 Thank you so much. A couple of things that I want to say. I have been married for 22 years. I have three kids So in all those years of experience I have had many many dinners at home where people are eating and not paying attention to what I say So please go ahead. I feel no guilty whatsoever and also, you know, I'm You know after David and Eric's doing their speeches and they have such a great hair and I have not That you know, that's why I wear a cap most of the time What I want to do is is to talk about the billion prices project and and actually I want to show you our new baby We have a new new Project that we're doing and indeed a little bit what Eric said is my goal I Mean we we faculty are very arrogant. You know that, you know, we are genetically designed to be arrogant and obnoxious in general So but my personal objective is that I want to change the way statistical offices do their job completely And I think that the procedures how we collect the data The the methodologies that we use Have been designed, you know, 50 70 years ago in the best of the cases and And most of those were designed with a particular purpose was to avoid manipulation and the rules were very different from what the rules that we would like today and Also technology has changed tremendously and therefore I think we can collect data differently And so I'm trying to change the way we measure life from the very personal aspect from You know, how you measure your self-awareness a lot of you and I have seen you have these devices that we put on our Dress that measures the number of steps that you take and the number of hours that you sleep and the deep sleep which supposed to be very good and But these devices actually provide a lot of data. They're really lousy Because they're not telling me if I'm healthier They provide no information whatsoever But the challenge that we have the where developing these technologies are tremendously good at collecting data We're not very good at actually providing better information So that actually goes to my first slide and My hope is that we can marry this too. There are two types of data in general People call it big data. I really hate When you Look at the data by their size. It has nothing to do with big. I mean the data on the census is humongous But it's not about the size is not what matters here What matters is how the data was collected? The data that we collect in a statistical office what we call design data has a particular purpose They are being collected to answer a particular question when I collect the financial statements or your you know Taxes you feel a lot of forms if I were able to collect the data that would be humongous data Okay, he has a particular purpose answers a particular question Which is what's your tax liability and it lets the governments to steal your money? Sorry to collect the money from you We have these devices that are fantastic devices. I mean most of you call this a phone No, it's interesting because you know in Latin America. This will be called a cell phone in Europe We will call a mobile phone So iPhone to interestingly they always use the word phone you see I Think this is actually like a small computer with a lot of sensors that by accident makes phone calls I mean like how many times do you actually make a phone with this stuff? You know, they actually are tremendous sensors that look at you know, how we speak they have our voice Recorded there. So by the tone of our voice. I could tell if you are sad or happy or angry By the by your emails, I will know what your network is So so we have these devices that measure tremendous aspect of our life We don't even know how powerful this is and this data we are generating this data without a particular purpose We really don't know why we're generating the data We don't even know we're generating the data every time you go to a mall And you're in a store and you have your cell phone on they are tracking you and they know where you are And they know that you stand in front of H&M But then you didn't go to bed except so that tells you something about the age of the person and if you actually go to you know Primark tells me something about you know, you know what the income level is tough And if you actually are standing on the back of the store that means that you're looking at the cells So you're probably a loser so it's a kind of you know We actually infer a lot by tracking your phone and we are not even knowing that we are we're actually generating that data There's an advantage of the of the data generated by these devices. It's very cheap. It's very large The access is usually quite open, especially in comparison to your tax records or your medical records, which will be impossible to get But there is a problem with that is that it's not necessarily representative It's a big aspect of that is that we're generating the data voluntarily So if you want to stop to be tracked when you go to a store you just have to turn off your cell phone And there you would be no no will be tracking more. He said it's kind of a decision So it's not representative in that sense and therefore the conclusions that you get from the data have to take into account that the data by itself is not representative If you treat this data as if we're the financial statements from the stocks From the companies that that aren't stock market you will make humongous mistakes I know more companies that have failed and go bankrupt because of misuse of data Then companies that have been successful in the managing of data and the reason again is that Sometimes you're mesmerized by the data Technology allows us to collect Tremendous amounts of data and the questions how we're gonna use it Let me give you an example. I think that today we write more words Sentences and paragraphs than ever in the history of humankind. I mean I have a horrible accent So when I dictate to the computer it always tells me you have a horrible Latino Accent can you can you type? You know, that's usually