 So, everyone, welcome back to the second keynote lecture of the 2021 TSC Digital Economics Conference. And today we're very, very happy to have you, Katarina Zyrowska, here. She is a professor of economics at the Paris School of Economics, and she's written like a truly staggering and amazing number of really great articles about the connection between media and politics and governance. And we're very happy to have her here to talk about, and again, so now she's moving to thinking about social media and the internet and how it affects politics and government. And so we're very happy to have you here to talk to us about three-gener and a conference in government. So, Katarina, I'll pass the floor to you, but just to let people know about the organization. So, if you have any questions throughout the talk, feel free to send them to me in the chat, and then I'll pass them through to Katya. So the floor is yours now. Thank you very much, Daniel. Thank you very much for the organization of this wonderful conference and for the invitation. It's a pleasure to be here. I'm just only said that it's not in person. So this is joint work with Sergei Guriev, who is in Sien-Spoil, Nikita Melnikov, who is a grad student in Princeton, and he will be on the market next year. And even though my keynote will be based on a paper, I will try to be a little bit less rigorous and maybe open up a little bit more compared to a normal paper presentation. And let me start with a very simple observation, which probably one doesn't even need to mention in this particular audience, but generally it's in use to quite a lot of people, which is that in the last decade, we have seen the ICT revolution, which has to do with the vast increase in broadband internet access across the world. And that probably most people know, but not most people know, not all people know that it has been due to the expansion of mobile broadband internet. And by mobile broadband internet, what I mean is 3G mobile technology and higher, 3G, 4G, and you know, the coming 5G. And here what I just show is the expansion of the total broadband internet access and separately broken by mobile and fixed. And as you can see, fixed really did not contribute to this global expansion. Moreover, and perhaps a little bit more surprisingly, it is true both for developed and developing countries, in particular in developed countries where we are usually thinking about internet as something which usually comes with optic fiber, the expansion of internet broadband has been tremendous and also was due to the mobile internet to 3G. And I should say also that the implications of 3G go well beyond just broadening broadband internet access. In particular, 3G was the key driver of a rapid expansion of social media. And the reason for this is that it was the first generation of mobile technology which allows users to freely transfer images and even more important to really stream and upload videos. And both of these features were absolutely instrumental for the development of social media. And here I just show you that indeed there was a really, really huge increase in particularly the use of Facebook and YouTube but other social media platforms as well which coincided with 3G. It is important to note of course that social media appeared before 3G, but the reason why it thrived and expanded so much was precisely this new technology which allowed people to just post stuff which they see right on the spot. So therefore, I would like just to right away argue that 3G not only changed how much time we spend online, not only just increasing access but what we do online. And social media of course is the key to this. And given this ICT evolution on the background, there has been a public debate and debate among political scientists and political economists about what are the political implications of the expansion of mobile broadband internet around the world. And broadly speaking, there are two views of this phenomenon. On the one hand, starting with the Arab Spring, there is a bunch of people who are very optimistic about the new social media role in politics, particularly internet has been branded and social media has been branded liberation technology. And there are very good reasons for this. In particular, social media and internet do improve informativeness of the voters in those places where there are no independent of the government political sources of information. And even more importantly perhaps, social media because it is a two-way information flow technology, it allows cardinating when people for example want to stage a protest. So all of these seem to be important ever since the Arab Spring has started the wave of protest which spread across the globe. That's on the one hand. On the other hand, however, pessimists consider both social media and internet media more generally as misinformation technology and we all of course know that these debates very well. Why would that be the case? Well, again, for very good reasons. First, it may facilitate dissemination of fake news. It may empower non-democratic regimes by reducing costs of surveillance and propaganda. They also help populists to connect to voters through social media as we have observed well in the last few days as well as over a longer period of time. So when we look at this debate from the academic perspective, actually these conjectures have found very nice and very solid empirical support in a number of single-country studies which look at various cases and identify them well. And here I would like to just take an opportunity to advertise a review which I wrote jointly with Maria Petrovna in Kalopov, which has come out last year in annual review of economics which summarizes all of this literature. It seems that one country's study perspectives are losing something because we really are thinking about a global phenomenon and to illustrate to you that this is a global phenomenon, I would like to quote from a very nice book called The Revolt of the Public written by Martin Buri who is a former CIA analyst. And he is very much interested in the political effects of social media and internet. And he writes, I quote, the rise of HOMA informaticals in person relying on social media for information places governments on a razor's edge where any mistake, any untold event can draw networked public into streets. This is the situation today for authoritarian governments and liberal democracies alike. The