 Good morning It's my pleasure this morning to introduce our plenary speaker Edwin Anderson Edwin grew up here in North Carolina Winston Salem. He attended NCSSM And he took a math modeling course with Dan Teague and in Edwin's word words that experience changed his life He's teaching has been a big part of his his career and his life After graduating from NCSSM in 1989. He earned mathematics and physics degrees in NC State There he served as a teaching assistant for the physics department for three of his four years and worked as a tutor For dyslexic students through student services He went on to earn his PhD in physics at Cornell where he also taught as a graduate student and was on the faculty for two years His careers have taken him in lots of different places In particular math modeling has played a big role in his careers He's currently a partner at the international consulting firm Oliver Wyman where he has remained involved in financial modeling including not only mortgages, but also climate change With United Nations and on housing challenges in foreign countries We're very excited that he's here with us today and please welcome Edwin Anderson as he talks about the razor in the bubble so a couple of comments the first is One of the kind of basics any Presentation is pictures diagrams pictures One of the basics of the business world is you don't know you can't take it with you. You can't show it to anybody So there are many things in here which I have seen in Google pictures and they are still the drive of some previous company I worked for and As much as I would like to buy a half a million dollar dataset and do nine months of work to show you the pretty picture I can't but I'll just start with this is a known pedagogical weakness So I want to do a little bit of in media res for this story Take you back to September 15th 2008 Subprime mortgage market had collapsed along with the associated bonds CDO bond market had collapsed Lehman it failed Bear Stearns had failed I was sitting with my team with the weird experience that this is pretty much exactly the list of events that we had predicted As so much crumbled it was a really weird weird experience We'd actually made money in the business world a lot, but not gobs and everyone else was losing gobs We'd done that while telling all of our clients sell your subprime mortgage bonds sell your CDO bonds and Through a role of interacting with senior leaders at companies We knew essentially no one was prepared a central banker had mocked me By pointing out some basic bits of standard economic theory that showed that subprime bonds could never collapse For what it's worth economics does not come out well in this talk Microeconomics behavioral economics fabulous macroeconomics mostly garbage What was actually a little darker is within a few days of all this? The merger with Merrill Lynch meant that basically we're all going to be fired So it's a little strange an important caveat to all of this This is a story about it's fundamentally about success in mathematical modeling and some of the things that led to it So it'd be easy for that to come across as a bit boastful. So I think it's incredibly important to caveat I have had an enormous amount of luck and privilege. I Have been in the right place in the right time and happened to know the right people at some kind of crazy points The Lehman Brothers thing came because I went to a conference on a lark and a Lehman Brothers guy Who probably shouldn't have explained one of their strategies to a group of people that he didn't expect one of their competitors to be in the audience Okay, just weird stuff and better than fair hit rates You expect on big financial predictions that if you get one in three you're doing well And we went through a period of about four years where we made zero mistakes. That's not fair That's like flipping heads eight times in a row Better than better than we deserved I Have had every privilege except money basically I White straight cis male from a stable family that valued education from a safe neighborhood with some unusual neighbors and Exceptional teachers and opportunities when I say unusual neighbors the guy behind me who his dad Wart on computers in the early 70s and He had built his own mainframe in the basement Just himself and it was apparently like the third one He would build them rip them down build them, but it meant that when I was three years old I remember having a large unique system to play with all this a huge advantage And then we have to recognize selection bias, right? If all of this had sure still crumbled into failure the odds I'd be standing here as your speaker this morning much lower unless I was giving you an object lesson in failure I won't claim I haven't worked hard, but as we all know that's rarely enough Today I'm going to try and link Some a critical moment in my career and some of the learnings that have followed from it to the mathematics I've learned and I'm really going to concentrate on some things. I've learned about the nature of problem solving and assumptions It's not going to be about a mathematical technique But about that structure. It's probably clear the bubble for my title is the Great Recession And a lot of the key mathematical ideas Are entwined with the razor title Occam's razor and now because I know at least some of you are thinking this they're going Why did he misspell Occam? So Occam is the name of a town in England. It's an English word But it got Latinized by people writing in Latin about this guy's writings later And they changed the spelling yet somehow when we reverse it back to English Most people leave the spelling the Latin one not the English one even though they've gone back to English That's weird So I'm going to try and show you how some of the basic good ideas I learned about math modeling about 30 years ago within these walls and they're going to be some side trips Another comment I could lay out a deeply carefully structured presentation and That is the second most effective way to communicate The most effective way to communicate study show is through character-driven narratives That hang in people's minds. So if I ear too far in that direction, please excuse me I'm hoping that some of the pieces of this is hang with you because I Want you to learn what I learned and hopefully it can color and help your students So I'm going to first talk a little bit about Occam's razor Then most of the talks to me in two sections about bubbles first Some base some the two big problems will hit on is they had very poor numerical and structural footings the models used and then Just a rethink of how academia approached it and then some closing thoughts Occam's razor So there's that dichotomy what was taught and what was learned and I'm gonna again another important caveat