what happens But for some that actually speaks English correctly They you can you know dictate the computer and the computer will just type whatever you're telling them So we write a lot of words more than ever before in the history a I Can tell you we don't produce better literature than 300 years ago The fact that the words the paragraphs and the papers are Cheaper to produce doesn't mean they are better They are not necessarily better and our challenge is to take these sources of data and make it information That big gap between data and information is becoming bigger and bigger And it's in the challenge of how we deal with the data and understand their limitations That will allow us to produce better information and I think that that's that's part of what this initiative is about I have to put a disclaimer before I start it's very important This disclaimer because when I say I I mean we okay That means Alberto Caballo and myself Alberto is a junior faculty at MIT. He was my student before that Also, very important when I say we I mean actually he okay So or or they or somebody else, but certainly not me Okay, so for example in a sentence like I am going to change the world means Alberto and I and when I said we Screw it up. It's Alberto. Okay so What do we do we I want to show you two things that we have done one is a billion prices project And the other one is I call it a thousand big max project the first one is we We started by thinking about how to collect data from to construct the inflation rate In a better way simpler way cheaper way And that's all some of the issues that have a statistical offices. It is very common for the statistical office that when they go to the store That the product is not found in the store that the day they have to go They go once a month for example in the United States They go the Wednesday close to the 15th of the month and they go to the stores physically and they collect the prices And if that is almost like a clandestine operation, so okay because stores don't like For people to take pictures of the items in fact you can do that go and you worry We're next to cause for street go down turn to the right after totem pole road You there's a primark on the right just enter primark and actually get your phone and start taking pictures of the tags You will see how fast you are gonna be kicked out of the store. Is that okay? And in fact a measurement of your cuteness is how long it takes for you to be kicked out It's okay if you are really really really really good looking well you can spend like you know 50 pictures Okay, if you are ugly looking like me after they see the phone out. They keep me out immediately. Okay, so But truly it's almost like a clandestine operation in which they have to collect prices They don't want the stores to know that the particular bottle of water is a very important item in the construction of the inflation rate of the UK And the reason is well imagine if you know that I collect only 10 bottles of water in the whole UK by the way They are 10 And imagine that you know that you are one of them and you know the brand that I'm collecting on the day I call you understand that you could affect the price of water now in the UK We collect 69,000 items per month, so it's a lot in Ghana. They collect 200 products a month 200 So if you want to trade on bonds at the same time You can you know you can make millions of dollars just by affecting the price of your item one day Not only that one afternoon when the guy shows up So truly they want to keep this information very confidential What now the problem is that when they show up to the store about 15 percent of the time 15 percent of the time the item doesn't exist and They have a horrific Procedures I mean they have a horrific problem because assuming that you don't drink water in London is a bad assumption No, the fact that I didn't find it doesn't mean that you don't drink it It's just that that day in that particular store that particular brand I couldn't find it So what we have decided was to actually go to online stores We started getting data from online stores first and food and and the first country We started was in Latin America the main reason is that Argentina at the moment when we started Argentina Intervened the statistical office and they started lying about the inflation rate. It's very simple the inflation rate used to be 12 The day after the intervention that the inflation rate dropped to eight a little bit less than eight and And it was very convenient for the government because the debt was indexed to the inflation rate So by lowering from 14 to 8 The interest payment just dropped by 6 I mean, they might be stupid, but they're not that stupid knows So they realize they can make a fortune and and lower the fiscal cost of the of the debt and So the first country actually we started was Argentina Brazil Uruguay Chile and Colombia those were the first five countries Where we started and a lot of these is because we knew people in this in these places When this project started I call it the billion prices project because I had the hope That we were going to be able at some point in time to collect 1 billion prices a quarter Today we have the ability to collect more than 1 billion prices a day. We track 1 billion prices a day We can track And again, this is the ability of what we do. We don't necessarily have to do that. We actually do way less than that But just these are orders of magnitude. We go to online stores first. We have to learn how to collect the price of a haircut I don't know how many of you have ever gotten a haircut online, but it's kind of very difficult Okay, I mean the only haircut you can get online is actually this one. Is that okay? This is the haircut you get online except for this one. Yeah, there's no haircut online But that doesn't mean that the prices are not online. You cannot buy gasoline online Well, you can buy gasoline online that will be mostly a terrorist webpage But but most of you cannot buy that doesn't mean the price cannot be found online So indeed we have the ability to collect a lot of prices for a lot of services online And when we learns to do that we decided to start constructing inflation rates for different countries These are all the places where we collect data today. We don't collect on all the sectors There's some countries where we collect like the United States the UK for example is one of the most fantastic ones in our data in terms of the The coverage and Venezuela we used to have but now there are a lot of promises I am from Venezuela. There's a lot of problems right now with price controls So we a lot of the stores have disappeared But give or take we have from everybody in India, for example, we will have food and electronics But that's it. We don't have we don't have all the sectors everywhere But for the ones that we have we construct inflation rates for those countries and we can construct an inflation rate every day Instead of going once a month we go every single day We have a basket that is incredibly big the day the iPhone is released I have an inflation rate of the iPhone the following day because I have two observations. That's it When the I've the the Apple watch was released actually I am already collecting prices from Apple watch from every web page on the world right now Not a single one has been sold and I already know for about a month What is the relative price of all the watches that exist? So This gives an advantage of what we compute let me just show you some of the indexes I'm going to talk about three things. One is the inflation rate. The other one is a true core And and the other one is a big Mac. The second one is a big conversation for central bankers I just want to show you how different it is. I'm not going to turn too much But these are the inflation rates of some of the countries the US the euro Australia and the UK We started collecting these at different times. So not all the countries start on the same day But but you can actually see the orange line is our daily price index the blue line is the official data So in some countries, this is remarkably close To the other one interesting thing that happens is that online prices always move earlier than offline prices always and the reason is that Try to remember the last time you purchased something online. Do you remember the price of what you purchase? In fact, let me do a race fan. How many of you actually purchased something online the last month raise your hand Let's see like everybody. Okay. Do you remember what you purchase if it's not an illegal item? Can you share with us? or not embarrassing 299 you know remember what was the price of that item a month before you purchase it? Isn't that interesting? You see we have here a gentleman that I mean look at him You know, he's kind of good-looking looks smart person. Okay, like you went to it. Yeah, yes And so yeah, he purchased something and he has no idea if he pays too much or too little is memory less And by the way every single online consumer in general is Memoryless you have no clue what the price of the item was a month before and Don't worry. There's only six and a half billion people like you. Okay But you know who knows that the stores knows that The store knows that you have no memory So if I have to increase the price of bread in France To whom will you increase it first to the guy that goes every day to the store or to the guy that goes online and has no clue So you increase the price on the clueless by the way in France, they don't increase the price of bread That happens like once every 200 years that okay What what they do what they do in France is that they take the bread when there's inflation They just make the bread smaller smaller smaller smaller smaller and that's how the croissant was invented. Is that okay? That's that's the procedure and then when they do have no other option then man I have to increase the price you increase the price and they burn like 24 stores Okay, the same happens in Italy with pasta by the way So so because of this feature online prices capture inflation way earlier So we turn in the United States always moves about two months before the official data in England always moves about three months in Australia is kind of I mean they produce one quarterly data is kind of irrelevant But we move that's five months before in Australia because of the frequency and in the euro is about a month and a half We have other countries where they're not that identical. I mean I want to show you some emerging markets Brazil, Colombia Chile and China I'm always asked about China if China is manipulating the data in China I can only produce we sorry Alberto and I we can only produce an inflation rate for food But we don't have that many sectors, but at least some food This is from 2009 and there's not a big difference between what we compute and what the statistical office computes I mean they're roughly the same the trends are very very similar and here you can see the recession in Brazil You see this decline tremendous decline of prices. Well, that's going to be announced in about a month and a half Okay, yes, you saw it here first anyway So now but this is a country that we always have in our data It's a country that now is used in the in the iLab and in my team because if you don't see the difference between the orange And the blue line you are legally blind This is this is Argentina always when I'm asked about a reporter they ask me Well, this is how many years you have I said it's about seven years of data Okay, so in seven years of data the official inflation rate is 220% Our inflation rate is just slightly different just one different digit is 520 just we just move the wrong digit anyway So some countries look like this Okay, some countries the inflation rate that we compute is just dramatically different from the official one and Russia Venezuela right now our inflation rate is twice what is reported in the in the official statistics There are these two properties that I said the first properties congruence that means that for some countries The inflation rate online is very similar to the inflation rate of line and the good thing is that is anticipation This has actually changed a little bit the way central bankers looks at our data Some central bankers use our data to try to improve the short run prediction of the inflation Policies and that's kind of the objective the true benefit here is between the first three six months after that There should be no no no any benefit from our data. Let me just talk a little bit about the core effect Almost every central anchor in the world talks about the core inflation rate I don't know if you have heard but the core inflation rate is that they want to take out of the inflation rate The things that they are out of their control Okay, so for example in the United States and by the way, this will be true in almost every country They said well, we don't control the price of oil because I have no control in the price of oil I have no control on the price of gasoline there will be inflation rate up and down on gasoline So what I want to do is to exclude gasoline from that index and I always Hated that procedure because they like okay, so if you don't control the price of gasoline Then what about transportation because transportation is heavy on gasoline and usually said well, you know what? Yeah, maybe I should exclude transportation to okay But if you exclude transportation, what about imported items that they all are transported by the transportation that you would have control Well, yeah, maybe we should actually take out also, you know imported items What about what about services because the price of energy will be affected by gasoline and that will affect every service of the country So if you take these two infinity, you are left with two items To the solution that's it truly. These are the only two that consumes a different form of energy Okay, except for that. I mean so that should be a very easy core And by the way in Italy and in Netherlands they have been collecting both for a long time So they actually they know It's a much better procedure to eliminate the indirect effect How much of the inflation rate of tomato depends on gasoline? How much of the inflation rate of transportation depends on gasoline if you want to take out out the Effect of gasoline you don't want to take the whole sector You want to take only the part that is affected by the price of gasoline and you know This this this is a statistical procedure is very difficult to run When you have very infrequent data You need to have very frequent data So the daily data made a big difference and this is for example the core of the United States I have been able to do this only for the United States The blue line is is the official core the orange line is the core that we have produced So I just want to show you how remarkably close they are in normal times But what happens is for example this year When the United States at the beginning of this year was detecting very negative inflation rate very negative outlook in the United States At the beginning of this year all of that was the indirect effect of gasoline affecting the sectors that they were not excluded and it's a and in some sense Then by the way the New York Fed knows this perfectly They know this absolutely perfectly some of the best researchers in the world are computing pass-through are in the New York Fed So they know this they they know the effect is there just that they just cannot Clean it out with the data they have It's a difference so do you know the effect is there everybody can do the exercise that I just perform They just cannot actually separate and it's and they knew that so much that that they were not Mentioning to the world that the inflation rate was negative in the United States And therefore we were in a massive deflation and we should actually be printing like crazy according to all the Keynesians Is that okay? Well, they they failed to mention that because they knew that the data that they were obtaining the signal That was coming from the statistical office was not the correct one And and by the way, I expect these two lines to catch up with each other by June of this year So by June they will should be at roughly the same So let me tell you about the thousand Big Macs project. These are new baby. This is the first place. I'm presenting this from MIT so What we did is because we have the same retailer around the world and we have the same products around the world We decided to construct new measures of macroeconomic Instability let me explain what it is For many many years the World Bank with a very big consortium with the United Nations the IMF The OECD and the statistical zero stat They have a program that is trying to compute how much ten dollars buys you in the US versus London versus Brussels versus, you know Beijing and versus a rural area in China So the idea is this is known as the purchasing power parity Whenever you see the statistics and you look at something that says GDP and then in parentheses says PPP adjusted what they're saying is well the ten dollars in London buys you Like, you know one cup of coffee. Okay, but ten dollars in India buys you a whole meal So the idea is that the ten dollars do not have the same purchasing power and therefore what we want to do is to compare Apples and apples know that's how much the ten dollars in India can buy and this program has been running for many many years the data Has it weaknesses because when you go to Germany? They collect a beer that is different from the beer that they collect in Belgium Which is different from the beer they collect in Japan, which is different from the beer they collect in United States So even though it's beer It's different brand and there's no reason why one should be the same as the other in fact in general. They are not So for many years they have been trying to to get prices the same way There was a brilliant