Bluegrass, Blight, and the Future of Cities
How a fiddler and an astrophysicist introduced predictive analytics to Cincinnati
Ed Cunningham is the front man and fiddle player for the Comet Bluegrass All-Stars, a band that has been playing every Sunday since 1996 (“except Easter,” he says) at the Comet, a bar in the Northside neighborhood of Cincinnati, Ohio. Thus the name. The All-Stars can be heard, along with Roseanne Cash, on a new album from the Cincinnati Pops Orchestra called “American Originals.”
He is also the head of the building code enforcement shop for the City of Cincinnati. That is not necessarily a likely launch pad for a thrilling data initiative that portends a tech renaissance that will shape the future of cities. But I’m here to tell you about Cunningham’s role in a project that does indeed provide that promise. With some help from an amazing program from the University of Chicago.
First some background. In April of this year, I wrote about how Cincinnati was leading the charge towards the future by integrating data into its daily operations. At that point, City Manager Harry Black and Chief Performance Officer Chad Kenney were doing something highly unusual in municipal governance — acting as if government should serve its citizens with the same effectiveness and accountability as successful companies serve their customers. In a broader sense, the two were determined to create a data-driven approach to civic operations aimed at addressing some of Cincinnati’s most pressing problems.
I’m happy to say there’s now more to the story — Cincinnati is on the brink of adding predictive analytics to its arsenal of digital tools. And, most importantly, its efforts are on the forefront of a larger movement that may help make life better for many other cities. Even if you aren’t immediately eager to read another column about Cincinnati, keep going. Like other good stories, this one has drama, memorable characters, sudden bursts of insight, and a cliffhanger ending that hints at future episodes. It also has a soundtrack.
The soundtrack, of course, is supplied by the All-Stars and its bureaucrat fiddler. Cunningham started working for Cincinnati 28 years ago, when the housing inspections department still used carbon paper and 3x5 cards and couldn’t effectively produce a report or opine on the state of the city. He’s always thought that using computers would be a good idea for management of his division, which protects public health and welfare through code enforcement. So he was more than intrigued when one day he heard Chad Kenney talk about predictive analytics and an innovative summer program at the University of Chicago called Data Science for Public Good (DSSG).
The rigorous DSSG program looks for city data questions or problems that are carefully scoped, that have social good aspects, and for which solutions can be implemented. In just its second year, it’s seeing great success: Hundreds of Ph.D.’s apply to be fellows during the summer in order to work on city issues. All of DSSG’s code is online, together with detailed descriptions of many of the projects; research papers based on these projects have been published and presented at major data mining conferences.
Kenney was talking about the program in hopes of figuring out a way he could draw on the expertise of those extraordinary post-docs for a pilot project that would support the case for bringing sophisticated data analytics personnel inside city government. “If corporations are going to have the analytical capacity to evaluate options and evaluate operations — and that’s what they need and use to survive in the 21st century — cities need the same thing if we’re going to be able to evolve,” says Kenney. But in order to justify hiring these people, Kenney had to demonstrate that having them around would objectively add value to city operations.
Luckily for Kenney and for Cincinnati, Ed Cunningham’s experience in bluegrass improvisation triggered an Aha! moment for the long-time civil servant. “You’re looking at stuff a little bit differently,” he points out about his fiddle playing. “You’re making a song your own.” And, he adds, “Like they say in Nashville, and I think the same applies to government, ‘A little creativity goes a long way.’”
So Cunningham saw things differently and came up with a great problem for DSSG fellows: Predictive Blight Prevention.
Following the nation’s foreclosure crisis, neighborhoods with abandoned, vacant, and dilapidated buildings have suffered. Cincinnati, which has lost 40 percent of its population since 1950, has several such neighborhoods. Vandals arrive to strip buildings of their copper wiring and break windows, the foreclosing banks don’t pay much attention to maintenance, and property values for the neighboring homes that aren’t vacant or abandoned decline mightily.
