SL Tribune

[berkshireskier]

It's always speculation what causes last minute swings like this. The best guess for 2012 is Hurricane Sandy,...

You're completely ignoring the near total alienation by Romney and the American Taliban of women, Latinos, African-Americans, and anyone who desires a candidate who actually has positions that s/he maintains and believes in. Sandy may have helped out a bit, but the prime reasons for Romney's resounding defeat are systemic in a party that has traveled so far to the radical right that even Regan could not be their candidate today because he would be considered too leftist. The American majority heard the same failed policies and economic remedies that favor the wealthy the Rethuglicans have been pushing for over a decade and wisely rejected them. Read what Frum, Sullivan, Jindal, Rubio, et al on the right have said about how Mittens went off the rails, and, especially, this piece by Bruce Bartlett in The American Conservative: http://www.theamericanconservative.com/articles/revenge-of-the-reality-based-community

I'm not sure how we ended up with a political discussion on a ski board, but I would dispute many of your assertions in the above quote. (and, for the record, I held my nose and voted for Obama, but, hey, I live in Massachusetts and would have been killed by my friends and neighbors and family members if I had admitted to voting for Romney). I'm certainly not defending the campaign that Romney ran or his many varying positions on the issues and mistatements and political missteps. I'm also skeptical that, even if Romney had been elected, that he would be any more successful at dealing with the myriad problems this country faces than Obama has been. However, I'm not sure that I would classify this election as "resounding defeat" for Romney. Obama won 26 states and Romney won 24 states; more interestingly, though, if you look at a map of the election results on a county by county basis across the entire US, Romney won a majority of the votes in over 90% of the country (on a geographic basis). And if we broke down the voting results even further by looking at voting precinct by voting precinct, Romney's "victory" across the entire country would be even much larger than 90%. Admittedly, he lost that other 10% - primarily the "inner city vote" by overwhelming majorities and this was enough to put Obama ahead in the larger states with many electoral votes. Furthermore, if there had been a switch of just over 340,000 votes (out of over 110 million total votes cast) in four of the swing states, Mitt Romney would be the new President. When you break down the vote by income segments, Obama won the "poor" vote - those making below $30,000 - with over 65% of that voting segment or by over 7 million votes and he won the overall election by about 3.5 million votes. Obama lost almost all of the other income segments, but not by enough of a margin to overcome that huge advantage he had in the "under $30,000" vote. With respect to the "woman vote", Romney won 45% of that total voting block (but, again, Obama won enormous, stupendous majorities of the black and hispanic "poor" woman vote, skewing his overall majority among women). I would not classify this a "total alienation" of women. Overall, this was an extremely narrow victory by Obama by almost every measure available. It was certainly not the overwhelming mandate that LBJ won in defeating Goldwater in 1964 or that Nixon attained in defeating McGovern in 1972 or even that Reagan won in beating Jimmy Carter in 1980.

[rfarren]

It was certainly not a landslide (i.e. resounding), but in actuality no election historically is truly resounding, i.e. 65-35 split. Most elections break between 50/50 and 45/55 with 45/55 being considered a landslide. This past one being 47/53 was in the middle of that scale. Furthermore, I find the argument that "if you look at a map of the election results on a county by county basis across the entire US, Romney won a majority of the votes in over 90% of the country (on a geographic basis)" specious at best. That argument completely avoids the fact that most of the US population live in cities.

I disagree with Tony strongly as far as Nate Silver is concerned. It would be foolish and overly simplistic to say: since, Obama won with 53% of the vote, he would hypothetically win the election only 53% of the time. It sounds to me that Tony's friend Harold is allowing emotion and out of date analysis to cloud his judgment (national polls vs. state polls, new available data, etc). MLB GM's of the 80s and 90s would find themselves outgunned vs. a modern GM in terms of available data, and how to value players. I see that as no different than a retired political consultant speaking from experience from his war chest days, when commenting on modern electioneering.

In the way that prioritizing and interpreting certain metrics changed major league sports, and the running of franchises, so do I believe it has changed electoral politics. The proof is in the pudding, just look at Mr. Silver's results in the past 3 elections. His results have only improved with each passing cycle.

[Tony Crocker]

I never watched Fox News. I followed the runup to this election less closely than in the past as I was leaving the country on Nov. 3.

You're completely ignoring the near total alienation by Romney and the American Taliban of women, Latinos, African-Americans, and anyone who desires a candidate who actually has positions that s/he maintains and believes in....

I'm not arguing those points and I think the Bruce Bartlett article is quite illuminating. But were those points less true 9 days before the election when long standing reputable national polls said the race was even? Those same polls said Obama led by 4 going into the election and the actual margin was 3.42%.

I agree with rfarren that the "90% of counties" argument is a complete red herring. Nonetheless Obama's 3.42% margin of victory is not impressive. Bush Jr's margin in 2004 was 2.46%. Carter/Ford was 2.06%. The 3 under 1% margin elections since WWII were 1960, 1968 and 2000, with 1960 being far the closest at 0.15%. The next closest elections after the 5 above were Truman 1948 at 4.48% and Clinton 1992 at 5.56%. So the reality is that the most similar margin of victory to 2012 was 2004.

