 Thank you all for attending this public lecture by Mathieu Jackson. Mathieu Jackson is Professor of Economics at Sanford University and is the recipient of the Jean-Jacques Lafon Prize 2020. So we are very happy and honoured that Mathieu Jackson has accepted this prize in memory of Jean-Jacques Lafon, Colette Lafon, Jean-Jacques Pauze and some of his relatives are with us today and we are very glad to be able to perpetuate the memory of Jean-Jacques with them in the context of this prize. The Jean-Jacques Lafon Prize was created in 2005 by the Toulouse School of Economics in partnership with the City of Toulouse. It is awarded annually and rewards a high-level international economist whose scientific work is in the spirit of Jean-Jacques Lafon's work, meaning it offers fundamental insights and is rooted in important real-world economic and social issues. I would like to thank all the people who at TSE and at the Town Hall have made this event possible. There are a lot of people but let me mention Stefanie Risser and Florence Chauvet who have provided a wonderful administrative support and this was not easy as you imagine given the many adaptations they had to work on due to the health situation. I would also like to thank Professors Karin van der Straten and Michel Le Breton who have taken very great care of the scientific organization of this prize. So the public lecture of Matthew Jackson will be followed by a prize ceremony that will be in French in presence of Mr. Maxime Boyer who is deputy of the Toulouse Mayor Jean-Luc Moudinck. Before leaving the floor to Karin van der Straten who will present Matthew Jackson, please let me congratulate you Matthew as laureate of the of the Jean-Jacques Lafon Prize 2020 and let me thank you very much for your availability, your kindness that greatly facilitated the organization of this event and again thank you very much for presenting us your fascinating work on the dynamics of social networks and their economy consequences. Thank you very much Karin, the floor is yours. Okay well thank you Sébastien, it has been a pleasure to prepare this short introduction together with my colleague Michel Le Breton who happens to know Matt well, has known Matt for a long time since he's been co-stores on some work in political economy. So it has been a pleasure but I must say that presenting Matthew is also a very intimidating task because his list of scientific contributions and honoruses is just extremely impressive and so maybe let's start with a few biographical facts. So as he said Sébastien, Matthew Jackson is professor of economics at Stanford University. He received his BA from Princeton and his PhD from Stanford. He spent the first 10 years also of his career at Northwestern, the following Internet Caltech and before joining Stanford some 15 years ago. His list of honoruses is already very long, he's a member of the National Academy of Sciences, a fellow of the American Academy of Arts and Sciences of the Economic Society of the Game Theory Society and the list is long. He has already received several awards among which the John von Neumann Award and the Arrow Prize for Senior Economist but so today we're here for the prize in the honor of Jean-Jacques Lafon and so Sébastien reminded us of the prize and Matthew is really an ideal recipient for this prize which is a tribute to economists whose research much in the spirit of the works undertaken by Jean-Jacques Lafon combines theory and empirics to address important economic and social questions and obviously Matthew Jackson excels at this and maybe one first thing that is particularly striking when contemplating is scientific output is that as Jean-Jacques Lafon is certainly not the scholar of a single field, his work spans many different areas with actually a stronger overlap with some of Jean-Jacques Lafon's own scientific interests so just to mention a few game theory, implementation, markets, auctions, political economy and network economics so these last topics as a study of economic and social networks has perhaps become over the years the most important topic in his research agenda with applications to questions as diverse as international trade, military coalition, disease control, microfinance, labor as we saw an example in the in the seminar Matt gave last Monday but also legislative activities, economics development and it would actually be the subject of today's lecture. Another distinctive feature I would say is that as much as in the case of Jean-Jacques Lafon, theory occupies a very central place in Matt's approach to the social sciences in a recent paper entitled the road of theory in an age of design and big data Matt asks the following question and I quote with the growing availability of large and detailed data sets and the improvement in the computing technology and methodology to mine and analyze those data is economic theory doomed to extinction and he had one can find those who claim that the data will do also speaking and that theory will become obsolete so obviously Matt doesn't seem to side with his people and he makes in this article a very convincing argument for a use of combination of theory, empirics and experimentation an argument that he actually put into practice in his own works some of these being done with a number of with a set of impressive courses and so using various tools like structural econometrics a tradition which is also represented as a tool for the economics but also using field experiments as for example in the recent series of work in development economics with a budget energy and the study flow inter alia so to conclude I think I would like to strongly encourage you to visit Matt's website not only to get access and and to read I think over 150 academic articles and books he wrote but also because you can find their wealth of information on the list of very eclectic topics I found a video explaining to school children why a game theory is fun and important you can find tips to prepare for a talk you can find instructions for cycling in Sphiathians well not that it was useful to me but we've also learned with when preparing with in production with with Michelle and so Michelle has been in contact with Matt's wife Sarah and he learned from her that Matthew at some point was number five in the Californian mountain biker ranking use if we got our information right so that's that's that's that's very impressive not only has the academy so impressive that he's also an accomplished athlete so I think with that I will I will stop the introduction just a few words about the format for today so this public lecture will last until 6 30 p.m then there will be the ceremony award in in French so there are many people in the room the microphones in in the audience will be kept switched off and we will keep the the questions for the ends so there will be no question during the talk but the Q&A session at the end but you can start asking questions in the in the chat if you have some and Sebastien I will keep track and at the end when we'll open the floor for session we will give a priority to questions that have been asked in in the chat okay so I think that we we can start now so Matthew the floor is your very happy and the honor to have you today well thank you very much for that kind introduction Karine and Sebastien and and also Michel behind the scenes and all of the it's it's wonderful to see the family la font here and let me just say a couple of words I think you know the the first time I visited Toulouse was in October of 1990 and staying in a small attic near Anatole de France actually was Salvador Barbera was teaching there at the time and I could see the excitement that was being built and the the quality that was being attracted and and what was really impressive was the the coalition that Jean-Jacques Lafont was building between researchers and government and industry and you know really trying to solve important problems for society and I I really am impressed by how how strong the the foundation he built has been and and how it has continued to thrive under the leadership over the years and it's it's a great honor to be here so let me share my screen here and we will start with the there can everybody see that hopefully yeah yes thank you very much okay so I'm going to talk about networks and their consequences and some of the dynamics and as Karine pointed out this will be a combination of theory empirics and some field experiments and I'm going to start with just a couple of pictures because I think this is really the way to start to understand why networks are helpful in understanding human behavior and what I'm going to do is I'm going to show you a high school I'm going to show you a picture of friendships in an American high school and each little dot is a student and there's a line between the the two students if they're if they did at least three activities together in a week so this is from the data set I explored with Sergio Karine and Paulo Pin a few years back and what I'm going to do is I'm going to show you two two different pictures this first one is a picture of a fictional high school not a real high school and here I drew this by taking the the same number of friendships that are going to be in the real high school but putting them down completely at random okay so this is the high school with no real social structure to it's just a it looks like a spaghetti bowl right so it's just a you know a bunch of students random friendships and and the next one I'm going to show you is a picture of the real high school and and I want you to see the differences that you can see between the first high you know the the fictional high school that has random friendships and the one that has real friendships and the the way that this picture is drawn is by what's known as a spring algorithm so what it does is it starts just with things at random and then if it picks two students and if they're if they're friends it moves them closer together and if they're not friends it moves them farther apart so it's a way of of having a computer find the patterns in the network and what you'll begin to notice is two things compared to the first one one is that there's a strong split here so if you sort of go across the middle of this see if I can actually you know so there's sort of a sort of a north south divide here where above and below the you see friendships densely packed in these two different places the other thing to notice is that there's more inequality in the friendships here so you see people with no friends it's sort of a sad fact of high school and then you see people with many friends so there's people who have many friends people who have no friends and that's quite different from this one where there's no divide and everybody has at least one friend and it's much more equal in terms of the distribution so this means that there's there's patterns here and then let me color the nodes and I'm coloring the nodes by race in this case self-reported ethnicity of the students so they could report that they were either in this case black Hispanic white Asian and so forth the the blue dots are the ones who reported themselves as black yellow are white pink are at least two different categories red are Hispanic and then blue ones are missing data but what you begin to see is that this this high school is split along these ethnic lines and so these things affect behavior so they'll affect who knows what what kind of information flows through the network what kind of access people had and on on Monday I talked a bit about how this affects labor markets but more generally you know that the fact that networks have these identifiable characteristics has consequences and it affects things like job information today I'm going to talk a little bit about microfinance participation polarization diffusion contagions how these things play out in networks with these kinds of shapes and another theme that we'll talk about is that the networks themselves are influenced by the institutions and markets and things around them so how who we interact with and how we choose to make those interactions depend on what's going on around us and so we're going to look at the impact some of the feedback and dynamics and what's