Rudy Giuliani, alleged Yankees fan, is rooting for the Red Sox to win the World Series.
P.S. In other Election 2008 news, Stephen Colbert is in a statistical tie for fourth place in the Democratic presidential race:
In the Democratic primary, Colbert takes 2.3 percent of the vote –
good for fifth place behind Sen. Hillary Rodham Clinton (40 percent),
Sen. Barack Obama (19 percent), former Sen. John Edwards (12 percent)
and Sen. Joe Biden (2.7 percent). Colbert finished ahead of Gov. Bill
Richardson (2.1 percent), Rep. Dennis Kucinich (2.1 percent) and former
Sen. Mike Gravel (less than 1 percent).
The poll has a 5 percent margin of error, so it’s meaningless to talk about Colbert’s 2.3 percent being "behind" Biden’s 2.7 or "ahead" of Richardson’s 2.1. As I said: he’s tied for fourth — with everyone else not named Clinton, Obama or Edwards. (He’s tied with me, at 0.0 percent, for instance.) The headline, really, ought to be simply: Colbert gets measurable support. (Hat tip: E&P, via InstaPundit.)
And in more serious election news, Hillary Clinton says she "would consider giving up some of the executive powers President Bush and Vice President Cheney have assumed since taking office." At least, that’s what the Associated Press says she said — and the AP headline turns that into "Clinton Says She’d Give Up Some Powers," which is clearly quite different than saying she "would consider" doing so. And even "would consider" might not be quite right. From the original Guardian article about their interview with Mrs. Clinton:
Ms Clinton said the accumulation of executive power [under Bush and Cheney] had put America
into "new territory" because Mr Bush and the vice president had taken
the view that what were previously extraordinary powers were now
inherent powers that belonged to the White House."I think I’m
going to have to review everything they’ve done, because I’ve been on
the receiving end of that," she said. Ms Clinton stated it was
"absolutely" conceivable that, as president, she would give up
executive powers in the name of constitutional principle."That has to be part of the review I undertake when I get to the White House, and I intend to do that," she said.
So, what she actually said was that she’ll "review everything they’ve done," as "part of the review I undertake" (which is functionally equivalent to saying she’ll set up a committee to study the issue), and that it’s "conceivable" that she would give up some executive powers. And somehow the AP headline-writer translated that into "Clinton Says She’d Give Up Some Powers." Hmm. Wishful thinking much? (Hat tip, again: InstaPundit, who declares himself "somewhat skeptical" that Hillary will follow through on her, uh, non-promise.)
By the way, the Guardian article, if you’re wondering, is headlined, "Clinton vows review of executive power." That is actually accurate.
|
Categories: Election 2008, Baseball
|
October 24th, 2007 at 7:50:44 am
Perhaps he should choose Richardson as a running mate; their baseball picks are eerily similar: http://slate.com/id/2167195/
October 24th, 2007 at 9:40:32 am
Sigh.
While you’re correct that these numbers don’t tell us anything particularly meaningful about how the minor candidates stand with respect to one another, you unfortunately perpetuate the myth that being “within the margin of error” means that you are in “a statistical tie.”
Margin of error refers to how precise the given data point can be, *independent of any other data point,* most of the time. With a 5% MOE (and assuming 95% confidence, which I am quite sure is the case*), that means that 19 out of 20 times, Hillary Clinton’s actual support among the population surveyed is somewhere between 35% and 45% based on this poll. But the 20th time, it could be outside that range. The same holds true for every other candidate on down the line.
So for example, 95% of the time, Edwards’s support could be as low as 7%, while any one or more of Biden, Colbert, Richardson, and Kucinich could actually be slightly above 7%, thus placing them ahead of Edwards, and in third place. And again, the fact that it’s “only” 95% confidence means that there’s a non-trivial chance that the disparity between the poll and reality is even greater.
