The new metrics of campaigns

Polls are as discredited as they should be. So I’m thinking about writing my Guardian column next week about all the new metrics we have to take the pulse of the nation on the internet. Please help me out with numbers you follow.

None of these is representative or certainly scientific. And many of them can be manipulated — which is just the point of them; they put metrics in the hands of movements that use them to make themselves known: witness Ron Paul’s devoted cult and how they played YouTube like an organ. I speculated after Iowa that one reason for Obama’s success there was the campaign’s ability to organize a critical mass of young supporters in the social services.

The new internet campaign metrics also let us sense trends that aren’t so manipulable, if we know where to look.

Among these metrics (many tracked by TechPresident):

* Mainstream media coverage: Here‘s Daylife’s track of Clinton v Obama v Edwards in the last 30 days. It shows Clinton coverage is ahead in coverage until a surge in Obama’s around Iowa with a dropoff in Edwards’. This kind of analysis is possible now that all the coverage is being gathered and analyzed. Before, we couldn’t so easily measure the perspectives and prejudices of media coverage; now we can. Last April, I used cruder measures to show that the MSM narrative emphasized Obama while the polls still gave more attention to Clinton.

* Google searches: Here, in a chart representing December 2007 in the U.S., we see Clinton generally ahead of Obama but with her falling and then showing a resurgence. What do searches indicate? I think they can at least measure interest if not affection or affiliation.

* AdWords demand: You can get an idea of the market value for a keyword on Google AdWords. Clinton shows an average cost-per-click of $2.30-3.15 vs. $1.04-1.30 for Obama and $1.02-1.27 for Edwards. I’m not 100 percent sure how to interpret this — that is, what factors influence this (such as relative traffic) — but there is higher demand for the Clinton keyword. That’s a market speaking.

* Blog mentions: OK, we’re not representative but there are lots of us and what we say can be tracked via Blogpulse and Technorati. The other day, I tracked the “change” meme in the Obama campaign here. This chart shows Clinton ahead of Obama and Edwards until Iowa and now they’re even.

* Textual analysis: We get to analyze the candidates’ language as well. In this post, I took the transcript of the Democratic debate in New Hampshire and used Tagcrowd to find out that they used the word “change” more than “health” or “economy” or “iraq.”

* Web traffic: This Alexa chart shows Obama solidly ahead in traffic to his site. Compete shows Clinton and Obama leapfrogging each other with a spurt in Obama traffic pre-Iowa.

* Video traffic: TubeMogul adds up YouTube views and shows Obama way ahead with 9.5m vs. 4.6m for Clinton and 4.5m for Edwards (on the GOP, Ron Paul beats them all with 10.5m followed by Huckabee with 4.8m — a surprise to me — Romney with 3.8m, Giuliani with 1.8m, and McCain with 1.2m)

* Microblogging traffic: Obama as 6,667 followers on Twitter, Edwards 4,167 (and his final post came two months ago), Clinton has a big 223 (not her medium, apparently). That is, two of the three said it’s not worth it.

* Social sites: TechPresident has been doing a great job tracking the candidates’ friends. On Facebook, Obama has long been ahead with 235k vs. 71k for Clinton and 35k for Edwards (on the Republican side, Paul is ahead with 72k and Huckabee comes in next with 40k). MySpace: Obama has long led with 225k vs. 162k for Clinton and 52k for Edwards (on the GOP side, Paul is again leader with 114k vs. 41k for McCain, 33k for Romney, and 28k for Huckabee).

* Donors: Of course, we always have money. A recent AP story said that Clinton raised $24 million in the latest quarter vs. $22.5m for Obama and $4.5m for Edwards (the Washington Post said that Clinton and Obama each raised more than $100m in 2007). The other important story is the number of donors. Obama was doing a great job getting lots of small donors — who, as I said here, became invested in the campaign. But USA Today said that for July to September, Clinton exceeded him, 100k to 93k.

* Prediction markets: NewsFuture’s market has a dead heat for the Democratic race — Obama 50% to Clinton 49% (on the Republican side, McCain leads at 44% with Giuliani next at 25%). By the way, whoever wins that race, NewsFuture’s predictors give the Democrats a 65% change of winning in the fall. InTrade is more complex with lots of contracts that get a market value. For the nomination, Clinton’s contracts are on top at 56.8 over Obama at 40.8 and Edwards 1.5. For the November election, Clinton’s contracts are selling for 37.2 vs. Obama for 25.0, McCain 17.6, Giuliani 6.5, Huckabee 4.4, Edwards 0.5.

* Odds: Linesmaker‘s rundown in another market: Clinton 7-5, Obama 2-1, McCain 7-2, Giuliani 8-1, Romney 15-1, Edwards 30-1, Thompson 50-1 — and Michael Bloomberg 5-1.

What other metrics do you know of and follow?

So what does all this tell us? Who the hell knows? The truth is that there is no reliable predictor — certainly not the polls. But there are many ways to try to sniff out trends and many ways to organize people to make those trends happen.

  • Pingback: » Blog Archive » Messbarkeit von Social Media()

  • you really think that google searches point to interest… how can you gauge sentiment from it? what if I am searching for hillary clinton to prove my point against her or to refute her claims.