been going on in terms of technology so at the end today I'll talk a little bit about what what kinds of trends we see due to technology in the world today okay so in terms of an outline I'm going to talk about how networks can impact a market first then we'll reverse things and talk about how networks end up being changed by that market and then I'll talk about two competing trends that we see in the world today one is that we see increasing density and interconnections so we're able to travel more we're able to communicate across great distances that affects spreading but it also we're also seeing increased polarization and increased segregation at the same time and so we'll talk about these two different trends so let me start with the diffusion on networks and how that impacts markets and I'm going to talk about the some of the work that Karin mentioned with Abhijit Banerjee, Rune Chandraseekar and S.J. Duflo and I'll start with two studies that we did looking at at microfinance spread in India and this became a more than a decade long study now we've been at it for I guess about 14 years and the latest study we're doing is sort of how the networks were evolving but I'll start with some of the some of the background information and then talk about what we know for how these networks are changing and so the first part is you know the way that this study started is we wanted to figure out how networks were helping information spread about microfinance in an area and how to best get information out and to spread that information and in terms of basic background the reason that we got involved with we've got involved with a bank in southern India that was trying to spread loans among poor people and the problems that they were facing were that microfinance participation was varying very widely across different villages that they would go into so some villages they would go in they wouldn't get anybody to participate other villages they were getting a lot of participation and the villages were fairly similar and they were trying to figure out why it was that the news was spreading well in some places and not in others and they were using essentially a word of mouth dissemination and I can talk about that if people are interested but the the idea was that they would go into a village and try to talk to important people in the village and ask those people to spread information and they were hoping that the information would spread organically through the village and some places it was spreading well and other places it wasn't spreading and so one question is you know how important are these initial positions in the network and as we saw in that high school network there are people who have very different characteristics in terms of how well connected they are and if you hit the right people sometimes you can get news to spread and not otherwise so I'm going to tell you a little bit about the dataset so we were working in Karnataka which is southern India we were working in villages around Bangalore villages on typically had about 200 households per village and we worked with 75 different villages and the bank eventually entered 43 of these villages and offered loans so these are places that didn't have access to loans before and 32 of them they did not enter due to the financial crisis so in 2009 and 10 they stopped lending and so one thing we'll be able to do is we surveyed the networks and mapped out the networks before they entered and then after and we'll be able to see how the villages they got microfinance differ from what happened in the villages that didn't so I'll talk about that in a little while but first of all let me just talk about how they tried to get information about these loans out and how that works and a little bit of the role of theory in networks together with the empirics so the the per capita income in these villages is roughly somewhere between one and five dollars a day depending on the villages you're in the loans are on the order of 10 000 rupees so roughly 200 dollars at that time they were loans given to women aged 18 to 57 for 50 weeks and at interest rates that are roughly high credit card interest rates so a little more than 30 percent and they were what are known as Grameen style loans so that these are loans where women were put into groups of five and they were jointly liable for each other's repayments okay so this is sort of a standard microfinance loan and these loans even though they're small numbers for for these individuals it's actually a large amount of money and enough money to help them spread smooth their income and so forth so we were in karnataka this part of india roughly in a band around vangalore just to give you pictures of what some of these villages look like you know this is a picture from one of the village you know they're sort of fairly small poor agricultural a little bit of sericulture silkworm production here's another picture of one of the villages so these are the sorts of areas that the loans were going out in and what we did is we went into these villages and then mapped out various forms of networks so for instance this is one of the villages each little dot in this picture is a an adult and the groups of dots are households and here this is the these lines between the different dots are we asked the people if you had to borrow or lend 50 rupees for a day who would you go to so we asked people who they were borrowing and lending from and then there's an arrow between two households if those households are borrowing and lending from each other and then we asked a bunch of other things so this is the borrowing network in more detail but we also asked you know who do you go to temple with who would you go to pray with who do you go to for important advice who comes to you to borrow kerosene or rice so these are heating oils and so forth who do you go to in an emergency for medical help and so so from all of these different things we can build a network and what's going to be important we'll work with households as the units and we'll keep track of whether two households are in contact with each other or not so did they did they do something together and so from all these things put together we can track which households are in touch with each other here also is just a picture of this is one of the villages village 26 this is the kerosene rice and now what I've done is I've put all those dots together into the households but I've also split it by the caste designation so the blue are the scheduled cast and scheduled tribes the relatively disadvantaged groups and the red are the general or otherwise backward cast so the relatively advantaged cast and one thing you can begin to see here is you see sort of you know a split roughly between the you know the the different cast groups so you have relatively advantaged cast and relatively disadvantaged cast not having many friendships with each other so here you have about a 15 times higher chance of having a friendship within those within your cast side than the other you also notice actually that the scheduled cast and scheduled tribes sort of split down here so you can sort of split this village into three different pieces and and that has consequences for sort of you know how these people behave and what kind of information so there could be information that's flowing from person to person in this part of the network that doesn't get to this part of the network or vice versa so these patterns in the network will make a difference in terms of how information flows okay so the information passing that we looked at the first part of this was you know do the initial injection points that the bank comes in and talks to does that help spread the news about microfinance does it matter and how should we measure their role okay and I think you know there's there's importance of evidence that you know actually going back to a your symbol in 1903 talking about how important people's position in the network can be for getting information out and so forth and that's been um studied in a in a long series of different works and in sociology economics and other places but here what we're going to try and do is is understand exactly how you measure this and what was going on in these particular villages and this is where some of the theory will come in so the the first question you know in in networks we get to is how central is somebody and the most obvious answer is um just counting how many friends people have so how how connected are they and and people that with the most connections are thought of as the most central and this is something you know what on twitter people count how many users you have followers you have and and generally in facebook people you know track how many friends somebody has and linked in and so forth so there's you know just measures of how many people could you reach by by shouting out and degree centrality this just counting degree degree is just the number of friends that somebody has would pick out you know for instance these two nodes in this network one with seven and one with six and it's sort of a first count of of how well connected somebody is but you know if if we have a network where there might be somebody here for instance this person in the middle with with two connections this person is actually very well situated even though they have very few friends and so if we wanted to spread information in this network it's possible that this person would actually be a good person to talk to because they're well connected to different parts of the network which wouldn't come across if we just were counting friends where we we would miss it and you know this person on the very edge of the network with a two looks just as central as this other person with a two and so there's certain aspects that just counting degree misses okay and the the basic way in which mathematicians at first sort of thought about capturing this is something that comes out which is known as eigenvector centrality it was brought into sociology in 1970s by Phil Bonacic but the the the notion here is that somebody's centrality you don't just count how many friends you have but you count you sum up all the the centrality of my friends so I get importance not from how many friends I have but by adding up the importance of all my friends okay and so if I have more important friends that can be better than having lots of friends and so this is a basic calculation eigenvector term refers to the fact that you can solve this as a system of equations so here you have to solve this and there's centrality you know I have to determine what my friend centrality is to determine my centrality so you've got a system of equations and unknowns but this has a very well defined and well studied solution to it and so in this case there's a unique vector which tells you what these centralities are and it's a it's a notion that in this particular network begins to tell us things that are quite different than just counting the friends so if you look at the eigenvector centralities of these nodes in this network then you would say okay look this this one here is actually a 0.3 compared to this one over here is a 0.11 so the one on the edge has much lower centrality than the one that's that has the same number of connections but their connections are much better connected the best connection in this whole network is this one here with the with the 6 the 0.5 more so than this one over here with a 7 and so eigenvector centrality paints a very different picture than you know just degree centrality and interestingly enough the algorithm that was behind the rise of google as a search engine was built off of something called back rub which was effectively a variation on an eigenvector centrality calculation so google search engine originally was was good because it was telling you what things you wanted to look at by looking at the graph of all web pages on the internet and then assigning scores to them and ranking them in this kind of iterative manner okay so those were those were two options for sort of measuring how central people were in a village and whether those people would be good people for spreading information and whether the bank was doing well because it was