Factional Voting in Local Elections: The Case of Cambridge, MA

This post is part of the Urban Affairs Forum* Local Elections Scholars Series.  If you are interested in writing about local elections in the places you live or study, contact Mirya Holman at mholman@tulane.edu.

By Jack Santucci

Cambridge (MA) is the last of 24 U.S. cities to elect its assembly with the single transferable vote (STV). The point of this system is for a group with, say, 30 percent of votes to end up winning 30 percent of seats — if voters sort into groups. But are voters actually sorting into groups under this STV system? Seventy, fifty, or even thirty years ago, those groups were political parties. As the city became overwhelmingly Democratic, that party system collapsed.

Where does party competition now stand in Cambridge? Are voters dividing themselves into groups? Ballot-level data back to 1997 allow us to check, and the entry last year of a controversial, new slate makes now a good time to do so. We will see that this Democratic city now has some organized factions.

1

Since 1941, voters in Cambridge have used STV, locally known as proportional representation (PR). The system asks voters to rank candidates in order of preference. Unlike single-winner ranked voting, such as the system in San Francisco or Maine, STV does not manufacture a 50-percent-plus-one majority. Rather, as the number of seats in a district goes up, the percent of votes needed to win one goes down (a “quota”). The quota in Cambridge is 10 percent: 100 percent divided by 9 + 1, as there are nine seats to fill. If your favorite candidate has 10 percent, your ballot goes to your next-ranked pick. Same if your candidate is hopeless.

Whether STV delivers party- or group-proportional results depends on three factors. First, there need to be clearly defined groups, which allows us to ask in the first place whether their seat shares are proportional to their vote shares. Second, groups cannot nominate too many candidates. If that happens, candidates may split each other’s first-choice support, causing them to get eliminated. Finally, voters from a group need to rank candidates just from their group. If they don’t, transfers risk flowing to some other group.

According to an old Cambridge quip, “PR” stands for “perpetual representation” — the same people elected over and over. Rather than present a common appeal, candidates often campaign for themselves, asking only for first-choice votes. This is not organized politics.

2

But at the last race, Bernie Sanders made headlines by throwing his weight behind a new slate. The local chapter of Our Revolution endorsed six candidates, three of which won. The election saw a record number of women running, and Council now has its first South Asian member. Another result was the in-council defeat of a long-time incumbent for Mayor.

One way to check for competitiveness is to see what the voters do with their rankings. Figure 1 gives the median number of candidates ranked in all Cambridge elections for which ballot-level data can be gotten. We see that from 1997 to 2015, the median number of rankings was about four or five. Last year, that surged to six.

So, voters on average are ranking more candidates. Are they keeping those rankings within defined groups? Using every possible pairing of candidates, I did a principal-components analysis of all 23,908 ballots cast in 2017. For each pair, a ballot is coded 0 if neither candidate appeared on it, 1 if one of them did, and 2 if the voter included both candidates in their preference ordering. The first dimension explains 22 percent of variance. A second explains another 11 percent. The gain from a third and successive dimensions falls off rapidly.

Figure 2 reveals three distinct clusters of voting behavior. One, at the top left, has candidate pairs in which both carried Our Revolution endorsement (red). A cluster in the middle almost exclusively captures pairs in which one candidate was Our Revolution-endorsed (pink). Finally, at the bottom right, is a cluster without many Our Revolution endorsements (black). I also labeled candidate pairs by the number of incumbents in each: a square for two, solid circle for one, and empty circle for none.

The result is a two-dimensional space. One dimension clearly reflects support/opposition to the Our Revolution slate. Within each grouping, however, we also see that pairs with incumbents tend to appear toward the bottom. Call this an anti-incumbent dimension.

These dimensions are not totally separate — we see incumbents in each of the blocs. This is consistent with strategic incentives in single transferable vote elections. If a party or group wants to win a majority, it needs to reach out to other groups.

It will be interesting to see what happens in 2019, especially as the ranked-choice movement gains steam. When STV first came to Cambridge, there was a clearly defined party system, which since has disappeared. Now, 77 years later, competition within the Democratic Party itself may give factional endorsement politics a new shot of excitement – at least within this STV system.

Author Biography

Jack Santucci is Assistant Teaching Professor at Drexel University, focused on electoral systems in the United States. He teaches classes on American politics, urban politics, comparative politics, research methods, and social movements. Website: www.jacksantucci.com. Twitter: @jacksantucci.

Photo from Pixabay

*The Urban Affairs Forum is presented by Urban Affairs Review, a a peer-reviewed, bi-monthly journal focused on questions of politics, governance, and public policy specifically as they relate to cities and/or their regions.

 

Leave a Reply

Please log in using one of these methods to post your comment:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s