idea by the Economist the magazine where they decided instead of Collecting all the prices to collect one item correctly and they went to the Big Mac and now they compare internationally the price of the big Mac Which I think is a brilliant idea. There's no we know that this item is Bread lettuce and something that smells like meat. Is that okay? And he's the same recipe everywhere is hated by all the nations equally We all consider that very unhealthy equally. So the price comparison by the way they are constructing two new indexes one is about the Latte from From a Starbucks a tall latte from a Starbucks worldwide and and the other one is an iPad What is the cost of an iPad worldwide? So so the idea is that that allows you for a price comparison Well in our data, we realize that I have Sarah and I have sat in 82 countries That means that I have the same t-shirt exactly in every place and Sarah sells about 4,000 products Then I go to H&M and H&M has the same t-shirt everywhere. I go to Nike same I go to Apple same products. I go to Dell same products. I go to actually all the tradeable sectors food clothing a personal care Gasoline and things that close to energy related And and we have all of them We actually collected the data from all of them So we were able to actually control these Matches from the world and instead of having just one big Mac. We have at several so this is the UK versus the United States These actually are the pictures from my web page. This is longer data I'm just showing you starting in January of 2014 just to show you this is one item And when you look at one item, this is this tells me that the item in England was twice the price of the of the American one Is that okay? So this is the pricing pounds divided by the exchange rate Divided by the price in the US. Okay, so I convert the pricing pounds to dollars And it's very interesting because this is how a big Mac will look like well sometimes is very cheap Sometimes is very expensive now. It's very interesting that when you actually start putting more items in this list Things change dramatically. This is one item. This is a different item. So this actually is cheaper in England By the way, this is food That is fuel. Okay, just to let you know and this is supposed to be confidential It's not like this is being broadcast on the web or something like that anyway, so So this is actually a food purchase on the same type of supermarket So this is comparing test code with an equivalent supermarket in your message will be for example Safeway, okay, same quality same type of customers So just happens to be that this this item of food is cheaper But then when you put more and more items, you know things start to be more convoluted It's very difficult to understand what it is and when you put all the thousand items looks like a very nice picture and You can see a trend here If you don't see a trend here, then you have you are prone in visualization What we do is we take all these massive set of items these actually are all the items that so I have no idea How many they are I guess I'm always glad that Alberto has a little tag at the top of that list Which is match IDs that says all and I just press all and whatever it comes. It is is a number and What we do is we construct this indexes and I just want to show you the difference. This is Brazil And in fact, this has made a big difference for emerging markets Not such a big difference for the UK yet in the sense that I am not reading something very different from the statistical offices for The moment, but in a country like Brazil and Australia, this has been very dramatic The blue line is the data constructed from the official From the official indexes you take the official indexes. This is what the World Bank or the IMF will construct according to that Brazil is getting cheaper and cheaper and cheaper and cheaper to time Okay, and this is actually not a small period is from 2012 to today And you can see that I normalize that are hundred at the beginning and Brazil is getting cheaper and cheaper and cheaper If you go to Brazil, you will realize that actually that's not the case It has no relationship with this and the reason is that these are some very particular basket This has a lot of domestic food and therefore you are comparing domestic food in Brazil with domestic food in the United States Well, you are not even comparing the same items When we actually do that tracking today Brazil is about the same place where it was three years ago In fact, I we construct an overvaluation you can look at the orange line if you want there's two ways of computing overvaluation But both of our measures will tell you that Brazil today is roughly in equilibrium by the way the recent devaluation. This is actually printed I think I sent this yesterday. So this is the data from yesterday morning. Okay, so which includes actually the previous day So I'm so in some sense is right about where it should be And and in fact when you look at these periods on the orange line those were the previous just preceded movements of exchange rate dramatically This is UK. I just wanted to show you the UK So if you look at the orange line, which is kind of my preferred one by the way The blue line is the preferred one by for for Alberto, but Alberto is from Argentina So he's always very negative. You see he's always very far from equilibrium as an Argentine And he has never known what equilibrium means is that okay? So to find a variable that is close to zero is impossible to understand so but but the top one You know, I actually pay attention to the to the bottom one. Okay, so actually the UK is roughly what it is And it's an interesting statement about About the UK is that we are gonna have a lot of volatility on the sterling dollar and sterling euro rate Due to the elections and these and these uncertainly that we're gonna see is gonna reflect it on a very volatile Sterling rate which already is happening now interestingly according to our data. This is