As part of my (sporadic and casual) work on how the UK police use Twitter, I’ve now set up a mini site so that people can see some basic statistics for all the official police Twitter accounts in the UK. The site is at lesscrime.info/policetweets/stats. You can filter the data in various ways to compare accounts in different parts of the country or relating to different types of policing. For example you can see all the Twitter accounts belonging to West Midlands Police, which chief constable has the most followers or which police helicopter was the first to respond to noise complaints via Twitter.
For each Twitter account, the site shows some basic information about the account, how I’ve categorised that account in different ways and four different metrics. The information is not-quite live because the underlying data are updated overnight, rather than in real time.
Basic account information
This section shows the account name and Twitter handle, as well as the number of tweets sent from the account and the number of followers the account has. It also shows the date on which the account was created, which may or may not be when the first tweet was sent.
Just below the account name is a short description of the account. Twitter accounts some with their descriptions, but these are free-text fields and not always useful. Indeed some of them didn’t even specify which force the account belonged to or where in the country the officers tweeting were based.
For every account, I wanted to know which force an account belonged to and what type of policing the account related to (neighbourhood reassurance, incident response, crime investigation, air support and so on). For individual accounts I also wanted to know the rank of the officer concerned. Since people get promoted and I did not want to have to change the details too often, I’ve grouped the ranks so that sergeants, inspectors and chief inspectors are listed as junior managers, superintendents and chief superintendents as middle managers and more-senior officers as senior managers.
I developed the categories for type of policing from the data, rather than imposing a pre-set list of types. This means I may need to add new categories in future, but for now the types are:
- air support,
- criminal justice (includes officers based at courts and in case-file management teams, but not investigators),
- dog handlers (and, in at least one case, their dogs),
- firearms policing,
- football policing,
- district/force main account,
- general or unspecified manager,
- marine support,
- mounted section,
- neighbourhood policing (includes community engagement and schools officers),
- proactive patrols,
- public-order policing (except mounted and dog sections),
- public protection (includes domestic-violence investigations and sex-offender management),
- response policing (includes control rooms),
- roads policing,
- specialist-crime investigation (includes crime-scene examiners) and
- staff associations (the Police Federation, Superintendents Association and so on).
In some cases I had to guess which was the correct category for a particular account because the Twitter description offered very little information. I based my guess on the content of the tweets sent from the account and on a Google search of the account name, but in a few cases there was no information on which to base a decision so I had to leave the category blank.
I’ve calculated four metrics for each account and given them light-hearted names to emphasise that they are approximate and readers should interpret them with caution:
This is the number of new followers the account attracts, on average, for every tweet sent. It may or may not be a rough indicator of the quality of the tweets sent, but some accounts (such as the main account for each force) attract lots of followers even if they don’t tweet very often and regardless of quality.
This is the number of weeks that the person had been active on Twitter, defined as the difference between the date they joined Twitter and the date of their most-recent tweet. This measure might be distorted if an account was set up as a placeholder and left untouched for some time before being put to use.
(that is a word, I checked)
This is the number of tweets the user sends, on average, each day (including retweets and @ replies).
This is the number of days since the user last tweeted. If no tweets are sent from an account for 90 days, it is excluded from the tables until a further tweet is sent – this stops abandoned accounts from clogging up the results.
To discourage people from fixating over small differences in metrics, I’ve presented them on a five-point scale, where each point represents one fifth of the difference between the minimum and maximum value present in the data. Tool tips are provided that show the actual value of each metric.
Why is everything wrong?
Some caveats. First, I’m only tracking official Twitter accounts that are on a list maintained by Nick Keane from the College of Policing. If an account isn’t on Nick’s list, it’s not included in my stats. If you think an account should be included and isn’t, send him a tweet. Bear in mind he only includes accounts that are run by an identifiable police officer or police-staff member, or an identifiable police team, so the list does not include anonymous or semi-anonymous accounts.
Second, the statistics site only includes active Twitter accounts, by which I mean accounts that have tweeted at least once in the past 90 days. If an account isn’t included because it’s dormant, just tweet from that account and it will appear on the site tomorrow after the site updates overnight.
Finally, in manually categorising over two thousand Twitter accounts I will have inevitably made some mistakes. The details of some accounts will also have changed since I categorised them, for example if an officer has been recently promoted. If you see any problems in the data (particularly my categorisation of accounts), please click on the ‘!?’ button to the right of each account to report it. Thanks!