration on the coasts could give a majority in the House but be a disadvantage in the Senate. So we have yet another conditional probability to worry about.

Finally, all three institutions have different terms of office. The House has 2 year terms. The Senate has 6 year terms with 1/3 standing for election every 2 years. The Presidency has 4 year terms with a 2 term limit. So the hypotheical 51% majority (and its optimized geographical distribution) has to be stable over time to produce simultaneous control. This is especially true with regard to the Supreme Court because of the timing issues in my second point above. So that makes one historical trend that Dr. Strangelove doesn't account for 3 conditional in his statistical analysis.

Leaving aside the historical ommission the conditional probability problem is enough to turn 6.25% into something "exceedingly small." My own arguments are just "off the top of my head" kind of things but it should at least convince you all that the chances should go down from 6.25% not up. Here is why...

Probability of getting simultaneous control of all branches of government = (6.25%)x(prob of favorable timing of justice retirement)x(prob of favorable geographical distribution of support)x(prob of both level and distribution of support being stable over time).

Since each of the probabilities I mentioned are between 0 and 1 the 6.25% starting point gets smaller not bigger. Furthermore, the probabilities I mention are each relatively unlikely events - but to err on the side of Dr. Strangelove's argument let's assume each has a 50% chance of happening (in my opinion the chances are far below that). That would give us (.0625)x(.5)x(.5)x(.5)= .0078 = p = 0.78% chance which I think a reasonable person would regard as "exceedingly unlikely."

Dr. Strangelove said...

The naive 1-in-16 probability (6.25%) calculation assumes that control of each branch is independent. And as both RbR and I agree, this is not true. There are a few interdependencies. The question is in which way they run when you put them all together, if any.

I cited a few factors (e.g. election-year coattails) that make it more likely that the branches will belong to the same party. RbR counters with a few factors (e.g., off-year turnover) that work against this. I see his point. (A few other factors mentioned though--differing terms of office, different times of judge retirement--are irrelevant because they do not address the issue of correlation.)

I had wrongly assumed that all the factors would be positive, so I was wrong about the "greater than 6.25%" claim. But I'm not convinced by the "less than 6.25%" claim yet either. What it comes down to is: if you believe in net positive correlation, you get a larger chance--if you believe in net negative correlation, you get a smaller chance. In other words, if you believe--all things being equal--that the system favors a more united government over a more divided government, you would get a larger chance instead of a smaller chance. (Of course, if you think all the factors end up in a wash you get, 1-in-16 again.)

RbR's argument concerning geographical distribution is the most interesting to me. He points out that the House and Senate represent two quite different projections of the political landscape onto a legislative map. It's like looking at the same picture through two different panes of warped glass.

On the one hand, if there were millions of tiny bubbles each pane of the glass, and the size and placement of the bubbles were wholly unrelated to the underlying picture, then--although the resulting views would be subtly different--the they would be pretty much the same. In that case, you would expect the House and Senate to yield very similar pictures.

But in our case, there are not millions of bubbles but 100 or 435--neither is "large" in a statistical sense (though the House obviously is more likely to yield an accurate reflection). Furthermore, as RbR points out, the shape/size of the districts and the underlying political landscape are correlated--the districting affect the landscape because likeminded people like to live together, and the landscape affects the districting through gerrymandering. So as any cartographer could tell you, it is unlikely that two such different projections will ever yield the same areas/shapes for each country--and neither will perfectly reflect the underlying landscape.

However, that's as far as it goes. If the underlying landscape is dead even, then I agree that neither branch is likely to reflect this accurately, so each will tilt in some direction. But they could tilt together as well as they could tilt oppositely. RbR claims a negative correlation: he claims that the landscape is skewed in a perverse way such that, when things are dead even, mild gains in the House will likely be paired with mild losses in the Senate (and vice-versa).

In other words, all things being equal, RbR says that the political landscape is such that divided government is favored. Do others agree? This is a question for the political people--and I'm not one of 'em. But I am a math-ish person, so I can say that RbR's arguments about South/Coasts etc. are not of the right kind, and the (0.5) factors he threw are very strong statements of negative correlation, and are not neutral guesses like the ones I started with.

Raised By Republicans said...

The differing terms of office are not irrelevant. Rather, they are critical to understanding the phenomenon. The question is not whether or not the four independent event ever turn up R instead of D. The question is will there ever be a moment when all four say R at the same time when the electorate is so closely divided (i.e. 51% vs 49% or so). Think of a yatzee game. By ignoring the varied terms and election periods, Dr. Strangelove is assuming that he can keep an R from dice throw 1 and keep rolling the other dice until they all come up Rs. But since we are concerned with simultaneous Rs and since the constitution requires that every stand for election periodically, you can't ignore the varied terms.

Another reason you can't ignore the varied terms is that people's opinions change over time. So it is critical that some account be taken of time in the model. Assuming that terms of office don't matter assumes preference stability over time on a mass scale (remember we're talking about millions of voters here not single individuals).

With regard to differences between House and Senate I didn't mean to argue (and I don't think I did) that the House and Senate ALWAYS bias in opposite directions. However, the House is based on population so populous states (like on the coasts) get lots of House seats while states in the South and on the plains don't get many House seats. So if your party is disproportionately concentrated on the coasts you'll have an easy time getting lots of House seats but you'll have diminished returns to the same level of support in the Senate. To speak in calculus terms I'm talking about sign difference on the second derivative not the first. Of course sign differences on the first derivative are possible as well.

