Modern Polling is incapable of successfully reaching individuals who are without landlines or those who do not opt-in to online surveys. A whole 59% of low-income individuals do not own a landline, and those who do are overly weighted when contacted.
What it does
This application allows campaigns and policy makers to text poll questions to individuals (86% of low-income households own a mobile phone) and applies post-stratification techniques to both the voter file and to census data for policy or campaign prescriptions.
How I built it
I worked on the post-stratification of the data, which was created in R. I built it using link 's work, and the wru package.
Challenges I ran into
Individuals can be more liable to lie over text than to someone on the phone.
Accomplishments that I'm proud of
To post-stratify, I needed to use an algorithm that would predict race by surname, party affiliation, and location.
What's next for FixPolling
The next step is to get a campaign to test it alongside other forms of polling to prove it is cheaper and easier.