#20 - Carlos Arrebola (University of Cambridge)

Do Advocate-General opinions influence decisions by the Court of Justice of the EU?


Interview with Carlos Arrebola, Ph.D candidate at the University of Cambridge and one of the authors of a paper on the potential impact Advocate General's opinions have on decisions taken by the Court of Justice of the European Union.  For the final episode of series 1 of the Public Procurement Podcast the interview duties were mostly left in the capable hands of Albert Sanchez-Graells (University of Bristol).



Pedro Telles (PT): For the 20th and last programme of the series we are doing something different. I have with me, not one, but two persons in the virtual studio, Carlos Arrebola a PhD candidate at University of Cambridge and Albert Sanchez-Graells from the University of Bristol. The purpose of our talk today is to probe a little bit more about Carlos’ research into the influence of the Advocate General’s opinion in subsequent decisions of the Court of Justice of the European Union. So Carlos and some of his colleagues put up a very interesting paper on SSRN and Albert decided to critique it and provide some very helpful feedback. And that is mainly the reason why I’m bringing Albert back because I want him to take the reins of the conversation today and effectively redo a little bit of his work of his critique that he did for the blog http://www.howtocrackanut.com.

Carlos and his colleagues replied to Albert's critique, by guest posting in Albert’s blog and as such this show is going to be slightly different from usual with me hosting the call more than anything else and Carlos and Albert hammering out their differences on a very reasonable and sensible conversation with a cup of tea or a cup of coffee as they prefer. Welcome to today’s show, can you start by introducing yourselves so that we can match the voices to the names?

Carlos Arrebola (CA): Oh thank you very much Pedro, I’m Carlos Arrebola.

Albert Sanchez-Graells (ASG): And I’m Albert Sanchez-Graells, it’s a pleasure to be on your show once again.


PT: Yeah, I couldn’t finish without you Albert, I started with you, I had to finish with you as well, so it’s great that we found out a way to do it.

ASG: That’s so kind Pedro.


PT: I’m going to kick off the interview, I’m going to kick off the talk with a low ball question for Carlos, Carlos could you summarise your research and what were the main findings please.

CA: Yes, so the research project that we devised was, as you said, about the influence of the Advocate General. It’s very interesting and thank you for giving me the opportunity to sort of give a bit more background information. Because how this came about was by Julia, who is another PhD candidate at the University of Cambridge, and me, we were talking about research methodologies. We were asking ourselves “what is it to do legal research?”, and comparing it with other sciences. So out of an example it came out this claim that people say all the time that is, well you know, “the Court follows the Advocate General” or “the Advocate General influences the Court”. And we are thinking, well, actually has that been proven or is there any evidence for it? So that’s how it came about. Then, after doing a bit of a literature review we saw that many people do some sort of statistics about it, but there was no econometric study where different variables are accounted to determine whether that’s accurate or not. So we decided to do it. And we called another friend of mine, who was doing MPhil in economics at the time at Cambridge as well, and asked him for some help in developing this model and testing whether the Advocate General actually was causing, as in whether we could establish causality in its opinions towards the decisions of the Court. So we did this. We put together a database of about 20 years of action for annulment and we came out with a response that was basically that yes, there is an influence. And the way we phrased the influence is that if the Advocate General asked the Court to annul an act, the Court will be 67% more likely to annul such act. To be honest, we were expecting the results to show that the Advocate General was less influential than we thought, than those claims about the Advocate General suggest. So that was a bit surprising. But, in a nutshell, to quote Albert’s blog, that’s what our research project was saying.


PT: Okay, Albert you read the paper and what were your main ideas or main thinking lines about it?

ASG: So I saw the paper on SSRN and I thought it was really interesting as a project, I think that definitely there is need to try to bring more econometric analysis to the analysis of case law to see whether we can identify patterns, but it’s been very interesting, it’s also extremely complicated to do so I tried to brush up my econometrics from when I studied the undergrad, which is certainly not yesterday and I looked at the model, you know, with the difficulty of trying to understand the model I just, I thought that there was some basic underlying assumption that I was not really buying it on the first reading so I started reading about more detail and I think that the main issues that we need to control in those models for noise or for factors that would explain what we’re trying to measure for different reasons and that’s a perspective from which I thought that some of the points that Carlos and his colleagues raised could be explored differently, particularly in terms of the tests that they run but of course that’s something quite in the detail so maybe it’s better for Carlos to go on and explain their claim and how they got their specifics and then maybe I can just pitch in when needed.

CA: So Albert is definitely right. That was the very first problem. When I approached my economics colleague I said “How can we do this?” And obviously you have to put together a database of different variables, different factors that you are going to account for that can explain why the Court decides in a particular way. To put it simply, we said we’re going to see if the Court decides to annul or partially annul or not. And the reason for why the Court or a particular judge may decide to annul, or not, are not necessarily related to the Advocate General at all. There can be many reasons, and one of them is because the law was very clear, it was ultra vires or whatever it might be. So we had to, as I said, account for all those other things that had nothing to do with the Advocate General, and we put together different variables, the ones that we thought were simple enough or relevant enough to be put into our database and so we chose something like who was the claimant? Was the claimant, an institution? Was the claimant a member state? etc, etc. Also, who was the Advocate General? Because we thought that maybe some Advocates General could be more influential than others and, so yeah, we compiled all that information. Of course we left out many variables that affect the decisions but there’s nothing you can really do about it. You can always refine your model, but there will always be something that you leave out. But then, and this was one of the main criticisms by Albert in his post, is that he was saying that, that makes it not reliable. We have to disagree a little bit here. We think that even if you leave out variables that are important or relevant for why the Court decides in a particular way your econometric model can be reliable. And that would be going a bit further into the econometrics. But, basically, it’s because your results are still statistically significant. And that’s what happened in our model. We ran different models and we came out with one that seemed to be, according to several econometric measures, reliable enough as to present the results.


PT: Albert.

ASG: Yeah, no, on that point I guess that that’s a link to the second of the criticisms I raise in the blog was then basically it’s not so much leading variables that could explain the decision of the Court out, it’s more about how the model is constructed in terms of which decisions the Court can adopt and this is a point that Carlos and his colleagues acknowledge in the paper and they say well it’s complex because the outcome in these proceedings basically could be either to dismiss the claim or not admit it and those would be negative results for the claimant or then there could be either partial or full annulment which would be resolved that are positive for the claimant and that’s how they construct their variables. So basically they give zero when the case is either not admitted or dismissed and one when there is partial or total annulment.

They explain quite well why partial and total annulment can be grouped together because basically I think that they’ve run tests and they see that there is no significant difference in the outcome but my main concern is with grouping, lack of admissibility and dismissal on the grounds of substance because basically from a legal point of view, it is not the same to say the case is not receivable by the Court for variable procedural reasons than to say it’s been received, it’s been assessed on the grounds of substance and then we dismiss it. And it’s not only that these are different steps but it’s also that the law that controls these two assessments are very different and I would claim that the rules controlling admissibility are much more limited and I would say are more straightforward because they are about whether there is sufficient reasoning behind the claim, legal standing, timing, so they are easier, they are more accepted by the parties and also I think that usually the filter of admissibility tends to end on the case not receiving the case and then discussing the grounds.

So I would maybe separate those two because I think that maybe otherwise it’s going to skew the results. Of course I would need to rerun a model with these differences and I haven’t had the time, or maybe not even the skill, at the moment to do it but my impression is that those two should not be grouped together and that’s something that Carlos replied on the blog, rebuttal by saying “Well that’s the way we did it”, so I don’t know Carlos, do you have any further thoughts on this?

CA: Yeah, I think it gives me the opportunity now to clarify. I tried to do it in the rebuttal post but it’s a bit complex, so I tried to keep it simple. Basically yes, you are right that we grouped those things together. But our results are not saying anything about inadmissibility or dismissal, because we are only looking at opinions when the Advocate General says or suggests to annul. So we are not saying anything... our claim or our final result are 67% more likely, it is not in the direction of negative opinions, it’s only in the direction of positive opinions. So basically when the Advocate General suggests to annul, how does the Court react to that? So basically that’s why our claim is that it doesn’t make any difference that we group together inadmissibility and dismissal. But there’s something interesting that you said is yeah, there are many other ways of looking at our model and maybe one of those ways, it’s looking at it in the negative like, what would happen if the Advocate General suggested to dismiss a case. Is he followed to the same percentage in that case? But that would require fiddling a lot with our database. Then there’s another problem and this is why we grouped total annulment with partial annulment, is that even we collected 20 years of data, when it comes to cases of partial annulment you might have no more than 20, and I’m not entirely sure whether there are more than 20 or not, but I remember that there were a very limited number. That’s difficult for econometrics because it means that the results are not going to be significant enough so your claims cannot be reliable. So basically that’s why we grouped them together and they were, we are missing some sort of information there. But we thought that going towards the direction of annulment was something that could be grouped together.

