A regulator that errs in either direction – by approving harmful transactions or by blocking benign deals – harms competition and consumers’ interests. However, decisions need to be thorough without being burdensome. Extensive enquiries come at a cost, potentially delaying beneficial agreements or causing them to be abandoned.
So, how effective is the European Commission (EC)’s review process? Experience tells us a great deal, but it can over-emphasise the importance of specific cases. What are the facts for the process as a whole? To answer this question, the Data Science team at Compass Lexecon built a unique dataset of EC merger decisions, which analyses 7,823 cases between 1990 and 2020,3 and uses machine learning to read and interpret almost 100,000 pages of decisions. The following analysis provides some preliminary insight into how the EC merger review process is working.

THE VAST MAJORITY OF CASES ARE APPROVED IN PHASE I
91% of all cases are approved in Phase I without commitments – i.e. they receive an Art. 6(1)(b) Decision. Relatively few cases require intervention or a detailed Phase II review – the EC approves a further 4% of cases in Phase I with commitments and only around 4% of all cases continue with an in-depth Phase II investigation (Figure 1 - https://www.compasslexecon.com/ecmergerdecisions/)
These outcomes may not be surprising, but they are important. Most potential mergers are non-problematic and simple. Understandably, the accuracy of detailed investigations receives attention, but most parties are affected by whether the EC reviews relatively straight-forward transactions quickly and efficiently.

THE EC INCREASINGLY USES ITS SIMPLIFIED PROCEDURE
Efficient decisions matter because the EC reviews many cases. From the post-recession low in 2009, the number of cases being reviewed by the EC has increased by over 50% – from 257 to 394 in 2018 (Figure 2).
An efficient process requires triage, not balance or compromise. It is therefore a positive sign that the EC processes only 4% of cases using its most in-depth investigation in Phase II and, since 2000, the EC has used its ‘Simplified Procedure’ to triage Phase I cases. The EC uses its ‘Simplified Procedure’ when the entities proposing to merge have circumstances that are unlikely to threaten competition in any case – either because the merging parties do not compete in overlapping markets, or the combined entity would have revenue, assets, or market shares below specified thresholds.4 The procedure accounts for a high proportion of cases – nearly three quarters of all cases since
2015 – and Figure 2 shows that proportion has been growing. In particular, the number and proportion of mergers assessed using the Simplified Procedure has grown significantly since 2014, when the EC relaxed its thresholds.
These trends suggest the merger review process is increasingly efficient. It seems the EC can focus its resources where detailed investigations are needed and rely on a leaner process where matters are straightforward.
Moreover, it is possible that a more efficient process has encouraged more parties in straightforward circumstances to propose beneficial mergers – when they might otherwise have been deterred by transactions costs imposed by a less efficient review process.4

PHASE II DECISIONS ARE TAKING LONGER
Has triage made decision-making quicker? On average, yes. The EC makes Phase I decisions more quickly than it used to, reducing from around 5 weeks (the target time) to around 4.5 weeks. However, that average masks two stories.
The simplified process is quick. Introducing the Simplified Procedure, and increasing its use, has reduced the average time the EC takes to process Phase I cases.
Figure 2: Cases over time by notification date (quarterly)5 - https://www.compasslexecon.com/ecmergerdecisions/
On the other hand, complex investigations are getting longer. After notification, non-simplified Phase I cases remain at the target time, five weeks – although this obscures the full picture, as the EC can engage with merging parties prior to notification, before the clock starts. Phase II investigations take considerably longer than they used to; the time that elapses between notification and a Phase II decision has increased from
around 25 weeks in the mid-2000s to over 30 weeks in the last few years (Figure 3).
Figure 3: Duration in weeks between notification and final decision6 -
https://www.compasslexecon.com/ecmergerdecisions/
  
Part of the reason for this is the increased extent to which the EC pauses its complex investigations, using its power to “stop the clock”. Between 2015 and 2019, the EC paused 60% of its Phase II investigations. Whereas, it paused 50% in the previous five-year period, and only 17% between 2005 and 2009. Consequently, the time the EC spends ‘on the clock’ is becoming less indicative of the real time that elapses for merging parties between the beginning and end of a review. Although the EC paused a higher proportion of its Phase II investigations, it did not pause those investigations more frequently. Of the cases where the EC ‘stopped the clock’ between 2010 and 2019, on average cases were stopped around 1.5 times. There was some slight variation across cases and years: it paused 54% of stopped cases once, 35% twice, and 11% three times.

There are many potential reasons why the EC has taken longer to investigate complex mergers. Some are substantive – i.e. it is possible that more cases have complicated circumstances than they used to. For example, they may involve more complex products or affect a larger number of markets. Some reasons could be procedural – i.e. the way those cases are selected and assessed is more complicated. For example, the EC
may face a higher standard of proof and greater judicial scrutiny; or it might acquire, process, and consider more information than it used to – either due to technology making more evidence available, or a shift in preferences within the EC.
In any case, these trends serve as a warning to a firm facing a potentially complex merger review that it may take longer and cost more than expected.
BOTH PHASE I AND II DECISIONS ARE BECOMING MORE DETAILED
Given that investigations take longer, we might expect the decisions they inform to be more detailed. They are. Decisions are getting lengthier. Phase I decisions have increased from around 10 pages in the early 2000s to around 35 pages in the last few years. Phase II decisions have increased by even more, from around 50 pages
in the early 2000s to around 250 pages in recent years (Figure 4).
Figure 4: Decision length, excluding Simplified Procedure - https://www.compasslexecon.com/ecmergerdecisions/
In general, this trend should indicate that the decision-making process is becoming more thorough, although that is not necessarily so, as greater detail can obscure the pertinent issues and facts. Similarly, longer decisions should make decision-making more transparent, allowing stakeholders to better understand judgements and
react accordingly.

