Mapping Bay Area Ballot Drop-Off Boxes

We recognized the need for accurate and accessible information on safe early voting options in the Bay Area. In response to this need, unBox’s BayAreaCommunity.org (BAC) team mapped every official ballot drop-off location in the 9 counties (Alameda, Contra Costa, Marin, Napa, San Francisco, San Mateo, Santa Clara, Solano, and Sonoma)*. Like every feature in our website, building this map took a village. It involved tackling unique data and design puzzles by pooling our diverse skill sets, backgrounds, and lived experiences. 

To build the map, we 1) identified a need for information, 2) added the best available information to our database and kept it up to date through both web scraping and manual data entry, 3) translated key information, 4) ported our data through to our Mapbox-sponsored user interface, and 5) got the word out! Here’s how we did it. 

1. We identified a need for information. 

We recognized a need for accurate, up-to-date information on safe, socially distant, early voting opportunities. Each of the 9 Bay Area county websites conveys this information, but some of their interfaces are not mobile-accessible, or display ballot drop-off box locations as a list view instead of as a map. Given that 26% of Americans earning less than $30,000 per year do not have broadband at home but rely on their smartphone for Internet access, it is essential that online information be accessible on mobile devices (Pew Research, 2019). We added ballot drop-off locations to our mobile-friendly interface to help low-income residents access this information on a map, from their phone. 

Building our map for mobile phones, on a customizable platform easily adaptable to user feedback, would not have been possible without MapBox. The Mapbox Community Team sponsored our project at its earliest stages, providing a generous coupon on all Mapbox features. They were also readily supportive for technical help, feedback on accessibility, and publicity ideas. Mapbox support staff are active members of our unBox Slack community, popping in with great event updates and encouragement.

Each of the 9 Bay Area county websites conveys this information, but some of their interfaces are not mobile-accessible, or display ballot drop-off box locations as a list view instead of as a map.

Furthermore, surveys projected that this election would be a turning-point for low-income voter engagement. A nationwide poll by Hunger Free America found that 80% of respondents earning less than $50,000 per year were “certain” or “very likely” to vote in this election. In 2012 and 2016, less than 60% of low-income Americans voted. The survey found that one of the most effective messages to inspire low-income voting was: “It’s now quicker, easier and safer to vote by mail or vote early in most states.” By mapping those quick, easy, and safe early voting options, we sought to help mobilize voter engagement in the Bay Area. 

2. We built our database through both savvy code and human data entry. 

The process of gleaning data from trusted sources, adding it to our database, and keeping it updated involves both data engineering and brute force manual data entry. COVID-19 era social service data changes rapidly (ballots weren’t too bad, but don’t even get me started on food pantries), and is formatted in myriad inconsistent data structures across information providers.** For these reasons, it is often necessary to check the quality of scraped sources with human eyes, or enter the data entirely manually.

Moving ballot drop-off box location data from county websites to our database involved both automated and manual processes. Alameda, Contra Costa, San Mateo, and Santa Clara counties provided easily scrapable ArcGIS maps or lists. Napa, Marin, Solano, Sonoma, and San Francisco were either difficult to scrape, or contained few enough sites to be simpler manually. 

Anjali Katta stepped up to the data engineering plate. Anjali graduated Stanford in 2019 with a BS in Engineering Physics and a minor in Human Rights. At Stanford, Anjali honed her skills in social impact data science through the Data Challenge Lab (DCL). After COVID-19 imposed an unexpected end to her post-graduation 7 month trip in Southeast Asia, Anjali returned to her Vancouver home, began a fellowship with Humanity in Action, and signed on to volunteer for BAC as data engineer. Anjali scraped the four county ballot drop-off web pages, transformed the data to fit BayAreaCommunity.org’s Human Services Data Specification data structure, and uploaded the data to our Airtable. As with many social service web scraping tasks, the path wasn’t a straight one. To parse maps that used JSON files, Anjali had to puzzle out their different data structures, extract information by iterating through all characters, and use many regular expressions. 

Angelina Polselli put the other five counties on the map. Politics major at University of San Francisco, staff for the SF Board of Supervisors, and former Volunteer Director for Laura Oatman’s 2018 District 48 Congressional campaign, Angelina joined unBox as one of its first members. As Policy Team Lead, she coordinates unBox’s policy research and advocacy work at the state and national level. She is firmly rooted, however, in the needs of her favorite corner of Northern California, and was eager to help mobilize local voter engagement through BAC.

Once the data was in our Airtable database, fellow DCL alum Christopher LeBoa helped ensure it was high quality, and up-to-date. By layering Anjali’s multiple webscraping pulls over time, and identifying mismatches, he ensured our data remained current as ballot drop-off box locations shifted slightly in the weeks preceding November 3. 

3. We translated key information into five languages. 

Our website supports the Bay Area’s 4 most common non-English languages — Spanish, Mandarin, Tagalog, and Vietnamese (SF Gate, 2012). People often ask why we also support Malay. Our founder and dev team lead, Joyce Tagal, proudly reps her home country Malaysia by translating our resource into her native language! 

