A Brief Analysis of the Cambridge City Council Election

A data dive into the Cambridge City Council Election of 2019

What's with all the yard signs?

The Cambridge, MA local elections took place on November 5, 2019. In the weeks leading up to it, I walked by an increasing number of yard signs and arrived home to pamphlets supporting various candidates. I realized that unless I made an effort to research the candidates (which I did), the main way someone like me would know anything about the candidates is via those pamphlets or by name recognition from the number of signs everywhere.

pamphletsimage

So I started to wonder: What matters for this election? Is it pure name recognition? Is it incumbency? I only have so much time on my hands and some data is lower hanging fruit, so I thought I'd narrow my scope and focus on campaign finance data. So we can ask the following questions:

  • How much does fundraising impact the outcome of the election?
  • How much does incumbency impact the outcome of the election?

Some quick online research brought up an article from 538 Politics that discusses the impact of money in elections, albeit at a congressional scale. A paper titled Political Experience and Fundraising in City Council Elections discusses what factors are most important in fundraising for city council elections. The paper finds that "fundraising is a function of incumbency and prior experience", plus "political endorsements are also important".

Though I'm sure this is a well researched topic in academia and in some political publications, I think adding another data point to the corpus of analysis couldn't hurt! Note that I am only analyzing the city council election, not the school committee one. Let's see what we can find!

Deep dive: What we can glean from public data

Let's set the scene. The people who won the election have their names written in green. The "starting balance" is how much money the candidate already had coming into 2019 (which matters for accurately tracking incumbent wallets), "raised" is how much money they raised between January 1st and November 5th, and "expenses" is how much they spent in the same time period.

Some helpful context: There are nine seats on the council, and eight incumbents in the race. See here for official election results, and see here for a list of incumbents and candidates. Notably, Craig A. Kelley was the only incumbent running who did not get re-elected. Patricia M. Nolan and Jivan Sobrinho-Wheeler are the two newcomers who did get elected.

Overall fundraising numbers: Incumbents tend to raise more

election fundraising

Some interesting observations: Both newcomers fundraised far less than many others in the race, yet still won. They did, however, fundraise more than many of the other newcomers. Craig A. Kelley and Adriane Musgrave are in the top half with their fundraising numbers, but fell short. Adriane's fundraising as a non-incumbent stuck out to me, and it makes more sense once I realize that they have done a bunch of work to benefit the community and have also worked in business circles. This rings consistent with the aforementioned paper's findings on how prior experience can contribute to better fundraising numbers in a local election. Patricia's victory despite their fundraising also stuck out, and makes more sense once I realize that they were previously on school committee, making them a pseudo-incumbent.

Categorized contributions: Individual donors reign supreme

I thought it would also be interesting to look at the breakdown of contributions, mostly because I wanted to see if any candidate received a substantial amount of money from unions, trade organizations, PACs, or any other larger body.

donation breakdown

Most of the donations seem to come from individual contributions, with relatively small amounts of committee and union/association contributions. A number of candidates also financed their campaigns with loans, and I think it is interesting to note that the incumbents borrowed far less money. That seems consistent with the fundraising advantage for those with political experience. Keener eyes might notice that these fundraising numbers are not the exact same as the totals from the first chart. I dug into the data and realized that not all receipts are itemized thoroughly, so this contribution-by-type chart misses some of the money. For example, Ilan Levy has $0 fundraised on this chart, but that isn't accurate. Since there aren't itemized, categorized receipts for all of their fundraising, I sadly can't gather them and display them here.

Number of donors vs. size of donations: Number of donors more significant

I apologize in advance for this plot but I really wanted to make it:

boxplot

OKAY SO this details the distribution of individual donations (the blue bars on the chart before this one) to each candidate. The left y-axis corresponds to the transparent blue bars to show the number of individual donors, and the right y-axis corresponds to the boxplot of the range of donations for each candidate. Note: Timothy J. Toomey Jr. has two rather large outliers, where one is about a $2k donation and one is about a $5k donation. I omitted them from this graph because it is already squished enough as is.

The reason I wanted to see the number of donations and the distribution of donation dollar amounts is to see what matters more, if anything. Most candidates have a median donation of under $100, barring Sukia Akiba. So the more powerful piece here seems to be the number of donations, since most of the higher blue bars correspond to election winners (barring Kelley and Musgrave).

As one would expect, those who won had (in total) a higher mean donation ($178 vs $155), a higher median donation($100 vs $50), a larger total number of donors (1904 people vs 1347 people), and a larger overall amount of money ($339k vs $210k).

So how much does fundraising impact the outcome?

We've explored the data, but how can we measure the impact of fundraising? We can perform a logistic regression where the input variable is the amount fundraised and the output variable is whether or not the candidate was elected:

Fundraising vs election outcome

We see some relationship, but it isn't a very strong one. Again, we are limited by our 21 candidate sample. We can also apply a linear regression where the input variable is the amount fundraised and the output variable is how many votes the candidate received in the first count (Cambridge employs a ranked choice voting system):

Fundraising vs number of votes received

Again, we see a relationship and some correlation (I found an r-squared value of 0.51), but it is not a strong one. Now lets take a quick look at incumbency before trying to answer how much fundraising impacts the election.

How much does incumbency matter?

fundraising violin plot

Since seven out of eight incumbents won, it is no surprise that the "elected" violin above looks very similar to the "incumbent" one. If I had to predict the election outcome based on a simple heuristic, given the above plot I would likely just predict all the incumbents to win. It is a clear, simple correlation that has lots of easy arguments to support it.

The unsurprising takeaway: Incumbency tied to stronger fundraising and winning

How much does fundraising impact the outcome, and how much does incumbency impact the election? It is hard to split them apart. We've seen that incumbents tend to raise more money, but perhaps they were initially elected and are incumbents because (in part) of their fundraising ability. The two variables are inextricably tied together. We have seen the moderate relationship between fundraising and acquiring votes/winning, and we have seen that 7/8 of the incumbents won their seats back (and if you look at the count-by-count of the ranked choice voting steps, the eighth incumbent very narrowly lost).

My takeaway is consistent with the paper mentioned before: incumbency seems moderately correlated with fundraising as well as with winning the election. I know, I know, this is not a surprise, it is just one election, I hear you. However, I think it is still useful to try to validate such obvious hypotheses just in case they're wrong, or if there is anything interesting we can unearth that furthers our understanding. And there's nothing wrong with confirming a prior! Remember that this is simply a correlation; proving causality requires a lot more work than just one data dive into one election.

If you made it this far, I hope you enjoyed this! Feel free to contact me at raaidarshad@gmail.com if you have any comments, questions, or want to (politely) point out that I did something wrong.