Here’s a look at how predictive data analytics is being applied to tweets, Fitbit data and social media comments in U.S. cities to detect wellbeing
By Matt Hamblen
More than 25 years ago, predictive data analytics technology began guiding grocery chain marketers to display seemingly unrelated products together on supermarket shelves to achieve maximum sales. In an early 1990s example, one Washington, D.C., supermarket chain reported a sudden revenue boost after bananas were displayed next to disposable baby diapers.
That kind of data analytics insight has evolved and grown through the years for all manner of businesses, especially those selling consumer products. Predictions have been made for the hottest smartphones and savory pizzas to help guide tactical and strategic marketing decisions.
Recently, city and other local governments have started to use public opinion surveys and other publicly available data gleaned from social networks to gain insights into the ways that cities can plan policies and how revenues should be spent for maximum impact.
One unusual twist of late relies on publicly-posted online social networking comments found on Twitter, Reddit, Instagram and others to determine factors such as community wellbeing, or happiness. Names and other personal information are stripped away before the data reaches the cities. Comments on sites like Twitter such as, “Jobs in my field aren’t easy to get in this town,” or, “I love Central Park at sunrise” can now be indexed, sorted and scored across an entire zip code or neighborhood to measure civic wellbeing.
In related fashion, Santa Monica, Calif., will soon launch an unusual Fitbit user project with nearly 200 volunteers wearing the digital wristbands to see how active certain residents are, and how they respond to future city programs intended to promote exercise and active lifestyles. A city survey found in 2015 that 48% of residents were active at least 20 minutes a day, a number that dropped to just 38% in 2017—something of a surprise given the city’s location on a famous beach and pier with abundant sunshine.
Separately, the city learned through its surveys that many people are eating less than half the national daily average of fruits and vegetables. As a result, the city staff went to farmer’s markets to enroll low-income residents in a state-run CalFresh program that offers them financial aid to help purchase fresh produce from nearby farms. Library cooking classes were also kicked off to help residents prepare those fresh ingredients.
The Fitbit findings on activity levels could possibly help the city in planning places to add street lighting and street trees or where to install bike lanes. A joint project with the University of Southern California in early 2017 looked into sentiments about Santa Monica shared in comments on Instagram and Twitter. Researchers hope to discover the factors that contributed to those feelings by looking at some data inside anonymous user profiles on various social networking apps.
Early Days for City Wellbeing Analytics
“We’ve done some social networking analysis and we’re testing the waters, but it’s still early days” for predictive analytics in cities, said Lisa Parson, project manager for the Santa Monica’s Office of Civic Wellbeing. “Predictive analytics is good for marketing, but for public policy it’s more challenging. But if you get it right, it can be valuable. There’s such a large population of data available that we need to find ways to harness it.”
Julie Rusk, the city’s chief of Civic Wellbeing, said her job function and the use of data analysis is intended to be forward-looking, perhaps many years into the future. “It’s important to think about our purpose, the why of our city’s existence and how to create a 21st century government,” she said. “Instead of being a collection of services, how can our city act together with the community?”
Santa Monica is committed to using types of data that go beyond typical administrative and survey resources, she said.
Hers is an unusual role for any size city, much less one of 93,000 people. Santa Monica is pioneering a number of ways to measure and improve wellbeing, as described on its wellbeing website.
“I am not aware of other civic wellbeing offices or chiefs,” she said. Just as the theme of environmental sustainability emerged 20 years ago and is now being advanced at all levels of government, she believes a city’s “wellbeing” is taking a similar trajectory, even though it is still early days.
“Clearly, sectors such as economics, healthcare, education and governance have got to work in more interrelated ways and transform to be effective and relevant,” she added. “Local government in particular can get very trapped in legacy operations and services and lose sight of its purpose and key outcomes. Part of why we developed a framework for wellbeing is to help focus local government on its purpose.”
The city’s online survey, created with a panel of international experts, is open source and will be shared with other cities. Meanwhile, Rusk said she plans to discuss the analysis of social media content for insights into city wellbeing with Evolve24, a predictive analytics firm.
Smart Cities and Evolve24 Analyzing Six Major Cities
Working in partnership with the Smart Cities Council, the 13-year-old Evolve24 has been analyzing anonymous Twitter and Reddit content in six large cities for nearly a year. An enormous volume of such content has been studied in six major cities: New York, Chicago, Houston, Los Angeles, Philadelphia and Washington, D.C.
The project is both ambitious and complex.
By studying the online content related to the subjective wellbeing of citizens, Evolve24 has actually derived a correlation between emotional wellbeing and a city’s economic growth. More specifically, the company has found that each one point increase in a 100-point emotional wellbeing scale correlated with a $10 million increase in new construction about two months later, Evolve24 CEO Anthony Sardella said in an interview.
