Bringing Macro Data Onchain

Transcript

Well, hello everybody. Uh, thank you for staying behind. It's always a tough tough time following Vitalic, but I do have a couple of words for him in a second. I'm Brenda. I'm the CEO of Teller. And Teller is the only oracle that was designed to be decentralized from the beginning. We launched in 2019. We were a fair launch. From the beginning, reporters were required to stake. They could be disputed on chain and anybody could request any type of information as long as they tipped on chain. So yes, on his last talk, if you guys didn't catch it, he said that one of the most important things that we could be doing in the space is actually improving oracles. I couldn't agree more

and I would love for you Vitalik if you ever see this talk um to actually come out and talk to me about it and actually call out why and why and what um what oracles have been have been doing wrong or centralized oracles make a distinction between centralized oracles decentralized oracles and help us grow. So with that a ve very direct message um I'm going to move on to what I'm I'm actually here to talk about. So before crypto I was an economist with the Bureau of Labor Statistics and the Bure of Labor Statistics puts out a lot of the major statistics for the US government. Today I'm going to talk about bridging macroeconomic data on

chain and why that has been so difficult. The Oracle problem for market prices and macroeconomic statistics is very different. So market prices in general, you can observe the majority of the um the the markets, you can observe them, right? In general, you can find where the majority of the volume is happening and you can come up with some price discovery that is observable. um when you're talking about macroeconomic statistics, it's a lot harder because you're talking about all different a lot of different aspects of the economy and it's very very hard to observe. It's very hard to observe for anybody. Um also they have two different manipulation surfaces for market prices.

you you have the markets. So that is observable for um macroeconomic statistics. is a lot more subtle and it can be within their methodology and it can also be just political capture or coercion because politicians fund um the very u statistical agencies that produce them and people don't always make the connection between the CPI or inflation and how it directly affects them. The CPI is used for so many things by the government from adjusting um pensions and payments that go to the public like u um yeah like different different things like in the

US like the uh for safety net networks. Um they'll they'll be changed based on inflation. So that could mean that the government has to give out more money uh in assistance payments. Also wage escalation. So most government employees will have um their their pay increase based on inflation at the end of each year. Um also they use it to inform their monetary policy and the IMF actually uses inflation or the inflation in in any country as part of their risk assessment. The higher risk you have the more conditions the IMF might put in your country when they lend you money or

when they lend them money. So with that um with that said, people don't always understand how politicized this number is. From the inception of the CPI in the US in the 1890s, it was it was commissioned because Republicans and Dem Democrats were arguing about who was right about the McKinley tariffs. And then it took 90 years for them to actually create or for them to actually include blacks and Hispanic families into the survey. So they didn't have the entire population. Um and then uh in 1999, the Baskin Commissioner report had all of these changes to the CPI

methodology that was generally geared towards decreasing it. And once implemented, they actually saved the government over a trillion dollars of um money that would have otherwise gone out to the public. And Argentina, it's a very special case that we will talk a little bit more in in the later slides, but they basically openly admitted to modifying distorting the CPI for uh for that period of time. And this is a period of time where inflation was so high that people clearly knew that the government was lying. Eventually they stopped producing it in 20 in 2015 and out of that though um some economists

from MIT did a lot of research to recreate that period of time for them and we're gonna so we're trying to I'm getting a lot of the of of the methodology that we would like to implement to create an independent CPI from what the research that they did and just to put it into perspective how politicized this is Turkey in 2022 that's not very long ago they actually filed criminal charges against an independent group that was trying to recreate an independent CPI number. I want to repeat that. Criminal charges, not civil, criminal. So, this is how scared governments are of oversight. But before I can talk about the challenges

that we run into on recreating an independent CPI, um I need to talk to you a little bit more about how it's actually created and be and go a little bit deep in it because that's where where the challenges lie. So I hope you had had coffee because it's going to be deep. Uh so first of all, the government gathers Oh no, this is the wrong the wrong presentation. Anyway, so uh these are the wrong slides by the way. Anyway, so the government gathers um so gathers information from administrative data and two surveys, household surveys and business surveys and from that they create a universe

that represents the population of items that we all uh consume. Right? So this is a lot of information and this universe does don't only contain prices um they contain quantity so that way they can figure out the market share of all the categories that are imple that of of things that you spend money on and then once they have this universe that hopefully has everybody represented in it they can get a random sample from it. And this is so this universe is so important because that's how you come out with weights. how you're going to weight each category. So, in general, people will spend about uh almost 45% of their money on housing um and then 15%

on food and food and beverages. So, so how you spend your money matters a lot, but them knowing how you spend it allows them to get a proper weight so that when they calculate the CPI, it actually represents how you spend your money. and um is representative of of inflation. So after that then um they take this sample and then they start collecting some information and you know they they'll take into consideration um you know non-response either because of no contact or maybe some people just refuse to provide data and then um you know eventually they they they uh they also

have to they have a privacy agreement with anybody that provides them data. So they'll suppress some sales if necessary if they if if for some reason they people can reverse engineer pricing of you know uh items within them within the sales. And then after they do that they'll go back rewe and then they'll start collecting information. And then while they collect information, they'll go into they'll be a be able to observe item non-response as well as um attrition rate. And so it item non-response is like people might stop reporting what they spend on something.

Um and attrition rate is basically people just get tired of responding to the survey. And so over time they just start uh providing information. All of these they take into consideration the next time they get a sample of goods and they just over sample because they know that they're going to eventually have so much attrition that the sales will become uh impossible to to publish. And then finally, you know, um once they're ready to aggregate this information, they collect it. They run it through outlier detection. They have analysts go through it just in case any any strugglers were left in there. Then um and then they that's it and they do some imputation for some things for things that are not reported and then finally

they do the aggregation which is the easiest part of it all. Um also all of this methodology is very very transparent. what's not transparent it's you know uh the administrative data and the the the universe that they use and then the micro data that is used for um creating the actual the the actual number there are three different CPIs that are published every every month the main one is the CPIU um and and and even though it has had tweaks in methodology it is very difficult to do tweaks in methodology on it because it's so politicized. No, generally administrations don't want to change

anything. So, they just don't um and that's the reason the CPI the C the change CPIU exists is because they couldn't change that because politicians didn't allow it. So, they just created a different measure that had improvements in methodology. But in general, um everybody pays attention to the CPIU. All of all of this data is actually collected for free. Nobody gets paid. It's all voluntary data except for the administrative data that the government requires businesses to provide. Um, and so that's how that works.

And guys, I really hate that I don't have my slides. Um but so I talked about all the challenges within the within like the statistical challenges of getting a proper sample base and making sure that you have those weights properly allocated because that is what drives whether or not the CPI is going to be correct. the the next the big criticisms of the CPI however are generally substitution bias lacked inclusion of items uh poor quality adjustments uh so basically for substitution bias is basically the government assumes that you are just as well off if you're eating a steak or if you're eating McDonald's if you get 20

grams of protein a day they're like well you know you're you're just as well off uh they also uh another criticism is that and and and and that's what they want. So they're always trying to make sure that uh that substitution bias that doesn't creep in and it's generally uh an an adjustment in methodology that leads the CPI to usually be lower. Rarely actually it has never I've never seen any adjustments done for it to actually be higher. I like the inclusion of items. Uh if you choose a basket of goods at the beginning of the period, over time people change preferences or new items come into play. Um people, you know, yeah, just preferences or new

things and so that that becomes um you you because you start at a period of time, they don't continuously update that. So until the the period ends and then they re uh they get another sample then you know the old stuff is going to be there for a while. Port quality adjustments is basically um you have you they're collecting data for an iPhone 12 and then next year is the 13 and they have to account for the change in the technology because otherwise it's not just an inflation change but actually a technology change. The camera might be better. So those type of adjustments need to be be done. Uh weights uh lag comes stem from the fact that uh there's

