Partisan Gerrymandering

[muon2]

There have been a couple of subthreads on the philosophy of partisan gerrymandering that popped up over the holiday weekend. Unfortunately some parts devolved into flamewars, and became a distraction to the underlying thread.

I admit that I have applied a blunt instrument to clean up some of these, but I do want to leave a thread for a continued discussion that can avoid parsonal attacks. To that end I have attached part of the subthread from the MI discussion as a starting point.

[muon2]

On Hispanics and VRA:

1) Solomon Ortiz, an Hispanic Democrat, was beaten in 2010 by an Anglo Republican in a district that is, presumably, protected by the VRA. In that election, the "prefered" candidate of Hispanic voters was Ortiz. Was this outcome illegal under the VRA?

The VRA is not out any specific outcome. It's about the opportunity for the minority to elect the candidate of its choice. Elections are not strictly deterministic, and an opportunity is realistically a statistical statement of likelihood. If a specific outcome recurs despite other electoral factors, it probably show whether an opportunity exists or not.

[muon2]

On Hispanics and VRA:

1) Solomon Ortiz, an Hispanic Democrat, was beaten in 2010 by an Anglo Republican in a district that is, presumably, protected by the VRA. In that election, the "prefered" candidate of Hispanic voters was Ortiz. Was this outcome illegal under the VRA?

The VRA is not out any specific outcome. It's about the opportunity for the minority to elect the candidate of its choice. Elections are not strictly deterministic, and an opportunity is realistically a statistical statement of likelihood. If a specific outcome recurs despite other electoral factors, it probably show whether an opportunity exists or not.

Question: what percentage do you believe reflects that "opportunity?" 90%? 80%? 70%? 60%? 50%? 40%? 30%? ["statistical  statement of likelihood"]

The percentage depends on the particular area. The best analysis would look at election results in both the primary and general election. Data by precinct can be fit through regression or similar techniques to model voter preference based on race or ethnicity. The model can then be applied to determine the most probable outcome for a particular district in the absence of external factors like incumbency.

Since this type of modeling requires data for a particular district it can't easily be converted into a fixed percentage. Some advocacy groups will substitute a percentage to use as a rule of thumb during the map making process. For example, MALDEF has frequently used 65% of the total population as a benchmark for Hispanic districts. However a court will generally want a more substantial analysis should a map face a legal challenge.

[muon2]

On Hispanics and VRA:

1) Solomon Ortiz, an Hispanic Democrat, was beaten in 2010 by an Anglo Republican in a district that is, presumably, protected by the VRA. In that election, the "prefered" candidate of Hispanic voters was Ortiz. Was this outcome illegal under the VRA?

The VRA is not out any specific outcome. It's about the opportunity for the minority to elect the candidate of its choice. Elections are not strictly deterministic, and an opportunity is realistically a statistical statement of likelihood. If a specific outcome recurs despite other electoral factors, it probably show whether an opportunity exists or not.

Question: what percentage do you believe reflects that "opportunity?" 90%? 80%? 70%? 60%? 50%? 40%? 30%? ["statistical  statement of likelihood"]

The percentage depends on the particular area.

No, as a simple matter of logic, the percentage is exactly the same in every area.The percentage of minority voting age population could vary dramatically, but, the likelihood that that minority candidate wins ought to be identical in every area. Does the minority candidate have a reasonably fair shot at winning if his odds of winning are 50%, or does the VRA mean that he must have a greater than even chance of winnings, say 60%, 70%, 80%, 90%, 95% etc?

You must be logged in to read this quote.

This isn't an answer to the question I asked. I didn't ask what the minority population ought to be to insure a  fair shoot at electing the prefered candidate. I asked, what probability defines a fair chance.

As much as you would like a hard and fast probability, the courts have made it clear that it won't be defined that way. The SCOTUS will not give the states a safe harbor for redistricting minority populations, since past history leads them to believe that the states will inevitably find a loophole to deny minorities the ability to elect candidates of choice.

My background is in mathematics and I would love to offer a clear statistical test. However, the only answer that is applicable to the VRA is a legal answer, which is what I provided.

[muon2]

On Hispanics and VRA:

1) Solomon Ortiz, an Hispanic Democrat, was beaten in 2010 by an Anglo Republican in a district that is, presumably, protected by the VRA. In that election, the "prefered" candidate of Hispanic voters was Ortiz. Was this outcome illegal under the VRA?

The VRA is not out any specific outcome. It's about the opportunity for the minority to elect the candidate of its choice. Elections are not strictly deterministic, and an opportunity is realistically a statistical statement of likelihood. If a specific outcome recurs despite other electoral factors, it probably show whether an opportunity exists or not.

Question: what percentage do you believe reflects that "opportunity?" 90%? 80%? 70%? 60%? 50%? 40%? 30%? ["statistical  statement of likelihood"]

The percentage depends on the particular area.

No, as a simple matter of logic, the percentage is exactly the same in every area.The percentage of minority voting age population could vary dramatically, but, the likelihood that that minority candidate wins ought to be identical in every area. Does the minority candidate have a reasonably fair shot at winning if his odds of winning are 50%, or does the VRA mean that he must have a greater than even chance of winnings, say 60%, 70%, 80%, 90%, 95% etc?

You must be logged in to read this quote.

This isn't an answer to the question I asked. I didn't ask what the minority population ought to be to insure a  fair shoot at electing the prefered candidate. I asked, what probability defines a fair chance.

As much as you would like a hard and fast probability, the courts have made it clear that it won't be defined that way. The SCOTUS will not give the states a safe harbor for redistricting minority populations, since past history leads them to believe that the states will inevitably find a loophole to deny minorities the ability to elect candidates of choice.

My background is in mathematics and I would love to offer a clear statistical test. However, the only answer that is applicable to the VRA is a legal answer, which is what I provided.

"Realistically a statistical statement of likelihood," given my background in math sounds something on the order of a two sigma event, or more. While courts may have been deliberately vague, surely, their words can be estimated by some ballpark figure.

One point is that "realistically, a statistical statement of likelihood" is a standard that a party could use to create favorable seats in redistricting. In other words, courts are demanding gerrymandered districts in which the outcome is, barring extraordinary circumstances, predecided before election day. [Except in gerrymandering the majority party might cut the seats a little more closely to maximize the number of seats.]

The second point is whatever that standard is, it isn't resulting in the prefered candidate of the protected class winning the nomination of the prefer party in the primary. Davis beat Hilliard in Alabama on the strength of White voters, while Hilliard was the "prefered" candidate of Blacks in the district. Sure Davis, like Hilliard, was Black, but he wasn't the "prefered" candidates of Blacks.

Because the courts have been vague on the math, there have been more than one type of statistical analysis of voting preference that has been used to justify or overturn minority districts. At a conference earlier this year I heard about at least three such methods. Needless to say, competing methods can give different results for the same district.

There's no question that the parties will attempt to use VRA districts as part of their strategy to draw the map in their favor. The use of competing methods to determine the ability of a district to perform electorally for the minority party is part of the legal strategy. Consideration of minority voting strength in a primary is also part of the strategy, but how much weight it is given will depend on which party is using it.

It is also the case that the behavior of the parties on these question will be different for different minority groups. The cohesiveness and turnout rates of the minority groups are not the same, and this impacts whichever statistical analysis is performed. The one likelihood is that the parties will select a model and data set that best supports their desired goals for a map. Then it's up to the lawyers to sell that model to the court.