crisis in the world concerns the loss of trust in government in general. The greater the diffusion of information to the public through social media, the more illegitimate any political status quo will appear. It poses an existential challenge to the legitimacy of every government and I should add the legitimacy of democracy as well. So this quote from Martin Buri really is at the core of what we are doing in this paper and what we are trying to understand in this paper. In particular, this paper tries to look at the global perspective. So we would like to try to document political effects of the expansion of mobile broadband internet across the whole world. And to do this, we are combining two sources of data which cover the whole world. One is the data on the expansion of 3G internet and that we have a very, very high resolution and we have that for the whole decade basically, last decade. And we merge that data to the survey data on attitudes towards government which come from the Gallup world poll for the same time period. And I'll talk about the data more, but I should say that here we have only localization at the sub-national level. But it's still not bad and it allows us to do quite a bit as hopefully I will convince you. So and then finally, we try to understand what are the electoral implications of the 3G expansion on democracies. What's important here is that the global setting allows us to shed light on at least some of the mechanisms at play because we are able to compare the effect of 3G on government approval across different institutional environments. And that's something which can't be done with single country studies. And let me write away, given the time is short, give you a few takeaways. First of all, we do find that access to 3G internet significantly and slowly decreases government approval on average. And there is very important heterogeneity which points to the fact that it is the content which is available on internet, which does the job. Particularly, it's the availability of independent of the government political information, which is definitely one of the channels at play. And importantly, it's not all about fake news. Some of this information is accurate. And we show that internet helps to expose actual corruption. And finally, we also obviously know that fake news do circulate on social media. And I'm going to talk about that a little bit too. But overall, let me just, before we dive into the story, tell you that we also show that in Europe. And that's where we have data on elections. We also find electoral implications. In other words, we see that incumbents lose votes with the expansion of 3G. And overall, from psychologists, we do know that one story is more persuasive than statistics with 1 million observations. I'm going to show you some statistics with 1 million observations as we go along. But before I would like to start with three stories, which basically give us a picture of what are we talking about? What should we expect from the data? So the first story is the story of the triumph of this gentleman whom you see on the slides. This is Klaus Joannis, who is a Romanian president, who will be remembered in history books as a Facebook president. And in particular, he won for the first time presidential election in 2014 when he was a complete outsider to politics. He's a former physics teacher. And at the time, and still unfortunately, Romania is a pretty corrupt country. And it was the second most corrupt country in the European Union after only Bulgaria. And he was campaigning on the anti-corruption ticket and won. And won despite the fact that his campaign was essentially on Facebook. He was really, really not doing anything more than just social media campaigning. And in fact, he himself on the election night wrote a post to celebrate his victory where he said, together, we have won this battle here on Facebook. For the first time online has made a difference. I'm not sure if it was the first time, but certainly the first time in Romania. And indeed, there are a number of post-election surveys which show that for a large number of Romanians, social media played a major role in their decision in this elections. And let me just give you a punchline of this story, that it seems to be a happy story, because Joannis sticks to his election promises. He tries to fight corruption in Romania from his office. This is a parliamentary democracy. So there isn't really much the president can do, but he can do some and he's trying to do this. And he actually got re-elected in 2019. So it seems to be a happy story where social media leads to some kind of political change. But unfortunately, it's not always so great as a story. And in order to illustrate this, I would like to give you an example of the story of this gentleman. This is Jair Botsanar who will go down in history with all sorts of different things. But one thing is definitely is important for my story is that his campaign was 100% on WhatsApp. So he also just as Joannis was a complete outsider but unlike Joannis, he, as you know, is a right-wing populist. And interestingly, due to the very strict rules on electoral campaigning in Brazil, which very much favor incumbents, he didn't get any access to TV. So he had to run for the digital campaign. He didn't choose it. He just didn't have any other choice. And the digital campaign in Brazil means that you need to find social media through which you can reach out to most of the Brazilians. And the only social media like that is WhatsApp. Why? Because vast majority of Brazilians have the so-called zero rating mobile plans which do not allow free access to internet or don't allow access to many of the social media apps just because these are much cheaper plans. And that's most of them, like almost all of them have WhatsApp. So what Sanar did, he ran the campaign on WhatsApp with a lot of his supporters who penetrated the groups because the WhatsApp communication goes through the limited groups to about 300 people with encrypted chats. And what's important about his story is that apart from the completely legitimate criticism of entrenched elites, he also disseminated, he based his campaign on disseminating false stories about incumbent. And there's quite a lot of observers who showed that in Brazil these fake news about incumbents were decisive for the results of this election indeed with Sanaro won the second round with 55% of the vote. And interestingly