WHAT EVIDENCE BASES INFORM DECISIONS?
Partly, longer decisions reflect changes in the evidence bases that inform them. What can we tell about the information the EC considers in its decisions?

To answer that question, we used automated text analysis, to analyse and categorise the types of evidence each decision referred to. First, we parsed and cleaned the text in each decision and used an algorithm to identify keywords that related to the evidence sources we were interested in. We grouped specific sources of evidence and analysis into broader categories. For instance, we allocated analyses such as the Upward Pricing
Pressure (UPP) test and econometric analyses to the “Economic Evidence” category. Second, the algorithm “adjusted” the keywords counts – using machine learning techniques to ensure that references were not double-counted as a result of differences in drafting style – to determine the number of times each decision referred to each evidence category.

Decisions refer to all types of evidence more often than they used to (Figure 5). We should expect this, given that decisions are lengthier than they once were. In part, evidence is more available than it was, enabling authorities to retrieve, analyse, and consider more information. Also, legal and evidential burdens also develop over time, as authorities must demonstrate that they have considered various factors and positions put
forward by parties.
Figure 5: Average number of references to different types of evidence used8 - https://www.compasslexecon.com/ecmergerdecisions/
The EC typically relies on several sources of evidence. On average, decisions refer most frequently to information gathered through the EC’s own market investigation: written Requests for Information to customers, suppliers, competitors and other relevant parties.7 This reflects how common the process is and, as it applies in all investigations, references have grown broadly in line with decision lengths.

Similarly, all decisions assess and refer to information submitted by the Parties in the Form CO, which in practice, focusses on market shares. References to market shares have plateaued even though the decisions referring to them have continued to get longer, which although essential, implies a ceiling on the issues that
simple inspection of market shares can speak to.

Prior to 2010, decisions rarely referred to supplementary evidence, but since then it has become more common. For instance, decisions increasingly refer to the internal documents – variants of internal email exchanges, correspondence, and estimates – the EC retrieves from the Parties. Decisions also increasingly referred to economic evidence from 2010, but references declined in the second half of the decade. It is not clear why. Economic evidence provides additional insight that can either corroborate or challenge findings from market investigations and, in complex cases, economic insight is essential to understand underlying dynamics and potential impact of a proposed merger. It is unlikely that recent cases were simpler than those between 2010 and 2015. Although it could to some extent reflect changes in conventions in referring to economic analysis in decisions, it nonetheless demonstrates a shift in the types of evidence that the EC considers.

IS POLICY TIGHTENING AS A RESULT?
So, investigations take longer, decisions are more detailed, and they refer to more sources of evidence. A natural question to consider is whether those trends affect outcomes – i.e. have the EC’s process changes tightened merger control?
No, merger control does not appear to have tightened. Figure 6 shows that, since 2000, the EC has blocked, or approved with commitments, a broadly constant proportion of the cases it assesses. It has increasingly blocked or approved only with commitments more of the cases that it reviews in detail, but that trend reflects the fact that it reviews fewer cases in detail than it used to, as it siphons a growing proportion to its Simplified Procedure.
Figure 6: Share of cases blocked or approved with commitments9 - https://www.compasslexecon.com/ecmergerdecisions/
BRINGING CLARITY AND TRANSPARENCY
Analysing our dataset of EC merger decisions helps unpick how the merger review process works. It reveals some positive messages: the EC has streamlined its decision making effectively for simple non-problematic cases. This matters. These make up the vast majority of cases, where speed and low transaction costs are critical. Although we must be careful about inferring causation, it is possible that a leaner process deserves
some credit for the increase in total cases – as unproblematic mergers faced fewer procedural barriers.
For potentially problematic cases, the EC’s review process appears to be getting more intensive and complex: timelines are longer, producing more detailed decisions, after considering a greater amount of evidence from a broader range of sources. Whether this reflects more thorough processes or more complex circumstances, it is likely that increasingly detailed reviews are here to stay.
For that reason, rigorous analysis and transparency will be crucial. Growing piles of information are available to competition authorities and merging parties. That availability can bewilder or mislead, seemingly confirming almost anything one expects to find. However, with skills to interpret it, the increasing availability of data is an opportunity to cut through complexity, to reveal the facts that we might otherwise fail to see.
The opinions presented here are those of the authors, Rashid Muhamedrahimov and Andrew Tuffin, who are employees of Compass Lexecon, and do not necessarily reflect the views of Compass Lexecon LLC or its management, its subsidiaries, its affiliates, its other professionals or clients.