Facebook is the New Rasmussen (in 2-4 years)

Just before the election for MN Senate, the Degausser pointed out that Franken had a handy lead in Google Search trends in Duluth that could translate to a win for him in the area. After doing a little bit of research into the emerging world of political analytics, I have seen the future. And man, is it scary/fascinating/scary.

Just a quick primer: analytics is the process of using internet data such as searches, page views, and unique visitors to identify trends. These data are being refined in the business world to better market products to buyers. In the political realm, these data will become the new polling numbers.

In 2006 mid-term elections, researchers looked at candidates Facebook entries and created models that equated supporters to voters. They found that challengers were more likely to use Facebook as an organizing tool and that it could potentially add 1%-3% to the candidate's total vote.

Analytics are just beginning to be used by political scientists and campaigns. Most of the data is scattered and more research needs to be done to discover total impacts, but in a few election cycles, the internet will become (or maybe it has already) a 24/7 polling place.

Just for fun, let's look at some data collected from Al Franken's and Norm Coleman''s web presence over the course of 2008.

NOTE: Data presented here are from January 1, 2008 to November 2, 2008.

I collected Google Search trends, Facebook supporters, Youtube Channel subscribers, and basic website visitor stats collected from compete.com to see if there were any interesting patterns happening.
Throughout the course of the election, search trends were pretty similar between Franken and Coleman. However, right before the election, more people started searching for Franken at rates far beyond Coleman's. The graph above represents average search volume from Jan - Nov. 2. Anything above 1 represents larger than normal search volume for this period. So, by Nov. 2, there was a 24% larger search volume for Franken than Coleman. It is unclear what affect, if any, these searches had on voting outcome.

Looking at the top 8 MN cities by search volume, Franken lead in each except Prior Lake.


Election Results

	Franken	Coleman
Rochester	48%	51%
Minneapolis	69%	19%
St. Cloud	54%	42%
St. Paul	61%	25%
Hopkins	57%	38%
Mankato	60%	39%
Prior Lake	38%	57%
Duluth	62%	37%

There appears to be a relation between final election results and the volume of searches.

Looking at search trends through the election season and comparing them to the average of the polls conducted through the same time show even more relation. The last data point for the poll line is the final election result.

One interesting point however, is that there is a statistically significant and positive relationship between search volume and poll results. The more search volume, the higher the poll numbers.





It's obvious that Franken had a larger web presence than Coleman, but it is unclear how this presence translated into votes. For now, look at the data and make your own conclusions.

All of the Floridian Drama, Without the Warm Weather

Minnesotans have backstage passes to one of the greatest political dramas since the 2000 Presidential race. Interestingly, this drama is seemingly more a product of weak candidates than a great partisan split. Recent articles in the Strib and on MPR suggest that many people who voted Obama did not cast a vote for any candidate in the Senate race. This non-decision means that either there was a mistake made in the marking of the ballot or there was simply no vote made. Either way, election officials are determining the intention of each ballot. Here’s some map pron via the AP that shows the concentrations of people that opted out of voting in the Senate race. St. Louis Co. leads the way. Way to go Northland, we’re finally #1 in something.

A good portion of the attention from this ‘no-vote paradox’ is being focused on St. Louis County. Coleman’s campaign recently challenged 100 ballots coming out of Mt. Iron that all went for Franken. Coleman’s campaign manager also called the recount results from the Iron Range ’statistical dubious’. Is there some trickery going on up north or is there some other explanation? Let’s do a cursory analysis (insert cheers or jeers).

Last week The Degausser wrote about how the Northland delivered heavily for Franken. St. Louis County, the County that the Coleman campaign is questioning ballot totals from, went to Franken by 22 points. However, the totals from last week to today have changed as election officials are certifying results. Look at how things have changed. Franken has gained 395 total votes from Lake, St. Louis, and Carlton Counties. Coleman actually lost 2 votes.

Votes Gained or Lost Since Nov. 5
	Franken	Coleman	Barkley
Aitken	-5	-2	0
Carlton	54	0	0
Cook	0	0	0
Itasca	0	0	0
Koochiching	0	0	0
Lake	246	0	0
Saint Louis	100	0	0
TOTAL	395	-2	0

We can see how Coleman’s campaign manager would be upset, but what are the probabilities of evil shenanigans?

The precinct in question is Mt. Iron P1. According to the MN Secretary of State, there were 891 registered voters on Nov. 4 in this precinct. Here are the breakdowns:

Mt. Iron P1	Franken	Coleman	Barkley	TOTAL
New	506	211	95	812
Old	406	211	95	712
New %	62%	26%	12%
Old %	57%	30%	13%

The election was never in doubt for Franken in this precinct, he was up nearly 2 to 1 before the re-count. Also it appears that that the 100 vote gain was simply human error in totalling rather than lost votes being found, which changes the probability of fraud.

For arguments sake, however, let’s assume the new votes were found votes. What would the probability that they would all be Franken votes? Here is what we need to know to do the calculation

• The number of votes left (179) = Total registered voters (891) - Total votes cast (812) + 100
• Changed votes = 100
• Probability votes going to Franken = 57% (original win percentage)
• Probability votes going to Coleman = 43%

The expected score out of these assumptions, according to a binomial probability equation is 102. This means that the 100 votes Franken got is actually less than the statistical expectation.

One thing to be aware of is that this assumes that the Mt. Iron P1 district has a normal distribution of partisanship. Considering the heavy Obama vote and long standing DFL affiliations in the area, partisanship is not split 50/50. Although the 100 vote windfall was a calculation error and not found votes, numbers suggest Franken would have fared better if the ballots were found in the trunk of somebody’s car and entirely within the realm of possibility given the political make-up of the precinct. So, yes, the 100 vote appearance is strange, but surprisingly, it is statistical aberration that did less damage to Coleman than odds would suggest.

The Duluth News-Tribune is Wrong OR Why Forrum Communications Cannot be Trusted

I was reading my hometown paper just now and they are saying that Northland Dems dissed Franken. In it, they posit that the Northland didn’t deliver for Franken and that Barkley played spoiler. They also say that if just 600 more people would have voted for Franken, everything would be different. I disagree with their assesment.

The DNT suggests that because 77% of St. Louis County went for Obama and only 55% went for Franken that Franken failed. Comparing Obama to Franken is like comparing a Ferrari to a Yugo. No way is this an apples-to-apples comparison. Franken’s campaign never really got its shit together and is lucky to be where they are now. As the Degausser showed yesterday, he still managed to destroy Coleman in the Northland. This shouldn’t have happened and it is more a testament to people hating Coleman more than liking Franken.

If the DNT wants to blame somebody, which is retarded because Coleman shouldn’t have made it such a close race, they should blame Hennepin or Ramsey County voters. They voted less for Franken, had less turnout, and have a higher concentration of voters. There were more opportunities to win in the Metro-Area than in the Northland. Franken didn’t pull it out, even in a heavy DFL/Dem year.

If Ciresi was the DFL Senate candidate, he would have pulled more votes up North and probably would have won the election outright.

I have a sneaking suspicion that this article in the DNT is Forrum Communication’s way of saying that the Northland is moving more into a Republican/Independent mindedness. This has been Forrum Communication’s stated mission for the newspaper ever since they took control of it. Forrum Communications endorsed Coleman and owns 22 papers in MN. Moreover, their endorsement was made on the basis of one man’s (Steve McLister) opinion and forced onto the opinion pages of all affiliates. Could it be that Forrum Communication’s endorsement swayed enough voters for Coleman? Instead of blaming the Northland, the DNT should look at their own.

Why Franken Should Kiss Duluth’s Ass

Yesterday on the Degausser we suggested that MN’s Arrowhead region could tip in favor of Franken. Boy howdy did it.

Because of the impending recount, all results reported here are technically unofficial, but Franken carried the region by 18 points (33,560 votes).

For those of you metro-centric folks who refer to anything outside of Minneapolis/St. Paul as outstate, here’s a map of the Arrowhead region (shaded in red). The counties that make up this region are Aitken, Carlton, Cook, Itasca, Koochicing (Kooch for those in the know. Or, if you want to get fancy, combine it with Itasca to get Kootasca), Lake, and Saint Louis. The major city here is Duluth (my home town, woot!).

