Comments on FiveThirtyEight: Politics Done Right: Tipping Point v2.0

酒店經紀人, 菲梵酒店經紀, 酒店經紀, 禮服酒店上班, 酒店小姐兼職, 便服酒店經紀, ...

2009-10-08T07:47:12.708-04:00

酒店經紀人,
菲梵酒店經紀,
酒店經紀,
禮服酒店上班,
酒店小姐兼職,
便服酒店經紀,
酒店打工經紀,
制服酒店工作,
專業酒店經紀,
合法酒店經紀,
酒店暑假打工,
酒店寒假打工,
酒店經紀人,
菲梵酒店經紀,
酒店經紀,
禮服酒店上班,
酒店經紀人,
菲梵酒店經紀,
酒店經紀,
禮服酒店上班,
酒店小姐兼職,
便服酒店工作,
酒店打工經紀,
制服酒店經紀,
專業酒店經紀,
合法酒店經紀,
酒店暑假打工,
酒店寒假打工,
酒店經紀人,
菲梵酒店經紀,
酒店經紀,
禮服酒店上班,
酒店小姐兼職,
便服酒店工作,
酒店打工經紀,
制服酒店經紀,
酒店經紀,
菲
梵,

艾葳酒店經紀公司提供專業的酒店經紀, 酒店上班小姐,八大行業,酒店兼職,傳播妹,或者想要到打工兼差、...

2009-09-11T02:04:26.703-04:00

艾葳酒店經紀公司提供專業的酒店經紀, 酒店上班小姐,八大行業,酒店兼職,傳播妹,或者想要到打工兼差、打工,兼差,或者八大行業,酒店兼職,想去酒店上班, 日式酒店,制服酒店,ktv酒店,禮服店,整天穿得水水漂漂的,還是想去制服店當上班小姐,水水們如果想要擁有打工工作、晚上兼差工作、兼差打工、假日兼職、兼職工作、酒店兼差、兼差、打工兼差、日領工作、晚上兼差工作、酒店工作、酒店上班、酒店打工、兼職、兼差、兼差工作、酒店上班等,想了解酒店相關工作和特種行業內容,想兼職工作日領、假日兼職、兼差打工、或晚班兼職想擁有快速賺錢又有保障的工作嗎???又可以現領請找專業又有保障的艾葳酒店經紀公司!

艾葳酒店經紀是合法的公司工作環境高雅時尚，無業績壓力，無脫秀無喝酒壓力，高層次會員制客源，工作輕鬆，可日領、現領。
一般的酒店經紀只會在水水們第一次上班和領薪水時出現而已，對水水們的上班安全一點保障都沒有！艾葳酒店經紀公司的水水們上班時全程媽咪作陪，不需擔心！只提供最優質的酒店上班,酒店上班,酒店打工環境、上班條件給水水們。心動嗎!? 趕快來填寫你的酒店上班履歷表

水水們妳有缺現領、有兼職、缺錢卡奴的煩腦嗎?想到日本留學缺錢嗎?妳是傳播妹??想要擁有高時薪又輕鬆的夜間兼職工作,打工機會和,假日打工,假日兼職賺錢的機會嗎??想實現夢想卻又缺錢沒錢嗎!??
艾葳酒店台北酒店經紀招兵買馬!!徵專業的酒店打工,想要去酒店的水水,想要短期日領,酒店日領,禮服酒店,制服店,酒店經紀,ktv酒店,便服店,酒店工作,禮服店,酒店小姐,酒店經紀人,
等相關服務幫您快速的實現您的夢想~!!

a片,成人網站,美女交友,美女視訊,a片,成人網站,美女交友,美女視訊,a片,成人網站,美女交友,美...

2009-03-20T11:31:00.000-04:00

^^ nice blog!! thanks a lot! ^^徵信, 徵信社,徵信, 徵信社, 徵信...

2009-03-11T06:05:00.000-04:00

^^ nice blog!! ^@^徵信, 徵信, 徵信, 徵信社, 徵信社, 徵信社...

2009-03-11T06:04:00.000-04:00

情趣用品,情趣用品,情趣用品,情趣用品,情趣用品,情趣用品,情趣用品,情趣用品,情趣用品,情趣用品,...