It’s simply not true that everyone in this poll within 5% of each other is “in a statistical tie.” Even that conclusion is drawing too much from these numbers. Especially because, as noted, the full extension of that logic would be to say that everybody except Gravel, but including Edwards, is in a “statistical tie” for third, which is clearly silly.
Not to mention, technically Obama and Edwards would be “in a statistical tie for second.” It’s true that 95% of the time, and notwithstanding the numbers reported in this poll, Edwards might actually be ahead of Obama 17% to 14%, or thereabouts. So now by this logic we have Edwards simultaneously in a tie for second and a tie for third.
The whole concept of “ties” is simply inapplicable when we’re talking about MOE and confidence level in a poll that is, after all, nothing more than an imperfect snapshot of a moment in time. The higher the confidence level and the smaller the MOE, the more “in focus” the snapshot is, but that still doesn’t mean people are “tied” because they are within the margin of error.
(I should point out that my sigh of frustration is not meant to criticize Brendan individually — he is far from the only person who makes this mistake. Indeed, the MSM repeats it in almost every story they write about any poll whatsoever, anywhere.)
In sum :), being within the MOE means something, yes — but it doesn’t mean it’s “a statistical tie.” That concept is virtually meaningless.
* Full disclosure: I briefly worked as a Project Director for the survey firm that conducted this poll.
October 24th, 2007 at 9:51:51 am
Ms. Clinton stated it was “absolutely” conceivable that, as president, she would give up executive powers in the name of constitutional principle.
Where I live, you get fined for not cleaning that up of the sidewalk.
October 24th, 2007 at 10:08:37 am
. . . for not cleaning that up off the sidewalk :-)
October 24th, 2007 at 10:12:55 am
No self-respecting Yankees fan will root for the Red Sox. Similarly, no self-respecting Sox fan would root for the Yankees. Rudi Giuliani is only showing that he is out of touch with the (Red Sox) Nation and does not deserve our vote because of it. As a matter of fact, I would go so far as to call on ALL Sox fans to vote for somone other than Giuliani.
October 24th, 2007 at 10:19:12 am
And furthermore, when politicians say that they will “have to consider” whether or not they will voluntarily give up power, it does not bode well for us little folk. The momentum Hillary has scares me a little bit right now. Since some states have gone so far as to force their votes in the first week of January, we’re going to have most competitors dropping out early. And that will lead to the selection of candidates who, under a later-primary-voting model might drop out for one reason or another. But I digress.
October 24th, 2007 at 10:19:15 am
I once, and only once, considered rooting for the Yankees. ‘Twas 2001, right after 9/11, when the city was rallying around the team, the country was rallying around the city, yada yada yada.
I ended up rooting for the Diamondbacks, and talking trash to Becky’s R.A., who was a Yankees fan from NYC.
:)
October 24th, 2007 at 10:26:27 am
I like Rudy, but his stock just plummeted.
October 24th, 2007 at 11:17:07 am
This is depressing. Colbert is beating my favorite fake candidate, Mike Gravel.
Now THIS is what I call comedy…
http://www.youtube.com/watch?v=0rZdAB4V_j8
October 24th, 2007 at 11:22:23 am
Considering Rudy isn’t going to carry either Massachusetts or New York, a smart political pander would be to grab a hat for a team from a sometimes Red State, the Colorado Rockies. Considering that the Sox and Yankees are mortal enemies, it would only make sense (my enemy’s enemy is my friend).
October 24th, 2007 at 12:26:43 pm
Has anyone checked polls on Hillary v. Rudy in NY? I haven’t, but it hardly seems a forgone conclusion that Rudy won’t carry NY against Hillary.
October 24th, 2007 at 12:47:34 pm
Angrier, that video is friggin’ hilarious.
October 24th, 2007 at 12:58:31 pm
Joe Mama,
It’s not looking good so far:
http://www.realclearpolitics.com/epolls/2008/president/ny/new_york_giuliani_vs_clinton-323.html
But it’s early. And none of these polls or pollsters jump out at me as being particularly reliable. (Quinnipiac in particular is always suspect.)
