Accessing Demographic Clusters with CartoDB's Segmentation Layers
The U.S. Census is an amazing resource of data and information. The U.S. Census performs a number of regular as well as ongoing surveys that document many facets of people and life in the U.S. These data can often be used to help learn about dimensions of a location and what it might contain.
As humans though when asked about what a neighborhood is like we don't rhyme off a series of census variables for that neighborhood. Instead whether it be the hipsters of Williamsburg or the stroller traffic jams of Noe Valley we tend to describe neighborhoods in terms of archetypes that we can more easily relate to. These kinds of neighborhood descriptions can add meaningful value and context to your location data.
We want to make this kind of contextual data available and easier to use in CartoDB.
Releasing segmentation layers
Today we are releasing inside the Data Observatory a new set of layers created from demographic segmentation. Demographic segments provide a kind of grouping of people that we then apply a data-driven naming method that makes them easily readable and recognizable when analyzing your data. By releasing them in the Data Observatory we are making them available for users to use for augmenting their own data quickly.
The segmentation is generated through a clustering procedure that we'll cover in more depth in a forthcoming blog post. The output gives us two granularities of clustering one that produces 10 unique groupings of people across the USA and a second that creates 55 unique groupings of people.
10 cluster resolution
To generate these clusters we used the algorithum proposed by Spielman and Singleton and for the x10 clusters we were able to adopt their naming structure.
The names of the 10 different neighborhood types are:
55 cluster resolution
The 55 cluster layer was more difficult as the names of each group had not been previously published. For these more detailed categories we generated names based on the dominant traits of the populations within that cluster (or the dominant omission in a few cases). For example if an area within a city population is found to be highly dominated by college age adults with some college education it was given the name "City center university campuses".
Take a look at all 55 proposed group names:
On the map
You can explore both on this map and the deep insights dashboard here or take a look at the simple map version here:
Accessing demographic segments
Using the awesome power of the Data Observatory to bring these segments into your data is as easy as calling a quick SQL statement.
To query these segments at a single point location simply use the function
10 clusters
55 clusters
Augmenting your data
Another interesting use of the segmentation data is to augment your tables. You can do so by adding a new column to any table called segment (or any other unique name).
Next augment your table with the segment description:
To create similar or even better visualizations you can watch our Data Observatory webinar as many times as you need to!
Next steps
Today we wanted to announce the availability of this exciting set of layers in the Data Observatory. In future blog posts we will explore some of these groupings what they can tell us about the U.S. and how they can add context and insight into your data. We will also detail how these segments
were created and how we plan to improve and expand on them in the future.
For further reading checkout the data services-api docs and the Data Observatory.
For now happy demographic segment mapping!