ASG: Yeah, no, as I said before I agree with the direction of annulment and particularly if we only have 20 observations or 20 something in 20 years, it’s not going to be significant. But the issue on the model being only one direction is the one that the third point I was trying to raise in the blog was that I think the big advantage that econometric studies of case law have, is they allow for intuitions to be either proven or disproven so in other words whether the results can be understood without deep model and would still make sense or if there is a counter-intuitive point, if they explained the counter-intuitive point very well. At the beginning your research, as you said, is basically about trying to give support to whether the Advocate General actually influences it or do not so the research question is what is the influence of the Advocate General? And I think that the results or the model would be much more credible or would convince the sceptics like myself, if you could make the claim both ways. So if you could say, it is not only that the Advocate General influences a Court when it decides to suggest that something should be annulled but also when it decides that something should not annulled or even when it submits that it should be dismissed. Because in that case you cannot have the counter explanation that the Court of Justice has for instance an activist or intergrationist agenda in which case it is more likely to annul on substance whenever there is any issue concerning restriction of freedoms for instance. So I think that that’s why it’s so difficult to build models, I think that maybe what we could agree is that you have developed half of the model and that for your claim to be solid, you would need to test it the other way round. Also because this is one of the problems with econometrics, we don’t know what’s going to come out. So if you were to say, well the Advocate General influenced the Court which suggests that it should annul but it doesn’t really suggest that it shouldn’t annul then I think that we will be conflating two issues which is the problem in maybe bias of the Court and the position of the Advocate General.


PT: Yeah, can I add something on there which I think would be interesting, which is you just looked at the actions for annulment in one big batch, am I correct Carlos?

CA: Yeah, yeah.


PT: So what I would like to know is if the influence of the Advocate General actually changes depending on the formation of the Court, i.e., the Court can meet in many configurations from a bench of what five Judges, that’s the smallest configuration Albert?

ASG: Yeah, or three.


PT: Or three, so my question would be does the influence of Advocate General actually decrease as the complexity of the case increases and, for example, if you have decisions that are taken by the Grand Chamber of Judges, does he have any influence at all?

CA: Yes, so coming to the last question by Pedro, we did include the composition of the Court. That was one of our variables. And, to be honest I don’t really remember, I’m scrolling through the paper to see if that showed us anything because I don’t want to summarise my own paper incorrectly.


PT: Sure.

CA: But I think we came out with something like the grand chamber was a variable that could explain the decision of the Court to a great degree but I’m not completely sure that it was a very relevant variable.


PT: Having said that I agree with Albert’s critique that your model to be sustainable needs to work the other way around as well.

CA: Yes, absolutely, that’s something we most definitely agree that, well we present a project and the results and obviously those have to be tested again and again to... that’s the basis of science, isn’t it? When you, you make a claim and it has to be an observation that can be done again and again by different tests and that’s when the claim becomes accurate and valid. And yes, as I said, I invite anyone, and I am glad Albert is willing to get the database and try to play with it a little bit and see what comes out. But I invite anyone who is listening to us, if they want the database, they just have to drop us an email and they can carry on with the study.

ASG: Yeah, no, that’s really generous, I definitely plan to do that, I just need to find some time which is probably not soon. But I think that brings me to the only other part that I found that the paper maybe could be strengthened or maybe was making a bit too much of a claim because the issue Carlos is that so you have your results and it was significative and the results need to be challenged, they need to be tested, but then in your discussion in the paper, which is quite interesting of course, you explore issues of judicial independence and reform of the Court of Justice together on the role of the Advocate General and I think that that’s where I am a bit sceptical about using econometrics that are not really rock solid. My impression is that on the basis of your finding, the argument of Judicial independence and reform of the Advocate General are a bit worrying because there is no non-econometric consideration you take into account so you don’t really look at how would the influence of the Advocate General compare with, for instance, the influence of public prosecutors in criminal courts and domestic settings where there is also a similar role on trying to clarify the facts in the law, but you don’t really look at any other international Court to see if amicus briefings play similar roles. So I think that my only issue with your discussion is that maybe it tries to find too much support on your finding for discussions that are really far-fetched and then by reasons of space are quite limited in your paper. So from the methodological point of view I would simply say it’s good to keep the econometric papers in econometrics and then maybe, you know, build longer, more considered papers on implications.


PT: That’s a good idea.

CA: Yeah, absolutely and we were very, or we tried to be very careful when drafting the discussion section and we started it all by saying “If we accept our results in the previous sections, then these are the implications”. I mean it was more, rather than saying, “Well this is what’s happened, and we need judicial reform”, we were saying, “Well if this is true, and we want people to sort of research a bit more and ourselves in the future,, then it has very important implications and that’s why this paper should be taken seriously”. And also there’s another comment, which normally happens in academia. We had two referees when we submitted the paper and we don’t know who they are obviously. But one of them was saying “Oh please be more bold about your claims in the discussion”, and the other one was saying “I don’t buy it”, like why are you saying all this? So it was difficult to strike a balance but definitely our own intention was to be cautious in what we were saying and just pointing out the potential implications rather than saying “Oh this is what should happen”. We don’t ask the Court of Justice to call for a reform tomorrow and look at our paper and say, “Oh actually we have to take the role of the Advocate General”, but, you know, those are things to consider towards the future.