hitting the right people in the village so for instance if it went into a village and it talked to this 0.31 person you would expect the information to spread much more than if it talked to the 0.11 person and so maybe that was explaining what was going on okay so those are two different measures and i think you know now everybody will be much more familiar with this part than than he were a few years ago but what we decided to do was say look neither of those measures measures actually pick up what we think of in terms of spreading behavior so spreading behavior of information looks a lot more like a contagion process and so we thought well why don't we define centrality off of a contagion process and this contagion process actually looks a lot like what's known as an sir model which now i think people are much more familiar with given covid but it looks something like a spreading behavior and so what we ask is how many nodes end up informed if some initial person in the in the network is initially informed and then each person who has that information randomly bumps into its neighbors and talks to them with some probability say p in each period and we run that for for some number of periods so there's some amount of time that people talk about this subject they randomly bump into people and happen to talk how does that spread okay and so let's think of let's suppose that you have a half a chance to talking to any given one of your friends and say in a particular week and we talk about something spreading for four weeks what would that look like so if this is the initial node and this was the network we could just simulate this we could say who do we you know how do we expect this to spread and so for instance you know with the point five probability and four different iterations this person might tell a friend this person now knows about microfinance so this person initially knew about it now this person knows this person tells a few friends this person tells another friend so after two periods it spread a bit after three periods it spread a bit more after four periods we would have 13 so we would say okay look this person you know told 13 people if this was the sort of process that was working and so for a given type of contagion process we could estimate how much information flow would this person be responsible for over a given time period and we could do something for a different person right so we go through we do the same kind of calculation and we end up with a six and so we would say this person is is different than the other one and so by by running these kinds of processes what we end up with is estimates of how important each person would be in a spreading process okay and so we called this diffusion centrality to capture the idea of how good is somebody at diffusing this kind of information okay so one thing that's interesting about diffusion centrality there's this number of periods that we're talking about how long is this going on and if it goes on just once then effectively all you get to do is talk to your immediate friends and so people who have more immediate friends are going to be more central than other people so if t was just one and we just did this once then that would be the you know it would look like degree centrality and if t goes on infinitely often then things have a chance to keep moving through the whole network and it turns out that if communication happens once then diffusion centrality is actually just proportional directly to degree centrality so on the one hand with very short periods degree centrality looks like how quickly things get out and you don't have a chance to iterate if communication occurs many times so as t becomes very very long then eventually things start percolating through the network and and diffusion centrality actually converges to eigenvector centrality and in between it's different so it spans these different ones and it looks different in the middle okay so what we what we're thinking then is the bank went into different villages and the bank's strategy was to try and find central people but they didn't have network information so the way that they thought of finding central people was to identify people that they thought were important so they looked for shopkeepers they thought they'd be in touch with a lot of people teachers and self-help group leaders so those were the categories of people that they looked for so when they went into a village they looked for those people and they told them look we're a micro finance organization we're coming in we're going to offer loans please tell your friends tell them to spread the news and then we'll come back and and meet with you and try and spread micro finance information and then we'll offer loans so what was happening was in some villages the teacher happened to be very central in other villages the teacher happened not to be very central and so by by going through these villages we can see who they told and then see how things spread and just to give you an idea so what we can do is is look at whoops how how well the information spread through the the villages and how that depended on different things and so we can ask how much of a micro finance participation can we explain by looking at these different characteristics just pure the village characteristics so some villages are were less segregated than others some were more segregated in terms of that that homophily kind of splits that I showed you earlier then we can use the degree centrality of the leaders so we can explain a little more than 25 percent of the variation by just looking at the village characteristics when we add how how many friends the teacher self-help group leader and shopkeepers had then we can explain up to a little more than 30 percent same thing with eigenvector centrality but then if we use diffusion centrality we can explain a little more than 45 about 47 percent of what's going on in the cross villages so we get much more explanatory power and in fact by using the diffusion centrality if we actually fit the the process in the t so we estimate how long people talk about things you can get this up to about almost 70 percent so if we are actually fitting the diffusion centrality and fitting that P&T so here this is with a P&T that are calculated in a manner that is sort of we explain to be robust across different applications but if we actually fit that we can get even a better fit so what's the what's the sort of message here the position of people makes a big difference in information flows and it makes a difference in terms of explaining whether or not you're going to man manage to get information out in in this case about a a loan program to the to the people in the villages okay um let me say just a couple of words about how you might so so one um issue that we faced in doing this was that we went into these villages and were able to map out these networks quite extensively that was a fairly expensive undertaking and if the bank was going to go to every household and ask them and survey them they could just offer the my micro finance directly so what they also wanted was um how how can we actually find central individuals without going into the villages and mapping out all the networks so if we don't have network information is there a way to find central individuals and so what we thought about doing was let's just ask people okay so so we're going to go into the villages and we're just going to ask people who would be a good person for diffusing information you know we weren't going to ask people who's diffusion central but we we wanted to ask them you know how do we find people who would be good news spreaders okay and so we went into 33 of the villages um and asked people who are the people that you would suggest that we talk to if we want to spread information okay and just to show you um we call those as gossips so we refer to these people as the gossips in the village so who would be a good person for us to spread news who are good people that for for spreading information and here's the you know so different households could be named so we went into and asked a lot of the households and so some of the the households were named by almost everybody in the village so there's you know one particular household everybody said look you have to go to that person that's the person you want to talk to if you want to get information out and here's the diffusion centrality of the of the households and then how many nominations they got and so what you can begin to see is if you find households that got lots of nominations they tend to be um fairly central in terms of diffusion centrality so even though we can't necessarily go in you know if i asked who should i talk to a two school of economics in order to spread news uh most people there would have a pretty good idea of who's a person who talks a lot and is well connected and would be a good news spreader um you could tell me and it might not be you know just sort of an obvious person in terms of position or something it it might be somebody who everybody knows is is sort of socially well connected and a good person for news spreading and so it seems that people are able to do that quite well and so then we went ahead and tested that and um we actually did an experiment so we've been working with the Haryana government and Haryana is trying to get um its villages to vaccinate their children and so we have been doing experiments on trying to get information about vaccination programs out inside these villages and so we went into 521 villages and what we did is we did four different ways in which we spread information about the vaccine program one we just picked households at random and we said okay look tell your friends others we went and did this gossip procedure so we went in and we said look who should we talk to and then we we went and found those households and uh who should we talk if we want to spread information then we found those households and said look we have a vaccine program tell your friends about it then we also tried just trusted so we asked who are people in your village that you think are really trusted individuals that everybody if they say something everybody believes it and would follow their advice so we had trusted and then we had trusted gossips so people who are both trusted and were named in the gossip as a as a person that would be central and then we we put in six seeds in each village and then we also had uh you know we were doing a bunch of other things too so we had reminders that went out and payments and so forth but in terms of the spread here's what happened in terms of when we went into the villages so this is sort of the per person we talked to how many um extra people end up showing up as part of the program a trusted person would generate about two two other people coming to the program on average um a gossip person would generate almost five and a trusted gossip would generate over seven so by going through and asking not just who is trusted but who do you think is central they seem to be picking out people who have high diffusion centrality and when we see this you know we see a marked increase in the participation rates that we got in the immunization program by figuring out who these people were that were both had a trust level and gossip okay so just a few lessons from these this first studies network positions have economic consequences there's lots of different ways to be central and they they matter in different different situations and in particular as diffusion centrality seems to be an important one for getting information to spread in in different settings and um people can identify these individuals so um people are pretty good at doing that and you know actually let me mention one thing this is a little bit surprising to me and partly because the there's a whole series of studies in the sociology literature that ask people to map out their networks so we give you a pencil and paper and then we give you that you suppose you go into a company and we say who's friends with whom or who communicates with whom you know draw out the network so um david crackard and others have done a bunch of work on this i give you that pen and pen uh paper it's