 We have a very exciting seminar today which is on trademarks and we are very fortunate to have with us Professor Georg von Grevenets from the University of East Anglia and Georg is one of the relatively few economists who work on trademarks and especially if you compare the research that exists in the field of trademarks to the research that exists in the field of patterns it is much more at its infancy but hopefully this is going to change you know one of the things that really has driven research in the patent area was the availability of new patent databases and I think something similar is happening in the field of trademarks where now researchers increasingly have access to unit record trademark filing and registration data from trademark offices and Georg is one of the economists who has employed these data for a number of studies the study that he's going to present today is on the phenomenon of trademark cluttering and he will give the full motivation but what is interesting if you read one of the I guess seminal economic treatments on trademark law by Landers and Posner which was written in the late 1980s they essentially say that you know the name is unlimited and as such you know there's no shortage of possible names that could be converted into brands if you contrast that with you know the registries in many trademark offices and you know if you talk to some of the practitioners in the field they very much say well you know they're concerned about the namespace possibly running out and their increase in costs company phase and coming up with new trade names whether or not you know we phase situations of of clutter trademark registers you know I don't want to pass judgment on this I think that's ultimately an empirical question but I think the first at least as far as I'm aware of the first study that sort of tries to rigorously look at this is Georg's study on trademark cluttering in Europe and he uses a natural experiments that was given by the enlargement of the European Union to study what effect it had on possible cluttering behavior. Before I give the floor to Georg just one or two words of biographical background Professor van Gravenet is a professor at the University of East Anglia in the United Kingdom he's also affiliated with a Center for Competition Policy at that university as well as with the Oxford Intellectual Property Research Center at Oxford University he was previously a professor at the Ludwig Maximilian's University of Munich and he has studied and taught economics in both Germany and in the United Kingdom and as I mentioned he's one of the relatively few economists and hopefully it's a growing community and that does research on trademarks so we are very happy you're you're here and we look forward to your presentation. I'm very pleased to speak about the topic here because I hope that there are people in the room who may know something about this topic that I haven't found out yet it's it's as Carsten said a very new topic and I've been going through a review process and in this process my referees have motivated me to look at certain aspects of this more and more and so I keep learning and I'm sure that some of you will have additional points to make so please if there's anything you don't understand or anything where you think I've got it wrong or whether you have information then please just interrupt me. Yeah so let's start so the the the data I'm going to use comes from the European Trademark Office which started in 1996 in Alicante and they've been very generous in allowing me to use their data for various studies and so I'm very grateful for that so this particular study let me just oh wait this is not connected of course so I'm actually gonna have to go like this so this I'm going to jump over this you're going to so this particular study was motivated really by a piece of work that we did for the European Commission several years ago so the Max Planck Institute in Munich was asked by the European Commission to evaluate the European trademark system after it had been set up and had been running for a while and I was asked to supply some economic analysis and so I was party to some of the meetings that the Max Planck Institute organized with various people stakeholders of the system and so at one of these meetings the representatives of the pharmaceutical industry made plain this worry that Carsten just pointed out that they had the feeling that in their particular domain if you like it was getting more difficult to register trademarks and that this had something to do with the fact that there were maybe too many trademarks or a lot of trademarks on the register and so I got interested in that and we are we tried to put that into the study and this is really where this began so as as you just pointed out Landers and Posner argue that in principle the supply of names for trademarks is unlimited we can you know we can come up with new variations letters figurative elements infinitely and so in in a sense there's no issue here of of too much being on the register there's always space for more what what changes in this particular context a little bit is that in the context of medicines the the regulators became aware in the early 90s mid 90s of cases where particular patients are given example in a minute were prescribed the wrong medicine because somebody made a mistake and it wasn't the doctor it was actually at the hand at the process where the medicine was handed out and gave them the wrong medicine and this can be very dangerous for the individual involved and so the the authorities have tried to limit the scope for this kind of thing by basically regulating the namespace and making sure that the names are distinct enough for these kinds of mistakes not to happen and this introduces an element of scarcity suddenly we have a second authority it's not just the trademark office now but there's an independent agency which has totally different considerations which comes along and says you cannot have that particular name for this particular medical product so you might be able to get a trademark for it but the trademark is useless to you as a pharmaceutical company because you can't employ it in the course of trade okay so the the response and I've got quite a lot of anecdotal evidence for this and also some evidence in the data that the companies have come up with the obvious responses to file multiple names if you don't know whether a particular name is going to run into trouble then you just file a lot of them and hope for the best and typically that's what happens in the firms have worked out how many they need to file so then the interesting question is well what happens to all the names you didn't end up using after you've gone through the process that end up on the trademark register because contrary to what I've heard from some people at this at the medical agencies it's not that the firms go to the medical regulator first and wait for the medical regulator to make a decision and then file the trademark they will file the trademarks first to make sure that those are covered you know and then they go to the medical regulator so as a consequence we have excess names on the register and in some cases maybe the firms will reuse the names for other projects but in many cases they will just be there and be unused and so the interesting question is does this matter and when we did the study for the commission with the numbers I could come up with at the time the the lawyers involved basically said well this is trivial this is not important does not you know the clutter might be there but it's more an academic issue it's not really a big empirical problem okay so in this particular paper what I'm doing is I'm using the expansion of the European trademark system in 2004 which was just a consequence of the fact that the European Union was enlarged 10 extra members joined the 15 at the time to see what happens because essentially what happens in the regulatory space here is that before you're facing one European regulator who farms out the actual decision to 15 member states and all of a sudden there are 10 more and each one of those local agencies could say well this particular name in our particular context doesn't work because it's misleading or there could be a miss you know people could make a mistake so essentially it becomes more difficult to get the names through this particular approval process and as a consequence you might expect that the firms will apply for more trademarks simultaneously that's essentially what I'm trying to find out okay so first of all let's just have a look at how this office operates and how it sort of works so in 96 you see the office beginning and there's a big spike of applications simply because people had lined up they knew this office was going to start so they all sent their trademarks there and then we can see that over time there's an increase obviously in demand for these types of European trademarks there's a big bump here which reflects the internet bubble because a lot of companies that went into business at the time were taking out names and there was a big discussion about the connection between the domains and the and the patents and the trademarks then you have this very distinct spike here and this is your EU enlargement and the basic context here is that the way the European Union the accession countries the negotiations worked in such a way that essentially the new countries had to accept all the trademarks that were in the system there was no re-examination of existing trademarks so essentially if you could get into the Ohim system before the deadline then your trademark automatically got extended to these new countries without further examination and so obviously it's very interesting to meet that particular deadline and that's why you get this spike okay and then we get an increase and the decline here at the end is a is a truncation issue that's just because the data end there and so I don't have more recent data which would tell me what's happening there okay so as I already mentioned the the work we did for for the commission basically suggested that there might be an issue yeah I found this study by Lalmont which is a is basically a piece of commissioned work for by by a company that supplies trademark data where he analyzes this particular market and says and I find the analysis very interesting he says that five to ten different trademarks for each trial drugs are sort of routinely filed okay and I talked to the farmer the trademark council of one of the German companies that works in this air in this context and and he said to me it's very simple if I don't have a trademark when the product is ready I lose my job so I will do everything to prevent that from happening and so I will file as many trademarks as is necessary and he completely confirmed the numbers he said that that was the same so what we see in 2004 is the following countries as I'm sure you're aware joined the European Union and thereby they also joined the European trademark office and the European Medicines Agency became the agency that governed the pharmaceutical names in these countries as well okay so the interesting question now is is can can we see any effects of this kind of behavior and so before we go on to that let me show you a little bit why this really is an issue so here's a particular example we have a product which is taken by people who have problems with their stomach so if you have you know ulcers or a danger of ulcers then you would be prescribed this losec and then lasex is a product which is a product prescribed against hypertension so high blood pressure so as you can imagine these are very different conditions and if both of these conditions are potentially very dangerous if they're extreme and so if you've got the wrong medicine you would you know you wouldn't alleviate the symptoms this could get very bad for you so the FDA actually forced AstraZeneca to change the name losec to prelosec because there were concerns about confusion here and what then happened is that somebody mistake prelosec for prozac in a pharmacy and the person who was supposed to get prelosec had a gastric ulcer so again prozac wasn't going to do much good there so um so you know you I mean when you and I guess when you look at these things yeah when you look at these things when we look at them I I guess it doesn't seem likely that you would confuse this yeah but what does happen often in many countries still is that the doctor actually writes the prescription in handwriting on a piece of paper they don't type it yeah and then you take it to the pharmacy and the pharmacist has to work out what this means so I guess that's where these things happen you know and and um these kinds of mistakes happen also in hospitals and you know there's a lot of evidence for this kind of mix ups happening um so the naming committees were established to prevent these mistakes and and they work on the premise that these mistakes can and will happen and they are very restrictive so they're very strict what's quite interestingly is that when you look at the context of the european uh trademark office uh the opposition chambers at the ohem actually when they're dealing with opposition cases between uh names for pharmaceuticals argue that because this is such a vital kind of space where people are really you know in danger of making uh fatal mistakes people are going