None of this has been checked with dantique he may hear this and go. Oh my goodness. Like is this what was he remembering here? And actually all my I live in a New York apartment So there's of course not room for the four enormous tubs of all my academic materials They are in my brother's garage in Washington DC So even though I have carefully kept and now it's in a blue binding with a white label on the front my old notebook from from Dantique's class I couldn't go I couldn't get my hands on it before today. So What do I remember? We weren't lectured to that much. We spent a lot of time doing in the class and We spent a lot of time solving problems and groups. I remember voting power, which I think you've spoken on at this conference in recent years auction systems spreads of diseases Suez canal usage optimization forestry and tree growth predator prey models and more I Remember enjoying it and I remember being engaged because of the amount of doing now. I Certainly remember that enjoyment of doing these things But what did I actually learn? I? Came out of it with a process and I don't know how explicitly any pieces of this process were taught to me because I can't look Back in my notebooks, but when I started trying to teach some of my employees and students through the years I found myself putting it in a framework What's the real question? How can I get an initial answer making the best set of assumptions possible? And I'll touch on best as we go forward How can I improve this answer through relaxing my assumptions or additional data? How trustworthy is my answer and what are the implications again? I'll focus mostly on the two about assumptions and I want to comment a little about what is the real question? I'm amazed how often this one is really wrong so There's a case that's from the military and a friend of mine was a modeler for the army before he came in worked for me in economics So the military asked its modelers could you predict when terrorist attacks in Baghdad are going to happen? Can you help us predict that and They worked really hard on it and they said look you know It's not working Which in the end you kind of know it can't work Right because if it was predictable then The attackers could essentially psych out their own patterns. There's a human feedback It's essentially doomed to be an at least pseudo chaotic system and At about the point that everyone was getting angry with each other. It was kind of revealed that they Everybody knew that it was dumb to ask when the attacks happened. That was a stupid question But the question that they actually had was Even if they weren't posing it well How do we prepare ourselves and that meant understanding the clustering of attacks? How many medical response teams do we need? How many military response teams do we need? What kind of distribution of those might make sense and in the end that problem the model answered well What are the implications of my answer was the one I took the longest really to fully see it I would think in terms of the implications in terms of oh, how does this change how people should do things as times gone on? seeing the political and organizational implications Has become clear how incredibly important it is and how those dysfunctions can feed back One of the other things I learned from that class Was a deeper understanding and appreciation for math in the world. I already liked math There's no question of that. I had many good teachers. I enjoyed what I did the other comment out those steps you essentially almost never see all these steps done well and Occasionally they are truly critically mangled Another classic business world one senior executive asked question No one dares go back and clarify ask the senior executive what they mean. No, I'm not gonna go I'm not gonna go talk to Mary Ann actually actual very senior executive that's come up in the last month I don't ask Mary Ann. What are her question and I'm terrified of that well So how does this fit in with Occam's razor? It's an old idea Occam great marketing move here because the idea Was 1500 years old when he wrote about it goes all the way back to Aristotle and it was basically that simpler explanations are better than more complicated ones What does it really mean people have written volumes? I'm not even gonna scratch the surface in most ways, but let's talk about it in terms of math modeling Solutions are best when they're most consistent with the trustworthy data and require fewer strong assumptions I'll clarify trustworthy data is nuanced Most data has not just random noisy errors Which everyone learns how to deal with how to measure how to understand how to do those basic equations for standard error And can look up in the internet techniques for doing fancier stuff But most of us systemic messy errors in it Strong assumptions and I mean this there are a lot of ways you can define this technically but very loosely Strong assumptions are ones that greatly affect the final answer. They box in where the solution is headed You want to have fewer of those? It's often abused Two are particularly insidious and common. The first is people pick The easy answer the one that involves them having to do the least work It's usually easy to spot, but it's sometimes hard to fight When time and resources are limited and sometimes it's your only option because your resources are limited, but Often that's not why it's happening The scary one is the least additional assumptions What do I mean by that? So say you're an economics modeler and You're looking at a bunch of securities and you're like, oh, they're rating agencies who tell me how good these are I'll build a model that inputs rating agencies data and Comes up with an answer Well, you've made very few assumptions You can probably make a really tight great set of assumptions But you have just built on top of everything the rating agency assumed and You know what you've you built on to many many assumptions and For what it's worth that example. I just gave is how a lot of people got the financial crisis wrong They used rating agencies and economists models without going all the way back to the basic data It's most common in academic disciplines and I'm going to tell you a fairly tragic story layer about that and in established businesses We've always done it this way. This formula has always worked. Why would we ever think back at this formula or look at the original data? so The key idea that I learned through high school math was really I think about thinking about my process and in particular assumptions and I think that they line up well with kind of the prudent use of Occam's razor The Rayleigh applied well, and I'm gonna talk you through how big these implications can be So let's talk about bubbles So let's back up a little bit for my earlier snapshot in 