very difficult for people to draw their networks so people are not very good they can name their own friends but naming which of their friends talk to each other and who talks to other people so if you ask me to actually map the network people are not very good at um and yet they're pretty good for picking out central individuals and in in one of these papers we actually build a theory for that and the theory is based off diffusion centrality which is you know if somebody's really central i'm going to hear news from them quite often so news from them will make it to me if people are really central and so then i will be hearing about people who are more central and i won't be hearing stuff from people who are less central and so the extent that i can keep track of that i'll be able to find those central people just by by being a good listener in a network so even though i can't map the network i can still identify who's spreading information well okay so you know diffusion on networks sort of impacts the market positions matter so the networks can help us understand what's going on in a market setting and the the part that was sort of serendipitous in this study was that once the financial crisis came and the bank stopped going into villages um roughly half these villages got microfinance and the other half did not and so what we could begin to trace was does the you know the availability of microfinance change the social structure of these villages and so that's the next part i want to talk about is just a little bit of the dynamics of can we see how a market changes social structure and you know there's a quote from ken aro from 1999 i think you know ken aro is one of these people that it's it is pretty much impossible to find an area of economics that he did not have um had not looked at at some point in time and so he has a quote here this leads to an important and long-standing question does the market or for that matter the large efficient bureaucratic state destroy social links that have positive implications for efficiency so once we put markets in place does that mean that people don't need to interact with each other as much and does that sort of destroy the social fabric which can be important in making sure that we um are cohesive in other ways and so um what we've done is this is a work with abhijit arun and ester again but also with emily breza and syntony synti kin and and here what we're looking at is um work where that we look in the same villages and we have the networks from 2006 so these are villages that we that you know we went into these 75 villages we surveyed them all 2007 to 10 this bank entered 43 of the villages and offered loans but it did not enter the other 32 so 32 of the of the villages did not get microfinance 43 did and then we went back in afterwards and we resurveyed the villages to get the before and after networks okay so we can see um how bigger the networks who's connected to whom how did these networks change between 2006 and 2012 and how did it depend on whether you had microfinance present in the village or not okay so what happens so these are the non-microfinance villages and a before and after and these are the microfinance villages on the right before and after and this is just the the frequency of links so this is just the you know on average how many other households if you look at two households what's the chance they're connected um in the non-microfinance villages before it was almost 10 percent um afterwards so in all these villages there was a little bit of decay in the network so all these villages saw some loss in their networks and and one thing that was going on at the time these are villages around Bangalore and there was some outward migration um so some of the villages were losing people to uh you know people were starting to work in in Bangalore or commuting around so so there was some general degradation of these villages just during this time period but what you'll see is that there's about a twice as large loss in the microfinance villages as in the non-microfinance villages so the microfinance villages get loans and people no longer need to borrow and lend from each other and that decreases some of the activity in those villages and i want to spend time you know digging a little deeper into that to explain sort of how that was changing and what the patterns of that change look like so um one thing we want to do is is uh in the microfinance villages not everybody got a loan it's not as if the whole village ended up participating in this so there are some people who ended up participating and other people who didn't and we want to be able to trace not only how does the network change but how does it change among the people who got loans compared to the people who did not get loans okay and so what we did was what's known in in the medical literature is propensity scoring and this is a technique that made it into economics i think in the 1990s but what we needed to do is the following so we have villages that got microfinance and villages that did not and let's suppose we have a household that got a loan in the microfinance village we'd like to know what would have happened to them in the alternative universe where they didn't get a loan they were in a village that didn't get the microfinance so we need to be able to match them so we want to look at a household in the microfinance village and say here's a comparison household that looks the same in a non-microfinance village that also would have gotten a loan had the had it been in in their village and so we want to be able to make these comparisons so we want to be able to match up which are the households that are very likely to get microfinance which are the households that aren't and so what we used is we used a machine learning technique known as a random forest algorithm to actually assign to go through and look at a household and say okay look we can see all kinds of things about this household we can see what cast they are what religion they are how many people are in a household what employment they have you know what the ages are we have all this information then we can use that to predict are they going to get a loan or are they not going to get a loan and you know their education levels and so forth and so what we could do is then break these into two different categories people who are highly likely to get loans and people who are unlikely to get loans so we call these highs and lows and in the microfinance villages the ones that we designate as highs so we run this algorithm and it picks out the highs they they have about a 46 to 47 percent chance of getting a loan the ones that we categorize as lows about 4 percent chance of getting loans so the demographics are pretty good at picking out and we also use network position and so forth so we can pretty well predict who's going to end up getting a loan and who's not going to get up get a loan and then we can do is compare the networks of highs and the networks of lows and see how they're changing in the microfinance villages compared to the non microfinance villages okay so it's a mouthful there's a lot going on here but we're trying to figure out you know which households got loans which ones didn't and how did it would what would have happened to them had they not had microfinance around would their network have changed in different ways okay so here's what happens in terms of the link probabilities so first of all we can look at the how many we look at the connections between highs and highs these are people that are likely to get loans and the the blue ones on the left are the non microfinance villages how how likely are these households to be connected to each other and then the green are the microfinance villages how likely are they to be connected to each other and what you'll see is pretty clearly all of the households are losing links in the microfinance villages compared to the non microfinance villages so the borrowing and lending network disappears in among the villages but it's not just among the high highs these are the people who are actually getting the loans it's also disappearing among the highs and lows which you would imagine but the low lows are also losing links and they're actually it's hard to tell from this picture but they lose them at the at the highest rate okay so in in terms of what's happening here we have a situation where the the highs end up getting access to loans we would imagine you know it's it's pretty clear that now they have formal loans from a bank they don't need to borrow and lend from each other anymore but it's it's less obvious why that should impact the rest of the village why are these people who are very unlikely to get loans many of whom are some of the poorest people in the village why are they not connecting to each other and anymore and this taught us some lessons so we spend a lot of time doing some some structural issues in terms of trying to figure out you know how can we explain this disappearance and what seems to be going on and also anecdotally when we talk to people in these villages there's a sort of an effort of socializing and what tends to happen is that you know people spend time in the central square just talking to each other they go to tea shops getting tea together is something very social and before microfinance everybody was doing this fairly actively and then once microfinance came into a village the people who actually got loans would tend to come less frequently to the square or to the tea shops and so forth and that made these places less attractive for everybody and so that means that other people started spending less time so they talk about oh you know we don't spend as much time in the square or the or the tea shop and so forth anymore and what happens then is you get less socializing not just among the people who are getting loans these sort of high probability people but also among the people who did not get loans and that carryover means that there's a broader effect on the village and you know this socializing maintains these old relationships and the way you meet new ones and there's complementarities here the more others socialize the easier it is to maintain your relationships the more active people are and and that means that you see things spreading not just from the highs but also to the lows and as one sort of extra piece of this we can look at the borrowing and lending network and then look at the non-microfinance and microfinance villages that you're seeing the drop in the highs borrowing and lending activity the lows borrowing and lending activity and then here we've got the advice networks as well and and so the fact that you see this happening almost equally as much in the advice network as the borrowing and lending network means that you see a change that is having spillovers not only beyond the people who got the loans but also in relationships beyond the types of relationships that are just financial relationships so in these villages a lot of people you know if I'm talking to you and I ask you for a loan or something or I'm also talking to you and maybe helping you out with your kids or giving you advice and so forth and so there's all kinds of other things that happen through these relationships and those things are disappearing as well okay so um one last piece to this puzzle so um there were two things about the karnataka experiments that were um somewhat uh that that we wanted to find out more about one is that we did not control which villages the bank went into in which villages it did not so it was not a randomized experiment so the bank chose which 43 villages it went into first and so we were worried that maybe there was some kind of contamination there um the other thing is that we did not collect information in in the end about how much income people had and how much consumption they had and so forth so we couldn't tell whether the disappearance of the network was actually having an effect on people's well-being and overall consumption patterns and so we also have um additional another study then that we are able to use data from where we went back into we there's the in Hyderabad um a microfinance study that was randomized where 104 villages got um