2009-02-14T01:43:00.000-05:00

Nate noted: "The closest Kennedy states were ...Ha...

2008-07-11T10:23:00.000-04:00

Nate noted: "The closest Kennedy states were ...
Hawaii, Illinois, Missouri, New Mexico, New Jersey. So, we start with a 51-EV gap and begin whittling .. down. ... But it also gives him 12 extra electoral votes. So we go back through the list in reverse order and see if there is any wastage. ... This was the most efficient possible combination of states that would give him a winning electoral margin."

Why bother with the whole going back through the list. I can see doing that if you have a sample of one, i.e. election is perfectly correct simulation. But you are doing 10,000 simulation. I'm not sure there is a need to look in each simulation for the most efficient combination of tippers. The "tippers" will just emerge in the course of so many sims.

To say that Ohio WAS the tipping point state, I think misses the point that Paul Bradford made @July 10, 2008 9:55 PM. Flipping 40K voters efficiently in NM, IA, NV deny Bush a victory "more efficiently" than flipping 136.5K OH voters.

To summarize, in Version 1.x You looked for what state put the WINNER over the top. But I gather really what we want to know is what states the LOSERs needed to flip in a simulation in order to win. Hence, V 2.x.

So you need to order not by percentage closeness by number of votes closest. But that then might put DC and WY into play.
Solution:
Chop out the tails of safe Dems and safe Republicans learned from looking at totality of the simulations which gets rid of DC, HI, CA, UT, WY, etc. Then it doesn't matter about cutting out landslides.

Then order the left over middle by closeness of number of votes needed to flip.

Then flip until the winner is no longer a winner.

Two tables/maps one for each candidate.

Finally, separate out wheat from chaffe.

written as usual in haste without proofreading.

This is very close to what I was hoping for in the...

2008-07-11T01:01:00.000-04:00

This is very close to what I was hoping for in the way of a change to the tipping-point state analysis. I would be very interested to see these states broken down among the two candidates. For instance, I suspect that Ohio is much more often a tipping point state for McCain than it is for Obama, since McCain has far fewer paths to victory which don't include Ohio than does Obama. I suspect each candidate's tipping-point state map would look very different.

Paul Bradford,Interesting data. Thanks.

2008-07-11T00:08:00.000-04:00

Paul Bradford,

Interesting data. Thanks.

Tipping States Through the Years1960: This is int...

2008-07-10T22:55:00.000-04:00

Tipping States Through the Years

1960: This is interesting. While, as Nate described in the post, Nixon would have needed IL, MO and NJ to get the requisite 269 electoral votes to win the election outright, Nixon would only have had to flip MO (9,980 vote plurality) and IL (8,858 vote plurality) to prevent Kennedy from winning. Kennedy only won 303 Electoral Votes because 8 MS electors and 6 AL electors supported Harry Byrd for president. If you subtract IL (27) and MO (13) from Kennedy's tally you get 263 which is less than the number needed for a majority.

1968: Humphrey would have had to flip MO(12), NJ(17), OH(26) and IL(26) to raise his electoral tally from 191 to 272. But he could have stopped Nixon if he'd merely flipped MO(20,488 votes), NJ (61,261 votes) and AK (2,189 votes). This is because George Wallace won 46 Electoral Votes.

1976: If Ford had flipped OH(25) and WI (11) he would have won the election. In fact, if he'd flipped OH (11,116 votes) and MS (14,463) he'd have gotten 272 electors and the election.

2000: Duh! Gore needed only 537 votes in Florida to reach 291 Electoral Votes and win the election

2004: 136,483 votes in OH would have given Kerry 272 EV's. Of course, flips in NM (6,047 votes), IA (13,498 votes) and NV (21,567 votes) would have brought him to 269 -- which would have been good enough for a tie.

Isn't it interesting how the same states keep popping up. In the USA, some votes count a lot more than others!

> "A Tipping-Point state is now defined as a state...

2008-07-10T21:33:00.000-04:00

> "A Tipping-Point state is now defined as a states that would be most likely to alter the outcome of a close election if it were decided differently"

"Likely" is equivalent to a probability, but one thing that the computation of the Tipping Point States would seem to NOT be is a probability. In order to be a probability, it seems to me you would have to take the probability of getting to the base outcome and multiply that by the probability of getting to the "tipped" outcome. Since the base outcomes are all equally likely, the factor that has not been included is the probability that the combination of states that "tip" the election would actually tip it. Furthermore, it seems to me you would have to consider all combinations of states that could tip the election, not just the highest probability state or states. I think what you have come up with may just be a heuristic that roughly corresponds to the "battleground states", but it is not a measure of likelihood.

