Legitimate Microtargeting

Today I want to clarify a point about the concept of "micro-targeting" raised in my posts two weeks ago about those mysterious Roboscam calls that appear to be an effort to improperly "harvest data" under the guise of a survey.  In those posts, I used the term "micro-targeting" and linked to a 2004 Washington Post article that referenced the legitimate micro-targeting efforts of the Bush campaign.  In doing so, I may have left the impression that all micro-targeting involves similarly shady data collection.  It does not.  In fact, vendors for both political parties conduct a form of survey based micro-targeting that is both legal and consistent with the ethical standards of survey research. 

Let’s consider the techniques described in the Post story:

[T]he Bush campaign’s $3.25 million contract with the firm TargetPoint Consulting…[that]produced dramatic innovations in direct mail and voter technology, enabling Bush to identify and target potential voters with pinpoint precision.

The article by Post reporters Tom Edsall and James Grimaldi went on to describe the micro-targeting effort in greater detail:

Republican firms, including TargetPoint Consultants and National Media Inc., delved into commercial databases that pinpointed consumer buying patterns and television-watching habits to unearth such information as Coors beer and bourbon drinkers skewing Republican, brandy and cognac drinkers tilting Democratic; college football TV viewers were more Republican than those who watch professional football; viewers of Fox News were overwhelmingly committed to vote for Bush; homes with telephone caller ID tended to be Republican; people interested in gambling, fashion and theater tended to be Democratic.

Surveys of people on these consumer data lists were then used to determine "anger points" (late-term abortion, trial lawyer fees, estate taxes) that coincided with the Bush agenda for as many as 32 categories of voters, each identifiable by income, magazine subscriptions, favorite television shows and other "flags." Merging this data, in turn, enabled those running direct mail, precinct walking and phone bank programs to target each voter with a tailored message.

Soon after my posts appeared I heard from Alex Lundry, a senior project manager at TargetPoint, who wanted to assure MP readers that his firm "had absolutely nothing to do" with the Roboscam calls and "never use[s] push polling or data harvesting methods to conduct our research."  I also received a similar message from Keith Goodman, the president of a Democratic micro-targeting firm, Bulleye Political Group, who adds a different theory about the possible motives of the mysterious calls (the full text of both email messages appears below, after the jump).

Again, to be crystal clear, the targeting efforts described in the Edsall/Grimaldi article were very different from the Roboscam calls discussed here.  Legitimate survey-based micro-targeting has been conducted by both Republican and Democratic firms over the last several years.  Some may confuse these efforts with Roboscam or even "push polling" calls, but those labels are unwarranted.  The legitimate micro-targeting technique is worth some explanation to try to head off that confusion.

The legitimate micro-targeting done by these firms combines individual-level consumer data with legitimate survey research.  They take computerized registered voter lists available from election officials in each state and merge it with data on "consumer buying patterns" from commercial vendors. The micro-targeters then use their lists to conduct surveys with very large sample sizes (5,000 respondents and up) and ask standard political questions.  Some micro-targeters ask just a few questions (sometimes just the vote and a certainty follow-up), while others ask thirty questions or more.  But rather than run the standard pollster cross-tabulations (involve gender, age, race, and similar subgroups) they use statistical modeling programs (with names like CHAID and C&ART) to identify very small subgroups that seem like good targets for further persuasion or mobilization (a hypothetical example: white men between the ages of 35 to 45 years old who drink Coors or Bourbon – insert your own joke here).   The firms use very large samples to allow for statistically meaningful results from very small subgroups.  Hence the term "micro-targeting." 

MP friend and Democratic direct mail consultant Hal Malchow (of the firm MSHC Partners) describes his use of this technique in his book, The New Political Targeting, parts of which are available for download at the MSHC website.

The key distinction between legitimate micro-targeting and Roboscam is that micro-targeters survey large random samples to target subgroups of voters for further contact.  They honor the researcher’s duty to safeguard the privacy of individual level data.  If the respondent in a micro-targeter’s survey receives a follow-up contact based on the results of the survey it is a coincidence .  Respondents in the subgroup are contacted along with everyone else in the subgroup. 

The Roboscammers, on the other hand, collect data under the guise of a survey to target individual voters for further contact.  If the respondent of a data harvester receives further contact, it is the direct result of they information they provided on the call.  These "data harvesters" violate the implied pledge of confidentiality of a poll or survey. 

One quick tip for telling a legitimate survey from Roboscam:  Legitimate automated surveys (whether done by micro-targeters or public pollsters like SurveyUSA) will identify themselves and provide a working telephone number for recipients to call to request placement on that company’s do-not-call list as required by the Telephone Consumer Protection Act (TCPA).  The Roboscammers do not. 

The full text of the messages I received from TargetPoint and Bullseye Consulting appears on the jump. 

Message from Targetpoint:

Mark,

I’m a regular reader of your blog and I’ve been following your posts on the "roboscam" calls with great interest.  In particular I noted your speculation that this could be some kind of illicit data harvest done for the purposes of microtargeting. 

It just so happens that I work for TargetPoint Consulting, the leading MicroTargeting firm for conservative candidates and groups.  I can assure you of a few things:

*We have absolutely nothing to do with these calls.

*We never use push polling or data harvesting methods to conduct our research.

*Our surveys adhere to the strictest industry standards on public opinion research.

I’m sorry to see that a thing such as this is happening, but I’m glad to report that it has nothing to do with the MicroTargeting done by our firm or for our clients. 

Hope this gets us all a little closer to some answers on this thing.  Best of luck and keep up the great blog!

Best,

Alex Lundry

TargetPoint Consulting

Message from  Bullseye Political Group: 

Mark,

As someone who builds microtargeting models, I have never been involved
in anything like this before.  I don’t think it is likely that this data
is being collected for use in a microtargeting project. – I’m betting
this is more traditional, i.e., you get placed on a non-modeled list if
you respond a certain way.  If they really wanted to be collecting
individual level response data they probably would have included a third
question – gender.  Then they could actually append the robo responses
back to a voter file with a relatively good degree of accuracy.  Plus,
the "push" poll aspect of these calls would certainly result in a skewed
dataset and might cause a lot of problems in the resulting model.  For
instance, if I was interested in differentiating between soft supporters
of wiretapping and hard supporters of wiretapping, I never would have
thrown in the Al Gore part.

The one caveat here is that it is possible that the data is being
collected to microtarget households rather than individuals but that
doesn’t seem likely.  From a modeling perspective, it would be valuable
to know which individuals are for and against wiretapping, but the
design of these calls doesn’t seem very modeling-oriented.  My
experience has been that traditional research techniques are not only
ethical, but also the most valuable for accuracy purposes.

Keith Goodman

President

Bullseye Political Group LLC