Because property taxes fund Ohio schools, a large decline in property values leads to a large decline in school funding — as well as the loss of life savings tied up in houses, which triggers other problems for the people who live there.
Cunningham asked himself: What if the City of Cincinnati could get ahead of these problems and stave them off by maintaining properties that were likely to become vacant and abandoned? “If you just maintain something — and that’s our job, property maintenance — it will last a very long time,” he says. What if the city could be proactive and intervene, rather than waiting for complaints to come in?
Cunningham was frustrated with his department’s reactive approach. Typically, for homes that will one day be vacated, the Department of Buildings and Inspections receives complaints only about a quarter of the time. But his housing inspectors could be doing so much more to avoid vacated houses.
After all, housing inspectors don’t just issue fines for noncompliance with city housing codes. They offer assistance for people who don’t have the wherewithal to make repairs, partner with community development people and for-profit developers, and, Cunningham says, “try to get everybody on the same page and moving in the same direction to try to save an area or make it more viable.” He adds that “Some days you feel like a social worker where you’re trying to get people the help they need.”
He figured that if Cincinnati could predict which areas were likely to become blighted, intervene there, and demonstrate to investors that property values were likely to stay steady or improve rather than decline, the tax base could be saved.
Cunningham spent a month honing Cincinnati’s proposal to be one of DSSG’s projects. He pointed out that Cincinnati had 14 years of data about the individual parcels of land in the city — including inspections compliance reporting, tax data, water shut-off records, information about the nature of the parcel’s ownership, and many other forms of data — but no way to glean insights from the data that might help the city focus and prioritize its intervention efforts by identifying troubled homes before they fell into serious disrepair and were vacated. DSSG, he was confident, had the technical chops to help the city use what it already knew.
Kenney says that hard pre-work made all the difference: “If it were not for Ed Cunningham,” he says, “this project would not have gotten off the ground. He sent this proposal that was just so well put together.”
The proposal drew, as hoped for, the participation of the DSSG, who this past summer assigned three fellows to work on the Cincinnati blight prevention problem. They were Katharina Rasch, a Ph.D. in computer science; Talia Kaufmann, a social scientist/city planner from Tel Aviv; and Jen Helsby, who had to delay her start by a couple of weeks because she was busy defending her dissertation for her Ph.D. in astrophysics from the University of Chicago.
Helsby had studied the large-scale distribution of galaxies in the universe — a subject she notes is “very disconnected from cities.” Nonetheless, such celestial calculations called on the same data analysis techniques, machine learning, and predictive modeling that Cunningham and Cincinnati needed. And Helsby was delighted to help: She wanted to work on things that had “more direct impact on people’s lives,” she says. “Not that astrophysics isn’t important,” she quickly adds. “There are a lot of really great people already doing basic research in astrophysics, and I thought that my skills could be better used in another arena.” She had taken a hard look at the DSSG program and its prior projects and found it to be a “very rigorous and successful program,” so she went through DSSG’s highly-selective application process — just 120 applicants were interviewed, and only 42 were chosen — and reported for work four days after her dissertation defense took place. One of the first things she did was to Google “blight” so she could understand the problem she was trying to address.
The three were paid a modest stipend to work as teams in a large co-working space in downtown Chicago on city projects. The collaborative, start-up hum of the workspace “took some getting used to,” Helsby says. Over the next few months, Cunningham and others in Cincinnati talked to the fellows every week or so, explaining what the fields in the datasets that had been sent up to Chicago were. The fellows cleaned and extracted the data, moved it into a centralized database, wrote a data dictionary clarifying what the data was, and set to work looking for proxies for “blight” that could be predicted based on correlations.
Most importantly, they went to Cincinnati in mid-July for two days to do a deep dive into the housing inspections world. Their first stop, of course, was to the Comet to see Ed’s band play. (Cue soundtrack.) Then they went to a host of meetings, including a ride-along with actual housing inspectors.