Anyway, this region is reliable DFL territory. However, due to Franken’s rocky start and lack of strong ties outside of the metro-area, it was up in the air as to how he would perform in the North Country. Apparently, Franken went over well. Here are the numbers:

	Franken	Coleman	Barkley
Saint Louis	64,319	38,267	14,275
Lake	3,304	2,434	863
Cook	1,620	1,206	458
Carlton	9,517	6,099	2,555
Aitken	3,897	3,617	1,770
Itasca	11,544	9,264	3,182
Koochiching	3,091	2,845	721
TOTAL	97,292	63,732	23,824
Percent	52.6%	34.5%	12.9%

In Duluth, Franken won by more than 20 points and St. Louis County had the highest percentage of Franken voters than any other county in the state, including Hennepin. Duluth also had a whopping 85% turnout rate, which will probably be one of the highest in the country.

The Arrowhead region provided Franken with over 8% of his total votes. Given the turnout and wide margin that he beat Coleman by in this area, Franken owes his recount to Northeastern MN, and specifically to Duluth.

The Importance of Being Analytic

Web analytics, the use of internet search traffic and page visit data, is just beginning to be used by campaigns to determine marketing efficacy. This practice is still in its infancy and the data is a little sketchy, but compiling data from the internet will be huge in future elections.

The candidates that can effectively use web analytics in the next election cycle (or this one, for that matter) will have a huge advantage over an opponent that cannot use the data. After the election, the Degausseer will attempt to unravel web trends of the MN Senate race to see if there was anything happening on the internet that could have evidenced the final outcome.

In the meantime, here is some quick info pr0n.

Charts show the google search indexes of Franken and Coleman through Oct 4 to Nov 2. Higher index means more searches. Again, since there is an error associated even with this data, searches are statistcally tied. Franken leads in internet searches by city, with more Duluthians searching for him (per capita) than any other city in MN.

Is Franken going to carry the Arrowhead region?

Could Coleman Eke One Out?

The final round of Senate polling before the election was completed over the holiday weekend. While little kids were learning the finer points of bartering and Halloween tricks were being played (or turned), the rest of MN was getting sampled.

The results of the these final polls show a shift toward Coleman. There have now been a total of 49 polls conducted since last year. Although the results of the final few polls show that there is a statistical tie between the two candidates, there are a few data points that suggest Coleman may eke out a victory tomorrow.

First off, here is the line graph of all polls.



Coleman has come back from mid-October lows. Franken has hovered around 40% pretty consistantly.

Looking at the overall trends, Coleman has taken a minute lead, but because of variation in the data these trends are too close to call. The take away point, however, is that Coleman has made a come back and Franken has stagnated.



Looking at spread advantage (the difference between poll numbers), the trend line has broken in favor of Coleman for the first time since mid-September. Again, this is too close to call, but trends have shifted in Coleman’s favor.

Finally, looking at average highs and lows for all polls, Coleman has a 3-point advantage. This advantage is again within a statistical margin of error, so the race is still a toss-up, but considering the overall data trends, Coleman appears on the verge of eking out a victory.

Poll Averages .......High .....Low .....Total
Coleman..................48....... 40......... 44
Franken.................. 45....... 37......... 41

In races this close, campaign organization become the key to winning. The election will be decided on who has the best behind the scenes operation. Who knocked on more doors? Who had the most face to face contact with voters? Was it the GOP or DFL?

If you’re a Franken fan, you hope to God that Franken has hitched his wagon onto the organizational juggernaut that is Obama. The GOTV effort from that campaign may push enough DFL minded folks out to the polls to vote for Franken giving him the edge.

If you’re a Coleman fan, you hope that the storied GOP voter mobilization efforts of the last few presidential election cycles still work. MN is clearly going to fall in Obama’s favor, but will the GOP mobilize enough people to come out for Coleman? We will know shortly.

If you’re a Barkley fan, you are justifiably bitter. The political egos of Franken nd Coleman dominated an election season that should have been a policy wonk’s wet dream. Instead of measured responses to very serious issues, you got a high school pissing contest.