Ask Bob: How Vermont's 'Mirror Towns' Reflect Statewide Election Results
Roughly 16 years ago, VPR senior political reporter Bob Kinzel set off on a massive research journey to see if he could find a town in Vermont that always mirrored the statewide results.
Not only did a town have to pick the winner, the margin of victory in the town had to be pretty close to the statewide outcome. And in his search, Bob actually found five communities that he calls "the mirror towns."
For the last 75 years, these towns have been very accurate barometers of the races for governor and lieutenant governor.
What are "mirror towns," and what prompted your fascination with them?
Bob Kinzel: First of all, I love to look at election data: how different candidates did in different towns in different years. So while I'm looking at the results of the 2002 general election in Vermont, this thought crosses my mind: "I wonder if there's a town somewhere in the state, that reflected or mirrored the statewide results on a regular basis."
I don't know why I thought there would be a town in Vermont that reflected the statewide totals, but I was determined that if it was there, I was going to find it. And I was also certainly open to the possibility that there wasn't a town that mirrored the statewide results every year.
What was the process, and how did you do your research?
Bob Kinzel: Well, I went through the state elections book for 2002 to start with. That has the results of every statewide race listed town by town. I wanted to see how many towns supported Jim Douglas — that was his first race for governor — and Brian Dubie in his first race for lieutenant governor.
That was a year where Douglas won a close race against Democrat Doug Racine, and Dubie defeated Democratic candidate for lieutenant governor Peter Shumlin, and Progressive candidate Anthony Pollina.
After combing through all these results, I had about 35 towns that fit the bill. But then I started going backwards. I did the same thing for 2000, and 1998 and 1996. How did these 35 towns do, as we looked at more and more elections? And they started to drop out.
In the end, I went back to 1960, and of those original 35 towns, there were five towns left standing: Bethel, Bristol, Cambridge, Jericho and Randolph. I have followed these towns in every election since 2002, and they have an amazing track record.
Have these mirror towns ever been wrong?
Bob Kinzel: The only times they've been wrong is what I call "the exception policy to the mirror town concept." If a statewide candidate comes from one of these towns, or if they live in a neighboring town, then we don't count that town that year.
I'll give you two examples of this. In 2008, Jericho Rep. Gaye Symington was the Democratic candidate for governor. Not surprisingly, she did much better with Jericho voters than she did on a statewide basis. And in 2014, Republican gubernatorial candidate Scott Milne had a very strong family base in Orange County, so I didn't use Bethel or Randolph in the governor's race that year.
But that's really the exception, what we think of as "the geographical factor."
What would someone who studies political science have to say about this mirror towns concept?
Tyler Ballard is a 2020 graduate of Castleton University who made the mirror town theory the subject of his political science major thesis. Ballard is now a Ph.D. student at Brown University, where he studies political science. His conversation with Bob Kinzel and Mitch Wertlieb has been edited and condensed for clarity.
Mitch Wertlieb: Tyler, this mirror town concept ... I'm wondering what got you interested in this subject?
Tyler Ballard: I was really enthralled with election forecasting models, and I had posited that idea to professor Rich Clark, and he's like, "Well, it's a good idea, it's just, there's just a lot of components that go into it that really make it difficult." And as social scientists – this is his favorite quote – “we're really good at predicting the past.”
He had brought up Bob's mirror town findings. And I sort of ran with that, just because I wanted to see if I could posit a theory to explain if there are any decisive factors that would contribute to these towns being able to mirror the statewide gubernatorial election outcomes.
Bob Kinzel: How did you go about researching this concept? What were some of the factors that you were going to look at?
Tyler Ballard: Well, there were a lot. I ended up doing some regression analyses. There was age, there was partisan identification... Partisan identification, obviously in Vermont, is very tough in a state where you don't have to register for a party. So there were some operationalization issues that come with that. There was race, there was median household income. There were a plethora of variables.
While it's useful to have all these sort of controls, you hit a certain point where they don't really add anything to the model. Partisanship ended up being a real huge factor, I would say, most of the time.
My theory going into it was that the towns that perform best in gubernatorial outcomes would also match or mirror the state on a demographic basis in terms of educational attainment, they would have a similar median household income, they would have a similar median age.
My theory was that they would not just perform the same as the state on electoral outcomes, that they would look like the state, too. And that didn't necessarily hold up. So there's nothing conclusive to really explain why these municipalities perform as well as they do; they just do.
Mitch Wertlieb: Did you come to the conclusion that, yes, this five-town mirror town concept does in fact work, you just weren't exactly sure why?
Tyler Ballard: Well, it works until it doesn't. That was the conclusion I drew.
Mitch Wertlieb: But to this point, it has held up?
Tyler Ballard: Yup.
This research seems to suggest that there is no predictive value with these towns, that just because something has been true in the past, it doesn't mean it will be true in the future. What do you think?
Bob Kinzel: You know, I'm reminded what Yogi Berra said, that predicting is really difficult, especially in the future.
I think Tyler really did hit on some stuff that is very important. And that would be ... the demographic profile of these communities.
My guess would be that if we spent some time really searching and comparing the demographic levels as he did in 10 or 12 cases, with those of the state of Vermont, we're going to see some overlapping demographics. But again, there's nothing conclusive there.
I think, as he summed it up best: it works until it doesn't.
We've closed our comments. Read about ways to get in touch here.