The Most Interesting Political Scientist in the World, Keith Poole
Today, I continue to discuss my book, Left Turn: How Liberal Media Bias Distorts the American Mind. (For a summary of the book, including some audio and video interviews of me discussing the book, please see yesterday’s post, or see my web site, www.timgroseclose.com.)
Actually, today I’ll only indirectly discuss my book. Instead, I’ll focus on (i) a remarkable professor of political science, Keith Poole, and (ii) one of his great achievements, NOMINATE scores.
Here is how I described Poole and NOMINATE in Left Turn:
The Primary Dimension of Political Conflict in the U.S., Why PQ Scores Reflect It
If you ask friends of Keith Poole which celebrity he most looks like, they will all tell you the same thing—Santa Claus. And although he doesn’t literally say “Ho ho ho,” he has a strong, distinct, and infectious laugh.
Poole, who currently teaches at the University of Georgia, is one of the most esteemed and prolific political scientists in the country. He has written well over 50 scholarly articles. Plus, he’s written six monographs or books, including Congress: A Political Economic History of Roll Call Voting, one of the most widely read and cited books in all political science.
Along with Howard Rosenthal, currently an emeritus professor at Princeton and an adjunct professor at New York University, Poole developed “Nominate,” which stands for NOMINAl Three-step Estimation program. Some people—and I am one of them— believe that NOMINATE is the most impressive research achievement ever in political science.
The program—which originally was a several-page set of Fortran computer instructions—estimates politicians’ “ideological positions in a spatial framework.” To estimate the positions, Poole and Rosenthal used the votes that politicians cast on roll calls in the House and Senate.
NOMINATE has estimated a position for every legislator who has ever served in the House or Senate. This means that NOMINATE can do things such as predict how James Madison or Henry Clay or Daniel Webster would have voted on, say, NAFTA, the Iraq War Resolution, or President Obama’s health care package.
One of NOMINATE’s most noteworthy accolades occurred in 1994, when the Smithsonian Institution, as part of a display about applications of super-computers, demonstrated some of its results.
Although Poole is universally recognized as one of the brightest and most acclaimed political scientists, you would not have predicted that if you’d known him in his youth. He spent most of his formative years in Ontario, Oregon. “Ontario,” said Poole, “is not in the glamorous part of the state, like Portland or Eugene. It’s in the rural, eastern part—like almost Idaho.”
He did not excel at school. However, after high school, he enrolled at Portland State University, partly so he could obtain a draft deferment from the Vietnam War. “But during Fall and Winter quarter of my second year,” said Poole, “I basically flunked every class.”
“No, it was none of that ‘I was depressed’ kind pop-psychology [expletive]. I just had fun. It’s like that song, ‘You gotta fight. For your right. To paaaarty.’ That was me.”
After flunking the classes, he received a letter from his local draft board, notifying him that his draft status had changed to “1A.” To avoid the draft, he decided to enlist. He chose the signal corps, and enrolled for a three-year stint, which included one year in Vietnam. “We were mainly doing things like operating the radio towers,” said Poole. “The radio men in the bush would send their signals; then our tower would relay them to headquarters.” Poole was technically never in combat. However, during the second wave of the Tet Offensive, the Viet Cong fired 8 to 10 122 mm rockets at his unit. “One landed by the fence and never exploded,” said Poole. “The others exploded in the compound, but none hit a building or even injured anybody. To this day, I think about what might have happened if that guy had had better aim. A shell could have landed in the barracks and killed us all. Sometimes I think about the 40 years of life I would have missed. When I get those notices in the mail for things like wounded veterans, I drop some money in the envelope and send it in. That could have been me.”
Vietnam changed his attitude toward school. “After I came back,” said Poole, “I re-enrolled at Portland State, and I nearly made all A’s. And after I re-took the courses I flunked, I finished with something like a 3.5 [grade-point average].”
