45.4 or higher is needed for a candidate to be
viable.
Dean: 100.0
Clark: 62.2
Gephardt: 32.0
Kerry: 30.5
Lieberman: 30.3
Edwards: 24.5
Sharpton: 9.6
Kucinich: 6.6
Moseley-Braun: 6.6
Clark surged yesterday. He seems to be rising at just the right time. He is, however, still pretty far behind, and needs to keep surging to truly threaten Dean's position. Still, this final week is more exciting then I thought it would be.
Clark is also closer to Dean than Gephardt is to Clark:
Clark: 100.0
Gephardt: 51.5
Importantly, the first of the final numbers have arrived. Yesterday's media coverage results will be included in the final ECC on Monday. Those results were:
Dean: 100.00
Clark: 40.74
Gephardt: 39.64
Kerry: 38.22
Edwards: 25.02
Lieberman: 26.76
Sharpton: 20.67
Kucinich: 18.23
Moseley-Braun: 13.24
I have a feeling that Dean's victory in the DC primary will prove to be a significant boost to his campaign. We will have to see.
Also, as promised, I have included an updated description of my methodology.
The Method and Theory Behind the ECC
The ECC arose out of last summer's flame wars on the old moveable type dailykos. My basic goal was to develop a system where the candidates, especially Dean and Clark, could be discussed free of the mindless rhetoric and vacuous talking points that dominated those discussions. Simply put, I hoped to find a way to return at least some of the discussion on the Democratic primary on this site back to actual political analysis.
The theory behind the ECC is that "momentum" generated from early state victories plays a much smaller role than commonly assumed in the primary season, and that the nominee can be predicted through an analysis of several broad indicators of success in the invisible primary. There is nothing particularly new about this theory, and several political scientists have also adopted it as a justification for developing their own predictive models of primary season success.
What is new and different about the ECC, if I may be so bold, is the sheer breadth and depth of the information it considers. While other predictors focus on one or two narrow aspects of the invisible primary season above all else (the Gallup poll, total fundraising, results in NH, etc.), the ECC is a compilation of every broad measure / predictor of primary season success in the invisible primaries. By using six categories and numerous sub-categories, I attempt to include every measurable, proven predictor of primary season strength in the invisible primaries.
(In order to combine disparate sets of raw numbers, (how can $12.487M Cash on hand be compared to having 13 points in a national poll?), every category and subcategory is translated in a 100.00 point scale before it is combined with other subcategories and categories. For example, if candidate A had the most cash on hand with $15M, and candidate B had the second most with $12M, then candidate A receives a score of 100.00 for that category and candidate B receives 80.00. If in the only current national poll, the same candidate A has 18 while the same candidate B leads the poll with 20, then A receives a 90.00 for national polls and B receives a 100.00. Combining the two categories gives A an overall score of 190.00, and B 180.00. This translates into a 100.0 final score for A and 94.7 for B. The candidate with the highest total score from the six categories always receives a 100.0 overall score.)
Category #1, Cash on Hand: In almost every post-reform primary season, the candidate who raised the most money won the nomination. There is an obvious causal link here: more money means more staff, more offices, more advertising and more almost everything needed to succeed in a primary season. One difference between the way I measure money and the way other predictive models measure it is that I only include Cash on Hand. The way I figure it, if money already spent hasn't helped a candidate yet, it isn't going to help a candidate in the future.
Category #2, IA + NH: While the importance of these two states in winning the nomination can be overstated, they still wield significant influence over the states that follow them. Thus, the most recent poll results from these states deserve their own category.
Iowa and New Hampshire are weighted equally in this category. A poll is considered "current" if it ended before the most recent poll began. (For example, if the most recent Iowa poll was conducted from January 10-12, than any poll than ended on the January 9th or earlier is removed from the calculations).
Category #3, Mini-Tuesday: I don't think it's much of a stretch to say that Mini-Tuesday will be roughly as important as IA + NH combined, so I have given it its own category. In this category, current polls from AZ, DE, ND, NM, OK and SC are combined to produce to overall category score (MO is left out because it is Gephardt's home state). The number of delegates from each state determines its weighting in this category:
AZ: 55/195
SC: 45/195
OK: 40/195
NM: 26/195
DE: 15/195
ND: 14/195
Since there are no polls from ND, and no post-Clark polls from NM, currently they are not included in this category. As with polls from IA + NH, in order for a poll to be considered "current," it must have been conducted at least in part before the most recent poll began.
Category #4, National polls: National polls have proven to be a decent predictor of primary success. While dismissed by many, I believe there is a causal link to be found here. While having support in the right states is the best way to determine how someone will do in the primary season, a candidate is far more likely to have support in all important primary states if they have a large national following. The smaller the national support for a given candidate, the more precisely it needs to be distributed into the primary and caucus states. The same definition of "current" for IA, NH and mini-Tuesday polls applies here.
Category #5, Delegates: At the latest, this race will be over by Southern Tuesday. While momentum will change the outcome of the later states somewhat, starting with a lead in the 30-odd states to hold either a primary or caucus on March 9th or earlier gives a candidate an important edge. Thus, I try to calculate that head start by estimating the projected delegate counts for every candidate in every state on March 9th or earlier. I use the 15% threshold for the estimations, and factor out undecideds (so, a candidate currently at 13% in the polls can still receives delegates in my projection if the undecideds are large enough). The same definition of "current" polls used in other categories applies to this category.
Category #6, Media Coverage: If a campaign isn't covered in the press, it will be very difficult for that campaign to make much impact with a large amount of voters. The basic rationale behind the "momentum" theory has always been that winning early states garners huge amounts of free press for the winning candidate. In turn, this free press leads to a large rise in state and national polls. Media coverage clearly plays an important role in any campaigns success. I have chosen to measure only coverage from the past week. Although somewhat arbitrary, we do still live in a world of weekend political talk shows, weekly print and TV news magazines, and a short-attention span media culture that now focuses on the flavor of the week rather than the flavor of the month. Also, the free media from winning, say, the Iowa caucus will have disappeared after the NH primary takes place one week later (as the coverage for winning NH will disappear after mini-Tuesday). So, while its somewhat arbitrary, measuring only coverage from the past week is also fairly reasonable.
I measure media coverage by doing a series of Google news searches. Unfortunately, this means that I cannot measure television coverage, which remains the primary medium of political news in this country. However, the searches should produce at least a decent estimate of the national coverage for each candidate. Unfortunately, one thing my searches cannot take into account is the tone of the coverage, both because it would be such a huge task and because determining tone is a little too subjective to fit into the ECC. But hey, some people believe that all coverage is good coverage, so maybe there's no need to measure tone.
Media coverage is split into two subcategories. The first subcategory is "total feature stories for the week" and the second is "total mentions for the week." The two categories are equally weighted to produce the overall media score. Here are the searches I do for each candidate:
Feature stories
Allintitle: "Braun" OR "Moseley-Braun"
Allintitle: "Clark"
Allintitle: "Dean"
Allintitle: "Edwards"
Allintitle: "Gephardt" OR "Gep"
Allintitle: "Kerry"
Allintitle: "Kucinich"
Allintitle: "Lieberman"
Allintitle: "Shaprton"
Except for Gephardt, Sharpton and Kucinich, who have such distinct names, I scan the results of every search to make sure the articles refer to the candidate in question. Iafter subtracting the articles talking about Herman Edwards, Lewis and Clark, a college deab or Kerry Wood, I record them and move forward with my day.
Mention searches:
Moseley-Braun
"Wesley Clark" OR "Wes Clark"
"Howard Dean"
"John Edwards"
Gephardt
"John Kerry"
Kucinich
"Joe Lieberman" OR "Joseph Lieberman"
Sharpton
The media category takes up about 80% of the total time of the ECC, since I need to recalculate it every single day. All six categories are equally weighted in the final projection.
So, that's how the ECC works. I have tried to make it as broad in indicator of the current state of the race as possible. If you have any suggestions, I'd still like to hear them.