How the VVI is calculated, in plain English

The VoteValue Index (VVI) is a single number between 0 and 100 that quantifies how much a single vote in a given district actually weighs. A 90 means structural conditions in that district give a single ballot serious leverage. A 30 means the system is doing most of the work and the voter is doing very little. Same election, same kind of ballot, different math.

This post walks through how the number is built. If you want the exact formula, it's on the methodology page. This is the in-between version: enough to know what you're looking at, not enough to need a political science degree.

The five factors and their weights

The VVI is a weighted sum of five subscores. Each subscore is itself a number between 0 and 1. The weights add up to exactly 1.

The weights aren't arbitrary. They were chosen based on what the political science literature says actually drives individual vote power. The biggest weight goes to competitiveness because in published vote-power models (Banzhaf 1965, Gelman et al. 2004), marginal-seat effects dominate everything else. The two structural factors (mobilization and integrity) are next because they affect every vote in the district even when the margin is close. Electoral leverage and race significance are smaller because they're either mostly fixed (district size) or apply uniformly across an entire chamber (whether the chamber is close to flipping).

Factor 1: Competitiveness (35%)

This is the simplest one. If the last election in your district was decided by 1 percentage point, your vote was, in a real sense, decisive. If it was decided by 30 points, the result was effectively foregone.

We build the competitiveness subscore from three inputs:

• Margin closeness. How close was the most recent general election? Closer races score higher.
• Volatility. How much have the margins moved over recent cycles? A district that's flipped between parties scores higher than one that's been safely one-party for decades.
• Flip history. Has the seat actually changed hands? An actual flip is a much stronger signal of competitiveness than a single close race could be.

Data source: MIT Election Data and Science Lab results from 2018 onward, supplemented by VEST 2020 precinct-level data. Most state legislative races use Ballotpedia for the most recent cycle when MIT hasn't updated.

Factor 2: Mobilization Potential (20%)

A close election only matters if people show up. Mobilization Potential captures how much turnout headroom exists in a district.

The intuition: a 10-percentage-point turnout swing in a district where 30% of eligible voters showed up is a much bigger absolute number of voters than the same percentage swing in a district where 70% showed up. The first district has more slack. A campaign or organizer working in the first district can move the outcome more easily than one working in the second.

We measure two things:

• Turnout gap. How does the district's turnout compare to the national median in comparable elections?
• Contestedness. Was the race actually contested? An uncontested race depresses turnout in ways that aren't really about the voters.

A district with a 50-point turnout gap and a contested race scores near 1.0 on this subscore.

Data source: Census voting-eligible population estimates and certified election turnout from state secretaries of state.

Factor 3: Electoral Leverage (15%)

This one is the closest to academic vote-power theory. It measures how much, mathematically, your single vote moves the outcome.

There are two inputs:

• Vote decisiveness. In a tight race, the probability that any one vote tips the result is small but non-zero. In a blowout, it's essentially zero. The expected decisiveness scales with margin and electorate size.
• District size factor. Smaller districts magnify per-vote leverage. A state house seat with 80,000 voters gives each voter more mathematical leverage than a congressional seat with 700,000.

The reason this factor's weight is only 0.15 is that for any given voter, this number is mostly determined by where they live and what office is being elected, not by anything that's changing cycle to cycle. It matters, but it's less of a lever than competitiveness or turnout.

Factor 4: District Integrity (20%)

This is the "is your district gerrymandered" subscore. We use two well-established measurements:

• Polsby-Popper compactness. The geometric ratio of district area to perimeter squared. Higher means more compact (rounder). Lower means more contorted. See our post on Polsby-Popper for the full explanation.
• Efficiency gap. From Stephanopoulos and McGhee's 2015 paper. It measures whether one party's votes are being "wasted" at a higher rate than the other's, by comparing surplus votes in won districts and losing votes in lost districts. An efficiency gap above 7% is generally considered evidence of partisan gerrymandering.

Both of these have known limitations (compactness in particular has a whole post worth of caveats). That's why we use them together instead of either alone, and why the User-driven Gerrymander Guesser feeds a separate community-derived rating that complements the geometry.

Data source: Official precinct shapefiles for boundary geometry, MIT election results for partisan vote totals.

Factor 5: Race Significance (10%)

This is the smallest weight, and it's the most "outside the district" of the five.

Race Significance measures how much the chamber that this seat sits in is close to flipping. If your state senate is 21-19 with two seats up, every one of those seats is heavily weighted, because the chamber's control is on the line. If it's 36-4 and a single seat flipping wouldn't change anything, the chamber-level stakes are lower.

We use three inputs:

• Chamber margin. How close is the chamber to the majority threshold?
• Uncontested penalty. If the specific race isn't contested, this factor is muted.
• Incumbency. Open seats score higher than incumbent seats, because the floor of voter familiarity is lower and the seat is more genuinely up for grabs.

The reason this factor only gets 10% of the total weight is that it applies relatively uniformly across all districts in the same chamber. It's important to whether a vote matters in the broader political picture, but it's less specific to any one address than the other four factors.

The exponential curve at the end

Here's something most people don't notice when they look at the score.

After we compute the weighted sum of the five subscores, we don't just multiply by 100. We apply an exponential curve:

VVI = 100 × (weighted_base) ^ 0.4

The ^ 0.4 raises the weighted base to the 0.4 power.

Why? Because without it, virtually no district would ever score above 50.

Even the most competitive, fair, high-turnout districts in the country have a weighted base of around 0.4 to 0.5 (out of a theoretical maximum of 1.0). The raw 0.5 ceiling is real, because no single district can be perfectly competitive AND perfectly fair AND perfectly high-turnout AND perfectly chamber-decisive AND perfectly small all at once. The factors trade off.

Without the curve, the most competitive seat in America might score 47 and the average safe seat might score 18. That spread compresses everything into a narrow band where nobody can tell the difference between "your district is structurally meaningful" and "your district is structurally noisy."

The exponential curve (x^0.4) preserves the ranking, but spreads the scores into the full 0-to-100 range that voters intuitively expect. A high-impact district lands in the 70s or 80s, where it looks high-impact. A low-impact district lands in the 20s or 30s, where it looks low-impact. The relative comparisons between districts are mathematically identical to using the raw weighted base, just easier to read.

This is documented in the methodology footnote on the home page, but it's worth calling out here: the underlying factor values are unscaled. Only the final number you see has been curved. If you want the raw weighted base, the API returns it.

What the VVI does not measure

The VVI doesn't tell you:

• Whether your representative is good at their job.
• Whether the candidates running this cycle are credible.
• Whether the issues being voted on are important to you.
• Whether you should vote (you should).
• Who you should vote for.

It only tells you how much structural leverage your vote carries. That's a precondition for voting having impact. It is not the same as voting having moral or civic value, which is a separate thing entirely.

The whole point of putting a number on the structural piece is to free voters to think clearly about the rest.

Try it: type any U.S. address into VoteValue and the VVI for each of your districts will appear with all five subscores broken out. Every input value, every formula, and every data source is documented on the methodology page.