so, if you just look at the data on micro regions in Brazil and look at the results of Bolsonaro's victory, you see that there is a very strong cross-sectional correlation between the Bolsonaro's vote share and the share of micro regions territory covered by 3G. And it's particularly surprising given that mobile internet coverage is much higher in Brazil in more developed areas and more urban areas where a priori it is much less likely that people would vote for a populist right-wing candidate. So again, the story is that mobile internet and WhatsApp in particular played a very important political role. But this time maybe not as positive as my first story. And the third story would like to highlight the fact that the effects which we are looking at are not just limited at democracies. In particular, we can think about examples when exposing corruption or in autocracies also plays a role. Of course, there are no free elections in autocracies so autocrats cannot just lose power in elections. Yet still they can be affected by what's going on social media. And here is my example of the screenshot from a documentary which was in Russia only available on YouTube which was done by the leader of the opposition, I should say, in Russia, Alexei Navalny, who is now world famous of course, which exposed corruption of the former Russia's prime minister and before that quote-unquote president, Dmitry Medvedev. And that documentary essentially led to a very sharp and unexpected fall in rankings of Dmitry Medvedev and eventually forced his resignation. So that video was watched by 20 million in the first week after it was posted. And as of yesterday, this film was watched by almost 40 million people on YouTube which is quite a big reach out. And it's very important that not a word about this was out in any of the traditional media. So these stories basically just highlight that we should expect some effect of a mobile broadband internet arrival. And let me now come to what we actually do in this paper. In particular, the central outcome variable which we're looking at comes from the Gallup World poll. I already said that. And this is a survey of about one million people. And in over 120 countries of the world, they ask a series of the same questions about people's attitudes towards the government, towards the incumbent in each of their particular countries. In particular, the questions concern other following. Do you have confidence in the national government, in the judicial system and courts in your country? Do you believe in honesty of the elections in your country? Do you think that corruption is widespread throughout the government in your country? And for all of these questions, you only have a binary answer. Yes or no? So we cannot look at the degree of the dissatisfaction of people but we can see the majority opinion basically is the overall government approval rankings. And this information is available at the individual level with the sub-national region localization. So we know which sub-national region, usually it's a NUTS2 level where the respondent lives in each of these countries. And we merge these data to the data on the 3G network coverage by year from 2007 up until 2018. And in this particular map, we just highlight the fact that at the beginning of our study in 2007, there was almost no 3G coverage in the world. And if we jump 10 years later, you see with these dark color, we indicate the 3G coverage. So starting in our first year of our observation of our sample, there was only 0.4 active mobile broadband subscriptions per capita. And by 2018, there were 0.7 mobile broadband subscriptions per capita, which is a big expansion. On top of that, this map highlights the sample of countries. So our sample, when we merge the two data sets, is comprised of 116 countries and 2,323 sub-national regions. And just to show you the size of the sub-national regions, we can see that on these maps with this, I don't know how visible it is for you, but you can see that we look at within country localization. However, there are some important countries, which unfortunately are not in our data set. For example, neither in China nor in Iran, Gallup World is allowed. So we cannot do that analysis there. So with this data, what do we do? We basically do a very, very simple thing. The first empirical exercise is just a standard difference and difference regression, let's say, where we look at how government approval changes with the arrival of mobile internet into sub-national regions, taking into account other factors that can affect government approval, such as economic development, demographics, education, income, and region and year fixed effects, and also in some specifications, regions by year fixed effects. So we are just basically trying to partial everything, all the potential confounds out, and then see what the relationship is. And I can basically illustrate this relationship with one graph of them. And that's the graph which you see in front of you, where on the vertical axis, we plot the government approval on average, net of all controls, which I just listed, and on the horizontal axis, we plot an increase in the sub-national regions, 3G coverage. And by 3G coverage, I mean the percentage of population who lives in areas which are covered by 3G. And what we see is that as 3G expands into the region, the government approval falls. And here it's just essentially raw data, net of controls, where the dots illustrate the equal size means by equal size means. So these are these one million observations which I mentioned before. But to be more precise, let me give you the actual estimates which come from this analysis, and I illustrate them in this graph. Here, the dots are the point estimates of the government approval, and there are several characteristics which we're looking at. The confidence in government, confidence in judicial system, and honesty of elections. The belief that there is no corruption in government and the overall government approval, which is the aggregate of the other four, which is actually the first principal component of them. So the blue dot is what the average level of all of these measures is before the arrival of 3G to an average region across the globe. And the red dot is what happens after the region gets covered fully with the 3G. And maybe it's worth noting that the rectangular shapes, they show the confidence intervals. So essentially what we see is that