Proportionality

When you click on the “Analytics” command in DRA 2020, the “Proportionality” section helps you understand the degree to which the map is fair politically.

You’ll see three parts.

Metric

The first part presents the metric:

The disproportionality is the percentage deviation from what would be proportional representation, so smaller is better. It uses fractional seat probabilities, as opposed to first-past-the-post accounting.¹ The ideal is no disproportionality, i.e., proportional delegations.²

Rating

The second part presents the rating:

Using historical data and a two times “winner’s bonus” (like the efficiency gap), the rating normalizes the raw measure to a [0–100] scale where bigger is better.³ The thermometer shows that rating which is further categorized below using a 5-point scale.

Sometimes a map is labelled “antimajoritarian” which means “One party receives more than half the votes but wins less than half the seats.”

• Because the past election data are estimates of how people will vote in the future though, we don’t label all maps that are nominally antimajoritarian as likely antimajoritarian. Maps get labelled as antimajoritarian only when the vote share is more than 2% less than 50% while the seat share is more than 50%. This accommodates a reasonable approximation of the average uncertainty in the data.
• Antimajoritarian results are so un- little ‘d’ democratic that this overrides how proportional a map is (or not) and automatically sets the rating to zero.

Notes

The notes in the third part elaborate on political characteristics of the map:

• The statewide Democratic vote share — This note describes how balanced (50–50) or unbalanced a state is politically. By convention, two-party vote shares are expressed as Democratic vote shares.
• The seat split — This note tells you the number of whole seats that is closest to proportional, given the statewide vote share (above). Then it reports the likely number of seats Democrats will win. This is a fractional number that estimates wins/losses using the same probability distribution as described above, instead of simple first-past-the-post accounting. Next, it reports the difference between the proportional and likely number of Democratic seats, i.e., the impact of the bias noted above.

Note: When one is available, we use a composite of multiple elections to assess this, rather than any single election. You can choose a different election dataset, using the Data Selector.

Methodology

To estimate the partisan characteristics of a map, we use:

• A composite of statewide elections, and
• Fractional seat probabilities instead of all-or-nothing accounting, except as noted.

Footnotes

1. For details on the probability function, see this white paper.
2. The Advanced view reports many measures of bias. See Two Notions of “Fair” for some background on the difference.
3. For more details and context, see Ratings: Deep Dive.