This post sets out one way to analyse evidence in cartel cases. It is probably of most interest to people in competition authorities but may also be of interest to anyone who has to put a complex story together from many pieces of evidence. There are certainly ways this can be improved, and completely different – and likely better – ways to do this, but this is the best I know so far.

This approach works well for three distinct steps of handling a cartel case:

  • the investigative stage where you need to analyse, summarise and further investigate the evidence;
  • the review stage where you need to look at the evidence, check for consistency, gaps so on; and
  • the writing up stage where you need to document the cartel.

These look broadly chronological, but in practice overlap heavily as you need to keep iterating over the different steps. Having a system that allows you to move easily from one step to another therefore makes handling the case easier.

An aspect of analysing evidence in cartel cases which is more complex than most other antitrust cases is that you need to see what it says about both the cartel as a whole, and each undertaking’s participation in the cartel. Although we in practice refer to cartel cases as, for example, the Power Cables decision or the Libor decision, each is in fact a bundle of decisions addressed to individual legal entities with adequate proof set out for the cartel as a whole, and the particular legal entity addressed by each individual decision.

This makes evidence analysis and subsequent drafting complex. Legal entities may start and end their participation on different dates, discuss one product but not another, or one customer but not another, participate in some conduct but not all, and so on. The more companies are implicated and the more documents that need to be analysed, the more complex this becomes.

The disadvantages of a word processor

Fundamentally a word processor is not well suited for this task, and using one typically means that you are using nothing more than an expensive typewriter: multiple text documents – perhaps one per undertaking, each then needing to be cross-referenced with the others – quickly become unwieldy; a single document that tries to cover everything becomes complex and difficult to maintain.

And once you’ve created one or more text documents, then you can’t easily ask your word processor to “show me the evidence relating to Company A”, or “show me discussions about Product X”. Nor can a word processor easily sort sentences or paragraphs into chronological order, so you have to do that yourself as you are going along.

Text documents are also slow to create, even if you can type quickly: a typical summary of an allegedly anti-competitive contact may read, “Company A met Company B on 1/6/2010 and discussed future prices, their response to Customer 1’s request for bids, and general market trends.” That’s already a lot of typing. Add more companies? And an explicit reference to a particular product if this is a multi-product cartel? And the geographic area? Repeat for a thousand pieces of evidence? Two thousand?

Spreadsheets as a step up from a word processor

Spreadsheets help to structure the evidence, and although I’ve seen them used in the past, I haven’t seen anything much more than using a date column to sort the evidence into chronological order. Spreadsheets can do a lot more than that. Here’s an evidence table created in a spreadsheet for you to download that I’ve been putting together with the help of some very talented colleagues in DG Competition. (All of the data is fake – the dates were randomly generated and then adjusted to make a more coherent story.)

In the rest of this post I’ll explain a little about this spreadsheet, and the advantages of using a spreadsheet for the analysis, review and writing up of cartel cases.

I don’t want to imply that a spreadsheet is the perfect solution. There are other approaches. A relational database, for example, can keep track of many to many relationships better than a spreadsheet – for example multiple individuals in relation to multiple companies in relation to multiple anti-competitive contacts. But using a spreadsheet is quick and easy, and by adding, deleting or modifying columns in the spreadsheet, it is easily adaptable as the case develops. It is not necessarily the most sophisticated approach, but the balance of sophistication, ease of use, and simplicity of output works well in practice. That said, if you know of better approaches, please either comment below or get in touch by email.

An example of a spreadsheet evidence table

The spreadsheet is filled with data using an example of a set of multilateral contacts. It looks fairly  elaborate, but in practice it is simple and adaptable, and this is a simplified version of one we are using in a real case. The key to designing the spreadsheet is to think about how you might want to present the information in a Statement of Objections, and then to capture relevant information accordingly often in a dedicated column of the spreadsheet. In more technical terms, as you analyse each piece of evidence you are creating metadata for it, and ensuring there’s a permanent link between the evidence and that metadata.

