This blog is still alive, just in semi-hibernation.
When I want to write something longer than a tweet about something other than math or sci-fi, here is where I'll write it.

Sunday, September 30, 2012

Sunday numbers, 9/30, part b:
The Senate races


I call my system Confidence of Victory, but even I am not completely confident the good news for the Democrats is as good as it looks. Unlike the presidential races, the polling in the Senate is somewhat sparse and a new poll in a state where the most recent poll is eight days stale is treated as gospel.  A lot of Senate races have been announced by polling companies specifically designated as Democratic, and what's worse, these are polls with very little track record. (In contrast, PPP is a Democratic polling company, but they are as busy as beavers, and I have some idea of how much they skew leftward, which is to say not that badly.)

In any case, I publish these numbers. The median result for the Dems is 54 seats before Angus King, still the very likely winner in Maine, make his decision in which caucus he will sit. There is now a tiny chance of 58 in the Democratic caucus, but even King would not give them a filibuster proof majority under the current rules. The odds of at least 50 Democratic seats is now a comfortable 99.94%.

These numbers could change before the election and I certainly expect them to change some by next Sunday. Stay tuned.


Sunday Numbers, 9/30, Part a:
The Electoral College

Let me put this nicely for fans of Romney-Ryan.

Umm... September is over? That's good because this month has sucked something fierce.

The bad news? Well, obviously... November is coming.

Here's the state-by-state probabilities as of this Sunday using my Confidence of Victory system.

Solidly Republican.
AK 3 AL 9 AR 6 GA 16 ID 4 KS 6 KY 8 LA 8 MS 6 OK 7 SC 9 SD 3 TX 38 UT 6 WV 5 WY 3
EV total = 137

===
Leaning Republican.
IN 11    0.9998
NE 4    0.9995
MT 3    0.9905
TN 11    0.9806
ND 3    0.9991
MO 10    0.9589
AZ 11    0.8301
EV total = 53

===
Coin flips.
N2 1    .5000
EV total = 1

===
Leaning Democratic.
NV 6    0.7391
NC 15    0.7466
VA 13    0.7821
IA 6    0.8538
FL 29    0.8689
CO 9    0.8714
NM 5    0.9537 GOLD
NH 4    0.9596
PA 20    0.9669
WI 10    0.9806
OR 7    0.9870
OH 18    0.9880
MN 10    0.9922
MI 16    0.9971
EV total = 168

===
Solidly Democratic.
CA 55 CT 7 DC 3 DE 3 HI 4 IL 20 MA 11 MD 10 ME 4 NJ 14 NY 29 RI 4 VT 3 WA 12
EV total = 179


No contest changed from the Republican side to the Democratic side, or even changed for Solid to Leaning.  All the changes were in the leaning categories, and the most important were Democratic leaning states leaning more Democratic. Ohio is now farther out of reach than Pennsylvania, and Pennsylvania already looked nearly a lock if the election were held today.



Unlike McCain-Palin, Romney Ryan has NEVER been in a flat-footed tie with Obama-Biden. Sarah Palin was a real sugar rush for the Republicans. Paul Ryan... not so much.


The median result is now 336-202. That's 4 electoral votes worse than last week.  Even worse, the most likely results, all landslide victories for the Democrats, got more likely still.

348 or 347 EV for Obama/Biden: 18.5% (was 12.4%)
341 or 342 EV for Obama/Biden:  10.4% (not a top choice last week)






The real problem is 47%. Romney was behind when that tape was released, and since then he has lost ground when there didn't even didn't look to be ground to be lost. The chance for an Obama victory last week was 99.9%. This week it would round to 100% (if the election were held today, never forget that) if I only rounded to a tenth of a percent, so I round to a hundredth of a percent and get 99.97%

Let me repeat. These numbers do NOT predict the result on November 6. Nate Silver says he can do that and that's one of the reasons I look down my nose at him. Polls are NEVER meant to predict the future.  That said, when late October rolls around and every state's early voting procedure kicks in, it may become too late for Romney-Ryan to turn this around.

Back next Sunday with newer data. Stay tuned.


Saturday, September 29, 2012

I gotcher unskewed polls RIGHT HERE!