No, they now no longer have to add up to 100 perce...

2008-07-10T20:48:00.000-04:00

No, they now no longer have to add up to 100 percent. There can now be more than one swing state in a given election.

Just one question, Nate: aren't the percentages su...

2008-07-10T20:47:00.000-04:00

Just one question, Nate: aren't the percentages supposed to add up to 100%?

Nate,I would really enjoy seeing the mode shown on...

2008-07-10T20:28:00.000-04:00

Nate,

I would really enjoy seeing the mode shown on the Electoral Vote Distribution chart. I am wondering how sensitive that is to changes in other data, most notably winning percentage (thought I would say EVs didnt you?)

Nate,It is a well known fact that, ever since FL b...

2008-07-10T20:18:00.000-04:00

Nate,

It is a well known fact that, ever since FL became a state, the winner of every presidential election has been the candidate who has won at least two (2) states out of {FL, OH, PA}. There have been no exceptions. None. This predictor is better than the national popular vote.

Furthermore, there is no other group of three states for which the above statement is true. Therefore, in the scenario analysis I would like to see how this plays out in your simulation runs

Oh yeah - I'm looking forward to seeing how this a...

2008-07-10T19:53:00.000-04:00

Oh yeah - I'm looking forward to seeing how this affects the ROI chart.

Brilliant! I love it!!In the ROI thread, I'd sugg...

2008-07-10T19:52:00.000-04:00

Brilliant! I love it!!

In the ROI thread, I'd suggested ditching the landslides, but I'd proposed a simple cutoff point. This is much more elegant.

Very, very nice.

James,I agree that you would need to use another a...

2008-07-10T19:33:00.000-04:00

James,

I agree that you would need to use another approach if you didn't want to calculate the probability of each state flipping. In that case, Nate's approach is probably a reasonable heuristic for determining the states most likely to flip. However Nate has the tools to calculate the probability fairly easily.

If you look for the minimum number of voters that needed to filp, then you get some interesting results. For example, as you mentioned the winning combination in 2004 would have been NM + IA + CO, where flipping 57,785 voters would have done the trick. In OH, 59,388 needed to flip.

It's still hard to say that the NM + IA + CO combo would have had better ROI in 2004, because even if Kerry had managed to flip 60,000 voters in those three states, it would have been hard to make sure they were distributed properly to actually affect all three election outcomes.

jdk,

I really like the idea of using the similarity score to compare states to the nation.

1. There really needs to be separate maps one for ...

2008-07-10T18:02:00.000-04:00

1. There really needs to be separate maps one for McCain and one for Obama.

2. I'm still not clear what you are trying to get at. Is it the "Florida, Florida, Florida" of 2000 or the "Ohio, Ohio, Ohio" of 2004?

The problem with that is that it never was FL, FL, FL; it was always TN, TN, TN. And since TN was that close, TN may not have come up by this mathematical definition. But probablistically speaking if you don't win your home state -- you will almost certainly LOSE!

In other words, once TN goes against GORE the odds are so staked against him that needs a miracle to win.

It seems to me the better metric we are after is "battleground". If we look at the current map, what states does McCain have to flip back to red to win: IN 11, VA 13, CO 9, OH 20.

It seems silly to call MI and PA tipping states when they're already tipped!

One thing that your State Similarity thing could have used was to take the Nation as a State and say what States are most similar to it. Maybe you did this but didn't post.

I used a little bit different State Similarity method based on 2004 results and my "special sauce" Ancestry (upon reflection, I certainly don't think geographic lat/long should be used; I have some concerns about overfit you've got quite a few variables, and DC, HI, and UT are just special cases).

In any case, my "closest" States to the Nation are FL, NV, CO IL, VA, MI, MO. Which might be interesting

I think that you have to chop of the tails of safe Dems and safe Reps, and only look at close elections in terms of those states, in other words don't let DC, CA, NY, IL, MA, etc. make things seem closer because the "wasted" margin of victory.

