One of the most pernicious flaws in American democracy is gerrymandering. Political gerrymandering for partisan advantage has been with us for roughly two hundred years, and the Supreme Court has decided that it’s perfectly acceptable.1 Gerrymandering reduces political competition and also results in legislatures that do not adequately represent the political views of citizens.
One highly effective solution to the distortions of popular will caused by gerrymandering is to use a proportional representation system. This ensures that the balance of power between political parties closely matches the balance of support by voters. This article will focus on mixed-member proportional systems, most notably used in Germany (since 1949), New Zealand (since 1994), and Bolivia (since 1994). It is also used in the Scottish and Welsh parliaments.
Voting for parties
With a few exceptions, most election systems that can be classified as proportional representation require voters to vote for a political party, not a particular person. Others use large multi-member districts.
Americans generally distrust political parties and - outside of a few areas - have limited experience with voting methods more complex than plurality or approval voting. Is there a way that Americans can cast their votes directly for a local candidate in a simple plurality election and still have proportional outcomes?
Voting for parties by voting for people
In what follows, I will use the 1992 elections for the United States House carried out in North Carolina as a convenient example because it involves a relatively small number of seats and a significant amount of gerrymandering. Please note, however, that in the US context, the main target for proportional representation as an electoral reform should be state legislatures.2
As I’ve discussed previously, the 1992 Congressional district map for North Carolina was gerrymandered by a Democratic legislature. Republican candidates won four seats with 1,204,983 votes while Democratic candidates won eight seats with 1,282,474 votes. Based on the fraction of votes earned, Democrats deserved 6.1 seats, while Republicans deserved 5.7 seats. A fair proportional allocation would have been six seats each. To go from 8-4 to 6-6 requires flipping two seats.
The easy way to force the results to be proportionate is to take the two races in which the Republicans were closest to winning and declare the Republican candidates the winners of those races. These would be the 3rd and 5th districts. The problem with this is that Tommy Pollard and Richard Burr only had the support of a minority of the voters of their district, meaning they may not represent that district’s local interests very well.
This problem gets even more severe if a third party earns enough of the vote to earn a single seat; for example, if the Libertarian Party had earned three times as many votes, this method of proportional allocation would have put J. Wendell Drye in Congress as the representative for the 8th district while earning less than 15% of the vote in his district.
Mixed-member proportional representation
Mixed-member systems finesse this problem by splitting the seats between district seats and at-large seats, reserving somewhere between a quarter and a half of the seats as at-large seats. If North Carolina’s twelve seats were divided into eight district seats in 1992, for example, the result could easily have been Democrats winning five or six seats due to gerrymandering - something like this:
However, if Democrats won the district seats 6-2, then Republicans would be assigned all four at-large seats to make up for the disproportionate outcome. If Democrats won the district seats 5-3, then Republicans would get three at-large seats and Democrats would get one. These at-large seats could either be assigned to the top vote-getters in the party or taken from another list. In either case, the Democratic Party would not have benefited from gerrymandering, and thus might not have gerrymandered the districts in the first place.
Similarly, if the Libertarians had earned a large enough share of the vote to deserve a single seat without earning a plurality of the vote in any single district, they could be assigned an at-large seat to either “list” candidates or to their top-performing district candidate.
Variations
The version of mixed-member proportionality I’ve described above uses votes for individual candidates to determine party votes. Other mixed-member systems give voters a separate party vote, allowing them to vote for a party that isn’t running a candidate in their district, or to support a party even though they dislike the party’s local nominee for their district. New Zealand uses this type of system.
Highly localized parties and voters splitting their electorate and party votes can lead to “overhang” seats, where a party wins more seats at the district level than their proportionate share of the legislature, even including at-large seats. Some versions of MMP add extra seats in order to compensate for overhang problems.
Others allow the overhang seats to create modest disproportionality - or strip the overhang seats away from some local winners, as in my first example where Burr and Pollard would have been awarded the 3rd and 5th district seats. In this last case, the use of at-large seats limits rather than eliminates the possibility of electing a less suitable district representative.
The dual-member proportional system
A closely related system is the dual-member proportional system. Each district has two seats, one being a local seat and the other being a party seat. Parties nominate up to two candidates for each district, a primary candidate and a secondary candidate. Just as in mixed-member systems, each party is assigned a quota of seats based on their share of the vote.
The primary candidate of the winning ticket in each district automatically gets the local seat for that district. The party seats are then assigned to fill out the proportionate quota of each party, using the party’s top-performing candidates.3 This means that every candidate is linked to a specific district, and every district has equal representation in the legislature - important if there are regional issues that cut across existing party lines.
I would say there are some significant flaws in the logic offered in Rucho v. Common Cause, and I discuss those flaws here.
First, elections to state legislatures are governed only by state laws, which makes it pragmatically easier to implement. Second, state legislators are usually individually very obscure, making their party identity the main factor in most voters’ decisions outside of Nebraska. Nebraska has non-partisan elections for its unicameral state legislature. Note that electoral reforms are easier to get into effect in states with functional referendum systems.
For this purpose, in districts the party won, secondary candidates are assigned 50% of the vote for the primary candidate - meaning that a secondary candidate in a district where the party won 70% is treated as having the same base priority as a primary candidate in a district where the party lost with 35%.