2002. I'm looking for a new job After I'd been a PhD at Cornell I Looked at what it meant to be a lecturer. I loved real-world problems didn't really like academic problem solving And I was like Okay, I kind of see the game with a lecture when I watched a lecturer with 15 years of experience Every teaching award at the University be told well There's a visiting adjunct professor who I'd met and could not actually speak the English language coherently and Guess what enrollment is down 2% so we think we're gonna fire you that was a wake-up call I'd spent two years at McKinsey going through their MBA program and I Can touch a layer. It's another big management consulting firm But what I would say is that they are masters of communication over substance And even though I learned some substance there the most important thing I learned there was about packaging things up And it had given me I think it's safe to say a little bit of a cynical view on some things in the business world I didn't offer from the think tank Rand It was a joint appointment to do education and economics work. I'm like this sounds like a lot of fun I think I'm gonna go do this and I told them I've got a couple more interviews to do but I'm probably gonna come join a friend of my aunt mine Asked me to interview a Bank of America at the time. I lived in Charlotte. I was like, I don't want to be an investment banker He's like one hour. Just give him one hour Okay, they have one hour and you know, what's an hour as a favor to a friend So I sat down with a department head. He was a young Duke math PhD who actually just by happenstance. I saw this past week And a fascinating breakfast with charade and he showed me how two models worked one for residential mortgages So single-family homes condos and one for commercial mortgages office buildings apartments Places where a landlord owns a building and collects rent He asked me what do you think of these? When consulting, there's a very standard type of interview and in that interview It's called a case interview. They give you a question You tell them what you think and show your skills by digging in and I'm like this looks exciting So I quickly took apart the residential model. I'm like look it's massively over fit for the following reasons It's gonna look pretty on back tests, but there's no way it's gonna work basically the whole the Tor problem was the following Imagine that I have ten thousand Phenomena that all follow the same sign curve How many degrees of freedom do I have? Well a sign curve has An amplitude Phase and a frequency so three, but I got ten thousand things doing this Do I have thirty thousand do I have three well if they're moving pretty much in lockstep? You have three and these folks were acting like they had hundreds of thousands of degrees of freedom and They were fitting twenty variables to it and they maybe had eight The commercial model was Beautiful it was so as a person trained in physics. It was so graceful and beautiful But also I remember things some of them explicitly from dance class and about some of the chaotic systems we saw and I was like This thing's chaotic It's gonna be like a random number generator any meaningful a time in the future I'm like, so I took apart your two models. I was feeling good and then it turns out this wasn't a case interview question These were the standard models for mortgages My jaw dropped. I looked at him incredulously and I was like so everybody's using these I tried to think of why because normally the assumption that people are being stupid is You not seeing them you not putting yourself in their shoes. So I assumed I had to be missing something I'm like, oh, is there a gross shortage of data where they forced to this by the fact and then he showed me the volumes of data available and I was like If I come work here, can I do something radically different? It was like walking into an academic discipline and discovering that Everything people had done previously was using crayon on walls and you're like and here's a stack of data Would you like to start a whole new field? It was kind of crazy and thus seven years of my life started Why were the models so bad don't we have a whole field of economics studying this whole enterprises The heart of these models were for academic institutions. We're gonna go to this Additional life lesson if you park in a two-hour space and you end up staying for eight plus hours You can accumulate a lot of expensive parking tickets, especially when you're currently unemployed a Gross simplification of what was wrong There are a couple of big things that kind of drive drove the crisis and for that matter sadly some of the modelers People were being paid to act in ways that encouraged Them to ignore problems and wreck the economy Sadly, this is largely still the case and it's well outside the scope of this talk But in the modeling way the academic grounding was terrible There were numerical structures that made doing it right very very difficult and the study of economic bubbles was and mostly still is an Unmitigated disaster Things are better now, but the pressure to do poor economic modeling remains enormous Poor numerical footings, so we take a vote here. How would you think about the economy? would you think about it as each industry having its own ups and downs and Some correlations between them you might measure that lead up to booms and crises or The economy moves up and down on moss and there are variations where an economy might do Industry might do better or worse, so how many people want to vote for the first one? Two three four a handful. How many people want to vote for the second? Okay, most of the room So though some people do pick the first answer most people pick the latter the latter is certainly the simpler way to think Turns out that from a mathematical point of view They're identical and the first one allows you to let every modeler build their own specific model for the individual disciplines and Use a correlation structure to link them, so it's politically easier chosen by both modeling groups But in a way if you think about it one of them sounds like an oversimplification We're choosing the easy route the other were letting all these modelers do their own thing and then putting something on top of it So we're building up a lot of assumptions on top of assumptions. That sounds like the other error so You know you build the distribution set up the relationship and add them up to understand it There's a dominant form that I'm not going to get technical in this talk on called Gaussian cupola's It was the standard academic approach And in fact there was math like sclaris theorem that showed that it's identical All you had to do is get the distributions and the cupola is correct and you were fine But it didn't work the results were