This post compares participation in state and local elections among likely renters and likely homeowners in Brookline, MA, where I live. I will examine voting in three elections: first, the 2019 local election, which included an open-seat contested Select Board race and property tax increases (overrides of Proposition 2 1/2) to fund the major renovation of a school and controversial building of a new elementary school (the override failed); second, the high-turnout November 2018 state elections, which included contests for Governor and Senate; third, the May 2018 local elections, which had a contested race for Select Board with two incumbents running and two less controversial but still contested property tax increases involving expanding Brookline High School and an operating override to prevent school employee layoffs (both measures passed). I’m using registered voters data from October 2019 and property ownership data from 2019. I requested more recent voting data from the Town Clerk and can update with 2020 elections when I get it.
Brookline’s population is split about evenly between renters and homeowners. But these groups differ substantially in their political participation, especially in local politics. As a rule of thumb, about three quarters of votes cast in local elections are cast by people who own their homes. In state and national elections, this number is more like 55-45. I summarized these in the chart below. Just 18% of Town Meeting Members are renters, according to the analyses I did a while ago. Turning out to vote in local elections also skews heavily towards homeowners, though less so than Town Meeting Members.

The analyses in this post look at how property ownership shapes participation in politics in different ways. First I look at the population of registered voters. Then I look at the share of voters, by election, who are renters. Next I look at the share of renters who vote. Finally, I describe the procedure I used to identify homeowners and mention some caveats.
How Many Registered voters are homeowners
My voter file from the Town Clerk from October 2019 has records for 31,705 registered voters. Of these, 16,421 (about 52%) appear to be homeowners and 15,284 renters according to the methodology described in the Methodology section at the end of the post. The map below shows the locations of registered voters in Brookline colored by homeownership (red are renters and blue are homeowners).

The map looks reasonable. Hancock Village, a large apartment complex in South Brookline on the border with Boston, shows a cluster of renter voters. Much of the rest of South Brookline, which is mostly single family homes, shows homeowners. Village Way, an apartment complex near Mission Hill is a similar cluster of renters, as are many of the Brookline Housing Authority Propertis. Beacon Street has a lot of renter voters, while the streets off of Beacon tend to have more homeowners.
This also suggests the matching procedure worked pretty well. The share of likely homeowners among registered voters is pretty similar to their share in the population; it is slightly higher, which makes sense since many student renters may be registered at a different address, such as their parents’ (likely in a more competitive state in national elections!).
How MANy Votes are cast by renters and homeowners, by election
In the May 2019 election, there were 9808 votes cast (26% of registered voters). In my registered voters data file, there are 9299 votes cast in May 2019 — the 509 vote difference likely comes from movers who have left town and had their voter registration purged from the Town Clerk when they registered to vote elsewhere, as described later, or some other recording error. Of the 9,299 votes, 6,875 were cast by likely homeowners, while 2,424 were cast by likely renters. So about 74% of votes cast were by homeowners. In the May 2018 election, there were 8,124 votes cast according to the Town Clerk (22% of registered voters). I have 7,647 registered voters recorded as voting in May 2018, of whom 5,882 (77%) are likely homeowners and 1,765 (23%) likely renters. About three quarters of votes cast in local elections are cast by people who own their homes.
National elections, which are much higher turnout in general, tell a somewhat different story. In the November 2018 election, there were 25,507 votes cast (68% of registered voters). I have records of 24,496 registered voters who cast ballots in November 2018. Of them, 13,615 (55.6%) are likely homeowners while 10,881 (44.6%) are likely renters. National elections look much more like the registered voters population than local ones. This is illustrated in the juxtaposition of maps below: on the left, each point represents a vote cast in the November 2018 election, with red indicating renters and blue homeowners; the right map does the same with May 2018 votes. The takeaway is that the map on the left is much more balanced between renters and homeowners than the right (look at Beacon St, where swathes of red disappear in the right map).

What share of renters and homeowners vote, by election
The last section looked at the probability that a given voter was a renter. Now I’ll look at the probability that a given renter (or homeowner) votes. I’m only going to look at those voters who, given their registration dates, could possibly have voted in a given election (discussed below in movers). The table below shows these proportions.
Election | Renter Turnout | Homeowner Turnout | Difference |
May 2019 | 15.9% | 41.9% | 26.0% |
May 2018 | 13.6% | 36.9% | 23.3% |
November 2018 | 71.2% | 83.0% | 11.8% |
The table illustrates how renter turnout is consistently lower than homeowner turnout, especially in local elections. There are two more subtle takeaways. First, the more controversial overrides in 2019 appear to have driven turnout to a greater extent among homeowners than renters. Second, the turnout among registered voters in November 2018 are, for both renters and homeowners, higher than the overall population’s turnout (68%)! How can this be? My registered voter data is from October 2019. When people do not vote for a certain amount of time, they are removed from the Town Clerk’s list of registered voters. Hence, many of the non-voters in 2018 have been removed from my data, and the people who are left on the list in 2019 (but registered to vote prior to November 2018) are more likely to have voted in November 2018 than the people who were on the registered voter list in 2018.
methodology: Identifying homeowners and renters
The first step in comparing the participation of homeowners and renters is identifying who is a homeowner and who is a renter. I matched the 2019 Town Assessor’s Database — which details facts about every property owned in Town, including the owner of record — with a 2019 database of registered voters obtained by the Town Clerk. For every registered voter in the database, I checked whether there was a corresponding property owner recorded in the Assessor’s Database. This matching procedure had two parts. First, for every registered voter, I identified the set of properties whose addresses were within 100 meters of the voter’s address (I used a distance based measurement rather than direct address matching because, it turns out, the addresses in the two datasets are not directly comparable, particularly for multifamily properties. For example, 75 Longwood Avenue in the voter data is part of a bigger parcel recorded as 218 St Paul Street in the Assessor’s database). I geocoded the addresses in the voter file to latitude/longitude coordinates using the ArcMap “USA_LocalComposite” Geocoder and used Christopher Schmidt‘s geocoded parcel data to get locations of parcels. Next, I checked whether the last name of the voter was very similar to the last name of one of the nearby properties in the Assessor’s Database (very similar, rather than exactly matched, because sometimes people hold property in trusts or LLCs which carry their names). For computational reasons, I actually implemented this in the opposite order — first looking for similar names, and then seeing whether there was a similarly named property owner in the immediate vicinity of the registered voter.
This match method is imperfect. A property owner who keeps their property in a trust or corporation that does not have their name in it would not match. A voter who does not own property but happens to live very close to a person who shares their last name might falsely be identified as a property owner (e.g. imagine a triple decker owned by Jane Smith who rents one floor to a tenant named Joe Smith). Another issue, albeit a small one, would be a person who very recently bought a house, who may not be recorded in the Assessor’s database.
Some considerations about movers
One issue with registered voter data is the problem of movers. Some people who left Town may have been purged from the database of registered voters but voted in a previous election before they did. These people won’t appear in my data. This is not a problem that can be addressed directly with the data I have. What I can do is compare the number of votes cast among the voters in my registered voters file with the actual number of votes cast in the election, which can give a sense of the scope of this problem. Another issue is that some individuals may have moved to Brookline after an election took place. I can address this by looking at a field containing when a person registered to vote at their current address and not include people who couldn’t possibly have voted in an election because they weren’t here yet for analyses of that election. I’m not sure how this would be affected by individuals who move from one Brookline residence to another.