Broadly speaking every parameter of the evidence needed to understand the case gets a dedicated column: for example a column for each company, a column for each type of conduct, a column for the date of the incident and so on. A spreadsheet is almost infinitely expandable so can capture as many parameters – as much metadata – as you need.

Some columns will be obvious – a column for the date of the incident, a column for every company involved, a column for each type of potentially anti-competitive conduct and so on.  Some columns will be less obvious: there may be more administrative columns that are not directly related to understanding the evidence, but rather relate to more procedural issues  – for example capturing the source and the date of submission makes it easier to identify the documents relevant to assessing the significant added value of leniency applications.

This doesn’t mean that you need to understand the case before starting to analyse it. Any system for analysing the evidence must be adaptable as your understanding of the case develops. Merging, splitting or adding new columns as you go along is straightforward.

At the start of a complex case, you may not know exactly what you want to capture.  Analysis is an iterative process – you capture the information you think relevant, review what you have captured and see if a structure is emerging. That may highlight additional information that you need, in which case you’ll need to go back over the evidence you’ve already included and add it, or go back to the parties with questions if it’s not already on the file. (You probably won’t need to re-analyse the evidence you’ve already excluded as if it showed anti-competitive conduct you would have already tried to note that in the existing structure.)

More complex cases will need more complex approaches than that set out in the example, and I’ve added some additional – blank – columns to give an idea of other data that could be captured. If the cartel seems to be organised around particular customers, then columns for the customer name and / or the request for quotations (RFQ)  may be useful. If the individuals participating are important, then the names of the individuals could be put into the company name column (in place of the simple  “x”) that is used here, or rows could be duplicated for each separate individual involved.  Columns could be added if the cartel was organised geographically, for types of meetings (multilateral, bilateral etc), and so on. And of course, you could add a column to highlight exculpatory information.

Analysing the Evidence

One practical advantage of this approach is that it’s a lot less typing. That may not sound significant, but the sentence, “Company A, Company B and Company C met on 1/6/2010 and discussed future prices, their response to Customer 1’s request for bids and general market trends.” – can be reduced to “x  x x 1/6/2010 x x x” plus a few tabs or mouse-clicks to move from one column to the next. If you’re processing evidence in relation to 1000 anti-competitive contacts, each involving multiple combinations of companies, then just typing “x” in the appropriate columns saves a lot of work compared to typing out the company name and the type of infringing conduct. This means that evidence analysis takes less time, and taking less time makes for faster cases – and happier case handlers.

Quicker evidence analysis has other advantages.

If a company submits evidence, and the case team can quickly analyse it, understand it, and compare it to pre-existing evidence then they can also quickly go back to the submitting companies with requests for clarifications, or go back to the other parties asking questions about new, specific, incidents.

Because the columns are structured, gaps are highly visible.  For example if the product being discussed is unclear then there will be blank product columns. If it’s not clear which companies were present at a particular meeting, then there will be blank company columns. This makes it easy to see where more information is needed, and you can quickly go back to the submitting companies to get it. Gaps are much less obvious in text documents.

There are also benefits, as we shall see in a moment, of adding the data in a structured way such that the spreadsheet can “understand” it. For example, always adding dates or document ID numbers in the right format means that the spreadsheet can sort and filter the evidence based on those dates or ID numbers. More on this below.

But it is also possible to highlight data in a non-structured way, that the spreadsheet does not necessarily “understand”. So, in this example, I’ve coloured the column for the immunity applicant blue and the leniency applicants yellow. It doesn’t particularly add to the understanding of the case, but when reviewing the evidence it’s helpful to understand who has supplied what.  I’ve also highlighted a few individual cells in red: this could represent, for example, a second reviewer who has looked at the evidence and does not understand or has doubts about these highlighted points.

