Number of service requests to NYC’s 311 from 2004 to 2016: 24.54 million
Min. 1st Qu. Median Mean 3rd Qu. Max.
0.00 0.16 3.00 12.87 10.00 365.00
Without the 365 days filter for Days.To.Close a ticket, the median was 3 and the mean 19.5 days. With the filter, the mean dropped by 34%, with no change in the median. I dropped the max filter to 180 days, and the mean moved down to 10.46, that is, by about 19%, and the median to 2.92 days, Not being a wonky Mayor (can we say Bloomberg?) determined to put the best spin on things, it was not clear to me that 180 days is a better cut off than 365. To keep data exclusion to a minimum, I’m sticking with 365 days as the cut off. Is there a particular department that is causing this mayhem? Unlike the DOT which is the main culprit carelessly closing out tickets before they were created, four agencies account for 75% of tickets closed out after more than a year. See further below.
A median of 3 days to resolve a service request is not bad for a city of 8 million? Maybe. Based on my anecdotal observation of NYPD investigating non-emergency things like “loud talking” or ice-cream truck jingles (yes, ice-cream trucks are actually a category at NYC 311!), I’d say a lot of those tickets get closed out without a “site visit” – they just let them ‘age’ for a couple of hours and close them out with “conditions not observed” or, less often, something to the effect ‘we took care of it’. In any case, NYPD has a mere 8 hours to attend to your non-emergency complaints so my hypothesis is they skew the median downward quite a bit. An exercise for another day.
The top 5 agencies closing out service requests to 311 after they languished for a year or more:
1: DPR 92384 (Parks & Recreation, the real thing, not the TV show)
2: DOB 34508 (Dept of Buildings)
3: HPD 30331 (Housing Preservation & Development)
4: DCA 14275 (Consumer Affairs)
5: DOT 9739 (Transportation)
311 data from NYC Open Data downloaded at various times. As of Feb 19, 2017, 24.5 million rows, 52 variables till 2009, 53 variables from 2010. Parsed using R v3.3.1 in RStudio v1.0.136