A fellow in Virginia names Dean Chambers made news with his idea that the polls were all messed up.  Romney couldn't possibly be behind, could he?  Chambers thought it impossible, so he came up with his own data sets and called the UnskewedPolls.com. Right wing news sites paid attention and the rest of the media followed suit. The difference, of course, is that the right considers him a hero and everyone else - including Scott Rasmussen - considers him a boob.


Well, since all the cool kids are doing it, I decided to unskew some polls as well, though unlike Mr. Chambers, I have some basic understanding of mathematics.  Let's take the daily tracking polls, Gallup, Rasmussen and Reuters/Ipsos, and consider September 1 the starting date.  They did not agree on this date, but I skew the results to make this the baseline.  Being a dirty Commie, movement up shows improvement for Obama and movement down shows improvement for Romney. The main takeaway from this way of showing the data is that all these polls think Obama has gained about 5% or 6% on where he was at the beginning of the month.

They all agree that things started looking up on September 5, the date of Bill Clinton's speech. Gallup and Rasmussen thought that was a bounce, Reuters thought it was the start of a trend.

The death of the Libyan ambassador prompted Romney's comments on Sept. 12, but the polls disagree on whether those comments helped or hurt.

Romney's hidden camera tape is released on Sept. 17 and takes a few days to be processed by the entire news media. Ten days later, all the polls agree that Romney was in better shape before the public heard him speak candidly to the people who give him massive bundles of cash.


And just because I hate hiding the objective truth, here how the polls have actually behaved in September. Rasmussen gave Romney a 4% lead on September 1, while Gallup and Reuters/Ipsos thought Obama was ahead by 1%.  Now, Rasmussen has Obama slightly ahead while the other two think it's a very comfortable lead of 6% to 7%. The yellow line shows the median, which is the polling result I trust the most.

Once again, let me say that daily tracking polls are much less useful than keeping tabs of the electoral college vote, which is the real way we choose a president. It is possible to win a landslide and lose the electoral vote, but that's never actually happened. With the exception of Bush-Gore, the electoral victor has won the popular vote for well over 100 years, When it happens the popular vote winner usually has less than a percentage point lead, with the exception of the 1876 election, the stinkiest voting scandal our country ever saw, even worse than Bush-Gore.

This does not mean the election is over. The most these numbers really say is what every honest person is telling you right now. Romney is behind and the trend is going against him. He still has more than a month to pull out of this tailspin, but for all the money that is being spent, he's having a hell of a time getting any traction and opening his mouth is not helping right now.

Electoral college numbers tomorrow. Stay tuned.


Sunday, September 23, 2012

Sunday Numbers, 9/23, part b:
The Senate numbers

Politics is personal. Whatever the positions of the two parties on the issues, it often comes down to the particular man or woman on the ballot for the particular office.  Many Republican Senatorial candidates are much better politicians than Mitt Romney. Many Democratic candidates are not as good at politics as Barack Obama.

But sometimes the guy at the top of the ticket opens his mouth and gets the whole party in trouble.

This was one of those weeks.

The median number of seats for the Dems is right now 53 before Angus King, the independent candidate in Maine who is still a prohibitive favorite, decides with whom he will caucus. Last week it was 51.

The odds of the Dems having at least 50 seats is 99.6%, up from 91.4% last week.

The glimmer of hope for the Republicans is that the Democrats do not appear to have any chance at present of getting a filibuster-proof 60 seat majority. What makes that glimmer a little dimmer is that the Republicans have abused the filibuster so much since Scott Brown was elected to take Teddy Kennedy's seat. There is a lot of talk of filibuster reform. Notably, it is up on Angus King's website, so even if he caucuses with the Republicans, he can't be counted to be in lock step.

The other conservative glimmer is the Republicans currently hold the House.  I have a way of predicting the House if there is enough data, but right now I have info on only 20 of 435 races, which isn't anywhere near close enough for a prediction.  I've seen a tweet from Nate Silver that he feels much the same way, so the House outcome is likely to be the great suspense of election night, barring some great tidal wave of data.

Stay tuned.

Sunday numbers, 9/23, part a:
The electoral college

A quick recap for who might be coming here for the first time. I kept track of the race for quite a while, but decided the selection of Paul Ryan was a good time to start fresh.

Like shaking an Etch-A-Sketch, if you will.