Poole applied to and was accepted to the highly prestigious political science PhD program at the University of Rochester. In early 1978 he graduated from Rochester, and he took a visiting teaching position at the University of Oregon. Rochester had impressed upon him the importance of using mathematics to study politics. At Oregon—while holding a full-time teaching job—he began to take upper-level courses in calculus, topology, real analysis, and statistics. He found that he had a special talent for topology—a branch of mathematics that analyzes shapes and mappings from one space to another. “Some of those problems were a bear,” he said. “Things like ‘Suppose two planes intersect a sphere. Compute the volume between the two planes.’”
Poole thinks in spatial terms. His language is peppered with phrases such as, “once you project the vector onto the hyperplane,” and “if you rotate the entire space.” Believe it or not, when he says things like that, he is usually talking about politics.
His political views are conservative, yet, unlike some academic conservatives, he is not bashful about revealing them. “I read it in the New York Communist Times this morning,” he will occasionally say. “Slick” was his favorite moniker for Bill Clinton during his presidency.
Poole may be the most well-read person in America. If you know him, you will often learn information that the rest of the world won’t learn for several days or weeks, sometimes months. During the 2004 elections, for instance, he explained why John Kerry would not withdraw troops from Afghanistan or Iraq if he won the election. “Even the Democrats understand that we have to win these wars,” he said. “Otherwise it could be real bad—like the end of Western Civilization. Just read Bernard Lewis. You’ll realize what’s at stake.” Poole’s insight, I believe, presaged President Obama’s Afghan and Iraq policy. Although, during his campaign, Obama fooled his leftwing base into believing that he’d withdraw nearly immediately from the wars, those of us who’ve benefitted from Poole’s insight, doubted that he’d really do that.
It was Poole who first informed me about Bill Clinton’s affair with Monica Lewinsky. I vividly remember the time he told me, and I also remember the clairvoyant feeling that I had the next couple days. I knew some information that few people knew. Yet in a few days, most of America would learn that information and consider it explosive.
“Oh man, this is good,” he said on the phone one day. Are you by your computer?” “Yeah,” I said.
“Okay, you gotta type this URL: w-w-w- dot-d-r-u-d-g-e-r-e-p-o-r-t, yeah all one word. …. Okay are you reading it?”
“Amazing,” I said. “Could this be true?”
“I don’t know, but her co-worker’s cooperating with the special prosecutor,” said Poole, whose excitement was about to overwhelm him. “And she waaar a WAAAR!” he bellowed.
“She what?” I asked.
“She wore a WIRE!!!” he exclaimed.
Although Poole is conservative, there is nothing conservative—or liberal—about NOMINATE. It uses standard mathematical tools, albeit applied in some brilliant ways, to measure the ideology of politicians. Indeed, Poole’s co-creator of NOMINATE, Howard Rosenthal, is liberal.
With the NOMINATE procedure, Poole and Rosenthal record the yea and nay votes of legislators. Their computer program then estimates a number for each legislator. That number indicates the position of the legislator on the dimension that best describes the main conflict between legislators on the roll-call votes.
In theory that dimension could represent any sort of conflict—e.g. a rural vs. urban, South vs. North, libertarian vs. anti-libertarian, pro-Civil rights vs. anti-Civil rights, etc. In practice, however, the numbers indicate a very particular type of conflict—liberal vs. conservative.
That is, once the NOMINATE computer program spits out its estimates, the legislators with the lowest scores are people—such as Nancy Pelosi and Barney Frank—who are commonly called liberals. And the politicians with the highest scores are people—such as Michele Bachman and Jim DeMint—who are commonly called conservatives.
There is nothing in NOMINATE that forces this result. For instance, if issues and votes in Congress had been different, then the program might have generated numbers with, e.g., (i) Southerners were at one end and Northerners at the other end, or (ii) libertarians and anti-libertarian at the other end. The fact that the latter type results did not occur means that we can infer the following about American politics: The conflict that we commonly call liberal-versus-conservative is indeed the main conflict within Congress.
At least two other sets of scholars have examined the same research question as NOMINATE— to estimate legislators’ preferences on the dimension of maximal conflict, using roll call votes as data. These scholars and the scores they have produced are:
- “Heckman-Snyder” scores, named after James Heckman, the Nobel-winning economist at the University of Chicago, and James Snyder, my occasional co-author, who is now a professor of government at Harvard University.
- “CJR” scores, created by Josh Clinton, a political scientist at Vanderbilt University, and by Simon Jackman and Doug Rivers, both political scientists at Stanford University.
These efforts make technical assumptions that are slightly different from NOMINATE’s. However, their main results are nearly identical to NOMINATE’s. Most important, they find that the dimension of maximal conflict is what people commonly call liberal/conservative.
Now back to PQ scores. It happens that PQ scores are very highly correlated with NOMINATE, as well as the latter two sets of scores. Thus, while the previous section argued that PQs capture the conflict that is commonly called liberal/conservative, the latter finding means that PQ scores are also meaningful—that is, they indeed reflect the main conflict in American politics.
[From Left Turn by Tim Groseclose, PhD. Copyright © 2011 by the author and reprinted by kind permission of St. Martin’s Press, LLC. All rights reserved.]
Yesterday, I said that I’d provide evidence that American political views are basically one dimensional—that is, if you’re conservative on one issue, say taxes, then it’s a good bet that you’re conservative on most other issues.
Here’s how NOMINATE provides evidence of that. When Poole and his co-author, Howard Rosenthal, run their NOMINATE computer program, they first tell it whether to estimate one-dimensional scores, two-dimensional scores, or whatever dimensional scores they decide to estimate that day.
In the above passage I described only the one-dimensional version of NOMINATE.
Now, it happens that the one-dimensional version of NOMINATE does very well at predicting roll call votes in Congress. For instance, since about World War II, the one-dimensional scores predict about 80% of legislators’ votes correctly.
Next, it happens that if you use the two-dimensional version, you do even better at predicting roll call votes. Specifically, if you look at roll calls that occurred in the period 1940-1980, the two-dimensional scores allow you to predict something like 86% of the votes correctly. That is, the second dimension buys you an additional 6% of accuracy.
As I mentioned above, the first dimension seems to measure legislators’ left-right ideology—that is, how liberal or conservative they are. As Poole and Rosenthal note, the second dimension seems to measure racial preferences—that is, how the legislators vote on issues like affirmative action and civil rights. As Poole and Rosenthal note, the second dimension of NOMINATE captures the area of conflict that is the second-most important in Congress. Thus, according to NOMINATE, general liberal-conservative ideology is the most important area of conflict in Congress. Second-most important are racial issues that are not well explained by general liberal-conservative ideology.
But here’s what‘s interesting. After 1980, the second dimension became much less important. That is, in terms of predicting roll call votes, the second dimension would only buy you an addition 1-2% of additional accuracy.
The reason is that, before 1980, there were lots of Southern Democrats, like Sam Ervin, who would vote liberally on most issues. However, on racial issues, their ideology could be called conservative. That is, they would tend to oppose affirmative action and civil-rights issues. But sometime around 1980 such legislators began to disappear. That is, anyone who voted liberally on, say, economic issues would tend to vote liberally on racial issues. That’s why the second dimension became less important.
Because the second dimension gives only a tiny amount of additional accuracy (1-2%), Poole and Rosenthal suggest that American political views are basically one-dimensional. Their results suggest that the Sam Ervins (and also Ron Pauls, and any other legislator who does not seem to fit within the standard liberal-conservative framework) of the world are very rare—at least in Congress. That is, if a legislator is conservative on one issue, say taxes, then he or she will probably also be conservative on all the other issues, including race and foreign policy.
If you’re interested in learning more about NOMINATE, or if you’d like to download NOMINATE scores, click here for Keith Poole’s web site. Also, here is an excellent Slate article that describes NOMINATE.
If you’re curious to see the extent to which Keith Poole really looks like Santa Claus, click here.
- Comment (11)
- · Quote
- · UnfollowFollow (2)
Comments :
Re: The Most Interesting Political Scientist in the World, Keith Poole
Tim, thanks. I know how crazy-making it is to be asked a question that you provide a good answer to in the book itself. I unfortunately can't get my hands on a copy quickly to see if you've already answered it, so please indulge me--exactly how are you defining "liberal-conservative?" How does it differ from "Democrat-Republican?"
Re: The Most Interesting Political Scientist in the World, Keith Poole
Claire, I let the Americans for Democratic Action choose the questions for me. Thus, like lots of other political scientists, I allow the ADA to define "liberal" for me. To compute Nominate, Poole and Rosenthal do not explicitly define "liberal" and "conservative". Instead, Nominate's first dimension is defined as the "primary dimension of conflict in Congress." However, once Nominate spits out its results, you see that that dimension looks like a liberal-conservative dimension. That is, the legislators with the lowest Nominate scores are people like Ted Kennedy, Maxine Waters, and Barney Frank - people we tend to think of as very liberal. The legislators with the highest Nominate scores are people like Michele Bachmann and Jim DeMint - people we tend to think of as very conservative. Hope this helps to clarify things. -Tim
Re: The Most Interesting Political Scientist in the World, Keith Poole
Tim, I'm sorry--I'm sure it's on your site, but I'm not seeing it. End of a long, interesting day here. Can you point me to the link for the ADA's definition?
Mar '11
Re: The Most Interesting Political Scientist in the World, Keith Poole
"I unfortunately can't get my hands on a copy quickly to see if you've already answered it"
You've got to get a Kindle, Claire.
Oct '10
Re: The Most Interesting Political Scientist in the World, Keith Poole
Would this be the ADA's LQ = Liberal Quotient?
You can measure your own. A half-joking genferei was 5, as was Mrs Bachmann. Paul Ryan was 0.
Re: The Most Interesting Political Scientist in the World, Keith Poole
Claire: Well, when I say how the ADA "defines" liberalism, I just mean the votes they choose to compile their scores and their choices whether to make the yea or nay side the liberal side. Here's a link to the roll calls they've chosen, as well voting records of politicians: http://www.adaction.org/pages/publications/voting-records.php.
Re: The Most Interesting Political Scientist in the World, Keith Poole
Mr. genferei: The PQs that I calculate are very similar to the Liberal Quotients that the ADA discusses. The difference is that Liberal Quotients aren't designed to be comparable over time. But PQs are. PQs are like the "curved" scores on a test. LQs are like the raw scores. You'll have to see the book for the full details. But it just so happens that I discuss a few of the details in the first segment of my interview on Uncommon Knowledge (which was just posted maybe a couple hours ago!). Here's the link.
Sep '10
Re: The Most Interesting Political Scientist in the World, Keith Poole
One dimensional PQ scores are predictive with 80% accuracy. How accurate is party affiliation as a predictive element? I would have guessed it was somewhere around the same figure.
Oct '10
Re: The Most Interesting Political Scientist in the World, Keith Poole
Tim Groseclose:
The reason is that, before 1980, there were lots of Southern Democrats, like Sam Ervin, who would vote liberally on most issues. However, on racial issues, their ideology could be called conservative. That is, they would tend to oppose affirmative action and civil-rights issues. But sometime around 1980 such legislators began to disappear. That is, anyone who voted liberally on, say, economic issues would tend to vote liberally on racial issues. That’s why the second dimension became less important.
Goodness knows I haven't looked at calculus since that old AP test, however there is something in the argument above that seems inconsistent to me and probably Ann Coulter. In Demonic, she discusses extensively in Chapter 10 the current mischaracterization of the conservative position preceeding and during the civil rights movement. Although interwoven with hyperbole and snark, the historical points are there. So, if the ADA defines the baseline, wouldn't all findings be relative to their own interpretation of history in the present day instead of platform issues that might evolve and change within a party over time? I understand that prevents comparisions over time but how can one ideology determine the baseline for both?
Dec '10
Re: The Most Interesting Political Scientist in the World, Keith Poole
Tim thanks for the post and the bio info. It's always nice to hear about about a hometown boy who's done so well.
Jun '10
Re: The Most Interesting Political Scientist in the World, Keith Poole
Prof. Groseclose,
I can accept that your methods give a meaningful ordinal ranking (x is more conservative than y) without necessarily accepting that there is meaning in a cardinal ranking (a numerical measure of how much more conservative x is than y). What's the argument for that? Another way of putting the question: How sensitive are your results to the way in which the ordering is mapped onto a numerical scale?