October 24th, 2007 at 12:58:51 pm
Sorry, that was me providing the RCP link.
October 24th, 2007 at 1:21:09 pm
Fair enough. Thanks, Brian.
October 24th, 2007 at 2:15:51 pm
Scientizzle-
Yeah. That Gravel has that “am I crazy enough for you to vote for me?” messaging down.
October 24th, 2007 at 3:09:03 pm
I’m going to disagree with you on the concept of statistical ties, Brian. The concept of a statistical tie is not that the two numbers are exactly the same, but that they cannot be meaningfully distinguished from each other based upon the precision capable from a given sampling method. The use of the word statistical indicates that it’s the statistical parameters which are tied to the extent that we can ascertain them, not the underlying values that they represent are the same.
I would indeed say from these poll results that Clinton’s in the lead, Obama and Edwards are tied for second, and that Edwards and those lower than him are in a statistical tie with each other. The transitivity principle does not always apply to statistical analysis; A can be indistinguishable from B, and B from C, while A and C are distinguishable from each other.
To me, the mistake made by the media is not in stating that those within a margin of error are statistically tied, but in not extending that to those within two margins of error.
October 24th, 2007 at 4:11:08 pm
Reading your comment, Mike, I think we actually agree pretty much in toto — the only difference is the significance we attach to the concept of a “tie.” For example, you say you see these results as indicating Obama and Edwards are tied for second. I don’t reach that conclusion, because the reality is that *either* Obama *or* Edwards truly is in second, and the other is in third, and this poll doesn’t / can’t reliably tell us which one.
I guess what *really* peeves me is not so much the “statistical tie” as it is the “within the margin of error.” MSM types would look at a poll like this and think Obama is clearly in second, because Edwards is not “within the margin of error.” And that’s just stupid.
Is this what you mean by their “not extending that to those within two margins of error”? I’d never thought of it that way before, but I guess it really is a reliable “shortcut” to figuring out the true significance of the range.
October 24th, 2007 at 4:25:14 pm
Brian, while it may be true that one or the other is actually ahead of the other, it can also be true (see WA governors race in 2004) that its next to impossible to tell WHO is actually ahead. It is times like that where I think its fine to refer to statistical ties.
Mike often argues from the “words have meanings” camp and that can be valid, but sometimes the understood meaning, even if slightly innacurate is more useful in conveying meaning to people, especially those not well versed in some of these matters.
October 24th, 2007 at 4:25:45 pm
It’s a rare day indeed that I, of all people, get to watch two commenters on my blog, neither of whom is me, argue back and forth about an issue, and think to myself, “Haha! NERDS!”
Thank you, Mike and Brian, for making today such a day.
:)
October 24th, 2007 at 9:21:58 pm
“I think I’m going to have to” does not equate to “vows”. Sorry.
October 24th, 2007 at 9:28:13 pm
A Republican will never again win New York. Upstate is in perpetual economic decline and needs South-type business-friendly taxes and regulations to boost growth again, but with Albany so dominated by NYC and its politicians with big-tax instincts and pandering spending habits, that’ll never happen. The result is Upstate is hemorrhaging middle-class families and the businesses they work at, while NYC is dominated by ethnic urban voters (i.e., they know no life but falling for Democratic pandering and sucking on government’s teat) and ultra liberals.
October 25th, 2007 at 12:13:36 am
“the reality is that *either* Obama *or* Edwards truly is in second, and the other is in third”
I disagree. The proportional allocation of delegates in Democratic primaries means that there’s a good chance that Obama and Edwards would get exactly the same number of delegates from South Carolina’s total allocation. And that is what most folks call a “tie”.
In the Republican primary, winner-take-all rules mean that if a candidate gets the most votes in each Congressional district, he gets all the delegates. Everyone else (including non-candidates like me) is then tied for second place with 0 delegates.
