Ternary plot of Bush versus Gore as run in the precincts constituting Texas Congressional District 24 in 2000.  Larger dots indicate more populous precincts; brighter shades indicate a greater Gore share of the two-party vote.  If one drops an imaginary vertical line from the Hispanic vertex to the bisector of the triangle’s bottom leg, one sees that, generally, most non-Hispanic voters in predominantly Hispanic districts were white voters.  If white voters in predominantly Hispanic precincts voted more Democratic than white voters in predominantly white precincts, aggregation bias could affect the estimates of the preferences of Hispanic voters, who turned out in low percentages. Image courtesy of Crimson Grid

A few months after arriving at Harvard Law School, Assistant Professor Jim Greiner turned to computer scientists at the nearby School of Engineering and Applied Sciences for some legal power—of a sort.

He sought the computational muscle of Crimson Grid, based at SEAS, and other grid systems including the Research Computing Environment at the Harvard-MIT Data Center, to analyze and uncover the often hidden complexity of the redistricting process in elections.

Typically, after a census, the boundaries of various U.S. elective districts are “redrawn” to ensure that equal representation is maintained relative to any changes in the population.

Some creative state legislators redraw the lines in ways that disfavor racial and ethnic minorities or favor their own party, a tactic called partisan gerrymandering.

Good boundaries make good tactics

Greiner points out that because voting happens behind closed curtains, finding the truth between the lines—the way redistricting may affect results relative to how people voted—proves difficult, especially as data from standard methods like exit polls may not be available or reliable.

Grid computing—relying on drips and drabs of idle cycle time from hundreds of individual computer processors—has the oomph to crunch through decades of census and voting data and the ability to run sophisticated Bayesian algorithms.

Working with Government Department Professor Kevin Quinn, Greiner has harnessed this power to compute a voter’s eye view of an election.

In short, Greiner and Quinn are using clever computation to reveal how individuals voted—without violating their rights or revealing their identities.

Combining a knowledge of law with statistics, Greiner can then determine, for example, if a district is racially polarized along party lines.

How much advantage can politicians gain by redrawing electoral boundaries? Jim Greiner and Kevin Quinn are using grid computing to crunch political and legal numbers. Stock image from sxc.hu

A novel marriage of statistics with law

Over the next few years, Greiner envisions creating an online clearinghouse for redistricting information: a site where legal scholars or advocates may enter or view election data and parse the results using statistical tools backed by the computational power of the grid.

He imagines such a tool would be invaluable, providing solid evidence for redistricting trials and limiting unfair election practices.  

The novel marriage of statistics with law makes Greiner one of the few scholars in an emerging field. He hopes, however, that he will soon be in good company.

“There’s currently only a nascent tradition of interfacing law with good pure empirical quantitative techniques,” he says. “The biologists know they need the statisticians. But in law what we do is guess. And if we ever gather data, which we usually don’t, we try to slap it into existing software that is almost never a good fit.”

Greiner suggests lawyers could learn a lot from their colleagues in the sciences, even those looking skyward rather than dealing with the minutia of legal matters.

“All that stuff that comes off the Hubble comes straight to the statisticians before it reaches the astronomers,” he says.

If Greiner has his way, relying on the law of numbers will become standard practice in ensuring that elections remain fair and open.

- Michael Rutter, Harvard University