to be very attentive and very careful and therefore their standard of proof for confusion is actually very high in this context so they're coming at it from exactly the opposite uh sort of uh so let's say viewpoint you know um which maybe you know from the from the two agencies uh perspectives make sense but it's interesting when you can contrast that you know so they're both regulating this space and um I had a very interesting I tried at some stage when I was starting this work I called up the european medicines agency and tried to contact them to see whether they had any statements to make about what you know what was going on here and they weren't even aware that people take out trademarks for these names before the names come to them so they didn't believe that this was the case so it shows you that there's a total disconnect between these two um operators here okay so here's an interesting bit of data from lalmo I don't know whether it's terribly reliable because it really is quite difficult to get data from from the european medicines agency they they publish only the the the things they approve so all the the rejected names you don't actually get to see they're not on the internet you can't um see them so how he got this data I'm not quite sure but um if we take it at face value then it suggests that before 2004 it was indeed slightly easier to get trademarks so to get names approved then it became after 2004 which suggests that this idea that extending to 10 additional countries actually did make it more difficult okay so you know a big international pharmaceuticals company is going to face not only the european medicines agency but they're going to face the europe the american trademark office the european trademark office the american food and drug administration and the european medicines agency so these are the four principal sort of hurdles but there are probably many more you know in terms of the other jurisdictions um so and just to point this out again the rejection of a name would mean that the product launch would be delayed which in in context of pharmaceutical products is really uh can can quickly add up to very large numbers so it can be very expensive now one of the things that the refereeing process threw up that I didn't initially investigate very heavily was well what does it actually cost to create these names you know if we go back to landerson posner who say it's actually very easy to make names and you know the variation here is is is very sort of there's a lot of possibilities what's the big deal well it so I I initially found only evidence from the web and then finally I actually did find a few uh publications so this one uh by kenage and stein and then more recent one by wick who suggest various numbers for the cost of producing a set of names so this would be your five or 10 names that you're trying to generate to protect your pharmaceutical product and and these numbers are not you know they may seem high if you're coming from the perspective that it's fairly easy for any of us to doodle around and you know make up fake names um but it's not extreme when you compare it to some of the numbers that let's say the olympic committee pays for a for a for a you know the device for the olympic games in london or that you know oxfam recently paid to have a new logo for that it could use uniformly across its shops in the uk so this is a an NGO and and you know so they all spend around about this kind of money for for these types of trade marks or figurative devices um so what actually goes on here at the pharmaceutical sort of in the pharmaceutical context is that the the consultants who will work in this context will hire people who work in the uh in the medical context in different countries and ask them to perform prescription simulation exercises so exactly the kind of thing we just said they will write down the names and then we will take it to the pharmacy and see what the pharmacist thinks this is and whether there any there is any confusion they will perform tests of name similarity tests of implied claims so this is also something that's very important to the agencies that the name of the pharmaceutical product shouldn't suggest that the pharmaceutical product is going to do anything particularly well or particularly badly you know so it shouldn't imply that it's good for this or that particular ailment and it should be a neutral name tests of visual and verbal similarity linguistic analysis so there's a fairly large number of tests that these things have to run through and i guess the costs really come from the fact that the people you hire here to perform these tests with are professionals you just you don't just go on the street and take a random focus group so there's probably an element of cost of time so now the interesting question of course is and this is one that i'm not able to really admit sort of address in this study because i don't really see how to identify it is is what proportion of these costs is due to the fact that there are many other names on the register you know so it stands to reason that there's going to be an issue here if there are more names on the register then this group of people is going to have more work to do to figure out whether there could be confusion but saying exactly which proportion of these costs is due to that is difficult so the only other thing that i found is another study by these two gentlemen here who sent out some questionnaires to us companies in in the mid 90s and they asked them to try and figure out how much it costs them to create trademark names and i adjusted this number so the number they report in the paper is slightly lower but i've adjusted it because it came out in 97 so i took a sort of inflation adjustment here so essentially the you can see that there's a factor of 10 here if not more difference some proportion of that factor of 10 difference must be coming from this cluttering but as i say it remains open to debate which proportion that is okay so the the idea that i had was that i could use the european union enlargement in 2004 as a kind of shock and economists like studying these shocks because if they are if they're unanticipated then they have direct sort of responses on the parties which reveal something about underlying economic mechanisms that we normally can't see we can we can see a sort of causal effect that's very hard to detect otherwise so what i wanted to try and do is actually to try and use this particular shock to see whether i could see what the difference in cluttering was as a consequence of this enlargement okay so there are a couple of questions you have to answer here so first of all is this an experiment so one way this could maybe