The thing is, Romney/Ryan did not start on an even footing with Obama/Biden. Part was the advantage of incumbency, even in an awful economy, and part was the thing a lot of people are coming to grasp, regardless of ideology: Mitt Romney is not a good candidate.

In any case, Mitt had some good weeks, Mitt had some bad weeks. The problems were fourfold.
1) He was behind.
2) Time is finite.
3) His good weeks were modest gains.
4) His bad weeks were drastic drops.

Here's how things shook out this week.
Solidly Republican.
AK 3 AL 9 AR 6 ID 4 KS 6 KY 8 LA 8 MS 6 OK 7 SC 9 TX 38 UT 6 WV 5 WY 3
SD 3  was Leaning Republican
GA 16 was Leaning Republican
EV total = 137

===
Leaning Republican.
ND 3    0.9995
NE 4    0.9995
AZ 11    0.9968
TN 11    0.9806
MO 10    0.9627
IN 11    0.9504
MT 3    0.9059
EV total = 53

===
Coin flips.
N2 1    .5000 Never polled before, considered Leaning Republican
EV total = 1

===
Leaning Democratic.
NC 15    0.6311 was Leaning Republican
NH 4    0.6870
CO 9    0.6915
NV 6    0.7991
OH 18    0.8467
FL 29    0.8847 GOLD
VA 13    0.9436
WI 10    0.9450
IA 6    0.9603
MN 10    0.9806
OR 7    0.9870
MI 16    0.9878
PA 20    0.9918
NM 5    0.9996
EV total = 168

===
Solidly Democratic.
CA 55 CT 7 DC 3 DE 3 HI 4 IL 20 MA 11 MD 10 ME 4 NJ 14 NY 29 RI 4 VT 3 WA 12
EV total = 179



Some Democratic partisans might feel trepidation that the Dems need to win Florida to maintain the presidency. Note that the Republicans need North Carolina, New Hampshire, Colorado, Nevada, Ohio AND Florida to pull off the victory and they currenly trail in all those places according to the median polling results. They also lost all those states in 2008



Simply put, last week was a bad week for Mitt, and last week the numbers were worse than any Romney/Ryan had seen. The median result has risen to 332 to 206.

Here's some worse news than that for Romney/Ryan.  The median result is not the most likely result. Because someone has FINALLY polled Nebraska district by district, the Omaha electoral vote looks like a toss-up, though Romney looks likely to win the rest of the state.  This means that the most likely results are in pairs one EV apart.

348 or 347 EV for Obama/Biden: 12.4%
333 or 332 EV for Obama/Biden:   9.7%
344 or 343 EV for Obama/Biden:   6.5%

If this were the Sunday before the election, I'd choose the top number as my prediction. But this is NOT a prediction because the election is too far away.  This is just saying what the polls look like right now.

Note: Obama beat McCain by even more than this because he took Indiana - where he now trails badly - and the red states took a few EV away from the blue states in the last census.




And now comes the probability of victory if the election were held today.

Obama: 99.92%
Romney: 0.08%

There appears to be a lesson in this week's numbers, though I admit I'm a mathematician and not a political scientist.

Shitting on 47% of the population is not a winning strategy, especially when many of them support you.

Tuesday, September 18, 2012

The tracking polls this month.

 Gallup, Rasmussen and Reuters/Ipsos are all doing national tracking polls. Here are the results of this month up through yesterday, Sept. 17. All to polls agree that things turned in Obama's favor as of Sept. 5, the day of Bill Clinton's speech, but Rasmussen and Gallup think it was a bounce, while Reuters/Ipsos thinks it's a trend.

I still think tracking polls are a lot of math and hard work for not much purpose, not unlike a very accurate horoscope. The main use is to calibrate polling companies against others, and here we can see Rasmussen is almost always the most right leaning of the companies, sometimes to a very noticeable amount. If we take the median of the three, we see a bounce instead of a trend.  Looking at the individual swing state polls gives us a very different view, and to my mind a more accurate one.


Sunday, September 16, 2012

Sunday numbers 9/16, part b:
The Senate races


There's good news for the Democrats in this week's polls of the Senate. The median result is now 51 seats from the Dems, up from 50 last week.  This is before Angus King of Maine, an independent who appears to be a shoo-in, decides on which party's caucus he will join. In other good news, King's website states that he thinks the filibuster has been sadly misused by the Senate and he would like to see changes in it. This sounds like a position that aligns him much more with the Democrats that the Republicans at this point in history.

In a continuation of the good news for the Democrats, the odds of at least 50 seats has risen to 91.4%, up from 71.4% last week and 54.6% the week before.  Again, let me re-iterate that these numbers always carry the proviso "if the election were held today".

More polls analyzed next week. Stay tuned.

Sunday numbers 9/16, part a:
The electoral college

Last week saw the polls reflecting the bounce from the Republican convention. It may seem like an eternity now, but all the polls that were counted last Sunday were taken before Bill Clinton's speech at the Democratic convention.  In this week's,  all the movement from one category to another was in favor of the Obama-Biden ticket, and that movement was significant.

Solidly Republican.
AK 3 AL 9 AR 6 ID 4 KS 6 KY 8 LA 8 MS 6 OK 7 SC 9 TX 38 UT 6 WV 5 WY 3
EV total = 118

===
Leaning Republican.
IN 11    0.9998
ND 3    0.9995
NE 5    0.9995
AZ 11    0.9981 was Solidly Republican
GA 16    0.9978
TN 11    0.9804
SD 3    0.9285
NC 15    0.9150
MT 3    0.9059
MO 10    0.7568
EV total = 88

===
No coin flips.

===
Leaning Democratic.
FL 29    0.6568
CO 9    0.6915
WI 10    0.7594
IA 6    0.7690
NV 6    0.8102
NH 4    0.9069 GOLD
VA 13    0.9478 was coin flip
OH 18    0.9516 was coin flip
MN 10    0.9806
OR 7    0.9858
MI 16    0.9886
CT 7    0.9969
PA 20    0.9980
NM 5    0.9996
EV total = 160

===
Solidly Democratic.
CA 55 DC 3 DE 3 HI 4 IL 20 MA 11 MD 10 ME 4 NY 29 RI 4 VT 3
NJ 14 was leaning Democratic
WA 12 was leaning Democratic
EV total = 172


The median result as of this evening is 319 electoral votes for Obama. As always, let me add that this is NOT a prediction of what will happen on election day in November, but a prediction of what would happen if the election were held today.

If the five weeks since the introduction of Paul Ryan, the GOP ticket has had three good weeks and two bad ones.  The problem is that they started out well behind, the good weeks have been modest and the bad ones severe.  We are two weeks into the ad onslaught and they haven't moved the needle for GOP ticket. There's still plenty of time, but right now Romney-Ryan has no traction.
 

If all this wasn't bad enough, the odds for victory have climbed to 99.6% for Obama, which says if the election was held today, Romney would be a 250-1 longshot. This is much worse than things were for John McCain at the same point in 2008, but Sarah Palin is very different from Paul Ryan. Her introduction was a much bigger shot in the arm and the eventual disappointment in her abilities much more pronounced than any reaction to Paul Ryan the electorate is likely to have.  It's very likely Romney will get more electoral votes than McCain did, but it will take some doing to bring Obama under 300 electoral votes, let alone under 270.

Next update a week from now. Stay tuned.



We missed a Caturday.


One of my most reliable blog buddies, Victor over at Cat In The Bag, has not been feeling well. He posts nearly every day, but he's missed over a week.  Please take a moment to head over to his blog and leave your best wishes.  You might want to do the same at his wife's blog, Zoolatry.

Get well soon, buddy. We all miss you.

Sunday, September 9, 2012

Sunday numbers 9/9, part b:
The Senate races


For several weeks, the trend has been good for the Republicans in terms of the Senate races. For the first time in a month, the trend moved slightly in Democratic favor. The numbers for this week are given by the thick purple curve, slightly farther to the right hand side of the picture, which means better for the Democrats. The median number of likely seats in the Democratic caucus is still only 50, which means Angus King of Maine, nominally an independent and not yet decided as to which caucus he will join, could be very important. The good news for the Democrats is that the odds of getting 50 or more seats is now at 71.6%, up from 54.2% last week.  This is due in large part because the odds of 51 Democrats is now more that the odds of 49 Democrats.

