Where Do Adoptable Dogs in Your State Come From?

Category: Best data-driven reporting (small and large newsrooms)

Country/area: United States

Organisation: The Pudding

Organisation size: Small

Publication date: 15/10/2019

Credit: Amber Thomas, Sacha Maxim

Project description:

If you’re looking to add a new furry friend to your family, you may be encouraged to “adopt not shop”. That is, to find a new dog at a local shelter or rescue organization rather than a pet store or breeder. But where do adoptable dogs come from? We looked at the PetFinder profiles of all 58,000 dogs available for adoption across the US on a single day and found 2,460 dogs whose travel was described in enough detail to follow. This piece explores the trends and patterns.

Impact reached:

This project revealed trends that we didn’t know existed – that adoptbale dogs often move from the south to the north. The most rewarding projects often come when you not only share something with the audience, but also when you learn something new too. This project also expanded our thinking about what you can get from traditional API calls.

Techniques/technologies used:

Using the PetFinder API, details about all 58,180 dogs available for adoption in the 50 US states and Washington DC on September 20, 2019 were collected. Since PetFinder does not provide an entry field for an animal’s location before arriving at its current organization, we parsed the text of each pet’s “description”. We started by limiting text to anything that came after the word “from” but before the word “to”, or after “located in”. We then analyzed the remaining text using entity recognition from the spacyR package. We manually checked the data for anything mislabeled. The front-end experience was built using HTML/CSS, Javascript, and D3.js.

What was the hardest part of this project?

The data collection was the toughest part of this project. The APIs weren’t really set up to do exactly what we wanted to do, so we needed to figure out some workarounds. Then it was a process of programmatically filtering through the text (described more in the answer about tools and technologies) and manually filling in where the computer couldn’t. Data work often needs that tag-team approach of computer + human, but that means the process can take longer. 

What can others learn from this project?

Sometimes a simple question (Where Do Adoptable Dogs in Your State Come From?) can yield some surprising answers. We didn’t expect to find such a stark geographic pattern when we first started this analysis, but you can clearly see that northern states import more does and southern states export more dogs.

Project links: