Building the Foundations of a Decentralized CPI: Infrastructure Requirements & Design Challenges

Transcript

[Music] so my name is Brenda I'm one of the co-founders of teller and but before that I was actually an economist for the US Bureau of Labor Statistics and I used to focus mainly on some of the pfis the uh some principal Federal economic indicators such as the employment numbers that you see come out every month the um producer price index and the unemployment insurance claims that you see come out every every week and although I believe that the people at the bureau are doing the best that they

can with the data that they can collect and the methodology that they're allowed to use I believe that we can make all of those statistics a lot better um specifically the one that I would like to focus first on improving upon is the CPI and that's because the CPI is used so broadly that it affects everybody's life so directly but rarely do people make the connection between these random number that the government puts out every month and how it directly affects them some of the things that are it is used for is just wage escalations if you might you guys might have some in your contracts or you may not um cost of living

adjustments adjustments to social programs and changes in monetary policy mon monary policy whether or not you understand the CPI it directly affects you and it doesn't matter if you're are rich or poor um it affects you Ben Franklin one of the um founding fathers of the US uh once said but in this world nothing can be said to be certain except death and taxes and um I went ahead and corrected his uh that's the wrong correction area but like I corrected his sentence to be actually inflation because as we all know if you have enough money in the US you can very likely be able to find enough enough loopholes that you don't have to pay taxes but no matter how

richual poor you are you're going to face inflation it's just it doesn't discriminate basically and the thing is that people rarely make that connection they also rarely make the connection between what the CPI is this measurement and how politicized it has always been from its Inception the very first time it was commissioned it was in 19 1890 and it was 19 1891 it was commissioned because of the Tariff Act of 1890 and it was commissioned because Republicans and Democrats wanted to use these numbers on the next election cycle to talk bad about the other one right like whoever made the big mistake of imp implementing

these tariffs and obviously because they were focused on their voter base the only people that the survey included was white guys that lived in the city that's it if you were a woman if you were of any other race you were not counted you did not matter you did not have a voting right anyway right so it was just not necessary in their minds then uh it took about 50 years for the government to finally say hey you know what this doesn't actually represent the population of the US so maybe we should include black and Latino households in this survey and they purposefully included them unre ented or not fully represented so each one of those

households represented a lot less than what a a white household represented and then it took another almost 30 years for the CPI to go through another revamp of instead of them collecting a basket of goods that might be only Goods that one specific group of uh people uh consumed to a basket of goods so now instead of just collecting bread and eggs they're collecting everything that falls into the food category and that helped alleviate this bias that was in 1978 that was the last time the us updated their CPI methodology their official CPI methodology and the way that this has been created for you know

the longest time is basically the government has this administrative data that they use as their universe and from there they take a representative sample of it so that they can uh estimate what the population value is and then they take the sample they also have uh an expenditure survey that they do every 3 to four three to four years and they use that expenditure survey to um to weigh each one of those categories so that they can create estimates out of it and then they try to keep you know the phrase of set set parabus or all else equal so they try to control for almost everything and to able to extract just the price change so every month they

collect information on the week of the 13th of the month and they collect information from businesses and households and they they take it into the B Labor Statistics or um uh yeah the B Labor Statistics they give it to the analyst and then the analysts do the best they can with this data the way that it is collected and with the methodology they're allowed to use they're not allowed to change the methodology by themselves literally and act up Congress to change it so it's not going to really change for a very long time currently they publish three main CPI um values the CPU which is what it's the official value and then the CPI W which only includes clerical workers and

hourly uh wage employees it doesn't include retirees or people that that are um unemployed and then the ccpi which is the the chain CPI this methodology was introduced in in 2002 um to sort of address some of the criticisms for the CPI but it is so politicized that they could not replace even though it's better or it's supposed to address the criticisms that everybody complains about it could not replace the main CPI value then it's actually it's actually mostly or more probably used for the majority of the decisionmaking

um but they at least added it right um and I want to I want to point out that the government doesn't actually pay to collect the data where we spend or where they spend the money is basically when it comes in how do you clean it up how do you classify it how do you aggregate it which aggregation is pretty simple because they don't really change it but then they do spend a bunch of money on methodology research that never sees a light of day because you know it has to go through all of these politicians to be able to approve anything now we have antitrust laws in the US that protect consumers against monopolies so it only makes sense to me that we would create this separate

verifiable economic measure to actually act as an oversight for governments and protect people from corrupt governments I'm not saying all governments are corrupt but there is a value of people having this alternative some examples of where this could have been useful was Argentina from 2007 to 2015 they openly admitted to manipulating their CPI values instead of reporting the whole value they were reporting only a third of it and one of the reasons that countries may want to do this is because when they go to the IMF or the World Bank to ask for loans there if um if they are seen as high risk they have low

purchasing power then their the rate their lending rate is going to be a lot higher than a country that has a lower CPI so there is this conflict of interest because they're producing the number but they're also the ones mostly affected by any changes on it um Venezuela um most people would say it's a very corrupt government they don't really trust any of their statistics um Zimbabwe in 2023 the new elected uh president or they decided to you to not use the new CPI method Theology and stick to the old one for their decision making I'd say that seems to be pretty uh politically driven but these are just some examples of how how this is much

needed uh the only things I can say that Argentina and Venezuela the really good things that came out of it was really good research by Alberto cavalo and Manuel bertolotto where they actually recreated that period of time for them and that's who the because of their research they were able to find out that basically they were not manipulating the values at the micro dat level but rather like they would produce it all and then they be like yeah we're just reporting a third of it and also they found out this was prior to the to uh Co they found out that there's very high correlation between online prices and local prices apparently we're really good at uh or companies are

really good at making sure they uh do price discrimination on their markets so that was great right because now we can actually use those findings to use that same methodology by scraping a bunch of data and creating a similar measure but even though we have all of these examples most of the criticisms that you're going to see on online and on the web it's always going to be substitution bias lack the inclusion of items and po quality adjustments and that's be and and I just want to talk about a little bit substitution bias it basically means that um you so think about how it's used and um the if I'm an employer and I'm told that or we have a wage Clause escalation Clause based on inflation and

I'm supposed to increase your pay at the end of the year based on this well you know what if you're eating 20 grams of protein I could argue that you're just as well off whether you're eating burgers or you're eating steak and there was an actually really uh funny tweet uh last year specifically talking about that somebody was complaining was like I'm not you know it's not my standard of living is different a steak and a burger are not the same but this is what uh substitution bias argues it argues that because people are tend to shift items in their basket of good to to whatever is cheaper within it then you're just as better off um and that's one of the criticisms but all of these criticisms

basically stem from the fact that were using um a static basket of goods that was selected 3 to four years years ago to keep track or to attempt to um to create an inflation value the one thing that nobody really talks about it's the conflict of interest of the government I really have not seen any research on it I rarely see uh there are some paid um paid services that create their own CPI values but then again you end up with a very similar product to what the government provides the government is very open about their method te ology even when they've been super biased about how they've implemented it um but they don't

they don't and cannot disclose the underlying data because they actually provide anybody that gives them data a guarantee of confidentiality and that their data will only be used for statistical purposes so what we need to do and what I think decentralized systems can bring to the table is actually try uh making it possible for us to create a transparent verifiable that um and also obviously we want to make sure that we have some sound methodology but we have plenty of research that we can actually incorporate that the government is going to take a long long time before they can approve any of it so the main goals obviously is going to be uh bringing in transparency and for

this um the actual technical um the actual smart contracts are very simple they they basically keep track of the decentralized store addresses where the description of the of the CPI is the the methodology for outlier detection for aggregation and for Price quality or for for data quality are kept and then um all of these and and then obviously you need a very decentralized Oracle because the moment that these numbers start diverging and people start asking questions that's when you run the risk of censorship so obviously I'm using for that but like but the main

purpose of this entire system is to is to encourage open peer review right now we all basically depend on the government um wanting to update their methodology which obviously they haven't been want to since 1978 um and there's a lot of research that goes in in inhouse I was I mean I worked there for 7even years I did a lot of research in it that never Sid the day like three two three three years of research and it was good job and nothing else nothing came out of it um so now with this open review system we have the data we have the methodology and if

anybody disagrees or think they can do it better they can actually iterate on it and improve upon it without the political agendas having you know say a say on what goes or what doesn't and that allows us to basically as a Community improve improve upon it the main things that I think that we can that that we benefit from uh opening this up and that we can actually bring in efficiencies is the data collection because we don't have to depend on the government approving our data scrapers like just get them done get as much dat as we can also product classification because the government spends so much money on analysts making sure that dat is clean and the DAT is classifi correctly we can actually

experiment with AI and make those processes a lot more efficient and less uh human interaction will also lead to hopefully less cost in actually in recreating this and out detection aggregation methodology that will eventually even if you think about a lot of the times that people do research it's not even I mean especially like academic research you don't really get paid lot right you you probably get a grant and you're just happy you keep getting those grants uh it's a lot of ego but honestly we can actually harness that into making our methodology better over time and making it open makes us more agile than the government uh we obviously have to keep some um or try to keep all the

equal so we have to make sure that we collect similar to the government we have a schedule similar to the government so we can uh so we can actually be comparable but we can use other better methods uh we still have to face some collection challenges just like any other survey um we also have some extra challenges because we don't have a universe of administrative data that we can use or uh for for randomly sampling or uh creating weights so we can we can initially send this off by sort of using part of the government information we also cannot collect taxes uh I wish it could and I guess uh what was it um Guild protocol was talking sort of like something like this for

common goods uh or for ethereum uh devs at least I know there's some other things out there but ultimately incentives are something that is be difficult to do just because of the nature of of of what we're doing um but common goods can actually create the foundation for very or for much more of profitable Goods like a CPI could create the foundation for a a sorry uh for a stable coin or even even some uh prediction markets everybody loves to bet on those um we don't have Universe data but we have the ability to collect more and more data the more data that

you collect the closer you are going to be to the population and that can help a lot reduce any any of that bias that you might be having incentives are a little bit more difficult to to implement at the individual basis so the way that they would be implemented it would be at the um whoever's creating those and maintaining those data scrapers and that's because you don't want to introduce self selection bias where like if you say if you collect this information will give you a dollar per price or whatever then you're going to get the part of the population and that's probably they probably needs the money and you're only going to get sample from the places where they actually go and and transact um the way that I'm separating this is into four

different categories of of people that would be participating in the network and the way that I'm structuring it is that they are not exclusive but they can be and that's just because I want to minimize the amount of specialization that is needed to be able to increase the amount of um of people that that participate in the network so for the data reporter you need somebody to collect the data uh to and those collectors are basically the ones uh could be the ones creating the actual data scrapers uh the data aggregators which should be sort of the easiest job because you already have the methodology you just have to apply it data quality research that is a lot more

specialized and that requires somebody that know with a statistical knowledge and economic um so they can do you know this open peer review and iterate on the methodology and then the data monitors which actually those are a little bit more difficult than just arguing that the data is wrong because any data that collects uh millions of rows of data is going to have outliers so there's got to be a threshold for those outliers an allowable Thresh threshold and then from that you have to have a a measure of quality so that you can then reward these um data collectors after the fact so it works a little bit different than how you would do uh a lot of how crypto

works that they want an immediate reward because you have to wait for the data quality to be assessed um this goes into each one of the of the roles yeah I don't want to go through that because I don't have a lot of time but mainly right now what I'm F focusing on is basically creating the proof of concept through the Big Mac index the Big Mac index um basically tracks prices of the Big Mac all around the world and that's because the Big Mac is so standardized it doesn't matter where you go a Big Mac's going to taste like a Big Mac and so uh McDonald's did a really good job at standardizing that and they've been publishing this uh information at the on The Economist for a very long time they're very open about

their methodology they don't have all of their micro dat available but this is the first of the proof of Concepts that I'm doing right now we're already working we already did uh data scrapers but we're also uh starting to play with AI uh to see how we can actually harness that and instead of maintaining all of these um data scrapers because they can't change country to Country um then obviously the uspi which is a bigger much bigger lift because it's a lot more data and then hopefully eventually turning it into a global index and uh applying better statistical methods to reference prices so we can benef the government can benefit from crypto but crypto can actually benefit from the statistical

methods that exist out there to improve data since the majority of the hacks that happen out there especially for oracles are because of thata of the manipulation of the price of the asset not the actual Oracle and with that I ran out of time so uh thank you guys for Te fantastic Brenda thank you so much um we have time for one question maybe quick question from the audience is anyone like one there yeah can you hold it okay uh I have actually two points uh what about the challenges with the take put it closer to the yeah okay uh when you mention about the challenge with the

government with all the politicians it's more about the sovereignity issue relatively the datas are now which are we which you are mentioning are more more sensitive datas of the each and every country so that would be an issue of sovereignity because nowadays we are seeing a lot of countries fighting over this kind of issues and also the tech companies which have a huge marketplaces which have uh data uh which have a big data uh on their data centers which are which they are in it's more a profitable uh area nowadays uh I didn't see that you mentioned about the take the the

take organizations because they are now the one of the main uh role players in this industry yeah so there there's price stats there's true flation there's a lot of companies that try that to provide some CPI data but they but they recreate basically what the government does which is uh they have some portion of it behind a pay wall and so you cannot absolutely verify what they're doing um that I don't see it as a challenge because I don't see this as a competitor to them I see this as um I I I see this as a common good and for in terms of like sovereignty and like dealing with with governments that's one of the reason it has to be decentralized

everything has to be kept on decentralized storage is because there is going to be that conflict there is going to be this uh these politicians that are going to say say no this this methodology does not make sense this is not real basically basically guys light you because they want to keep their own methodology and be able to control what people see so I don't know if that answers your question but I'm happy to talk about it later because I know they're trying to get somebody else up here thank you so much Brenda for your time and insights big round of applause [Music]