Armed with novel statistic, researchers push back against partisan bias in redistricting

Jun 01, 2011 By Sarah Anderson

Voting districts are redrawn every 10 years — after each U.S. census — to reflect changes in population. The process can be manipulated by "gerrymandering," drawing districts that have irregular shapes to effectively "stack the deck" in favor of one party or another. Today, as mapping capabilities have evolved, political insiders can create districts with unremarkable shapes that still lock in safe seats for one party.

In recent years, voters have attempted to rein in such abuses, with Californians passing two ballot measures, each declaring, "In the current process, politicians are choosing the voters instead of voters having a real choice. This reform will put the voters back in charge."

Still, the cross-cutting interests at play and the complexity of the process can make it difficult to determine whether partisan bias will be better or worse with a given redistricting plan.

Now, researchers at UCLA have developed an innovative one-number summary for evaluating redistricting plans, relying solely on census data and geographical information. The measure makes comparing plans as transparent as assessing the gas mileage of different cars or the energy efficiency of various appliances.

Tom Belin, a professor of biostatistics at the UCLA School of Public Health, led the research effort that gave rise to the idea of a "density-variation/compactness," or DVC, measure, showing that district plans that have less variation in population density and greater average geometric compactness tend to have less partisan bias. 

The research is published in the current issue of the online journal Statistics, Politics, and Policy (bit.ly/lAKIHL).  ;

Despite landmark "one-person, one-vote" Supreme Court rulings in the 1960s, which held that districts of unequal size underrepresent some voters' interests and overrepresent others, it is still possible to "pack" voters of the opposition party into a small number of districts while maintaining a partisan advantage in other districts.

The researchers compare current redistricting dynamics to a scenario in which one of the teams vying for the NBA championship would be allowed to dictate that several games in the seven-game series would be played six-on-four rather than five-on-five — as long as the average number of players across games was five to a side.

Not every legislative district will be as competitive as a five-on-five basketball game even with the DVC information, the researchers note, but plans with higher DVC scores tend to contain more districts in which close elections would be expected.

The researchers developed the DVC measure using voter registration data from California, then tested its performance on election data from Texas, where there have been several high-profile redistricting battles over the past decade. As with grade-point averages, a DVC value of 4.0 would be a high score, but the DVC score can also be a negative number if a plan has less balance in population density than there was in the year 2000, which was taken as a baseline year. As it happened, all of the actual redistricting plans since 2000 in California and Texas had negative DVC scores.

The approach could have implications for polarized politics in many states and in the nation as a whole. 

"Would Californians have embraced the redistricting done in 2002 knowing that its DVC score was -2.92, when a DVC score of 3.00 was possible?" Belin asked. "Maybe. But maybe a lot of people would have been embarrassed to think that -2.92 is the best they could do by way of self-governance."

Although it might seem surprising that publishing a one-number summary for each proposed redistricting plan would have much impact, Belin cited a familiar example of how a statistic can have a game-changing effect. 

"In the 1980's, the airline reservation systems used by travel agents favored the shortest flights, so airlines gamed the system by publishing schedules with unrealistic arrival times. All it took to bring reality back into the system was to require airlines to publish on-time arrival statistics."

Would publication of DVC scores have a game-changing impact? 

Belin thinks so, and when the newly created California Citizens Redistricting Commission held a public-input hearing in Los Angeles in late April, he made the case that constituencies favoring less gridlock in government would easily qualify as "communities of interest." 

"For the sake of transparency, it is imperative that this commission report a DVC score for each plan you submit for public consideration," he told the commissioners. "The 2002 redistricting had a negative DVC score — it didn't have to — and it had consequences."

In addition to Belin, members of the research team included Heidi Fischer, a current Ph.D. student in the UCLA Department of Biostatistics, and Cory Zigler, who received his Ph.D. from the UCLA Department of Biostatistics in 2010 and is now a postdoctoral fellow at Harvard University.

Explore further: Digital native fallacy: Teachers still know better when it comes to using technology

add to favorites email to friend print save as pdf

Related Stories

Study: 'Carved-out' voters often struggle at ballot box

Aug 06, 2010

The age-old practice of dividing congressional districts evenly by population speaks to such American ideals as fairness and equality. But when a county's residents are carved into separate districts simply to maintain that ...

Study: Local newspapers keep politicians accountable

May 07, 2010

In areas where members of Congress get lots of ink in local newspapers, voters are more informed and representatives do more to serve local interests, according to a study to be published next week in the Journal of Political Ec ...

Recommended for you

Gypsies and travellers on the English Green Belt

Oct 17, 2014

The battle between Gypsies, Travellers and the settled community over how land can be used has moved to the Green Belt, observes Peter Kabachnik of the City University of New York.

Cadavers beat computers for learning anatomy

Oct 16, 2014

Despite the growing popularity of using computer simulation to help teach college anatomy, students learn much better through the traditional use of human cadavers, according to new research that has implications ...

User comments : 5

Adjust slider to filter visible comments by rank

Display comments: newest first

FrankHerbert
0.7 / 5 (48) Jun 01, 2011
Gerrymandering is a symptom of large districts. If districts were to be capped at a reasonable size (30,000-50,000 people), gerrymandering would become virtually impossible.
Yogaman
not rated yet Jun 02, 2011
@FrankHerbert: In CA, your district size would result in 600-1000 legislators. I'd rather not help to pay for the extra salaries and expense accounts and infrastructure and etc. I also suspect that getting legislation done would be more difficult and prone to influence by special interests in the larger body. So, your solution to gerrymandering seems less than ideal. Neutral scoring sounds way better.
Thrasymachus
5 / 5 (1) Jun 02, 2011
Larger legislative bodies are less prone to influence by special interests than smaller ones, though you are right in that a larger legislative body is less likely to get anything done than a smaller body. The flip side of that is that what does get passed is less likely to be the pet legislation of some special interest that harms the public, and more likely to be something that meets the needs and wants of a larger proportion of the population.
FrankHerbert
0.7 / 5 (48) Jun 02, 2011
I'm not opposed to eliminating the first past the post system, but your fears are misplaced. It's exactly because of the very large districts we have today that special interests are a problem.

Because there are so few legislators, essentially their value goes up. The "price" to influence a legislator as well as the rewards for being a legislator are inversely proportional to how many their are. Just look at the executive branch. Compare the power of the President to the power of an individual lawmaker. The power of each branch is constant in the sense that they are set by the Constitution and law. The power among the individual members of branch can be changed by adjusting their numbers.

If the lawmakers had less individual power, they'd have less ability to abuse their offices. Not just that, but I believe those less likely to abuse office would be attracted to running, because they wouldn't be competing with such power hungry people as are attracted to office today.
FrankHerbert
0.7 / 5 (48) Jun 02, 2011
Also, as far as infrastructure and travel, there is no reason the United States can't have a decentralized legislative capital. Technology would make this trivial today. With biometric scans and such it would be impossible to impersonate a legislator via teleconferencing or something similar.

I would like to see US representatives dial into Congress at their state capitals. Also if this were the case, legislation may become a part-time job as intended. I'm sure salaries would reflect this.