not end up being an experiment if if actually things that were happening in the trademark context were driving EU enlargement yeah then the causal effect that i'm trying to identify is actually sort of not there because it's actually the trademark filing which is causing european enlargement and not the other way around that it's european enlargement which is causing a change in trademark filing so in this particular case i think it's it's pretty safe to argue that european enlargement had nothing to do with what was happening in trademarks okay that was a political process there are lots of other things going on but i don't think there's much of a link to the trademark system so in that sense we can probably argue that there's an ex you know that there's a sort of an exogenous shock here okay now the other thing that's interesting is to ask well what are the outcomes here so if we're looking at this in a particular economic framework the interesting question is if there hadn't been the shock what would pharmaceutical firms have done in the alternative case and so that's also important to bear in mind exactly what we're studying so what we're studying here now is how does the behavior of the pharmaceutical firms change as a consequence of enlargement and that's not just as trivial as saying they apply for more or less trademarks it's a little bit more complicated than that we may have to take into account that if they anticipate that enlargement is coming and if it's sufficiently costly to them that may have effects on their decisions to actually continue operating in this particular space you know so if you thought that the cost increment was very large some of the firms might switch out of the whole thing entirely and say we give up you know now again that's probably not likely here although the costs are there they're not that big you know relative to the profits that pharmaceutical companies make these costs are still relatively small okay but we have to take the reactions of the pharmaceutical firms into account and and deal with them and so the last question here is so what happens how does the pharmaceutical firm decide to stay a pharmaceutical firm or not and I think the the mainly so the main point you have to bear in mind here is when does the pharmaceutical firm make which decision so essentially the pharmaceutical firm will do r&d quite a number of years before the product comes to market yeah and there are studies which show that this delay can be anything from five to eight years it might even be longer than that you know so if you think the r&d decision is taken um in the 90s for products that come out around about when EU enlargement happens okay so now the interesting question is so at that time did they know that European enlargement would happen in 2004 so is it conceivable that they factored that into their decision of how much money to spend on a particular class of drugs you know and I think that's very unlikely they may have and you know obviously there was a discussion of whether this could happen or not but the actual decisions the political decisions that were taken for for exactly when it would happen were taken later so what we can assume is that the pharmaceutical firms had a program of research which was running and then at some stage the political decision gets taken for enlargement and then the pharmaceutical firm has to respond to the implications of that which is that it becomes more difficult to get trademarks or names accepted for particular products that are in the pipeline so what I've done uh this was also in response to some of these uh referee comments was that I I haven't done I haven't given you the full formal model here because it that I think that would take us too far into sort of a fairly non-productive territory right now but essentially I'm making the argument I just made so firms make a long-run decision to undertake R&D and they make short-run marketing and regulatory compliance investments so here these are the investments in the names okay so essentially if you think that R&D investments pay off in period T they are made sorry they've they've undertaken period T the they pay off two periods later and the name investment is done somewhere in the middle okay and then what I can so show is that a regulatory regulatory change in period capital T that is announced one period before will lead to increase trademark applications for products that were in the cohort of R&D investments made one period before that so really what I'm trying to say is there is a group of pharmaceutical research uh products which essentially the firm has spent the R&D on so it can't adjust there anymore but what it can adjust is the number of names that it might apply for yeah but there's no other adjustment that's going to happen okay that's the essential point because for for later uh cohorts what could happen is that not only do they change the number of names they apply for but they could also adjust the R&D investment and so then I have two effects going on in the data and these may counteract each other so it's not so clear anymore what I'm seeing when I'm looking at the data okay okay so and this is the second result yeah so if if uh sort of later on the firms may factor in these additional costs into their uh decisions okay so how do economists study this well there's this this very scary looking expression here which is the empirical specification that I'm going to estimate yeah but it's actually fairly simple this is a differences in differences model so the idea is this you have two groups of companies you have the pharmaceutical companies and you have all the others yeah and you look at those before and after the shock yeah and so the idea is that the other companies will give me the baseline they tell me what would have happened to the pharmaceutical companies if there hadn't been a shock yeah and so then I can work out the the the so the difference for the non-pharmaceutical firms tells me what is the underlying trend yeah so are the firms anyway applying for more trademarks or less trademarks simultaneously yeah and then the difference between that and the difference for the pharmaceutical companies tells me what is the actual effect of the shock okay that's that's what what I'm trying to identify here and so what this is really doing is that I'm here on the left hand side what I have is a measure which is the simultaneous applications for names on a given day by a particular firm that's what I'm actually looking at okay and when I've discussed this with various people they've they've said well you know there are all sorts of issues how can you be sure that that all these names are actually going for the same product and the answer is I can't be so there's going to be some some measurement error as we call that in there but it still seems to be the case that in many cases the firms are actually making these applications for one particular product on a given day yeah I've checked that a little bit if you look at the data it looks quite good there are a couple of outliers so there are a couple of cases in my data where a firm is putting in several hundred names on a given day yeah and that's quite clearly not for one particular product this is a firm entering the system and so I've run specifications where those particular instances are not part of them you know not that they're removed from the data it doesn't make any difference okay so down here I don't know how well you can see this there are a couple of things that are important when you run this kind of model and you'll see that I'll show it in the data again so the first one is that the two groups are actually if you like affected by the same forces yeah if there are totally different forces affecting the incentives of these two types of firms to make investments in names then the comparison of the two groups doesn't give me anything yeah because they're they did they're just not comparable yeah essentially so the technical term for this is common trends so we'll have to look at the data to see that whether somewhere before the shock became sort of part of the you know before we started discussing European enlargement seriously are the firms behaving more or less in the same way if that's not the case then then we have to worry about the quality of the empirical results okay now another thing that's important here is that the pharmaceutical firms tend to be quite different from the vast majority of the other trademark applicants because they're much larger you know they operate in much more concentrated markets so that's going to be an issue do I have to control for that how do I how do I deal with this so if that doesn't affect the trend but it just affects the absolute number of simultaneous applications it doesn't matter okay now there's another way of doing this so this this difference in differences model was one way of doing it another way of doing this is is essentially what's called a matching process so there what I do is I take the pharmaceutical company and I look at its characteristics and then I try and look in my data for another company that has the same characteristics so same size you know has been active in pharmaceutical has been active in trademarks for the same duration you know maybe comes from the same country okay that these kinds of things has a similar size stock of trademarks all sorts of characteristics of the company that you might think were important in this context and then I just compare those two companies before and after you know and so here what I'm doing is I'm essentially moving into a framework where we really think of this as an experiment you know now in a in a natural science experiment it's a controlled experiment you know and I normally do what's called random assignment so I pick randomly the people who get the pill and the other people don't get the pill and then I can do a comparison right so obviously in this context here I don't have that you know pharmaceutical firms are not randomly picked to be pharmaceutical firms so this is the assignment issue again and so the interesting question is can I can I apply the the sort of methodology of a random experiment to this particular context or is it actually not possible because the group of firms I'm looking at is selected in a specific specific way they've become pharmaceutical firms for particular reasons so there's a there's a sort of set of very technical sort of discussions in the economics literature where people essentially argue and this is immense here he says so if the participation so being a pharmaceutical firm here is separated from the outcome so in this case how many names we apply for then we can argue that we can use the analogy to a natural science experiment anyway even though assignment isn't random you know okay so the question is whether treatment uptake depends on variables that have no effect on outcomes okay even if we don't see those variables so we then we go back to the model and we can have a long discussion about what motivated these companies to become pharmaceutical companies at the time when they were doing the r&d or to stay pharmaceutical companies and whether this is in any way related to the decision about how many names to apply for several years later and so my argument here would be that actually these decisions are fairly far apart and probably the reasons for doing the two things are fairly different so I'm quite happy to apply this particular analogy okay that's that's the kind of implicit story here and I'll show you the results from the two it's quite interesting to compare those right so when you do this kind of study what's important now is to determine who are the groups the group which is the group of firms that is being affected by enlargement and which is the group of firms because they're pharmaceutical companies and who are all the others and here the problem is that I just have registered data and so what I can do in the registered data and this is what I do do is I can go into the niche classes so the trademark system is separated up into different classes and each class is specific for a particular group of products and what I can do is is say well here's a class that is clearly contains the pharmaceutical applications but it may also contain applications for pharmaceutical products that are not regulated by this name regulator so it will contain other pharmaceutical products like this bandage where the name of the bandage is is not regulated in any sense okay so that's an issue with the data so in my group of treated firms those ones which are affected I will have some which are actually not really treated they shouldn't be in that particular group but I can't perform the identification who is really producing medicines that are subject to name regulation and who is not okay what I can do is I can identify companies who are applying for trademarks in a completely different space so for ice cream or cheese or cars or something like that you know so I can separate so essentially what I did was that I started out saying there's class five which has all these pharmaceutical products in them and then I looked at all the other niche classes and said well how often do firms simultaneously apply for that particular niche class in conjunction with class five so a pharmaceutical firm may go for class five class three and class one fairly often whereas other class firms will go for class 11 24 and 32 in a combination yeah and so I separated the classes into two set of three sets those which are really closely related to class five those are that never really have any connection to class five and then sort of more amorphous intermediate group okay and then using that I created four different groups I said if the application falls into into classes that are very closely connected to pharmaceuticals then it's a pharmaceuticals application if it falls into classes that are very different from pharmaceuticals only then it's an artifacts this is a name that I made up application if it falls into that intermediate group then it's this and then there's also an overlap between the pharmaceuticals and the artifacts so there's a sort of mixed group yeah so these are basically the sets of groups that I identified and I'm really only interested in how the applications that go into this group differ from any of these okay the differences between these don't really matter they're helpful in the sense that in my study I don't expect very big differences between these two groups so if I did find big differences between these two groups that would open up questions about the the way I'm looking at the data and there's obviously something going on that I don't necessarily understand so just to take you back to what is it I'm looking at so the fundamentally it's quite simple I'm just looking at the number of names that a firm puts to the trademark office on a given day in a particular set of niece classes so the pharmaceutical firm will take out five eight ten simultaneously and other companies may take out two or three simultaneously or only just one and so what you have here on the left hand is the logarithm of the average number of simultaneous applications in a particular class at a given date okay and the red line here are the pharmaceutical applicants the black line are the artifacts and these are the two other groups and I have put around these numbers these these lines the confidence intervals okay so these this is basically the range of 95 percent of the observations instead of in that particular class okay um all right so what you can see is that until just before the internet bubble well I don't know whether you want to call this common trends but essentially the things seem to be moving more or less in the same direction you know it's fairly flat you know there's not much variation and then you can see that the internet bubble has an effect in all of these areas the number of simultaneous applications seems to go up a little bit it goes up particularly in pharmaceuticals and then it stays up there at a higher range you know increases again just before enlargement and then drops down and that difference between the groups stays the same okay so now there's something that's important here which is that the the shock I'm interested in is the one that happens here in 2004 this is where enlargement happens so in in the canonical type of application of this particular method what you would really like to see is this red line more or less on a par with the black line until about here and then suddenly jumping up and then staying up here okay so that's quite clearly not what we see you know so what we see is the two lines the two sets of lines are separating much earlier so when I saw that I started thinking about this I thought well you know does this mean that the whole thing is is is sort of that is we can forget about this and then I realized well actually what's important here is the Treaty of Nice which is the treaty that established that in 2004 we would have enlargement and that's really the political statement where all the countries come together and say okay now we make a legal binding agreement that we will enlarge in three years time so then you could say well actually the shock if you like for the pharmaceutical companies is here and it's important to bear in mind here that the pharmaceutical companies will take out the name several years in advance they will anticipate that they want the product on the market two or three years from now or four years from now and will take out the names then okay so then then it looks more likely if you like that there was actually an effect here okay so statistically what we now do is we just transfer this visual idea into a statistical model which allows us to filter out a couple more effects and we can add sort of variables that may filter out effects for the size of the company to some extent and and maybe the country the company's coming from and things like that which might also affect the tendency of the company to apply for more or less names okay this is a sort of descriptive data we were just describing the the data here so what I did is once as I told you before I did I've done it for the whole sample and then I threw out all these outliers so the cases where the firm applies for 127 marks at the same time and you can see here there's an extreme outlier here where one company is applying for 634 trade marks on one day yeah this is quite I see this sometimes this is clearly a company which has decided they you know what often happens is the companies watch this new trademark system for a while and they may even participate to the extent that they oppose other applicants but they don't file and then at some date they suddenly decide okay now it's it's important for us to take out trademarks within this system and then they enter and then it's an avalanche they come in with everything they have on one day yeah and so that's what we're seeing over here okay now but as I said in this in this in this particular sample here where I'm giving you these means I've basically filtered those out okay and these are some of the variables that I try and use as additional controls to kind of make sure that the comparison I'm making is actually even handed as far as possible with this data okay so here are the results the the statistical results and what we're looking for now is the interaction of the dummy which identifies the particular shock I'm interested in so here we have anticipation so this is the effect of applications between 2001 and 2004 and the applicant coming from the group of pharmaceutical firms okay and this and this this effect here is a comparison against a baseline which is all those companies that are not in pharmaceuticals not in artifacts and not in food and household yeah in that period okay so it's a comparison of the difference in the behavior of the pharmaceutical firm before and after 2001 relative to the difference in the behavior of the sort of comparison group and so what you can see is there's a significant positive increase in the number of simultaneous applications