there is a pretty large decrease in confidence in government, and particularly if we look at the government approval overall, the magnitude of it is 5.7 percentage points, which is quite a lot, given that it starts at about 46, 47 percent of people approving their government in an average region, an average country. And interestingly so, the effect is even bigger for rural residents. So here we are talking about the magnitude of about 8 percentage points. So we jump from zero to 100 percent increase in 3G. It's worth noting that the standard deviation is about half of that actually. So it's not true that everywhere we see over the period of 10 years in which we look at where, that everywhere the jump is 100 percent, but yet it is pretty large. So the question then which I'd like to address is, you know, first, what is this, should we believe this relationship? Is this relationship causal? In other words, could it be that the extension of 3G networks is driven by a change in, let's say government approval or in some other factors which are related to government approval? And the answer of course is yes, it could be, but it's unlikely for a number of reasons. And in particular one reason which we highlight is that we can look at the dynamics of government approval around the sharp changes in regional 3G coverage. Generally, most of the time the 3G coverage increases gradually. So we of course identify the effect from differential increases in different places, but we would like to see what happens if suddenly, you know, 3G gets installed in the whole region. And how do we do this? We define a sharp increase as more than half of the population of a region, of a subnational region, which suddenly gets 3G in one year. And by definition it could only happen once in history, if it happens at all. And it turns out that there are 452 regions in 62 countries where we can observe that. And that gives us the possibility to see if there is anything going on before this arrival of 3G to region happens. And in particular, this gray line basically illustrates this event, the experiment which we're looking at here. So what happened with 3G coverage before the sudden increase and after? And we see that nothing was happening before, nothing is happening after, and this one event happened in one year suddenly. So what we'd like to see is what is the dynamics of government approval? And that's what we're showing this blue schedule. In particular, we see that there are no pre-trends. So it doesn't seem like there was anything going on before the technology was installed in the region. And right after it was, government approval falls and it stays low and actually even increases with time. Of course, nowadays, you cannot just show the results of different diff without understanding that if it's a staggered treatment, you need to use a robust estimator. And for those of you who are concerned about this, here is the schedule, the same estimates, but using the estimates of the Schismann, Mark-Hannon or Föy, which understand that there could be some negative weights when you have a started treatment and you get exactly the same picture. So this event study analysis is something which helps us to make sure that it doesn't seem like there's other confounding factors around the same time with the expansion of 3G. The other very important question which we need to think about is that would it be that it's not about broadband mobile technology? Could it be that it's any ICT technology when it arrives, then it somehow is related to government approval or dynamics in it? And in order to do this, we look at the placebo using the expansion of 2G technology, the second generation. And the second generation compared to third generation is the one which allowed SMS, allows texting, but it doesn't allow broadband internet access. And what we see using 2G, and this is a graph for our estimates of what's happening when 2G arrives to a subnational region, we see that the effect not always significant, but it is actually reversed. So with the arrival of 2G in the region, the government approval increases. So that actually means two things for us. One is that our effect is really about broadband internet, which is mobile rather than any ICT technology because the expansion of 2G actually followed the same rules as the expansion of 3G but a little bit earlier. But second, what's important is that there is another paper which is an excellent paper, which is now forthcoming in Kinematrica by Manacorda and Tesse, who look at Africa, African continent and who show that 2G is associated with the higher incidence of protests. And here it is important to compare the results of this analysis and the results of Manacorda. They say to really understand what's going on. So basically, indeed 2G was enough to help essentially minorities who are usually staging protests to organize and sometimes also to get informed, but it took social media and the next generation of mobile technology for that discontent with the government to spread to the majority of the population. So it is very important that what we're looking at here is just average ranking for the average voter. And that's not the person who usually participates in protests. So however, in order to completely nail causality, we need an instrument, of course. In particular, we need to find a factor that affects 3G expansion and is sure not to affect government approval other than through 3G expansion. And what we use is the frequency of lightning strikes which usually cause power surges and therefore they substantially increase costs in providing mobile services. And therefore, they may hind the rollout of telecommunication technology. And that instrument actually was also used by Manacorda and TSA for Africa. Just for you to illustrate that indeed frequency of lightning strikes is a very important predictor of the speed of 3G expansion, particularly in those places where there's no power surge protection technology easily available. I can show you this graph where we look at the rollout of 3G in regions with low and high frequency of lightning strikes per square kilometer in the same countries. So I keep the set of countries the same where you have both types of regions and you see that in places where there's no the lightning strikes frequency is low, that's where 3G rolls out much faster. And what we see overall and this is now we can look at the dynamics of government approval in