My original plan for Project 3 was to research election results in Cary, NC, the town I sent postcards to and focused on in Project 2. However, that data hadn’t been released by Thanksgiving, so I chose a different route. Another 2020 race I followed closely was Sara Gideon’s challenge to Susan Collins in the US Senate. I participated in this race by making donations and by switching my voter registration back from New York to Maine, since I resided there for the majority of 2020 and was therefore legally able to do so. 

Collins has historically presented herself as a centrist who doesn’t always vote along Republican party lines, and had been endorsed by Planned Parenthood of Maine for many years until they switched their endorsement to Sara Gideon. Many Mainers have grown disillusioned with Collins’ ostensible centrism, particularly after she cast a decisive vote in 2018 to confirm Justice Brett Kavanaugh to the Supreme Court. Nationally, Democrats counted Gideon as as an important (and relatively safe) step towards taking back control of the U.S. Senate. The poll aggregator FiveThirtyEight gave Sara Gideon 2:1 odds of winning the seat in its final prediction before the election. In the end, however, while Maine went for Biden with over 52% of its vote, Gideon lost her bid for the Senate seat with less than 42% of the vote. Collins even secured a majority, with just over 50%.

This Senate race gained national attention because people saw it as a chance to flip the Senate to a Democratic majority, so I think the nature of the results could be interesting to anyone who was interested in the race. That being said, I was trying to work with a Maine audience in mind. I think, to some extent, I accomplished that — I’ve been looking at different political visualizations of Maine’s results since I finished my own, and I haven’t seen any that explicitly compare the presidential-senate pairings.

Data was available for Maine’s elections at all levels, and I was interested in a comparison between Maine’s presidential and Senatorial races. I wanted to follow up on the reports that Collins benefited from split tickets that chose Joe Biden for President but then voted for her at the Congressional level. My initial research question was which towns and counties Maine voters were most aligned along party lines in their Presidential and Senate choices, and where they were most split.

I planned to do this using a scatterplot and a map, and I think these ended up being the right choices for this data. The scatterplot allows viewers to appreciate the strong trend in the data, and to see every town ordered by its Senate and Presidential choice, from most Democratic-leaning to most Republican-leaning. It’s clear that the dots are skewed upwards, towards the “Democratic President” part of the axis, and to the right, towards the “Republican Senator” part of the axis. Overall, the trend is pushed about 20 percentage points in favor of Biden and Collins — that is, in towns where Biden was beating Trump by 20%, Susan Collins was toeing the line for win or loss. If Susan Collins was beating Sara Gideon by 20%, Biden was toeing the win-loss line. (That, as opposed to Gideon always winning where Biden won, or Collins always losing where Trump lost, in which case the points would all be along the line Y=X.) There are towns in the quadrant that went Biden-Gideon, Biden-Collins, and Trump-Collins. There are no towns that went Trump-Gideon. The larger towns are in the blue, Democratic part of the plot, and smaller towns are mostly in the red, Republican part of the plot.

The map lends geographical context and completes the visual thesis that the scatterplot begins by showing that the pairings are clustered geographically by town size and by distance up, and from, the coast. Of the map, I had said: “This is harder for me to visualize in advance. I know that coastal Maine is where most of the state’s Democrats live and the large inland swaths are more Republican. However, given the background narrative of Collins as a bipartisan/ “not-like-other-Republicans” Republican, her hold on the coast might be a little tighter than one would expect.“

Indeed, the map shows Republicans’ strong hold in small inland towns– these went overwhelmingly for both Collins and Trump. And the larger coastal towns were almost entirely blue, voting for Gideon and Biden. What surprised me in its consistency, however, is that the “purple,” centrist part of the state is also geographically between the red and blue parts.

In order to answer my initial question, I’d have to get data from previous elections — I’m looking now and I see that it’s available in identical Excel format for 2014 and 2008, so perhaps I’ll add some context to the 2020 Election map just for fun. I’m particularly curious about 2008’s Presidential-Senate pairings now! It seems like an obvious Part II to this project, and would lend valuable context to understanding how Maine’s political momentum may have shifted over the course of Susan Collins’ career.