I Think if you compared them to actual history kind of silly in fact How did it go wrong? Well, the correlation structure necessary to make the answers identical was really complicated It was so complicated that people essentially threw up their hands and said let's just measure historical ones We'll say the correlation between income and forestry and changes in common corn farming is a point six We looked at it historically point six The problem is the actual correlation structure with a complex function Who is going to figure out thousands of complex functions wasn't even doable fill in point six? problem with that There's a mathematical nugget that's often talked about in modelers, which is that it all correlations go to one in a crisis So having all of them said at point three actually essentially in point six actually did not allow Crises to happen in the models. It was impossible so Isn't the alternative to oversimplify well the test was to look at the data and it turns out that the simple approach One big curve for the economy perturbations fits phenomenally better. You're actually not giving up much by taking that and When you relax the assumption and allow the variations by industry to be more complicated turns out It works pretty well So the simpler model was more robust which changes in assumptions Partially just because there were fewer complex ones built on I want to tell a side trip here and It is a piece of tragedy I would say So it's about 2004 I don't have the records to be certain of the year and I was doing a job interview with a student who is about to finish their PhD in economics There's an Ivy League school and it was one of the saddest moments of my career Students reaction. I mean it's still solid in my mind and The reason I'm telling this story is because this student had gotten themselves out of place You never want any of your students to get into To make this make sense. I'm actually I have to talk about The assumptions behind a bit of commercial real estate theory And so we're going to take a detour off our detour So there's an idea theory called double trigger default which you can write in two different logically equivalent forms One says a borrower will not default on their mortgage on a commercial building. So building where they collect rent if The rent is greater than their expenses plus paying the mortgage that makes sense, right? If I'm getting a check each month, why am I defaulting on this building? And even if they're not making money As long as the mortgage is less than what the property can be sold for I'm going to hang on to it and at least try And sell it to unlock the difference That makes sense defaults these are pretty common sense rules You you'd like to have a check and if you have a thing even if it's losing you a little bit of money each month If you can sell it quickly and make enough money you'd also hang on to it Well, that's logically equivalent to saying they will default If the rent is less and the mortgage is greater I partially did this just to have fun with basic logic, but also The one on the left is the one that usually fits people's common sense better The one on the right is actually how it's applied normally and structurally in models It's pair of common sense triggers. What could go wrong? Let's do a thought experiment. Imagine you're the property owner You've got the nice building in your commercial district It's worth a hundred bucks Which is more than the 70 bucks most of the other buildings are and why is that because you've got a really nice restaurant And everyone's like you have got the best tenant. They've got a good solid income. They send you a big check It's really great. You make seven bucks a year At $12 in rent minus a buck of expenses most of the time tenants pay most expenses But there's some you're probably having to pay the mow the lawn and maybe you're including water Who knows and a four dollar payment on your 70 dollar mortgage Pretty straightforward. Suddenly family events happen. It's a little family business and they shut down abruptly Suddenly you're cause it's costing you five bucks a year, right? You still got to pay the mortgage You still got to keep the water on What are you gonna do? Well, you've met the first default trigger right Then you're like, I just lost the tenant that makes my building More valuable than everyone else's so my building is probably worth at most 70 bucks and the other people they've got tenants I'm probably worth even less because I got an empty building right now So my mortgage is now under water. I've now met both triggers of double trigger. So we're gonna take another vote To default or not to default How many people here would default on the mortgage at this point? Okay, uh half a dozen. How many people would not default? Maybe 30 and a lot of people not putting hands on So here's a question Not sure or i'm being unclear Not sure. Okay So what do people actually do? Well tenants think about it this way All right. I'm sorry how landlords think about it. They go What maybe i'm just not going to get a tenant Well, then i'm in trouble and i'm going to need to default or i'm going to have to come out with money outside It's going to be ugly I could get an okay tenant Just like everybody else That's the case I'll be making less money each month, but i'll still be getting a check each month The mortgage will probably kind of a wipe right my building's not worth 70 my mortgage is 70 I'm gonna have trouble refinancing it, but it'll probably be okay But i've probably lost a little money while i was finding a tenant, but you know a year or so Of income i'll make up for that What if i get another great tenant? I'm back. It's sunshine and roses In reality most landlords gamble on the second two Sometimes even for years looking for that next tenant Nobody Essentially defaults immediately Well, why am i telling you a story of an inconsistent logical framework? And i'm telling you this is a tragedy this does not sound like a plausible tragedy Well, the key is that double trigger is entirely accepted by academia Everything in academia uses it it assumes it it starts from it and dare you not question it Student had done a great piece of modeling They've made reasonable assumptions Tested the validity, but he made an error making the fewest additional assumptions All of his work was based on his thesis advisor and a group of about half a dozen other academics Who all of their work was built on double trigger Because all the and he what he why shouldn't he all the academics he knew used it Everybody knew it was true. It's common sense right i put up that first one It makes a lot of sense until you think through all the other options But like putting a well-built wooden box on top of a house of cards It all kind of easily collapsed I was doing a