Fresh from a job in management consulting and eager to unlock whatever efficiencies could be found, I had promised during the campaign to set up a 311 system, so residents wouldn’t have to figure out the relevant department and its own contact information in order to report a pothole or get a streetlight fixed. When the 311 center opened, a year after I took office, we gained something even more valuable than a new mechanism for customer service; for the first time, South Bend had a central, constantly updated data set on what people were calling about. Using the data, the city was able to make countless operational improvements, from cutting the time it took to get a large item picked up by our trash crews, to simplifying the way residents paid their water bills.
ARRIVING IN OFFICE, ESPECIALLY with my consulting background, I took it as a given that more data was a good thing—the more objective and analytically driven our work, the better. There was an emerging bipartisan consensus about this style of government, and I bought in. Just as Martin O’Malley had gained a reputation for excellent work modernizing Baltimore’s government with improvements on everything from overtime costs to pothole patching, Republican Mayor Steve Goldsmith of Indianapolis racked up a number of wins from increased child support collection rates to the reduction of sixty-eight thousand pieces of unnecessary paperwork per year.
But this style of government also had its detractors, as I saw when Councilman John Voorde stopped by my office one day to discuss an upcoming budget vote. “Have you seen that documentary on Vietnam?” he began, in what I assumed was just small talk as he settled into a seat at the conference table. The office of mayor had once belonged to his father, Edward “Babe” Voorde, whose term ended tragically when he was killed in a car accident in 1960. Wearing his usual sweater vest over a shirt and tie, John smiled benignly and leaned forward a little in his chair. I liked him, though our styles were certainly different. John was mainly a creature of the old school—he had worked various city jobs, beginning on the street department, and was city clerk at the time I first got elected. He was one of many people of my parents’ generation to support me when I first ran, and we got along well. But as a council member he was sometimes unpredictable, and I couldn’t count on his vote without spending time with him to make sure I had made my case and asked where he stood.
To answer his question: I had not seen Ken Burns’s new PBS documentary series on Vietnam, but it was getting a lot of attention. It seemed to be especially resonant for the generation that had experienced it as the dominant issue of their coming-of-age. As John reminisced about various people in his St. Joseph High School Class of 1962 affected by the draft, I worried that it might be a while before we came to the topic of my budget proposal. Then he said something that made clear he had been thinking of city affairs all along:
“Sometimes, Pete, when you talk about your data-driven government, I think of Robert McNamara.”
I just smiled, not sure exactly where to take the conversation from here. Coming from John, or really anyone who looked back on the Vietnam War with anguish, being likened to LBJ’s Secretary of Defense was not exactly a compliment. By all accounts, McNamara had been a brilliant individual, a genius even, his rimless glasses and sharp gaze embodying modern technocracy at its finest. But the outcome of the war—and David Halberstam’s book The Best and the Brightest—made the sum total of his brilliance seem dark and ironic, as he and the other geniuses of the national security establishment led our nation into quagmire and defeat.
I could also see where the comparison was going. Before serving in public office, McNamara had been the CEO of Ford Motor Company, and the use of data and metrics on his watch escalated almost to a kind of fetish. After the Vietnam War collapsed into chaos, historians and journalists inquired into how the most brilliant minds of their generation could have led the country into such a lethal blunder, and the image emerged of McNamara as a data-obsessed manager who missed the forest for the trees. “Statistics and force ratios came pouring out of him like a great uncapped faucet,” Halberstam wrote. Yet, for all the statistical brilliance of McNamara and the rest of President Johnson’s inner circle, all of them were tragically late to the obvious fact that the war was a losing one, keeping America entangled there at a cost of thousands more American lives.
This must have been on Voorde’s mind when he had turned the subject of our conversation to the road surface app that he had heard about. My administration was in the process of creating the first objective asset map of the city, cataloging the quantity and quality of streets, fire hydrants, signs, and anything else in a public right-of-way. This work even included an app, which could be run on an iPhone mounted on the dashboard of a supervisor’s vehicle, to scan the conditions of the road and report cracks, potholes, and other deterioration.
Thinking back to his youth on the street department, John was skeptical. “You have this technology to tell you which streets need repair,” he said. “But if your foreman’s any good, he ought to already know that off the top of his head!”
The technology had other capabilities, and I’m glad we use it—but, admittedly, the councilman had a point. One of the reasons we have qualified, experienced individuals in organizations is to use their intuition and expertise to solve problems. If the foreman of a street crew knows every crack on Lincoln Way West and every pothole on North Shore Drive like the back of his hand, why do I need to spend money on an app to tell us where the problems are?
For all the power that data analysis represents—and I’ve worked to build a reputation for running one of the country’s most data-oriented city administrations—it also has its limitations, and the potential for mischief. You might spend lots of time and resources gathering data that will never be used, or accumulate data that winds up telling you things you already know.
SO HOW DOES A TECH-ORIENTED MAYOR make sure that the data is serving the administration, rather than becoming an end in itself? Put another way, how does a government official interested in data come to be viewed more like Goldsmith or O’Malley, and less like McNamara? Over time, I’ve learned a number of rules that have helped us to make sure the use of data makes sense, and does good.
First, know the difference between reporting an issue and resolving it. In some cases, the two go so closely together that you can lose track of the distinction. For example, when we installed ShotSpotter technology using microphones to acoustically pinpoint gunshots, we were enhancing our ability to deal with gun violence. An officer could be immediately dispatched to the scene of a shooting, be it an outdoor fight or a domestic violence case, whether someone called it in or not. And this, in turn, would help in the long run to deter gun violence. But in other cases, knowing more doesn’t help. At a tech conference, I once saw a pitch from a start-up that would automatically detect patterns of opioid use by scanning for trace amounts in sewage. The technology is brilliant, and may do a great deal of good in some places. But in South Bend, our problem wasn’t knowing how much opioid use was prevalent in this neighborhood compared to that one; it was a lack of mental health and addiction resources to deal with the issue wherever we found it. Financing a project to tell us more about the problem could even come at the expense of treatment options, which are grossly underfunded in our county and state health systems. In cases where we have ample means to fix a problem, then we need only to find it. The rest of the time, reporting an issue is necessary, but not sufficient, for resolving it.
A second rule we learned quickly was to recognize that responsiveness and efficiency are not the same thing; in fact, they can sometimes pull against each other. Consider the example of snowplowing. The most responsive thing to do would be to ensure that anytime someone called about an impassable street, a plow crew was immediately dispatched to take care of that block. It would be an attractive thing to be able to do (think of the political credit)—but it’s also clearly not the most efficient; far better to use a zone system, covering the city as quickly as possible, starting with main roads and then movi
ng to residential streets, with added input from a parametric model that takes temperature and precipitation rates into account. Any other approach would take longer, and ultimately mean less quality of service and/or more cost. Local officials often feel pressure to deal with a squeaky wheel right away, when stepping back and considering a big-picture solution would serve people better.
Under the wrong balance of responsiveness and efficiency, data can actually make us worse at our job. This is one reason I eventually backed off from my enthusiasm for the idea of publicizing a twenty-four-hour pothole guarantee. It seemed at first like a great way to show how responsive the city was to road concerns—and doable, because in peak patching season we already get to most potholes within a day or two of them being called in. But after reviewing the concept with engineers, it became clear to me that if I instructed the staff to make sure every hole got taken care of as soon as we knew about it, I could actually reduce the efficiency of the operation. Crews on the West Side might have to drop what they were doing to go deal with a pothole on the North Side, then go chase another work order downtown, all coming to them in order of appearance. An expensive vehicle and work crew would zigzag through the city according to real-time data on which residents were first to call and complain, with little regard for whether it made more sense to have Harter Heights wait a couple days while we systematically took care of the Keller Park area for the season.
At other times, the reverse is true and responsiveness really is more important than efficiency, as in the case of graffiti. It might seem that the most efficient thing would be to treat graffiti like snow—take whatever resources we have for repainting, and have them work the city, street by street, systematically. But if a stop sign gets tagged with graffiti, leaving it there even for a couple days might motivate someone to tag something else nearby. Whether it’s a gang sign or a cartoon bunny, what shows up on Falcon Street may soon be copied on Walnut, and the longer it’s there, the more likely someone will seek to imitate or outdo it. So clearing it right away is the most important thing, and the team works to fix any reported graffiti almost immediately (except on a dedicated graffiti wall opposite the Emporium Restaurant, where artists are welcome to do whatever they like). The result is that people who might be motivated to deface public property find it’s not worth the effort, and now it is less likely to happen in the first place.
A third lesson on data and efficiency is to be honest at the beginning about whether you are willing to follow the data where it leads. When I asked Eric Horvath, our public works director, to get creative on ways to keep solid waste billing rates under control, his team came back with options from selling ad space on city trash bins to charging customers differently by how much they throw away. The most usable idea involved a technology for partly automated trash trucks that can pick up a bin with a robotic arm, eliminating the need for a human “picker” on solid waste crews. This meant a savings for the city, keeping rates lower—and, we learned, led to lower injury rates as well. But buying the technology was only worth it if we were prepared to eliminate the jobs. It wasn’t an easy thing to do, because our solid waste workers were likable and hardworking. In the end I decided to go ahead, because the city could offer the workers other jobs, provided they earned a commercial driver’s license. Half the affected workers did so, and half left city employment altogether.