**There is no Tableau Public link; all project visualizations are in this post.**

This project is about quantifying both my election-related anxiety and some of my participation in electoral politics this year. I know that when I’m anxious about politics, I tend to give more money to candidates I want to have in office. I give in increments of about $5 to $25, so am definitely what the fundraising office kindly calls a “grassroots donor.” After noticing in the middle of September that I had donated about $150 worth of small donations in less than two weeks, I decided I should probably find another way of getting involved, or my wallet would be noticeably lighter by Election Day. I began researching get-out-the-vote postcard writing, which felt like a manageable place to jump in.

My original plan was to make my own postcards (see below for more on that), but I quickly found that there is a science to GOTV postcarding. The last thing I wanted to do was turn off a potential voter with my DIY GOTV efforts, so I decided to go with an established source for my postcards. Additionally, since I was starting pretty late in the game with only 6 weeks until Election Day and lots of early voting expected, I wanted to maximize the number of postcards I could send as soon as possible.

On September 23rd, I woke up, as I sometimes do, around 5am having a climate apocalypse nightmare. So fun! That prompted me to 1. donate money to a couple politicians who believe in science and 2. order 200 postcards and stamps from The Sunrise Movement, a youth-led organization that aims to tackle climate change and inequality as the massively overlapping issues they are, through direct action. A lot of the negative emotions I experience about climate change relate to being young and feeling like my future is being taken away from me; on the flip side, it feels very positive to me to participate in a movement led by other young people. In the end, Sunrise received many more orders than expected (woohoo!) and supply chain issues + me changing locations meant that I only received 100 postcards.

But I wrote them all! The rectangles on the chart below represent postcards I wrote, each red dot loosely measures the anxiety level I felt that day, and the little dollar symbols appear on each day I made a (small) contribution to a campaign. (FYI, the visualization is postcard-sized and definitely inspired by Dear Data.)

FRONT:

BACK:

A challenge of handmade data viz was not being able to edit once I committed to the pen lines — for one thing, I ended up without a good spot for a title, so I left it off to keep things uncluttered. Another piece of data I collected that I didn’t find a satisfying way to display is the time it took me per postcard. The fastest I could write and address a card was just under 3 minutes, but most cards took me on average about 5 minutes. This is in part because I knew I was more likely to finish all 100 cards if I made it more enjoyable by watching bad TV while doing it, hence the Gilmore Girls cards. (Side note: there is nothing like mindlessly absorbing hours of Gilmore Girls to make me appreciate just how awful and derivative TV was allowed to be when you had to wait a week between episodes and every creative decision was funneled through a major network. We are so lucky now *sob emoji*)

I got 100 postcards and stamps from Sunrise, and messed up on three of the postcards. I still had those three stamps, though, so I decided to send just three homemade postcards and make it to 100. Although I was hesitant to send a lot of homemade postcards, the design I made had actually been vetted by another and organizer from Postcards to Voters, which also does GOTV postcarding and has each volunteer provide their own postcards. I figured it was better to send 3 homemade, politically neutral cards (with a sneaky data viz!), than to send nothing at all.

I also only made 3 postcards because of my less-than-ideal printing conditions. The postcards are linoleum prints (see below for the print block). I didn’t have any printing ink with me and used acrylic paint instead, and there is already another carving on the back of this piece of linoleum. Unfortunately, both of those factors contribute to making the print quite muddy, but hopefully the three people who got these ones appreciate the DIY!

The prints were a lot of work — it took me about 5-6 hours to design and carve my linoleum block, and about 1.5 hours to print and paint 5 postcards. But, like the postcard-writing, those hours actually helped me to feel better and less anxious, so it didn’t feel like a burden to spend more time on them. (I decided to carve this linoleum block instead of watching the first presidential debate, for one thing.)

And for those keeping track (me), I put on Gossip Girl while carving the linoleum block. Common theme: the TV has to be bad enough that I’m not tempted to actually watch it while I work.