Nonetheless the fact that Obama won at all (plus the Senate results) with the economy of the past 4 years should be a big wake-up call to the Republicans for the reasons many of us have mentioned.

It would be foolish and overly simplistic to say: since, Obama won with 53% of the vote, he would hypothetically win the election only 53% of the time.

Neither Harold or I ever said that. Harold's contention is that the race was even 9 days before the election and moved 4 points toward Obama after that, a contention supported by the national polls.

It sounds to me that Tony's friend Harold is allowing emotion and out of date analysis to cloud his judgment (national polls vs. state polls, new available data, etc)....I see that as no different than a retired political consultant speaking from experience from his war chest days, when commenting on modern electioneering.

Harold may be fairly conservative, but he's definitely "reality based" and would agree with many points in the Bruce Bartlett piece. If you think I'm a numbers geek, Harold is on a completely different level and his knowledge of American political history is encyclopedic.

I disagree with Tony strongly as far as Nate Silver is concerned....The proof is in the pudding, just look at Mr. Silver's results in the past 3 elections. His results have only improved with each passing cycle.

That's a very small sample size. One reason for Harold's suspicions about Silver is that he has made statements about past elections that were not true. The 2008 and 2012 elections were not that hard to call, especially the electoral vote with the current level of political polarization. The Washington Post just ran a piece about an 8th grader who called all 50 states. I have not analyzed Silver's work like Harold has, but I can say from other stuff I have done that 3 election cycles is way too short a time frame to be anointing someone as the definitive authority in the business.

Part of Harold's historical perspective is that he shares the unease about political polarization. In the close elections of 1960, 1968 and 1976 there were lots of close states, so campaigning was more broad based across the country. One reason Nixon looked like crap in those debates with Kennedy is that he pledged to campaign in all 50 states. When Romney wins 23 states by 9 points or more, it pretty much guarantees those states and their counterparts on the Democratic side will continue to be ignored in future national presidential campaigns.

[jamesdeluxe]

The 2008 and 2012 elections were not that hard to call.

I don't even know where to begin with the comment about 2012. You're taking credit for the correct call three weeks after the fact?

[rfarren]

The Washington Post just ran a piece about an 8th grader who called all 50 states.

He also predicted all the senate races except one? Also, being very close on the margin of victory within each state? Also, Mr. Silver's algorithm is about taking polling data and weighting it based on their past performances. How he interprets the data doesn't affect the algorithm's predictive ability, as he doesn't change weighting as the cycle goes on. Harold's contention may have been based on his written entries not on the solid numbers that Silver's model was using.

Part of Harold's historical perspective is that he shares the unease about political polarization. In the close elections of 1960, 1968 and 1976 there were lots of close states, so campaigning was more broad based across the country. One reason Nixon looked like crap in those debates with Kennedy is that he pledged to campaign in all 50 states. When Romney wins 23 states by 9 points or more, it pretty much guarantees those states and their counterparts on the Democratic side will continue to be ignored in future national presidential campaigns.

Your point proves exactly my thought, Harold's experience was ill-equipped to grasp this election. 9 days out Silver predicted Obama would win the election 75% of the time, at that point the right wing press was excoriating him. Furthermore, your friend Harold said he was wrong and thought the election was all but a toss up. The truth is that the red state/blue state paradigm has ossified, and that there are only a few states which dictate the outcome of an election. This meant that the predictive power was in figuring out how those states would fall. Mr. Silver figured out that Ohio was polling better than what the national polls should be predicting for Ohio (This is a horrible oversimplification of Mr. Silver's algorithm). It was information such as that responsible for the prediction that 75% of the time Obama would get at least 271 electoral votes (9 days out). That prediction now appears prescient, and the common wisdom that the election was a tossup at that point appears to be all but just that "common wisdom."

Finally, it is myopic to blame Mr. Romney's loss of momentum on Sandy. Certainly, Sandy helped Obama, but not to the extent that it pushed him over the edge. If you look at the polling data it shows that from October 15th till the 29th the election was trending towards Obama. It is my view that Sandy was the cherry on the top, and by the point she made landfall the election was in hand barring a major gaffe.

[Tony Crocker]

We don't know for sure why the vote moved 4 points in Obama's direction during those 9 days. Sandy was the major news event during that time, so just a reasonable speculation that Sandy was at least part of it.

Mr. Silver figured out that Ohio was polling better than what the national polls should be predicting for Ohio

Obama won the national vote by 3.42% and won Ohio by 1.9%, so the above statement is wrong any way you look at it. Move every state that 3.42% and Florida, Ohio and Virginia switch to Romney and Obama's electoral margin is 272-266.

If you look at the polling data it shows that from October 15th till the 29th the election was trending towards Obama.

