When you’re from Maine, it’s hard to watch people pooh-pooh this year’s turn-out. Mainers turned out in extremely high numbers for a non-presidential year, topping this year’s election turnout contest. Why did they come out in those numbers, despite having just been through a ridiculously unseasonal snowstorm with inch-accumulation in the double-digits?

Why WOULDN’T you come out, would be the Mainer’s answer, given the closeness of most of the state’s major races? Not only did the state have a gubernatorial election that multiple sources touted as the tightest election in the nation, but there was an extremely tight race for Maine’s second Congressional district — a representative who constitutes half of Maine’s delegation to the US House — and then there was this unfortunately timed, recurring, yet extremely hotly contested ballot measure about bear hunting. (Personally, I am now considering this to be the election where the bears finally figured out how to get back at us.)

Competitiveness is a classic reason why people get excited about turning out for elections. When people know that their vote could be decisive, they have a greater interest in participating. This is a big reason why political scientists think gerrymandering depresses voter turnout.

So I knew personally about Maine. But how well does competitiveness predict the scale of other states’ turnout figures?

To find out, I used the last week of polls posted at Real Clear Politics. Since we’re talking about state-level turnout, most individual US House seat races wouldn’t be effective to count, but I did include all polls for state races covering at least half of the state. This added up to 66 races for governor, US senator, or US representative for states with two or fewer US House reps like Maine and New Hampshire. It did not include bear baiting or any other exciting state-level ballot initiative, because RCP didn’t include those polls.

I sorted all polls taken in the last week of these 66 races and, where there was more than one poll, chose the poll with the narrowest projected victory. I then scored all of these polls on a 0-5 scale according to their competitiveness: I gave a score of “5” to projected ties, “4” to races where the front runner was a single point ahead, and so on down to “0” for a projected vote spread of 6 points or more.  This is what that looked like (click to sort.)

Because several states (like Maine) had multiple competitive races, I totaled their competitiveness scores together into a composite competitiveness index. This helped reveal that while states could have a few big statewide contests, it was possible that none of them was competitive. This was the case in South Dakota, for example, where the closest of the three statewide races was projected to be won by 11 points. Other states, meanwhile, had multiple competitive statewide races.

Here’s what the ranking of states according to the competitiveness index looks like:

Based on a highly scientific eyeball test, this set of scores bears a passing resemblance to the initial state turnout rates collected by Michael McDonald and visualized at FiveThirtyEight. States with competitive elections were more likely to turn out in higher rates.

The correlation is certainly not perfect but it looks to explain a good chunk of the variation.

The biggest outlier in terms of high competitiveness–but relatively low turnout–was Georgia. The state has come under attack for “losing” some 40,000 voter registrations, making it the focus of the major election day condemnation from the Lawyers’ Committee for Civil Rights Election Protection project.

Georgia’s outlier status therefore doesn’t disprove the voter interest theory behind increased turnout. Instead, it suggests that voting is a two-way street. Even if you’re interested in turning out to vote, your state also has to be interested in permitting you to do so.