at that time okay that's what this blue thing is showing you and even if we compare after 2004 with the period before 2001 we still have that effect it's a little bit weaker but there's still an effect okay so essentially what this shows is that European enlargement had a effect on the simultaneous applications of the pharmaceutical companies relative to all other companies okay now there are various things you can do to see whether this effect is real you could you could argue them there may be something wrong with the econometric model that I'm using so first I threw out all these outliers I kicked out all these companies applying for large numbers of fur trade marks to see whether this was robust and essentially you see the effect doesn't change very much then you might say well the way I've controlled for time effects so for macroeconomic shocks that are happening over time might be uh wrong or not sort of fine enough so I introduced instead of controlling for years I and controlled for the quarter that also has no effect and doesn't change anything very much and then there's an important test which you can do which is to include a time trend so you just say that instead of allowing things to vary from quarter to quarter you could argue there's an underlying trend upward or downward in the data okay so it's interesting what happens when you put that in because in the context of this kind of study that often means that the effect vanishes the one that you're studying yeah and in fact here we can see that now um the comparison of uh sort of what happened before 2001 and what happens after 2004 is no longer sort of statistically significant there's no the effect isn't strong enough for an economist to say okay this matters yeah um but still the period between 2001 and 2004 we still have that effect okay so that there was something at least for a short period of time if we include the time trend and it's important to notice that if you're sort of happier with this particular specification then there is still the effect of time there is still that trend so it's clearly and that is if it's significant so it's it's clear that something is happening to simultaneous applications it may not be as radical as the the shock at enlargement but it's sort of something did happen okay so this is one set of results and now the interesting question becomes okay we've seen there was an effect is it economically important do we care about this is this does it matter you know should we should we worry about it um and I'll I'll talk about that in a minute so before that I have one last test I'm sorry I'd forgotten about this so one thing you can do is you can make this model terribly spent sort of flexible in the sense that you can allow every quarter to have its own effect yeah so what you're looking for here now again is is basically you would want the the average of this to be below zero and these bars here indicate a range of statistical sort of significance so essentially I'm saying that if this effect here if this bar crosses zero then this effect that I'm measuring here is statistically not distinguishable from zero so every time the zero is inside this particular range that's indicated here by the by these by the lines you know then the effect that I'm measuring for that particular quarter could also be zero and we couldn't distinguish the two so it turns out that before 2001 there are a couple of quarters where we get significant sort of negative effects and after 2001 we get a couple of quarters where there are significant positive effects so again the data suggests there was some increase in simultaneous applications um but it's perhaps not quite as pretty as I would like it to be yeah so if I could choose of course then I would have all the error bars below zero here and all of them way above afterwards and then I'd be really confident that there was a significant shift okay here it looks more incremental okay right so now let's try and figure out whether we can say a little bit more than just okay there was a positive effect and it was significant so as I told you before what I'm measuring in the class of pharmaceuticals is actually a mixture of the pharmaceutical applicants who are applying for names that are regulated and other companies that are applying for names for pharmaceutical products that are not regulated by name regulation yeah so that's this number here this is the average and it's an average of a group of unregulated firms and their simultaneous applications and the regulated firms and their simultaneous applications so capital N here's the number of regulated firms capital O is the number of non-regulated firms and this expression is basically just giving me this average and then what I did was I introduced a factor which says that the proportion of these two groups within this set of pharmaceutical applicants is constant over time yeah I can rewrite the whole thing like this and what I'm doing by blue here is I'm marking numbers that I can observe in the data yeah so I know this number for before and after the shock and I know this number if I'm willing to make the assumption that the firms that are inside my pharmaceutical group but that are not regulated by the name regulator are behaving in exactly the same way as all the other firms in the system that are outside pharmaceuticals yeah then I can also get this number okay and then I did a couple of more things I assumed that name review is equally tough in each of the countries and then I used this data that you saw earlier by Lalmore to calculate how difficult it is actually to convince the regulator in each country yeah then I assume that these two probabilities here don't or proportions and probabilities don't change at the shock and then finally I put that all back into my economic model and so what I can then work out from that is that between 2001 and 2004 every year we spent about 17.7 million US dollars on inventing surplus names okay so that's you can do this with this with this economic model now there's quite a lot we have to just you know I wouldn't put too much sort of emphasis on this number there are a lot of very strong assumptions going into this calculation here okay but I did it because basically one of my referees was asking me well can you tell me a little bit more about what your results imply you know is there anything any number that can come out of these and and so I was really tempted to try and figure out something and so I pushed the model as far as I could go yeah and I would say that for example the model using the estimates that I have is suggesting that this proportion of pharmaceutical real pharmaceutical applicants in that group of firms applying in that