A Sketch Map of Part of the G4S Corporate Structure and a Light Shone on Some Public Payments to Them

In case you haven’t noticed, huge amounts of public money are spent with (profit making) private companies to deliver what we might otherwise think of as public services. Today, it seems as if the Army has been called up to cover the back of security firm G4S in providing security staff to, erm, police the Olympic Games (e.g. Guardian:Olympic security: army reinforcements called in to fill G4S shortfall, FT: Army steps in after Olympic ‘shambles’).

I’ve done a couple of sketches with G4S related data before, so not wanting to miss a little stirring opportunity, I thought it might be worth posting them here…

As with most large companies, G4S is not a single entity (that would be too easy…) Rather, complex corporate structures are developed in order to separate concerns and facilitate group wide, tax efficient relationships between different legal entities. WHilst we might think of G4S as a single company (“G4S”), it is actually a corporate sprawl, made up of a myriad of other companies that often share directors as well as shares in each other.

Some time ago, I started exploring how we might be able to use directors’ dealings, as collated by OpenCorporates, to sketch out maps of how corporate sprawls are structured (Mapping the Tesco Corporate Organisational Sprawl – An Initial Sketch. Here’s a recap of the recipe I used:

– grab a list of companies that may be associated with “Tesco” by querying the OpenCorporates reconciliation API for tesco
– grab the filings for each of those companies
– trawl through the filings looking for director appointments or terminations
– store a row for each directorial appointment or termination including the company name and the director.

At the time, I grabbed data for a handful of companies, including Serco (you have heard of Serco, haven’t you…?!) and… G4S. You can find the data here: OpenCorporates data for G4S on Scraperwiki. The two relevant tables are companydetails_g4s and directors_g4s. The Scraperwiki view here can be used to generate a GEXF graph that describes the bipartite netwrok of G4S companies and the directors who have had dealings with them. A key argument is used to grab the data from the required directors table on the other scraper. So to graph the director dealings capture by the table directors_g4s we use the URL https://views.scraperwiki.com/run/tesco_sprawl_demo_graph/?key=g4s. If you save this file with a .gexf suffix you can load it into graph visualisation tool Gephi and generate something like this:

What this shows is company and director nodes that map out part of the G4S corporate sprawl. The nodes are sized according to a statistical measure calculated across the network known as eigenvector centrality, which is related to the number of directors associated with the company, as well as the number of companies they are associated with. For this graph, I’m not sure how informative the measure actually is…

Note that are several other caveats that should be attached to this sketch (and it is just that: a sketch). Firstly, not all G4S group companies are necessarily included within it (there is still work to be done in providing reliable ways of identifying all of the entities that we might thing of as being meaningful corporate group members). Secondly, the sketch may show companies that no longer exist (for example, it may include companies that have been wound up). Thirdly, only recent directors’ dealings are included, and then, not all of them, (appointments and terminations are included, but other dealings that are: a) a matter of public record, and b) available as open data) are not necessarily in the database view I scraped from OpenCorporates (which itself may be gappy).

On my to do list is capture the date of appointment/termination of a director so that we can see an animated/timeslider view over the directorial network that holds the sprawl together at any particular time.

Caveats aside, what the sketch does show is that there are or have been a wide variety of companies that we might think of as G4S, although each of them may actually provide very different services. There may also be names you recognise amongst the directors…

As well as the corporate sprawl, I have, in the past, has a quick peak at some of the monies G4S has received from public bodies (you might recall that public bodies, depending on what they are, must disclose spend over £500 or £25k). So for example, in Sketching Substantial Council Spending Flows to Serco Using OpenlyLocal Aggregated Spending Data, I also did a quick sketch of how council spending flowed to G4S companies:

If you’re interested in the data, the scraper is here and a breakdown of spend items by local councils as reported by OpenlyLocal (some time ago, it has to be admitted…) is here (so for example, Milton Keynes’ Children and Young People’s Service spend a fair bit with G4S’ fluffy bunny outfit, G4S Care & Justice Services(UK) Ltd).