Helsby found the ride-along “very exciting, a great experience.” As a computer scientist, she had thought that a housing inspector would be “more of an enforcement person.” She continues: “But that’s not really what they seem to be doing. They would go to a home, and they would have a friendly conversation with the homeowner, and they seemed to be more like educators, because the homeowners often didn’t know what the building codes were, and the building code inspectors were working with these homeowners to teach them.” She watched the inspectors log in detail into a tablet PC the data they gathered as they were working, and she asked questions.
And it was a good thing she did: There were some aspects of the data the fellows had misunderstood, and the hours of interactions with the inspectors changed significantly what the fellows ended up doing. “They were very helpful and had a lot of suggestions for features for our predictive model — things they had noticed that might be predictive of blight,” Helsby says.
Cunningham says this trip was essential: “It made the data more meaningful” for the fellows. Kenney agrees: “The inspectors know on the first citation if a property is going to be blighted in two years from now or not. They have knowledge that maybe doesn’t come through when you’re just looking at tables and tables of data.”
The fellows then went back to Chicago, and Helsby personally wrote many lines of code for the predictive model the fellows built based on three years of historical data. The proxy they chose for “blight” was, essentially, “likelihood of serious housing code violations.”
Then they tried to get a sense of how one might use data to predict the areas where intervention might best mitigate that blight. They used about 50 variables (led by home value, crime rate, and other factors) gleaned from those three years of data to synthetically target home inspections that would have taken place in the fourth year, and then looked at actual serious housing code violations in the fourth year of data. That gave them a 78 percent hit rate — serious housing code violations would have been found in 78 percent of inspected houses. But if the inspectors had proceeded in that fourth year based on their ordinary practices — responding to calls from citizens — just 53 percent of inspections would have revealed serious housing code violations. In government-speak, this is a huge improvement in effectiveness.
Here’s the cliffhanger: this improvement in inspection accuracy was based on historical data alone. It’s just a static tool; it can’t be “retrained” based on additional data. The city still needs to do rigorous A-B field testing to validate the model the DSSG fellows came up with. (“If inspectors, not knowing whether they’re using a predicted list or an ordinary list, inspect properties, what’s their hit rate for serious housing code violations?”) Chad Kenney and the City of Cincinnati need resources with which to carry out that testing, and to go on to test which interventions — repair? assistance? fines? creation of additional green spaces? — actually result in improved neighborhoods. “We did a first pass,” says Jen Helsby.
Nonetheless, Kenney says the summer was a win. It showed the potential power of data projects to the city; in a sense, it “started creating a market for this type of analysis within Cincinnati city government.” Also, he says, the fellowship program “gave us a cheap way of getting a really good understanding of how how to manage these projects so that we can be successful as we scale up.”
And using fellows for a summer provided an agile process aimed at building the case for increased analytic talent in the ranks of city employees: the summer engagement was lightweight, easy to understand, and quick. The analytic talent Cincinnati hires has to come with the ability to talk to other people clearly — without human skills, the fellows could not have asked the right questions of the inspectors that led to vast improvement in the model.
Next steps: Cincinnati — like many other cities — needs increased funding for more data analytics people. Community foundations, with their deep ties to community and government, could usefully fill this gap. Ed Cunningham, who worked to define a specific and simple data problem that would yield insights that could in turn be fed back into his operation, deserves praise for being willing to change his department’s stripes and implement a new way of doing business.
DSSG, with its high standards and well-scoped issues, should be imitated at every major university. Helsby, the astrophysics Ph.D. turned warrior against urban blight, makes the case: “I think that there are very few opportunities available for these technical skills and working on projects that really improve people’s lives. There are a lot of programs out there for data science training, generally, but there’s not a lot working directly with government agencies and non-profits. Directly from a Ph.D., I was able to go and work on a project with a city, and that’s a very unique opportunity, which I think there should be a lot more of. Clearly, there’s a lot of demand for this training and the sort of opportunities that one gets working with the great project partners that DSSG has.”
And you should really listen to the Comet Bluegrass All-Stars.