Tottenham riots: Data journalists and social scientists should join forces

In the wake of some of the worst riots in London in more than a decade, Ben Goldacre has said on Twitter:

[blackbirdpie id=”100156268534181888″]

Yes, we’re now going to have to suffer through lots of ill-informed speculation from columnists. Brace yourself yet again as they take out their favourite axe from the kitchen cupboard and grind away on it just a bit more until the head is gone and they’re whittling the handle into a toothpick. It will enrage more than enlighten.

I have a better suggestion. With the current interest in data journalism, this would be a great time to revisit one of the seminal moments of data journalism carried out by Philip Meyer in the wake of the 1967 riots in Detroit. As a Nieman fellow at Harvard, Meyer studied not only how social science could be applied to journalism, but he also explored how main frame computers could be used to quickly analyse data. (For data journalists, if you don´t already own it, you should buy a copy of  Meyer´s book, Precision Journalism, first published in 1973 and since updated.) As a national correspondent for Knight-Ridder newspapers, Meyer was sent to Detroit to help cover the riots.

The 1967 Detroit riots stand as the third worst in the history of the US, only eclipsed by the 1992 riots in the wake of the acquittal of police officers in the beating of Rodney King and draft riots in New York during the US Civil War. As the Institute of Social Research at the University of Michigan said:

The Detroit Free Press and the Detroit News threw every resource they had into covering the uprising. And as the disturbance died down, journalists and commentators, most of them white, struggled to understand who the rioters were and why they had taken to the streets. One theory was that those who looted and burned buildings were on the bottom rung of society—riff raff with no money and no education. A second theory speculated that rioters were recent arrivals from the South who had failed to assimilate and were venting their frustrations on the city.

But for many, those theories rang false.

A survey had been done following the 1965 Watts riots. Meyer approached Nathan Caplan, a friend from graduate school at the University of Michigan in Ann Arbor. They both had a similar idea to see if a survey similar to the one done after Watts could be done in Detroit. One challenge was that the Watts study took two years, but Meyer wanted it done in three weeks. The ISR has an article that looks at the process in great depth, and what is clear is that the study of the 1967 Detroit riots and the journalism that followed had a lot of support not only from the newspapers but from the university, government and local foundations. They recruited and trained 30 teachers to conduct the surveys, drew up a random sample and interviewed 437 black residents.

The survey debunked a number of theories put forward to explain the violence.

  • One theory was that the rioters were poor and uneducated. No, the survey found otherwise. ¨There was no correlation between economic status and participation in the disturbance. College-educated residents were as likely as high school dropouts to have taken part.¨
  • Another theory laid the blame at recent arrivals from the south who had little connections to the community. That theory was also wrong. ¨Recent immigrants from the South had not played a major role; in fact, Northerners were three times as likely to have rioted.¨

Like Ben, I´m sure that we´ll see hours of speculation on television and acres of newsprint positing theories. However, theories need to be tested. The Detroit riots showed that a partnership amongst social scientists, foundations, the local community and journalists can prove or disprove these theories and hopefully provide solutions rather than recriminations.

11 thoughts on “Tottenham riots: Data journalists and social scientists should join forces

  1. While the interest in the call is good; there are fundamental problems with it as it stands. This comment covers gross-generalisations in order to be (still not very) short 🙂

    The cited study was possible because the team was already together, and was able to (mostly) repeat what had been done before in a previous, just finished study. Does one exist? Doing a random sample poll of tottenham phone numbers is relatively cheap, easy, and probably massively inaccurate.

    The fundamental issue is not analysing the data – it’s collecting representative, reliable (and lots of other words like that which make it usable) data in the first place. It takes some time (especially if you don’t know the area) – @benGoldacre tweets issues here. The data journalism focus is all on analysing data and writing about it; that’s great; but it’s a very different beast to collecting representative and accurate data to be analysed.

    The Guardian seems to have done an excellent job of getting reports from Tottenham (Paul Lewis went out there for a start), and twitter provides others, but those will not be a representative sample. They are deeply personal (qualitative) stories, which focus on their experience; and are great for news pieces; but such a survey would also need a large sample of people asked the same questions (quantitative data).

    Currently, we have only seen one aspect of the public side of this. As Paul Lewis mentioned in a tweet last night, lots of other bits of the story are passing between friends on BBM, and not in the public sphere. A story given to researchers will be quite different to the story given to a journalist (traditionally framed). Working together is important; but you’re after independent data. Informed Consent matters here. Anonymisation matters here. The equivalent of showing up tomorrow morning with a clipboard and a stack of questions will probably get you mistaken for a police officer. That needs aforethought to be avoided.

    Many new approaches can make it easier to get long form responses; but they can’t easily get comparable responses to the same questions and also be representative of the community.

    That is not to say that it is impossible; but to say that it is difficult at this point of time. However, given that this is likely to happen again in the next few years, doing something, limited, flawed, now and putting a group somewhat together ready for next time may provide the ability to do that for then.

    This is a conversation worthy of discussion and consideration for how this could happen after what the police term a Major Incident. I get the feeling this will not be the only time such an event occurs in the next few years.

  2. Sam,

    Your comment has a few assumptions that aren’t accurate. The team wasn’t already together. They did use a previous survey, but they had no team. Philip Meyer had a personal contact at the University of Michigan, but that’s about it.

    In terms of the randomised sample, that was generated by the research team. I’m sure that a good social research team here in London could come up with a statistical sampling method. In terms of the Detroit survey, the sample size was 437, relatively small but deemed to be representative. I guess a key question is: How big of a sample is needed to be representative?

    As for folks with clipboards being perceived as a threat, that is why the original survey team recruited and trained black teachers from the city who wouldn’t be greeted with suspicion by the community.

    I’m not entirely clear why you’re bringing BBM and Twitter into this other than to say that there is still reporting to be done. I’m not suggesting that a survey and analysis is a substitute for reporting. I did suggest that we put some more energy into a survey like the one done in 1967 rather than wasting gallons of ink of speculative commentary, but I didn’t suggest it was a substitute to reporting.

    As for the time required, this was accomplished in 1967 in 3 weeks. A smartphone probably has more compute power than the mainframe they used in 1967, and they also wouldn’t have to convert the data into punch cards as they did in 1967. That’s one less speed bump.

    Even if it might prove difficult or impossible to do this in three weeks in this case, I still think there is value in building up the links between journalists and social scientists to support this kind of work.

  3. Having reread your post, my comment, and yours, and your later comments on twitter. The scale of this is very different, now (Tuesday am), to the scale on Sunday lunchtime. That change has probably made somewhat easier (but not necessarily easy) the parts that aren’t related to the amount of computer power or communications available. Those areas were the focus of my comment.

    Not wishing to comment further apart from one note: we are probably still in the middle of this. Things will probably change more.

  4. Sam,

    Reading this most recent comment, it seems that you believe I’m suggesting a network analysis of Twitter or social media communications (I’m especially noting when you say “the communications available). I’m not. I’m suggesting a proper social survey that would inform coverage and public discussion about the riots. Surely, there has to be people with social survey expertise in the UK that could help with this, just as the University of Michigan played a significant role in the Detroit riot survey.

  5. Hi there,
    I enjoyed your piece. I plan on doing some research into these riots in under two months. Is a survey really necessary or is a media analysis sufficient?
    Your opinions are appreciated

Comments are closed.