Rowdy neighbors, honking traffic, or loud leafblowers are accepted as part of modern life. The nuissance seems random, inconsequential, and perhaps even silly. But noise pollution is, quite literally, a textbook example of an externality. You’ll find it somewhere on the “The more you know” sidebar in the market failure chapter. But! Apparently, it’s a serious problem that’s only getting worse.
AC/DC hit the subject with their 1981 ballad “Rock and Roll Ain’t Noise Pollution.” Inspiration for the song came from Londoners unhappy with a rowdy club in a residential neighborhood.
“Nah, that was one that Angus came up with,” he says with a dismissive wave of his hand. “We were in London at the time and there were all those problems with the old Marquee Club because it was in a built-up area and there was this whole thing about noise pollution in the news, the environmental health thing that you couldn’t have your stereo up loud after 11 at night, it all came from that.”
Begs the question: what role does rock music play in noise pollution? How often do people really complain about loud things? To answer those curiosities I turned to Kansas City and their 311 data.
Loud music is certainly noise pollution and rock seems to be the top hated genre. It beats country, rap, and pop as the most identified genre when people issue complaints for loud music. But other sources of noise pollution are also a problem. Traffic, construction, trash collection, bar and club loiterers all registered as other sources of frustration. This jives with broader, national studies on the subject.
The data for this study came from KC - a town that has birthed a range of acts from Puddle of Mudd to Tech N9ne. Complaint records spanned three years and all days/hours of the week. Rock registered 3x the frequency of other genres, so next time you’re in KCMO don’t play AC/DC.
311 data is an easy entry point for cities to join the “open data” movement. Calls are subject to most state FOIA laws and almost every municipality has paid for technology that assists call center interactions and records the data in relational format. That’s when Socrata steps in. They help move your data from the call center application into their platform for the public to access. It’s a nice service to share things quickly, and Socrata has done a good job creating open APIs for access.
Cities don’t share call center notes in their file transfers. There’s too much text, and it’s relatively unstructured, so it’s not really worth posting. But that’s where all the good data is - that’s where people cite specific examples and stories that allow folks like me to do analysis deeper than noise complaints vs. unkept yards.
Kansas City was the only local government that shared case-specific URLs in their Socrata dataset. Each URL pointed towards a call-specific webpage in their case management system. On that page included the call center agent’s notes which were located in the the html tag “rc_descrlong.”
The workflow was simple:
Ping the KC 311 Socrata API for records where “Noise Control” was flagged using the fromJSON argument in R to create the data frame
Create a new variable in that dataframe that uses the R library rvest which scrapes according to HTML tag
Build dictionaries for each classifier using terms and some regular expressions
Use those dictionaries as new variables to do the counting, and then sum the variable columns
Last step, I used a Dimple.JS chart and their API for a quick visualization in D3
All the code is below. A few notes:
The scraping took about 20 minutes for around ~1,200 scrapes. The pull was fortunately not rate limited by KC 311’s CRM system. It took awhile because that’s a lot of webpages to open, read, pull, and store.
R’s rvest library is stupid simple to use and deploy for webscraping
D3 is pretty great to use if you find something that’s prebuilt, it just really limits your ability to customize.
Dimple.js D3 script
(kcnoise.tsv is a three column dataset based off summaryframe from the R code)