phone interview and early in that phone call before i understood really totally what the student had done He said oh, I did some work on double trigger default. I said, you know, that's really interesting Um, because you know someone needs to do some nuances on that since it doesn't work And he was like, um, what? I started talking to him about the data and then as I talked to him more it became clear what his thesis was And um at that point i'm trying to in some ways soften the below point out what's okay The student was to say he was crushed would be going to make an understatement He'd spent three years of his life writing a thesis He had pride in that work. He had trusted his advisors And suddenly he learned that his advisors had ignored the data because it made a beautiful theory His work Doesn't have any applicability. It's just an exercise And he was still going to get his phd, but it was clear he felt bad about it The world of getting caught onto layered previous assumptions is it can be a dark one And i'll keep remembering this and it helps me remember To not make this mistake and help my students not make this mistake So we're gonna go back to academia We're gonna attempt to rethink it Um a little aside. I remember going to one of my first economics conferences And I sat down with some grad students and postdocs. I thought this would be just interesting Let's hear about what they're doing and they just chatted with them And then as we were getting to the end of it was a break and uh, we were getting to the end of snacking and drinking our coffee One of the guys was like Said like kyle kyle's committing academic suicide today. I was like what? I came there. I got yep all over for you kyle like what? Kyle's showing data Why would that be? Why would You don't think about that as a problem psychology uses data chemistry uses data. Why would economics not use data? We'll come back to that Because there's an enormous amount of modeling work on economic bubbles Scads of books Papers complex work There's a large amount of funding for it The state of prediction And even for that matter understanding is incredibly poor I'll show how much this is a case of taking the easiest answer and the danger of poor external motivations I'll show a simple approach that looks at simpler assumptions fits with history And gives us some insights that help us actually manage bubbles And as an aside I've actually given a more detailed talk on this to the central banks of europe They've they've had interest in this because in the end A lot of them are pretty aware of this mess as people trying to actually apply things to make the economy as the world run a little better So let's start by actually trying to define what we even mean by an economic bubble we have two Slightly confusing and fancy definitions from nasaq or investopedia I think the academic one's the slightly funnier one because typically the academic definitions are circular The best way to forward your theory about something is to define a way of the problem Say you're from the university of chicago You're a free market economist who believes the economy can do wrong. What is the definition of a bubble? It's a thing that people want to call a bubble because there's no such thing as bubbles because everything is efficiently priced always And values of things are always as they are up or down Yeah, I mean, why do we really want to talk about that? Or you go to someone from princeton and they're like It's a phenomenon in which evil people try and fool stupid people and they try and fool other more stupid people And it's a corporate of fools that result in unnaturally high prices Well, if I'm trying to prove that Markets are infinitely efficient or That it's a greater fool theory I can take care of that quick But what people can agree on is the examples Some of the most cited ones are here to the left The south seas bubble from 1720 is where the term comes from And there are tons of examples For me, I just tried to pick a basic definition A significant price in pulses prices that might necessarily fall. There's weaknesses to this. It's a little glib But it's a lot more useful than the complex definitions And I kind of poke here at garber who is actually kind of the key chicago person who's pushed most Bubbles are not real. Um, he has a really great book in which he goes through a lot of the classic data from bubbles And he does this beautiful historical work and he cleans up the data and he shows you all this great stuff And then and then a miracle occurs and his theory is true So it's a tragic book to read. But if you stick with the history part, it's pretty cool So let's talk about what makes an appealing explanation Because that's what wins often in the marketplace of ideas The first is people like a single easy to understand cause and effect People like simple answers and academia respects theories when they cover more things Single umbrella theory that covers everything is better than three or four or five separate ones Supporting experiments A cynical is going to be about some things people believe things more when there's experiments showing that it's true Especially kinds that are repeatable, which are hard because the financial world is messy In many cases we're running one experiment. We don't get a control group It's hard Supporting economic theory and math Seeing something that happens and lining it up because you can't do that as well building mathematical structures helps a lot And then a controversial inflammatory or politically pandering explanation That there's a deep human appeal and exciting answers and pundits and experts Are not rewarded for being right There's a famous guy named tetlock who Used to be an obscure name. I've tossed out in these talks, but now some folks have run into him in Nate Silver's work He's a sociologist who when he got tenure said i'm using tenure for what it's for i'm now going to run a 30 year experiment starting now And what he did is he studied experts And what was correlated with him being right and what was related to success in their careers Being right had nothing to do with success in careers because no one goes back and challenges pundits and whether they're right or not Few people go back and challenge academics on if they're right or not. Many academics are Horrible at predicting anything, but that's not what they're judged on On the other hand, how do you get famous? You say big exciting inflammatory things that get you front of a newspaper reporter on a national finance show And making it even more extreme is economics is embroiled with politics So an answer that sounds really republican or really democratic Gets you all kinds of support Oh, everyone should be listening to edwin because his results are here on the political spectrum and therefore he's right In the end the literature of economic bubbles is full of oversimplified politically pandering ideas There's another thought from tetlocks work that has become again I think probably the biggest popularizing recently has been by Nate silver though I actually ran into it in an article in a news week decades ago by a New York newspaper reporter named Sharon begley And it's this idea of the hedgehog in the fox And it comes from an ancient greek warrior poet who said the fox knows many things But the hedgehog knows one big thing hedgehogs build a structure They fit things in and they understand it holistically foxes know lots of individual things and solve it in pieces We love hedgehog theories Politics and academia love hedgehog theories Hedgehogs are much much less good at predicting things than foxes Most theories are p that work Are piecemeal and messy because the world is piecemeal and messy So what kind of answer am I going to give you about bubbles? I'm going to give you one. Does it does it have a single easy to understand cause-effect relationship? No, no, it definitely doesn't Does it have export supporting experiments? Actually it does It's one where we can do some What about supporting mathematical economic theory? In the end, I'm going to give you a fox answer Or a little at the surface of one given time today and only in one of the cases. Is there any solid math? Turns out it's a very useful case, but the others no What about a controversial inflammatory or politically pandering answer? The answer will make no one in those camps happy So I spent four years helping regulators after the crisis get to a better understanding how to take apart companies And as part of that I used to teach their employees Teach their employees classes on a regular basis and I used to go into each of them And I would say I am an equal opportunity offender Whichever tribe you belong to let's be clear their tribe democrat or republican your economic theory Is bunk. It does not line up well with facts and I'm going to offend you by showing you facts So The way I'm typically inflammatory is by not being appropriately pandering for whatever audience so One of the things I'm going to point out is that what I'm going to line up do is definitely a fox answer Which hopefully is a good sign And in the end trading the financial markets is akin to good experimental science It allows you to test things in all sorts of interesting ways To do that I first said I can not sure I believe this all bubbles of the same phenomenon But what if bubbles are like a fever? I know there's a fever but saying like I'm now going to model all disease by saying there's a single thing called the fever In the end fevers are outcomes of lots of underlying causes in the end if you don't think about the underlying causes You don't get it right So I started asking questions just really basic history ones to characterize and structure the problem Asked was a dominant player role played by leverage just basically borrowing or over printing of currency Was a dominant role played by fraud in the bubble And there's a note that's a dominant role Bubbles attract fraud even when they so there's usually somewhere on the edge was the boom kicked off by technology Was picking individual assets important So in other words did some things win and some things lose instead of everything seeming to win and then everything seeming to lose Was a large role played by human tendency to assume that prices are going up They're going to keep going up and finally we're meaningful things of value left behind and you may go in what economic bubbles leaving value behind I'm going to touch on a couple, but What about the internet bubble? We get anything out of the internet bubble Yeah, we got all sorts of interesting companies new services. Was it a mess at the time? Yeah, but value is left behind We can ask a lot of other questions What I did is I this is part of a much larger spreadsheet I made a list of bubbles and then I just started checking the boxes Just read a lot of history And I went, okay, you find that momentum plays a role in all of them Dominant role of fraud only in a few I know a lot of people think of the subprime in the larger bubble as having a large roll of fraud Fraud makes good newspaper and television, but in the end it was mostly around the edges Had a lot more to do with the leverage one So I tried to say the leverage what about a dominant role of tech change And I looked at these checkmarks for a little bit and I said this doesn't look like one phenomenon It looks like three with an exception so There are fraud based bubbles the ones in red It is perhaps should make us feel good the last major fraud bubble was before the great depression in 1920s Have you ever heard the phrase want to buy some swamp land in florida? Anybody hear that? That's from that It's hard to travel. There's no internet a lot of parts of florida were hard to get to there's a boom happening there People would buy and sell florida real estate sight unseen and sometimes it was literally water And in the end There's a longer story here We don't have time for but the bubble lasted until somebody took an already old schooner and accidentally wrecked it in the miami harbor weird things and bubbles We also have leverage bubbles the dot-com bubble is uh, it is one That we all know well, but for that matter Black tuesday at the great depression and for that matter The experience of the japanese all based on that and then tech bubbles There have been a lot of them anyone who tells you this time is different You should basically start not trusting them Most of these things have happened lots of times before and in fact If I were to tell you the story of the auto bubble in the u.s. You think it would sound like silicon valley A town where all the talent suddenly goes Detroit And as people gather there thousands of small companies start fighting over what's clearly going to be a meaningful market There's a melee. There's a boom and bust as thousands of the thousands fail Are absorbed and in the end Few dozen companies come out the other side eventually winnowing down to the eight that the early modern era saw Same things happening over and over. So I said, okay What if there are three things I need to model here? I have now lost any plausibility in academia because I'm not going to give you one theory I'm going to suggest that there are three And I'm going to tell you modeling fraud modeling Behavior of fraud detecting it is its own fun thing and we could have a lot of fun talking about bedford's law And other exciting stuff like that But it's really hard to do because the point of fraud is to go in a direction people haven't anticipated before Tech bubbles there's some things we can understand but as we'll touch on tech bubbles are weird and