I'm going to be presenting the numbers as several curves, with the most recent curve much bolder than the others and the last two curves have the check marks at each data point.  Let me know if the graphic is clear or not in the comments.

More numbers next week, the first week after both conventions and the start of the serious media buys brought to us by the Citizens United decision by the Supreme Court.

Stay tuned.

Sunday Numbers 9/9, part a:
The Electoral College

The electoral college polls this week were taken almost entirely before Bill Clinton's speech on Wednesday, so any movement should be seen as the bounce from the Republican convention.  The movement was all to the good for the Romney-Ryan ticket, with Ohio going into the toss-up column and New Jersey bouncing from solidly Democratic to leaning Democratic. To be fair, the New Jersey change was not much, but in my system the Confidence of Victory has to be over 99.99% to be counted as solid, and New Jersey fell to 99.71% chance of an Obama victory if the election were held today.

With further ado, this week's list with changes noted. (Both changes were pro-Romney.)

Solidly Republican.
AK 3 AL 9 AR 6 AZ 11 ID 4 KS 6 KY 8 LA 8 MS 6 MT 3 OK 7 SC 9 TX 38 UT 6 WV 5 WY 3
EV total = 132

===
Leaning Republican.
IN 11    0.9998
ND 3    0.9995
NE 5    0.9995
GA 16    0.9978
MO 10    0.9656
TN 11    0.9804
SD 3    0.9285
NC 15    0.7708
EV total = 74

===
Coin flips.
OH 18    .5000 was leaning Democratic
VA 13    .5000
EV total = 31

===
Leaning Democratic.
FL 29    0.6568
WI 10    0.7594 GOLD
NH 4    0.7659
IA 6    0.7690
NV 6    0.8102
CO 9    0.8350
MI 16    0.8570
MN 10    0.9362
PA 20    0.9756
OR 7    0.9858
WA 12    0.9937
CT 7    0.9969
NJ 14    0.9971 was solidly Democratic
NM 5    0.9996
EV total = 155

===
Solidly Democratic.
CA 55 DC 3 DE 3 HI 4 IL 20 MA 11 MD 10 ME 4 NY 29 RI 4 VT 3
EV total = 146


While the movement early in the week was in the Republican direction, the median electoral college result right now is Obama 301, Romney 237.  The GOLD label shows the easiest path to victory, so right now Romney would need to win all the states where he is favored, the current toss-up states Ohio and Virginia, and then both Florida and Wisconsin, states where Obama currently is favored.


Winning all those battles would be hard as the numbers stand right now, and it shows in the current probability using the Confidence of Victory numbers. My computer program chugged through two to the twenty fourth power possible outcomes - a tidy little 16,777,216 variations - and gives Obama a 91.0% chance of being re-elected if the election were held today.





While I have serious qualms about the usefulness of national tracking polls - you'll recall Gore won the popular vote in 2000 - it is worth noting that since Clinton's speech all three of the daily trackers have been trending towards the Democratic side in a big way. Gallup marks Obama's lead at 5%, while Rasmussen and Reuters/Ipsos give him a 4% lead.  We will see if these numbers translate to a good week in the electoral college count for the Democratic ticket when I publish the new numbers next Sunday.

Stay tuned.

Saturday, September 8, 2012

A little polling wonkery:
PPP vs. Rasmussen

There are a lot of polling companies out there and some tend to skew left and others right. Given the new paradigm that says everyone IS entitled to their own facts, it only stands to reason that the two busiest polling companies in the business are PPP, a Democratic polling company, and Rasmussen Reports, which does not call itself a Republican polling company, even though Scott Rasmussen is the go-to guy for spin on polling numbers on Fox News, which does not call itself the propaganda arm of the Republican Party, though that is the conventional wisdom.


As of this Saturday, looking at the polls I'm using to produce the Confidence of Victory numbers I report every Sunday, PPP is the somewhat busier bee of the two. There are some states with very little polling data, so I use the most recent poll, ignoring polls that were many weeks old when the new data came out. PPP is the poll of record in six of these states, while Rasmussen is the poll of record in four of these states.


There are also states where there has been so much polling I use the median poll as the poll of record.  A poll can either be the leftmost, the median or the rightmost in these cases.