Modeler,I understand what you're saying and at lea...

2008-07-10T17:34:00.000-04:00

Modeler,

I understand what you're saying and at least from the point of view of perceptions you're absolutely right. No one looks back on the 2004 election and says "if only we'd won Iowa, New Mexico, and Colorado." <- though Colorado was less close than Ohio, so by Nate's metric this wouldn't be true anyway.

I think the fundamental difference in how we're looking at this comes down to:

"For every state you should have the expected result and the variance. From these you can estimate the probability of the state voting for the losing candidate using an error function."

The way I'm looking at the most likely way to swing is looking at only the data on how close the states electoral outcome was in that particular simulation. You're looking at the actual outcome and the variance in other simulation (or looking at it another way, you're looking at the formula used to generate the outcome in that particular simulation). With info on the variance you can calculate probabilities. Looking at just the results without the formula, I think the best proxy is number of voters, or alternatively percent margin of victory.

I remain amazed that's it's possible to have such interesting and civil discussions on this site.

If NC goes Obama, we're essentially sure that VA h...

2008-07-10T17:20:00.000-04:00

If NC goes Obama, we're essentially sure that VA has already gone for Obama as well ...

Yes.

... and furthermore that the election as a whole has tipped so far to Obama that we're confident he's already won it, with or without NC's electoral votes.

No. This would suggest that all movement would be felt equally across all states -- we simply line them up a single axis from most Obama-friendly to least, and allow for movement along that axis.

But that misses the whole point of the state correlations. Suppose Obama's support collapses among non-Evangelical white working class voters, but he gains a lot of support among African Americans and white Evangelicals -- he could lose OH, MI, WI, but win VA, NC, and even GA. The first group are demographically similar, and we can expect them to move somewhat together, with OH turning red before WI does. Likewise VA, NC, and GA will likely move together, with VA turning blue first. But the two groups are likely to be relatively uncorrelated with each other.

How to implement this into the model? Currently, for each simulation a random number is generated for national movement (applied to all states) and an independent random movement is applied to each state. Instead, the 50 state terms should have a correlation structure based on their demographic similarities. Transform an independent random vector by a covariance matrix constructed from the distance measures that Nate has computed, I believe.

Rasmussen just posted: Obama 50, McCain 39 in Wisc...

2008-07-10T17:00:00.000-04:00

Rasmussen just posted: Obama 50, McCain 39 in Wisconsin (leaners, O 52, M 42)

http://www.rasmussenreports.com/public_content/politics/election_20082/2008_presidential_election/wisconsin/election_2008_wisconsin_presidential_election

Continuing this train of thought a little further....

2008-07-10T16:33:00.000-04:00

Continuing this train of thought a little further...

You could actually consider multiple "Minimum tipping combinations" for each run, and weight the states in each combination by the probability that they flip. So in my example, MT, NH, and NM would all have weights of 6.4, and MO would have a weight of 10.

The remaining question is "How do you calculate the probability of a state flipping?" For every state you should have the expected result and the variance. From these you can estimate the probability of the state voting for the losing candidate using an error function.

Good news for Obama in latest Pew Poll.It shows hi...

2008-07-10T16:27:00.000-04:00

Good news for Obama in latest Pew Poll.

It shows his supporters are much more likely than McCain to be "strong supporters".

James,You're certainly not speaking nonsense (as f...

2008-07-10T16:21:00.000-04:00

James,

You're certainly not speaking nonsense (as far as I'm concerned). If I understand you correctly, you are proposing the following metric:

1) Consider the minimum number of voters that would have needed to change their votes to alter the election.

2) Label the states in which those voters reside "tipping point states."

That would be a cool metric, and it would combine elements from both the Tipping Point and ROI analyses. It might actually favor small states because of the high EV/voter ratios in those states, as you pointed out. If that is Nate's intent, then he should be able to calculate it explicitly using estimated voter turnout. Maybe he can call it ROI v2.0.

However, with "Tipping Point" states I think Nate is trying to get at the combination of states most likely to have have changed the outcome of the election. It would be the the combination for which the losing side can most reasonably say "If we'd only won X, we would have won." In my example, it would be more realistic for them to say that about MO than the NH-MT-NM combo.