in the end By the time you're trying you can understand them often in hindsight But trying to predict them ahead of time is what gets people in trouble Leverage bubbles once you isolate them are the place where you actually can do some meaningful modeling There we are So fraud bubbles There's a famous author kindleberger who has one of the best books on echin on Bubbles and the reason why is because he doesn't try and put for any theories He just writes about every bubble in the world, but He points out that when everyone else is getting rich Or you think they are it's easier to believe someone's implausible pitch to you Some comes to you and say I can double your money in a month. You're like, okay Clearly clearly someone trying to rip me off But when two of your neighbors just said hey, I just doubled my money in a month And someone comes you want to go tumble your money in a month sounds real different and In some bubbles fraud was definitely the dominant effect And for regulators the real lesson here is about vigilance and when it really comes down to it Not saying they shouldn't keep working, but they've done a pretty good job It's important to note about how smaller frauds proliferate around both other kinds of bubbles and larger frauds There's a classic work that outlives some things about the south seas bubble the one that gave bubbles its name And People started stock companies then for things like a wheel for perpetual motion For employing poor artificers and furnishing merchants and others with watches for the transmutation of quicksilver Into malleable fine metal now there was you have to be careful about your own lens on things One of the items in that list often scoffed at was an exchange market for human hair But how many people wore war Wigs made of hair bought from poorer people that grew it out and cut it off Lots it actually was a good idea, but it sounds dumb through our modern lens leverage bubbles Increasing the supply of credit how many loans people can get increase in money They involve a cycle of asset prices go up a little bit for some natural maybe good reason Which in turn people say hey this is going to keep going probably so they Decide they're willing to pay even more it feeds on itself and the longer it goes up the more people are like Well, this is just the new normal everything's different now. They're going to keep going up This is the new trajectory. So people go up even further and then eventually Something cracks it Great depression great recession leverage bubbles are typically the ones that hurt human well-being the most One of the interesting things and it kind of comes from It started from kind of the money supply leverage bubble in is there's some great experimental work Uh, one of the kind of interesting bits of my career was getting to actually meet the guys who won the Nobel prize for their Experimental work and they've done tons of controlled experiments And I don't have time today to go through all of it, but they've done a fascinating series of experiments And now so many other people leveraged off of it relaxing different assumptions So we set up a fake marketplace. Well, the problem is it's a fake marketplace Well, let's do some experiments where it involves lots of real money. Well, the problem is is that you make the Experiment too short. Let's run totally huge ranges of links to the experiment and look at how the answer varies over time The problem is that everyone is weird. You guys heard that term Western educated Uh industrial it's a flaw with most social science studies. They basically are all using college students So they heard what if we use people from all over the world at different income levels? Results all have held and what they've tended to show is that bubbles are a natural human phenomenon We tend to create in the absence of anything else because we like momentum There are only a few things that allow you to change bubbles The leverage of money supply the experience of the traders What's interesting is if everyone's a neophyte or everyone's experienced bubbles Or sorry, or some people are experienced. Some aren't bubbles. It is interesting if actually I misstated if everyone is experienced Bubbles are much much smaller Because there's a definitely an element of greater fool to bubbles I have to have someone who I think is kind of stupid who's going to buy the thing when it's definitely gotten too high So if I know everyone else playing the game is really good at this I'm going to be a lot more careful and certain times of futures markets There are weird ways you can kind of almost regardless of the link to the experiment Arbitrage if they would say away the entire rest of the game What about tech dominant bubbles? Um, normally there's a big change and important thing is tech bubbles are often driven by equity They're not about there's not about loans. They're not about bonds. They're about investing in ownership in companies But because investors are rational weird crud can happen very weird things another basic There are two basic economic theories that you should just know going out of here Do not work at all. The first is almost all economic theory after a 101 course Assumes that money essentially flows like a fluid It goes where it needs to be it expands to fit it is frictionless This is bunk Another thing it assumes is that market prices Give you an idea of fair value This is also bunk. Well in the second one you might go hold on a minute Isn't it the fair value? Let's ask a question. So I'm going to pick a real world asset It's a major piece of infrastructure in the u.s And it went up for sale Because it had tolls associated with it What were the bids for it? I eventually got my hands on them one 450 million dollars The next 500 million dollars The next 700 million dollars The winning bid 2 billion dollars What do you think the fair market price was? Didn't take much to show that 2 billion This is one of the two times. I essentially ended up as a whistleblower in the system Something funny was happening But even the person who bid 700 was well above everyone else And in fact, I'm almost certain the mistake they were making making them think it was worth more Things are bought by the person who loves the most And they almost certainly are bidding too high And that creates weird effects. Think about a tech bubble Imagine there's some new technology that's going to create a thousand bucks of value New companies arise. We'll assume a hundred of them and maybe We think 10 of them are survived. So those 10 companies, let's say they get a hundred bucks each when everything ends As investor, I'm looking at all these different companies and I'm like Okay, they're only going to unlock a thousand dollars of value Say there are a hundred companies that get funding They show an average