pharmaceutical class is quite high it's saying it's 96% and my expectation is that that that number is too high I would I would have expected that proportion to be a bit lower and so I guess some of the assumptions I'm making here to get to that to identify that particular variable is are probably too strong but anyway so what we have is a number which gives us some indication of how expensive this is this has been and an important assumption going into that number is that each new pharmaceutical trademark costs $25,000 which is at the lower end of that range that I showed you before yeah so the journal suggested that this could go up by a factor of seven and so then we can scale up this number by a factor of seven as well okay right so if we do that if we say that we're spending about 50 million dollars on invented names that we don't necessarily really need yeah so then the interesting question is can we can we reduce that cost would that be something we could do and so there are two answers to that one answer is probably a little bit yeah because if we didn't have as much clutter then maybe we wouldn't have to invent quite so many names simultaneously but we're never going to get back to a situation where the pharmaceutical applicant applies for only one name because there's always the risk that the name regulator says this particular name doesn't work in my jurisdiction and I think there's I can't think of any intervention that we could create to regulate this system that would remove that you know the pharmaceutical firm is always going to face the risk that it's picked a name that doesn't work you know so all that we can do is we can try and make it easier for them to pick the names but we can't totally remove this this phenomenon of simultaneous applications I would think okay right so just to contrast the number that I came up with here before okay so this is the number that comes out of this diff and diff framework if you use the other empirical framework that I was talking about earlier on so I'll just jump over to the results here now you get significantly and I'd like you to just focus on this particular column here and all these numbers are derived under slightly different assumptions on how you use this estimator which are technical and we don't really have to go into because they're all more or less in the same range yeah they're significantly higher than the numbers you saw before yeah the numbers you saw before were 0.06 and now we've got numbers in the range between 14 and 18 yeah so you can see the effect here is bigger so what's happening here now is that the estimator is really picking the pharmaceutical firm with a particular set of characteristics and comparing it to another firm with similar characteristics and then trying to see how do they differ in their behavior okay and there is an argument to be made for the fact that this estimator is more appropriate here because the two groups of firms are just so different in their characteristics to start out with okay so all that I think we can take away from this exercise is that there's another reason here to believe that the effect is probably larger than the number I gave you earlier on yeah because the estimate is probably you know the estimate of the shock is higher so that's that's what I would take out of that okay and there's some more robustness checks here okay so now that got a little bit technical at the end but really this is a my attempt to summarize what I've got in the data okay so I would say that that graphical evidence I showed you at the beginning or in the middle this picture where I was comparing the different groups of firms doesn't really favor the idea that there's a big jump in cluttering that you can see a signal but it's not like it's overwhelming once you start applying empiric a sort of econometric results then you do get some indication that there was an effect and also when you control for specific periods yeah so when you allow the the estimator to be very flexible about the effects of individual points in time we get some evidence of a jump yeah if we include this time trend in the difference in difference model then the treatment effect is no longer significant the trend is also not significant so here the results are a little bit more ambivalent but I would point out this is something I didn't put here that the the effect for the period 2001 to 2004 was still significant the matching estimator suggests that enlargement did have a big effect and that those effects could be four times bigger than the ones we got from this other method yeah so if you're wondering why the economist is doing this why is he using one model and then he's using another model this is because I tried to be a little bit open about this we can see that there are various assumptions going into each of the models and we can have long discussions about whether these assumptions are justified or not so what we try and do is we apply as many models as possible and try and see whether the results are sort of you know fit together and if they do we have more confidence in them that's essentially what what we try to do here yeah okay so so if the membership in these groups you know so if the fact that you're a pharmaceutical company is exogenous to European expansion of the European Union so it's not affected by that yeah then the results either show that there was a significant increase in simultaneous applications in 2004 or that there is a positive time trend here both of the results suggest that something is happening the matching estimator shows that the effects are pretty significant yeah and the rest here is really just ideas of how we could try and extend the the this particular empirical exercise to try and have more confidence in the results yeah so one thing we could do is we could pull in an extra jurisdiction so we could try and compare let's say to the United States because then we would have more comparisons in the data yeah yeah we could try and be more specific about how we control for pharmaceutical firms which would require much more effort on identifying the individual companies as companies that are regulated or not regulated we might want to control for the rate of conversion of trademarks in the particular areas that's something that I haven't done yet and reassignments of trademarks for different applicants so these are some of the things that might still be done sort of in this work or in other work okay and that's that's it thank you very much Gail for this interesting presentation and particularly appreciated that you know you I think gave quite an extensive motivation of why this is relevant but then went through the details of your approach and your results