PT: If I can jump in a make a comment there, I am always reminded of a piece of advice I was given by my then teacher supervisor, Professor Sue Arrowsmith that she said to me that far-fetched claims are where the good thesis die in a viva so we may have a very good piece of research and then we get carried away and start making suggestions and assumptions that are yet to be tested and that are not entirely supported by our actual research. And as time goes on I have come to appreciate the comment and the way that she put it, and I’ve seen that in a few vivas, students that were a little bit too ambitious then on the comments in the conclusions or in the extrapolations of their research and they were just ripped to shreds, so to speak, in the viva. So I agree with Albert on this one as well which is it’s probably better to build up the econometrics and then leave it standing as a piece of research on its own and then further down the line take into consideration the wider implications of introducing changes to the Court and including future research that may or may not have similar research methods.

CA: No, yeah, definitely, I also agree and I thank the opportunity to highlight that we tried to do it in the conditional but saying, if we accept the results that we put forward rather than saying “This is what should happen”. But yeah, sometimes it comes across like actually we are suggesting that there should be a judicial reform. So thank you for letting me clarify that.


PT: Okay, Albert, any further questions?

ASG: No, I think that, I mean overall what I think is the only final thoughts I would have is that it... I think it’s commendable to see that young PhD researchers, well first they are interested in topics that are not within the core of a specific thesis because as far as I know Carlos is not researching on this.

CA: No.

ASG: And I’m not very familiar with Julia’s research but I think she’s not either looking at the judicial reform so that’s interesting for PhD is to know that you can have some side projects and that’s going to improve the quality of your thesis as well because it gives you brain space and you know, you’ll get things with different views. And also that they are engaging in inter-disciplinary and collaborative projects which I think is where we are headed. So just to say well done in the setting of the study and in pushing these trends forwards even if we may disagree on the outcome but that’s an ongoing discussion we can keep in future years but it’s something difficult to do and I think it should be appreciated that you are engaging these things, so well done to you and your colleagues.

CA: Thanks, thanks a lot and obviously thank you for engaging with the debate. It obviously, as a young researcher, it comes as a surprise that people are really interested and engaged with your work. So that was a very positive to see.


PT: Very well, I do have one final question, what’s next for you Carlos?

CA: What’s next for me? Well at the moment I’m very, very busy. In a way I’m happy that we submitted this paper and that’s one thing less to do. I’m trying to finish my PhD. I’m writing up. As Albert said, it is nothing to do with the Advocate General or the Court, it’s on the European Economic Constitution and it’s a bit broader and theoretical. And I’m also doing a GDL, a Graduate Diploma in Law, because I’m heading to the city to be a Solicitor, hopefully in competition law. So that’s it. And when I’ve finished that, I’m planning to cycle across Europe from Finland to Portugal. It’s a bit to raise awareness or to ask ourselves a bit more about the European Union identity, which I think is a very important topic right now, in the UK especially. And I’m going to be recording all that in a Youtube channel that I have that is called Cycle across the EU.


PT: It’s not a bad way to do it, I think it’s quite interesting, are you going to go through the UK or are you going to avoid the UK on the trip?

CA: I decided not to go through the UK and not to go through Belgium basically because I already started making a lot of videos to give a bit of context to the trip and all of those videos take place in the UK because it’s my current place of residence. So I thought it is not really necessary to come back to the UK. And also I decided not to go to Belgium because this year I’m going to Belgium twice and that’s also going to be on the Youtube channel. So I thought it would be a bit more interesting to show some other places of the Union, which will allow as well the audience from the UK to see those places.


PT: Okay, I think this is a very fitting way to end not only the programme but also the series. Thank you gentleman for coming.

CA: Thank you.

ASG: Thank you Pedro and congratulations on the series because I’m not sure you hear this regularly but it’s really, really interesting so thank you for putting the project together and having these 20 podcasts.


PT: My pleasure.

ASG: Well … you know you got me on the first and the last so …


PT: Sorry Albert, are you saying that you should be in all of them, is that what you’re saying?

ASG: No, no, I’m saying probably I should just be in one, even in the middle, you know, for the next series.


PT: Yeah, I don’t know yet if we’re going to have a next series or not so if you’re listening to this and want to sponsor another 20 episodes of the Public Procurement Podcast, you know, just drop me a line. Where can we find you, so what’s your online presences guys?

ASG: So for me, you can follow me on Twitter at @asanchezgraells or at @how2crackanut, there two is a number or in http://www.howtocrackanut.com/ which is the new refurbished blog, and I hope to see you there.

CA:Yeah, and for me it’s mainly, for academic purposes mainly Twitter, it’s @carrebola and that’s with a B, and well if you want to follow the trip it’s @EUcycle or also on https://www.facebook.com/EUcycle.


You can find me at my blog, Telles.eu or on Twitter where I use two handles, @Detig for general discussion and @publicprocure for public procurement related topics. As ever, and for one final time, I am very grateful for the support of the British Academy Rising Star Engagement Awards which made possible this project.