So how did the GOP's vaunted microtargeting-driven 72-hour plan work? How did the Democratic program work? I'll be posting data on over the next few months as it becomes available.

GOTV and a Republican Microtargeting Disadvantage

One of the interesting features of microtargeting is that the intelligence is a snapshot in time. That is, microtargeting is built from a large poll that is, in most cases, conducted MONTHS before an election.

If the environment is static, the microtargeting will be pretty accurate on Election Day. But what happens when conditions change between the date the microtargeting is completed and the election? This is precisely the problem facing the Republican 72-hour plan.

Here is an example. Many campaigns use 65% ( as the threshold for GOTV. That is, if someone is likely to support your party/candidate at a probability of 65% or higher, that individual is rolled into a GOTV universe.

Consider a conservative independent who, when the microtargeting was constructed, had a 66.7% chance of voting Republican. What this means is that for each voter a campaign turns out, you are wrong 1/3 of the time. Therefore, you must GOTV THREE voters of this type to gain a net vote.

But, if a result of changing conditions, that voter in reality is only 60% likely to vote Republican, you have to GOTV FIVE voters to gain a net vote. It is amazing that a mere 7% trial heat shift can cut Republican GOTV output by nearly half for this type of voter. Of course, Democrats will benefit from this phenomenon this cycle. Small differences really do matter when it comes to GOTV, and right now the Republican operation is struggling against gale-force winds.

I just thought of a reason why Republican microtargeting is at a disadvantage in 2006 that I think it mathematically interesting. I'm working on a post on this and will have it up soon.

Myths about Republican Microtargeting: Myth 3

3. Republicans have microtargeting down to a science.

If Republican phonebankers and canvassers only knew that they are making calls based on faulty microtargeting, I don't think they would be very happy. But this is possibly true, at least if you believe the book Applebee's America, which claimed that in 2004 Republican microtargeting interviewed a maximum of of 5,000 in key states.

Now, I'm about to move into boring statistics land, so cover your eyes if the word regression sends you running for the exit. One important principle when building microtargeting models is that you should have a hold-out sample that allows you to verify your work. Sometimes, if you build a bad model, it "overfits" and includes certain pieces of information as important that are either irrelevant, or worse, have an effect opposite to that the model predicts.

Back to 2004: In most states, the Republicans appear to have segmented the electorate into about 25 pieces. This means that each segment has about 200 voters in it, sans holdout. Or, with a holdout of equal size, each segment has 100 voters, with a margin of error of approximately 10%! Suddenly, the models don't seem so accurate.

Earlier, I doubted that Republicans actually use car type and liquor choice in their models because these variables have not been proven to be significant when I have tested them in a rigorous statistical environment. If in fact these variables are being used, it is a sure sign of overfitting.

Conclusions

Hopefully this post helps to debunk the Republican mythology around microtargeting. Of course, if you have information that any of my assertions regarding Republican work is incorrect, please let me know, although all the evidence I have seen supports what I have written above. I'll post again soon with more of a look at why microtargeting is critical and what it can accomplish once you strip away the false mythology.

Myths about Republican Microtargeting: Myth 2

2. Republicans are able to use consumer data information like what kind of car you drive, whether you drink bourbon or gin, and what magazines you subscribe to develop sophisticated narrow-cast messages.

Regardless of your political leanings, you have to hand it to the Republican press operation - they've done a masterful job in selling their program to the press. Each new article about how the GOP could target based on car ownership or liquor choices reinforced Republican donors' belief that their money was being well spent by a sophisticated political operation. Meanwhile, Democratic consternation grew as donors and many operatives wondered why we weren't using the same data. The Republican spin is mainly oriented to mazimize press interest, entice donors, and confuse the other side in place of an accurate description of their approach,

The data that Republicans cite is available through commercial vendors such as InfoUSA and Acxciom. This information is what Republicans are talking about when they say they have "four or five hundred" pieces of information about each voter. Democrats and progressives have access to this information as well.

Most of this data, however, is useless. It may be true that 60% of Bourbon drinkers are Republicans, and 60% of Gin drinkers are Democrats. But this doesn't mean the information is useful. Why? There are dozens of other, less sexy pieces of information that are far more predictive about someone's partisan leanings - such as age, gender, marital status, number of children, income, and place of residence, to name a few. Combined with information on voter files, this is the information that Republicans and Democrats alike use to microtarget.

So why does the media fixate on the Starbucks-drinking, Volkswagon driving voter profiles, and why do Republicans actively encourage such coverage? Because it is a lot more interesting than talking about logistic regression, database systems, and "boring" information tidbits.

Myths about Republican Microtargeting: Myth 1

I've been doing microtargeting for progressive and Democratic organizations since 2004. In early 2006, I founded Bullseye Political Group LLC and now do microtargeting work for several clients.

Over the past several years, I've watched as the media has run a series of articles regarding the success of Republican microtargeting efforts. Just today, I watched an Anderson Cooper story on microtargeting repeat the same stylized story about what microtargeting is and how it is being used. Now, just before the 2006 mid-terms, talk of Republican skill in this area seems to have reached a fever pitch, with many progressives expressing concern over the perceived Republican advantage. The truth is that Republicans do have a slight advantage, but Democrats and progressive groups have rapidly caught up and in some areas surpass Republican capabilities. This post is my attempt to correct some of the misinformation (as I see it at least) that is out there. In the coming days I hope to add to this post, although updates may have to wait until after the election.

I will also state that a lot of these myths will be more apparent to the media once the votes are counted in less than a week. Perhaps then the media will understand that there is a lot of smoke in what the Republicans say they are doing with microtargeting. The same can be said of the 72-hour plan, but that is a post for another time. Could I be wrong? Yes. But I think the explanations below will help clarify what I am trying to say. Here is the first myth:

1. Democrats are way behind Republicans in their use of microtargeting.

We may have been behind at one point, but we have rapidly caught up and in some respects threaten to surpass Republican capabilities.

The community of people working on microtargeting on our side is impressive. I know of at least 6 private firms who are engaged in microtargeting projects as well as several internal programs at major organizations. I've had direct experience working with microtargeters at Hart Research, Copernicus Analytics, the DNC, and the AFL-CIO and I can say that the capabilities are impressive. The research techniques developed by these groups have created a robust ideas laboratory.

To my knowledge, most of the work on the Republican side is concentrated with TargetPoint. (Please let me know if you know otherwise.) I tend to believe that innovation is now occuring at a slower pace on their side because their market is somewhat monopolistic. Now, I could be wrong on this and we may see some really impressive new efforts from TargetPoint, but what I have seen so far is running the old playbook.

Overall, Republicans probably are still ahead, which I attribute almost entirely to the vast centralization of the GOP political operation. Karl Rove and Ken Mehlman call the shots, and they serve the President. When top GOP staff decide they want every battleground state to use microtargeting, their decision is backed by the authority of the president. With a Democratic president, would microtargeting be easier to conduct? I think the answer is yes. But so would fundraising and messaging! I think there has been increasing support for microtargeting and fundamental political research at all levels, and that is very positive.

Now, I also think there are some disadvantages to the Republican approach. Republicans, while they might be good at finding messages that move voters, have a messenger problem. If Republicans identify a segment of the electorate that can be moved by an enviromental message, they are out of luck - the Republican party is not going to be a credible messenger on this issue.

Sometimes "disorganization" (or better put, a more Grassroots approach) has its advantages. On our side, there are a lot of different messengers using microtargeting, making it more likely that there is a match between the best message and a good messenger. Of course, there is an inefficiency cost, but at least part of the Republican centralization advantage can be nullified by this factor.