A quick trawl for G4S on OpenSpending didn’t turn up anything, but never fear, UK gov departements are happily spending with G4S companies. So for example, Home Office spend with G4S CARE & JUSTICE SERVICES (UK) LTD in May 2012 was getting on for quarter of a million via the UK Border Agency on “Refugee repatriation.serv”; or how about this: using the OKFN recline.js explorer tool, we can grab the MoJ (ARAMIS) April 2012 Spending over £25k data and view the payments to G4S INTEGRATED SERVICES (UK) LTD:

Police authorities also divert public funds into G4S coffers. So for example, the Hampshire Police Authority (Scraperwiki data) regularly bung G4s Forensic & Medical Services Ltd the odd 180 grand, sometimes twice within the same month (at least, according to data I scraped… so treat this with caution…! Remember – all I do are sketches that are intended to act as a starting point for further investigation…)

So… that’s a quick tour of some of the open data that’s out there that we can use to shine a torch on elements of the corporate sprawl that is G4S. Hmm… thinks… presumably the complicated corporate structure can also act as a firewall to limit any liabilities arising from the oops around the Olympics contract… (which is possibly only one of many contracts to a variety of G4S companies from different public bodies incurring direct Olympics related expenditure? Anyone care to investigate…?!;-)

PS As a precursor to a little more digging around this, I thought I’d check out the G4S PLC annual report for 2011. This included a mention of major shareholder holdings identified via DTR 5 declarations. It strikes me that a map of the extent to which major investors own significant chunks of large companies might be interesting? The question then is: is there an aggregation of DTR5 declarations anywhere, or would it have to be cobbled together by looking at each of the annual reports of the FTSE 100 companies, for example? Which would be a pain…

PPS WHilst looking through the report, it also struck me that on transparency grounds, a statement of the total revenues received from invoices settled by the public sector might be interesting? This would give a number to aim for when aggregating payments to a particular company or corporate group from separate public sector spending data releases…;-)

Author: Tony Hirst

I'm a Senior Lecturer at The Open University, with an interest in #opendata policy and practice, as well as general web tinkering...

8 thoughts on “A Sketch Map of Part of the G4S Corporate Structure and a Light Shone on Some Public Payments to Them”

  1. “Some time ago, I started exploring how we might be able to use directors’ dealings, as collated by OpenCorporates, to sketch out maps of how corporate sprawls are structured.”

    Wouldn’t it have been easier to just go & look at the annual report, which lists all subsidiaries which “in the opinion of the directors, significantly affected the group’s results and net assets during the year” (see p. 122 of the 2011 annual report), or the the Companies’ House annual return, which lists all subsidiaries?

    The annual report by the way identifies a number of joint ventures which you seem not to have picked up (see page 121).

    G4S are clearly a bunch of tossers, but thanks to the body of regulation which governs companies’ reporting requirements there isn’t any great mystery to or difficulty in identifying the companies which make up a corporate group entity.

    I’m really not sure what point is being made here.

    1. @oliver The point of the post is largely to demonstrate a method using free tools and open data. It also provides a tool for generating views over OpenCorporates data that may be useful in respect of cleaning or improving the quality of data in OpenCorporates, as well as suggesting possible omissions. It also acts as a stepping stone demonstration towards other things as more of the director-company graph is opened up and made available (for free..) in a machine accessible way. For example, if I could interrogate OpenCorporates by director, or address, or geofenced areas, I might be able to identify companies that are in close proximity by virtue having one or more shared directors, or registered offices within 50m of each other.
      On the to do list is the ability to use more than one search term to identify target companies and add the data to the same table, or as a way of trying to find director mediated links between corporate sprawls, which would also serve to generalise the technique a little more.
      In the meantime, ropey as it is, I can generate these sort of sketches using this approach in not much time at all…

      PS erm, how come you didn’t post a link to a more comprehensive depiction of the directed mediated links between G4S companies? As you suggested, the data is there… If you can let me have the data as a graph file, I’ll try to do a comparison with mine to see what sort of coverage it offers, and where the gaps are…

    2. Could you define “corporate sprawl”? Perhaps I’m missing something, but I don’t see what’s surprising about the fact that G4S has a large number of subsidiaries. For one thing, it’s a multinational company; many of those subsidiaries exist so that it can do business in other countries.

      “There is still work to be done in providing reliable ways of identifying all of the entities that we might thing of as being meaningful corporate group members.”

      There really isn’t. The work’s been done for you by G4S, and it’s in their Companies’ House annual return. Companies which fail to provide this data on an annual basis are breaking the law.

      I didn’t link to the data I cited as a graph file because it’s not in graph file form. It’s a list of subsidiaries. If you burrow through the Companies’ House return you’ll get info about who sits on each subsidiary’s board, and who (if any) the minority external shareholders are. If you want to convey something useful and intelligible about the group, that’s the kind of info that’s best set out in tabular form; directors in rows, subsidiaries in columns. A simple count of the number of directors on any one subsidiary’s board will give you some crude information about its relative importance to the group (more directors = more important subsidiary), and a count of the number of boards a director sits on will likewise tell you something about that director’s actual power or importance within the group (more boards = more important director). Ownership structures are best shown with a family tree type representation. A swirly hard-to-read picture conveys almost no information, but does rather seem to send the message “OMG conspiracies!!!”