But in other cases, we are not prepared to capture an efficiency when we find it. For example, we continue to operate a walk-in center for paying your water bill, even though this can be done online, over the phone, and by mail. Part of me (the consultant part of me, naturally) finds this maddeningly inefficient: Why pay for a brick-and-mortar structure, and staff, at a facility whose work can be done more quickly, efficiently, and easily by other means? But the more I looked into the issue, the clearer it became that low-income residents who did not have bank accounts relied on the facility so they could pay in cash. The long-term solution would be to help them to get banked, but the reality is that this will not happen for some of our residents. The right thing to do here, it seems, is to tolerate an inefficiency for now, even though the data tells us how it could be eliminated.
A fourth data lesson came from the ShotSpotter experience: follow the data where it leads, and recognize that it could show you the answers to questions you never even asked. When we adopted the technology, the obvious appeal was that police could be dispatched immediately to the site of a shooting, without having to rely on someone quickly calling it in to 911. We knew there would be a tactical advantage, but only slowly did we realize this would also be a powerful tool for both measuring and changing the relationship between community members and the police department.
As our police chief at the time, Ron Teachman, explained, “Law enforcement projects an air of omniscience. If residents hear a gunshot and don’t see an officer coming to the scene, they don’t think it’s because we don’t know about it. They assume we know about it, and that we’re not there because we don’t care.” With the new technology, officers appeared on the scene of shootings we simply didn’t know about before. And gains in police legitimacy could be achieved by using the community policing method of “knock and talk” in concert with the technology. When the system detected gunshots in a residential area, officers would work that block the next day, letting residents know they were concerned and leaving door hangers for those who were not there, explaining why they had visited and how to follow up.
Soon the ShotSpotter data became a measure for tracking something completely different from gunshot rates: perceived police legitimacy. Since we knew from the sensors how many gunshots were fired in the coverage area, and we also knew how many times someone in the same area called 911 for shots fired, we could now tell what proportion of the time people heard gunshots but didn’t bother letting us know. For the first time, we had an index of how many people thought it was worthwhile to call the police about gunfire near them, a hard number to help us measure something very difficult to quantify: trust in the police department.
Initially, we had assumed that a small fraction of gunshots went unreported, perhaps 20 percent. Instead, we learned that, shockingly, the reverse was the case. Now, we watch the ratio closely; I can log on to a law enforcement dashboard that will tell me, on a monthly basis, what proportion of shots are being called in. It’s only one measure, and an imperfect one, but I use it to help get a sense of how much residents think it’s worthwhile to call the police.
This leads to another concern when it comes to data-driven government, or government in general: the confusion of technical problems with moral ones. In many ways, it is psychologically easier to deal with technical problems, ones with right and wrong answers. In these cases—how to make pothole patching more efficient, or get more children tested for lead exposure—it is clear that if we find a more efficient way to proceed, by definition it should be done. But in many ways, political leadership isn’t required for these technical gains, other than to give a green light to staff who identify ways to make them. Elected officials earn our keep by settling moral questions, ones where there is no way to make someone better off without making someone else worse off.
Even the most ground-level decisions can have this character, as when we switched the trucks for trash pickup. Not only did it mean some city workers losing a job, it also meant that we had to get residents to haul their trash bins to the front curb once a week, since the newer trucks couldn’t operate well in narrow alleys. Moving the bins is a pain, and there’s no getting around this when interacting with a resident who would rather it all stayed in the alleys. We were presented with a trade-off: keep it in the alleys and let trash pickup be more expensive for everyone, or move it to the front and make it more inconvenient for some. No math could solve this problem or present an obvious right answer; we just had to make a call, and then be willing to explain it to those affected.
A more serious version of this trade-off came up when an ethanol plant went out o
f business, and houses nearby found their basements full of water. It turned out the plant was such a big water user that it artificially depressed the water table—something the home builders did not take into account when deciding how deep to dig the basements. Now both the ethanol plant and the home builders were gone, and a bunch of homeowners were flooding the council chamber demanding to know what I would do to fix their problem. The city was not technically involved, but these homeowners had done nothing wrong, and it seemed we needed to do something to help.
Ultimately, we decided to re-create the effect of the depressed water table by pumping water into a ditch, to get the homeowners some relief. In the end I got lucky: a new operator took on the ethanol plant and began pumping again. But in the meantime, it seemed that we were faced with a problem that no amount of data could solve: Do we undertake a deliberate waste of water and energy costing every city resident a few cents, or let a handful of homeowners lose thousands of dollars of value on their homes through no fault of their own? Again, there was no technical answer to this problem; it was a question of who would suffer and how much.
The question of suffering brings me to one last concern around the use of data-driven techniques to bring about better government: the importance of exceptions, otherwise known as mercy. Efficiency, almost by definition, has to do with following rules and patterns; if there is an inefficiency within a rule, it can be ironed out by making a sub-rule. But sometimes our moral intuition just tells us that making an exception is the right thing to do, even if we can’t explain or defend the precedent.
Shortest Way Home Page 19