Reviewing the Evidence

The advantages of this approach are not just faster input and faster questions to the parties. A spreadsheet gives a visual overview of the evidence. Simply by looking at the x’s in each column, you can get some sense of which companies, products or conduct were heavily implicated, and which were peripheral.

Going beyond getting a quicker overview of the evidence, if you put the information into the spreadsheet in this structured manner, then you can get the software to do a lot of work for you.

The spreadsheet can sort and filter the data:

I said at the start that an authority needs to document the story of the cartel, and the story of each company’s involvement in the cartel. With a spreadsheet, if you want to see the evidence that implicates Company B, then you can simply filter the spreadsheet to display only those rows that have an “x” in the Company B column . So starting from seeing the whole of the evidence for all companies, you now see that subset of the evidence that shows just Company B’s involvement in the cartel.

If you then sort on the date of incident column you can see a chronology of Company B’s involvement. Scan down the incident column and you can easily see if the chronology is consistent, or if there are gaps that might be a problem for demonstrating a continuous infringement.  Then you can look at the first row – that’s the start date of Company B’s involvement – and the last – for the end date. If in the future – say after having read Company B’s Reply to the Statement of Objections – you decide that the evidence for the start date is not as strong as you originally thought, then you can delete that piece of evidence from the table and immediately see the new start date.

If you need – in EU competition law terms – to assess whether a leniency submission provided on a particular date gave significant added value, the table can help with that. Go to the submission date column and filter out – exclude – evidence gathered after the date of a particular submission. Then sort on the incident date. You can then see the evidence submitted on that date compared to what was already on the file. It doesn’t replace the need to look at the evidence and analyse its value, but it shows you where to look.

How about if you want to assess the leniency applications and see if any deserve partial immunity (because – in terms of European Commission policy – they extended the knowledge of the duration of the cartel as a whole compared to what was already on file)? You can create a separate worksheet that summarises the start and end dates demonstrated by the evidence on file at any given date. Again, you need to look at the evidence to assess it fully, but you can at least automatically see the relevant dates.

And then the writing up…

Ultimately you need to write up the analysis – in the EU, in a Statement of Objections. A spreadsheet can make that quicker as well.

By filtering and sorting the spreadsheet, you can organise the evidence to allow you to write up a factual description. So from a row of the spreadsheet, reading across the columns, you can see and then type out the following information.

On 19/4/05, companies A, B, C, E, H and J met to discuss customer allocation. [Footnote: ID 1, page 1, Meeting Minute of 19/04/05, submitted by Company C on 4/1/14]

Repeat that for every row of the spreadsheet and you have a complete chronology of the cartel.You can then also filter on an individual company, and repeat the exercise to create a company-specific chronology.

You can adapt this approach to however you want to structure the presentation of the evidence. If you want to organise it by customer, and then chronologically, you can sort on those columns; the same if you want to organise the evidence by RFQ, or by any other parameter.

But however you choose to organise the evidence, these are very repetitive tasks and, given that the basic analysis is now done, the work is essentially just repetitive typing. Computers are very, very good at repetitive tasks – they are quick and, assuming the source data is correct, they are accurate. So why not let the computer do the work rather than the case handler? Most spreadsheets have some kind of scripting language, such as Visual Basic for Applications if you use Excel or AppleScript if you use Numbers. You can use these scripting languages to iterate through each row of the spreadsheet, and create a text document that contains all of the information contained in the spreadsheet.

Essentially you ask the spreadsheet to sort the evidence however you wish, and then  read the first  row of the table and create the relevant text as in the sentence and footnote above. Then move to the next row. Similarly you can ask the spreadsheet to filter the evidence that’s relevant only to Company A, and go through the same process. A text document with a fully documented chronology of 1000 anti-competitive contacts for the cartel, and a subset of that chronology for each company’s participation in the cartel, can be created in a few minutes.

That’s rather quicker, and a lot more accurate, than typing it up manually.

That is a long, long way from the end of the work. Examples and explanations are still needed. But it’s  a good start – a structured, accurate foundation upon which you can build.