That correlation is being tested. Future research will seek to establish a firmer causal relationship between emotional wellbeing, or happiness, and economic growth, Sardella said. It may seem fairly obvious already to a layman that happy, satisfied and optimistic people will spend more or borrow more, which would likely be behind the reasons for developers to build new homes, stores and other buildings. The advantage of using the social media data is to detect ups and downs during a particular time so a city can respond in a thoughtful, agile manner while also laying groundwork for long-term changes.
“We’re hoping to provide city officials ways to measure subjective wellbeing so they can make really good decisions about how to allocate resources,” Sardella said. If a mayor knows there’s a sudden surge in goodwill, for example, then he or she could make sure to hire more city workers to review building permits before the demand for permits and inspections hits.
“There’s been so much focus on technology and Internet of Things to measure the efficiencies of a city, but we think equal focus must be on wellbeing of citizens,” Sardella added.
Santa Monica’s Rusk called Evolve24’s work “intriguing” and she hopes to learn more about it. Several officials from different cities said that the cost of analyzing social networking content promises to be far cheaper than commissioning large scientifically grounded surveys on an annual basis.
Social networking content is even arguably a more reliable indicator of true sentiment than public opinion surveys, Sardella contended. That’s because surveys judge perceptions and opinions at one particular moment in time, often just once a year, while social media content can stretch across days or weeks to give a more comprehensive picture. “Greater than 70% of a score of a survey is relative to the moment you ask the question, and if you ask on a Monday morning versus a Friday, it’s not a good distribution of data,” he said.
Also, social media content can offer insights to nuances that people won’t answer in public opinion polling, Sardella said. Comments on social media aren’t normally prompted by a specifically worded question, as in a poll, and can be broader ranging—even hitting areas a city official might not have thought to ask.
Secrets behind City Wellbeing
The secret sauce of Evolve24’s analytics work is how well the company’s researchers search and then interpret comments online and score them on a 100-point emotional wellbeing scale. That wellbeing scale is also averaged with several other measures to create an uber-wellbeing 100-point scale. It seems like a daunting task as Sardella explained it:
“You get the content and the words from Twitter and Reddit. We had bins of words over 10 months. We’re not really looking at the words as much as what does a comment mean when somebody uses that language. Are they communicating in an angry state, for example. It’s a heavy lift.”
For example, a comment such as “There are no jobs in this city” or “It’s not worth investing here” or “I’m not going to buy a house or get a job next month” would produce a lower emotional wellbeing score. By contrast, more optimistic comments like “I love living by the lakeside” or “I feel really good about where things are going for me” or “Amazon might bring a lot of jobs to the city” would bring a higher score.
Evolve24 used dimensions of emotional wellbeing derived from years of academic literature in psychology to build its system for scoring social media content. In 2016, noted University of Pennsylvania psychology researchers Alejandro Adler and Martin E.P. Seligman described the measurement of wellbeing for use in public policy in a scholarly article, “Using Wellbeing for Public Policy.” The researchers argued that the “science of wellbeing is mature enough to complement economic assessments of national progress.”
Sardella wants to apply the resolve that Adler and Seligman found to cities and counties and their planning processes.
The University of Michigan Consumer Sentiment Index, a telephone survey, already is widely used by U.S. government officials to get a sense of people’s attitudes on whether the economy is improving or not.
The index is based on the premise that “emotion moves us first, which drives decisions on investments,” Sardella said. “If we’re in a pessimistic state, the principle is we’ll spend less. “
As ambitious as their project is, Smart Cities Council and Evolve24 don’t plan to use their methodology to rank various cities on wellbeing. “We want to provide insights to cities and avoid rankings,” Sardella said. Houston, after Hurricane Harvey hit the city in August, couldn’t be expected to rank very high on an overall measure of wellbeing, but social networking results from there in a year or two after the storm hit could show leaders how residents feel about reparations.
Social media content posted online by tourists could also distort the true feelings of residents and workers in a city. Evolve24 can filter out such content based on the location-–perhaps if it is a museum or a restaurant or other gathering spot known to be especially popular with tourists. Or, a comment such as “My fantastic first trip to the Santa Monica Pier and California” could be judged on its face as applying to an out-of-town guest.
Limitations of Social Analytics for Policymakers
Evolve24’s work with cities and the Smart Cities Council is still in the early stages. Three technology industry analysts who study the data analytics industry said data analysis using content from social networking sites has been used for several years to judge popular sentiment on a wide range of topics–mainly consumer products–although it is not widely used thus far in cities.
Santa Monica and other cities are wise to regard social media commentary as another tool in their toolbox, one analyst said, alongside scientific surveys and some good old-fashioned gut feelings by city administrators and elected officials honed after years of running for election and serving in office. Big city mayors and their deputies have been known to hang out at neighborhood bars and civic gatherings while wearing a disguise to get the pulse of the community.
“Using Twitter and Facebook to look for positive and negative comments is a good indicator of customer satisfaction—within limits,” said Jack Gold, an analyst at J. Gold Associates. “There’s always the risk that disgruntled people will overweigh the results or the positive side could too. And bots could also be used to skew the results, so you have to be cautious.”
