Saturday, April 23, 2011

Geodemographics & the Canadian Election:: The Issue of Microtargeting Audiences

Environics data on Surrey North, from Globe & Mail

The Globe & Mail has an interesting article on how the Conservatives are using data to "microtarget" voters within ridings. This use of technology is not new in politics and this 2004 PBS Frontline,"The Persuaders" [streamed, see Chapter 5: "Giving Us What We Want"], talks about the use of data and narrowcasting messages. The chapter features Frank Luntz and his semiotics of semantics, as well as highlighting some gems from the Karl Rove playbook. The article features a breakdown of Surrey North, a swing riding, by a market research firm, Environics. The idea is to first break down ridings into types or clusters based on market research lifestyle profiles, i.e., geodemographics. 
"Environics Analytics’ Prizm segmentation system breaks the Canadian population down into 66 types. These types, which are narrower, more sophisticated versions of composites such as the soccer mom or NASCAR dad, are based on a host of factors, from where you live to what you buy and what you believe."
In the US, Claritas does this with PRIZM and you can punch in ZipCodes to see which profiles are in it. The next step is to superimpose the election results at the polling district level over the geodemographics. This allows the crafting of specific messaging towards different segments based on demographics, lifestyle, and voting behaviour in a polling district.

I've done a few posts mapping 2008 Canadian federal election results at the polling district level and developing an index of partisanship, akin to the Cook PVI used in the States. [See:: Edmonton—Strathcona, Whitby—Oshawa, and Nunavut]. What would make this index richer would be the incorporation of Census data using a clustering algorithm, as well as multivariate modeling with partisanship, political volatility {variation/stability in voting patterns}, demographics, and behaviours aimed at increasing predictive ability. {As an aside, Punditsguide.ca has a great interactive feature that allows lookups of Census data at the riding level.} While this metric wouldn't be as comprehensive as those developed by Environics or market researchers of that ilk, it might be "good enough" for all practical strategic purposes. Moreover, one of the metrics I've found party effects for in selected Ontario ridings in 2008 was turnout. The NDP fared worse in polling districts that had high turnout; curiously, an effect that wasn't there in 2006.

Nevertheless, market research that taps into context and history can be very powerful. Knowing where a demographic group lives in a riding and how they perceive key issues can indeed help refine messaging.

Sliced Too Thinly?
There are a few issues here. The main one is one that The Persuaders brings up...if you divide up the polity into small slices and tell the people what they want to hear, how can you have a coherent and shared vision of what the party stands for? The party means too many things to different people. I think the Conservatives and their consultants have these shiny tools to identify and customize messages to 66 types of Canadians, but I'm not sure in this era of social media that you can target votes from new Canadians in one riding, while appealing to {or turning a blind eye to} xenophobic sentiments on immigration in another. The Conservatives are facing this very issue. The Tamils understood the "code" behind this Conservative ad featuring Tamil "boat people" and cried foul::


The fallout? Statements like this::
“While chameleon Kenney tries to buy the immigrant vote by watching cricket and eating in ethnic restaurants, his anti-immigrant record speaks for itself. Kenney has drastically expanded slave-like temporary worker programs, under which migrant workers are exploited as cheap labour without basic rights. Meanwhile, the number of accepted refugees and permanent residents has plummeted.” —Sozan Savehilaghi, member of No One Is Illegal
While the geodemographics/micro-targeting is supposed to increase communication efficiencies, it can also present challenges when a party is talking from both sides of its mouth. In essence, there's still a disconnect here, but it may be an artifact of where the Conservatives are in 2011 under Harper. The communication model is still one-to-many rather than being conversational. What should be informing decisions and platforms is feedback from the neighbourhoods, not telling the neighbourhoods what you want to hear. A huge risk for the Conservatives is they've banked on a "stay the course" message and have no degrees of freedom in their budget and their stance. They were swinging for the fences with a majority, which is understandable since I think Harper's future is dubious without one.

I think the politics of the near future will see parties using social media to keep a finger on the pulse of what's going on in each riding on a neighbourhood level. This can circumvent, to a certain extent, the lock market research firms have on expensive data and provide grassroots feedback from the communities themselves. The next generation of web technologies will better enable more civic engagement {and provide insights through data mining}, but let's face it, power, politics, and engagement tend to make awkward bedfellows.

Twitterversion:: [blog+link] Globe&Mail discusses #CPC use of micro-targeting, but is it a problem if a party has 66 different messages?  @Prof_K

No comments: