The Right To Vote

Voting is the cornerstone of democracy. It cements the right to elect representatives, to weigh in on state and federal issues, and to have your hand in the course of your country.

Everyone has a vote**.

But who actually gets to cast it?

That’s the question that determines whether a proposition passes or fails. Not how many people in the voting population support it, but how many of those people vote for it. If 60% of people are for something, and 40% are against it, but 90% of one group and only 10% of the other group get to vote, that’s going to affect the outcome.

Of course, there is already prolific documentation about the unequal accessibility of this fundamental right. This is most often framed in the light of voter turnout, meaning the percentage of possible voters who actually go to the polls. The United States has a startlingly low voter turnout as democracies go. Policy researchers are avidly exploring explanations for this discrepancy with the rest of the world.

But while I was voting by mail this election, I started asking another question.

“Does everyone vote for everything?”

I live and vote in California. My ballot is 5 pages long. I’ve got federal, state-level, and local representatives to research and vote for; I’ve got county measures and city measures to comb through. I’ve spent a couple nights this week working on my civic duty, much to the chagrin of my laundry and dishes!

It’s one thing to show up to the polls in a country that doesn’t have an election day as a mandatory holiday, or a compulsory vote — that’s already hard, and we know that. Polling place locations and lines are long. Sometimes polling places close and it’s hard to know where to go.

But once you get there, does everyone make it all the way through that long ballot? How many people leave things blank?

I decided to find out.

Welcome to King County, WA, where voter turnout is over 80%!

This is actually about the average turn-out in the rest of the democratic first world outside of America. (That makes the ~55% American turnout in 2016 look staggeringly behind.)

King County is a mostly urban and sub-urban county — one of the three that make up the larger Seattle-Tacoma-Bellevue area. According to the 2010 Census, King County is majority white; according to the 2017 American Community Survey (see fig 10) a majority of the population has had some college education. Last week, the Washington Post reported the connection between education status and turnout, so it shouldn’t be surprising that such an educated place has such high voter turnout.

I downloaded precinct-level reporting data on every measure on the 2016 ballots of King County. I validated that data with the overall information to make sure the numbers matched up, but the data I worked with has data for individual polling places, so I have a lot more granularity to work with.

But this post isn’t about getting to the polls. This post is about what happens after.

Voter turnout is high in King County, but not everyone votes for everything.

Screenshot 2018-11-04 10.04.15.png

People overwhelmingly vote for federal level positions.

I categorize the different races on the ballot in order to group ones together, and order by the precinct-level average % of voters voting. You can see that (in 2016) the highest percentages of voters vote for President & Vice President, with most of the federal positions ranking high (US Senator, US Congress). But state-level representatives recieve votes from about 10% less of the population on average — meaning about 10% of King County residents make it to the polls, but don’t vote for the candidates.

More people abstain when candidates run unopposed.

Why don’t people vote for an issue? In some case that’s because their vote may literally not really count. In every circumstance, unopposed candidates for state representatives or for judges (more on them later) get about 15-25% less of the overall population voting for them.

Voting is way lower for judges, of which there’s lots more than the other low positions.

Now back to the judges. Far and away, fewer residents vote for judges, even in contested races! Of course, they also vote at similar rates for the state treasurer and superintendent, but here’s the difference: there were 13 elected judicial appointments on ballots across King County, compared to just one treasurer and superintendent race. Across all these races, people just plain old vote less.

The distributions of vote share between two sides overlap each other — a lot.

Zooming in on four county-level races (we’ll get back to these particular races later) — two judge positions and two propositions — we can look at the distributions of the vote share per precinct that each candidate or choice got. The more distinct these distributions, the clearer the gap between the two choices.

Screenshot 2018-11-05 20.43.11.png

But instead, these particular races — which are varying degrees of contested, as you can tell by where the mean in the boxplot lies — have vastly overlapping distributions. Are they significantly distinct from one-another? Votes are just the polls that matter. Are these outcomes separated by more than the margin of error of a poll?

(This should trouble us as good statisticians. If these distributions overlap so much how confident can we be that the mean coming out higher in one or the other is “right”? More on how perturbable this is later.)

Do people really vote less for ‘downballot’ races?

I actually hadn’t heard of this word until I made this figure and was trying to understand. I showed the plot to a friend. “Oh, downballot races, that’s a known thing,” he said.

Screenshot 2018-11-04 11.28.49.png

And indeed, the literature suggests that both ballot order and choice fatigue have an effect on absention. How does that effect work? Do people just vote less often as they vote for more things?

I got a hold of the county-wide ballot, and luckily, the first page is almost all propositions, which have similar abstention rates. Then I split it out by where the proposition was on the page

More people abstain as the measure moves down the page.

Screenshot 2018-11-04 11.35.32.png

Here, the Y axis is ordred in the same way as the measures on the ballot, with each box representing one of the 3 columns of page 1. You’ll notice that in each one, the closer you are to the bottom, the less people are voting for you — it’s a literal physical down ballot effect.

With one HUGE GLARING EXCEPTION: in column 3, this effect instead centers around the LEAST abstained vote, i.e., the vote for US President and Vice President.

The further you are from the item of interest, the fewer people vote, i.e., the more abstain.

The effects on actual outcomes

“These trends are cute,” you might say. “A neat little experiment. But what’s the consequence?”

As before, the importance of voting is really in representation. If a non-representative group of voters votes, the election doesn’t represent what people want.

Often this is the argument for voter registration drives. “If we get out the vote, and people show up, our propositions will pass.” “Our representatives will be elected.”

So how big of a difference can abstention really make alongside getting the vote out?

30% of county-wide races and 20% of municipal races could be flipped by abstainers.

I did a simple analysis, asking just the question, “if everyone who turned out to vote but abstained on an issue voted for the losing side, how often would the outcome reverse?”

Here’s a visualization of this analysis for a few county-wide races — the ones we looked at before, to see how well the vote share distributions per precinct overlapped.

Screenshot 2018-11-05 07.41.48.png

The margin of actual counted votes was within about 50-100k (5-10% of turn-out) of one-another for these disparate races. But that turn-out includes a similar number (or in the cases of judges, sometimes a drastically higher number) of ballots where that particular box was just left blank.

In each of these cases, the blank votes would be more than enough to reverse the outcome — and in the case of the judges, they could reverse them by a serious landslide.

And these reversals aren’t always on the levels of hundreds of thousands of people — in at least one case, the margin between one legislator and the other is literally one actual votePeople’s votes literally do matter, and in a simplistic model, they could have a big impact.

If their votes could mean so much, what’s making people abstain?

So if votes are so valuable, why aren’t people who actually DO make it to the polls … using them on every ballot measure?

There have been a lot of good pieces of interviewing journalism on why people aren’t voting this year (or at all). But I wanted to look at this question from a different perspective, by examining my dataset rather than asking voters for their personal explanations.

The wider the margin, the more abstentions (for local races)

Screenshot 2018-11-04 21.58.40.png

I’ve continued to mark out races that are flippable (by my test above) here, but am focused on something else: the correlation between the number of blank ballots and the winning margin of the winning side, though only for municipal races. (The correlation for county-wide is likely underpowered and therefore insignificant).

By Spearman’s rho, there’s a statistically significant relationship indicating that the bigger the vote margin  gets, the bigger the number of blank ballot lines gets (p = 3.04\times10^{-5}).

Why is this?

Is this because people feel they need to vote less on less contentious issues? Is it because people who feel really extremely vote, and others don’t, so the vote looks extremely split?

That much we can’t tell from just this data (that’s where population-wide polling data comes in…), but we can say there is a strong relationship between these two quantities.

So we’ve learned:

  1. Voters don’t vote for judges.
  2. Voters lose steam as they move down the physical ballot, or away from items they’re interested in voting in.
  3. Blank votes are not a trivial number of votes — in 30% of county and 20% of municipal races they could outright flip the direction.
  4. The less competitive a municipal racethe more blank ballots there likely were.

Turnout is the beginning, but it isn’t the end.

If you’ve had all the time in the world you could possibly need to understand a measure and you say, “I’ve done all the research I want, I can’t possibly choose [candidate | some choice | yes | no ],” abstaining is a valid exercise of your right to vote.

But are voters abstaining for that reason? Would they really feel empowered to make informed decisions about every candidate, including the decision not to decide? Even if 100% of people went to the polls, do they have the opportunity to learn and think through what they’re voting for?

I’d argue the hints of ballot fatigue suggest that isn’t what’s going on.

Getting to the polls is just the first step. 

It’s what we do once we get there that fulfills that civic duty and cements the act of participating in a democracy — whether that’s participating by voting, or by abstaining.

And if citizens don’t feel ready for that, even in a world of perfect, 100% turnout, we won’t have a representative voting process.

It isn’t just about turning up to the polls. It isn’t about checking a box. It’s about feeling empowered to make choices, not feeling distracted, frustrated, and throwing up your hands and saying: “Whatever, I guess I just won’t vote on this measure”.

It is not enough to just show up. 

**This does not reflect felons and others who are deprived of their votes.