into the study half of them got microfinance so it and this was roughly in the same period and what we did is we went into there afterwards in 2012 and gathered data in these Hyderabad villages but we also um measured consumption and income so we went in and measured the networks but we also measured the consumption and income and so from that we got um several things first we see extremely similar effects in sign and magnitude of the disappearance of the relationships so when we look across Hyderabad when we look at the 52 villages that got microfinance compared to the 52 that don't we see a similar drop in the percentage of links in the microfinance villages compared to the non but then we also have um information about consumption and income and when we look at the lows in Hyderabad what we see is um they don't lose income but what we see is an increased correlation between their income and consumption almost a two-thirds increase okay so what does this mean this means that these people um who are not getting loans but are losing relationships are actually seeing that their income um fluctuations are are pushing their consumption fluctuations more than they would have otherwise so we see um a situation where here we end up uh having an increased relationship between consumption and income which means they're they're they're not risk sharing as well so we see a decrease in risk sharing in the Hyderabad villages that got the um uh microfinance and among households that did not get the microfinance the non microfinance takers are um are sort of less able to smooth income here the H's don't see this increase so the people who actually get the loans um don't you know their their correlation between consumption and income doesn't change but it's the ones that don't get the loans so sort of you know lessons of this we see complementarities spillovers or economists we would call this GE effects sort of generally glibrium effects we're seeing movements um you know the the loans come in and that changes the social interactions that changes it among people who aren't involved in that market at all and it's um you know it's partly because these links are not independent people socialize and it helps us understand what's going on and sort of imperfect search for others to to relate to them is sort of why these low lows are affected um there's something called multiplexing in sociology and it refers to the fact that our relationships are multifaceted so it's not just that we borrow and lend money together but we also share advice or you know with your colleagues over lunch you might be talking about all kinds of things um and giving them information not just about uh market opportunities but other opportunities and so the fact that we lose some of the social fabric then we we care about this not just because of the social fabric but also because of the implications for other activities so let me just say you know some of the implications of this what it says is exposure to formal loans changes social fabric in measurable ways and it has unintended consequences so the lows are losing relationships even at a higher rate um that's increasing their income variants they're not the income I should say consumption variance for the lows and non-money relationships are are affected as well um it could increase inequality we don't have a good measure of this but we know that it's at least increasing inequality in the network itself um and I think the the message is not that we want to stop markets from moving into these places or we want to stop microfinance but that we need to take this into account in policy and mechanism design so we do need to understand how one one play uh one type of intervention can impact other things that we didn't intend it to and make sure that when we go in that we take care of the individuals for instance it might not be involved in that program because they might be affected um as well okay so um just in terms of outlines you know I think what we've looked at it shows that these networks have some consequences the networks are changeable they do change in reaction to the to the world and I want to talk just sort of in these you know last say 10 15 minutes about competing trends that we see in the world so there are you know sort of two things happening to our world at once and one is that we see it's sort of an increase in density and interactions and spreading around the world so you know we can call this globalization but basically we're able to keep in touch with each other right now we're having a zoom meeting you know across continents 20 or 30 years ago we would not have had this technology right we could not have this going on so there's an ability for people to connect at greater distances and that means that there's more interaction and more possibility of information spreading or other kinds of things but at the same time there's also a possibility of increased segregation and homophily so the ability to connect also comes with it some selectivity and I'll show you some data on that you know the ability of people to choose who they're interacting with sort of moves us in the opposite direction so we can become more connected and more segregated at the same time and I sort of want to go through some of these and I'll start with just some pictures and it's hard to find data sets that span centuries and networks so what I've done here is worked with what's known as the ATOP data set and this is from a study I did with Stephen Nye and this is going to be interactions across countries so now instead of people each node is a country and two countries have a relationship together if they were allies that means that they had some kind of treaty in place and we'll start right after the Napoleonic period so 1815 and then we'll just sort of go forward and I'll just show you pictures of what these look like at 10-year snapshots right so we start in 1815 and so here we have Russia Austria-Hungary Germany and so forth so we have a bunch of different countries they're all having relationships with each other here's France and then we'll just track this over time right so we have 1815 1825 35 45 right so we get into 1850s 60s 70s were a period where there was not much alliance going on 80s and what you can begin to notice is this network is bouncing around a lot right it's it's changing a lot we get to 1910 20 30s 40s and now you begin to see a big change 1950s 60s 70s 80s 90s 2000 and so you begin to see that you know by France down here you get different positions and so forth but you see a much much denser network okay and so these alliances have been getting much denser over time there's many more countries that are allied now than they were before and one thing that this correlates with quite strongly is the trade increase in trade so right roughly in the 1950s is also where you see trade costs decrease container shipping shipping improves you see a dramatic decrease in trade costs and you see more and more countries trading with each other and so beforehand pre-world war two if you go back you know before the 1940s and 50s countries had on average two and a half allies two-thirds chance they last five years post-world war two they're much more connected 10 and a half allies about a 95 percent chance that any given alliance will still be there five years later so you see a denser network and a much more stable network and that you know this increased trade relationships has led to increased as we see in these networks we see increased alliance networks and the other aspect that came with it this is from our our data with with Stephen that again these are what are known as courts of war mid-fives which are military interstate disputes of the level five so these are conflicts between countries that involve at least a thousand casualties so think of them as it is some kind of war and what you see is you see about a tenth as many post-1950 as you saw pre-1950 so then you know you saw this period right now we're in the most peaceful period that that humanity has seen at least in the last two centuries by a lot and and part of this is this you know globalization has led to increased trade and in the paper we we go through a lot of sort of you know time varying econometrics we don't have exogenous variation here that allows you to to do a causal inference but what we can show is that the timing of trade and so forth correlate pretty strongly with a hypothesis of you know trade leads to countries to have common interests which means that they don't want to attack each other and they'd even want to protect each other and and that once you start putting it into a network context means that you get a more global society which ends up being much safer in terms of interstate wars so increased trade alliances decreased conflict these things are all closely tied to each other and you can see that in the networks so these global effects are sort of interesting on the one hand you get this increased trade alliances decreased conflict that's all great you also get increased ability of shocks to travel greater distances so defaults by banks due to subprime mortgage in the US can influence you know french balance sheets or greeks default on loans can affect french banks balance sheets and so you have you know interplay between different countries and movements of economic shocks across borders now and as we've seen recently with covid you know sort of I think a fascinating number the black plague if you look in the 14th century and you look at the time it took from the for the plague to get from marseille to stock home it took four years for it to get that distance covid it took about a month for it to spread pretty much around the world right so so you know along with this kind of increased interaction and increased trade you also get transmission of shocks and so it means that we're really in a global world now we're not in a world where you've got separate separate different areas interacting on their own so we've got you know this these networks are spreading more widely they're spreading around the world and the last thing I want to talk about then is that we're also seeing more and more of this homophily and segregation in these patterns at the same time and I can talk a little bit about what some of those pressures are in in those directions so let's have a look here you know we've looked at two different forms of homophily right we saw this in the in a high school network we saw it among castes and indian villages it's pretty much you know everywhere you go you will begin to see these kinds of splits in networks and you see the these splits quite poignantly when you begin to look in networks I talked about on on Monday a little bit some data you know you can see them in Facebook and so forth let me show you one study which I think is a very telling one this is by a team from Facebook this is a lot of Adamic Bakshi and Messing it just came out in Science a few months ago and what they did was they went and characterized people as either conservative or liberal in their political leanings so they were looking at at Facebook data and they're looking at people who are either conservative or liberals and they want to see how much access does a given person if I'm a liberal how much conservative information do I see postings that would be conservative in nature and how much liberal stuff do I see and if I'm a conservative how much you know cross cutting so cross cutting here percent cross cutting content means if I'm a liberal how much conservative material could I see and do I see and if I'm conservative how much a liberal content could I see and do I see so am I seeing stuff from the other side of the political aisle essentially and what what they're so the categorization was roughly 40 percent of the people they could look at as conservative I won't go through all the techniques they used to categorize people but so they they have 45 40 percent conservatives about 45 percent liberals and there's about 15 percent independence and now so if you are just randomly you know spreading information and shares you know things that stories that people share then everybody should see about 45 percent liberal content and about 40 percent conservative content okay so that would be sort of the mix you would see and then what they do is they track um what's what would you actually see given your friendships so the what are your friends posting and now you see this drop so as a conservative I'm only going to see about 35 percent of my friends would ever be posting liberal material and as a liberal only about 25 percent of my friends would ever post conservative material and then when you look at what I'm actually exposed to from the algorithms and then what you see what I actually click on by the time you get to clicking the liberals are seeing only about 20 percent conservative and the conservatives are seeing about 28 percent liberal and the reason that these are differing a bit is because the um independence who aren't shown here are are sharing more liberal content than conservative uh content so the the independence tend to be friends with both of these and they tend to share more liberal content but what you see is you know by the time you get through the network and then through the algorithms that are directing the the information both through the network you begin to see a lot more segregation of the information flows than you would have seen had people just been broadly exposed to stuff in the network and so this is sort of a I think a study which points to these ideas of echo chambers quite clearly in the fact that the people around you and what you get to see through your social media might be a very different slice of what's out there than what's um you know what's available in the broader world and just one picture which I always find fascinating to see whether or not this is playing out in politics um so this is a picture of the U.S. Senate and they're color coded by political party so the blues on the left are the democrats the reds they looked a little pink on the right are the republicans and this is code I got from a computer scientist and um what this does what I did is there's there's a connection here between two senators if they voted at the same way on bills at least half the time so they're agreeing more often than they're disagreeing and then they'll be connected and in um this is 1990 about 82 percent of the senators were linked to each other so most of the senators were agreeing more than than disagreeing um this is 2015 so this is before trump was elected and now you see and again this is you know the the computer drew this picture in terms of figuring out where the nodes should go and where the split is so I didn't pull these apart um now you see only 53 percent linked and many more of the links are within party than a cross party and so you know you can begin to put in your favorite Mitch McConnell and fine saying here's Sanders Durbin Schumer Graham Rubio so you begin to see them splitting and they they form positions in these um in this network you see a much more split network and certainly this isn't unique to american politics um finding french political maps is a little harder but you know this is just a a quick picture of you know if you look around France on the 2017 election who is the biggest vote-getter and by region and you know Macron in certain regions the pen in other regions fionne uh Mélenchon you know you can look in different regions and and so you do see splits and so forth and in where people are located and who they're interacting with and what they might be exposed to in terms of political views and that plays out in the politics so just as a last slide um one thing I find fascinating to see you know we we see more um algorithms interacting in our life um this is a study from Mark Rosenwald uh and colleagues a sociologist at Stanford and what they did is they tracked they've been tracking people in a survey for many years they look at people and their romantic partner and ask how did you first meet and if you go back to the 40s 50s and 60s it used to be friends and family and then it's school and if you look starting in the 1990s online takes off so more and more people first met their partner online I mean now it's the majority and the plurality and in fact um you know bar actually goes up as well um which has to do with some of these apps are used you know like people actually use the apps in bars to meet people but uh you know this is saying that that things are more mediated and I think some of the technological changes and challenges are that these platforms they benefit from our attention and they're competing with each other and those algorithms are built to sort of find people that look just like you to be your friends and try and put you together to match you up along you know dimensions that you are looking similar to that individual they want to offer news that resonates with you so they they try and detect your news your political leaning put the news in that direction and that can exacerbate the homophily in these sort of echo chamber tendencies so you know what um you know these networks matter the information flows we talked about jobs on monday microfinance trade and peace they have serious economic consequences they're being changed by the advent of markets and institutions we don't have a good understanding of that yet and I think even something we understand less well is the fact that technology is mediating more and more of our interactions and like it or not the the the instructions and the algorithms that those platforms have of how to connect people and what to share are making a difference in what we see and and who we interact with and we need better understanding of these dynamics and feedbacks so i think that's a good place to stop and uh merci beaucoup thank you for your attention okay thank you i think you're you're muted carine sorry i sort of muted okay so thank you for for this uh fascinating talk i think we would have love to have it go on and continue um so i've seen that there are a number of questions in uh in the chat um so maybe it's best if people ask questions directly um so i see that there is one from john i see i see john on my screen too can i so i think should be a mutated john are you or not we are not you're muted so yeah i'm not muted anymore so i can read the questions i can read it uh so it's on the normative front going back to the villages in india and the idea that actually socializing is a good thing which by and large i will agree with but of course we also know that we leave uh villages for cities because villages also are very oppressive you have to signal all the time you have to be nice to everyone you have to conform and so on is there a way in your um in your sample uh in your data to to actually tell apart what's good and bad in terms of socialization i mean the links you are cutting and i guess you are going to tell me about the low low or something like that but still even the low people they they might still feel that because the high people are in the village square they have to go because otherwise they wouldn't get something like advice or or or loan maybe sometimes uh i don't know i mean you know and there is a question also from peter which is similar actually on homophily um is that is that a bad thing i mean obviously from what you're telling us it's it's right to be bad thing but there might be cases also where it's a good thing right yes yes yes certainly so i think um the homophily you know and and these links have multiple purposes and it's very difficult to to to map out the overall welfare implications and i think you're exactly right so we do see situations where for instance the lows are seeing higher higher variance in their consumption so we do know that we can have one measure by which it looks like they're getting hurt more in the microfinance villages than non that doesn't mean that there there's you know overall welfare is is worse or that other aspects of their life aren't going to get better and you know that the the there's always these transition periods where for instance in in these villages you know bangalore is expanding opportunities to to get employment outside the villages is increasing and that changes the social fabric and in basic ways that we don't understand and i think you know the the main purpose of this study or the only thing we can conclude is that there's non-trivial spillovers and sort of general equilibrium effects that are not second order they're really sort of first order effects that are happening in these areas and so we need better measures of all the different things that happen through these relationships and and certainly homophily can be a very good thing in some regards right so you know if if i'm a young parent talking to other young parents are probably the best people i could talk to about advice for what to do when my child's sick or you know how to take care of my children and what to teach them and so forth so you know homophily can be very helpful in connecting people who have useful information for each other and exchanges but it also means that then we're also insulated and i think one of the key things to to get out of this is that these networks serve many many different purposes at the same time and so that person that you're sharing parental information with could also be the person that could be giving you a job opportunity or something and and maybe they're great for getting you know tips about the kid but then they're not very well connected to to give you information about a job and so forth and so so the homophily cuts both ways and you know i i think having a deeper understanding of all the multitude that of things that these networks influence in terms of our lives and how they interact with each other is something that we we're just beginning to scratch the surface of okay thank you thank you matt i think we have a couple of other questions gems maybe i think you're on youtube oh hi thanks my question was a little bit technical about the the eigenvector connection the centrality seems to be a count of the number of friends that my direct friends have and it seems to me it would be important whether my friends have the same set of friends or have disjoint sets of friends and so there could be some double counting going on there and then how does that relate to the the other measure of centrality you had yes a good question so so you know what but the beauty of of eigenvector centrality and diffusion centrality in terms of i think some of what they're capturing about the network is that there's other things going on as you're pointing out not just how many friends i have and how many friends my friends have and so forth but are my friends friends pointing right back to me are they pointing to each other or are they branching outwards and if you really want to maximize eigenvector centrality or diffusion centrality the the point at which it would maximize would be something that looks like a tree where it's actually growing outwards and none of the links are coming back to each other so both of these things in terms of looking at how many people you reach and so forth are maximized and sort of expandographs that look like trees and and grow outwards rather than ones that have lots of links that come back in on each other and so those are implicit in those calculations and part of the reason that i think you know you see a lot of different measures for centrality is all of these things matter and it's tricky to you know to keep track of how they're all interacting and and they're they're embedded in in those calculations okay thank you so maybe we've got time for one one last question maybe jack hi uh thank you very much mad this was a great talk there was one point which i found a bit disappointing uh it was when you began speaking about the policy lessons from your work on microfinance i mean it sounded a bit like no i participate too much too much round tables on digital firms and everybody says we need to study more we need to think more about it and we need more research so if you had to make a i know you can't do a scientific work on this but if you had to do a bet you know how would you use what to know now in order to modify the way microfinance is distributed yes um i think what i would do is try to make the loans so i think you know one one issue about microfinance is it comes in one size