regions with high and lower lightning frequency. What we see is that indeed only in those regions where frequency of lightning strikes is low, that's where government approval has declined. And of course we do the rigorous analysis with the second stage and we confirm our estimates which I showed you earlier using just a simple definitive. So now let me come maybe to a more interesting part of the story to try to understand what is going on. And the first thing we want to understand is whether it is indeed the content which is available online which is behind our estimates. And in particular when we think about it if 3G provides citizens with negative information on government performance and in this particular case doesn't even matter whether it's real or fake, then we should expect two things. One, we should expect that the effect should be weaker if the internet content is censored and the effect should be stronger if alternative channels of information, for example traditional media assets. And that's exactly what we find. In particular here we just split all our 116 countries into two groups, the countries which do censor internet and it turns out that there are quite a lot of those countries, particularly in Asia but also in Europe and in Africa and the countries which do not censor internet and we see that all of this our average negative effect comes uniquely from the countries where the content of the internet is not censored. And we see absolutely nothing for the countries with censored internet. At the same time if we focus on those countries which do not censor internet there is a very important heterogeneity with respect to how free the traditional process how much of the information is available for other channels. In particular if we split the countries which have free internet into those with above and below media censorship of the traditional media we see that the effect is stronger in other words steeper in places where traditional media is pro-government it's not independent of the government. It's worth maybe noting that actually it's technically and for logistics reasons it's much easier to censor traditional press than the internet. In particular most of the ownership of traditional press in the world is very concentrated so it's much easier for the governments to control it. So with this let me try to understand a little bit so what does it mean? So why would negative information about the government actually affect government approval? So on the one hand of course indeed elites may control traditional media in many countries and an average country in particular and this means that pre-mobile broadband internet may be the view of the median voter and the average voter was too positive with the government because they just didn't know about corruption scandals and didn't know about the all the negative things about the government which means that this positive bias would be corrected with 3G. That's of course a very positive view of what's going on but one could think about it differently which is that actually it could be that the bias is rather coming with 3G. In particular we know very well that the quote from Martin Buri actually talks about this that social media is more prone to disseminating negative messages and in particular the whole business model is based on engaging the viewers and engaging viewers with negative emotions as usually much users I mean it's much simpler than engaging them with positive emotions and we do see indeed that negative messages disseminate faster and deeper than positive ones so if that's the case then that would mean that with the arrival of 3G there is a bias in public opinion towards being focused on more negative things and indeed social media were called outrage machines by many observers yet irrespective of whether which of these stories is true and actually in fact probably both of these stories are true to some extent you know if the government is super clean think about let's say Denmark or Switzerland then it's not clear why we would expect a negative effect of 3G right we would expect it to be an average negative but not in all countries particularly even if in Denmark people read on social media something bad about their government they also have a possibility to read about other countries and then they would learn that actually situation in other countries is much worse right so that may actually lead to an improvement of the confidence in the own government and the question is whether it's the case or not and indeed it turns out that in the super clean countries like Denmark, Germany, Japan, the Netherlands Norway, Sweden, Switzerland, New Zealand and the United Kingdom the effect on average is positive so then if you expand the set of countries further then you don't get that anymore and particularly in this slide what we show is the average effect of the expansion of 3G on various measures of government approval by groups of countries where we split all these 120 or so countries into 14 equal size groups by the overall corruptness level and we see that for the least corrupt countries it's consistently positive and it's not positive for any billions in fact when you enlarge this group of clean countries up until including US it immediately becomes negative so before US it seems like you could try to tease out these positive effects so so far what I told you is about this evidence seems to point out very strongly to the idea that it's the content which is available online about the incumbent governments which seem to be driving this effect but I told you absolutely nothing about whether this content is accurate or not and we are in particular very interested in whether internet actually helps to expose actual corruption in order to do this we use two measures of actual corruption and both are quite interesting the first measure comes from the wonderful work by Furseri and others who read through the Economist Intelligence Unit reports and basically collect data on corruption incidents so we use these data this is a panel at country level and we try to look at the following experiment let's say suppose that there is a corruption incident which suddenly happens in the country we would like to know how the actual corruption incident translates into the perceptions of corruption in government depending on whether the people live in places where 3G is available and or where 3G is not available and this is what I