  • i’m going to agree with the skeptic, you can’t really gauge sentiment

  • P.J. Hinton

    I’ve always found the Iowa Electronic Markets to be an interesting metric.

    P.J. Hinton

  • “Polls are as discredited as they should be.”

    Primaries are more difficult to predict than general elections.

    Just because more Undecideds voted for Hillary Clinton in New Hampshire than expected doesn’t mean polls are discredited.

    I’d be more interested in reading a poll about the South Carolina primary than reading how many Google searches for a candidate’s name came from South Carolina.

  • Cooler Heads

    Jeff, you are generally so insightful. Why do you periodically go off the deep end with silly ideas?

    A poll is MUCH more reliable than, say, the number of Google searches. You’ve gotten your knickers all in a twist over the mistakes in New Hampshire in the Democratic poll, and completely overlooked the fact that the polls were quite reliable in the GOP race there. And in both races in Iowa. Perhaps your takeaway here is the 25 percent of polls (one of the four I mentioned above) are unreliable.

    Think about this: Polling is not absolute science, but a kind prediction. It assumes that whatever prediction is made it will definitely be wrong 5% of the time or thereabouts just because that’s the amount of random error in the world. And then there is additional margin of error depending on the sample size, the type of questions asked, the variation in the results.

    So having a handful of polls go wrong doesn’t invalidate all polling; the statistics admit that such a thing is quite possible.

    What’s more interesting is what you posted about earlier, that there may be something specifically related to the Democratic election in NH that caused the pattern in polling errors. You were much closer to the real issues there.

    You do this kind of anti-polling burp every four years. Look back at your old posts and see if you find something similar from the last big election.

  • Pingback: The New Metrics of Campaigns | Jeff Jarvis | Voices | AllThingsD()

  • Pingback: InTrade is no psychic —but what if that bit of truth is systematically said BEFORE, as opposed to AFTER. | Midas Oracle .ORG()

  • Colin Kerr

    The bookmakers tend to be uncannily accurate, probably because they do so much research and also because they can access inside information.

  • Hi Jeff,

    I find Technorati and Blogpulse too unwieldy, and rarely precise enpough. I recommend ( Doesn’t work for everything under the sun, but it’s pretty interesting when it comes to the presidential race itself. Not that blog trends are significant per se, but they got it about right, for Iowa and NH for the republicans at least.

  • eric

    The constant skeptic gets it right, the majority of your mentioned metrics are not measuring behaviors they are measuring actions. Just because someone jumps in a search ranking does not mean that voter minds are being changed. maybe they are looking to refute an argument. Maybe they are seeing why their candidate was being attacked. the metric says nothing about what the person was thinking after they performed the search.

    Further, the markets you mention are not representative. How many families from east LA and Harlem, WV hill country, TX ranchers, college students from city/state college have funded up their InTrade accounts?

  • Jeff,

    Your use of the CPC found in Google to show demand seems to be very off. While I see what you are trying to do, CPCs are not just based on bid amount and has to do with relevancy and historical CTRs among others. Back in the old days when relative position was based on bid, you could have maybe learned something from average CPC. Perhaps looking at the number of bidders in the morning a few days might be a better judge of which names are garnering demand.



  • Daniel

    Prediction markets are becoming increasingly more interesting to me, now that such markets are proving themselves to be of surprising predictive value in many contexts. (For instance, the prediction market at has been an eerily prescient forecast of MLB’s final standings than any poll of experts, two years running.)

  • Pingback: links for 2008-01-15 « Kevin Bondelli’s YD Blog()

  • Pingback: La Tejedora » Blog Archive » Cara a cara en la red()

  • Pingback: State of Local()

  • Pingback: Looking Wider Than Opinion Polls | Political Futures()

  • Fascinating. Thank you Jeff.

    I need to explore this some more but I’m interested in the varying responses through different media, for example, Obama’s stronger showing on You Tube than through blogs, It’s telling us something significant, I think, about how the use of internet media is changing amongst a young generation.

  • Pingback: Novas métricas para campanhas « Atrium - Media e Cidadania()

  • Pingback: Knitware Blog » Blog Archive » The new metrics of campaigns()

  • Pingback: Nel 2012 internet deciderà le elezioni USA? « MediaMeter()

  • Jarvis —

    An implication of this list is that election campaigns nowadays function as spectacles rather than contests. An opinion poll measures the level of support for a candidate — who is likely to win or lose. This list tries to capture its intensity instead — the viral activity a candidate stimulates.

    The corollary in campaign coverage is the discrediting of so-called Horse Race journalism and the development of a replacement that I call Reality Gameshow journalism. In 2008 it isSurvivor not Seabiscuit.

    Regards — Tyndall

  • Pingback: Novas ferramentas para campanhas política « Herdeiro do Caos()

  • Pingback: Novas ferramentas para campanhas política «()

  • Pingback: HAGAKURE | Internet PR()

  • Pingback: Ardi XIV()

  • Pingback: Presidential Candidature, part deux | pixelblog()