      Trying to map spending by councils with G4S by drawing lines between two columns of data really isn’t a useful way of conveying information. What’s wrong with a grid? – rows for councils, columns for each G4S subsidiary, with each cell containing the amount spent by a given council with a given subsidiary. Fill that in, and you’ve got some useful info presented in an intelligible fashion. And once you’ve done that spade work, then you could think about graphical representations to make the info more digestible–band the spending, and then represent by colour-scaling maybe?

      1. @oliver I was using ‘corporate’ sprawl as a phrase that is evocative of, as you say, the number of subsidiary companies that make up a corporate group and the mess of possible relations between them. As you suggest, these may be an accounting convenience to manage the (financial) interests in different countries, or different operational areas. Or they may be there to facilitate tax efficient accounting practices; or introduce traps against transparency (such as routing accounts through the Cayman Islands). From a comprehensive analysis of director make up of the different companies, or the progression of a director through different companies, we might be able to learn something about the strategic make up of the company.

        “The work’s been done for you”. I’m trying to build tools that explore the ways in which we can make use of freely available, open data that’s out there that’s being expressed through machine readable APIs, in part as a way of trying to identify where gaps or errors may be in the public data through trying to use it. As to “meaningful corporate groupings”: for us conspiracy theorists (;-) as far as G4S goes, if three of the directors of G4S are directors of a company (that maybe incidentally provides services on behalf of G4S, but is not part of the G4S group), would a look at the G4S report mention that company? I’d say that might be an interesting to spot, and maybe use as the basis for a further bit of digging (which of course may turn out not to be that interesting). As the data gets better linked, it should be easy to spider things like director-company graphs (finding directors associated with companies, then companies associated with directors, and so on). I’m just exploring bits of tooling around the edges of that, as well as demonstrating that it’s not necessarily possible to do that at that moment because the data isn’t openly available via machine readable APIs (or maybe it is – is a complete list of directors available under public open license anywhere?)

        Regarding different ways of displaying co-director or director-company relations, yes, there are probably dozens of ways of visualising it to pull out different perspectives (I can think of several..). The data’s on Scraperwiki – feel free to try some of them out and post a link back:-) I do a lot of my learning from others, as well as by trying things out and sharing the fragments… (Just bear in mind that actually the data is a bit all over the place – this is all still very much at the level of sketching. In particular, you might want to take care about appointment and termination states, which I need to figure into a streaming graph representation.)

        As to corporate structure – yes, I’m sure a family tree might be interesting. If the data was published in a semantic form (companyX isSubdiaryOf companyY) it’d be 5 minute trivial to write a script a to plot the tree of any corporate group. But at the moment, we make do with what we have… By the by, from a governance point of view, don’t you think it might be interesting to see how companies relate by virtue of shared directors? (Similarly within an institution – a committee structure may suggest one formal reporting pathway, but committee membership may reveal a different communications structure. As might an analysis of corporate email correspondence.) Of course, it might equally not be very interesting. But the automated route makes it 5 minutes easy to sketch the graph, whereas trawling Companies House (at a financial cost, as well as in terms of time), getting the data out of PDFs or unstructured docs into a data form etc etc is a pain. And not something I’m in… I’m looking to build tools that help demonstrate how to get at and start to work with data that’s out there… (how many times…….?!?!?!;-)

        As to “drawing lines” – are you referring to the Sankey diagram? That fell out of a demo of how to use the d3js widget (and I was contriving data sets that at a push could be forced into that sort of visualisation). As it is, I think it’s a reasonable view for flagging up issues with the OpenCorporates data and provides a quick and dirty tool that might support a bit of data cleaning activity. Again, it’s not specific to G4S – it was originally just a quick hack that I thought might have some utility, as well as being a demonstration of how to glue various tools and datasets together… And it’s in this post because I was just using G4s as a tag to pull various tools and techniques that I’ve blogged before, and that *I* know can be used as part of a mosaic approach though that fact might have escaped the attention of others, into the same place…

        PS re OMG – maybe that has some function too? How many people know much if anything about the structure of large corporations and why they often are so complex? Pretty much everything I post here is posted with exploratory, rather than explanatory, intent. I use visualisations as part of an ongoing process, as tools for helping me ask the next question (that looks odd? Why’s that? Does it mean something or is it nothing?), rarely as a presentation graphic. Which is one reason why I tend not to post images that other folk might be tempted to reuse. Because they’re only snapshots, taken at a momentary point in time, of a conversation I was having with the data…

  2. Bl**dy hell; I actually gasped a bit when I saw the graph. That puts a context on the G4S thing that is entirely missing from the various traditional / mass media news coverage. Great work.

Comments are closed.

%d bloggers like this: