Hunter Knifton on Data, the Future of Politics, and How to Make an Election Model
I talked to the campaign data expert about how data is operationalized into actionable insights in the modern Canadian political campaign.
This is part of a series of posts discussing best practices in political staffing. If you want to participate or suggest topics, you can provide input here.
I’m gifting two months of free access to paid content for current political staffers - email me at northpollstrategies@gmail.com to redeem if this applies to you.
Hunter Knifton, currently a senior consultant and data scientist at Crestview Strategies, is one of the innovators in the Canadian political world working at the exciting intersection of data and politics.
His past political work includes data and communications roles in a wide variety of federal, provincial and municipal campaigns, including in the 2022 Ontario Liberal campaign war room, two Ontario Liberal leadership campaigns, and the Olivia Chow Toronto mayoral campaign.
We talked about the growing and changing role of data in Canadian political campaigns, the future of campaigns, and the work that goes into creating election models:
AK: You’ve worked at the intersection between data and politics, which is one of the areas of politics that has changed the most recently. What can you tell us about how data can be used intelligently to supercharge modern campaigns?
HK: I think that you would attest to this as well, that in Canada - given how stringent spending is - there are resource allocation problems. In the US, you can just throw millions of dollars at paying people to knock doors, paying for ads, paying for pretty much everything. In Canada, even with a fully funded campaign, you still don't have nearly as much money as you could spend if you wanted to. And so campaigns are really about optimizing resources.
I think the area that it has changed most recently is that data used to be just about who are my voters and how do I talk to them through traditional surveys or focus groups or something like that: you just figure out who your voters are and how do I convince them to go vote and vote for me. Whereas now, that's obviously still a big piece of the puzzle. But now there's data everywhere.
And so campaigns are using data to figure out:
What their mailer should say.
How do they optimize their volunteer recruitment?
What platforms are they advertising?
How do they talk to their voters through texts and emails?
And so I think that bringing in all those sets of all those different pieces of data and figuring out how to help you optimize the $30 million you have if you're a federal campaign, like where to spend that money? How to do it efficiently? How to make all those micro decisions in the most efficient way possible? That's probably the area that it has changed and will continue to change the most.
AK: Yeah, you talked about resource allocation, which is always a struggle for people in political offices and campaigns. And to that point, I think we end up in a spot where we see a lot of things that we can do with data, a lot of things that are potentially really interesting to try, but we have to prioritize. We see so many innovations, so many things that are being done south of the border where they have these infinite budgets. But then we have to pick and choose ones that we think are the most promising or the most likely to deliver results.
Are there any specific innovations that you've been looking at recently that you think if someone is entering politics and they're looking to be a data and politics person, they should really be getting in ahead of the curve and learning how to do this thing because it's going to be the future of campaigns in Canada?
HK: Yeah, probably a couple areas. The first one builds off my previous point, but figuring out the best way to bring all those pieces of data together. Previously, when the only person that cared about data was your pollster and they would just put together a deck and have crosstabs, that's a pretty easy sort of data source to look at. Whereas now you have data coming in from door knocking, your fundraising team. the folks that are sending out emails and sending out texts and sending out mailers, and then, of course, you have your public opinion research.
And so I think the liberals in particular do a really good job of this, bringing together all those pieces of data and putting them in a dashboard or whatever the output is, but bringing all those pieces of data into a digestible format so that campaigns can make strategic decisions based on all of those pieces combined. I think that's an area where there's going to be a lot of development and if folks are interested in getting really good at something that's going to be really important in that scenario.
A second one, and it's sort of related to this, is that survey research is getting more precarious for a number of reasons. It's just more difficult to find different groups of folks to respond to surveys. There's questions around do we weight for past vote or do we not? How do we reach non-English speaking communities? All these sorts of things.
Of course, survey research is going to continue to be really important to figure out who your voters are, but down in the US and increasingly in Canada, campaigns are using modeling to figure out, first of all, what are the writings we can win? And then even down to what does a voter look like from an income or demographic or psychographic lens?
If people can figure out how to build these really good voter models and how to apply those models to political campaigns, that's probably another area.
And then we talked about AI off the top of the call before we started this officially, but that's also an obvious one. AI is changing how most things in Canada work. There's probably going to be some tools that are new based on AI, but more importantly, it makes all these things a lot easier and cost efficient to do. And so, if you're a young person, and you have time on your hands, you're probably more AI literate than most other folks that are interested in politics, and it's probably a lot easier for you to figure out how to build a voter model or how to bring in a bunch of data into a dashboard so a campaign manager can read it. The barriers to doing that are just so much lower with AI, and that's something I'm focusing on and probably a lot of young people can and should spend some time on.
AK: You mentioned modeling. I feel like that's something the two of us have had some experience with that, but for most people in politics its a very mysterious and mystifying thing that happens behind the scenes. I know you've made a model for the federal election that I got a chance to take a peek at, and you've made some models as you've worked on various campaigns. Do you want to explain to the readers what a model actually is and, without giving away your secret sauce, what types of inputs go into it and how you use a model to make actual strategic decisions in a campaign?