October 25th, 2007 at 8:50:17 am
David:
Largely true, but largely irrelevant. When talking about actual election results on Election Day, it’s no longer a question of how accurately the survey sample represents the actual preferences of the electorate, because the election itself measures the actual preferences of the electorate. In theory, the tally on election day will produce the one, clear, undisputed winner, in every case whatsoever, *except* the exceedingly unlikely instance where the two leading candidates received the *exact* same number of votes. That, and *only* that, is truly a tie, when talking about actual results.
Of course, in practice, as we saw in WA for the gov’s race and in FL for the presidential, even the counting of actual votes is susceptible to error, so sure, absolutely, there will occasionally be “ties” in the sense that the margin between two candidates is so small that we cannot be confident that the result was unaffected by the inevitable human errors made in counting (and casting!) the votes along the way.
But this is an entirely separate and distinct concept from the MOE of a statistical sample at a given confidence level.
The Monster:
I was not aware that anyone was discussing the allocation of delegates according to national party rules. I had perceived the discussion to be about what conclusions could properly be drawn about candidates’ relative support among the overall population of registered voters, based on the results of the statistical survey reported in the post.
If you can direct me to the reference in either Brendan’s post, or the story to which he links, making clear that these numbers were actually meant to represent proportional shares of delegates at the convention, I will happily retract my earlier statements and replace them with a statement of how mind-bogglingly foolish it was to try and get an accurate measure of delegate distribution, which is determined on the basis of a series of *state-by-state* primaries, by taking a survey of a *national* sample.
October 25th, 2007 at 8:53:00 am
Correction:
“among the overall population of LIKELY voters”
October 25th, 2007 at 10:24:12 am
No one has to explicitly be talking about that. You’ve objected to the phrase “statistical tie” (also seen as “statistical dead heat”) under the theory that when an election is finally held, one of the “tied” people actually gets a few more votes than the other. My point (not Brendan’s) is that the very process of voting knocks some precision out of those numbers.
The word “tie” implies the accuracy of the method of measurement. I was taught in science classes not to show more significant digits in a computed quantity than the constituent measurements could justify. That often puts numbers at two significant digits or even one. There are even cases where scientists are thrilled to get an order of magnitude, without even a single significant digit of precision.
Back before we had sophisticated electronic timing devices, there were more ties in Olympic races than there are now. At some philosophical level, the judges may have thought that one of the competitors finished the race some then-immeasurable instant sooner than the other, but they still awarded them the same medal. That is a “tie”.
Two NFL teams can be tied in the standings at the end of the season, requiring “tie breakers” to determine playoff participation. After a tie is “broken”, is it still a tie? I think that in a sense it is, justifying the continued use of the word.
Local elections have ended in absolute ties, even after recounts, requiring the candidates to meet with the election officials to cast lots to award the office randomly (coincidentally the final NFL tie-breaker). When such a tie is broken, would you object to anyone saying that the candidates had tied?
October 25th, 2007 at 11:20:14 am
“No one has to explicitly be talking about that. ”
Respectfully, yes, someone does, *if* my comments are going to be taken as having any relevance to that topic. Given that I wrote on the topic of whether these poll results truly show that certain candidates are in a “statistical tie” in the general population sampled, there is no basis for assuming that my comments would have any bearing at all on the question of whether any such candidates might eventually end up with the same number of delegates at the Democratic convention — or even the question whether, if the Democratic convention were held today and the candidates received support at the convention in exact proportion to the results of this poll, they would end up tied in the delegate count.
That is completely and totally not what I, or anyone else, was talking about, and your bringing it up is the very essence of a non-sequitur.
“You’ve objected to the phrase “statistical tie” (also seen as “statistical dead heat”) under the theory that when an election is finally held, one of the “tied” people actually gets a few more votes than the other. ”
This is false. I objected to the phrase “statistical tie” when used to express the notion that two candidates are “tied” because their performances in a statistical survey are “within the margin of error.” I objected to the phrase in this context because merely being “within the margin of error” has no bearing on whether the candidates are actually close to one another in the population being sampled.