summarize on this very very simple graph we see that when corruption incident happens that leads to a decrease significant and large decrease and we believe that there is no corruption in home government in regions which are covered by 3G and much small and insignificant declined in this belief in regions which are without 3G which actually strongly suggests that 3G is important in delivering this message to the population so what we also do we use another measure which unfortunately available is only as a cross-section but it certainly was a surprise measure so that comes of actual corruption that comes from the revelations from Panama Papers many of you remember that in April 2016 we learned about the offshore accounts of many of the people in governments in different parts of the world for example it led to the dismissal or resignation of the Icelandic Prime Minister for example but there were many other stories as well so we use Panama Papers as a measure of this news about actual corruption and we again see that this news led to a much bigger and more significant decline in the belief of that there is no corruption in government for those places where of course these were the revelations of true corruption compared in regions with 3G compared to regions without 3G so again what we see is that when there is real corruption 3G has exposed it of course this is not the full story and the story of Jair Bantz-Belsanar it definitely illustrates that and I would like to mention it right away so the other important question to understand is could it be something else about social media which we are picking up in particular in the age of social media when we spend so much time online people may get just depressed generally may get isolated it's hard to think about this now in the age of lockdown because of the COVID but we were isolated even before that because of the social media so then we would like to we need to be sure that our effect on the government doesn't come through the general effect on unhappiness or let's say less of a positive outlook for the future and what we do for this we use the different measures of life satisfaction and the expectation for the dynamics of life satisfaction of people for the future and we see that there is absolutely no effect on any of these characteristics from the expansion of 3G so which means that this is really truly the political effect which matters it's not just the some kind of implication of the overall decline of unhappiness so with this let me very very quickly go through the other part of the analysis which we do in the paper which tried to ask the question what are the electoral implications of this in democracy so we have basically I hope I have convinced you that there is a disillusionment with the government in general which is caused by the expansion of 3G but you might want to ask so what is it real does it have real effects and for that we unfortunately need to scale down our operation and instead of looking at the whole world for the because of the reasons of data availability we look just at Europe so we look at full set of European democracies in the same decade which we've studied already so and we look at all parliamentary elections happening in these 33 countries which all together are 102 parliamentary elections and we look at data again at the sub-national level and it so happens that the time on the study coincides with the previous large increase in the popularity of populist parties in Europe and also not only the increase in popularity but actually rise to power of populists in many of the European countries so we also like to know whether these phenomena are related just before I come to showing you results I would like to also highlight the fact that even in Europe you know in the European democracies which are very developed countries during this decade on the study there was a very large expansion most of the time not from zero but still very large expansion of 3G so we have quite a lot of variation in the data to be able to estimate this with this we asked the question so are the incumbents heard by 3G and the answer to this is definitely yes so here I illustrate again our main result where we plot the vote share for the incumbent as a function of increase in 3G coverage at some national level in these countries and we see that these disillusionments overall with the government translates into lower vote shares for the incumbents next question is who gains and I'm aware of the fact that I need to wrap up soon to be able to find to leave time for the discussion so let me be very quick we try to understand what kind of position gains from the expansion of 3G and what we find is that it's in particular populists both on the right and on the left of the particular of the of the political spectrum who gain from the expansion of 3G and non-populist opposition there's no effect and if we for example focus on green parties which are also anti-establishment parties but not populists you also still have absolutely no effect however importantly if we look at subset of European democracy where populists actually already are in power we see that the populist incumbents are also heard by the criticism of the government online so 3G leads to decline in vote shares for the populist incumbents and that means that our story is not about populists our story is indeed about loss of trust in incumbents populist or not at the same time it is interesting why populist opposition gains and here unfortunately we can't say much because it could be both coincidental and causal so on the one hand it could be that 3G expansion coincided with the time and populist message resonates the most with the voters but it could also be the case that the some political messages of opposition are better suited for format of social media than others in particular populist messages are simpler they blame someone else with the problems they blame elite while for example Green Party is they ask voters to take responsibility so it might be much more complex and more difficult so we don't know why populist opposition gains and this is the we think a very important avenue for future research but nonetheless we are pretty confident that what we see is that indeed the expansion of access to mobile broadband internet is what reduces government approval and reduces votes for the incumbents and let me just stop here so that we have time for discussion thank you very much