to be worth 10 dollars each But the problem is each person buying a company believes it's going to be one of the 10 winners And they're like, well, I should only pay 10 bucks and buy some of this company I'm going to bid up and get more of this company I'm going to pay 30 bucks. I'm still going to more than triple my money when I suddenly get a hundred So everyone thinks they have the winner when they bid So what's funny is an index of these companies would be 3,000 Everyone has acted rationally. Everyone has made reasonable guesses and the market is triple overvalued This is the standard pattern of equity bubbles I'm going to skip this one just given time, but a couple of ones that's worth noting UK Railway Mania makes some good reading There's actually uh If you think the dot-com bubble was big imagine a bubble that gets to 8 percent of the economy And in fact it had such a set of weird events that so discredited its regulators that things almost fell apart The dot-com bubble is an interesting one because we think we know what happened. It's studied tons Lots of people like to study where data is available. It's a recent bubble Therefore it gets lots of study and people act like it's every other bubble It's amazing people who are part of the greater fool camp love to only use the dot-com bubble and nothing else Now the greater fool part is that tech bubbles have a general phenomenon If you've heard of it, you're going to lose money investing in it All the money is made by people early on before it's publicly known Normally by the time you launch off into valuations that are never going to be sustained That's about when you're hearing about it on the financial news or from your neighbor Though it's interesting the tech bubble green span got out in front of it and called it Overvalued at a point where it actually wasn't yet. So it was interesting. This one has a little bit of exception I'm gonna have to touch on as I get to the end tulip mania I won't get to fully explain it, but it's the weird bubble and yet it's the one everyone likes to write about And it's the only one that doesn't fit any pattern because it's not a tech bubble It's not a leverage bubble. It's not a fraud bubble. In fact, if you look into it, it was a fashion bubble tulips The patterns are unique and associated with viruses You can't breed them the fancy tulips. These are not the single color tulips. You can go buy whole foods These are crazy patterned beautiful things and they were caused by viruses So the only way to replicate a beautiful tulip was essentially to clone it by splitting off the bulb Owning the bulb of a variety was essentially owning the factory for a or a fashion item turns out also the stories of people paid the many years of salary To buy one bulb. No, they didn't They paid a tiny percentage up front and it was a fully cancelable contract They paid a tiny fraction of that and they were like, I'm gonna watch in particular the french fashion market And if this becomes big, I'm going to go ahead and pay the money because I'm going to make big money on this And if it looks like it's not I just won't pay any more money So they essentially as you might say bought an option on a fashion item. It was very very strange And it doesn't fit anything we've had since But it doesn't stop people from liking to write about it and in some way the attention on it distracts us from the answers that matter more I make a little side trip as I reach the end mathematics I found useful in a career of doing mathematical modeling. The short answer is all of it And the most extreme case I can say is I once used the first fundamental theorem of calculus To help a pulp and paper plant save save several million dollars a year It's all useful if you're going to go into mathematics There are things outside of basic arithmetic and geometry that have been the most useful and impactful in this order stats probability and combinatorics linear algebra And the basics of calculus the idea of areas under curves slopes, etc There's some things that I have to highlight that people grossly misuse constantly game theory Single-play games versus repeat games everyone loves the answers from single-play games almost everything's a repeat game chaos theory Good excuse for acting like we know nothing But actually if leveraged right allows us to know some fascinating stuff and most big data techniques because most people's data is terrible And in the end big data techniques act like data is pristine and often gives us enormous rounds of false knowledge and bogus garbage Now there's an important thing There are things that aren't on this list There are lots of other things we learn that are important and I want to make a really important distinction I'm a believer that there's a big when you learn something it changes the scaffolding in your mind These are the techniques I used but that's different From what things help grow the fact of that scaffolding in my mind to think about other problems And I think that that's a critical thing So some closing thoughts The most important thing I've learned from my modeling class was things about the problem solving process and how to think about assumptions In the end they will probably forget most of the mathematical techniques you teach them But hopefully you'll build a scaffolding and if you teach them problem solving techniques They'll probably use that no matter where their life takes them the spirit of Occam's razor It's really about trustworthy data and strong assumptions and avoiding strong assumptions And to do that they have to understand how to test that how to work with that and that's something that's normally not taught but More useful than probably any individual thing about conic sections The razors typically misapplied a lot of ways and you should be ready to spot it and help them spot it Seek in the easy answer and trusting too many other assumptions Economic bubbles It's pretty troubled people use overly complex structures They're hard to use and understand even if they can mathematically be proven to be equivalent in rigorous And poor academic footing has been encouraged by poor incentives for experts to be inflammatory rather than right And one of the things I think is interesting is that relatively simple framings Have allowed me to actually predict things about bubbles that Really have not taken that hard of math But what it took was looking at a problem that everyone had a lot of accepted answers They were entrenched camps and just take a fresh look at it And I cannot but wonder how many other age old problems that we act like are messy and hard just require a fresh look require new framing And i'm hoping that the next generation of modelers Can be better armed to go find those And that the folks in this room will help prepare them. Thanks everybody