Leftmost cases: PPP 5, Rasmussen 2
Median poll: PPP 3, Rasmussen 1
Rightmost cases: PPP 1, Rasmussen 3

So as we might expect, Rasmussen polling methods lean towards the conservative side and PPP leans to the liberal side. PPP is in the middle about 20%, while Rasmussen only hits the middle 10% of the time. PPP is not one of the companies that produces a daily tracking poll and I commend them for it. The popular vote doesn't decide the presidency, as anyone whose memory can stretch back twelve years will recall.  Rasmussen, Gallup and Ipsos do the daily tracking, and Rasmussen is often the odd poll out to the right.  Besides the waste of energy of producing daily tracking, Rasmussen has decided on a policy of not calling people with cell phones, which means they miss the growing part of the population without landlines.

All of this is commentary way too early to tell what will happen. The pudding gets proved in November and if we use the rules of polling, the polls today only give us a snapshot of the situation now. I will do a post-mortem after election day, but I will only pass judgment on the polls produced within a week or so of election day.

Stay tuned.

Sunday, September 2, 2012

Sunday numbers, 9/2 part b:
The Senate races


If conservatives are looking for a ray of bright light, it exists in the numbers for the Senate races from this week. The movement in the polls for the past three weeks has been a steady march towards a more even divided Senate.  Angus King in Maine is running as an independent and looks to be a shoo-in, but he has not declared which party he will caucus with as of yet. Right now, the Democrats have only a 54.2% chance of having 50 seats without King, down from 69.2% last week.

People on different sides of the spectrum believe the media has a bias one way or the other, but the overall biases of the media are for new things, controversial things and close races.  Right now, Obama-Romney is NOT a close race, but control of the Senate is, and it's getting closer each week.

Another report next Sunday.  Stay tuned.

Sunday numbers, 9/2, part a:
The Electoral College

Yet another week has drifted by and the numbers as judged by my methods have made a change in the likely results of the Electoral College. After two fairly good weeks for Romney-Ryan, the week of their convention was a disaster, wiping out all the gains they made and even digging a bit of a hole from where they started. Here are the states and how they currently stand, using a combination of latest poll number when there are very few polls and the median polls when there have been at least two in the past week.

Solidly Republican.
AK 3 AL 9 AR 6 AZ 11 ID 4 KS 6 KY 8 LA 8 MS 6 MT 3 OK 7 SC 9 TX 38 UT 6 WV 5 WY 3
EV total = 132

===
Leaning Republican.
IN 11 0.9998
ND 3 0.9995
NE 5 0.9995
GA 16 0.9978
TN 11 0.9804
MO 10 0.9656
SD 3 0.9285
NC 15 0.6118
EV total = 74
===
Coin flips.
VA 13 .5000
EV total = 13

===
Leaning Democratic.
WI 10 0.7594
NH 4 0.7659
IA 6 0.7690  was Leaning Republican
NV 6 0.8102
OH 18 0.8149
CO 9 0.8247  GOLD
MI 16 0.8570
FL 29 0.8610 was Leaning Republican
MN 10 0.9362
PA 20 0.9756
OR 7 0.9858
WA 12 0.9937
CT 7 0.9969
NM 5 0.9996
EV total = 159
===
Solidly Democratic.
CA 55 DC 3 DE 3 HI 4 IL 20 MA 11 MD 10 ME 4 NY 29 RI 4 VT 3
NJ 14  was Leaning Democratic
EV total = 160

Some states changed slightly but did not jump to a different category.  The three big moves were New Jersey going from Leaning to Solidly Democratic, and Iowa and Florida flipping for the Republican column to the Democratic column. As always, let me say that these results assume the election was held on the same day as the poll was taken.  There are still two months to go and an amount of money never before seen in going to be pumped into this election on both sides.


The median result right now is Obama getting 316 electoral votes and Romney getting 222. The Republican National Convention "bounce" has not yet been fully seen, nor has the "bounce" from the Democratic Convention, if any.  A lot of Americans are tuning out, realizing the conventions are pretty damn meaningless these days.




My program also tracks the probability of victory for each party given the current polling. Right now, Obama-Biden has about a 98.2% probability of victory according to the program I ran, taking into account two to the twenty third different combinations, which is 8,388,608 for all you "power of two" fans out there.

More numbers next Sunday. Stay tuned.