a lack of inclusion of new items and those and preferences can change and so the weights in the sample the weights in the categories can change but that's not really uh taken into consideration through the period of time because they're locked in. Now all of these are not going to go away because we create an independent all of these criticisms and things because are not going to go away because we create an independent CPI. With an independent CPI, what we are trying to address is actually the conflict of interest, right? So if the government benefits from the CPI being suppressed um and they continue to make methodically changes to suppress it u because either they want an IMF uh loan

or because it just saves them money. Um nobody really talks about the huge conflict of interest that exists right there and everybody just trust the numbers blindly. Um not so much in Argentina anymore but it it does happen. And so the conflict of interest, the censorship, censorship, suppression and distortion is what we're trying to address and um what I'm proposing and anyway some of the challenge so what I'm proposing is basically recreating this aside from the challenges that I already spoke about that are every survey has uh the biggest the biggest challenge to recreating a decentralized CPI is the fact that we don't have that universe data and it is very difficult to reconstruct

because we cannot compel people to give us their information, especially businesses for free, right? So, there's a couple of things that we can do to address that. Um, we can uh for one, we can buy the data. uh but uh going back to the fact that this would be a decentralized uh CPI there's uh we cannot really implement too many incentives because we want to avoid self- selection bias. So if I actually offer to pay people money for them to provide information, and this is one of the reasons the government does not pay individuals to provide information, it's

it's voluntary and free. It's because you don't want just poor people or people that need the money to provide the information because then you're not getting a full representative sample of the of the population. You can't infer from that. Um so that's a difficulty that we would have to deal with. Um but that's a difficulty that everybody deals with including the government. Um then if we do it independently and in the centralized ways we would have the cost of decentralized sto decentralized storage right and then if it's a public good who pays for it that's you know that's a huge question and there are ways that you can incentivize the mechanism uh as as a as a DA or like as a community um especially if you don't

trust your government but it it is it is it is a challenge it is not something that's going to be easily attainable and then getting an actual universe to have a proper sample. There's a couple of ways that you can do it if you actually have enough money and pay for it. There is um Visa and Mastercard actually have a more robust uh universe of how we spend our money. But there are privacy laws. So they can't just go about uh telling people what you spend your money on. But they do sell aggregated reports on it. And those aggregated reports are probably much much better than what the government has. But they also guard them very closely because that gives them it's a source of income for them. So it's

expensive. Uh the other source that you we could use is um it's there's these two other companies that collect scanner data directly from reporters and they have agreements, exclusive agreements by the way. So neither of them have everybody. So if if you're if you're providing the data to one of them, you're not going to be with the other. So um and and the value that these businesses get from them is that they get analytics back and so but their data is also very expensive. Now another thing that it's for future research and I think it's becoming uh more and more feasible is actually allowing people to voluntarily provide their data with uh

privacy preserving mechanisms like see proof that you could potentially be willing to provide your information if it was uh protected, right? and we could just use your numbers to actually do either weights or uh reproduce the the CPI. Um and then the last thing um really wish um and and the last the last thing because I think I'm getting kicked out of the stage um that I want to talk about is um what we can do currently. So currently there is no weights there is no fix to the weights problem and we've seen that even

with the recreation of the Argentina Argentina so from uh the research that was done through Argentina uh price stats um came about and it's u it's it's it's a product that provides different CPI numbers for different countries and they still apply weights from the government. So the weights problem is a very difficult problem to solve unless you have enough money to get all of this data from private companies or create a shadow uh survey from people and just ask anybody that got uh that is being surveyed by the government to provide the data also on chain and then now we have the macro data to actually validate

it but then again that comes with like a lot more research on how CK proofs can be used for that. Um, I want to close with a sentence that was probably earlier. Um, sorry. Why this matters is because Benjamin Franklin once said, "But in this world, nothing can be said to be certain except death and taxes." I went ahead and corrected that because in the US, if you have enough money, you can probably avoid taxes. But no matter who you are, you will you will experience inflation.

No matter who you are, how poor you are, it does unlike governments, it does not discriminate. And with that, thank you so much for your time and for staying behind. Sorry.