The Associated Press breathlessly reported that “Republicans had a real advantage” for 2016 elections in both statehouse and the U.S. House of Representatives as a result of drawn district boundaries. Combining its own analysis with some academic literature, it concluded that in both arenas the GOP won more seats than the distribution of votes among parties in states would have suggested, alleging that this came as a result of partisan gerrymandering.
That politicians have engaged in this strategy for two centuries certainly is not new, and unquestionably, as past studies have shown, at the margins it influences seat distribution. But in claiming a more significant role for partisan gerrymandering, the AP study resorts to dubious assertions and makes the classic error of confusing association with causation.
The report assumes that, given the nature of voting behavior for plenary institutions, that a one percent advantage in vote share represents a two percent advantage in seats attained. Accordingly, a higher ratio means something else operates, such as a partisan gerrymander. Using this computation, the analysts estimated that, in Congress, the GOP won 22 more seats than all other things equal, or almost their majority.
Delving into the mechanics of the effort, a number of problems surface, beginning with the 2:1 assumption. The authors present no justification for choosing that as their benchmark, other than unidentified “common political expectation.” But as the literature in the translation of votes to seats makes clear, elections are not one-size-fits-all events, with a number of factors impinging on how the translation occurs and therefore on that ratio. That the analysts chose 2:1 is a complete shot in the dark that may or may not accurately describe the underlying dynamics of the 2016 House and every state lower chamber election (and, therefore, even if the hit-or-miss tactic actually hits on one, it therefore very likely missed on most or all of the others).
The study contained other questionable assumptions. It included small-delegation results, a choice that tends to accentuate bias in the direction of the more successful party. It excluded all state upper chambers, where a reasonable tactic could have included only those where all members faced selection that year. It excluded states such as Louisiana that did not have legislative elections, where a comparison of those results could have served as a baseline to weed out idiosyncratic national factors that could interfere with understanding the seats-votes relationship. And it picked an arbitrary 75/25 division (majority/minority party vote expected) for uncontested districts.
As an example of how just one erroneous assumption can change significantly results, the article notes how Wisconsin played out to have one of the largest “efficiency gaps” among the states to favor the GOP. However, in the 5-3 Republican advantage over Democrats while the GOP candidates collectively did not even obtain 48 percent of the statewide vote, two Democrat wins came unopposed. Applying as a proxy the proportion of vote given to the Republican presidential candidate in each, that suggests the 75/25 split underestimates the actual majority party support (minority in those districts), thereby overestimating the gap. And keep in mind that a review of Wisconsin’s electoral history additionally may demonstrate that the 2:1 ratio considered normal for the gap may be far off base, rendering the study’s conclusions even more useless.
Yet the biggest problem with the analysis comes from the voluminous literature establishing that the greatest portion of the gap comes from extraneous considerations, such as Democratic voters’ concentrating disproportionately in overwhelmingly liberal urban districts with Republican voters spreading out more evenly in the suburban, exurban, small-town, and rural districts. Or, that this tendency becomes more marked in rapidly-urbanizing states – almost all of which have increasingly Republican statewide tendencies. (Some researchers dispute the prominence of the “human geography” factor, but only by having to make ludicrous, facially invalid claims like, for example, California’s allegedly nonpartisan districting commission has not produced consistently pro-Democrat biased maps or that Illinois does not present a Democrat gerrymander).
Also, the Voting Rights Act itself forces gerrymandering that can masquerade as partisan. This relates to the “human geography” argument in that residential patterns typically concentrate racially, and that blacks overwhelming vote for Democrats. As the VRA prohibits racial gerrymanders, map drawers have only a limited ability to divide concentrations of minorities lest they open themselves up to a legal challenge that they unconstitutionally try to dilute minority representation.
Louisiana provides interesting side information to the issue. Because of its unusual political geography, as demonstrated in its last redistricting attempt in 2011, of black residents either highly clustered in urban areas or very widely distributed outside of these, despite blacks’ proportion of the population of over 30 percent, it essentially is impossible to draw two of six districts with a majority-minority constituency (absent something like the very bizarre early 1990s “Zorro” district).
Additionally, at the state level Louisiana would prove most vexing for analysis, principally as a result of its blanket primary system. That tends to reduce inter-party competition at lower-level constituencies, since they likely are more homogenous, drawn specifically to aid whichever party the geography suits. (In an aggregated constituency with a competitive two-party system indicative at the statewide level, this effect would diminish.) Reflective of this, in 2015 only six of 39 Senate and 21 of 105 House seats had at least two major party candidates appear on a ballot on election day.
In short, the study makes a number of dubious choices and assumptions that make it difficult to accept its conclusions; different ones equally if not more plausible could lead to very different, otherwise unknown, conclusions. And the fundamental error of assigning any deviation from the expected as necessarily a product solely of partisan gerrymandering simply is unsustainable.
While pretty obvious that North Carolina for Republicans and Massachusetts for Democrats represent attempts at partisan gerrymandering for Congress, on net the impact from such efforts likely creates a gain of just a handful of seats now for Republicans or for Democrats in the past. However, more to the point, if this happens, there’s nothing inherently wrong in the process that produced this result. If one set of elites feel aggrieved that the party they oppose uses the tactic to pad its advantage in the House or in state chambers, these complainers have access to a simple solution: present an agenda to voters that wins elections for their side. No other recourse is needed.