coming from the retailers. You understand I'm not doing anything sophisticated here. I just take the tomato in Tesco, and I divide the tomato by the tomato in In safe way or Giants, that's it Like it's wow super sophisticated. This cannot be less professional. Is that okay? Okay, so so so our data is telling us that the retailers that are trading all the items In in in this between these two countries Telling us that the rate is about to write one. So maybe there's a lot of volatility, but there doesn't seem to be a very big trend here What is my goal? And let me just finish with this. I am I am working a lot on retail sales and consumer demand I am we have done a big strides, especially Alberto on supply disruptions So after an earthquake or a or a flooding to understand and measure better the economic impact of those natural disasters We're working on employment and labor markets to try to understand demand It's a little bit a very important theme of the of the digital economy Is that this is going to change dramatically the demands for a skill versus own skilled workers? And therefore we need to understand where the holes will be and where the high demands What are the professions that are in high demand and and finally I'm working really hard on real estate and trying to get a Daily measure of GDP Some people say like what do you need to do this? I always said well, you know, I usually need like one billion prices, but but truly I think I need more than that I need like one trillion prices. Okay, and I don't know but you know, you can see the similarities Thank you so much. Let me pick out these questions The question is how you develop the weights how you construct the prices, so We cannot weight the price the items by just looking at the web pages It cannot be done that way and in fact I have spent about 15 years going to the stores I arrived today in the morning to London at 6 a.m I took a shower and the first thing I did was then take my train Paddington then Paddington to actually Oxford Circus and I walk through Oxford Street And and what I do is I go to the stores to try to understand what they are selling and how and what They are selling is shown on the web Because it's the how they present the products on the web tells me a lot what they're selling because when you think about it Do if you go and if I sometimes have to buy things for them to tell me, okay, which I do So my backpack was empty and now I have like there are two shirts for my daughter and candy for my wife I mean makes it so they love these research that I do because you know I take all the money that I spend on all these materials and I give it to David because it's research You know so so he pays for the research and so we get a free clothing and free equipment all the time So but but I have to understand that and there's a big mapping there There's some stores that are very badly behaved that what they show you online is not what they are gonna sell And that's the part the very important part of transforming this data into information if you don't do that homework You will make a mistake now. There are some honest very honest web pages So let me tell you some of them IKEA H&M Sarah Apple these are web pages that are exceedingly honest in terms of showing you What they are gonna sell and roughly how they show you tells you a lot what matters and that's that's kind of the way We do it so it's a story specific in that sense And and we have so when I construct the price of tomato in in the UK It's not that I constructed one price of tomato actually is one per store So science very will have a different index and test cons so on and then I aggregate into tomato and then tomato to let us And then I put some bread and some bacon BLT whoo-hoo the next item. That's how more or less how we proceed No, the question. Yes, I don't think a great Equally well in a deflationary scenario versus inflation because you've only talked about inflation so far so indeed indeed so What changes between deflations or recessions and booms what changes is how? Anticipatory is the online. Okay, so for example when you have a massive recession Usually what occurs is that prices tend to go down and therefore the online prices go down a little bit faster than the other ones But it will be a month but but no more than a month and the reason is Now there's no reason not to lower the prices You see if the price of bread is gonna be cheaper consumers if anything they're gonna be happier So there's an issue with the timing so on the on the downturns The degree of anticipation is a little bit smaller than the upturns on the upturns the anticipation can be three months But yeah, that's the only difference now. I show you some data on the United States and you can see the recession We kind of pick the same deflation Both of us so and then the recent decline on prices due to oil also for example in all the countries Look very similar to the statistical office. So in that sense It's the same now not all the price declines are the same for example oil prices when they decline they decline very fast on both Both data, okay But recessions that come from lower demand will be a little bit more anticipated on the online But that's the difference good question Yes, hi. Yeah, have you looked at this in relation to more complex bundled product like the way telecoms bundles your TV and your voice and your broadband and so on it's very hard to separate out the individual elements. Oh The very good question. No, no, I have not so so when I look at for example, you know TV contracts or You know internet and contracts that you will have for your home electronics and Actually, I just look at at what is the bundle that they are selling so I don't disentangle It's a very good question But no, I don't I have not paid attention to that Yes Last question actually it was more of a request. Could you please remove that those pictures are driving people crazy on the web? Thanks Roberto very much good. Thank you so much for everything