NYU Researchers Devise Method To Identify ‘311’ Underreporting Of Heat And Hot Water…
Complaint lines such as New York City’s 311 let people report quality-of-life problems in their building or neighborhood, from excessive noise to illegal parking. But resident-generated data typically suffer from reporting bias, with some neighborhoods and addresses calling attention to problems at lower rates than others.
A team of New York University researchers has developed an automated modeling tool to help the New York City government estimate 311 under-reporting by building, neighborhood, and subpopulation. In a new study, published today [May 30] in Annals of Applied Statistics, the researchers describe a method that, using machine learning, can estimate the potential under-reporting of heat and hot water problems. If adopted, this tool would help the city’s Department of Housing Preservation and Development (HPD) identify which buildings or locales may be placing a lower-than-expected number of 311 calls. The agency could take steps to better ensure that heat and hot water issues are not going unaddressed.
The city’s 311 system generates more than 8 million calls of all kinds annually, of which more than a third are related to concerns about heating and hot water supply. Because most of New York City’s building inspections are conducted in response to 311 resident complaints, “a problem that is not reported to 311 is both less likely to be observed in the data, and less likely to be addressed in a timely manner by the city,” according to the study.
Titled “Estimating Reporting Bias in 311 Complaint Data,” the study was co-authored by Kate S. Boxer and Daniel B. Neill of the Machine Learning for Good Laboratory at New York University’s Center for Urban Science and Progress, and Boyeong Hong and Constantine E. Kontokosta of NYU’s Marron Institute of Urban Management.
The authors addressed the challenge of identifying 311 under-reporting via two methodological approaches. The first looked for apartment buildings where no heat and hot water problems were reported during the heating season, but which had building and population characteristics similar to residential buildings throughout the city where 311-reported heat and hot water problems were frequently called in by the tenants.
The second approach involved examining buildings with fewer 311 calls than expected as compared to similarly sized buildings with similar estimated problem durations.
In comparing buildings, some of the characteristics examined included its age, whether it was a rental or coop, and the number of units. In addition, neighborhood-level demographic and socioeconomic factors touched on the number of limited-English speakers, proportion of elderly individuals or households with children under 18, median rent, and the proportion of population that votes in elections, among other factors.
The authors of the study noted that obtaining “ground truth” is neither easy nor inexpensive. “Nevertheless,” they wrote, “we hope that these methods and results will assist city agencies and advocacy groups improving access to 311 and increasing services to communities whose quality of life is impacted by heating and hot water problems.”
https://doi.org/10.1214/24-AOAS2003