A simple fix for gerrymandering

It is widely believed that the partisan gerrymandering of election districts is unfair, but revising district boundaries to alleviate the problem is easier said than done.

Proposed methods for doing so involve such abstruse concepts as “efficiency gaps.” Perhaps his characterization was too harsh, but Supreme Court Chief Justice John Roberts probably spoke for many when he derided such approaches as “sociological gobbledygook.”

{mosads}To prevent gerrymandering by politicians, the citizens of Colorado, Michigan and Missouri have just voted to have their election districts determined by experts. But this approach has not been conspicuously successful in California, which pioneered it.

There, the vote tallies for each political party diverge from their seat shares in both Congress and the state assembly as lopsidedly as in other states.

Yet, one can imagine a way to dispel the pernicious effects of gerrymandering without the need to modify any election districts. It can be described using data about the 2012 Wisconsin election for the state assembly, which was at the heart of a gerrymandering case that the Supreme Court considered this past term.

In that election, Republicans won in 60 of the 99 assembly seats even though they only got 48.6 percent of the combined votes for Republicans and Democrats. (Third-party vote totals were negligible).

Suppose that, mirroring the statewide vote split of 51.4 percent to 48.6 percent, Democrats were declared the winners in 51 of the 99 districts. These districts would consist of the 39 that they won outright plus the 12 additional districts in which they were the closest runners-up. The remaining 48 districts would be allotted to the Republicans.

This approach is simple and can be implemented quickly and unambiguously once election results are at hand. It would focus directly on the disparity between statewide and gerrymandered results that is the actual source of frustration. By definition, this “statewide match” proposal would eliminate that disparity.

As one explores the idea further, it becomes more attractive. The Wisconsin Republicans won 60.6 percent of the assembly seats in 2012. A total of 2.66 million two-party votes were cast in assembly races and, ideally, 60.6 percent of the seats would correspond to 60.6 percent of the votes, which is 1.61 million votes. 

Yet, the Republicans only amassed 1.28 million votes statewide. In effect, the Republicans were awarded 330,000 votes (1.61 million minus 1.28 million) that actually went to Democrats. Put another way, 330,000 Wisconsin voters essentially had their votes “reversed” and assigned to the other party. This would not happen under a “statewide match” proposal. 

Of course, declaring Democrats the winners in 12 Wisconsin districts where they were runners-up would also entail vote reversals. But such reversals would be relatively rare. Typical of these districts is District 50, in which the Republican candidate got 12,842 votes and the Democrat 11,945.  

Actually, to win that district, the Democrat would have needed just over half of the 24,787 votes cast, which works out to 449 additional votes. Thus, declaring the Democrat the winner in that district would be effectively to reverse 449 Republican votes and assign them to the Democrat.      

Over the 12 districts in which Wisconsin Democrats came closest to winning, the total number of such vote reversals would have been 7,711. In other words, eliminating the 330,000 statewide vote reversals tied to gerrymandering would have required 7,711 reversals at the local level.

Thus, the number of reversals would have dropped by nearly 98 percent. Calculations for other states accused of gerrymandering (e.g., Maryland, Pennsylvania, Michigan and North Carolina) yield similar outcomes, in both assembly and congressional elections. In California, reversals drop by over 97 percent compared to those arising in the districts generated by experts.

Admittedly, it seems less than ideal that “statewide match” outcome would deviate from “first past the post” results in some individual districts.

But “first past the post” is already under challenge by such approaches as ranked voting (which was recently used in San Francisco) and, as noted, there is evidence that the runners-up who get seated under “statewide match” would only slightly have fallen short. 

The approach combines simplicity, practicality and fairness. It deserves nationwide consideration. 

Arnold Barnett is the George Eastman professor of management science and a professor of statistics at the MIT Sloan School of Management.