fits all loans and so those loans are attractive to part of the population and not to the other part so for instance the highs end up getting the loans and the lows don't but you know part of it's a mismatch and so if you could increase the heterogeneity in the sets of loans that might help the the lows smooth their borrowing and i didn't actually show the borrowing numbers but the lows borrowing goes down and their consumption variance goes up and so part of the reason that they're seeing that increased variance is that they're not seeing the financial the money lender opportunities that they have are very high interest rates and they're seeing less social connections and they're not taking advantage of the microfinance and somehow you want to be able to get income smoothing to them so it would be targeted income smoothing and figuring out exactly how you can get um that that would be the most immediate economic stuff on the advice networks and so forth then i i just you know i don't have an answer as to how you replace that but you know the um i i think it's it's pretty clear that somehow they're not getting the money injections and the ability to smooth their income across time that they need and so it would be thinking that you know these programs are only going to be taken up by a subset of the population and you want to make sure that that is reaching a greater set of people and maybe one instrument is not enough to do that and it might also involve some sort of subsidies or consumption subsidies at certain times of year you know in these villages part of it is that these are farmers and so they have to put every all their income into investing in seeds and so forth one part of the year and then basically if they don't have some way of smoothing income they starve for part of the year until their crop is ready and so they go through this really down period in terms of consumption and that's where having a microfinance loan allows them to smooth it but the other you know the people who don't have those loans aren't able to smooth that income okay um so thank you Matt so i seen the chat we also had a question from Peter Bayer but it was kind of incorporated also in in Jean's question um there is also one from uh chat mason uh could you please raise your hand so i can find you among the participants um there were 100 people uh sorry i can seem to find you uh i mean raise your hand with uh with uh with a tap uh to raise hands otherwise i will just uh read the did that work did i it was that worked okay um so i'm i'm really interested first i should say thank this was this was a fabulous talk i had my choice of three uh webinars uh i'm in the mountain time zone in the us uh my choice of three seminars at nine o'clock and i picked well um so what what struck me here was the possible tie into uh wealth of income disparities over time the increasing income disparities that we seem to be observing microfinance strikes me as just increased economic opportunity and to the extent we've got these very wide differences in income opportunities going forward is this i'm wondering what the really dark implications might be for a place like flint michigan that gets left behind and do the is it beyond the economic effects is it is it a complete collapse of the social networks or a large-scale collapse of the social networks and then what kinds of uh implications might that have across the board along the lines of the torrent social fabric seems like that that's kind of the next that's got to be the next step in the in the progression of this logical argument right right and i think um that's a very important point and and uh we right now we're doing a study where we're looking at at people's social capital and i talked a little bit about that on monday where we have a large data set we can track how well people who are relatively poor are connected to people who are more advantaged and how well that does in terms of their mobility and their job opportunities and in fact um i think what what does happen is in settings where like flint michigan or something where people's networks are fairly cut off the kind of social capital for instance that robert putnam talked about which is very introspective you know so it's you know there's a community there's a community very dense well you can have a very dense network but if the if the network is very dense in a very poor community with no with no connections to the outside that doesn't bring in the information and the opportunities that are needed for that community itself to to do well and so you can see i think you know part of this the puzzle that about recent book i wrote the human network where i talk about the role of social networks and growing inequality and i think the social fabric does exacerbate a lot of the other influences that are pushing towards inequality and understanding the fact that people can be trapped in these networks it can be a very powerful device for constraining people and something that we should understand in policies going forward to try and improve this because it's it's not easy to to to fix and i think you know just putting money into these areas isn't the answer you really need to help the the people have the information they need and the connections they need to to break out of this and that's a much deeper puzzle okay so um thank you uh thank you very much to uh to to all um i think we're we're reaching the end of of the session so thank you again to Matthew and to all the participants and give back to Sebastian thank you thank you very much Karen yes thank you again Matthew for for for this wonderful talk it was really a great pleasure to to listen to you and and thank you all for participating in this event it was uh you were more than 120 people at one point and and i think it was very very interesting to have the opportunity thanks to this technological network that we that we built to have such a diverse crowd and and very wide access over the over the world um so now uh for those of you who would like to with us yeah excuse me Sebastian i just seen one message from Jennifer Stephenson uh she was asking whether it would be possible for everybody to switch on their video so that we can take a group picture even if you are in pyjama okay even if you are in pyjama that's fine it's just for the group picture so thank you very much oh uh just to have some candidiality also in this event even if we are not all together i think it's it's great thank you Jennifer for this great idea yeah it's just for one minute just so we can take a picture and uh keep it for other records yes thank you very much you tell us when when it's good okay we can i think we'll just let them click on the screens for two seconds okay great thank you long i just clicked through the different screens nice to see everybody oh i can do this too actually okay i think that's fine thanks everybody thank you very much ready thank you Jenny yeah so yeah i was just mentioning that we are now going to proceed with the uh with uh official ceremony for the for the Jean-Jacques LeFond Prize 2020 and so we are going to to switch to French so for those of you who would like to enjoy some foreign language or enjoy French you are welcome to do so and uh and if not we we hope to you enjoyed a lot this this event this this public lecture by by Matthew Jackson and we hope to see you soon uh at TSC hopefully uh the the sooner the better okay thank you thank you very much i'm going to maybe now uh check that uh our guests of the Mr Boyer is here yes indeed Mr Boyer is here it's great thank you very much thank you very much stéphanique i also see that Colette is doing it and there are many Colettes and it's really very nice to be among us um so uh i i'm i'm going to do a little word reminder i'll come up again so to welcome you all again i really want to thank you very much for attending the award of the Jean-Jacques LeFond Prize 2020 We are very happy, very honored that Matthew Jackson, who is a professor at the University of Sanford in the United States, accepted this award in the memory of Jean-Jacques Lafond. Collette Lafond, the wife of Jean-Jacques, who is with us and also with some of the relatives of Jean-Jacques and Collette, attended this event with us. We are very happy to be able to share the memory of Jean-Jacques together in the context of this award. The award, Jean-Jacques Lafond, was created in 2005 by the Toulouse School of Economics, in partnership with the city of Toulouse, and it is awarded every year and rewards an international Hauvold economist in the scientific work in the spirit of those of Jean-Jacques Lafond, offering fundamental contributions and which are inspired at the same time by economic and social issues and from the real world. So I sincerely want to thank the Toulouse's merit for its continued support over the years in the context of the award of Jean-Jacques Lafond, but also more broadly in terms of the landscape that you like of the Higher Education. Mr. Maxime Boyer, to my mother, from Toulouse, Jean-Luc Moudinck, it is our honor to be among us today and will intervene during this ceremony. So thank you very much, Mr. Boyer, for accepting to say a few words on the occasion of the award of Jean-Jacques Lafond 2020. I also want to thank all the people who at TSE at the Toulouse's merit have made this event possible. We see that it has worked very well until now, we hope that it will continue. There are a lot of people that I could quote, but I would like to mention Stéphanie Rissert and Florence Chauvet, who have brought a great support to certain organizations. It has not been easy given all the changes that we had during the last few months to adapt to the health situation. And I would also like to deeply thank the professors Karine van der Straten and Michel Le Breton, who have taken great care of the scientific organization at this price. I would also like to thank Jean Tirol, who is the president of the TSE and Christian Gaulier, the director-general of TSE, who contributes and works at the scientific excellence of the TSE project. It is a great pleasure to be gathered together tonight to celebrate Matthew Jackson and also Jean-Jacques Lafond's memory. So, as it happened just before the public intervention of Matthew Jackson, Karine van der Straten makes us happy to say a few words before the formal price. And I would like to thank you again, Matthew, and I address you in French because I know that you speak our language well. So I would like to thank you as the laureate of the Jean-Jacques Lafond 2020 price. And also, I would like to thank you for your availability, your kindness, which has greatly facilitated the organization of this event. And a big thank you to have shared with us the conclusions of your passionate work on social networks and their economic and social consequences in a more general way. So thank you very much. And so I will pass the floor to Karine, who will say a few words on Matthew's scientific journey. Thank you, thank you, Sébastien. It is obviously a pleasure and an honor to make this introduction. Let me say a few words in French of what we already had the opportunity to say earlier at the beginning of the presentation. So thank you also for the opportunity to dive even further than until now in the extremely impressive and very rich production of Matthew. So to just remember, at the beginning, a few elements of biography. So Matthew Jackson is a professor of economics at the Stanford University, who joined him after ten years at North Western and then at Caltech. The list of honors that have been given to him is already very, very long. He is a member of several very renowned scientific societies, like the National Academy of Sciences in the United States, or he is a member of the American Academy of Arts and Sciences, a member of many scientific societies that have already received several awards, including the Fundman award and the Harrow award. It's really a pleasure for us to give him the award again, Jean-Jacques Lafon, today, because by the multiple aspects of his work, Matthew excels at what Jean-Jacques Lafon saw as extremely important dimensions of the work of economics, was to help contribute by the theory and the empiric to the best