Next up? I definitely want to know how people are voting in Cary, NC. I want to know what youth voter turnout is like, particularly in swing states. I’d like to make more linoleum cuts of different designs and print them more cleanly. And next week, I’d really, really, really, like to be sending some postcards to government officials-elect who will soon be in office working to un-f*ck our future.

Please click through to view the Dashboard in Tableau, and select Fullscreen (in the bottom righthand corner) for the best viewing experience. The first draft can be viewed here.

My dashboard aims to answer the question, “where and when do New Yorkers interact most with the most fauna of the City: rodents, pigeons, and mosquitoes?” My goal was to remind viewers that New York is not only a city for its human residents, but an ecosystem for its animal residents. I imagine that most New Yorkers will be able to connect to this dashboard, since it’s hard to live here for any significant time without seeing a rat or a rug of pigeon droppings. The dashboard might also be helpful for city officials to keep track of trends—they might be able to answer, for example, why there have been 10 reported rodent bites in the first nine months of 2020, up from 5 reported last year and just 0 or 1 every other year since 2010.  

The largest component of my visualization is the small multiples map that shows each year of complaints side by side. My original goal there was to show change over time, but for the most part, the more striking thing is consistency over the last 10 years. Despite not displaying changing trends, as I had hoped, the maps are also helpful in that they subset the data into roughly 10% views of the whole set. Even so, there are still so many rat sightings each year that I had to turn the opacity of dots down to 25% to see any nuance beyond a big blob! On that note, another feature that helps with readability is the ability to click and highlight just one complaint description at a time—this will highlight across all visualizations, so it’s possible to see on the line chart how the number of complaints changed each month from 2010-2020, as well as to see the same complaint isolated on the map and in the bar chart. This gives an overall answer to how a complaint changes seasonally, which borough sees the most of that complaint, and where on the map it might be most or least common. 

I found my 311 data subset difficult to understand even when looking through the columns in Tableau, and I definitely began to appreciate the power of Tableau as an exploratory tool. Given my lack of prior experience with it, though, my visualizations changed several times as I discovered more effective ways of showing off the data. For example, all of my visualizations started by coloring according to “complaint type,” which corresponded to the three categories of 311 complaint I had filtered from the site: Rodent, Mosquitoes, and Unsanitary Pigeon Conditions. It took me until the last couple hours I was working on the project to realize that the “Descriptor” column contained more nuance— just enough that it made the charts much more interesting without losing cohesion as a graphic. By switching my x-variable and color pill to Descriptor, I was able to show the distinctions between different rodent-related complaints, which brought to the surface, for instance, the observation that rat sightings are very seasonal, with much higher numbers in the summer than in the winter, while mouse sightings are lower overall and stay constant throughout different months. 

The biggest design challenges I had were first, getting the small multiples map to correctly display the grid of different years (this was a querying problem on my part, since I had first written a valid query that didn’t do what I wanted, and couldn’t figure out where my error was), and second, killing my darlings to get everything I wanted to fit on my dashboard. In the end, I think I’ve chosen three graphs that inform and enhance each other, rather than 3 that tell disparate stories, and it feels cohesive and exploratory to me. I’ll be very curious to see how others respond in today’s critique, though!

My scope ended up only including animal data, although my original proposal had planned to ask, “How do areas with more animal complaints overlap (or not) with city services like street cleaning, or demographics like median income, population density, or building zoning?” I still think it’s a good question, and one I’d like to explore more. One lingering concern I have is about 311 reporting in general. Who uses 311? Who does not? I see large swathes of Queens that report many fewer mouse sightings than the Upper East or West Sides, for example. Is that because there are fewer mice in Queens or because there are fewer 311 users in Queens? Perhaps mice really do prefer Manhattan—in that case, I’d like to know why! (I might choose it too, if I could live rent free in someone else’s apartment.)

EDIT: After feedback from the critique, I acted upon the notion that the small multiples format showing different years didn’t really show differences over time. Instead, I kept the small multiples format for comparison, but made each map a different complaint description. I also made the timeline chart larger, since this chart seemed to garner more attention than its place in the corner suggested it should.