I was not reading polls closely during that time. But my impression was that Romney was actually ahead after the first debate, so even on Oct. 29 still fits with that.

I see 3 possibilities here:
1) If Silver was right both 9 days ahead and on election eve, the national polls were way off 9 days ahead but somehow corrected that by election eve.
2) The converse of the above, national polls right both times and Silver wrong 9 days ahead but corrected by election eve.
3) The national polls were close in both instances but Silver's numbers were biased toward Obama. His numbers could have been wildly biased for Obama and he would have only missed North Carolina. His state polling numbers rate to be more erratic than national data but perhaps the final North Carolina polling he used was one of the most accurate state predictions.

Since this debate is about who was right or wrong 9 days before the election, it is not possible to prove conclusively.

[socal]

National polls are irrelevant to predicting the winner.

Sent from my Galaxy Nexus using Tapatalk 2

[jamesdeluxe]

Not a surprise, but fascinating nonetheless:
http://fivethirtyeight.blogs.nytimes.co ... -to-blame/

[Tony Crocker]

The problem with national vs. state polling is the margin of error. Suppose you need 5,000 responses to get a low enough margin of error. You do a proper national poll and then use the relationship among states (which evolves slowly) to project electoral vote.

To get that accuracy out of state polling you will have to poll 5,000 people from several swing states, which costs a lot more. So the state polling may be using say 1,000 responses per state, with each state having twice or more the margin of error of the national poll. I'm not saying these are the real numbers, but it's definitely the principle involved. Also, the long time national polling organizations may have more expertise in constructing an accurate poll than some independent state polls do.

I would agree that internal pollsters are more vulnerable to groupthink and confirmation bias. That's why you want someone like Harold in that kind of organization.

[rfarren]

The sample size increases tremendously when you combine multiple state polls, as Mr. Silver does. He takes all the state polls ( per state), weights them based on each firm's historical trending, and combines them. This effectively lowers the margin of error. Certainly, weighting polls has its risks to skew the numbers, but apparently his methodology is solid given his track record (albeit only 3 cycles). National polls are just as prone to volatility look at Gallups numbers this past cycle. They were predicting Romney by 7 points!!! That is enormously off, and since it is historically a solid and reputable polling firm it was overly quoted to show Romney in the lead by major new sources. If you took an aggregate of the national polls however, Gallup was an outlier, and a major one in that. Simply said, I believe the science of election prediction has moved beyond Harold's understanding of it. There is far more information available now that even 12 years ago. Harold is simply skiing a straight old ski in an era of wider shaped skis.

Mr. Silver figured out that Ohio was polling better than what the national polls should be predicting for Ohio

Obama won the national vote by 3.42% and won Ohio by 1.9%, so the above statement is wrong any way you look at it. Move every state that 3.42% and Florida, Ohio and Virginia switch to Romney and Obama's electoral margin is 272-266.

I didn't say that Silver said Ohio was polling better than the national numbers, I said that Mr. Silver said that Ohio was polling better than what should be predicted for Ohio historically (a center right state traditionally) given the national poll numbers and economic situation.

[Tony Crocker]

I believe the science of election prediction has moved beyond Harold's understanding of it.

Not so sure about that. In his era polls got a 30% response rate. Now the best you can do is about 9%. This makes polling more expensive to conduct and more vulnerable to getting an unrepresentative sample unless you're very careful.

Ohio was 4.02% more Republican than national in 2000, 0.35% more Democratic in 2004, 2.69% more Republican in 2008 and 1.42% more Republican in 2012. That's more variation than I would expect, especially since it's fluctuating rather than gradually trending one way or the other.

[rfarren]

In his day understanding of demographics was far less developed than it is now, as well as voter habits etc... I just saw a segment on this, and it is unbelievable how far these electoral consultants have gone in the last 20 years.

[jamesdeluxe]

Harold is simply skiing a straight old ski in an era of wider shaped skis.

Ouch.

[berkshireskier]

See story below on efforts to deal with the point that I made above - that one candidate winning a large majority of the vote in limited areas - i.e., urban areas - can skew the election results on a geographic basis. It's still amazing to me that you can lose 90% of the country and win the election.

http://www.bloomberg.com/news/2012-12-0 ... votes.html

[socal]

See story below on efforts to deal with the point that I made above - that one candidate winning a large majority of the vote in limited areas - i.e., urban areas - can skew the election results on a geographic basis. It's still amazing to me that you can lose 90% of the country and win the election.

http://www.bloomberg.com/news/2012-12-0 ... votes.html

90% of the country didnt even vote. And of those who did something like 52% voted Obama vs 47% Romney.

Geography and population are very different things. States like North and South Dakota , Wyoming, etc are large land areas but sparsely populated. I posted an article in this thread earlier that's shows that votes in these low population states have a lot more influence in terms of votes per electoral college vote.

Because you live in a rural area your vote should count just as much as those in the cities, no more and no less.

Sent from my Galaxy Nexus using Tapatalk 2