understanding of economic and social phenomena and to give us the means to act, to propose more efficient policies. So his scientific work covers a lot of fields and, more particularly, in recent years, the economy of networks. So obviously, the networks are everywhere. The networks relate people to the networks of friendship, the networks of business, the networks of politics. And what we were lacking so far was to have the right economic tools to understand them, to analyze them. And what Matthew did in his work was to provide us with the tools to go further in the analysis of these networks. And by doing so, he was able to revisit a number of important questions in economics that brought new intuitions and new results. So today's public reading has shown a few examples, but it has allowed us to see how important questions are such as microfinance, public health, international trade, or the work economy, regulations and the work market, to take into account these analysis of the networks and what we were able to learn from other disciplines such as sociology, to change our vision. So once again, it's a really impressive work and it continues when we go to Matthew's site and we see the list of work documents in court. It's very impressive. So, in ten years, there will probably be a long list that will be added. Thank you. Thank you very much, Karine, for this presentation. It's Matthew's work. I think we can maybe ask Mr. Boyer to say a few words if you wish, Mr. Boyer. Thank you. Thank you, Mr. Director. So Mr. President of the Toulouse School of Economics, Mr. Jean-Tier Vole, Mr. Director, Mr. Sébastien Cougé, ladies and gentlemen, the members of the family, Jean-Jacques Lafon, Mr. Professor, Mr. Mathieu Jackson, ladies and gentlemen of the educational community and of the Toulouse research in particular. Hello or rather good evening. I would like to apologize to Jean-Luc Moudin, the mayor of Toulouse, president of Toulouse Metropole, who may not be ours tonight, but who warmly greetings you for this award. To you, organizers, to you, members of the jury, and you, of course, Mr. Professor, Mr. Mathieu Jackson. So I am very happy to represent Jean-Luc Moudin for two reasons. On the one hand, because I am joined by my mother in higher education and that this question of research in Toulouse is an important element of our territory and of our attractiveness. But also because I am an former student of the Toulouse School of Economics. And so indeed, it makes me happy to see a few professors that I was able to have when I was in the course of my teacher here on this vision. So, the price, Jean-Jacques Lafon, is an important price for Toulouse. It is an important price for Toulouse, which was created in 2005 in partnership with the Toulouse Merit and which values the Toulouse research in particular, but which also values all the research and all the economic science throughout the world and allows each year to compensate for the great economists who participated in their work in scientific research. We can mention a few names, Joseph Stiglitz, or last year, Marianne Bertrand, who was laureate in 2019. I do not list the whole list because it is quite long and it would be maybe a little too long. Mr. Haute-Jackson, we have perfectly described to you a few seconds ago, so I may not be able to make a small list, but you have a brilliant career. You have been and you are a professor in a prestigious university, Stanford, your work participates in the understanding of the world today, in particular the question of social networks, which is an important subject and which is a current and daily subject, and we know that its evolution will also be very important for the future, this subject of social networks. We therefore see the importance of your relationship, of your intellectual relationship, in reality to the understanding of the world today and also to understand how it can evolve tomorrow. You have been the author of a book which is considered by researchers as a reference book for the economy of networks. It is the Social and Economic Network. It would be, in quotes, allow me to express it without being able to express it to anyone. It is a bit like the Bible of the 21st century, since today we are still very digitalized, and it's even a bit of a shame that you are not able to be in Toulouse today because you see behind me, in reality, the illustrious room which is one of the most beautiful monuments of the city of Toulouse. So it's a bit special where you are virtually, but I hope that one day, if you are going to Toulouse, you will have the pleasure of visiting the hotel of life and in particular this beautiful room which illustrates the excellence that we have been able to preserve through history and which makes the mark of our city. So Toulouse is a city, is a territory, in reality, of excellence in economic matters. We are the top 6 European cities of the territory of the innovation economy. We created, over the last 20 years, about 150,000 jobs. The economy, indeed, is a theory, but it is also a reality of the daily life. The city of Toulouse supports the school of economic Toulouse, obviously because this school is one of the flora of academic research and the fundamental research of the T-shirt Toulouse and which makes the excellence of the higher education at Toulouse. And then, given the situation of the sanitary crisis, the city of Toulouse and Toulouse-Métropole is involved in a particular process because this sanitary crisis is not only a sanitary crisis, it is also a social and economic crisis. And so it took us, local collectivities, to invest. We are launching emergency plans to help companies, especially around 30 million euros. We are launching plans to fight precariousness for all the people who are going to suffer from this COVID-19 situation. We are also launching a plan to relaunch metropolitan. In short, when we make the sum of the investment that we are going to take on public investments, that on support through social performance, that on support to the local economy, whatever the forms of support, to tax relief, or in relation to money in the territory, we are on a total of 143 million euros. This is the moment that he is exiled because we have to be in a proactive and pragmatic policy. But it is also a moment, this situation of the COVID-19 2020, which is also a moment of reflection, which must also be considered as the moment where we can also, where we must collectively ask ourselves the question of how the world must evolve tomorrow. That's why, without political privilege, because Rodelga, the president of the region, and Jean-Luc Boudin, the president of Toulouse-Métropole and mother of Toulouse, have decided to launch and to actually ask a certain number of experts, specialists, to be able to help us to improve our public policies for tomorrow and to engage our territory towards better future, I would say. That's why an independent commission has been created, which is called Toulouse-Territory of the future, which made a first report under the presidency of Marion Guillou and under the high-paying role, I would say, with a certain number of experts and who made proposals, proposals to increase the Toulouse-Scientific region to develop the Toulouse-Métropole strategy for the climate, to control, to comfort, to transform, to diversify the Toulouse-Industrial activity or to develop the tourist attractiveness. In short, a certain number of experts who will influence our tomorrow's decisions at the level of our collectivities. In reality, all that to tell you, Mr. Professor, you have certainly participated in the development and you will participate and you will still participate tomorrow in the development of the scientific research, but your work, to you as an entire economist, also participates in influencing us when it comes to our economic decisions and our decisions of economic policy at the local level, like at the national level, by the way, must be able to sit down on reflections, on thoughts and on added value, intellectuals and economists to be able to gain efficiency and also gain social justice. In short, your relationship is not only scientific, it is also a relationship for the general interest and that's why in reality I am here today to participate in this award ceremony and congratulate you for all your contribution, which is certainly a scientific contribution, but also a general interest contribution for the whole of everyone in reality, here and elsewhere, I was going to say. I thank, of course, the whole of the participants and in particular you, Mr. Jean Tirol and you, Mr. General Director, with the whole of the staff that helped you a lot and that allowed the realization of this event. In any case, congratulations, Mr. Mathieu Jackson for the whole of your contribution. A big thank you, Mr. Boyer, for his warm words and very inspiring words. In normal times, if we had all been gathered in this magnificent hall of illustrations, we could have physically put to Mathieu Jackson his prize, in the presence of Colette Lafond. This prize is materialized, this materialized in a bust of Ariane that we could have offered. Mathieu Jackson, we're going to let you go through it, we're going to send it like that, you will be able to see it very soon. And of course, we would also, with Mr. Boyer, we would also give you the cheque that symbolizes, let's say, the price in Jacques Lafond, which is generously offered by the Toulouse's fair. So once again, thank you very much. In normal times, we would have also been able to go and share a little moment of convivialization together, but today it will not be possible, but we will perhaps have the pleasure of hearing the few words from Mathieu Jackson. Mathieu, do you want to... Yes, of course. Thank you very much, especially to Colette Lafond and Mr. Boyer for being here and to Sébastien, Karine, Michel, Jean-Jacques, all the people of Toulouse. It is a great honor to receive this price in memory of Jean-Jacques Lafond. And thank you very much to the Toulouse's economy school, to the foundation and to the city. As a theorist, to pronounce his work in the 80s and 90s, the works of Jean-Jacques Lafond have shown me how deep research could clarify important issues for the well-being of society. Jean-Jacques Lafond's efforts to build the TSE, whose department of origin and to attract the best talents of the world, are inspiring. The success continues and the influence of schools on local, national and global politics is a model of the impact that economic science can have on the world. I am deeply moved by this price and I am waiting impatiently for the moment when I will be able to come to Toulouse again to visit my colleagues and friends in person to see the hall and to discuss the future after the sanitary crisis. Thank you to the TSE, the foundation, the city, and also to my parents, brothers and sisters, my daughters and especially my wife, Sarah. Thank you very much. Thank you very much, Mathieu, for your very, very nice words. And so, as mentioned by Mr. Boyer, so Jean-Tiro, Christian Golié, the general director of the Foundation, Jean-Jacques Lafond, myself and all the researchers here would be delighted that you came to see us in person, you said it, there is the hall of history to visit. We will also be able to bring you other flurons of the culture and art of Toulouse. And there is also the gastronomy of Toulouse that we would really like to be able to share with you. Well, it will be for another time, but we keep it in mind that we still owe you something. So we will make sure to be able to welcome you as soon as possible. Listen, I would like to thank you all again for your participation. Thank you very much, Mr. Boyer, for making the pleasure and honor of being among us. And once again, congratulations to Mathieu Jackson for the obtaining of this prize, Jean-Jacques Lafond 2020. I wish you a good evening to all and good continuation and take care of yourselves. See you soon. Thank you, Mathieu. Thank you, thank you very much. Goodbye. Goodbye. Thank you. Thank you, goodbye. So much, Matt. That was great. Yes, thank you so much. It's a great pleasure. I look forward to being there again sometime. Enjoy a nice testimony and talk about the future.