So why would this kind of diminished returns matter? Because we are talking about simultaneous control under a very specific circumstance (51-49 split in the electorate), seemingly small changes in the overall level - or geographical distribution - of support for one side or the other can change the partisan control of the House in particular. Suppose that instead of winning a bunch of seats in New York by a point or two, the Republicans instead won a smaller number of House seats in Georgia by 30 points. It probably wouldn't change the results in the Senate but would change the results in the House (overall).

By the way if you ever wanted to know the kind of stuff that poli sci grad students talk about when they get drunk...this is it.

Raised By Republicans said...

Oh, one other thing. The founding fathers intended the institutions to favor a divided government. See especially Federalist 51.

Dr. Strangelove said...

With all due respect, RxR's "yahtzee" argument is wrong. A better way to frame the problem is to think in terms of functions and correlations. Imagine the political control of each of the 4 branches as 4 functions that vary with time. All the details of geography and election terms are all included in these functions. The sole question is: are these 4 functions positively correlated, negatively correlated, or completely uncorrelated?

The easiest way to approach the question is to see if we can figure out any direct correlations between the functions. I offered the "coat-tails" argument as one such direct positive correlation, but RxR effectively countered that with the "off-year" argument as a direct negative correlation. I also mentioned the appointment of Justices by the President as another positive correlation, but RxR points out that this is so weak and random as to be negligible. So no luck here.

The next tactic to approach the question is usually to find some fifth function of time that is correlated with each of the original four political functions in a way we understand--because that will tell us how they are correlated with each other. The most obvious choice is the popular will. We believe--we hope!--that control of each of the 4 brnaches is positively correlated (to some degree or other) with the popular will at the time. But unfortunately, in the question at hand, we were asked to consider the situation in which the popular will was dead even--and of course in that special case, these correlations won't matter. So we are forced to look for other ideas.

RxR brought up political geography as a possible correlator. Clearly, the locations of likeminded voters are clumped regionally (and sometimes within districts) and therefore might skew the election results toward one party or another. But in order for this to be relevant to the question at hand, this clumping must give rise to a positive or negative correlation between the House and Senate elections when the popular will is split. In other words, the geography must somehow make it more or less likely that the House and Senate will move together. This is a subtle effect and no one has asserted this either way. Pointing out that there is clumping, and arguing that clumping might in principle matter, is not the same as saying that it actually leads to a significant correlation one way or the other when the popular will is tied.

One more option is to consider correlations of present composition of the branches to the history of the popular will. RxR's argument about the changing of popular will over time would be better phrased in this language. But we must be careful to consider only what is relevant. Autocorrelations--for example, the fact that incumbents tend to stay in power--are irrelevant. On the other hand, an example of a relevant and reasonable negative correlation might be: the historical variance of the popular will which are more likely to lead to an evenly divided popular will at some moment in time are also more likely to generate a divided government at that same moment. Please read that last sentence carefully because it sounds more plausible than it is.

Anyhow, I don't know what the truth of the matter is and at this point I don't really care. By now, I just want to get the statistical argument right. RxR has convinced me that my naive argument that it might be more than a 1-in-16 chance was just that. But by the same token he has offered no good argument that there should be any significant overall negative correlation either. And anyhow, such "probability" statements are meaningless without a model, and it's obvious to me now that there's no simple model. In that case, actual historical studies are more illuminating and relevant. And as RxR pointed out in the very beginning, it's been a century since this happens.

So I suppose that's pretty rare. My point is only that I don't think it's a sign of anything broken in the system. That's all.

Raised By Republicans said...

Here is why time matters in the interests of getting the statistical model right:

Each of the four institutions are correlated but that relationship is spurious. Ultimately they are all functions of "popular will" for want a better term. But each institution is a different function of a different popular will. The fact that the model is constrained to the special case where we have a nearly even split does not assume away time factors. Indeed, I argue it means we should be hyper sensative to them.

For example, the House is a function of the popular will at time T. And that function is largely based - more or less - on a count of population and so favors opinions held in populous regions.

The President reflects the popular will from T-x where x is some number of years between 0 and 8. The electoral college has a slight bias in favor of opinions held in sparsely populated regions.

The Senate however is complicated; 1/3 of the Senate is a function of popular will at T, 1/3 is a function of popular will at T-2, and 1/3 is a function of popular will at T-4. What's more the function is based on states so the Senate's function favors opinions held in sparsely populated states.

The Supreme Court is even more complicated. To the extent that it is a function of popular will at all (debatable that it was ever intended to be so), elements in the court can reflect popular will from T-20 or even further in the past.

These time issues are a problem for contructing a model predicting the liklihood of simultaneous control of all four institutions if popular will is not stable over time. There are mountains of evidence that popular will changes over time and that these changes can be quite sudden. When popular is so close to 50-50, even minor changes in the overall distribution of opinion within the popular will can change the representation in one or more of the institutions. Furthermore because of the interaction of the timing issue and the differing methods of representation, each institution responds to these changes in popular will at different rates. The House responds very quickly. The Senate has a bit of a lag. The Presidency has a bit more of a lag and the Supreme Court is down right anachronistic at times.

So excluding time from the model will not produce accurate assesment.

Dr. Strangelove said...
This comment has been removed by a blog administrator.