We have seen firsthand how language can make or break one’s ability to access key information about essential resources. In June and July, we conducted 7 compensated user testing interviews amongst native Spanish-speaking low-income Bay Area residents. We are now completing our second round of interviews amongst both low-income Bay Area residents seeking resources, as well as social workers who refer clients, patients, and families to resources. We are scheduled to complete 14 interviews by the end of November (7 low-income Bay Area residents, and 7 social workers [from three different local organizations]). Of the 14 completed and scheduled interviews amongst low-income residents, 13 requested that their interview be conducted in Spanish. Hearing interviewees’ personal stories about resource information and access challenges has profoundly reaffirmed the need to make our website available in users’ preferred languages.

Native Spanish speaker Micah Trautwein translates all of our website headers and resource text descriptions into Spanish, and coordinates our translation team. She and Karen Amaya Aguirre conducted our Spanish-language interviews with note-taking and coordination support from Isabelle Foster, Sarah Kratzer, and Hannah Hudson. For our ballot drop-off map specifically, Celine Gandingco and Benjamin Topacio translated the website headers into Tagalog, Heather Nguyen into Vietnamese, Songnan Wang into Mandarin, and Joyce, of course, into Malay. 

4. Uploading our data to the website

We design every aspect of our site for maximum accessibility and usefulness. Pooling insights from user testing, conversations with partners, and the lived experiences of our own members, we puzzle out: which information is essential, which is unnecessary, and which may be non-essential but helpful to the extra curious user. These decisions determine how we craft checkbox filters, badges, and listview templates for each resource. After convening on these decisions for the ballot drop-off data, Joyce added the additional features to our user interface (shout-out to her great choice of ballot drop-off icons!), and our backend developer Lawrence Wu transformed our Airtable database variables to populate our CARTO DB backend. 

The push of a button. A celebratory Slack message. Our team reloaded our pages and over 300 ballot drop-off sites popped up on our screens. IT’S ALIVE!!!

5. We got the word out. 

We accelerated our social media launch date in order to showcase ballot drop-off locations. (Follow us on Twitter and Instagram!) Right before we posted, several articles appeared regarding unofficial ballot drop-off locations popping up across California. We ensured our social media emphasized “official ballot drop-off locations” and of course, our UI explicitly linked to the county website sources. 
As we publicized our resource to local partners and news media contacts, Derek Ouyang, Program Manager of the Stanford Future Bay Initiative (SFBI) and one of our unBox advisors, reached out. Interested in the potential of our map to highlight resource access barriers in the Bay Area, Derek assigned his SFBI students an accessibility mapping project using BAC. 

Jayne Stephenson, Stanford Earth Systems major, former intern for the California Energy Commission and Environmental Defense Fund, and co-Director of Students for a Sustainable Stanford, took on the assignment and mapped the accessibility of ballot drop-off locations: https://jaynestevenson.github.io/Assignment-5. She found that 84-86% of the Bay Area workforce lives within a 10 minute drive of a ballot drop-off box location, and access doesn’t vary greatly by income. Jayne chose to model driving accessibility instead of walking accessibility, given the assumption that more voters in the region would be driving to drop-off sites than walking, but acknowledges that either choice introduces subjectivity and limitations. 

I found Jayne’s insights surprising and affirming. I had assumed that access would be lower, simply because some areas like Sonoma appear visibly sparse. When analyzed by transit routes however, it is clear that Bay Area ballot drop off locations were chosen strategically to meet most voters where they are at. I had also assumed that access would vary by income level. We have observed this destructive correlation when analyzing Bay Area food retailer access (an analysis created by Samantha Liu, Teaching Assistant for SFBI and unBox Data Team Lead). In all our work at unBox, maps continue to challenge our assumptions. Sometimes zooming in, zooming out, or adding a new layer, changes your whole perspective on an issue. 

In the weeks leading up to election day, our team members mailed in ballots from across the globe, dropped them off at ballot box locations, or waited to vote in person. After spending an evening, the week before election day, sitting at the family dinner table and discussing local candidates and state propositions, Christopher LeBoa’s family used the BAC website to plan the following morning’s triumphant march to the nearest ballot drop-off location in San Leandro. 

Putting ballot drop-off box locations on a map won’t transform voter turnout statistics nor substantially bolster the pillars of democracy. If our map helped some Bay Area residents vote a little bit earlier or a little bit easier, we were successful. Even if some didn’t use our map to find a ballot box, but just seeing them on the website nudged them to vote, that’s a win. Ultimately, we built this map to further advocate for accessibility: essential information should be available to everyone, in a format that meets them where they are at.

*Our current website BayAreaCommunity.org does not show ballot drop-off boxes. We removed them after polls closed, because maintaining the recency of our data and interface is one of our top priorities. To browse a previous version of the site displaying ballot drop-off boxes, check out this link!

**We hope to contribute to an open source information ecosystem, in which service providers can share and receive data as easily as possible. We therefore adhere to Open Referral’s Human Services Data Specification, in an attempt to make our data as transferable to other HSDS compliant organizations as possible. We are also eager to transform our variables as necessary to suit partners with different structures.