Two candidates who poll precisely at 15% with a 5% MOE, could very well be separated by 10 points, 20% to 10%, and there’s a 5% chance they could be separated by an even greater amount. This is in no sense a tie, statistical or otherwise.
On the other hand, two candidates who poll at 11% and 14% could very well be separated by 13 points, 19% to 6%.
But, two candidates who poll at 10% and 16% could very well be *tied*, even though they are “outside the margin of error.”
Yet application of the phrase “statistical tie” based on being “within the margin of error” suggests that the first two pairs of candidates are “tied” while the last pair is not. As even an elementary understanding of the concept of margin of error reveals, this is not at all the case. The 10/16 candidates are likely to be tied, just as the 15/15 and 11/14 candidates are, so there is no principled basis for distinguishing two of these pairs as “statistical ties” but not the third.
In any event, my objection to the phrase has nothing whatsoever to do with anything related to “when the election is finally held,” as my response to David K made clear. While it is true that imprecision in the actual election sometimes prohibits a definite and objective declaration of the One True Winner in extremely close elections, that phenomenon is completely unrelated to the fact that the statistical methodology underlying political opinion pools is consistently and systematically oversimplified in virtually every media report thereof, such that an unsupportable concept of “statistical tie” has arisen that is related to being “within the margin of error” even though the state of so being has no bearing on whether or not one is truly in anything resembling a “tie,” statistical or otherwise.
“The word “tie” implies the accuracy of the method of measurement.”
I agree completely. My point is that using the word in reference to candidates “within the margin of error” inaccurately represents the accuracy of the method of measurement. It is both overbroad and underinclusive — the former because it treats candidates as “tied” even though they may be significantly separated, and the latter because it treats as safely “not tied” certain candidates that may in reality actually be tied.
Saying that two candidates in a poll are “statistically tied” on this basis thus fundamentally misrepresents both the methodology and the results of the poll.
Your reference to instances where an order of magnitude is a relevant and significant measure is inapposite in a context where ultimately, one vote is enough — that is to say, the scale of the measurement matters. More to the point, you implicitly describe an *actual measurement*, not a statistical sample, which (as I have repeatedly pointed out) is the proper context of my remarks.
Similarly, your examples of early Olympic races, NFL teams at the end of the season, and local elections post-recounts all refer to actual measurements of an actual event, phenomoneon or process constituting the entire relevant universe, rather than a measurement based on a statistical sample thereof. The concepts of margin of error and confidence level only apply with respect to attempts to project the results yielded from the sample to the general population. Specifically, they refer to the size of the range within which the actual measurement of the sampled population is likely to be found, and to how confident we can be that the actual measurement lies within that specified range. These have no bearing on the determination of the actual result pertaining to entire population itself, whether that population is runners, sports teams, or voters.
Finally, you seem to think I have objected to the very concept of a tie, or alternatively that I think a tie no longer exists once it has been broken, arbitrarily or otherwise. I cannot fathom how you could have arrived at such a notion. Perhaps it was by grossly misreading these words:
“the exceedingly unlikely instance where the two leading candidates received the *exact* same number of votes. That, and *only* that, is truly a tie, when talking about actual results”
It is quite clear that nothing in these words has any implications for whether or how to break a tie, much less whether the breaking of a tie somehow retroactively deprives the tie of existence.
If you believe I made other statements suggesting that I find tiebreakers to be so powerful that they can “do violence to history” in the fashion you suggest, please let me know — I’ll be curious to see how you derived that idea.
In the meantime, it suffices to say that of course all the ties you describe still exist, even after they are broken. That has nothing to do with my point that political survey results do not show “statistical ties” based on candidates being “within the margin of error,” no matter how often or loudly you, the media, or anyone else proclaims otherwise.