Concerns about Russian influence in social media and online advertising since the 2016 U.S. election might raise some red flags in the minds of city leaders. But that influence apparently hasn’t cooled the interest by data scientists in finding new ways to tap the feelings of city dwellers.
“In this day and age, everyone is questioning what’s fake and what’s real in terms of data,” said Justin Bibb, a senior adviser at Gallup, the global analytics and advisory company, in an interview.
“The best way to understand the state of mind of citizens is to conduct scientific analysis and data collection on things that matter most to city officials. We are a big believer in probability sampling. We also conduct analysis of social media and other third party data sets alongside existing Gallup metrics. We believe in taking a holistic view and having a multi-pronged approach to problem-solving.”
A Brief History of Social Analytics for Policymakers
Data analytics industry attempts to look at public sentiment in cities through social media and blogs go back at least as far as 2012. That year, IBM described its “Intelligent Operations Center for Smarter Cities,” an ambitious package of software which seemingly covered a gamut of things cities care about. One of the product’s 10 components was designed to provide “social media analytics [to] let city officials gather and analyze public opinion on city operations, plans and services.”
In 2014, a company called DataSift partnered with the United Nations’ Global Pulse initiative to analyze worldwide opinions on vaccinations and poverty, among other topics. Global Pulse built a social media monitoring dashboard to explore online conversations in multiple languages about climate change and more.
In 2015, a case study published by the Global Commission on the Economy and Climate and social analytics vendor Crimson Hexagon showed how search terms were used to analyze tweets and other social media to assess public opinion on the relationship between reducing carbon emissions and the economy.
Crimson Hexagon, based in Boston, is an example of a social analytics vendor that has performed data analytics for companies like Samsung, while also engaging in public policy work. A number of smaller social analytics vendors focus heavily on working with businesses, an indication of the growth in the field that stands to expand to the public sector. Such vendors include Hootsuite, Lithium, Sprout Social, Synthesio, Buffer, Brand24, and Zoho Social.
Big names like IBM, Adobe Sytems, Oracle, SAS, and Salesforce are also in the social analytics market primarily for business clients, and that market is expected to reach $2.7 billion in 2017, growing by a phenomenal 28% a year to $9.5 billion by 2022, according to a forecast by Research and Markets.
Balancing Pros and Cons of Social Analytics
One chief advantage of analyzing social media content is that the data is publicly available, said IDC analyst David Schubmehl. One potential downside is that concerns could arise about personal privacy if a government entity connects social media data about wellbeing with police or housing records, he added.
In 2016, the Brennan Center for Justice at the New York University School of Law posted a U.S. map showing dozens of cities, counties and law enforcement agencies that have spent more than $10,000 on social media monitoring software.
Despite the concerns some privacy advocates have raised about social media monitoring by police, there have been some obvious concrete beneficial outcomes from scouring online comments. In one example, in San Francisco in 2016, food safety got a boost with analytics from thousands of publicly-available Yelp reviews of restaurants. Researchers at the University of California at Irvine found more than 71,000 Yelp posts that contained keywords associated with foodborne illnesses, then built software to correctly identify health code violations in some restaurants.
Worries about privacy with analyzing social media feeds have to be weighed, Schubmehl said. “I don’t see anything wrong with listening to your citizens on Twitter or Facebook or other social media platforms to understand what they are saying about your city or how it can be improved or changed,” Schubmehl added.
“Like anything else, it’s caveat emptor,” he said. “But having said that, social media is another channel that governments need to watch and analyze.”
Targeted email campaigns and other traditional forms of letter-writing lobbying could be no less biased or backed by a single political group than online comments on social media, he noted. Even email campaigns to politicians can be conducted with the use of automated bots, skewing the numbers seen by elected leaders.
The Future for Social Analytics in Cities
For cities such as Santa Monica, outcomes from analyzing social media posts promise to become more concrete in future years, even as the definition of a city’s overall wellbeing can be somewhat abstract. The definition of wellbeing and happiness will certainly evolve and might never become standardized across all U.S. cities.
The analysis being conducted of social media in six major cities by Smart Cities Council and Evolve24 focuses on a community’s wellbeing, which likely incorporates social media references to concrete subjects such as food safety and dozens of other indicators about such things as air and water quality and traffic congestion as well. One major focus of Smart Cities Council’s project is to give communities a better say in their future.
“Santa Monica and Dubai are pioneers in deciding how to prioritize efforts based on citizen wellbeing,” said Philip Bane, managing director of Smart Cities Council. “Now that the science allows for the discovery of a community’s well-being, there is no reason not to consider using it as an indicator of livability, workability and sustainability. It is important to allow the community a voice in prioritizing city projects.”
I’ll be following this topic in coming months. If you have insights or comments, please email me at: firstname.lastname@example.org
Read more "Judging a city’s wellbeing using tweets and social commentary"